Democracy, Inequality, and Representation in Comparative Perspective 0871540886, 9780871540881

The gap between the richest and poorest Americans has grown steadily over the last thirty years, and economic inequality

127 15

English Pages 448 Year 2008

Report DMCA / Copyright

DOWNLOAD PDF FILE

Table of contents :
Title Page, Copyright
Contents
About the Authors
Acknowledgments
Part 1. The Context
1. Income Inequality and Democratic Representation
2. Inequality Patterns in Western Democracies: Cross-Country Differences and Changes over Time
3. Social Rights, Welfare Generosity, and Inequality
Part 2. How Democratic Politics Shapes Inequality
4. Electoral Institutions, Parties, and the Politics of Class: Explaining the Formation of Redistributive Coalitions
5. Economic Institutions, Partisanship, and Inequality
6. Political Agency and Institutions: Explaining the Influence of Left Government and Corporatism on Inequality
Part 3. How Inequality Shapes Democratic Politics
7. Economic Shocks, Inequality, and Popular Support for Redistribution
8. Inequality and Unemployment,Redistribution and Social Insurance,and Participation: A Theoretical Model and an Empirical System of Endogenous Equations
9. Income, Inequality, and Electoral Participation
10. Inequality as a Source of Political Polarization: A Comparative Analysis of Twelve OECD Countries
11. Inequality and Institutions: What Theory, History, and (Some) Data Tell Us
12. Inequality and Democratic Representation: The Road Traveled and the Path Ahead
Index
Recommend Papers

Democracy, Inequality, and Representation in Comparative Perspective
 0871540886, 9780871540881

  • 0 0 0
  • Like this paper and download? You can publish your own PDF file online for free in a few minutes! Sign Up
File loading please wait...
Citation preview

DEMOCRACY, INEQUALITY, AND REPRESENTATION

DEMOCRACY, INEQUALITY, AND REPRESENTATION A Comparative Perspective

Pablo Beramendi and Christopher J. Anderson EDITORS

Russell Sage Foundation  New York

The Russell Sage Foundation The Russell Sage Foundation, one of the oldest of America’s general purpose foundations, was established in 1907 by Mrs. Margaret Olivia Sage for “the improvement of social and living conditions in the United States.” The Foundation seeks to fulfill this mandate by fostering the development and dissemination of knowledge about the country’s political, social, and economic problems. While the Foundation endeavors to assure the accuracy and objectivity of each book it publishes, the conclusions and interpretations in Russell Sage Foundation publications are those of the authors and not of the Foundation, its Trustees, or its staff. Publication by Russell Sage, therefore, does not imply Foundation endorsement. BOARD OF TRUSTEES Thomas D. Cook, Chair Kenneth D. Brody W. Bowman Cutter, III Christopher Edley Jr. John A. Ferejohn Larry V. Hedges

Kathleen Hall Jamieson Melvin J. Konner Alan B. Krueger Cora B. Marrett

Nancy Rosenblum Richard H. Thaler Eric Wanner Mary C. Waters

Library of Congress Cataloging-in-Publication Data Beramendi, Pablo. Democracy, inequality, and representation : a comparative perspective / edited by Pablo Beramendi and Christopher J. Anderson. p. cm. ISBN 978-0-87154-088-1 1. Equality—OECD countries. 2. Democracy—OECD countries. 3. Representative government and representation—OECD countries. 4. OECD countries—Politics and government. I. Anderson, Christopher, 1966– II. Title. HM821.B46 2008 305.09182’1—dc22 2008005773 Copyright © 2008 by Russell Sage Foundation. All rights reserved. Printed in the United States of America. No part of this publication may be reproduced, stored in a retrieval system, or transmitted in any form or by any means, electronic, mechanical, photocopying, recording, or otherwise, without the prior written permission of the publisher. Reproduction by the United States Government in whole or in part is permitted for any purpose. The paper used in this publication meets the minimum requirements of American National Standard for Information Sciences—Permanence of Paper for Printed Library Materials. ANSI Z39.48-1992. Text design by Suzanne Nichols. RUSSELL SAGE FOUNDATION 112 East 64th Street, New York, New York 10021 10 9 8 7 6 5 4 3 2 1

Contents

About the Authors

vii

Acknowledgments

ix

PART I

THE CONTEXT

1

Chapter 1

Income Inequality and Democratic Representation Pablo Beramendi and Christopher J. Anderson

3

Chapter 2

Chapter 3

PART II Chapter 4

Chapter 5

Inequality Patterns in Western Democracies: Cross-Country Differences and Changes over Time Andrea Brandolini and Timothy M. Smeeding Social Rights, Welfare Generosity, and Inequality Lyle Scruggs HOW DEMOCRATIC POLITICS SHAPES INEQUALITY Electoral Institutions, Parties, and the Politics of Class: Explaining the Formation of Redistributive Coalitions Torben Iversen and David Soskice Economic Institutions, Partisanship, and Inequality Pablo Beramendi and Thomas R. Cusack

25

62

91

93

127

Chapter 6

PART III Chapter 7

Chapter 8

Political Agency and Institutions: Explaining the Influence of Left Government and Corporatism on Inequality David Rueda HOW INEQUALITY SHAPES DEMOCRATIC POLITICS Economic Shocks, Inequality, and Popular Support for Redistribution Thomas R. Cusack, Torben Iversen, and Philipp Rehm Inequality and Unemployment, Redistribution and Social Insurance, and Participation: A Theoretical Model and an Empirical System of Endogenous Equations Robert J. Franzese Jr. and Jude C. Hays

169

201 203

232

Chapter 9

Income, Inequality, and Electoral Participation Christopher J. Anderson and Pablo Beramendi

Chapter 10

Inequality as a Source of Political Polarization: A Comparative Analysis of Twelve OECD Countries Jonas Pontusson and David Rueda

312

Inequality and Institutions: What Theory, History, and (Some) Data Tell Us Ronald Rogowski and Duncan C. MacRae

354

Inequality and Democratic Representation: The Road Traveled and the Path Ahead Pablo Beramendi and Christopher J. Anderson

387

Chapter 11

Chapter 12

Index

278

417

About the Authors

PABLO BERAMENDI is assistant professor of political science at Duke University. CHRISTOPHER J. ANDERSON is professor of government, faculty affiliate of the Center for the Study of Inequality, and director of the Institute for European Studies at Cornell University.

ANDREA BRANDOLINI is head of the Economic Structure and Labour Market Division at the Department for Structural Economic Analysis at the Bank of Italy. THOMAS R. CUSACK is a senior research fellow in the research unit on market processes and governance at the Science Center Berlin (WZB). ROBERT J. FRANZESE JR. is associate professor of political science at the University of Michigan, Ann Arbor. JUDE C. HAYS is assistant professor of political science at the University of Illinois, Urbana-Champaign. TORBEN IVERSEN is Harold Hitchings Burbank Professor of Political Economy in the Department of Government at Harvard University. DUNCAN C. MCRAE received a Ph.D. in political science from UCLA. JONAS PONTUSSON is professor of politics at Princeton University. PHILIPP REHM is a postdoctoral research fellow at Nuffield College, University of Oxford, and assistant professor of political science at Ohio State University.

RONALD ROGOWSKI is professor and former chair in the Department of Political Science at the University of California, Los Angeles. DAVID RUEDA is professor of politics and fellow of Merton College at Oxford University. LYLE SCRUGGS is associate professor of political science at the University of Connecticut. TIMOTHY M. SMEEDING is Distinguished Professor of Economics and Public Administration at the Maxwell School of Syracuse University. DAVID SOSKICE is research professor in the Department of Political Science at Duke University, research professor of comparative political economy at Oxford University, and senior research fellow at Nuffield College.

Acknowledgments

This book examines how democratic politics helps to shape levels of income inequality in society, as well as how inequality affects the quantity and quality of democratic representation. Although all of us have been working on questions related to the theme of the book for some time, this volume has its most immediate origins in a conference organized in May 2005 by Chris Anderson and Pablo Beramendi at the Maxwell School of Syracuse University. The conference was generously supported by several organizations, and we are grateful for their support: the Maxwell European Union Center, the Center for European Studies, the Moynihan Institute of Global Affairs at Syracuse University, and the Institute for European Studies at Cornell University. We also would like to thank the director of the Moynihan Institute, Professor Peg Hermann, for her help in bringing together the group of authors as well as a number of conference participants. We would like to thank the following individuals for participating in the conference and providing thoughtful and constructive feedback during the sessions: Michael Cain, Michael D. McDonald, Branco Milanovic, Juan Rafael Morillas, Mitchell Orenstein, and Christopher Way.

Michael Wallerstein, in memoriam

Part I

The Context

Chapter 1

Income Inequality and Democratic Representation PABLO BERAMENDI AND CHRISTOPHER J. ANDERSON

Similar concerns haunt academics and policy makers throughout the old world as recent scholarship suggests that excessive inequalities attack the foundations of democratic political regimes (Acemoglu and Robinson 2006; Boix 2003) and the distributive consequences of markets become increasingly unequal. But these long-standing and unresolved debates have received renewed attention in recent years. Thus, in an essay entitled “The Rich, the Poor, and the Growing Gap Between Them,” The Economist recently commented on the growing gap in the United States between rich and poor by recalling John F. Kennedy’s saying that “a rising tide lifts all boats.” The essay goes on to note that the tide of America’s economic growth is no longer lifting all boats to similar heights. The political consequences of this trend “will depend on the pace of change and the economy’s general health. With luck, the offshoring of services will happen gradually, allowing time for workers to adapt their skills while strong growth will keep employment high. But if the economy slows, America’s skepticism of globalization is surely to rise. And even their famous tolerance of inequality may reach a limit” (2006, 25). Similar concerns haunt academics and policymakers throughout the old world as the distributive consequences of markets have become increasingly and more noticeably unequal. We share these concerns in Europe because excessive inequalities attack the very foundations of democratic political regimes (Acemoglu and Robinson 2006; Boix 2003). But inequality’s consequences are sometimes ambiguous, its trajectory is not

4

Democracy, Inequality, and Representation

always obvious, and its causes are complex. Thus, even though inequality is indeed on the rise in many countries, the magnitude and pace of the increase vary across forms of income, countries, and periods. For example, while the inequality of disposable incomes increased in the United States and the United Kingdom in the 1980s, and in Sweden and Finland in the 1990s, it changed little in Canada, France, and Germany, and over the past twenty years it showed no clear trend at all in Italy. Moreover, the nature of trends in inequality depends to a significant degree on the definition of income and, in particular, on whether taxes and benefits are included in calculating individual incomes. Establishing whether inequality is high or low and rising or declining is as critical as explaining differences and trends and as difficult as understanding the consequences of inequality. Forces such as the integration of capital and financial markets, skill-biased technological change, and deindustrialization are frequently called upon to explain rising inequalities. At the same time, they provide a similar environment for all advanced industrial societies and thus cannot be solely responsible for differences in trends and levels in economic inequality. To be sure, these structural economic transformations are engines behind the observable rise of market income inequalities throughout the Northern Hemisphere, and they have a great deal to do with how the tide rises, determining both who is lifted and who is likely to sink (Esping-Andersen 2007; Esping-Andersen et al. 2002; Rowthorn and Ramaswamy 1997, 1998; Wood 1994). Yet the extent to which these similar underlying forces shape the distribution of income within advanced societies varies significantly (Atkinson 1999, 2000; Schwabish, Smeeding, and Orsberg 2003). Indeed, while some tides bring about ever more inegalitarian outcomes, others are contained through a complex system of political and institutional channels. In fact, at a time when markets appear to be generating ever more unequal outcomes, there is striking variation in the willingness and ability of advanced societies to mute the impact that market forces have on the distribution of disposable income. Thus, in the United States, those at the very top of the income distribution continue to enjoy the fruits of economic growth (Piketty and Saez 2006), while the European political economies produce more egalitarian distributions of income (Kenworthy 2004; Pontusson 2005; Kenworthy and Pontusson 2005). We argue in the chapters that follow that the forces behind these cross-national differences are political and institutional—that is, that they are the product of the democratic political process and of how political institutions work to produce economic outcomes. In turn, levels of equality or inequality feed back into the processes of democratic representation. Insofar as politics is about “who gets what,” the distribution of income becomes an important factor shaping the pref-

Income Inequality and Democratic Representation

5

erences of voters, parties, and politicians as well as their dispositions toward the political process and their actions in it. Take, for example, the relationship between people’s relative income positions and their redistribution preferences for understanding how different distributions of income can engender different patterns of political behavior. In more unequal societies, the fact that a greater number of citizens are relatively far removed from the middle of the income distribution produces incentives to act in particular ways, whether by engaging in politics or exiting from the political process altogether. But inequality affects democratic political processes much more broadly and fundamentally, including the choice of political regime, the selection of fiscal structures, parties’ mobilization strategies, and the decision to turn out to vote. Thus, inequality is political and institutional not only in its origins but also in its consequences. This idea is at the core of this volume. The claim that inequality is as political as it is economic is hardly new, of course. More than fifty years ago, the economist Simon Kuznets concluded his famous address “Economic Growth and Income Inequality” by pointing to the centrality of politics and institutions for understanding inequality:

It is imperative that we become more familiar with findings in those related social disciplines that can help us understand population growth patterns, the nature and forces in technological change, the factors that determine the characteristics and trends in political institutions, and generally patterns of behavior of human beings—partly as biological species, partly as social animals. Effective work in this field necessarily calls for a shift from market economics to political and social economy (1955, 28).

The Political Foundations of Inequality: The Median Voter and Beyond This book takes up Kuznets’s call and builds on research in different corners of the social sciences that have provided important insights into the political processes shaping the diversity in income distributions among advanced industrial societies. Theoretically and very generally speaking, these research streams on the political economy of inequality have revolved around three elements: the median voter model of redistribution, the partisan motivations of incumbents and their impact on fiscal policy and distributive outcomes, and the role of institutions, particularly corporatism, as constraints on the set of policies available to policymakers. The median voter model of redistribution (Meltzer and Richard 1981;

6

Democracy, Inequality, and Representation

Roberts 1977; Romer 1975) has been very influential for understanding the democratic politics of redistribution and inequality. It applies the Downsian model of political competition to analyses of the redistributive conflicts among voters with different incomes. The model assumes that redistribution operates through a linear income tax with an intercept. This reduces redistribution to a single dimension, which in turn allows for single-peaked preferences and the application of the median voter theorem. Given the fact that income distributions are skewed to the right, the preferred amount of redistribution is a function of the relative position of the median voter in the income scale: the larger the distance between the income of the median voter and the average (mean) income in the society, the larger the preferred amount of redistribution. Thus, to predict voters’ preferences for more or less redistribution we need to know voters’ own relative income position, as well as society’s mean and median incomes. The model’s simplicity and the formal tractability of the arguments are well known and valuable. But, as subsequent research has shown, these advantages also come at a high analytical price, in large part because the model oversimplifies the nuts and bolts of the politics of redistribution. In fact, for a number of reasons, reducing distributive politics to contentions over one specific policy tool in a single dimension makes it difficult to account for the considerable variation in policies and distributive outcomes across advanced industrial societies. We discuss here three major analytical implications of the median voter approach to redistribution that get in the way of developing a comprehensive picture of the politics of inequality in democratic societies. A first major implication of the median voter model concerns the extent to which policy reflects contending preferences in society. If the median voter in fact rules the democratic game, there is no good reason why policies should exhibit partisan traits. Yet if there is one robust empirical finding in the comparative welfare state and political economy literatures, it is that government partisanship (ideology) has persistently shaped both how much of societal income is redistributed and, perhaps more important, how this share of income is redistributed. Since the seminal contributions of Douglas Hibbs (1977, 1987) and James Alt (1985) on the relationship between political parties and macroeconomic outcomes, this logic has been applied to many policy realms that relate directly to the distribution of income, including human capital formation policies (Boix 1998), fiscal policy (Alesina and Rosenthal 1995; Cusack 1999; Franzese 2002a), labor market policies and outcomes (Rueda 2005), social welfare spending (Allan and Scruggs 2004; Cameron 1978; Garrett 1998; Huber and Stephens 2001; Korpi and Palme 2003; Kwon and Pontusson 2007; Swank 2002; Wilensky 2002), and tax policy (Bera-

Income Inequality and Democratic Representation

7

mendi and Rueda 2007; Cusack and Beramendi 2006; Ganghof 2006; Swank and Steinmo 2002).1 In line with the power resource theory, and more directly focusing on income distribution and redistribution, a recent paper by David Bradley and his colleagues concluded that “leftist government very strongly drives the redistribution process directly by shaping the distributive contours of taxes and transfers and indirectly by increasing the proportion of GDP devoted to taxes and transfers” (2003, 225; see also Korpi 1983; Stephens 1979). Similarly, David Brady (2003) found that the strength of the left has a significant negative impact on the observable levels of poverty in advanced societies. Aside from assuming away the role of partisan preferences in distributive conflicts, median voter approaches also are notably institution-free. In particular, they neglect the influence of institutions as long-standing and sticky political settings that shape the conflicts among different political and economic interests and that translate these conflicts into public policies. This observation opens a second front on which the comparative research on inequality and redistribution departs from the median voter model of redistribution in important ways. As in the case of government partisanship, a considerable amount of research has argued that variations in inequality, both cross-nationally and over time, have pervasive and predictable institutional roots. By way of illustration, we focus on two types of institutions: economic institutions, that is, so-called corporatist arrangements for the representation of economic interests; and electoral institutions, the systems of representation that translate citizens’ preferences into political outcomes. Consider first the direct representation of economic interests in policymaking as an institutional feature that shapes redistributive outcomes—a feature that is common to the Scandinavian and continental European countries and that has been much written about by comparative politics scholars. It is well established that the degree of wage coordination between capital and labor constitutes a critical element of the organization of the economy. In fact, this is conventionally regarded as a crucial aspect of the difference between liberal and social market economies (Hall and Soskice 2001; Pontusson 2005). Wage coordination is a key aspect of a broader institutional setting in the form of corporatism,2 which is generally defined as “various types of institutional arrangements whereby important political-economic decisions are reached via negotiation between or in consultation with peak-level representatives of employees and employers (and/or other interest groups and the state)” (Kenworthy 2004, 11; see also Traxler 1999). Distributive issues are central to these agreements (Cameron 1984; Mares 2006; Regini 1984; Swenson and Pontusson 2000; Wallerstein, Golden, and Lange 1997). A highly stylized “ideal type” works as follows.

8

Democracy, Inequality, and Representation

Through national labor union organizations, and in negotiation with employers and government, labor agrees to restrain wage demands. This wage restraint contributes to lower inflation and better economic conditions, but, most important from the perspective of labor, it guarantees a degree of income insurance for workers. Government uses fiscal policy to compensate labor for its sacrifice and thereby reduces the costs of the compromise. It does this through a large welfare state that provides labor with an insurance system that guarantees both a good income during periods of economic downturns and predictable longer-term earnings (in the form of pensions). In addition, labor unions obtain the capacity to ensure an egalitarian wage distribution and political control over the implementation of a large number of public policies. As a result, employers agree to accept solidaristic wage policies and a large welfare state. Employers benefit from their coordination with labor and tolerance of welfare states because they are thus able to avoid the disruptions in production associated with industrial disputes. In addition, the welfare state benefits employers by contributing to the maintenance of a high level of human capital in the form of a labor force with highly developed specific skills (Estevez-Abe, Iversen, and Soskice 2001; Iversen and Soskice 2001). In short, high levels of wage coordination imply that, in exchange for wage moderation on the part of labor, capitalists accept that the government (together with the unions) develops a large, very costly, public insurance system. As several of the contributions to this volume analyze in detail, this kind of institutional arrangement affects the distribution of income through mechanisms that are largely overlooked by the simple median voter model. To begin with, the institutionalization of wage moderation through wage-setting institutions directly compresses the wage distribution. In turn, observable increases in wage inequality since the mid1990s can be seen as a direct result of the demise of corporatist wagesetting structures (Rueda and Pontusson 2000; Wallerstein 1999). In addition, broad cross-class support for public insurance systems largely financed by labor facilitates high levels of income redistribution between high- and low-earnings workers through the tax and transfer system (Beramendi and Rueda 2007; Cusack and Beramendi 2006; Kenworthy 2004; Moene and Wallerstein 1999; Pontusson 2005). Needless to say, these processes are not part of the median voter model. More important, the model is oblivious to both the general structure and the details of the political architecture behind the origins and evolution of this institutional setting, a topic to which we return later. Similarly, median voter accounts constrain our understanding of the role of other political institutions, such as electoral institutions, in the

Income Inequality and Democratic Representation

9

pursuit of equality through redistribution. The median voter model theorizes redistribution as a conflict between two parties under plurality electoral rules, such as those found in the United States. Yet most countries do not approximate the American scenario. What is more, given the diversity of electoral institutions around the world and the differential incentives they entail for voters, political parties, and governments, the applicability of the median voter model in international comparisons is severely limited, since it cannot explain why some countries redistribute more than others. In fact, considerable research showed that alternative systems for aggregating voters’ preferences generate policies that respond to the median voter’s preferences to different degrees (McDonald and Budge 2005; Hibbs 1992). And when it comes to the issue of income inequality, David Austen-Smith (2000) showed that, in proportional representation systems, it is not clear where in the income distribution the critical voter is. In a related paper, Austen-Smith and Jeffrey Banks (1988) also showed how proportional representation shapes the bargaining strategies of potential coalition parties and, in turn, the policy expectations and political choices of voters. In contrast to majoritarian systems, proportional representation is expected to result in policy outcomes that reflect policy positions further away from the median voter, thereby allowing for sharper partisan effects on public policy (but see McDonald and Budge 2005). And recent work that evaluates the relationship between electoral rules and redistribution concludes that majoritarian elections cut welfare spending by about 2 to 3 percent of GDP (Persson and Tabellini 2003, 179). Not surprisingly, the median voter model’s neglect of the role of partisanship and political and economic institutions speaks directly to the model’s ability to account for observable facts. Consider the chief comparative statics of the median voter model of redistribution, namely, that a more unequal (pretax) income distribution should lead to higher levels of redistribution. Contrary to this expectation, the empirical literature suggests that the countries with more equal income distributions actually have more generous and encompassing redistributive systems (Iversen and Soskice 2001; Lindert 2004; Moene and Wallerstein 2003). Although the literature on partisanship and institutions we have just reviewed has made substantial contributions to our understanding of this pattern, much work remains to be done to give a full account of the observed associations between political agency institutions, public policies, and distributive outcomes across advanced industrial societies. We turn now to a discussion of the most pressing theoretical and empirical issues and the ways in which the approach adopted in this book helps to illuminate them.

10

Democracy, Inequality, and Representation

The Next Step: Inequality, Representation, and Redistribution The inability of the median voter model of redistribution to explain the differences in income inequality across advanced industrial societies is a symptom of a more general and largely unresolved issue in the field, namely, the need to account, both theoretically and empirically, for the fact that the politics of redistribution is multidimensional. Consider, for instance, the explanatory scope of unidimensional partisan (left-right) scales. As reflected by a rich literature on welfare state regimes, reducing the size of government or the scope of redistribution to a simple left wing dimension limits our ability to account for the heterogeneity of fiscal policies and redistributive outcomes observable in continental Europe (Esping-Andersen 1990; Estevez-Abe, Iversen, and Soskice 2001; Iversen 2005). As Torben Iversen (2005) pointed out, this is a problem that concerns partisan and institutional approaches as well, and it is a problem that remains at the heart of current and future research efforts. The limitations of accounts based solely on government partisanship are several. First, partisan preferences are not necessarily unidimensional, since redistribution triggers multiple cleavages other than income, including gender, race, religion, and center-periphery relations, to name just a few. Indeed, recent theoretical work has shown that our ability to understand puzzling aspects of distributive politics increases once some of these dimensions are incorporated into the model. Examples of this include the long-standing puzzle of why the rich do not expropriate the poor under democracy (Roemer 1999, 2001), the roles of income, risk aversion, and skills specificity in shaping citizens’ preferences for the redistribution of workers (Iversen and Soskice 2001; Moene and Wallerstein 2001) and employers (Mares 2003; Swenson 2002), and the interplay between race and the redistributive preferences of different income groups (Alesina and Glaeser 2004; Austen-Smith and Wallerstein 2003). Yet, even if we were able to track analytically as many dimensions of partisan politics as there actually are, important aspects of the politics of distribution would remain out of reach. This brings us to a second, more important disadvantage of focusing our attention exclusively on partisanship. To the extent that institutions constrain the options and resources of political actors, it is reasonable to expect that political parties will vary their strategies under different institutional regimes. Yet institutions alone can hardly account for the variation in policies and outcomes across advanced industrial societies, as reflected, for instance, by the literature on wage inequality. Although wage-setting institutions have direct distributive effects (Wallerstein 1999), they also shape citizens’ will and ability to pursue alternative strategies and reconcile competing

Income Inequality and Democratic Representation

11

macroeconomic outcomes (Boix 1998; Iversen and Wren 1998; Lange and Garrett 1985; Rueda and Pontusson 2000). Overall, then, pure partisan models and strictly institutional accounts advance our understanding of the problems at hand considerably, but they both fall short of qualifying as the ultimate explanation of observable facts. Figure 1.1 places these two examples within a more general picture of the politics of distribution and outlines the main steps involved in the democratic process that determine the distribution of income in society that we address in this volume. For a variety of reasons ranging from income, labor market status, or gender to race, age, or skill level, citizens have very different interests at stake and thus develop very diverse preferences on this issue. Although some of them, by choice or by constraint, opt out of the democratic process, a majority (of varying size across countries) become the target of political parties’ mobilization efforts. Parties gather support for contending ideological platforms. Their goals include a mixture of rent-seeking and truly ideological motives, but they all seek office, either alone or as part of a broader coalition. Whether they seek office alone or with other parties largely depends on the incentives created by the electoral system and the other political institutions involved in the aggregation of citizens’ interest (for example, federalism). In turn, policy reflects not only the decisions reached within the system of political representation but also the pattern of relationships with both capital and labor, as reflected by different systems of representation of economic interests. Assuming that political actors know the political environment they operate in and have some anticipation of other actors’ expectations and strategies, this three-level interplay between political parties, political institutions, and the representation of economic interests constitutes the foundation of the democratic process of redistributive politics and is unlikely to be linear or straightforward. Indeed, the systematic analysis of this complex set of relationships involves figuring out how actors and institutions together shape the different components of the distribution of income. This puts processes of representation of varying political and economic interests at center stage of the analysis. The next hurdle, then, is to flesh out the causal sequence at work in the different linkages captured in figure 1.1. The literatures on growth and macroeconomic policies offer good examples of how to tackle some of these linkages. For example, Peter Lange and Geoffrey Garrett’s (1985) work on the politics of economic growth helps us to gain a better understanding of the political mechanisms behind the economic fortunes of advanced industrial societies by illustrating how partisanship and institutions together do the work of shaping economic outcomes. Likewise, following on Hibbs’s path-breaking con-

12 Figure 1.1

Democracy, Inequality, and Representation The Democratic Politics of Distribution

Citizens' Preferences Over n Dimensions

Political Involvement

Political Parties

Political Institutions

Economic Institutions

POLICY

Labor Market Income Market Income

Distribution of Disposable Income

Source: Authors’ compilation.

tributions on the partisan origins of macroeconomic outcomes, the next generation of scholarship turned its attention to several areas of institutional conditional effects (Alt 2002; Franzese 2002b), as illustrated, for instance, by research that shows how different combinations of monetary and labor market institutions generate different outcomes in terms of inflation and unemployment rates (Hall and Franzese 1998; Iversen 1999). In line with this institutional revival in comparative political economy, Torben Iversen and Ann Wren (1998) and David Rueda and Jonas Pon-

Income Inequality and Democratic Representation

13

tusson (2000) illustrated the benefits of unpacking the working of different political and institutional configurations as a way to shed light on the determinants of earnings inequality. Although Iversen and Wren were not primarily concerned with explaining earnings inequality, their argument speaks directly to its political and institutional foundations. In their argument, nations can maximize two of the following three goals: earnings equality, employment growth, and fiscal discipline. Their analysis reveals that, because of the interaction between these three dimensions, there is no combination of policies and institutional choices able to maximize all three. Neoliberal regimes maximize employment and growth at the expense of equality. Scandinavian nations sacrifice fiscal discipline to overcome the trade-off between growth and equality. Finally, continental European countries combine budgetary restraints and wage equality, at the expense of employment growth. After identifying these combinations, Iversen and Wren proceeded to unpack their links to different political economic configurations. The winners in neoliberal regimes are the middle and upper-middle classes in secure, paid jobs, whereas the losers are in the low-pay, low-skill service class (see also Esping-Andersen 1993). In turn, continental types are built around a growing divide between labor market insiders and outsiders.3 Finally, Scandinavian regimes are built around the interests of public sector workers and private sector employees at the lower half of the wage distribution, an orientation that generates increasing tensions over tax policy choices with those at the upper end of the wage distribution.4 In sum, different levels of earnings inequality reflect complex institutional configurations involving the organization of labor market institutions, labor market regulations, the size of the public sector, and taxation policy. In turn, Rueda and Pontusson’s (2000) investigation of the extent to which partisan effects are contingent on the institutional differences between liberal and social market economies points to two critical issues in the study of earnings inequality: first, the need to theorize the mechanism driving the interplay between political parties and economic institutions in shaping inequality; and second, the need to disentangle the marginal effects of parties’ actions from the cumulative effects of institutions. Several contributions to this volume take on these challenges and advance existing knowledge on these issues. These contributions focus not only on the interplay between partisanship and economic institutions (see the chapter by Beramendi and Cusack and the chapter by Rueda) but also on the interplay between partisanship and electoral institutions (Iversen and Soskice). As with economic institutions, the idea is to disentangle the mechanisms through which different political and institutional combinations shape the diversity of outcomes in advanced industrial societies. This includes improving our grasp of the causal sequence of the

14

Democracy, Inequality, and Representation

relationship between political actors and institutions with a focus not only on the distribution of earnings but also on the distributions of market and disposable income inequalities. By mapping out when and how different combinations of political parties and institutions matter for channeling political and economic interests, and for what type of inequality, this volume aims to advance the research frontier in the study of inequality. In their efforts to tackle the political and institutional complexity underlying the production of income inequality in the advanced economies, contributors to this volume focus primarily on the partisan mobilization of citizens’ interests and their interaction with different representative institutions, both political and economic. When it comes to the study of distributive conflicts, this opens a new and critical flank of research in that neither citizens’ involvement nor partisan positions nor the very choice of representative institutions themselves can be assumed to be ex ante independent of the existing distribution of income in society. Thus, income inequality may well be a factor shaping the political processes and the formation of specific institutional constellations that we see as shaping inequality. (Figure 1.1 includes this possibility through the arrow linking the distribution of income to the beginning of the process driving the democratic politics of distribution.) The issue of endogeneity is critical in the study of the political economy of inequality, and one that informs ongoing efforts in institutional theory (Przeworski 2004, 2007) because it suggests the need to disentangle the exogenous effects attributable to political factors such as voter turnout, government partisanship, and institutions (proportional representation, federalism, corporatism) on inequality from the conditions that influence these very factors and institutions. This new agenda is already bearing fruit in creating a better understanding of the origins of political regimes (Acemoglu and Robinson 2006; Boix 2003) and the relationship between federalism, decentralization, and inequality (Beramendi 2008). However, its full implications for advancing our understanding of the politics of inequality are yet to be developed. By treating the relationship between representation and inequality as a two-way street, this volume specifies the mechanisms at work in the feedback sequence of these relationships (see figure 1.1). We focus on how inequality shapes four crucial elements of the democratic process: preferences for redistribution, incentives to participate in politics, partisan polarization, and the formation and design of political institutions themselves. By focusing on these four elements, this book offers a systematic attempt to better identify the nuts and bolts through which inequality shapes the democratic process, thereby contributing to a better understanding of the relationship between democratic politics and distributive outcomes in advanced industrial societies.

Income Inequality and Democratic Representation

15

We turn now to a detailed outline of how the book copes with these issues of multidimensionality and endogeneity.

The Plan of the Book Aside from our theoretical understandings about the connections between inequality and democracy, our ability to compare levels of inequality and redistribution empirically across countries and over time hinges critically on issues of data quality (Atkinson and Brandolini 2001). Grand claims about the role of politics in inequality, or vice versa, in the absence of reliable data to substantiate such claims are simply insufficient. Yet far more often than is desirable, researchers base their conclusions on cross-national comparisons of a mix of different income concepts, measured across different units of analysis or definitions of the population of reference and at different moments of the politicoeconomic process (before or after taxes). If overlooked, such practices render inferences about the politics of distribution a futile exercise in which sophisticated theoretical arguments are evaluated against questionable evidence. Indeed, our collective goal in this book of addressing the issues of multidimensionality requires us to pay particular attention to how we define and measure inequality, not least because it is the central empirical regularity for all the contributions to this volume (as either the dependent or independent variable of interest). To perform the crucial task of capturing how different types of inequality vary across space and time, Andrea Brandolini and Timothy M. Smeeding (chapter 2) discuss differences in income concepts and measurement and compare trends in economic inequality across the industrialized nations. They show that the United States had the highest overall level of inequality of any rich OECD nation in the mid-1990s, while Northern and Central European countries had the lowest levels. Using a variety of series from published and unpublished data, Brandolini and Smeeding show that there was no common trend in the distribution of incomes during the last quarter-century, thereby undermining arguments for cross-national convergence in the dynamics of inequality. While the inequality of disposable incomes increased in the United States and the United Kingdom in the 1980s and in Sweden and Finland in the 1990s, it changed little in Canada, France, and West Germany and showed no clear trend in Italy. The trends that are observed appear to depend on the definition of income and, in particular, on whether taxes and benefits are included. Brandolini and Smeeding also offer estimates of the effects of government policies and social spending efforts on inequality.

16

Democracy, Inequality, and Representation

In line with these efforts to characterize more precisely the core concepts of interest in the comparative study of inequality and welfare states, Lyle Scruggs (chapter 3) traces the evolution of welfare generosity over the last three decades in eighteen OECD countries and offers a more accurate image of the dynamics of public policy interventions. This new measure of generosity suggests that, even though welfare states at the beginning of the twenty-first century were more generous than they were in the early 1970s, many have become much less generous since the mid-1980s. Indeed, the countries that have shown the most program retrenchment have been those traditionally considered the most generous. The second half of the chapter links this evolution of generosity to differences and changes in income inequality. Scruggs compares the effects of program generosity against conventional spending measures that have been used in several recent studies and shows that the benefit generosity index is not only conceptually a more valid approach to defining the welfare state than are spending measures, but also a better empirical predictor of income redistribution. The analyses by Brandolini and Smeeding and by Scruggs situate the redistributive impact of the public budget at center stage, paving the way toward more analytical questions on the political and institutional origins of budgetary choices across advanced industrial societies. This brings us back to the issues of multidimensionality and endogeneity. The first set of contributions (chapters 4 to 6) focuses on the political origins of inequality. Specifically, these chapters examine how different aspects of the representation of political and economic interests shape inequality. The authors of these chapters make an effort to clarify why and how these factors matter in shaping income inequality across nations and over time. Torben Iversen and David Soskice (chapter 4) focus on the role of electoral institutions in distributive politics. Noting that standard political economy models of redistribution—notably that of Allan Meltzer and Scott Richard (1981)—fail to account for the remarkable variance in government redistribution across democracies, Iversen and Soskice show that the electoral system shapes the nature of political parties and the composition of governing coalitions and thereby the levels and nature of redistribution. In particular, Iversen and Soskice contend that center-left governments dominate under proportional systems, while center-right governments dominate under majoritarian systems. As a result, proportional representation systems, they argue, redistribute more than do majoritarian systems. In chapter 5, Pablo Beramendi and Thomas R. Cusack examine the role of economic institutions and how they, together with competing

Income Inequality and Democratic Representation

17

partisan platforms, shape income distributions across countries. Starting from the observation that OECD countries continue to be more diverse in their distributions of labor earnings and disposable income than they are in their distributions of market income, Beramendi and Cusack show that the way in which political parties are able to pursue their goals varies across forms of income. Although political parties directly affect the distribution of disposable income through their choices about fiscal redistribution, their capacity to shape the distribution of earnings is contingent on the degree of wage-bargaining coordination. By establishing the direct and indirect effects of economic institutions on the different components of the distribution of income, this chapter advances our understanding of important institutional foundations of inequality. David Rueda, in chapter 6, takes one step forward to argue that two fundamental tasks remain before we can understand the relationship between partisan government and equality: separating the effects of partisanship on policy and of policy on the economy, and assessing the influence of political agency once we account for the mediating role of institutions. Rueda illustrates the benefits of this approach by concentrating on the lower half of the wage distribution. In so doing, the chapter sheds new light on how and why corporatism mitigates (or magnifies) the influence of government partisanship on income distributions among the (relatively) poor. The following chapters change the book’s perspective to focus on the possibility that democratic representation is shaped itself by inequality as much as it shapes inequality—that is, that representation is endogenous to inequality. In particular, these chapters analyze how crucial aspects of democratic politics are conditioned by the distribution of income (or the specific dimensions thereof). The first such aspect concerns the origins of preferences for redistribution. Focusing on the critical case of external shocks, Thomas R. Cusack, Torben Iversen, and Philipp Rehm take on the task, in chapter 7, of analyzing not only how different institutional configurations condition government responses but, more importantly, how the exposure to labor market risks as well as income shape preferences for redistribution. In disentangling the latter link, this chapter brings to the fore an important and often overlooked aspect of the democratic process. In chapter 8, Robert J. Franzese Jr. and Jude C. Hays tackle the endogeneity of inequality and participation, another crucial aspect of inequality’s impact on democratic politics, by examining the reciprocal relationships between redistribution, social insurance, and citizen participation. They argue that the generosity or miserliness of the social safety net may itself affect simultaneously the efficiency of the labor market and the po-

18

Democracy, Inequality, and Representation

litical participation of society’s less fortunate, whose political participation affects the identity and thereby the income and job security status of the median voter. This implies several endogenous relationships between economic performance (employment and distribution), the social safety net, and political participation. The chapter elaborates on the theoretically expected nature of these relationships and offers empirical estimates of the resulting system of equations. Next, we focus in chapter 9 more narrowly on the relationship between income inequality and political participation. Peter Lindert (2004) placed political voice at the center of his historical analysis explaining the expansion of public expenditures. Because of this prominent role for political voice, it is crucial to establish whether and how the distribution of income shapes the conditions under which citizens engage in political action; such an analysis contributes to a much needed distinction between the effects of political voice on inequality and the role of inequality in the set of conditions under which political voice itself emerges. Working on individual and macrolevel data collected in eighteen OECD democracies, we find that income inequality at the macro level depresses electoral participation, even when we account for the potential endogeneity of redistribution to turnout. Moreover, at the level of individual citizens, we find that the effects of income differentials are linear: individuals who are below the median income in society are less likely to participate in elections, and those above the median income are more likely to do so. This chapter also reveals that overall income inequality has similar effects on people at different ends of the income distribution. In addition to shaping preferences and participation, there is a third way in which inequality affects the democratic process: by conditioning the degree of polarization among contending political platforms. As inequality increases, the distance between different views on the role of government also increases, thereby fostering party polarization. Chapter 10, by Jonas Pontusson and David Rueda, analyzes this process and elaborates on its implications for a better understanding of the relationship between democratic politics and income inequality. Pontusson and Rueda argue that left parties are particularly responsive to the interests of low-income earners, while right parties are particularly responsive to high-income earners. As these two constituencies move further away from the median voter, party polarization increases. This chapter explores a number of alternative ways of defining inequality and polarization and offers comparative empirical evidence based on a sample of industrialized democracies. Finally, a crucial step in establishing the distributive effects of representative institutions is to explore whether and how structural factors

Income Inequality and Democratic Representation

19

and preexisting distributive tensions in society shape the selection and the way in which democratic institutions are linked to the politics of distribution. In chapter 11, Ronald Rogowski and Duncan C. McRae take a first step toward elucidating the relationship between structural transformations, representative institutions, and distributive outcomes. Their chapter suggests that exogenous changes in technology, trade, or demography alter the value of factor endowments and thus change both inequality and institutions. To develop their argument, Rogowski and McRae develop a welfare-maximizing model with endogenous institutional choice that they illustrate with a historical examination of franchise extension in Europe. The conclusion recapitulates the main contributions of the volume and outlines a number of directions for future research.

Notes 1. 2. 3. 4.

For a critical view on the effect of partisanship on public policy, see William Clark (2003). This is also sometimes referred to as neocorporatism (Lembruch and Schmitter 1982). For an analysis of the implications for social democracy of the insider-outsider divide, see Rueda (2005). For evidence on the extent to which labor bears the lion’s share of the tax burden in Scandinavian regimes, see Thomas Cusack and Pablo Beramendi (2006).

References Acemoglu, Daron, and James Robinson. 2006. Economic Origins of Democracy and Dictatorship. New York: Cambridge University Press. Alesina, Alberto, and Edward Glaeser. 2004. Fighting Poverty in the United States and Europe: A World of Difference. Oxford: Oxford University Press. Alesina, Alberto, and Howard Rosenthal. 1995. Partisan Politics, Divided Government, and the Economy. New York: Cambridge University Press. Allan, James P., and Lyle Scruggs. 2004. “Political Partisanship and Welfare State Reform in Advanced Industrial Societies.” American Journal of Political Science 48(3): 493–512. Alt, James. 1985. “Political Parties, World Demand, and Unemployment.” American Political Science Review 79(4): 1016–40. ———. 2002. “Comparative Political Economy: Credibility, Accountability, and

20

Democracy, Inequality, and Representation

Institutions.” In Political Science: State of the Discipline, edited by Ira Katznelson and Helen V. Milner. New York: W. W. Norton. Atkinson, Anthony. 1999. Is Rising Income Inequality Inevitable? A Critique of the Transatlantic Consensus. United Nations University World Institute Development Economic Research (UNU/WIDER) lecture series 3. Helsinki: UNU. ———. 2000. Increased Income Inequality in OECD Countries and the Distributive Impact of the Government Budget. Unpublished paper. Nuffield College, University of Oxford. Atkinson, Anthony B., and Andrea Brandolini. 2001. “Promise and Pitfalls in the Use of ‘Secondary’ Data Sets: Income Inequality in OECD Countries as a Case Study.” Journal of Economic Literature (American Economic Association) 39(3): 771–99. Austen-Smith, David. 2000. “Redistributing Income Under Proportional Representation.” Journal of Political Economy 108(6): 1235–69. Austen-Smith, David, and Jeffrey Banks. 1988. “Elections, Coalitions, and Legislative Outcomes.” American Political Science Review 82(2): 405–22. Austen-Smith, David, and Michael Wallerstein. 2003. “Redistribution in a Divided Society.” Discussion paper. Evanston, Ill.: Northwestern University, Center for Mathematical Studies in Economics and Management Science. Beramendi, Pablo. 2008. “Fragmented Solidarity: Inequality, Redistribution, and Endogenous Fiscal Institutions.” Book manuscript. Department of Political Science, Duke University. Beramendi, Pablo, and David Rueda. 2007. “Social Democracy Constrained: Indirect Taxation in Industrialized Democracies.” British Journal of Political Science 37(4): 619–41. Boix, Carles. 1998. Political Parties, Growth, and Equality. New York: Cambridge University Press. ———. 2003. Democracy and Redistribution. New York: Cambridge University Press. Bradley, David, Evelyne Huber, Stephanie Moller, Francois Nielsen, and John D. Stephens. 2003. “Distribution and Redistribution in Postindustrial Democracies.” World Politics 55(2): 193–228. Brady, David. 2003. “The Politics of Poverty: Left Political Institutions, the Welfare State and Poverty.” Social Forces 82(2): 557–88. Cameron, David. 1978. “The Expansion of the Public Economy: A Comparative Analysis.” American Political Science Review 72(4): 1243–61. ———. 1984. “Social Democracy, Corporatism, Labor Quiescence, and the Representation of Economic Interests in Advanced Capitalist Societies.” In Order and Conflict in Contemporary Capitalism, edited by John Goldthorpe. New York: Oxford University Press. Clark, William R. 2003. Capitalism, Not Globalism. Ann Arbor, Mich.: University of Michigan Press. Cusack, Thomas R. 1999. “Partisan Politics and Fiscal Policy.” Comparative Political Studies 32(4): 464–86.

Income Inequality and Democratic Representation

21

Cusack, Thomas R., and Pablo Beramendi. 2006. “Taxing Work.” European Journal of Political Research 45(1): 43–75. The Economist. 2006. “The Rich, the Poor, and the Growing Gap Between Them.” The Economist 379(8482): 25. Esping-Andersen, Gøsta. 1990. The Three Worlds of Welfare Capitalism. Princeton, N.J.: Princeton University Press. ———, editor. 1993. Changing Classes. London: Sage Publications. ———. 2007. “Sociological Explanations of Changing Income Distributions.” American Behavioral Scientist 50(5): 639–58. Esping-Andersen, Gøsta, Duncan Gallie, Anton Hemerijck, and John Myles. 2002. Why We Need a New Welfare State. Oxford: Oxford University Press. Estevez-Abe, Margarita, Torben Iversen, and David Soskice. 2001. “Social Protection and the Formation of Skills: A Reinterpretation of the Welfare State.” In Varieties of Capitalism, edited by Peter Hall and David Soskice. Oxford: Oxford University Press. Franzese, Robert J. 2002a. Macroeconomic Policies of Developed Democracies. New York: Cambridge University Press. ———. 2002b. “Electoral and Partisan Cycles in Economic Policies and Outcomes.” Annual Review of Political Science 5: 369–421. Ganghof, Steffen. 2006. The Politics of Income Taxation. Monograph series. Essex, UK: European Consortium for Political Research (ECPR). Garrett, Geoffrey. 1998. Partisan Politics in the Global Economy. New York: Cambridge University Press. Hall, Peter, and Robert J. Franzese. 1998. “Central Bank Independence, Coordinated Wage Bargaining, and European Monetary Union.” International Organization 52(Summer): 505–35. Hall, Peter, and David Soskice. 2001. Varieties of Capitalism. Oxford: Oxford University Press. Hibbs, Douglas. 1977. “Political Parties and Macroeconomic Theory.” American Political Science Review 71(4): 1467–87. ———. 1987. The Political Economy of Industrial Democracies. Cambridge, Mass.: Harvard University Press. ———. 1992. “Partisan Theory After Fifteen Years.” European Journal of Political Economy 8(3): 361–73. Huber, Evelyne, and John Stephens. 2001. Development and Crisis of the Welfare State. Chicago, Ill.: University of Chicago Press. Iversen, Torben. 1999. Contested Economic Institutions. New York: Cambridge University Press. ———. 2005. Capitalism, Democracy, and Welfare. New York: Cambridge University Press. Iversen, Torben, and David Soskice. 2001. “An Asset Theory of Social Policy Preferences.” American Political Science Review 95(4): 875–95. Iversen, Torben, and Ann Wren. 1998. “Equality, Employment, and Budgetary

22

Democracy, Inequality, and Representation

Restraint: The Trilemma of the Service Economy.” World Politics 50(July): 507–46. Kenworthy, Lane. 2004. Egalitarian Capitalism. New York: Russell Sage Foundation. Kenworthy, Lane, and Jonas Pontusson. 2005. “Rising Inequality and the Politics of Redistribution in Affluent Countries.” Perspectives on Politics 3(3): 449–71. Korpi, Walter. 1983. The Democratic Class Struggle. London: Routledge Press. Korpi, Walter, and Joakim Palme. 2003. “New Politics and Class Politics in the Context of Austerity and Globalization: Welfare State Regress in Eighteen Countries, 1975–95.”American Political Science Review 97(3): 425–46. Kuznets, Simon. 1955. “Economic Growth and Income Inequality.” American Economic Review 45(1): 1–28. Kwon, Hyeok Yong, and Jonas Pontusson. 2007. “Unions, Globalization, and the Politics of Social Spending Growth in OECD Countries, 1962–2000.” Unpublished paper. Department of Politics, Princeton University. Lange, Peter, and Geoffrey Garrett. 1985. “The Politics of Growth: Strategic Interaction and Economic Performance, 1974–1980.” Journal of Politics 47(3): 792–82. Lembruch, Gerhard, and Phillippe C. Schmitter. 1982. Patterns of Corporatist Policy Making. London: Sage Publications. Lindert, Peter. 2004. Growing Public. New York: Cambridge University Press. Mares, Isabela. 2003. The Politics of Social Risks. New York: Cambridge University Press. ———. 2006. Taxation, Wage Bargaining, and Unemployment. New York: Cambridge University Press. McDonald, Michael D., and Ian Budge. 2005. Elections, Parties, Democracy: Conferring the Median Mandate. New York: Oxford University Press. Meltzer, Allan H., and Scott F. Richard. 1981. “A Rational Theory of the Size of Government.” Journal of Political Economy 89(5): 914–27. Moene, Karl Ove, and Michael Wallerstein. 1999. “Social Democratic Labor Market Institutions.” In Continuity and Change in Contemporary Capitalism, edited by Herbert Kitschelt, Peter Lange, Gary Marks, and John Stephens. New York: Cambridge University Press. ———. 2001. “Inequality, Social Insurance, and Redistribution.” American Political Science Review 95(4): 859–75. ———. 2003. “Earnings Inequality and Welfare Spending: A Disaggregated Analysis.” World Politics 55(4): 485–517. Persson, Torsten, and Guido Tabellini. 2003. The Economic Effects of Constitutions. Cambridge, Mass.: MIT Press. Piketty, Thomas, and Emmanuel Saez. 2006. “The Evolution of Top Incomes: A Historical and International Perspective.” American Economic Review, Papers and Proceedings 96(2): 200–205.

Income Inequality and Democratic Representation

23

Pontusson, Jonas. 2005. Inequality and Prosperity. Ithaca, N.Y.: Cornell University Press. Przeworski, Adam. 2004. “Institutions Matter?” Government and Opposition 39(4): 527–40. ———. 2007. “Is the Science of Comparative Politics Possible?” In The Oxford Handbook of Comparative Politics, edited by Carles Boix and Susan C. Stokes. New York: Oxford University Press. Regini, Marino. 1984. “The Conditions for Political Exchange: How Concertation Emerged and Collapsed in Britain and Italy.” In Order and Conflict in Contemporary Capitalism, edited by John Goldthorpe. Oxford: Oxford University Press. Roberts, Kevin W. S. 1977. “Voting over Income Tax Schedules.” Journal of Public Economics 8(3): 329–40. Roemer, John. 1999. “The Democratic Political Economy of Progressive Income Taxation.” Econometrica 67(1): 1–19. ———. 2001. Political Competition. Cambridge, Mass.: Harvard University Press. Romer, Thomas. 1975. “Individual Welfare, Majority Voting, and the Properties of a Linear Income Tax.” Journal of Public Economics 4(2): 163–85. Rowthorn, Robert, and Ramana Ramaswamy. 1997. “Deindustrialization: Causes and Implications.” Working paper 97:42. Washington: International Monetary Fund (April). ———. 1998. “Growth, Trade, and Deindustrialization.” Working paper 98:60. Washington: International Monetary Fund (April). Rueda, David. 2005. “Insider-Outsider Politics in Industrialized Democracies: The Challenge to Social-Democratic Parties.” American Political Science Review 99(1): 61–74. Rueda, David, and Jonas Pontusson. 2000. “Wage Inequality and Varieties of Capitalism.” World Politics 52(3): 350–83. Schwabish, Jonathan, Timothy Smeeding, and Lars Orsberg. 2003. “Income Distribution and Social Expenditures: A Cross-National Perspective.” Working paper 350. Luxembourg: Luxembourg Income Study. Stephens, John. 1979. The Transition to Socialism. London: Macmillan. Swank, Duane. 2002. Global Capital, Political Institutions, and Policy Change in Developed Welfare States. New York: Cambridge University Press. Swank, Duane, and Sven Steinmo. 2002. “The New Political Economy of Taxation in Advanced Capitalist Democracies.” American Journal of Political Science 46(3): 642–55. Swenson, Peter. 2002. Capitalists Against Markets. Oxford: Oxford University Press. Swenson, Peter, and Jonas Pontusson. 2000. “The Swedish Employer Offensive Against Centralized Bargaining.” In Unions, Employers, and Central Banks, edited by Torben Iversen, Jonas Pontusson, and David Soskice. New York: Cambridge University Press.

24

Democracy, Inequality, and Representation

Traxler, Franz. 1999. “The State in Industrial Relations.” European Journal of Political Research 36(1): 55–85. Wallerstein, Michael. 1999. “Wage-Setting Institutions and Pay Inequality in Advanced Industrial Societies.” American Journal of Political Science 43(3): 649–80. Wallerstein, Michael, Miriam Golden, and Peter Lange. 1997. “Unions, Employers’ Associations, and Wage-Setting Institutions in Northern and Central Europe, 1950–1992.” Industrial and Labor Relations Review 50(3): 379–401. Wilensky, Harold. 2002. Rich Democracies: Political Economy, Public Policy, and Performance. Berkeley, Calif.: University of California Press. Wood, Adrian. 1994. North-South Trade, Employment, and Inequality. Oxford: Clarendon Press.

Chapter 2

Inequality Patterns in Western Democracies: Cross-Country Differences and Changes over Time ANDREA BRANDOLINI AND TIMOTHY M. SMEEDING

There is some intuitive appeal in the idea that democracy is associated with a more equal distribution of income. By allowing for a better representation of the interest of the poorest classes in the society, democratic institutions may be instrumental in the adoption of progressive redistributive policies. Thus, in his celebrated model of an inverted-U relationship between income inequality and economic development, Simon Kuznets explained the falling part of this relationship by observing that, “in democratic societies the growing political power of the urban lower-income groups led to a variety of protective and supporting legislation, much of it aimed to counteract the worst effects of rapid industrialization and urbanization and to support the claims of the broad masses for more adequate shares of the growing income of the country” (1955, 17, italics added). Half a century later, Torsten Persson and Guido Tabellini (1994) argued that income inequality is harmful for economic growth because it ultimately leads to higher taxation and hence to a larger distortion of agents’ investment decisions. This result relies on the existence of a democratic political process, whereby redistribution and taxation are set at the levels favored by the median voter (see also Benabou 1996). In political science, a strand of research pioneered by Gerhard Lenski (1966) also takes political democracy as leading to greater equality, although the distribution of political power within the society, not structural economic change, as in Kuznets, is seen as the driving force (see Shanahan and Tuma 1994, 748). But the causal link may work in the opposite direction. Low inequality

26

Democracy, Inequality, and Representation

in the distribution of income can be seen as a prerequisite for the working of a democracy, since extreme concentration of resources limits the ability of poor persons to exercise their political rights and may eventually lead to political instability. This view was echoed in the United States, for instance, in the 1994 Economic Report of the President, which stated: “This Administration sees the combination of stagnating average incomes and rising inequality as a threat to the social fabric that has long bound Americans together and made ours a society with minimal class distinctions” (U.S. Council of Economic Advisers 1994, 26). The idea is developed by Carles Boix in Democracy and Redistribution (2003, 10) by means of a model that predicts that “increasing levels of economic equality bolster the chances of democracy”: that is, lower inequality weakens the redistributive pressures from the poorest social classes, while it pushes the cost of taxation that the rich have to bear in a democratic regime below the cost of repression they should incur in an authoritarian regime. In their recent book, Economic Origins of Dictatorship and Democracy, Daron Acemoglu and James Robinson (2006) investigated how the conflict between the rich elites and the mass of poor people drives the development of democracy and suggest that the middle class can play an important role in both the emergence and the consolidation of democracy by operating as a buffer between the rich and the poor. These hypotheses on the links between democracy and inequality (and related variants) have been empirically tested in a cross-national environment in both political science and economics (for a survey of the former, see Castles 1999). But comparability problems are formidable. Indeed, Boix (2003, 11) argued that the failure to show convincingly the empirical validity of such a link is due to “the lack of broad and reliable data sets of income inequality until very recently.” But even then, what is “reliable”? Recent compilations like the one by Klaus Deininger and Lyn Squire (1996), used by Boix, are also not without problems, as discussed, for instance, by Anthony Atkinson and Andrea Brandolini (2001). Indeed, there is a need to be sensitive to the nuances of underlying data in the estimation of the relationship between democracy and (in)equality. At the same time, whatever the supposed direction of causality in this relationship, any theory must account for the basic fact that inequality levels and trends vary widely both across democratic countries and within each country. With these considerations in mind, our aim in this chapter is to set the stage for the analysis of the link between democracy and inequality in the following chapters by reviewing the cross-national pattern and the within-country trends in economic inequality. In the next section, we compare levels of inequality in nominal and real incomes in thirty-two countries with what can be seen as a Western-type democratic system.

Inequality Patterns in Western Democracies

27

We also provide some evidence on the impact of public redistribution, the pivotal variable in the relationship between democracy and inequality in all the theories mentioned earlier. In the following section, we discuss the trends in inequality and redistribution over a period of four decades in nine rich countries. We draw some conclusions in the last section. As we begin, we remind the reader that we are both economists who are not trained in the study of political regimes or the nuanced differences in legislative politics. These domains of political economy are left to the talented authors of the next chapters in this book.

Cross-National Differences in Income Inequality We begin with the widest cross-national comparison of income inequality we can present. Figure 2.1 compares the distribution of equivalent disposable money income across persons in thirty-two nations for various years around the turn of the century, or for the most recent year available. Figures are computed from the Luxembourg Income Study (LIS) database, which provides the best source of internationally comparable data on household incomes (Smeeding 2004); it is used in half of the chapters in this volume. These figures are integrated with estimates from the European Community Household Panel database (waves 1 to 8, December 2003) for Portugal and with statistics for Japan computed according to the same methodology as all other figures by Tsuneo Ishikawa (1996; see also Gottschalk and Smeeding 2000). Following World Bank (2005) categorization, countries are separated into high-income and middle-income economies according to their per capita gross national income in 2004. Disposable money income is given by the sum of all cash incomes earned by the household (wages, salaries, earnings from self-employment, cash receipts from property, unemployment compensation, welfare benefits, public and private pensions, child and family allowances, alimony), net of income taxes and social security contributions.1 However broad, this definition excludes capital gains, imputed rents, other unrealized types of capital income, home production, and in-kind income. These items may account for an important share of the economic resources at the disposal of the household, and their inclusion in the income definition may affect measured inequality, as discussed later in relation to public benefits for health care, education, and housing. To account for the economies of scale stemming from cohabitation, total household income is adjusted by a simple equivalence scale, the square root of the household size: for instance, the equivalent income of a household of four is obtained by dividing total household income by two.

Figure 2.1

The Distribution of Equivalent Disposable Income in Thirty-Two Countries P10 (Low Income)

Length of bars represents the gap between highand low-income individuals

P90 P90/P10 (High (Decile Gini Income) Ratio) Index

High-income economies Denmark 2000 57 Norway 2000 57 Finland 2000 57 Sweden 2000 57 Netherlands 1999 56 Slovenia 1999 53 Austria 2000 55 Luxembourg 2000 57 Belgium 2000 53 Switzerland 2000 55 Germany 2000 54 France 2000 55 Taiwan 2000 52 Canada 2000 48 Japan 1992 46 Australia 2001 47 Italy 2000 45 Ireland 2000 41 United Kingdom 1999 47 Greece 2000 43 Spain 2000 44 Israel 2001 43 Portugal 2000 45 United States 2000 37 Middle-income economies Slovak Republic 1996 56 Czech Republic 1996 59 Romania 1997 53 Hungary 1999 54 Poland 1999 52 Estonia 2000 46 Russia 2000 33 Mexico 2000 32 0

155 159 164 168 167 167 173 184 174 182 180 188 196 188 192 199 199 189 215 207 209 216 226 212

2.8 2.8 2.9 3.0 3.0 3.2 3.2 3.2 3.3 3.3 3.4 3.4 3.8 3.9 4.2 4.2 4.5 4.6 4.6 4.8 4.8 5.0 5.0 5.7

0.225 0.251 0.247 0.252 0.248 0.249 0.260 0.260 0.277 0.280 0.275 0.278 0.296 0.302 0.315 0.317 0.333 0.323 0.343 0.338 0.340 0.346 0.363 0.370

162 179 180 194 188 234 276 331

2.9 3.0 3.4 3.6 3.6 5.1 8.4 10.4

0.241 0.259 0.277 0.295 0.293 0.361 0.434 0.491

50 100 150 200 250 300 350

Source: Authors’ calculations from the Luxembourg Income Study (LIS) database, as of March 10, 2007 (figures coincided with those then reported in http://www.lisproject.org/keyfigures/ineqtable.htm), and, for Portugal, from the European Community Household Panel database (waves 1 to 8, December 2003); statistics for Japan were computed according to the same methodology as all other figures by Ishikawa for Gottschalk and Smeeding (2000). Note: P10 and P90 are the ratios to the median of the tenth and ninetieth percentiles, respectively. Observations are bottom-coded at 1 percent of the mean of equivalent disposable income and top-coded at ten times the median of unadjusted disposable income. Incomes are adjusted for household size by the square-root equivalence scale. Economies are classified by the World Bank (2005) according to 2004 per capita gross national income in the following income groups: low-income economies (LIC), $825 or less; lower-middle-income economies (LMC), $826 to $3,255; upper-middle-income economies (UMC), $3,256 to $10,065; and high-income economies (HIC), $10,066 or more.

Inequality Patterns in Western Democracies

29

This value is then attributed to each person in the household to derive the distribution among persons.2

Relative Differences There is a wide range of income inequality among the nations listed in figure 2.1. The United States is an outlier among rich nations, and only Russia and Mexico, two middle-income economies, have higher levels of inequality. A low-income American at the tenth percentile has an income that is only 37 percent of the median income (P10). By contrast, in most countries of Central, Northern, and Eastern Europe, the income of the poor exceeds 50 percent of the income of a middle-income person; in the other English-speaking nations and in the Southern European countries, plus Israel, it is above 40 percent. Only in Russia and Mexico do the poor fare relatively worse than in the United States. In Greece, Portugal, Spain, Israel, the United States, and the United Kingdom, rich persons, those at the ninetieth percentile, earn more than twice the national median income (P90). In poorer countries—for example, Mexico, Russia, and Estonia—the ninetieth percentile can also be very high in relative terms. The countries in figure 2.1 fall into some distinctive clusters. Inequality, as measured by the decile ratio (the ratio between P90 and P10), is least in the Nordic countries, the Netherlands, and the Czech and Slovak Republics, with values of 3.0 or less. The two other Benelux countries (Belgium and Luxembourg), the countries of Central Europe (France, Switzerland, Germany,3 Austria, and Slovenia), and three other Eastern European countries (Hungary, Poland, and Romania) come next at 3.2 to 3.6. These precede the four English-speaking nations (Canada, Australia, Ireland, and the United Kingdom), which have decile ratios between 3.9 and 4.6, and the Southern European countries (Italy, Spain, Greece, and Portugal) and Israel, whose ratios fall between 4.5 and 5.0. Only the United States, Estonia, Mexico, and Russia have values in excess of 5.0. With decile ratios around 4.0, the two Asian countries, Taiwan and Japan, are in an intermediate position. Inequality differs much more across middle-income than high-income economies. While Estonia, Russia, and Mexico show a very unequal distribution of income, the other five countries, all from Eastern Europe, exhibit moderate or low levels of inequality. The shape of the income distribution was already noticeably different in the mid-1980s across these formerly planned economies— that is, before they turned into Western-type democracies—with Czechoslovakia showing the least inequality and the Soviet Union the highest (Atkinson and Micklewright 1992).

30

Democracy, Inequality, and Representation

In figure 2.1, countries are arranged, within the two categories of high-income and middle-income, by decile ratio from lowest to highest. More important, the country rank order does not coincide with that based on other statistics reported in the same figure: P10, P90, and the Gini index. For instance, Sweden shows the second-highest P10 but the seventh-lowest Gini index. This follows from the fact that a Swede at the ninetieth percentile is not as close to the middle as the equivalent person in Denmark, Finland, or the Slovak Republic. Although these differences may be small and are likely to be within the bounds of sampling error, one should still be aware that the exact ranking of countries in international comparisons may well depend on which part of the distribution is being analyzed. For example, the rankings with a bottom measure, P10, or the top measure, P90, may differ from the rankings that might result from single-observation summary measures of inequality, like the Gini index or the Theil and Atkinson indices. Different summary measures may lead to different orderings, since they weight differently the top and bottom of the distribution. In the same vein, the results of empirical tests are sensitive to the choice of the inequality index—as shown by Sarah Voitchovsky (2005) for the relationship between inequality and growth and by Jonathan Schwabish, Timothy Smeeding, and Lars Osberg (2006) for the relationship between inequality and social spending. A more robust, if partial, ranking is provided by comparing the entire income distributions through the analysis of Lorenz dominance as developed by Atkinson (1970). For a wide class of inequality measures, including most of those commonly used, incomes are distributed less unequally in country A than in country B if the Lorenz curve of A always lies above—that is, dominates—that of B. If the Lorenz curves intersect, the two distributions cannot be unambiguously ordered, and their ranking varies with the inequality measure.4 We compare the Lorenz curves at cumulative decile points, assuming that two curves are distinct only when their difference at some point exceeds 0.3 percentage points in order to allow, however imperfectly, for sampling variability. The results are reported in table 2.1, where the sign “+” indicates that the country in the column dominates the country in the row, the sign “–” indicates the opposite, and the sign “?” indicates that the Lorenz curves cross or are not distinguishable at any decile point according to the mentioned criterion. Thus, the two “–” signs in the last row of table 2.1 tell us that Mexico and Russia are the only countries where incomes are more unequally distributed than in the United States, while the “?” sign suggests that inequality cannot be unequivocally stated to be higher in Portugal than in the United States, or vice versa, unless an inequality measure is specified. This complex pattern of bilateral comparisons is summarized in the Hasse diagram in figure 2.2. Income inequality falls by moving from the

Inequality Patterns in Western Democracies

31

top to the bottom: a traceable line downwards from country A to country B (for example, Mexico and Estonia) implies that the Lorenz curve for country A lies below that for country B; when this is not possible, as with Israel and the United Kingdom, countries A and B cannot be unequivocally ranked. The merit of figure 2.2 is to make it explicit that many comparisons are indeed ambiguous. At the same time, it confirms the basic pattern of international inequality sketched earlier: Mexico and Russia are at the top, followed by the English-speaking countries intertwined with the Southern European countries; the other continental European nations come next, while the Nordic countries show the lowest level of inequality; Eastern European countries are spread along the entire tree.

Absolute Differences It is often argued that the higher the average standard of living in a particular nation, the better off its citizens are. By this argument, the “average” U.S. resident is better off than the residents of Italy or Finland because the U.S. real gross domestic product (GDP) per capita in 2000 was 34,300 international dollars, compared to 25,300 international dollars in Italy and Finland (International Monetary Fund 2006). But it may also be important to know whether this higher average standard of living extends to all levels of the income distribution. Indeed, political systems respond not only to relative living standards and relative levels of inequality, but also to absolute differences in incomes and their growth. The absolute threshold used in the United States to calculate official poverty statistics is a clear example of such a concern. Moreover, richer countries can afford more redistribution, and redistribution to the poor is much easier to accomplish in growing economies than in stagnant ones (Smeeding 2006). To deal with this question we must compare real incomes, that is, incomes deflated by a purchasing power parity (PPP) index. This is a standard, but crude, way of measuring the amount of goods and services that a certain income can purchase. On the one side, it is questionable whether the same conversion factor should be applied across the entire distribution, although the same concern could be raised for within-country differences in the cost of living. On the other side, real disposable income does not account for goods and services, such as education and health care, that are provided at different prices and under different financing schemes in different nations. As low-income citizens in some countries need to spend more out of pocket for these goods than do lowincome citizens in other countries, their living standard is relatively lower than that measured by PPP-adjusted income (Smeeding and Rain-

Austria Belgium Canada Czech Republic Denmark Estonia Finland France Germany Greece Hungary Ireland Israel Italy Luxembourg

Australia

– – – – – + – – – + ? + + + –

Austria

+ + ? – + – + + + + + + + ?

Belgium

+ – – + – + – + + + + + –

Canada

– – + – – – + ? + + + –

Czech Republic

– + – + + + + + + + ?

Denmark

+ + + + + + + + + +

Estonia

– – – – – – – – –

Finland

+ + + + + + + +

France

– + + + + + –

Germany

+ + + + + –

Greece

– – + ? –

Hungary

+ + + –

Ireland

+ + –

Israel

– –

Italy



Romania

Portugal

Poland

Norway

Netherlands

Mexico

Luxembourg

Lorenz Comparison for the Distribution of Equivalent Disposable Income in Thirty-Two Countries

Russia

Table 2.1

United Kingdom Taiwan

Switzerland Sweden Spain

Slovenia

Slovak Republic

+ – – ? + – + – – + – – – + +

+ – ? + + + + – – + – + + + +

+ – – + + ? + – – + – + + + +

+ – – ? + – + – – + – – ? + +

+ ? ? + + + + ? ? + ? + + + +

+ + + + + + + + + + + + + + +

+ – – – + – + – – – – – – – +

+ ? ? + + + + ? ? + + + + + +

+ – – + + ? + – – + – ? + + +

+ – – + + + + – – + – + + + +

+ – – – + – + – – + – – – ? +

+ – – – + – + – – + – – ? + +

+ – – – + – + – – + – – – + +

+ – – – + – + – – – – – – ? +

+ – – – + – + – – + – – – + +

+ ? ? + + + + ? ? + ? + + + + – – – – – – – – – – – – – – ? + + + + – + + ? + + + + + + + + – ? + ? + + + + + – + – – + – – ? + + – + – – – – – – – ? + – – + – + + + +

– – – – – – – –

+ + + + + + +

+ ? + + + +

– – – ? +

+ + + +

+ + +

+ +

+

Source: Authors’ calculations. Note: Comparisons are based on cumulative decile points. A + (–) indicates that the Lorenz curve of the country shown in the row lies below (above) that of the country shown in the column, with a difference exceeding 0.3 percentage point for at least a cumulative decile group; a ? indicates that the Lorenz curves cross, or are not yet distinguishable at any point according to the defined criterion.

Mexico Netherlands Norway Poland Portugal Romania Russia Slovak Republic Slovenia Spain Sweden Switzerland Taiwan United Kingdom United States

Figure 2.2

Hasse Diagram for the Distribution of Equivalent Disposable Income in Thirty-Two Countries Mexico Russia Portugal

United States Estonia

Israel United Kingdom Spain Greece Italy Ireland Hungary

Australia

Poland Taiwan

Canada

Switzerland

France

Romania Belgium Germany Luxembourg

Austria

Norway Slovenia

Czech Republic Sweden

Netherlands Finland

Slovak Republic Denmark

Source: Authors’ calculations.

Inequality Patterns in Western Democracies

35

water 2004). Further complications arise because the PPP indices are available for various aggregates and from different sources5 and are computed for national accounts, which are intrinsically different from survey data (Deaton 2005).6 The statistics for real, equivalized incomes in 2000 international dollars are reported in figure 2.3. Original incomes are adjusted by the national consumer price indices in the case of non–base-year observations and are converted by means of PPP indices for GDP drawn from the International Monetary Fund (2006). In each country, the real P10, P90, and median are recomputed as a fraction of the U.S. median real income. Even if we are mostly considering rich nations, differences in average real living standards are huge. The median person in middle-income economies earns less than one-third of the median American—and the median Russian about one-tenth—but variation is also considerable among high-income economies: in Portugal, Slovenia, and Greece, median real income is below half of the U.S. value, while only in Luxembourg do we find that the median is higher than in the United States. But these differences do not necessarily carry forward to the rest of the income distribution. If the living standard of the median Belgian or Finn appears to be 63 and 72 percent, respectively, of that of the median American, the living standard of Belgian and Finnish poor people is roughly the same as that of their American counterparts, around 37 percent of the U.S. median. Low-income people in Denmark, Norway, Switzerland, and, especially, Luxembourg are much better off than elsewhere. In all Southern European countries, but also, to a lesser extent, in Australia, Ireland, and the United Kingdom, the living standards of lowincome households are lower than in the United States. Of course, they are a great deal lower in all middle-income economies. At the other extreme, rich Americans far surpass the rich in any other nation observed, save for the Luxembourgers. For instance, the rich American is sixty percentage points above the rich Canadian and seventy-one points above the rich Briton. The horizontal bars in figure 2.3 are proportional to the absolute distance between top and bottom incomes. The absolute gap in the United States is twice as high as in Switzerland, Taiwan, and Canada, and it is much higher than in any of the remaining countries. The claim that the United States enjoys the world’s highest living standard must be evaluated alongside the equally valid claim that the United States enjoys the greatest absolute inequality between the rich and the poor among developed countries. While the rich in the United States are truly well off by any measure of living standards, many poor Americans at the same time have living standards below those in other nations that are not as rich as the United States.

Figure 2.3

The Distribution of Real Disposable Income in Thirty-Two High- and Middle-Income Economies Length of bars represents the gap between highand low-income individuals

Real P10 (Low Income)

Real P90 P90/P10 (High (Decile Real Income) Ratio) Median

High-income economies Denmark 2000 45 Norway 2000 48 Finland 2000 36 Sweden 2000 34 Netherlands 1999 39 Slovenia 1999 23 Austria 2000 41 Luxembourg 2000 65 Belgium 2000 38 Switzerland 2000 45 Germany 2000 39 France 2000 34 Taiwan 2000 39 Canada 2000 39 Japan 1992 Australia 2001 30 Italy 2000 25 Ireland 2000 30 United Kingdom 1999 31 Greece 2000 20 Spain 2000 25 Israel 2001 23 Portugal 2000 18 United States 2000 37 Middle-income economies Slovak Republic 1996 14 Czech Republic 1996 18 Romania 1997 8 Hungary 1999 12 Poland 1999 12 Estonia 2000 9 Russia 2000 3 Mexico 2000 4

80 84 63 59 69 44 75 114 72 82 72 62 76 81

129 111 136 141 96 121 115 91 212

2.8 2.8 2.9 3.0 3.0 3.2 3.2 3.2 3.3 3.3 3.4 3.4 3.8 3.9 4.2 4.2 4.5 4.6 4.6 4.8 4.8 5.0 5.0 5.7

40 54 27 44 45 47 26 46

2.9 3.0 3.4 3.6 3.6 5.1 8.4 10.4

25 30 15 23 24 20 10 14

125 134 103 99 116 73 129 209 125 150 130 117 150 152

0

50

100

150

200

65 56 72 66 46 58 53 40 100

250

Source: Authors’ calculations from the Luxembourg Income Study (LIS) database, as of March 10, 2007, and, for Portugal, from the European Community Household Panel database (waves 1 to 8, December 2003); statistics for Japan were computed according to the same methodology as all other figures by Ishikawa for Gottschalk and Smeeding (2000). Note: Real P10 and P90 are the percentage ratios to the U.S. median of the tenth and ninetieth percentiles, respectively; real median is expressed as a percentage ratio of the U.S. median. Observations are bottom-coded at 1 percent of the mean of equivalent disposable income and top-coded at ten times the median of unadjusted disposable income. Incomes are adjusted for household size by the square-root equivalence scale. Consumer price indices and purchasing power parity conversion factors from local currency units to international dollars are from International Monetary Fund (2006).

Inequality Patterns in Western Democracies

37

Levels of Monetary Redistribution Every nation’s tax and benefit system reduces market income inequality, but not all are equally effective in doing so. The extent to which nations accomplish this task may vary over time as well as across space. A common “output” measure of the level of redistribution is represented by the difference between the Gini index for market incomes—that is, before public transfers are added and taxes and social security contributions are deducted—and the Gini index for disposable incomes. This difference provides only a crude estimate of the actual degree of public redistribution, since it assumes that market income inequality would remain the same if taxes and benefits did not exist. This is clearly unrealistic because it does not take into account how taxes and benefits encourage or discourage earnings or savings. Likewise, it disregards the different impact of programs designed to achieve redistribution: universal benefits, targeted means-tested assistance, or social insurance schemes (see also Mahler and Jesuit 2006; Smeeding 2006). Finally, this measure ignores how much redistribution is carried out through noncash programs, since it only reflects differences in cash programs (the importance of noncash programs is discussed later). Ideally, we would like to know how people would behave in a different environment with no taxes and benefits or different assistance schemes, but this would require bold assumptions and a rather complex behavioral microsimulation model. On the contrary, the difference in the Gini indices for market and disposable incomes is an intelligible, if imperfect, way to gauge the level of income redistribution in a country. As shown by LIS data, in all sixteen nations reported in figure 2.4, disposable incomes are more equally distributed than market incomes, suggesting that the tax and benefit system narrows the overall distribution. On average, inequality falls by about one-third, from a Gini index of 44 percent to one of 29 percent. Cross-country variation in original inequality is wider than after redistribution: the Gini index ranges from 33 to 52 percent for market incomes, and from 23 to 37 percent for disposable incomes. The United States has the highest inequality of disposable incomes; although the dispersion of market incomes is on the high side, it is not far from most other countries: it is as high as in Germany and Australia and below the values recorded for the United Kingdom, Poland, and Israel. The fact is that the percentage reduction in before-tax-andbenefit inequality in the United States is a mere 23 percent. If we exclude Taiwan, where redistribution has a tiny impact, only Switzerland shows a reduction as low as the United States, but the Swiss start from a much more equal distribution and end with a Gini index below the average. These percentage reductions are very consistent with the patterns of

38 Figure 2.4

Democracy, Inequality, and Representation Gini Indices of Market Income and Disposable Income in Sixteen OECD Countries (Percentage)

Denmark 2000 Finland 2000 Netherlands 1999 Norway 2000 Sweden 2000 Czech Republic 1996 Germany 2000 Romania 1997 Switzerland 2000 Poland 1999 Taiwan 2000 Canada 2000 Australia 2001 United Kingdom 1999 Israel 2001 United States 2000

23 25 25 25 25 26 28 28 28 29 30 30 32 34 35 37 0

10

20

Market Income

Reduction in Gini Indexa 42 47 36 38 36 39 39 41 45 46 41 44 43 48 27 38 22 36 41 50 9 33 28 42 34 48 33 51 52 33 23 48 30

40

50

Disposable Income

Source: Authors’ calculations from the Luxembourg Income Study (LIS) database, as of March 10, 2007. Note: Observations for disposable income are bottom-coded at 1 percent of the mean of equivalent disposable income and top-coded at ten times the median of unadjusted disposable income. Changes in disposable incomes due to bottom- and top-coding are entirely attributed to market incomes. Both market and disposable incomes are adjusted for household size by the square-root equivalence scale. a Difference between the Gini index for market income and the Gini index for disposable income expressed as a percentage of the former.

aggregate public expenditure (or non-elderly spending, see Smeeding 2005). High-spending Northern and Central European nations have the highest degree of inequality reduction, from 36 to 47 percent; the AngloSaxon nations (excluding the United States) and Israel are next, with 28 to 33 percent reductions; the United States and Switzerland are, as just seen, at the bottom of the scale. The degree of redistribution in Southern Europe is lower than in Ireland and the United Kingdom—especially if public pensions are not included among transfers—according to the EUROMOD estimates based on microsimulations rather than the records

Inequality Patterns in Western Democracies

39

of the original microdata sources (Immervoll et al. 2007). The nations that redistribute the most are not necessarily those with the greatest degree of market income inequality: before-tax-and-benefit incomes in Finland and the Netherlands are far more equally distributed than in the United States. In fact, Schwabish, Smeeding, and Osberg (2006) found almost no correlation between the P10 value for market income and the level of social spending.

In-Kind Benefits None of these estimates include benefits in kind or indirect taxes. How much difference do they make? In their study of the distribution in seven rich countries in the early 1980s, Timothy Smeeding and his colleagues (1993) found that including the value of noncash benefits in household income reinforced the redistributive impact of cash tax-and-transfer mechanisms in all countries but did not affect markedly the pattern of national differences in income inequality from that which emerged from the analysis of cash income alone. More recent analysis of ten rich countries in the late 1990s by Irwin Garfinkel, Lee Rainwater, and Timothy Smeeding (2006) confirms the egalitarian impact of noncash redistribution. After augmenting income to include the value of noncash benefits for health care and education net of both direct and indirect taxes, the income of the poor turns out to be much closer to the median, and the distance between the rich and the poor falls in all countries except Belgium and Finland. Changes are largest among the English-speaking nations, with the United States showing the greatest drop in the decile ratio. Differences across countries appear to shrink considerably. Two reasons can account for these results. First, compared to other advanced nations, the English-speaking nations tend to be short on cash and long on in-kind benefits. Thus, relatively equal noncash benefits can go a long way toward equalizing command over total resources, including more unequally distributed cash benefits and other incomes. Second, these countries rely less heavily than the big-spending national welfare states on indirect taxes and taxation of cash benefits. Together, these two factors explain the big shift when moving from cash disposable income to augmented income. These results are to be taken with caution because they depend crucially on the assumptions made to evaluate and impute noncash benefits. Although this caveat has to be borne in mind, it is clear conceptually that these benefits are worth some nontrivial amount to both rich and poor alike. Empirically, health and education transfers are as large a part of what the welfare state does for families as the provision of cash benefits

40

Democracy, Inequality, and Representation

in all nations—or an even larger part. This fact must be taken into consideration in studying the relative effectiveness and generosity of all welfare states and their effect on inequality.

Postwar Inequality Trends in Selected Rich Countries The previous section offered a snapshot of income inequality and redistribution around the turn of the century. As is well known, however, inequality has increased significantly in the last decades in several countries, considerably so in the United States and the United Kingdom (Gottschalk and Smeeding 2000). It is therefore worth considering the long-run patterns that have emerged in selected rich countries: three Anglo-Saxon nations (Canada, the United States, and the United Kingdom), two Nordic nations (Finland and Sweden), and four continental European countries (the Netherlands, West Germany, France, and Italy). The evidence at our disposal is summarized in one figure for each country. Before turning to this evidence, we must emphasize that reported series are a selection of those that are internally consistent for a sufficiently long span of time; they are not necessarily comparable across nations, nor is one comparable with another within the same nation. Furthermore, these statistics are somewhat more uncertain than those in other fields, such as national accounts.7

The United States, the United Kingdom, and Canada In the United States, pretax inequality exhibited a very sharp fall between 1929 and 1944, according to the Bureau of Economic Analysis (BEA) statistics (figure 2.5). This prewar change in income distribution was judged “for its magnitude and persistence . . . [to be] unmatched in the record” by Kuznets (1953, xxxvii)—the first to detect it on the basis of his estimates of the income shares of upper-income groups—and promptly described by Arthur Burns (quoted in Pechman 1958, 108) as “one of the great social revolutions of history.” In the following three decades, the Gini index showed some fluctuations around a flattened trend according to the BEA figures, or a moderately declining trend according to the Current Population Survey (CPS) series for families of two or more people.8 It is in the light of this evidence that we must read Robert Solow’s conclusion that the personal distribution of income “is a facet of economic life which changes slowly when it changes at all” (1960, 109–10), or Henry Joseph Aaron’s infamous remark that tracking changes in the distribution of income is “like watching the grass grow” (1978, 17).

Inequality Patterns in Western Democracies Figure 2.5

41

Gini Index in the United States (Percentage)

52 1. BEA, Gross Income 4a. CPS, Market Income

Gini Index

48

44 3. CPS, Gross Income 40

36 4b. CPS, Disposable Income

2. CPS, Gross Income 32 1925

1935

1945

1955

1965

1975

1985

1995

2005

Year Source: 1: Brandolini (1998, table Al): estimates from BEA grouped data for gross incomes of households. 2: U.S. Census Bureau (2006a), CPS data: gross money income of families; weighted by family; shown the major discontinuity between 1992 and 1993, but not other minor breaks. 3: U.S. Census Bureau (2006b), CPS data: gross money income of households (families and unattached individuals); weighted by household; shown the major discontinuity between 1992 and 1993 and the break in 2000 (for which two figures are given), but not other minor breaks. 4: U.S. Census Bureau (2006c), CPS data; a: market income including capital gains and health insurance supplements to wage and salary income of households (definition 4); b: disposable income including capital gains and health insurance supplements to wage and salary income of households (definition 15); in both cases, weighted by household; shown the major discontinuity between 1992 and 1993, but not other minor breaks.

The postwar relative immobility of the distribution lasted until the 1970s, when the United States entered a period of unrelenting increases in income inequality. According to CPS figures (both excluding and including unrelated individuals), the Gini index returned by 1980 and 1981 to the level of thirty years earlier and rose further in the following decade. An extensive methodological revision led to a major break in the CPS series between 1992 and 1993 (hence the interruptions in the figure; see Ryscavage 1995). The tendency of inequality to rise continued thereafter, however, even if at a more moderate pace. The most comprehensive CPS series for disposable income—including capital gains and noncash benefits—confirms the widening of the distribution since 1980. (The spike in 1986 was most probably driven by capital gains, which re-

42

Democracy, Inequality, and Representation

flected the performance of the stock market.) The increase in inequality from 1979 to 2001 is even more pronounced in Congressional Budget Office (CBO) data, which are adjusted for underreporting using register data and more accurate measures of capital accumulations by high-income persons (Smeeding 2005). Also in the United Kingdom, income inequality has traced a U-shape in the last sixty years, and the period around the Second World War seems to have been a watershed (figure 2.6). According to the Blue Book (BB) series relative to tax units, the Gini index fell by over five percentage points for incomes before taxes and by seven points for incomes after taxes between 1938 and 1949. The leveling of incomes continued, at a much slower pace and with some minor recrudescence, until the late 1970s, when the trend abruptly reversed. The values of the Gini index in 1984 and 1985, when the series was discontinued, were not too different from those prevailing in the aftermath of the war. The five years between 1985 and 1990 saw an unprecedented rise of income inequality, as testified by the seven-point increase of the Gini index for equivalent disposable income recorded by both the Economic Trends (ET) series and the series reconstructed for Great Britain at the Institute for Fiscal Studies (IFS). Ever since, the concentration of income has shown some changes but no sustained trend, as also confirmed by Stephen Jenkins’s (2000) estimates based on the British Household Panel Survey. As regards the distribution of market incomes, inequality steadily rose in the 1960s, 1970s, and especially 1980s, and then stabilized in the 1990s. Two features of distributive changes in the United Kingdom need to be highlighted (Atkinson 2003). The first is the mere size of the upsurge in inequality in the second half of the 1980s—far higher than in the United States. In part, this rise reflects the failure of public redistributive policies to counteract the pressure toward greater inequality generated in the economy. This is clearly illustrated by the comparison between the Gini index for equivalent market income and that for equivalent disposable income: between 1985 and 1990, the former rose by three percentage points, while the latter went up by seven points. The reforms of personal income taxation, unemployment benefits, and social assistance implemented in that period all went in the direction of widening the distribution of disposable income (see, for example, Atkinson 1993; Atkinson and Micklewright 1989; Johnson and Webb 1993). The second feature worth mentioning is that inequality stopped growing during the 1990s. What needs to be explained in the British historical pattern is not a lasting long-run tendency, but rather a single episode of extraordinary intensity that led to a new higher inequality plateau. The experience of Canada differs from that of the two other AngloSaxon countries (figure 2.7). The distribution of both pre- and post-tax

Inequality Patterns in Western Democracies Figure 2.6

43

Gini Index in the United Kingdom (Percentage)

55

2a. ET, Market Income

50 1a. BB, Gross Income

Gini Index

45 40 35

2b. ET, Disposable Income 1b. BB, Disposable Income

30 25

05 20

00 20

95 19

90 19

85 19

80 19

75 19

70 19

65 19

19

55 19

50 19

45 19

40 19

35 19

60

3. IFS, Equivalent Disposable Income

20

Year Source: 1: Official publications of the Royal Commission on the Distribution of Income and Wealth and of the National Statistical Office as detailed in Brandolini (1998, table A3), BB data; a: gross income of tax units; b: disposable income of tax units; in both cases, weighted by tax unit; the first series is for incomes net of amounts spent on mortgage interest (old basis), while the second is for incomes gross of those amounts (new basis); figures refer to calendar years until 1967 and to financial years afterwards (starting in the year indicated in the figure, for example, 1968 for 1968 to 1969); the figures for 1938 and 1949 are reconstructed and are less precisely estimated than subsequent values. 2: Official publications of the Royal Commission on the Distribution of Income and Wealth and of the National Statistical Office as detailed in Brandolini (1998, table A3) for data prior to 1980; Jones (2006,, table 27, 39) for 1980 to 2004 to 2005, data from Family Expenditure Survey (FES) until 2000 to 2001 and Expenditure and Food Survey (EFS) since 2001 to 2002; a: market income; b: disposable income; in both cases, weighted by household; the first series refers to unadjusted income, the second series to equivalent income; McClements equivalence scale; figures refer to calendar years until 1993 and to financial years afterwards. 3: Brewer et al. (2006), data from FES for 1961 to 1993 to 1994 and from Family Resources Survey (FRS) for 1994 to 1995 to 2004 to 2005: equivalent disposable income of households, before housing cost; weighted by person; McClements equivalent scale. Note: Figures refer to Great Britain alone, but in the period 1961 to 1991 they differ by at most 0.4 percentage point from the corresponding series for the whole United Kingdom computed by Goodman and Webb (1994).

monetary incomes among families and unrelated individuals did not vary much from 1965 through the late 1980s, although we may discern episodes when inequality was virtually stable (1984 to 1987), ascending (1965 to 1971, 1981 to 1983), or descending (second half of the 1970s). This “stasis amid change,” as Michael Wolfson (1986) labeled it, is note-

44

Democracy, Inequality, and Representation

Figure 2.7

Gini Index in Canada (Percentage)

54 1a. SCF/SLID, Market Income 50

Gini Index

46 42

1b. SCF/SLID, Gross Income

38 34 1c. SCF/SLID, Disposable Income 30 1965

1970

1975

1980

1985

1990

1995

2000

2005

Year Source: 1: Stark (1977, table 18, 33) for 1965 to 1975, Statistics Canada (1996, table 6, 34) for 1971 to 1994, and Statistics Canada (2007) for 1980 to 2004, data from Survey of Consumer Finances (SCF) for 1965 to 1995 and Survey of Labor and Income Dynamics (SLID) for 1996 to 2000; a: market money income of households; b: gross money income of households; c: disposable money income of households; weighted by household.

worthy when contrasted with the widening of the distribution of market incomes, which was particularly sharp in the early 1980s. What distinguishes Canada from the United Kingdom and the United States appears to be the redistributive role of public transfers in offsetting the increasing inequality generated by market mechanisms (Fritzell 1993; Osberg, Erksoy, and Phipps 1997). The 1990s marked a change as inequality began to rise. From 1989 to 2004, the Gini index for disposable income went up steadily overall by four percentage points. Initially, public policies counteracted underlying forces toward a more unequal income distribution, but their balancing role ceased after the mid-1990s. Marc Frenette, David Green, and Garnett Picot (2004) attributed the change to the reduction of personal tax rates, especially for the highest incomes, and the tightening of social benefits such as unemployment insurance, whose reform in 1996 led to a halving of the ratio of beneficiaries to the unemployed. In short, Canada has experienced a noticeable increase in disposable income inequality in

Inequality Patterns in Western Democracies Figure 2.8

45

Gini Index in Sweden (Percentage)

56 52 48

1a. LLS/IDS, Equivalent Market Income

Gini Index

44 40 36

1b. LLS/IDS, Equivalent Gross Income 2a. IDS, Equivalent Disposable Income

32 28 24 20

1c. LLS/IDS, Equivalent Disposable Income

16 1965

1970

1975

2b. IDS, Equivalent Disposable Income 1980

1985

1990

1995

2000

2005

Year Sources: 1: Gustafsson and Uusitalo (1990, table 2, 85; table 3, 89; table 4, 91) for 1967 to 1985 and Statistics Sweden (2006a) for 1975 to 2004, data from Level-of-Living Survey (LLS) for 1967 and Income Distribution Survey (IDS) for 1975 to 2O04; a: equivalent market income of families; b: equivalent gross income of families; c: equivalent disposable income of families; in all cases, weighted by person; social assistance equivalence scale; second and third series differ for the definition of income. 2: Statistics Sweden (2006b, 2006c); a: equivalent disposable income of households including capital gains; b: equivalent disposable income of households excluding capital gains; in both cases, weighted by person; social assistance equivalence scale.

the last decade, after a long period of virtual stability; the worsening has, however, been much less pronounced than the overall widening in the distribution of market income.

Sweden and Finland In Sweden, the dispersion of equivalent family incomes decreased considerably from 1967 to 1975 and kept falling, more moderately, until 1981 and 1982 (figure 2.8). Although both transfers and direct taxes contributed to the narrowing between 1967 and 1975, Bjorn Gustafsson and Edward Palmer (1997, 308) showed that, in subsequent years, “the development of transfers [was] a major cause of the decrease in inequality . . . and was instrumental in offsetting the tendency of direct taxes to

46

Democracy, Inequality, and Representation

work in the direction of increased inequality.” The trend reversed in the early 1980s. By 1990, inequality was back to the 1975 level, and it continued to rise throughout the 1990s.9 Both the ascending tendency and the greater variability of the last fifteen years reflect the broadening of the income tax base brought about by the tax reform of 1991, which caused a major break in the Income Distribution Survey (IDS) series. Moreover, the reform introduced incentives to realize capital gains on equities in 1991 and 1994, which, being now included in the tax base, produced sharp temporary increases of the Gini index (Björklund 1998; Eriksson and Pettersson 2000). Also, the peak of inequality in 2000 followed from high capital gains in that year (Statistics Sweden 2004). Excluding capital gains from the income definition smoothes out the movements of inequality and dampens down its tendency to rise, as shown by the comparison of lines 2a and 2b in figure 2.8. The Finnish distributions of gross and disposable income became substantially less unequal from 1966 to 1976 and then remained fairly stable over the 1980s and early 1990s (figure 2.9). This is confirmed by the analysis of Lorenz dominance by Markus Jäntti and Veli-Matti Ritakallio (1997). Although part of the fall reflected the compression of the wage distribution caused by income policies, these dynamics were mainly driven by the steady increase of government transfers (Uusitalo 1989). Their equalizing effect first amplified the decline in the inequality of market incomes from 1966 to 1976 and then offset the considerable rise from 1981 to 1994. Redistribution was particularly effective during the severe and prolonged recession of the early 1990s, when the unemployment rate rose from 3.1 percent in 1990 to 16.6 percent in 1994. These patterns changed dramatically in the last years of the past century as the redistributive impact of transfers and, to a much lesser extent, direct taxes weakened considerably: between 1994 and 2000, the Gini indices of gross and disposable incomes increased by around five percentage points and returned to the values recorded in 1971, despite a modest rise in the concentration of market incomes. Inequality indices flattened out over the first half of the 2000s.

The Netherlands, West Germany, France, and Italy In the last group of countries, all from continental Europe, the evidence is more mixed. In the Netherlands, there was a sharp rise in inequality toward the end of the 1980s; from 1977 to 1985, however, there was little change, nor has there been much change since 1990, once statistical breaks are taken into account (figure 2.10). The Gini index for equivalent disposable incomes fell in West Germany by four percentage points be-

Inequality Patterns in Western Democracies Figure 2.9

47

Gini Index in Finland (Percentage)

48 44 1a. HBS/IDS, Equivalent Market Income

Gini Index

40 36 32 1b. HBS/IDS, Equivalent Gross Income 28 24 20 1c. HBS/IDS, Equivalent Disposable Income 16 1965 1970 1975 1980 1985 1990

1995

2000

2005

Year Source: 1: Statistics Finland (2006), data from Household Budget Survey (HBS) for 1996 to 1981 and Income Distribution Survey (IDS) for 1987 to 2004; a: equivalent market income of households; b: equivalent gross income of households; c: equivalent disposable income of households. In all cases, weighted by person; OECD equivalence scale.

tween 1962 and 1973, but the tendency was reversed in the following quarter-century, according to the Income and Consumption Survey (EVS, Einkommens- und Verbrauchsstichproben; see figure 2.11). The findings of the Socio-Economic Panel (SOEP), which are broadly in line with the EVS results in overlapping years, suggest that inequality went up through 2004, with an overall increase of the Gini index by almost four points between 1983 and 2004. The concentration of market incomes rose markedly from 1973 to 1978, and more moderately until 1988, but remained stable in the 1990s. All in all, the impact of public redistribution in Germany remained high and fairly stable from 1978 to 1998. In France, the Gini index of gross income did not vary from 1956 to 1962, fell considerably from then until 1990, and then was unchanged between 1990 and 1997; the Gini index of equivalent disposable income decreased until 1997 and then stabilized through 2004 (figure 2.12). Thus, income inequality has not shown to date any upward trend in France. Note, however, that these estimates derive from the Tax Revenue Survey (ERF, Enquête Revenus Fiscaux), which is based on fiscal records

48

Democracy, Inequality, and Representation

Figure 2.10

Gini Index in the Netherlands (Percentage)

32 30

Gini Index

28 26

1a. Equivalent Disposable Income

24 22

1b. Equivalent Disposable Income

20 18 1970

1975

1980

1985

1990

1995

2000

2005

Year Source: 1: Personal communication from Wim Bos of the Central Bureau of Statistics (CBS), data from IDS for 1977, 1981, and 1985 and from Income Panel Survey (IPS) for 1989 to 2004; a: equivalent disposable income of households; weighted by household; b: equivalent disposable income of households; weighted by person; in both case, CBS equivalence scale.

supplemented with imputed social assistance benefits: since undeclared property incomes (that is, those that are not taxed or are subject to withholding tax) are not included in the income definition, inequality is understated (Guillemin and Roux 2003, 404) and the effect on time variations is uncertain. As regards Italy, a markedly egalitarian phase began in the 1970s and lasted until the early 1980s; it has been followed by a period of fluctuations around a stationary or slightly rising trend (figure 2.13). Post-tax income inequality rose sharply during the recession of 1992 and 1993, the worst downturn since the Second World War, but remained surprisingly stable afterwards in a period of many changes for the Italian economy, especially in the labor market (Boeri and Brandolini 2004).

Trends in Monetary Redistribution The previous summary has shown how the evolution of inequality may differ depending on whether it is measured for market income or for dis-

Inequality Patterns in Western Democracies Figure 2.11

49

Gini Index in West Germany (Percentage)

48 44

2. EVS, Equivalent Market Income

Gini Index

40 36 32 3. SOEP, Equivalent Disposable Income 28 24

1. EVS, Equivalent Disposable Income

20 1960

1965

1970

1975

1980

1985

1990

1995

2000

2005

Year Source: 1: Becker (1997, table 1, 47) for 1962 to 1988 and Becker et al. (2003, table 3.3, 78–80) for 1983 to 1998, data from Income and Consumption Survey (EVS): equivalent disposable income of households; weighted by person; OECD equivalence scale; only German population. 2: Hauser and Becker (2001, 86) for 1973 to 1998 and Becker et al. (2003, table 3. 1, 73–74) for 1983 to 1998, EVS data: equivalent market income of households; weighted by person; OECD equivalence scale; only German population, 3: SOEP (2006, 83–84), data from Socio-Economic Panel (SOEP): equivalent disposable income of households, including imputed rent; weighted by person; modified OECD equivalence scale.

posable income. This suggests that the trend in the redistributive impact of tax-and-transfer systems may also vary considerably across nations. These patterns are shown for six countries in figure 2.14, again by looking at the absolute difference between the Gini index for market income and that for disposable income. Note that these time series also reflect national practices, and so the level of redistribution is not completely comparable across nations. What emerges is a general pattern suggesting that the redistributive impact of taxes and transfers initially increased and then stabilized or dropped in all countries except for the United States, where it remained quite stable over time (but the series starts only in 1979). The United Kingdom stands out for having the most dramatic switch of regime: in the early 1980s, it apparently shifted from a situation not too different from the two Nordic countries to a model closer to that of the two North

50

Democracy, Inequality, and Representation

Figure 2.12

Gini Index in France (Percentage)

52 1. ERF, Gross Taxable Income 48

Gini Index

44 40 36

2. ERF, Equivalent Gross Income

32 28

3. ERF, Equivalent Disposable Income

24 1955 1960

1965 1970 1975 1980 1985 1990 1995 2000 2005 Year

Source: 1: United Nations (1981, 108, 110) for 1956 to 1975 and Concialdi (1997, table 11.11, 256) for 1962 to 1984, data from Tax Revenue Survey (ERF): gross taxable income of households, excluding nontaxable incomes (the majority of social benefits, some property income); weighted by household. 2: Hourriez and Roux (2001, table 1, 280), data from ERF: equivalent gross taxable income of households excluding property income and some social benefits; weighted by household; OECD modified equivalence scale; only households with non-negative taxable income and positive disposable income. 3: Chevalier et al. (2006, figure 4, 449); figures provided by Pascal Chevalier for 1970 to 2002 and INSEE (2006, table 2, 71) for 2003 to 2004, data from ERF: equivalent disposable taxable income of households excluding property income and some social benefits; weighted by person; OECD modified equivalence scale; only persons in households with non-negative taxable income and positive disposable income.

American countries. It is not possible to infer from this simple measure whether changes in redistribution are the automatic response of a progressive tax-and-benefit system to changes in the distribution of market incomes or are instead the product of explicit policy choices (Atkinson 2004). Nevertheless, they confirm that a widening of the market income distribution need not result in a drastic increase in the inequality of disposable incomes. Rising levels of redistribution in Finland, Sweden, and, to a lesser extent, Canada—where policies have been increasingly targeted to the poor—have been more effective in muting increasing market income inequality than have stable but low levels of redistribution in the United States, though periods do matter.

Inequality Patterns in Western Democracies Figure 2.13

51

Gini Index in Italy (Percentage)

46 2. SHIW, Disposable Income

Gini Index

42

38

1. SHIW, Disposable Income

34

30 3. SHIW, Equivalent Disposable Income 26 1965

1970

1975

1980

1985

1990

1995

2000

2005

Year Source: 1: Brandolini (2004, table l, col. 4, 14), data from the Bank of Italy’s Survey of Household Income and Wealth (SHIW): disposable income of households excluding imputed rents and interest and dividends; weighted by household; figures for 1968 to 1972 estimated from grouped data. 2: Brandolini (2004, table l, col. 5, l4), data from SHIW: disposable income of households excluding interest and dividends, weighted by household; figures for 1973 to 1975 estimated from grouped data. 3: Brandolini (2004, table 1, col. 8, 14), data from SHIW: equivalent disposable income of households; weighted by person; square-root equivalence scale.

Conclusions In this chapter, we have shown that there are considerable differences in the level of disposable income inequality across rich countries. Mexico and Russia have the most unequal distributions, followed by Englishspeaking countries intertwined with Southern European countries; other continental European nations come next, and the Nordic countries show the lowest level of inequality; most Eastern European countries show low to medium levels of inequality, while Taiwan and Japan are in an intermediate position. This clustering owes much to the working of the national tax-and-benefit systems, which play a considerable role in narrowing the original market income distribution. A further egalitarian impact is due to benefits in kind, although the evidence is still limited. National experiences have varied during the last four decades, and

52

Democracy, Inequality, and Representation

Figure 2.14

Equalizing Effect of Taxes and Transfers

30

Percentage

25

20

15

10

5 1965

1970

1975

1980

1985

1990

1995

2000

2005

Year Canada (Unadjusted Income) Finland (Equivalent Income) Germany (Equivalent Income)

United States (Unadjusted Income) Sweden (Equivalent Income) United Kingdom (UnadjustedEquivalent Income)

Source: Authors’ computation. Note: Absolute difference between the Gini index of market income and the Gini index of disposable income.

there is no one overarching common story. There was some tendency for the disposable income distribution to narrow until the mid-1970s. Then income inequality rose sharply in the United Kingdom in the 1980s and in the United States in the 1980s and 1990s (and is still continuing), but more moderately in Canada, Sweden, Finland, and West Germany in the 1990s. Moreover, the timing and magnitude of the increase differed widely across nations. Inequality did not show any persistent tendency to rise in the Netherlands, France, and Italy. Commonality seems to be greater for market income inequality: in five of the six countries for which we have data, we observe an increase in the 1980s and early 1990s and a substantial stability afterwards. Changing public monetary redistribution appears to be an important determinant of the time pattern of the inequality of disposable incomes.

Inequality Patterns in Western Democracies

53

Changes in inequality do not exhibit clear trajectories but rather irregular movements, with more substantial changes often concentrated in rather short lapses of time. Together with the lack of a common international pattern, this suggests that we should look at explanations based on the joint working of multiple factors that sometimes balance out and sometimes reinforce each other rather than focus on explanations centered on a single cause like deindustrialization, skill-biased technological progress, or globalization. Identifying and characterizing episodes and turning points in the dynamics of inequality may be more fruitful than searching for overarching general tendencies. All of the countries considered in this chapter share what we very loosely define as a Western-type democracy. The large cross-country variation in levels and trends of inequality suggests that the specific features of a democratic political system must be taken into account in order to explain its interconnections with income distribution. The electoral system, the division of power between the central and local administrations, and the political inclination of running coalitions, interacting with other institutional characteristics and the operation of economic forces (see the chapters by Iversen and Soskice, Beramendi and Cusack, and Rueda in this volume), may all be relevant factors behind observed changes in inequality. The variable, but always noticeable, effect on inequality that is attributable to public redistribution confirms the importance of understanding how the political system mediates people’s political views and brings them to realization. Whatever the link between democracy and inequality, it is unlikely to be a simple one.

This chapter is a revised version of the paper presented at the 2005 conference on democracy, inequality, and representation held at the Maxwell School of Syracuse University. Part of the chapter draws on our article “Inequality (International Evidence),” in The New Palgrave Dictionary of Economics, edited by Stephen N. Durlauf and Lawrence E. Blume, Basingstoke, Palgrave Macmillan 2008, pp. 273–82. Timothy Smeeding would like to thank the Ford, Russell Sage, and Sloan Foundations for their support, along with Mary Santy, Kati Foley, Karen Cimilluca, Kim Desmond, Jeff Thompson, and Coady Wing for their assistance. We thank Wim Bos of the Dutch Central Bureau of Statistics and Pascal Chevalier of INSEE for providing us with the Dutch and French income statistics, respectively. The statistical information on historical trends draws heavily on the joint work of Andrea Brandolini with Sir Anthony B. Atkinson. We also thank the conference participants and two anonymous reviewers for comments and suggestions that have improved this chapter. In the end, however, the

54

Democracy, Inequality, and Representation views expressed here are solely those of the authors; in particular, they do not necessarily reflect those of the Bank of Italy, Syracuse University, or any of our sponsors.

Notes 1.

2.

3.

4.

5.

6.

7.

In this chapter, we pay no attention to the distribution of labor earnings among employees. This distribution must be kept distinct from the distribution of income among households, as they need not move together (Atkinson and Brandolini 2006). Atkinson and Brandolini (2006) contains a brief discussion of the data in the OECD Structure of Earnings Database used in some of the subsequent chapters in this volume. To minimize the impact of outliers, all records with zero income are dropped, and observations are bottom-coded at 1 percent of the mean of equivalent disposable income and top-coded at ten times the median of unadjusted disposable income. Different definitions and computational assumptions affect measured inequality and hence international comparisons. See, for instance, Atkinson, Rainwater, and Smeeding (1995), Gottschalk and Smeeding (1997, 2000), and Atkinson and Brandolini (2001). Throughout the chapter, “Germany” refers to the Federal Republic of Germany after reunification in 1990, while “West Germany” refers to the Federal Republic of Germany until 1990 and to the Western Länder thereafter. The Lorenz curve plots the share of income of the bottom 100x percent of the population against the population share x; it is convex to the origin and always lies below the 45-degree line of perfect equality. PPP indices are routinely estimated by various international agencies, such as the Organization for Economic Cooperation and Development (OECD) or the World Bank, or by international research projects like the Penn World Table (Summers and Heston 1991); moreover, they are computed for various national accounts aggregates, like GDP or household final consumption expenditure. Methods to estimate PPPs also differ, as discussed, for instance, by Steve Dowrick and Muhammad Akmal (2005). This difference shows up in sizable shortfalls of total survey incomes from GDP aggregates. Because these shortfalls vary across countries, comparisons of living standards based on survey means may differ from those based on national accounts, although the correlation between per capita GDP and survey disposable income per person is positive, if less than one. The comparisons of real incomes discussed later would be affected should we align household-level data to aggregate statistics. The main sources on income distribution have increasingly become household sample surveys; tax returns and other administrative data are still extensively used, notably in Nordic countries, but often in conjunction with survey infor-

Inequality Patterns in Western Democracies

8.

9.

55

mation on family characteristics. Generally speaking, basic statistics have become more representative, since they have been extended to cover the whole household population. This fact alone points to being cautious in comparing postwar and prewar records, as well as the earlier postwar figures with the more recent ones. Another difficulty for temporal comparisons originates in the periodic revisions of statistical procedures to collect and elaborate the data. Moreover, series may differ for: the measure of income, namely, whether it is before taxes and transfers (market income), after transfers but before taxes (gross income), or after direct taxes and transfers (disposable income), and whether it includes capital gains or noncash items such as perks or in-kind benefits; the reference unit (household, family, person, or taxpayer); the weighting unit, the main alternatives being between weighting on a family basis (counting each family or household as one regardless of its size) or a person basis (replicating the observation as many times as the members of the family—the so-called person-weights used in figures 2.1 to 2.3); and the allowance for family composition, that is, whether income is adjusted by an equivalence scale to account for differences in needs and economies of scale. Lastly, the Gini index is chosen since it is the single measure most readily available in international statistics, and often the only one found in country sources, especially back in time. It must be kept in mind, however, that not only levels but also changes over time can differ across alternative inequality measures—although the conclusions on long-run trends are unlikely to be seriously affected. The BEA statistics differ from the CPS series for the more comprehensive income definition and for being based on “synthetic” methods: distribution is estimated “from a wide variety of sources, including—besides field surveys such as the CPS—tax returns, other business and governmental administrative records, and the income type aggregates as contained in the National Income Accounts” (Budd and Radner 1975, 451). All three series labeled 1 in figure 2.8 refer to families, which are narrowly defined to include only an adult or a couple and all children below age eighteen. The implication is to assume lower (statistical) income sharing within the household—for example, children older than eighteen living with their parents are treated as a separate family unit and are not attributed the whole (equivalent) household income. This drives up measured inequality, as shown by the comparison of lines 2a and 1c in the figure.

References Aaron, Henry Joseph. 1978. Politics and the Professors. Washington, D.C.: Brookings Institution Press. Acemoglu, Daron, and James A. Robinson. 2006. Economic Origins of Dictatorship and Democracy. Cambridge: Cambridge University Press.

56

Democracy, Inequality, and Representation

Atkinson, Anthony B. 1970. “On the Measurement of Inequality.” Journal of Economic Theory 2(3): 244–63. ———. 1993. “What Is Happening to the Distribution of Income in the U.K.?” Proceedings of the British Academy 82: 317–51. ———. 2003. “Income Inequality in OECD Countries: Data and Explanations.” CESifo Economic Studies 49(4): 479–513. ———. 2004. “Increased Income Inequality in OECD Countries and the Redistributive Impact of the Government Budget.” In Inequality, Growth, and Poverty in an Era of Liberalization and Globalization, edited by Giovanni A. Cornia. Oxford: Oxford University Press. Atkinson, Anthony B., and Andrea Brandolini. 2001. “Promises and Pitfalls in the Use of Secondary Data Sets: Income Inequality in OECD Countries as a Case Study.” Journal of Economic Literature 39(3): 771–800. ———. 2006. “From Earnings Dispersion to Income Inequality.” In Inequality and Economic Integration, edited by Francesco Farina and Ernesto Savaglio. London: Routledge. Atkinson, Anthony B., and John Micklewright. 1989. “Turning the Screw: Benefits for the Unemployed 1979–1988.” In The Economics of Social Security, edited by Andrew Dilnot and Ian Walker. Oxford: Oxford University Press. ———. 1992. Economic Transformation in Eastern Europe and the Distribution of Income. Cambridge: Cambridge University Press. Atkinson, Anthony B., Lee Rainwater, and Timothy M. Smeeding. 1995. Income Distribution in OECD Countries: The Evidence from the Luxembourg Income Study (LIS). Paris: Organization for Economic Cooperation and Development. Becker, Irene. 1997. “Die Entwicklung von Einkommensverteilung und Einkommensarmut in den alten Bundesländern von 1962 bis 1988” [“Trends in Income Distribution and Poverty in the Former German Federal Republic from 1962 to 1988”]. In Einkommensverteilung und Armut: Deutschland auf dem Weg zur Vierfünftel-Gesellschaft? [Income Distribution and Poverty: Germany Towards the FourFifth Society?], edited by Irene Becker and Richard Hauser. Frankfurt: Campus. Becker, Irene, Joachim R. Frick, Markus M. Grabka, Richard Hauser, Peter Krause, and Gert G. Wagner. 2003. “A Comparison of the Main Household Income Surveys for Germany: EVS and SOEP.” In Reporting on Income Distribution and Poverty, edited by Richard Hauser and Irene Becker. Berlin-Heidelberg: Springer. Benabou, Roland. 1996. “Inequality and Growth.” In National Bureau of Economic Research Macroeconomics Annual 1996, edited by Ben S. Bernanke and Julio J. Rotemberg. Cambridge, Mass.: MIT Press. Björklund, Anders. 1998. “Income Distribution in Sweden: What Is the Achievement of the Welfare State?” Swedish Economic Policy Review 5(1): 39–80. Boeri, Tito, and Andrea Brandolini. 2004. “The Age of Discontent: Italian Households at the Beginning of the Decade.” Giornale degli economisti e Annali di economia 63(3/4): 155–93.

Inequality Patterns in Western Democracies

57

Boix, Carles. 2003. Democracy and Redistribution. Cambridge: Cambridge University Press. Brandolini, Andrea. 1998. “A Bird’s-Eye View of Long-Run Changes in Income Inequality.” Unpublished paper. Bank of Italy. ———. 2004. “Income Inequality and Poverty in Italy: A Statistical Compendium.” Unpublished paper. Bank of Italy. Brewer, Mike, Alissa Goodman, Jonathan Shaw, and Luke Sibieta. 2006. “Poverty and Inequality in Britain: 2006.” Commentary 101. London: Institute for Fiscal Studies. Accessed at http://www.ifs.org.uk/bns/bn19figs.zip. Budd, Edward C., and Daniel B. Radner. 1975. “The Bureau of Economic Analysis and Current Population Survey Size Distributions: Some Comparisons for 1964.” In The Personal Distribution of Income and Wealth, Studies in Income and Wealth vol. 39, edited by James D. Smith. New York: Columbia University Press. Castles, Francis G. 1999. “On Income Inequality and Democracy: Theories and Data in Search of a Viable Data Base.” In Growth, Distribution, and Political Change: Asia and the Wider World, edited by Ryoshin Minami, Kwan S. Kim, and Malcom Falkus. Basingstoke, UK: Macmillan. Chevalier, Pascal, Olivier Guillemin, Aude Lapinte, and Jean-Paul Lorgnet. 2006. “Les Évolutions de niveau de vie entre 1970 et 2002” [“Evolutions of the Standard of Living Between 1970 and 2002”]. In INSEE, Données sociales—La Société française [Social Benefits—French Society]. Paris: INSEE. Concialdi, Pierre. 1997. “Income Distribution in France: The Mid-1980s Turning Point.” In Changing Patterns in the Distribution of Economic Welfare: An International Perspective, edited by Peter Gottschalk, Bjorn Gustafsson, and Edward Palmer. Cambridge: Cambridge University Press. Deaton, Angus. 2005. “Measuring Poverty in a Growing World (or Measuring Growth in a Poor World).” Review of Economics and Statistics 87(1): 1–19. Deininger, Klaus, and Lyn Squire. 1996. “A New Data Set Measuring Income Inequality.” World Bank Economic Review 10(3): 565–91. Dowrick, Steve, and Muhammad Akmal. 2005. “Contradictory Trends in Global Income Inequality: A Tale of Two Biases.” Review of Income and Wealth 51(2): 201–29. Eriksson, Ingemar, and Thomas Pettersson. 2000. “Income Distribution and Income Mobility: Recent Trends in Sweden.” In The Personal Distribution of Income in an International Perspective, edited by Richard Hauser and Irene Becker. Berlin: Springer. Frenette, Marc, David Green, and Garnett Picot. 2004. “Rising Income Inequality Amid the Economic Recovery of the 1990s: An Exploration of Three Data Sources.” Research Paper Series 219. Ottawa: Statistics Canada, Analytical Studies Branch (July). Fritzell, Johan. 1993. “Income Inequality Trends in the 1980s: A Five-Country Comparison.” Acta Sociologica 36(1): 47–62.

58

Democracy, Inequality, and Representation

Garfinkel, Irwin, Lee Rainwater, and Timothy M. Smeeding. 2006. “A Reexamination of Welfare States and Inequality in Rich Nations: How In-Kind Transfers and Indirect Taxes Change the Story.” Journal of Policy Analysis and Management 25(4): 897–919. Goodman, Alissa, and Steven Webb. 1994. “For Richer, for Poorer: The Changing Distribution of Income in the United Kingdom, 1961–1991.” Commentary 42. London: Institute for Fiscal Studies. Gottschalk, Peter, and Timothy M. Smeeding. 1997. “Cross-National Comparisons of Earnings and Income Inequality.” Journal of Economic Literature 35(2): 633–87. ———. 2000. “Empirical Evidence on Income Inequality in Industrialized Countries.” In Handbook of Income Distribution, edited by Anthony B. Atkinson and François Bourguignon. Cambridge: Cambridge University Press. Guillemin, Olivier, and Valérie Roux. 2003. “Le Niveau de vie des ménages de 1970 à 1999” [“The Standard of Living in Households from 1970 to 1999”]. In INSEE, Données sociales—La Société française [Social Benefits—French Society]. Paris: INSEE. Gustafsson, Bjorn, and Edward Palmer. 1997. “Changes in Swedish Inequality: A Study of Equivalent Income, 1975–1991.” In Changing Patterns in the Distribution of Economic Welfare: An International Perspective, edited by Peter Gottschalk, Bjorn Gustafsson, and Edward Palmer. Cambridge: Cambridge University Press. Gustafsson, Bjorn, and Hannu Uusitalo. 1990. “Income Distribution and Redistribution During Two Decades: Experiences from Finland and Sweden.” In Generating Equality in the Welfare State: The Swedish Experience, edited by Inga Persson. Oslo: Norwegian University Press. Hauser, Richard, and Irene Becker. 2001. Einkommensverteilung im Querschnitt und im Zeitverlauf 1973 bis 1998 [Income Distribution: A Longitudinal Analysis from 1973 to 1988]. Bonn: Bundesministeriums für Arbeit und Sozialordnung. Hourriez, Jean-Michel, and Valérie Roux. 2001. “Vue d’ensemble des inégalités de revenu et de patrimoine” [“An Overview of Inequalities of Income and Inheritance”]. In Conseil D’Analyse Économique, Inégalités économiques [Economic Inequalities]. Paris: La Documentation Française. Immervoll, Herwig, Horacio Levy, Christine Lietz, Daniela Mantovani, Cathal O’Donoghue, Holly Sutherland, and Gerlinde Verbist. 2007. “Household Incomes and Redistribution in the European Union: Quantifying the Equalizing Properties of Taxes and Benefits.” In The Distributional Effects of Government Spending and Taxation, edited by Dimitri B. Papadimitriov. Baskingstoke, UK: Palgrave Macmillan. INSEE. 2006. Les Revenus et le patrimoine des ménages: Édition 2006 [Income and Household Inheritance: 2006 Edition]. Paris : INSEE. International Monetary Fund. 2006. World Economic Outlook. Washington: IMF (September).

Inequality Patterns in Western Democracies

59

Ishikawa, Tsuneo. 1996. Data runs conducted by the Japanese Ministry of Welfare (November 26). Jäntti, Markus, and Veli-Matti Ritakallio. 1997. “Income Inequality and Poverty in Finland in the 1980s.” In Changing Patterns in the Distribution of Economic Welfare: An International Perspective, edited by Peter Gottschalk, Bjorn Gustafsson, and Edward Palmer. Cambridge: Cambridge University Press. Jenkins, Stephen P. 2000. “Trends in the U.K. Income Distribution.” In The Personal Distribution of Income in an International Perspective, edited by Irene Becker. London: Springer. Johnson, Paul, and Steven Webb. 1993. “Explaining the Growth in U.K. Income Inequality.” Economic Journal 103(417): 429–35. Jones, Francis. 2006. “The Effects of Taxes and Benefits on Household Income, 2004–2005.” London: Office for National Statistics. Accessed at http:// www.statistics.gov.uk/articles/nojournal/taxesbenefits200405/Taxesbene fits200405.pdf. Kuznets, Simon. 1953. Shares of Upper-Income Groups in Income and Savings. New York: National Bureau of Economic Research. ———. 1955. “Economic Growth and Income Inequality.” American Economic Review 45(1): 1–28. Lenski, Gerhard. 1966. Power and Privilege: A Theory of Social Stratification. New York: McGraw-Hill. Mahler, Vincent A., and David K. Jesuit. 2006. “Fiscal Redistribution in the Developed Countries: New Insights from the Luxembourg Income Study.” SocioEconomic Review 4(3): 483–511. Osberg, Lars, Sadettin Erksoy, and Shelley Phipps. 1997. “Unemployment, Unemployment Insurance, and the Distribution of Income in Canada in the 1980s.” In Changing Patterns in the Distribution of Economic Welfare: An International Perspective, edited by Peter Gottschalk, Bjorn Gustafsson, and Edward Palmer. Cambridge: Cambridge University Press. Pechman, Joseph A. 1958. “Comment.” In An Appraisal of the 1950 Census Income Data, Studies in Income and Wealth, vol. 23. Princeton, N.J.: Princeton University Press. Persson, Torsten, and Guido Tabellini. 1994. “Is Inequality Harmful for Growth?” American Economic Review 84(3): 600–621. Ryscavage, Paul. 1995. “A Surge in Growing Income Inequality?” Monthly Labor Review 118(8): 51–61. Schwabish, Jonathan, Timothy M. Smeeding, and Lars Osberg. 2006. “Income Distribution and Social Expenditures: A Cross-National Perspective.” In The Distributional Effects of Government Spending and Taxation, edited by Dimitri B. Papadimitriou. Northampton, Mass.: Edward Elgar. Shanahan, Suzanne Elise, and Nancy Brandon Tuma. 1994. “The Sociology of Distribution and Redistribution.” In The Handbook of Economic Sociology, edited

60

Democracy, Inequality, and Representation

by Neil J. Smelser and Richard Swedberg. Princeton, N.J.: Princeton University Press. Smeeding, Timothy M. 2004. “Twenty Years of Research on Income Inequality, Poverty, and Redistribution in the Developed World: Introduction and Overview.” Socio-Economic Review 2(2): 149–63. ———. 2005. “Public Policy, Economic Inequality, and Poverty: The United States in Comparative Perspective.” Social Science Quarterly 86 (supp.): 955–83. ———. 2006. “Poor People in Rich Nations: The United States in Comparative Perspective.” Journal of Economic Perspectives 20(1): 69–90. Smeeding, Timothy M., and Lee Rainwater. 2004. “Comparing Living Standards Across Nations: Real Incomes at the Top, the Bottom, and the Middle.” In What Has Happened to the Quality of Life in the Advanced Industrialized Nations?, edited by Edward N. Wolff. Northampton, Mass.: Edward Elgar. Smeeding, Timothy M., Peter Saunders, John Coder, Stephen Jenkins, Johan Fritzell, Aldi J. M. Hagenaars, Richard Hauser, and Michael Wolfson. 1993. “Poverty, Inequality, and Family Living Standards Impacts Across Seven Nations: The Effect of Noncash Subsidies for Health, Education, and Housing.” Review of Income and Wealth 39(3): 229–56. Socio-Economic Panel (SOEP). 2006. SOEP-Monitor: Beobachtungszeitraum: 1984–2005: Analyse-Ebene: Person [SOEP-Monitor: Observation Period: 1984–2005: Analysis Level: Person]. Berlin: SOEP. Accessed at http://www.diw.de/deutsch/ sop/service/soepmonitor/. Solow, Robert M. 1960. “Income Inequality Since the War.” In Postwar Economic Trends in the United States, edited by Ralph E. Freeman. New York: Harper & Brothers. Stark, Thomas. 1977. The Distribution of Income in Eight Countries. Royal Commission on the Distribution of Income and Wealth background paper 4. London: Her Majesty’s Stationery Office. Statistics Canada. 1996. Income After Tax, Distributions by Size in Canada: 1994. Ottawa: Statistics Canada. ———. 2007. “Table 202-0705: Gini Coefficients of Market, Total and After-Tax Income, by Economic Family Type, Annual (Number).” Accessed at http:// cansim2.statcan.ca/cgi-win/cnsmcgi.pgm?Lang=E&ArrayId= 2020705&Array _Pick=1&Detail=1&RootDir=CII/&ResultTemplate=CII\CII___. Statistics Finland. 2006. Tuloerojen kehitys Suomessa 1966–2004 [The Development of Income Difference in Finland 1966–2004]. Accessed at http://www.stat.fi/til/tjt/ 2004/tjt_2004_2006-06-16_tau_001.html. Statistics Sweden. 2004. Income Distribution Survey 2002. Accessed at http://www .scb.se/templates/Publikation____90903.asp. ———. 2006a. The Distribution of Income 1975–2004. Accessed at http://www .scb.se/templates/tableOrChart____163551.asp. ———. 2006b. Disposable Income Including Capital Gains per Consumption Unit for In-

Inequality Patterns in Western Democracies

61

dividuals by Deciles. Accessed at http://www.scb.se/templates/tableOrChart____ 163545.asp. ———. 2006c. Disposable Income per Consumption Unit Excluding Capital Gains by Deciles. Accessed at http://www.scb.se/templates/tableOrChart____163547.asp. Summers, Robert, and Alan Heston. 1991. “The Penn World Table (Mark 5): An Expanded Set of International Comparisons, 1950–1988.” Quarterly Journal of Economics 106(2): 327–68. United Nations. 1981. A Survey of National Sources of Income Distribution Statistics (First Report). Statistical papers, series M, no. 72. New York: UN. U.S. Census Bureau. 2006a. “Historical Income Tables—Families—Table F-4: Gini Ratios for Families, by Race and Hispanic Origin of Householder: 1947 to 2005.” Accessed at http://www.census.gov/hhes/www/income/histinc/f04 .html. ———. 2006b. “Historical Income Tables—Households—Table H-4: Gini Ratios for Households, by Race and Hispanic Origin of Householder: 1967 to 2005.” Accessed at http://www.census.gov/hhes/www/income/histinc/h04.html. ———. 2006c. “Historical Income Tables—Experimental Measures—Table RDI-5: Index of Income Concentration (Gini Index), by Definition of Income: 1979 to 2003.” Accessed at http://www.census.gov/hhes/www/income/histinc/rdi5 .html. U.S. Council of Economic Advisers. 1994. Economic Report of the President. Washington: U.S. Council of Economic Advisers. Uusitalo, Hannu. 1989. Income Distribution in Finland: The Effects of the Welfare State and the Structural Changes in Society on Income Distribution in Finland in 1966–1985. Studies 148. Helsinki: Central Statistical Office of Finland. Voitchovsky, Sarah. 2005. “Does the Profile of Income Inequality Matter for Economic Growth?” Journal of Economic Growth 10(3): 273–96. Wolfson, Michael C. 1986. “Stasis Amid Change: Income Inequality in Canada 1965–1983.” Review of Income and Wealth 32(4): 337–69. World Bank. 2005. World Development Report 2006: Equity and Development. New York: Oxford University Press.

Chapter 3

Social Rights, Welfare Generosity, and Inequality LYLE SCRUGGS

Comparative analyses of welfare state reform have relied overwhelmingly on public spending as the indicator of program commitment and change.1 Yet many welfare state scholars have long criticized the use of this type of data, emphasizing the importance of nonspending features of welfare state institutions for understanding the impact of national social programs. Despite this criticism, large-n comparative analyses of welfare state dynamics using alternative institutional measures of the national welfare state are surprisingly rare. Such alternatives are essential to any accurate assessment of the extent and impact of contemporary policy reform. The Comparative Welfare Entitlements Dataset (CWED) is intended to fill this gap. CWED contains systematic coding of important characteristics of three major social insurance programs that are central elements of modern welfare states: unemployment insurance, sick pay, and public pensions. These three programs make up one-half to two-thirds of nonhealth social spending in OECD countries. The program characteristics in CWED are tracked over eighteen OECD countries between 1970 and 2002. This chapter describes the data set’s main features, focusing particularly on the creation of an aggregate indicator, a benefit generosity index, which should be suitable for large-n (or small-n) comparisons. Following a description of the main features of the generosity indicator, the chapter presents an overview of national trends before turning to an analysis of the relationship between benefit generosity and redistribution. The results suggest three main points. First, though almost all welfare states were more generous at the start of the twenty-first century than

Social Rights, Welfare Generosity, and Inequality

63

they were three decades before (and probably ever before that), many countries have experienced some retrenchment in major social insurance programs from their peaks. In some cases, retrenchment has been a comparatively recent development. In others, the decline has been under way for more than a decade. Second, though the small Northern European countries did continue to be more generous at the start of the twenty-first century than most other countries we looked at, those differences were much less pronounced in 2002 than they were in the 1980s or early 1990s. In other words, there has been some downward convergence in recent years, albeit no “race to the bottom.” The third key point, which emerges from the empirical analysis at the end of the chapter, is that welfare benefit generosity is, on balance, a better overall predictor of redistribution than is public spending.

Problems with Comparative Social Spending There are several widely used cross-sectional time series indicators of welfare states in the comparative political economy literature. The most popular are general government spending and transfer spending, both of which have long been used in the comparative analysis of social policy and political economy (see, for example, Cameron 1978; Garrett 1998; Huber and Stephens 2001; Iversen and Cusack 2000; Korpi 1983; Lindert 2004; Swank 2002; Wilensky 1975; Wilensky and Lebeaux 1958). More recently, the OECD’s social expenditure database has been used to assess more fine-grained categories of social spending (Castles 2002, 2004; Lindert 2004; OECD 2004). Even case study comparisons often use this type of spending data. For example, Paul Pierson’s (1996) influential work frequently relies on spending trends to provide evidence for his main contention that the politics of welfare retrenchment differs from the politics of expansion. Undoubtedly, researchers rely on spending measures because they are available, appear directly comparable, and vary across countries and time (see De Deken and Kittel 2007). Yet researchers have also acknowledged that spending data have significant drawbacks as measures of the generosity of welfare states. These criticisms have become more pronounced in the evaluation of welfare state change (Castles and Mitchell 1993; Clasen and Sigel 2007; Pontusson and Clayton 1998; Esping-Andersen 1990; Gilbert and Moon 1988; Goodin et al. 1999). As a guide to understanding the impact of the welfare state on individual life chances or exposure to risk, spending data reveal little about the level of social protection against risk. As Gøsta Esping-Andersen famously remarked, “it is difficult to imagine that anyone struggled for spending per

64

Democracy, Inequality, and Representation

se” (1990, 21). But several other shortcomings of spending data as an indicator of welfare generosity are worth considering. Three are highlighted here: dependency structure, differential rates of economic growth, and tax systems.2

The Welfare Dependency Structure Aggregate spending as a measure of effort almost always relies on a spending ratio (for example, spending divided by GDP). That is, most empirical results account for variation in the amount of government spending relative to the size of the economy. Such measures do not typically account for the size of the dependent population, even though this is critical in determining the generosity of any spending level (or spending ratio). This problem arises both in cross-sectional views and in historical accounts. The implications of Esping-Andersen’s (1990) classic example of unemployment expenditure in the United Kingdom in the early 1980s—spending ratios went up even though benefits entitlements to individuals were severely and permanently restricted, because unemployment rates rose faster than the benefits were cut—have gone basically unheeded. Compared to unemployment rates, the retiree ratio presents a more severe challenge to measuring generosity with spending ratios. Almost every OECD country has recently experienced considerable growth in the ratio of individuals over age sixty-five to working-age adults. Stagnant spending ratios in the face of identical demographic shifts lead to very misleading conclusions about whether welfare retrenchment is taking place. For example, assume that the number of retirees doubles (from the same base) and the social spending ratio is unchanged. In a public pension system that provides for only 40 percent of retiree income, the retrenchment is considerably less than in a system that provides for 80 percent of retiree income.3

Differences in Economic Growth National time series of public spending ratios also underestimate real welfare expansion in poorer OECD countries compared with wealthier ones, insofar as the latter tend to grow faster. Compare Ireland and the United Kingdom. Based on spending ratios, Ireland has experienced significantly more welfare retrenchment since 1980 than has the United Kingdom. Irish spending ratios are lower as far back as the 1960s. Such cross-national differences are almost certainly more attributable, however, to differences in economic growth rates, structural change, and de-

Social Rights, Welfare Generosity, and Inequality

65

mographic developments than they are to more program retrenchment in Ireland. Ireland has grown at an average rate of 6.2 percent since 1983, the United Kingdom about 2.7 percent. Assume that real spending in both countries grew by 3 percent per year until 2007. The spending ratio in the United Kingdom would grow by around 10 percent, but in Ireland it would fall by more than 50 percent. Combined with the previous point, the contrast between the two countries becomes even starker. Ireland’s retiree-population ratio actually fell during this period, as did its unemployment rate. Higher numbers of retirement-age and unemployed people raise spending pressure and lower total output, putting upward pressure on spending for a given level of state commitment to individual social protection. Thus, it is possible that any observed spending declines (that is, in purchasing power, not ratio, terms) were more than compensated by a decline in the “needy” population, leaving generosity to rise considerably.

Taxation Differences in the tax treatment of transfers—due to special credits and exemptions or simply different tax rates—further distort the degree to which a given spending ratio translates into different real levels of disposable income for recipients. Although the tax system is increasingly being used as a transfer mechanism—the Earned Income Tax Credit (EITC) in the United States and the Working Family Tax Credit (WFTC) in the United Kingdom being notable examples—it can also be used to claw back social transfers (Howard 1997). Willem Adema and Maxime Ladaique (2005) find that increases in gross expenditure can be offset considerably by changing the tax treatment of transfers (for example, by making benefits taxable) or by increasing consumption taxes. It is true that these measurement biases in the spending data could be ameliorated with careful controls for relevant factors on the righthand side of a regression equation. However, this does not make spending a measure of generosity: these are still models that explain variation in spending ratios. Analyses of the implications of a spending model for retrenchment or generosity, even with appropriate controls, are virtually nonexistent in the literature because the belief that these are retrenchment models is so firmly embedded in the current approach to understanding welfare state dynamics. The idea that spending ratios are a good proxy for welfare state commitments is a bit like asserting that this year’s property insurance claims tell you the number of people with insurance.

66

Democracy, Inequality, and Representation

From Welfare Spending to Welfare Commitments An alternative to evaluating welfare state commitments is an entitlement or social rights approach. This approach assumes that the conditions stipulated in national social insurance programs—which are the most important sources of non-life insurance for the bulk of the population in all industrial countries—better encompass the extent of welfare state generosity. This is essentially the approach advocated by Korpi (1989), promoted in EspingAndersen (1990), and widely embraced as an ideal in comparative social policy (Castles 2002; Green-Pedersen and Haverland 2002; Hicks 1999; Kitschelt 2001). Taking this approach requires gathering information on institutional aspects of programs that are sometimes considered “qualitative details.” In fact, they are just the institutional features that are the “rules of the game”—and as subject to measurement as anything else.4 Social insurance regimes can be thought of as sets of commitments. These commitments extend beyond the number of actual claims being paid at any given time. The behavioral effect of these commitments may, of course, alter the balance of political and economic power between groups. Indirect behavioral effects are often explicitly considered in formal economic models, though such models tend to make somewhat narrow assumptions about the social welfare implications that such rights have on outcomes. Viewed as a commitment, the generosity measures presented here should be useful for understanding many traditional questions in the sociology, economics, and politics of labor markets. The generosity of commitments affects things that aggregate spending levels do not, such as bargained wages, quit rates, unemployment duration, labor market matching, and other microlevel phenomena. In other words, looking at institutional commitments rather than monetary outlays should be of greater use in thinking about and empirically testing many (if not most) political economy models of welfare state programs.

Comparative Welfare Generosity This section outlines our approach to measuring a benefit generosity index. The index functions as an alternative indicator of the extent of public commitment to welfare and is a more appropriate gauge of the extent of welfare generosity than aggregate spending. It is also similar in spirit to Esping-Andersen’s (1990) decommodification index—which has almost iconic status in the contemporary comparative social policy literature. Our major improvements on Esping-Andersen’s decommodification index are better accuracy and consistency of measurement, a more robust approach

Social Rights, Welfare Generosity, and Inequality

67

to aggregating different program dimensions, presentation of annual data over a long period of time, full documentation of the underlying data collection methods, and public availability and accessibility of the results.5 Table 3.1 provides a breakdown of the social insurance program characteristics that are used to compute the generosity index. Net income replacement rates (RRs) are calculated based on the legal rules for benefit calculation using a typical worker earning typical wages, incorporating details of the income tax and social insurance tax system. Data on program coverage address the universalism of the programs. The data currently cover eighteen countries: Australia, Austria, Belgium, Canada, Denmark, Finland, France, Germany, Ireland, Italy, Japan, the Netherlands, New Zealand, Norway, Sweden, Switzerland, the United Kingdom, and the United States.6 We combined the program characteristics described in table 3.1 to form indices of program generosity using the following procedure. We took the cross-country mean and standard deviation for each characteristic (except unemployment and sick pay insurance coverage and pension take-up) in 1980. (The choice of a base year is ultimately arbitrary, but 1980 is the year Esping-Andersen used to compute his influential welfare state decommodification index and is thus convenient for comparative purposes.) We obtained the standardized “g-score” for each characteristic in all years (1971 to 2002) by subtracting the benchmark year mean and dividing by the benchmark year standard deviation.7 We bounded these scores to be a maximum of two standard deviations from the mean. Then we added two to each score to make all values positive. [Value(i,t) – mean(Value)1980]/sd(Value)1980 + 2 = g-score(i,t)

(3.1)

There are separate program generosity indices for each of the three main social insurance programs: unemployment, sickness, and pensions. This is done by adding all of the computed g-scores for each social program characteristic from table 3.1 and multiplying that value by the coverage ratio. In the case of sick pay and unemployment, the coverage ratio is the number of insured workers divided by the total labor force; for pensions, it is the percentage of retirement-age people in receipt of some form of public pension.8 To further illustrate with an example, the unemployment insurance program score is calculated as follows: [Single replacement rate (RR)UE g-score + Family RRUE g-score + Qualifying PeriodUE g-score + Waiting PeriodUE g-score + Benefit DurationUE g-score] * UE coverage ratio

(3.2)

Table 3.1

Dimensions of the Decommodification Index

Core Program– Program Characteristics Unemployment insurance Single replacement rate

Family replacement rate

Qualifying period Waiting days Duration of benefit Coverage ratio Sickness benefit Retirement pension Minimum replacement rate (single) Minimum replacement rate (couple) Standard replacement rate (single)

Standard replacement rate (couple)

Qualifying period Contribution ratio

Take-up ratio Source: Author’s compilation.

Definition After-tax benefit for single, fully insured forty-year-old earning average production worker (APW) wage divided by after-tax wage of employed APW After-tax benefit for a family of four (one APW earner, nonworking spouse, and two children) divided by after-tax wage of employed APW Weeks of insurance-employment required to qualify for benefit Number of days before benefits start Weeks of benefits payable for fully insured (single) forty-year-old Percentage of the labor force covered by unemployment insurance All program characteristics defined the same as for unemployment insurance After-tax replacement rate at retirement for a single person with no work history (or income) After-tax replacement rate at retirement for a couple with no work history (or income) After-tax replacement rate for a single person with a full work history (maximum forty-five years) at APW wage After-tax replacement for a couple with one full work-history earner and spouse without a work history Years of insurance needed to qualify for single standard pension (defined above) Employee-employer + employee ratio of payroll taxes (at time pension is claimed) Portion of population above retirement age receiving pension

Social Rights, Welfare Generosity, and Inequality

69

What is referred to throughout this chapter as the “generosity index” is the sum of the three separate program indices. The decision to simply sum the three program scores is contentious, as it is with almost all efforts to create aggregates without an obvious numeraire. In part, the decision is based on precedent: Esping-Andersen’s previously mentioned decommodification index sums his scores for these programs. Our decision is also based on the fact that we have already “weighted” the score for each program based on its coverage of the target population in the country. (Indeed, we find that in some cases changing coverage of the population over time affects the generosity scores in nontrivial ways.) These scores avoid most of the problems associated with spending ratios. Since they are based on the conditions and wage replacement ratios facing a “typical” worker in each year, the measure of generosity is not conflated with the size of the dependent population or the rate of economic growth. Moreover, because benefits are based on post-tax wage replacement rates, they do not inflate the measured effort of countries with high nominal spending but high taxes on benefits. Of course, the index has its own shortcomings. First, the estimates rely on “notional” household types, which may be more or less representative in different countries or time periods. (Nonetheless, our analyses suggest that these notional households are reasonably representative.) Second, because the replacement rates are a function of trends in net benefits and net wages, it is possible that the benefit system allows absolute purchasing power to go up even though benefits replacement rates may stagnate or decline. On this account, our focus on relative replacement is consistent with the vast majority of the literature on inequality that focuses on relative, not absolute, living standards (for evidence that benefit generosity is also associated with improvements in absolute equality, see Scruggs and Allan 2006b). Figure 3.1 displays the average generosity index for three periods— 1971 to 1973, 1983 to 1986, and 2000 to 2002. We use three-year averages to focus attention on long-run trends. Most welfare states are at least as generous in the early twenty-first century as they were in the early 1970s. The mid-1980s represented a general shift toward greater generosity among all of the countries examined here, and there was some evidence of convergence. Mean generosity increased from 22.6 to 29.0, and convergence (measured by the coefficient of variation–standard deviation divided by the mean) for the eighteen countries declined from .30 to .27. During this time, historically less generous countries— Japan, Ireland, Switzerland, the United States—gained some ground on the more generous countries (the Scandinavian countries plus the Netherlands). Comparing the 1983 to 1986 period with 2000 to 2002, the generosity

70

Democracy, Inequality, and Representation

Figure 3.1

Trends in Benefit Generosity

Benefit Generosity (Three Year Average)

50 Sweden Norway

40

Denmark Norway Netherlands Sweden

30

20

10

Netherlands Norway Denmark Germany

Finland Belgium France Switzerland Germany New Zealand Austria

Belgium Finland France Canada Ireland Austria Italy Canada New Zealand United Kingdom Australia United States Switzerland Italy Japan Australia Ireland United States United Kingdom Japan

1973

1986

Sweden Netherlands Denmark Belgium Finland Austria Germany Canada

Ireland France Italy New Zealand

United Kingdom Japan Switzerland Australia United States

2002

Year Source: Scruggs, Comparative Welfare Entitlement Data (CWED).

index illustrates a clear pattern of downward convergence. Scores in three out of four countries are lower, the mean score declines, and the coefficient of variation shrinks still further. Retrenchment appears to be most pronounced in the most generous countries. Except for oil-rich Norway, which should probably be viewed as the unconstrained ideal of the social democratic vision, all of the countries with scores above the median in 1986 experienced retrenchment in generosity, eight of them by more than one point. Among the nine least generous countries, only four experienced retrenchment, and only two saw their scores fall by more than one point. Do these results suggest that contemporary welfare states will continue to experience retrenchment? On the one hand, generosity is higher in the early twenty-first century than it was in the 1970s in almost every country examined (Germany and Switzerland being the exceptions). On the other hand, generosity scores in the most recent period probably underestimate retrenchment for current workers. Since we calculated the pension index on the basis of what a current retiree receives, not what current workers can expect when they retire, reforms enacted during the

Social Rights, Welfare Generosity, and Inequality

71

1990s in several countries (for example, Japan, Sweden, and Italy) will probably result in reduced benefits to future retirees. Figure 3.2 illustrates trends over time in the generosity index and the social expenditure ratio for each of eighteen OECD countries. Reliable social expenditure figures are only available after 1980, so our comparative analysis is limited to this more recent period. Although we would prefer to compare trends during the period of expansion in welfare expansion, the post-1980 period does cover the era of welfare austerity. There is wide national diversity in these data, but several common patterns are notable. In most countries, welfare state generosity expanded considerably in the 1970s and early 1980s, for three main reasons. First, in many countries, some social benefits were established or greatly expanded in the 1970s. Canada instituted a sick pay program in 1971, and the United States established a national social pension (Supplemental Security Insurance, or SSI) program in 1972. A number of countries—Australia, Belgium, Finland, France, Ireland, Italy, Japan, the Netherlands, New Zealand, and Sweden—also considerably increased the replacement rates of their social pensions during the 1970s. In other cases, like Ireland, the program rules were simply made more generous across the board. Second, many countries increased the coverage of their social insurance programs during the 1970s, either through extension to previously uncovered categories of workers or owing to structural economic change. In the Nordic countries with Ghent unemployment insurance systems, coverage increased dramatically with rising union density in the 1970s (and throughout the 1980s). For example, in Denmark unemployment insurance coverage increased from 34 percent of the labor force in 1971 to 61 percent in 1980. Switzerland introduced compulsory unemployment insurance in 1976, increasing coverage from about 16 percent in 1974 to over 90 percent in 1977. The introduction of some other programs noted earlier also increased the universalism of pensions. For example, the introduction of Supplementary Security Insurance in the United States began to pay public pensions to some of those who did not receive traditional social security contributory pensions. A final explanation for increasing generosity scores is the maturation of earnings-related public pension schemes, which exist in almost all OECD countries. Often (re)created after the Second World War, these programs increased welfare state generosity during the 1970s for a couple of reasons. Earnings-related, defined-benefit pension systems pay benefits proportional to how long an individual has paid into the system, up to a maximum number of years and after a “vesting” period. Retirees in the 1950s and 1960s accrued benefits on only a fraction of that maximum, if at all, because they could not have built up enough contribution

72

Democracy, Inequality, and Representation

Figure 3.2

National Trends in Benefit Generosity and Social Spending Australia

Austria

Belgium

Canada

Denmark

Finland

France

Germany

Ireland

Italy

Japan

Netherlands

New Zealand

Norway

Sweden

Switzerland

United Kingdom

United States

40 30 20 10

40 30 20 10

40 30 20 10

40 30 20 10

40 30 20 10

40 30 20 10 1970

1980

1990

2000 1970

Benefit Generosity

1980

1990

2000 1970

1980

1990

2000

Year Social Spending (Percentage of GDP)

Source: OECD (2004); Scruggs, Comparative Welfare Entitlement Data (CWED).

Social Rights, Welfare Generosity, and Inequality

73

years. In some cases, when these pension systems were established, older workers were completely excluded from the system. However, by the 1970s almost all new retirees qualified for some pension benefits, and a typical worker came closer and closer to qualifying for full pension benefits under the law. It is notable that the main exceptions to the trend of increased generosity in the 1970s were those countries with more “mature” earnings-related pension systems: Austria, Belgium, Germany, Italy, and France. (These are all, of course, the “conservative” welfare states.) These pension programs could conceivably have paid “full” benefits already in the early 1970s.

Generosity Since the 1980s The individual country series in figure 3.2 demonstrate what is implied in figure 3.1: program generosity peaked in the 1980s for almost all countries. Declines were particularly marked in Denmark, Finland, the Netherlands, New Zealand, Sweden, and Switzerland, but they were also noticeable in France and Germany. Except for New Zealand, all of these countries are commonly considered to be the more generous welfare states. Ireland and Italy are the only two cases that appear to have increased the generosity of their programs in the last decade.

Spending Trends Trends in social (and other) public spending ratios are widely discussed in the rest of the literature. Figure 3.2 confirms that there is not a lot of evidence of spending retrenchment. Spending trended down after the 1980s only in the Netherlands and in Ireland. In the first case, this was due not only to some major reforms but also to welfare state “outsourcing.” In the second case, it was due to the country’s long-delayed economic miracle. (Ireland was a clear laggard during Europe’s golden age of economic growth through the 1970s.) In several countries—Australia, Austria, France, Germany, Italy, Japan, Norway, and Switzerland—social spending trended upward. Several countries had noticeable spending “bubbles” in the 1990s— Austria, Canada, Denmark, Finland, the Netherlands, Sweden, and the United Kingdom. It is impossible to tell from these series whether these are cyclical wobbles or evidence of spending retrenchment. For example, in none of the cases (save the Netherlands) were spending ratios lower after 2000 than they were in the late 1980s. Viewed by themselves, these spending trends might in fact justify optimism about the resilience of Europe’s generous social programs.

74

Democracy, Inequality, and Representation

We believe, however, that there are several reasons not to be overly sanguine. First, while current spending may not be below the levels of a decade or so ago, dependency ratios have tended to grow. Second, and more importantly, cuts in benefit generosity typically preceded recessioninduced spending bubbles in the 1990s. In Denmark, Finland, France, the Netherlands, New Zealand, and Sweden, large declines in generosity preceded declines in spending by five to ten years.

Are Spending and Generosity the Same? The use of spending data in the absence of other measures of welfare generosity has occasionally been justified by asserting that spending and benefit generosity are closely correlated empirically. The obvious problem is that there has been no comprehensive indicator of generosity to substantiate that claim. Our generosity index allows for one. We first looked at the correlation between the generosity index and three popular spending ratios in 1999: total government outlays, transfer spending, and total social expenditure. We would certainly expect some association between spending and generosity, if for no other reason than the fact that pension spending is a large part of state spending and most people can expect to draw a public pension. The correlations (for 1999) are, respectively, .74, .50, and .66.9 Thus, at best, spending explains only about half of the variation in the generosity index. Although we could argue that the deficiency lies in the generosity measure, the larger point persists: spending and our generosity index are not measuring the same thing. In our previous critique of spending ratios, we suggested that spending ratios cannot be transformed into measures of generosity simply by conditioning spending on the number of dependents in a regression equation. Using data on spending, generosity, and dependence, we can empirically evaluate that claim. If spending, controlling for dependency rates, is a good measure of generosity, we would expect a close positive relationship between it and our index. Figure 3.3 presents a scatterplot of the generosity index against the residuals from a regression of social spending on unemployment and the portion of the population over age sixtyfive. Spending adjusted for dependency and the current state of the economy is higher in generous Nordic countries and low in more “liberal” welfare states like Australia, Italy, Japan, the United Kingdom, and the United States.10 The correlation coefficient of .63 is indeed respectable, but hardly the kind of association expected of a good proxy variable.11 Moreover, the correlation is slightly lower than the unconditional correlation between generosity and spending. If spending were a good proxy for generosity, moreover, we would

Social Rights, Welfare Generosity, and Inequality Figure 3.3

75

Conditional Spending Ratios and Benefit Generosity

Spending Model Residuals

10

5

0

−5

−10 20

25

30

35

40

45

Benefit Generosity Source: Author’s calculations.

also expect to see a positive correlation within each country or across annual cross-sections. There is in fact a consistent correlation in year-toyear cross-sections, but not within individual countries. The within-correlation is effectively zero (rho < .30) in four cases—Austria, Canada, Ireland, and the United States; significantly positive (rho > .30) in eight cases—Finland, France, Germany, the Netherlands, Norway, New Zealand, Sweden, and the United Kingdom; and negative (rho < –.30) in six cases—Australia, Belgium, Denmark, Italy, Japan, and Switzerland. To summarize the analysis in this section, there are three main findings. First, there are some considerable differences, particularly within countries, between the generosity of the welfare state programs and the level of welfare spending ratios. These differences are, at best, only partly accounted for by differences in the portion of the dependent population. Second, based on the institutional details of the major social insurance programs, there is considerably more evidence of welfare state retrenchment during the last two decades than the spending data suggest. This program retrenchment is likely to continue into the near future, as recently enacted cuts in some public pension programs affect more and

76

Democracy, Inequality, and Representation

more retirees. Third, though it might be manifest in slightly different ways—for example, in higher or lower replacement rates or in less or more universal coverage—the generosity of social insurance programs has tended to converge internationally during the last three decades. This does not mean that the welfare state is everywhere undergoing retrenchment and convergence. But since the social programs we deal with in the generosity index have been, and remain, very important elements of welfare state protection, they are substantively important.

Inequality, Redistribution, and Welfare Generosity Establishing that welfare state generosity and need-adjusted spending are empirically distinct motivates the second part of this chapter: do differences in generosity matter for income redistribution? This section provides some empirical results based on a reestimation of two recent studies of fiscal redistribution (Bradley et al. 2003; Möller et al. 2003). The results suggest that not only is welfare generosity a more conceptually desirable indicator of welfare program commitments than is spending, but that it is also a better predictor of redistributive outcomes. All of the redistribution indicators evaluated here are derived from income data from the Luxembourg Income Study (LIS). The advantages of this data set over previous international income distribution data, and some of its limitations, are outlined by Brandolini and Smeeding (this volume).12 We concentrate here on two key indicators of inequality and redistribution:

1.

Reduction in relative poverty rates: Poverty is defined as the percentage of the household equivalents living below 50 percent of the median national income.13 Poverty reduction is computed as the difference between poverty rates (50 percent of national median) based on market income and on disposable income.

2.

Difference in the market and disposable income Gini coefficient: This indicator is derived from the LIS files and is discussed in several previous papers (Bradley et al. 2003; Hicks and Swank 1984; Korpi and Palme 1998; Milanovic 2000; Pontusson 2005) and throughout this volume. It is also defined using the equivalency conversions discussed earlier.

It is important to note that what has sometimes been referred to as “market income” by previous users of the LIS data is in fact post-tax (but

Social Rights, Welfare Generosity, and Inequality

77

pre-transfer) income. When this is the case, a country’s poverty measure is biased upward, for two reasons. First, the estimated poverty threshold has a downward bias: measured median income is lower because it excludes all taxes. This means that too few people in these cases are assigned to “living in poverty.” Second, individuals with market and posttax, post-transfer incomes that are above poverty may be measured as “poor” based on post-tax, pre-transfer income. Thus, they are counted as being poor in the market but removed from poverty by the tax-andtransfer system when in fact they are both taken into and lifted from poverty by the tax-and-transfer system. This is most likely to be the outcome where many households pay high taxes and receive high transfers. Moreover, in these cases of mismeasured market income, reductions in the Gini coefficient are also biased upwards. This may happen because measured incomes are bounded from below (for example, no one is taxed below zero income) and there is some progressiveness in the tax system. Both can make the computed “pre-fisc” Gini coefficient lower than the true pre-fisc Gini. To avoid these two problems, our estimates include only countryyears with truly pre-fisc income. This eliminates all values for Italy, Austria, and Ireland and the 2000 value for Belgium.14

Bivariate Correlations We begin by presenting scatterplots of the relationship between redistribution and benefit generosity. For each of the two main indicators, we use measures of redistribution among all households.15 Data for all figures in this section are from the most recent year for which LIS data were available, generally 1999 or 2000. Figure 3.4 depicts the relationship between benefit generosity and the reduction in the poverty rate via the tax-and-transfer system—that is, the percentage decline in market and post-transfer poverty. Reductions in poverty appear to be closely related to benefit generosity. A plot of post-fisc poverty against generosity (not shown) has a similar shape, suggesting that there is a negative association between post-fisc poverty and the percentage reduction in poverty but little association between prefisc poverty and subsequent poverty reductions. On average, income poverty rates decline by over 60 percent as a result of the tax-and-transfer system. The smallest reduction is in the United States (28 percent), followed by Australia (41 percent), while the largest reductions are in France (79 percent) and Sweden (78 percent). Figure 3.5 displays the relationship between benefit generosity and the reduction in the Gini coefficient as a result of income taxes and transfers. Benefit generosity is positively associated with a reduction in the Gini. These results are also consistent with expectations.

78

Democracy, Inequality, and Representation

Reduction in Relative Poverty (All Households)

Figure 3.4

Benefit Generosity and Relative Poverty Reduction France

80

Sweden

Belgium Germany Finland

70

Denmark

Norway

Netherlands United Kingdom

60

Switzerland Canada

50 Australia 40

30

United States 20

25

30

35

40

Benefit Generosity Source: Scruggs, Comparative Welfare Entitlement Data (CWED), Luxembourg Income Study (LIS).

Multivariate Models of Redistribution The preceding results suggest that more generous benefit systems are associated, as we might expect, with greater reductions in poverty and inequality. However, bivariate analysis does not take into account potentially confounding variables. In this section, we reestimate the multivariate models used in two recent studies of income redistribution, Stephanie Möller et al. (2003) and David Bradley et al. (2003).16 Table 3.2 provides a summary of the variables used in these models. Since both studies evaluate redistribution among the non-elderly population (specifically households headed by those age twenty-five to sixty), we report results only for that subset of the population. In addition to evaluating whether the relationships suggested in the bivariate analysis are robust, we can also evaluate whether generosity provides greater explanatory power than spending measures alone. As table 3.2 details, most of the independent variables in these models can be found in the Comparative Welfare States Data Set (Huber et al. 2004). “Per capita income” is from the Penn World Table and is measured

Social Rights, Welfare Generosity, and Inequality

Reduction in Gini Coefficient (All Households)

Figure 3.5

79

Benefit Generosity and Reduction in Income Inequality

.5

Belgium Germany

Denmark Sweden

France .4

Norway

Australia .3

Finland

Netherlands

Canada United Kingdom Switzerland

.2

United States 20

25

30

35

40

Benefit Generosity Source: Scruggs, Comparative Welfare Entitlement Data (CWED), Luxembourg Income Study (LIS).

in thousands of dollars. “Wage dispersion” is the ratio between pre-fisc incomes at the ninetieth and tenth percentiles.17 “Unemployment rate” and “female-headed households with children” are self-explanatory. “Capital market openness” refers to the absence of four legal restrictions to the capital account, as discussed in Dennis Quinn (1997). The range is between two and four. “Wage coordination” is the degree of centralization of wage bargaining in a given year. “Christian Democratic cabinet” and “left cabinet” each refer to the cumulative portion of cabinet seats held by a Christian Democratic or left party. “Constitutional veto points” refers to the number of power-dispersing institutional arrangements in a country’s basic political system (Huber, Ragin, and Stephens 1993). “Welfare spending” is the sum of the standardized values of total revenue as a portion of GDP and transfer spending as a portion of GDP. In estimating the Gini redistribution model, we use the full generosity index as a regressor. In the poverty reduction model, we use the sum of unemployment and sickness program scores only. We make this distinction for two reasons. First, some portion of the pension system is funded from general taxes, which are transferred from the working population

80

Democracy, Inequality, and Representation

Table 3.2

Variables Used in Different Regression Models

Variable Dependent variables Relative poverty reduction (under sixty-fiveyear-old households) Reduction in Gini coefficient (twenty-five to fifty-nine households) Independent variables Per capita income Wage disperison Unemplyoment rate Female-headed households with children Capital market openness Wage coordination Christian Democratic cabinet (cumulative from 1945) Constitutitional veto points Left cabinet (cumulative from 1945) Welfare spending Benefit generosity

Source

Bradley et al. (2003) (Gini)

LIS data files (author)

x

LIS data files (author)

x

Penn World Table (in HRS) LIS data files (author) OECD LIS data tables Quinn (in HRS)

x

Kenworthy (in HRS) Swank (in HRS) Lijphart (in HRS) Swank (in HRS) OECD (in HRS) CWED (author)

Moller et al. (2005) (Poverty Rate)

x x x

x

x x x

x x

x

x (x)

x (x)

Note: CWED—Scruggs, Comparative Welfare Entitlement Data; HRS—Huber, Ragin, and Stephens, Comparative Welfare State Data.

to those in retirement. Since both sets of results that we are replicating estimate models wherein redistribution is measured only in the workingage population, and since general taxes are at least moderately progressive, a more generous pension system should exert a negative effect on the post-fisc overall income distribution in the working-age population.

Social Rights, Welfare Generosity, and Inequality

81

Second, more generous public systems substitute for some private saving, which shows up in the LIS data as less inequality.18 However, the generosity of the pension system should not be particularly relevant for measuring the reduction in the head-count poverty rate in the workingage population. The first part of table 3.3 provides estimates of the Bradley et al. (2003) model. Following their approach, all of our parameters are estimated with ordinary least squares (OLS) and robust-cluster standard errors, which correct, among other things, for within-unit error correlation. The first column shows their reported estimates. The next two sets of estimates suggest that the generosity index provides (somewhat) better results than the spending measure. Using the generosity index increases the fit of the model and the standardized beta. When the effects of both variables are estimated simultaneously, the fit of the model is unchanged (actually it is slightly worse), the spending coefficient is substantively small and statistically insignificant, and the benefit generosity coefficient is more or less unchanged. All of these results suggest that generosity is a better predictor of redistribution than spending. Figure 3.6 compares the residuals from the spending and generosity models plotted against the actual values for redistribution. The generosity model reduces the average squared residual considerably for seven countries (Australia, Belgium, Canada, Denmark, Italy, the Netherlands, Sweden, and the United Kingdom), but also increases it in four (Finland, Ireland, Switzerland, and Norway). Overall, the root mean squared error in the generosity model is 7 to 8 percent lower than it is in the spending model. Table 3.4 presents estimates based on Möller et al.’s (2003) model of reduction in poverty rates via the tax-and-transfer system. Their model is slightly different from that of Bradley et al. (2003), but the results show a similar pattern to what we observed in table 3.3. Substituting the generosity index for spending improves the fit of the model and gives a larger standardized beta coefficient. Entering both variables (spending and the generosity index) also supports the idea that generosity is a stronger predictor: the spending coefficient drops considerably and is not statistically different from zero, while the generosity index coefficient drops only slightly. However, the uncertainty about the generosity estimate is higher (the p value is about .11).19

Conclusions The main goal of this chapter has been to present an alternative indicator of welfare state commitment than traditional spending ratios. The benefit generosity index introduced here provides comparative welfare state re-

.82

20.28 (3.87)** 59

Source: Author’s calculations from Bradley et al. (2003). Note: Robust t statistics (absolute value) are in parentheses + p < .10; * p < .05; ** p < .01

Observations Root mean squared error R-squared

Constant

Benefit generosity

Spending

Christian Democratic cabinet share Left cabinet share

Single female households with children Capital market openness

Unemployment

Income per capita

–.12 (.10) –.17 (1.16) 0.74 (3.26)** .39 (2.41)* –1.23 (1.13) –.21 (2.22)* .21 (1.73)+ 2.83 (5.88)**

Coefficient

.61

.27

–.29

–.09

.23

.29

–.11

–.01

Beta

Bradley et al. (2003) Estimates

Percentage Reduction in Gini Coefficient

23.7 (5.85)** 57 5.19 .73

.002 (.06) –.12 (.38) .92 (2.53)* .36 (1.55) –.83 (1.75) –.14 (.76) .30 (2.62)* 2.27 (3.27)**

Coefficient

.47

.37

–.18

–.15

.19

.31

–.04

.01

Beta

.59 (3.91)** 7.30 (1.25) 57 4.81 .76

–.005 (.25) –.60 (1.91)+ 1.36 (4.71)** .61 (2.74)* –.48 (.99) –.10 (.70) .26 (2.58)*

Coefficient

.54

.31

–.13

–.09

.32

.46

–.22

–.03

Beta –.004 (.22) –.56 (1.99)+ 1.32 (4.85)** .60 (2.62)* –.52 (1.10) –.11 (.71) .25 (2.07)+ .23 (.21) .55 (2.22)* 8.59 (.97) 57 4.86 .77

Coefficient

(Head of Household Aged Twenty-Five to Fifty-Nine)

Results for Pre- and Post-Fisc Reductions in Gini Coefficents

Income dispersion

Table 3.3

.50

.05

.30

–.14

–.10

.31

.44

–.20

–.02

Beta

Social Rights, Welfare Generosity, and Inequality Figure 3.6

Residuals from Spending and Generosity Models

20

be be

nl

10

0

ch ch ge it it it

nl be

nl

sw be fi fi nl us ca fr uk ca ca ca dk us ca cage ca sw no usus ge ge ge sw ge uk ir no fr uk dk fr ge dk no iruk uk no ir as asas

−10

83

fi sw

sw

ir −20 20

10

30

40

50

Reduction in Gini Coefficient (Households, Aged Twenty-Five to Fifty-Nine) Spending Residuals

Generosity Residuals

Source: Author’s calculations.

search with a conceptually superior way to account for the extent of, and changes in, social insurance against risks—arguably the most essential purpose of a welfare state. Insofar as a commitment to insurance against risk does not show up in the level of welfare spending, relying primarily on what welfare states spend misses an essential component of what welfare states do. One can argue that the index does not encompass enough welfare programs, though it does capture very important ones. Our efforts have convinced us that the best answer to such objections is to extend institutional measures like these rather than to continue to rely on spending outputs. More substantively, the results in this chapter establish two things about welfare states in advanced democracies. First, the era of the 1970s to the mid-1980s was generally one of considerable expansion in the generosity of major social insurance benefit programs. Since then, however, welfare commitments have stagnated or contracted, and they have done so more in those welfare states that were, in the 1980s at least, the

.91

21.75 (5.56)** 61

42.83 (5.90)** 57 9.19 .81

1.68 (3.06)** 2.47 (1.88)+ –4.29 (4.81)** –.212 (.77) 3.93 (3.13)**

Coefficient

.37

–.12

–.50

.18

.25

Beta

1.62 (2.88)* 20.64 (2.36)* 57 8.24 .85

2.62 (6.39)** .21 (.12) –4.84 (5.22)** –.32 (1.42)

Coefficient

.53

–.18

–.57

.01

.39

Beta

2.43 (4.58)** .37 (.19) –4.60 (5.77)** –.33 (1.43) .94 (.48) 1.43 (1.71) 23.56 (1.79) 57 8.28 .85

Coefficient

(Head of Household Age Less than Sixty-Five)

Reduction in Relative Poverty

.47

.09

–.18

–.54

.03

.36

Beta

Source: Author’s calculations from Möller et al. (2003). Note: Robust t-statistics in parentheses. Möller et al. (2003) results are based on households headed by twenty-five- to fifty-nine-year olds, not all under sixty-five. p < .10; * p < .05; ** p < .01

Observations Root mean squared error R-squared

Constant

Benefit generosity (labor market)

Spending

Left cabinets

Veto points

Wage coordination

1.52 (5.07)** 1.49 (1.79) –2.1 (3.96)** .63 (5.25)** 4.00 (6.35)**

Coefficient

Möller et al. (2003) Results

Results for Percentage Pre- and Post-Fisc Reductions in Poverty Rates

Unemployment

Table 3.4

Social Rights, Welfare Generosity, and Inequality

85

most generous. We hope that these results will open a new chapter in the debate about causes of stagnation and retrenchment, since they should make us less sanguine about the strength of welfare state commitments in the face of contemporary challenges like aging, globalization, and more “ideological” assaults on state intervention. The second major substantive finding of the chapter is that countries with more generous welfare commitments, as indicated by the generosity of their social insurance programs, have fiscal systems that reduce poverty and overall inequality more than do countries with less generous benefit systems. We hope that these findings will provoke both a theoretical and empirical reexamination of the determinants and effects of the welfare state. An increasing body of historical and theoretical work is honing in on the fact that welfare programs are often fundamentally about the distribution of risks. By moving past a focus on the size of the welfare budget, we can move a step closer to examining patterns of social protection, not simply patterns of social spending.20

Notes 1.

2.

3.

4.

5.

The program data are available at the CWED website, http://sp.uconn .edu/~scruggs/welproj.htm. Support for the collection of this data was provided by the National Science Foundation (SES-0095367). Another factor that is not discussed here is the role of forced private social spending (Adema and Ladaique 2005). As Johan De Deken and Bernhard Kittel (2007) note, this is a category that casts serious doubt on the comparability of social spending series like the OECD Social Expenditure Database. Note that simply controlling for the over-sixty-five population on the righthand side of a regression equation does not do anything to differentiate the two cases. By construction, the retiree population is the same; what differs is the replacement rate of the public system. The Social Citizenship Indicator Project, under the direction of Walter Korpi at the Swedish Institute for Social Research, reportedly has collected very similar data for a similar set of countries for more than two decades (Kangas 1991; Korpi 1989; Korpi and Palme 1998, 2003). However, because these data are not publicly available to the scholarly community, they have not been widely used in the comparative welfare state literature. Nonstatutory benefits, even if applicable to most of the workforce (for example, supplementary pensions negotiated and implemented by social partners and extended by decree), are excluded, but legal requirements, like compulsory sick pay from employers, are included. Lyle Scruggs and James

86

6. 7.

8.

9. 10.

11.

12.

13.

14.

Democracy, Inequality, and Representation Allan (2006a) and Scruggs (2007) provided a comparative analysis of important substantive and methodological problems with Esping-Andersen’s index. Details about the elements of the index, sources, and details of the calculation of replacement rates are available in the data files and codebook, all of which are available at the CWED website. Eventually, the data set will be expanded to include similar information on family and maternity benefits and extended to cover more OECD countries. In a few cases, I dropped extreme values in computing the benchmark values. Precise information on the benchmark values and computation of benchmark scores can be found in the source code used to generate the scores (available from the CWED site) and also in Scruggs (2007) and Scruggs and Allan (2006a). It is important to emphasize that the coverage ratio for unemployment and sick pay does not refer to the percentage of the unemployed or sick in receipt of benefits (that is, what we call benefit “take-up”), but to the percentage of the labor force currently insured. Because some currently unemployed or sick individuals have exhausted their benefit entitlement, take-up rates for unemployment and sick pay may be lower than the coverage rate. One reason we did not use take-up rates for these programs is that they are extremely hard to compute and interpret over time with the available data. It is important to point out that at least two things that reduce measured take-up rates—more stringent qualifying conditions and shorter benefit periods—reduce a country’s generosity score in our measure. The correlation coefficients vary somewhat (higher and lower) depending on the year and spending measure used, but the 1999 results are representative. Though the benefit generosity index produces scores consistent with Esping-Andersen’s decommodification index, there are also a number of important discrepancies (Scruggs and Allan 2006a). A similar analysis using all government spending instead of just social spending produced even poorer results. The overall correlation coefficient there was .55. We generally follow coding procedures discussed in Brandolini and Smeeding’s chapter: for example, income is top-coded at ten times the mean and bottom-coded to eliminate negative gross income. Income is computed on an individual-equivalent basis from household reporting, that is, an individual’s income is assumed to be the total household income divided by the square root of the number of people in that household. Household equivalency assigns each member of a household with household income X and an income level equal to X/sqrt(N), where N is the number of people in the household (see Bartolini and Smeeding, this volume; Beramendi and Cusack, this volume). Results that we replicate later in the chapter include some of these cases.

Social Rights, Welfare Generosity, and Inequality

15.

16.

17.

18.

19. 20.

87

Where we identify them as having been included, we also include them in that replication analysis. The reasons for using “working-age” households—which is prevalent in the literature and which we do here—are not really well grounded conceptually and empirically. One-third or more of the electorate (those over age sixty and under twenty-five) is excluded by that definition. Like Baramendi and Cusack (this volume), we note that ignoring these welfare-dependent (and enfranchised) segments of the population removes an important source of variation, both across countries and time, that is relevant to democratic politics. The set of independent variables included is based on the “combined” model in Bradley et al. (2003) and model 6 in Möller et al. (2003). In all of these estimates, we rely on the available panel data for all countries considered in our earlier cross-sectional scatterplots. To maximize the comparability of these results with those in previous work, we try to include the country-years used in those studies. (There is a discrepancy between the number of cases reported in regression tables in Bradley et al. [2003] and the country-years presented in their tables. We included only those years specifically listed.) Vocational training is omitted because the data were not available. The low salience of this variable in the study by Bradley and his colleagues—its substantive and statistical significance is low and is omitted in their “reduced” model—suggests that the omission should not have a severe impact on our results. There is some ambiguity in Bradley et al. (2003) about what this ratio is. The description in their paper (207) suggests that it is the 90/10 wage ratio for full-time workers, but figures for this variable are not, to our knowledge, available for many of the observations in their data set. The text of the paper (212) suggests that they used the post-fisc 90/10 income ratio derived from the LIS data files. However, our regression results were close to theirs only if we used the pre-fisc 90/10 income ratio using the LIS files. Since that measure is, in any case, the most appropriate measure available based on their description of what they were interested in controlling for, this is what we used. LIS counts any money that might go to private savings as disposable income, but money going to a public pension is a tax deducted from disposable income. Thus, a more generous public pension for the middle class means lower measured disposable income for them. (Whether the net effect is more or less lifetime income is not relevant.) The variance inflation factor score for the two welfare variables is over five. For analyses of the political foundations of redistribution from a social protection perspective, see Thomas R. Cusack, Torben Iversen, and Philipp Rehm (this volume).

88

Democracy, Inequality, and Representation

References Adema, Willem, and Maxime Ladaique. 2005. Net Social Expenditure: 2005 Edition. Social Employment and Migration Working Papers 29. Paris: Organization for Economic Cooperation and Development. Bradley, David, Evelyne Huber, Stephanie Möller, François Nielsen, and John D. Stephens. 2003. “Distribution and Redistribution in Post-industrial Democracies.” World Politics 55(2): 193–228. Cameron, David. 1978. “The Expansion of the Public Economy: A Comparative Analysis.” American Political Science Review 72(4): 1243–61. Castles, Frank. 2002. “Developing New Measures of Welfare State Change and Reform.” European Journal of Political Research 41(5): 613–41. ———. 2004. The Future of the Welfare State: Crisis Myths and Crisis Realities. Oxford: Oxford University Press. Castles, Frank, and Deborah Mitchell. 1993. “Three Worlds of Welfare Capitalism or Four?” In Families of Nations, edited by Frank Castles. Brookfield, Vt.: Dartmouth. Clasen, Jochen, and Nico Sigel, editors. 2007. Investigating Welfare State Change: The “Dependent Variable Problem” in Comparative Analysis. Cheltenham, UK: Edward Elgar. De Deken, Johan, and Bernhard Kittel. 2007. “Putting the Chainsaw into Social Expenditure.” In Investigating Welfare State Change: The “Dependent Variable Problem” in Comparative Analysis, edited by Jochen Clasen and Nico Sigel. Cheltenham, UK: Edward Elgar. Esping-Andersen, Gøsta. 1990. The Three Worlds of Welfare Capitalism. Princeton, N.J.: Princeton University Press. Garrett, Geoffrey. 1998. Partisan Politics in the Global Economy. Cambridge: Cambridge University Press. Gilbert, Neil, and Ailee Moon. 1988. “Analyzing Welfare Effort: An Appraisal of Comparative Methods.” Journal of Policy Analysis and Management 7(2): 326–40. Goodin, Robert, Bruce Headey, Ruud Muffels, and Henk-Jan Dirven. 1999. The Real Worlds of Welfare Capitalism. Cambridge: Cambridge University Press. Green-Pedersen, Christoffer, and Markus Haverland. 2002. “The New Politics of the Welfare State and the New Scholarship of the Welfare State.” Journal of European Social Policy 12(1): 43–51. Hicks, Alexander. 1999. Social Democracy and Welfare Capitalism: A Century of Income Security Politics. Ithaca, N.Y.: Cornell University Press. Hicks, Alexander, and Duane Swank. 1984. “Governmental Redistribution in Rich Capitalist Democracies.” Policy Studies Journal 13(2): 265–86. Howard, Christopher. 1997. The Hidden Welfare State: Tax Expenditures and Social Policy in the United States. Princeton, N.J.: Princeton University Press. Huber, Evelyn, Charles Ragin, and John Stephens. 1993. “Social Democracy,

Social Rights, Welfare Generosity, and Inequality

89

Christian Democracy, Constitutional Structure, and the Welfare State.” American Journal of Sociology 99(3): 711–49. Huber, Evelyne, Charles Ragin, John D. Stephens, David Brady, and Jason Beckfield. 2004. Comparative Welfare States Data Set. Northwestern University, University of North Carolina, Duke University, and Indiana University. Huber, Evelyn, and John Stephens. 2001. Development and Crisis of the Welfare State. Chicago: University of Chicago Press. Iversen, Torben, and Thomas Cusack. 2000. “The Causes of Welfare State Expansion: Deindustrialization or Globalization?” World Politics 52(3): 313–49. Kangas, Olli. 1991. “The Bigger the Better?: On the Dimensions of Welfare State Development: Social Expenditures Versus Social Rights.” Acta Sociologica 34(1): 33–44. Kitschelt, Herbert. 2001. “Partisan Competition and Welfare State Retrenchment: When Do Politicians Choose Unpopular Policies?” In The New Politics of the Welfare State, edited by Paul Pierson. Oxford: Oxford University Press. Korpi, Walter. 1983. The Democratic Class Struggle. Boston: Routledge & Kegan Paul. ———. 1989. “Power, Politics, and State Autonomy in the Development of Social Citizenship: Social Rights During Sickness in Eighteen OECD Countries Since 1930.” American Sociological Review 54(3): 309–28. Korpi, Walter, and Joakim Palme. 1998. “The Paradox of Redistribution and Strategies of Equality: Welfare State Institutions, Inequality, and Poverty in the Western Countries.” American Sociological Review 63(5): 661–87. ———. 2003. “New Politics and Class Politics in the Context of Austerity and Globalization: Welfare State Regress in Eighteen Countries, 1975–1995.” American Political Science Review 97(3): 425–46. Lindert, Peter. 2004. Growing Public. Cambridge: Cambridge University Press. Milanovic, Branko. 2000. “The Median Voter Hypothesis, Income Inequality, and Income Redistribution: An Empirical Test with the Required Data.” European Journal of Political Economy 16(3): 367–410. Möller, Stephanie, David Bradley, Evelyn Huber, François Nielsen, and John Stephens. 2003. “Determinants of Relative Poverty in Advanced Capitalist Democracies.” American Sociological Review 68(1): 22–51. Organization for Economic Cooperation and Development (OECD). 2004. Social Expenditure Database. Paris: OECD. Pierson, Paul. 1996. “The New Politics of the Welfare State.” World Politics 48(2): 143–79. Pontusson, Jonas. 2005. Inequality and Prosperity: Social Europe Versus Liberal America. Ithaca, N.Y.: Cornell University Press. Pontusson, Jonas, and Richard Clayton. 1998. “Welfare-State Retrenchment Revisited: Entitlement Cuts, Public-Sector Restructuring, and Inegalitarian Trends in Advanced Capitalist Societies.” World Politics 51(1): 67–98.

90

Democracy, Inequality, and Representation

Quinn, Dennis. 1997. “Correlates of Change in International Financial Regulation.” American Political Science Review 91(3): 531–51. Scruggs, Lyle. 2007. “Welfare State Generosity Across Space and Time.” In Investigating Welfare State Change: The “Dependent Variable Problem” in Comparative Analysis, edited by Jochen Clasen and Nico Sigel. Cheltenham, UK: Edward Elgar. Scruggs, Lyle, and James Allan. 2006a. “Welfare State Decommodification in Eighteen OECD Countries: A Replication and Revision.” Journal of European Social Policy 16(1): 55–72. ———. 2006b. “The Material Consequences of Welfare States Benefit Generosity and Absolute Poverty in Sixteen OECD Countries.” Comparative Political Studies 39(7): 880–904. Swank, Duane. 2002. Globalization, Political Institutions, and Policy Change in Developed Welfare States. New York: Cambridge University Press. Wilensky, Harold. 1975. The Welfare State and Equality. Berkeley, Calif.: University of California Press. Wilensky, Harold, and Charles Lebeaux. 1958. Industrial Society and Social Welfare. New York: Russell Sage Foundation.

Part II

How Democratic Politics Shapes Inequality

Chapter 4

Electoral Institutions, Parties, and the Politics of Class: Explaining the Formation of Redistributive Coalitions TORBEN IVERSEN AND DAVID SOSKICE

There is considerable variation in the extent to which governments redistribute income, and there is broad agreement that the explanation for such redistribution lies in the design of political institutions and partisan responses to inequality (see also the chapters by Brandolini and Smeeding, Beramendi and Cusack, and Rueda, this volume). But just how politics shapes distributive politics is still not well understood. Allan Meltzer and Scott Richard’s (1981) political economy model of redistribution, which is the best known, captures the key intuition that democratic institutions empower those who stand to benefit from redistribution and that redistribution is greater the more unequal the distribution of income. There is some evidence to support the first implication, although it is disputed (see Ross 2006), but most of the variance in redistribution is probably within the same regime type. According to data from the Luxembourg Income Study (LIS), for example, the reduction in the poverty rate in the United States as a result of taxation and transfers was 13 percent in 1994, whereas the comparable figure for Sweden was 82 percent. (The poverty rate is the percentage of households below 50 percent of the median income.) To explain this variance, we have to look at political and economic differences among democracies, and here the second implication—that inegalitarian societies redistribute more—turns out to be of little help. In fact, the empirical relationship between inequality and redistribution is the opposite of the predicted one (see Benabou 1996; Moene

94

Democracy, Inequality, and Representation

and Wallerstein 2001; Perotti 1996). Not only does Sweden redistribute more than the United States does, but it is also a much more egalitarian society. So the explanation for why some democracies redistribute more than others would seem to lie more or less wholly outside the standard framework in political economy to explain democratic redistribution. This chapter seeks to help fill this gap in our understanding. One possible explanation is that the power of the working class and left political parties varies across countries (see, for example, Hicks and Swank 1992; Huber and Stephens 2001; Korpi 1983, 1989). Since it is plausible that redistribution is a function of government policies, and since such policies reflect the preferences of those who govern, looking for differences in government partisanship is a promising avenue. Furthermore, if left governments not only redistribute more but also reduce the inequality of earnings by, say, investing heavily in public education, partisanship may also explain why equality and redistribution tend to covary. Indeed, there is much evidence to the effect that government partisanship helps explain cross-national differences in redistribution (Boix 1998; Bradley et al. 2003; Kwon and Pontusson 2003), and our findings corroborate this evidence (see also the chapters by Beramendi and Cusack and by Rueda, this volume). But looking at government partisanship raises another puzzle: why are some democracies dominated by left governments while others are dominated by right governments? Although government partisanship is often assumed to reflect the level of working class mobilization, we argue that it is in fact mainly determined by the differences in coalitional dynamics associated with particular electoral systems. Table 4.1 shows the strong empirical relationship using a new data set on parties and legislatures (see Cusack and Engelhardt 2002; Cusack and Fuchs 2002). The figures are the total number of years with right and left governments in seventeen advanced democracies between 1945 and 1998, organized by type of electoral system. Mirroring a similar finding by Bingham Powell (2002), we find that about three-quarters of governments in majoritarian systems were center-right, while three-quarters of governments under PR (proportional representation) were center-left (excluding here “pure” center governments). The numbers in parentheses convey a sense of the evidence at the level of countries, which are classified according to whether they had an overweight (more than 50 percent) of center-left or center-right governments during the 1945 to 1998 period. Our explanation for the association in table 4.1 builds on an emerging literature on the effects of electoral formula on economic policies and outcomes (see, for example, Austen-Smith 2000; Persson and Tabellini 1999, 2000, 2003; Rogowski and Kayser 2002). In particular, we argue that the electoral formula affects coalition behavior and leads to system-

Electoral Institutions, Parties, and the Politics of Class Table 4.1

95

Electoral System and Number of Years with Left or Right Governments, 1948 to 1998 Government Partisanship

Electoral system Proportional Majoritarian

Left

Right

342 (8) 86 (0)

120 (1) 256 (8)

Proportion of Right Governments 0.26 0.75

Source: electoral system, Lijphart (1994); government partisanship: Cusack and Fuchs (2002), Cusack and Engelhardt (2002). Note: Excludes governments that are classified as “centrist” on the Castles and Mair scale (Castles and Mair 1984).

atic differences in the partisan composition of governments—hence to different distributive outcomes. The explanation we propose assumes that parties represent classes, or coalitions of classes, and that it is difficult for parties to commit credibly to electoral platforms that deviate from the preferences of their constituents. We make a critical departure from the standard models based on Meltzer-Richard by allowing taxes and transfers to vary across classes, thereby transforming redistributive politics into a multidimensional game. In particular, we move away from a simple rich-poor model to one in which the middle class fears taxation by the poor, even as it faces an incentive to ally with the poor to take from the rich. The only constraint is that the rich cannot “soak” the middle class and poor under democracy—a condition that can be justified on empirical, normative, and institutional grounds. Based on these very general assumptions, our argument is that in a two-party majoritarian system the center-right party is more likely to win government power, and redistribute less, than would be true in a multiparty PR system, where the center party is more likely to ally with parties to its left. The intuition is that in a majoritarian system where parties cannot fully commit, the median voter faces low taxes if a centerright party deviates to the right if elected, but high taxes and redistribution to low-income groups if a center-left party in government deviates to the left. With PR, on the other hand, the middle class party has an incentive to form a coalition with the left party because together they can “exploit” the rich. No such exploitation of the poor is feasible under realistic assumptions. Remarkably, therefore, the same set of assumptions about redistributive policies leads to opposite predictions about government partisanship depending on the electoral system. We also discuss

96

Democracy, Inequality, and Representation

how these results are modified by the presence of large Christian Democratic (CD) parties, which organize groups from different classes, and we derive implications for understanding the distribution of pre-fisc income. We test the model on postwar data for redistribution and government partisanship for advanced democracies since the Second World War.

The Argument In a 2006 article in the American Political Science Review (Iversen and Soskice 2006), we spelled out a formal model of coalition formation and redistribution. Here we provide an informal version of this model and discuss two implications that were not part of the original model: the role of Christian democracy and the implications of the argument for educational spending and distribution of earnings. Assume that there are three equally sized income classes in an economy, L, M, and H (going from low to high income). Under PR, there are three parties, L, M, and H, each representing one of the groups and sharing its goals (parties are distinguished from classes by italics). Because there is no reason for parties in a PR system to deviate from the preferences of their constituents, we refer to them as representative parties. No party has an absolute majority, so a government can be formed only through a coalition of two parties. Assume, for simplicity, that M is the formateur and has to choose a coalition partner. The implications of our argument also apply when the formateur is randomly chosen, but it is easier to explain the logic when that distinction always falls on the middle party.1 Since there is evidence that center parties are more likely to be chosen as the formateur, controlling for party size, it is also not an unrealistic assumption (Warwick 1996).2 The key intuition of our PR coalition argument is that a party is less capable of looking after its interest if it is excluded from the coalition. Since M benefits more from taxing an unprotected H than from taxing an unprotected L, M will choose L as its coalition partner. This can be modeled in a number of different ways; the only bargaining structure that is excluded is a take-it-or-leave-it offer from M.3 The basic point is that it pays for L and M to form a coalition and take resources (in the form of tax revenues) from the excluded H party rather than for H and M to form a coalition to take resources from an excluded L. If we assume that the division of the pie between the coalition partners conforms to Rubinstein bargaining theory, both coalition members will end up with approximately the same shares under the reasonable assumptions that both discount the future at the same rate and that there is no significant first-mover advantage. All that matters for our

Electoral Institutions, Parties, and the Politics of Class

97

conclusion is that the pie is larger when the excluded party has more resources. The key implication for coalition behavior also follows if resources (tax revenues) are coming from both the middle- and upper-income groups (which is the assumption in the original formal model), as long as redistribution is subject to a “nonregressivity” constraint. This constraint, which also plays a role in the majoritarian case, is that the poor cannot be left with nothing if the middle and upper classes get something, and that the middle class also has to get something if the upper class does. If it holds, the size of the pie that is being split between the coalition partners will be smaller if M allies with H than if M allies with L, since L must get something whenever higher-income classes do. The nonregressivity assumption can be justified on a number of grounds. Empirically, it is an accurate description of reality since available evidence shows that redistribution is always at least mildly progressive, and it is also common to assume that democratic governments are constrained by a basic notion of fairness (see, for example, Roemer 2004).4 In our own view, the constraint can also be understood as reflecting the wider institutions of advanced democracies. These include a free press, free trade unions and other forms of association, the ability to demonstrate collectively, and so on. These wider institutions underwrite the ability of both lower and middle classes to take collective action if right wing domination of the legislature or the executive leads to attempts to exploit less-privileged groups. But again, the key intuition for the PR case is that there is a cost of being excluded from the government, and that this cost is rising in income. As long as this is the case, PR systems will privilege center-left coalitions and such coalitions will redistribute more than center-right coalitions. Majoritarian systems operate quite differently. If we assume that Duverger’s law holds, the three parties are now replaced by two, a centerleft (LM) party and a center-right (MH) party, both competing for the vote of M. If both parties could credibly commit to an M platform, then each would win 50 percent of the time. But modern approaches to political parties avoid this restrictive assumption and instead allow the possibility that either party will deviate from its platform after the election (see, for example, Persson and Tabellini 2003).5 Specifically, if M-voters believe that there is some possibility that an LM government will be tempted to move left and an MH government to move right, then a fundamental center-right bias in majoritarian systems arises. This is because, ceteris paribus, M has less to fear from an MH government moving right than from an LM government moving left, assuming that fiscal policies are nonregressive. In the former case, the benefits going to M will be reduced, but there will be an offsetting reduction in taxes as H maximizes

98

Democracy, Inequality, and Representation

its own net income (while avoiding regressive taxation) by setting taxes and benefits to zero. In the latter case, M will be taxed harder by L, even as benefits are reduced, since this maximizes transfers to L. Hence, if the probability of deviating from the median voter platform is the same in both parties, M has a higher expected payoff from an MH government than from an LM government. Because parties understand the importance of convincing voters that they are truly committed to middle class interests—especially left parties, since they cannot win without convincing voters they are more committed to the middle class—they are prone to elect strong leaders who are willing to ignore the pressures from the party base and are capable of doing so. We therefore refer to majoritarian parties as leadership parties. Yet, without the possibility of writing legally binding contracts between parties and M, platform commitment can never be complete, and there will always be a risk of deviation from the platform. Of course, center-left parties may sometimes succeed in electing leaders who are believed to be more credibly committed to an M platform than center-right parties (think of Tony Blair), but over longer periods of time we expect a center-right bias on average.6 Note that the insights of this model are completely lost in one-dimensional models such as Meltzer-Richard’s, or indeed in power resource theory, where the contestation occurs along a single dimension. The reason is that these models artificially impose a symmetry on the distributive game where the interests of M are always equally well aligned with the interests of L and M. With three parties in a PR system, this means that M is equally likely to ally with H as it is to ally with L. Likewise, in a majoritarian system, any deviation from an M platform is equally threatening to M whether it comes from the center-left or center-right party (for example, the center-left party is forced to share with M even if L sets policies). There is one important qualification to our argument. The center-left bias of PR systems is less pronounced in countries with large Christian Democratic parties. Among the latter, the proportion of center-left governments, measured as in table 4.1, is reduced to 57 percent, whereas it is 63 percent for the sample as a whole. This also implies that for PR countries without strong CD parties, notably Scandinavia, the center-left advantage is more pronounced: 71 percent. The reason for this difference, we believe, has to do with the cross-class nature of CD parties. Because these parties include constituencies from L and M as well as H, differences in distributive preferences between these groups must be negotiated out within the party. This means that the parties can still be said to be representative of their constituents, but also produces a more center-oriented platform than we would usually associate with a centerright party. This in turn makes CD parties more attractive coalition partners for “pure” center, or middle class, parties. The logic that leads center

Electoral Institutions, Parties, and the Politics of Class

99

parties to ally with the left is therefore broken, and in countries where CD and center parties have at times held a majority of seats (such as Germany and Italy), the influence of the left has been reduced. Where such CD-center majority coalitions have not been feasible, as has often been the case in Belgium and the Netherlands, we observe frequent coalitions between CD and left parties that produce a unique blend of policies where transfers are high and redistributive but some of these nevertheless are directed to those with high incomes. Where Christian democracy has been politically dominant, it stands to reason that the focus in social policy has been on insurance than on redistribution, since the former is more palatable across classes than redistribution. Finally, our coalition argument is not restricted to redistribution through transfers but can also help explain variance in the pre-fisc distribution. Standard microeconomic theory says that the relative wages of two individuals will be equal to the ratio of their marginal productivities (absent any influences that might result from market imperfections). Since the ratio of marginal productivities is closely related to the human capital ratio, the distribution of educational attainments plays a large part in determining the underlying distribution of earnings from employment. Our argument suggests that electoral systems also affect the distribution of educational spending, and hence relative marginal productivities and earnings. In particular, center-left governments have an incentive to spend more on L’s education than do center-right or middleof-the-road governments. And they have a lesser incentive to spend on H’s education. Indeed, if H opts for private education, and if there are positive externalities for M from educational expenditure on L (for example, economies of scale in school buildings), then M has an increased incentive to opt for an LM coalition.7 The argument is bolstered by evidence in Charles Boix (1998), Ben Ansell (2005), and Marius Busemeyer (2007). All of these studies find that left governments spend notably more on primary and secondary education than right governments do, and in Torben Iversen and John D. Stephens (2008) this spending is linked to better literacy test scores at the low end of the distribution (using OECD adult literacy surveys). This study also finds that governments dominated by Christian Democratic parties spend less on public education, which again translates into lower international literacy test scores. If PR affects partisanship, therefore, there is a prima facie case that the electoral system is an important determinant of the compactness of the skill distribution. In turn, this means that PR produces both a more equal pre-fisc distribution of income and more redistribution—an association that runs counter to the second core implication of the Meltzer-Richard model that redistribution is tied to pre-fisc inequality (Meltzer and Richard 1981).

100

Democracy, Inequality, and Representation

Table 4.2

The Payoffs for the Middle Party (or the Middle Class) from Different Party or Coalition Choices, Depending on the Electoral System Choose LM Party or Coalition

PR system

Half the proceeds from taxing H or half the proceeds from taxing, both M and H

Majoritarian system

Pr(LM party represents M)*targeted spending on M – Pr(LM party represents L)*taxes on M

Choose MH Party or Coalition Half the proceeds from taxing L or half the proceeds from taxing both M and H minus the share going to L Pr(MH party represents M)*targeted spending on M

Source: Authors’ compilation. Note: Optimal choices are shaded.

Table 4.2 summarizes the key ideas in the theoretical argument. It shows M’s (or M’s) expected payoffs from supporting different parties, or coalitions, contingent on the electoral rule. Since we know that M’s party choice determines the outcome of majoritarian elections, and since M’s choice of coalition partner in a PR system determines whether LM governments will be more frequent, the key results of the model are captured by mapping M’s (or M’s) choice of coalition partner (or party). Note that in a PR system the incentive of M to pick L as a coalition partner holds whether we assume that revenues come only from taxing the excluded party or from taxing both the middle and upper classes (M and H). In the latter case, the result follows from the fact that L can never be entirely shut out from sharing in redistributive spending, even when L is not in the coalition. This implies that M has to share with both L and H in an MH coalition, whereas M only has to share with L in an LM coalition. M therefore has a common interest with L in soaking the rich. In a majoritarian system, by contrast, the main concern of M is to avoid being soaked by the poor. Although both parties present the same median voter platform—targeted benefits to M (and some sharing with L to satisfy the nonregressivity constraint)—if there is a probability that LM will become dominated by L, LM has every reason to soak both the rich and the middle class (this is the subtraction from M’s expected payoff), whereas an H-dominated MH party is constrained by nonregressivity to leave L and M no worse off. That means cutting taxes—and hence its own losses—to zero (and thus making no subtraction from M’s payoff). With an equal chance of the parties deviating from their electoral plat-

Electoral Institutions, Parties, and the Politics of Class

101

forms, M is therefore predisposed to vote for MH. The same nonregressivity assumption that leads the middle class to support center-left governments under PR rules thus causes it to support the center-right under majoritarian rules (the shaded cells). The model therefore implies that the electoral system is associated with both different tendencies in government partisanship and different levels of redistribution. The next section tests these propositions.

The Evidence The purpose of this chapter is to show that electoral systems (E) explain the partisan composition of government (P), and in turn that P explains redistribution (R). The basic forms of the structural estimating equations are therefore: P = ƒ(E) with ƒ’ > 0

(4.1)

R = g(P) with g’ > 0

(4.2)

Equation 4.1 follows if the probability of deviating from a median voter platform under majoritarian rules is the same for the center-left and center-right, and assuming nonregressivity. Equation 4.2 clearly holds when redistribution refers to transfers to the poor. (We report results for the poverty rate later in the section.) However, we also want to define redistribution more broadly as the percentage change in the Gini coefficients from before taxes and transfers to after taxes and transfers. With this definition, it is easy to see that equation 4.2 holds when the electoral system is PR, since we have already established that L and M are both better off under an LM than under an MH government. It is also the case that L is always better off under an LM government in a majoritarian system as long as there is some probability that L will set policies in the LM party but not in the MH party. On the other hand, if L sets policies in LM, M will be worse off under a center-left government (which is why M is inclined to vote for MH). Whether the Gini falls when going from an LM to an MH government therefore depends on whether the gain to L exceeds the loss to M if the two governments are run by L and H, respectively. (There are obviously no differences if they are both run by M.) This is clearly the case, since L gains the tax revenues from both M and H while M loses only the tax that M has to pay. Therefore, regardless of whether we focus on redistribution to the poor or overall redistribution in terms of the percentage reduction in the Gini coefficient, center-left government always redistributes more than center-right governments do. Note that equation 4.2 does not necessarily imply that an LM (or MH)

102

Democracy, Inequality, and Representation

government redistributes the same under different electoral systems. We cannot say anything about that in general, since it depends on the probability of deviating from the median voter platform. Yet, since M will avoid voting for LM unless the probability of deviation is thought to be very low, we should expect LM in a majoritarian system to be more centrist than LM in a PR system. That would imply an effect of electoral system on redistribution that is independent of the effect of partisanship.

The Data We base our analysis of redistribution on the Luxembourg Income Study, which has been compiling a large database on pre- and post-tax-andtransfer income inequality during the past three decades. The LIS data used for this study cover fourteen countries from the late 1960s (the first observation is 1967) to the late 1990s (the last observation is 1997). All fourteen countries have been democracies since the Second World War. There are a total of sixty-one observations, with the number of observations for each country ranging from two to seven. About one-fifth of the observations are from the 1970s and late 1960s, about 40 percent from the 1980s, and the remainder from the 1990s. The data are based on separate national surveys, but considerable effort has gone into harmonizing (or “Lissifying”) them to ensure comparability across countries and time. The LIS data are widely considered to be of high quality and the best available for the purposes of studying distribution and redistribution (see Brady 2003; OECD 1995). As noted earlier, we use the data specifically to explore the determinants of redistribution as measured by the percentage reduction in the Gini coefficient from before to after taxes and transfers. The Gini coefficient, which is perhaps the best summary measure of inequality, varies from zero (when there is a perfectly even distribution of income) to one (when all income goes to the top decile). Using an adjusted version of the LIS data—constructed by Evelyne Huber, John Stephens, and their associates (Bradley et al. 2003)8—we include only working-age families, primarily because generous public pension systems (especially in Scandinavia) discourage private savings and therefore exaggerate the degree of redistribution among older people. Furthermore, because data are only available at the household level, income is adjusted for household size using a standard square root divisor (see OECD 1995). On the independent side, the key variables for explaining redistribution are government partisanship and electoral system. The first is an index of the partisan left-right “center of gravity” of the cabinet based on the average of three expert classifications of government parties’ placement on a left-right scale, weighted by their decimal share of cabinet portfolios. The index goes from left to right and is standardized to vary

Electoral Institutions, Parties, and the Politics of Class

103

between zero and one. The measure was conceived by Donald Gross and Lee Sigelman (1984) and has been applied to OECD countries by Thomas Cusack in a recent comprehensive data set on parties and partisanship (see Cusack and Fuchs 2002; for details, see Cusack and Engelhardt 2002). The expert codings are from Francis Castles and Peter Mair (1984), Michael Laver and Ben Hunt (1992), and John Huber and Ronald Inglehart (1995). One issue raised by this measure is how we can be sure that partisan effects are due to differences in “who governs” as opposed to differences in voter preferences. Our argument is that the electoral system affects the party composition of governments and hence government policies—not that electorates in different countries want different governments and policies (although that might also, of course, be the case). One way of making sure is to compare the ideological center of gravity of the government to the ideological position of the median voter. Since the position of each party represented in the legislature is known, we can use the position of the party with the median legislator as a proxy for the median voter preference. Hence, we also test our model using this relative center of gravity measure. In cases with single-party majority governments (such as the current British Labour government) where the government party controls the median legislator by definition, we use the mean position of the legislative parties weighted by the parties’ seat shares (so that the Labour government would be recorded as being left of center).9 Turning to measurement of electoral system, the theoretical distinction between majoritarian two-party systems and proportional multiparty systems is roughly matched by differences in actual electoral systems (see table 4.3). With the partial exception of Austria (because of the historically strong position of the two main parties), all PR systems tend to have multiple parties and coalition governments, whereas the non-PR systems have few parties and frequent single-party majority governments (although Australia and Ireland have experienced several instances of coalition governments).10 This is indicated in the third column of table 4.3 using Markku Laasko and Rein Taagepera’s (1979) measure of the effective number of parties in parliament.11 France is somewhat of an outlier among the majoritarian cases, but the second round of voting in the French runoff system usually involves candidates from only two parties. The division of countries into two electoral systems is bolstered by the quantitative proportionality measure in the last column. This is a composite index based on Arend Lijphart’s measure of the effective threshold of representation and Michael Gallagher’s measure of the disproportionality between votes and seats (data are from Lijphart 1994). Note that the index is consistent with the division into a majoritarian and a proportional group: there are no cases that should be “switched” based on their

104

Democracy, Inequality, and Representation

Table 4.3

Key Indicators of Party and Electoral Systems Effective Number of Legislative Parties

Proportionality of Electoral System

majoritaritya SMP run offb STVc SNTVd SMP SMP SMP

2.5 2.2 3.8 2.8 2.7 2.0 2.1 1.9 2.5

0.19 0.13 0.16 0.70 0.61 0.00 0.16 0.39 0.30

PR PR PR PR PR PR PR PR PR

2.4 5.2 4.4 5.1 2.6 4.0 4.6 3.3 3.3 3.9

0.89 0.86 0.96 0.87 0.91 0.91 1.00 0.76 0.90 0.90

Electoral System Majoritarian Australia Canada France Ireland Japan New Zealand United Kingdom United States Average Proportional Austria Belgium Denmark Finland Germany Italy Netherlands Norway Sweden Average

Source: electoral system: Lijphart (1994); effective number of legislative parties: Laakso and Taagepera (1979); proportionality of electoral system: Lijphart (1994). a The use of the single transferrable vote (STV) in single-member constituencies makes the Australian electoral system a majority rather than plurality system. b The two-round runoff system has been in place for most of the postwar period, with short interruptions of PR (1945 until early 1950s and 1986 to 1988). c The Irish STV system is unique. While sometimes classified as a PR system, the low constituency size (five or less) and the strong centripetal incentives for parties in the system makes it similar to a median voter-dominated single member plurality (SMP) system. d The single nontransferrable voting (SNTV) system in Japan (until 1994) deviates from SMP in that more than one candidate is elected from each district, but small district size and nontransferrability makes it clearly distinct from PR list systems.

value on the index. All our results go through if we use this index instead of the PR-majoritarian dichotomy. We also controlled for variables that are commonly assumed to affect redistribution or partisanship. Here we provide these variables, with definitions, sources, and a short discussion of causal logic. Country means and a correlation matrix are provided in table 4A.2.

Electoral Institutions, Parties, and the Politics of Class

105

Pre-tax-and-transfer inequality: This variable is included to capture the Meltzer-Richard logic that more inequality leads to more redistribution. It is measured as the earnings of a worker in the ninetieth percentile of the earnings distribution as a share of the earnings of the worker with a median income. We are using earnings data, despite their limitations, because the Meltzer-Richard model applies to individuals, not households. The data are from OECD’s wage dispersion data set (unpublished electronic data). Constitutional veto points: This is a composite measure of federalism, presidentialism, bicameralism, and the frequency of referenda, based on Evelyne Huber, Charles Ragin, and John Stephens (1993). The more independent decision nodes there are, the more veto points. The left in countries with many veto points may have found it harder to overcome opposition to redistributive spending. Unionization: According to power resource theory, high union density should lead to more political pressure for redistribution and a stronger left, while simultaneously reducing primary income inequality. The data are from Jelle Visser (1989, 1996). Voter turnout: Lijphart (1997) argues that there is much evidence to the effect that voter non-turnout is concentrated among the poor. Higher turnout may therefore be associated with less redistribution. The turnout data are from annual records in Thomas Mackie and Richard Rose (1991) and International Institute for Democracy and Electoral Assistance (1997). Unemployment: Since the unemployed receive no wage income, they are typically poor in the absence of transfers. All countries have public unemployment insurance, so higher unemployment is “automatically” linked to more redistribution. We use standardized rates from OECD (various years). Real per capita income: This is a standard control to capture “Wagner’s Law,” which says that demand for social insurance is income-elastic. The data are expressed in constant 1985 dollars (Heston, Summers, and Aten 2002). Female labor force participation: Women’s participation in the labor market is likely to affect redistributive spending because it entitles some women to benefits (unemployment insurance, health insurance, and so on) for which they would otherwise be ineligible. Since women tend to be lower-paid, their labor force participation may also increase support for the left and for redistributive policies. The measure is female labor force participation as a percentage of the working-age population and is taken from OECD (various years).

106

Democracy, Inequality, and Representation

The Statistical Model Our starting point for estimating the first structural equation (equation 4.1) is a simple error correction model. In this model, current redistribution, Ri,t, is equal to past redistribution plus a contribution from redistributive partisan policies, Pi,t (and potentially other factors), that deviate from policies that would preserve the status quo level of redistribution: Ri,t = λ · [α + β · Pi,t – Ri,t–1] + Ri,t–1 + ui,t

(4.3)

where λ is speed with which redistribution changes in response to changes in policy, and u is identically and independently distributed with mean 0 and variance su2. With our data on redistribution, however, we cannot estimate this model directly, since the observations on the dependent variable for each country are unequally spaced, varying between two and as many as ten years. To deal with this missing data problem, we develop a modified version of the model where we substitute the above expression for Ri,t–1, Ri,t–2, and so on, until we get to another observation of the lagged dependent variable. This procedure yields the following expression: N

N

Ri ,t = λ ⋅ α ⋅ ∑ (1 − λ ) + λ ⋅ β ⋅ ∑ (1 − λ ) s ⋅ Pi ,t − s + (1 − λ ) N +1 s

s =0

s =0

(4.4)

N

⋅Ri ,t − N +1 + ∑ (1 − λ ) ⋅ ui ,t − s s

s =0

or N

N

s =0

s =0

Ri ,t − (1 − λ ) N +1 ⋅ Ri ,t − N +1 = λ ⋅ α ⋅ ∑ (1 − λ ) s + λ ⋅ β ⋅ ∑ (1 − λ ) s N

(4.5)

⋅Pi ,t − s + ∑ (1 − λ ) ⋅ ui ,t − s s

s =0

The second term in the last expression is a measure of the cumulative effect of partisanship over a period of N years, where N is the gap between the current and previous observation (s is the lag in years). Of

Electoral Institutions, Parties, and the Politics of Class

107

course, insofar as other variables affect redistribution, we need to calculate the cumulative effects of these in precisely the same manner as for partisanship. Since we have annual observations for partisanship and all control variables, the estimated model is based on complete time series except for the dependent variable. The model is estimated by choosing a value for λ that maximizes the explained variance. Given our assumptions, the composite errors are serially uncorrelated,12 but because the error term depends on N, there is heteroscedasticity. The reported standard errors adjust for such heteroscedasticity, but not for contemporaneous correlation of errors because the latter tends to be inaccurate when there are few observations over time (Wallerstein and Moene 2003). In practice, however, the results are very similar when also adjusting for contemporaneous correlation (known as panelcorrected standard errors; see Beck and Katz 1995), and we therefore do not report them here. The model used to explain partisanship in the second part of the analysis (equation 4.2) is a straightforward OLS regression that is explained later.

The Findings We present the findings in two parts. In the first part, we use partisanship and electoral system as explanatory variables to account for differences in the level of redistribution (equation 4.1). In the second part, we use partisanship as the dependent variable, testing the proposition that the electoral system shapes coalition behavior and therefore the composition of governments (equation 4.2).

Redistribution We begin our presentation with the results from estimating a simple baseline model with economic variables only (model 1 in table 4.4). As expected, female labor force participation and unemployment are associated with more redistribution. Contrary to Wagner’s Law, higher per capita income slightly reduces redistribution, although the result is not statistically significant across model specifications. As in other studies, we also find that inequality of pre-tax-and-transfer earnings has a negative effect on redistribution, contrary to the Meltzer-Richard model expectation. This negative effect is statistically significant at a .01 level, and the substantive impact is strong: a one-standard-deviation increase in inequality is associated with a 0.3-standarddeviation reduction in redistribution. Yet the effect of inequality reverses (though the positive effect is not

108

Democracy, Inequality, and Representation

Table 4.4

Regression Results for Reduction in Inequality (Standard Errors in Parentheses)

Inequality Political-institutional variables Government partisanship(right)

(1)

(2)

–16.75*** (5.68)

13.17 (9.36)



Government partisanship relative to median legislator Voter turnout



Unionization



Number of veto points



Electoral system (PR)



Controls Per capita income Female labor force participation Unemployment λ R-squared Observations

–2.38*** (0.73) —

(3) 12.48 (8.96) —

0.01 (0.10) 0.16* (0.09) –1.57** (0.62) 5.00** (2.15)

–2.93*** (0.75) –0.06 (0.10) 0.15* (0.09) –1.79*** (0.59) 4.44** (2.06)

–0.001*** (0.00) 0.73*** (0.11) 0.81*** (0.27)

–0.001 (0.00) 0.36* (0.20) 0.99*** (0.27)

–0.001 (0.000) 0.45** (0.20) 1.08*** (0.26)

.4 0.648 47

.7 0.746 47

.7 0.765 47



Source: Luxembourg Income Survey (LIS). Note: Standard errors are in parentheses. All independent variables are measures of the cumulative effect of these variables between observations on the dependent variable. *p < .10; ** p < .05; *** p < .01 (two-tailed tests)

significant) when we include controls for the political-institutional variables (models 2 and 3). One likely reason for this change was suggested earlier: left governments, as well as strong unions and PR, not only cause an increase in redistribution but also reduce inequality through public spending on primary and secondary education. Iversen and Stephens (2008), show that there is a strong positive effect of left partisanship on spending on all types of public education, and demonstrate that this has

Electoral Institutions, Parties, and the Politics of Class

109

the effect of compressing the skill distribution at the low end if we measure skills by performance on standardized adult literacy tests (for additional evidence, see Ansell 2005; Busemeyer 2007). Given this, excluding partisanship produces an omitted variable bias on the coefficient for inequality. The most important result in table 4.4 is that right partisanship has a strong and statistically significant negative effect on redistribution, regardless of whether we use the absolute (column 2) or the relative (column 3) measure of partisanship. A one-standard-deviation shift to the right reduces redistribution by about one-third of a standard deviation. This confirms previous research, especially David Bradley et al. (2003), and it adds the finding that partisanship matters even when measured relative to the ideological center of the legislature. This is important to our story because it implies that political parties, and the coalitions they form, matter for redistribution—not just differences in the preferences of electorates. The results also suggest that multiple veto points, as expected, reduce redistribution and that PR has a direct (positive) effect on redistribution. The latter effect holds regardless of which measure of electoral system in table 4.3 we use. Our model does suggest one possible reason for this, because if the probability of left deviation from a median voter platform is not too high, center-left governments always redistribute more to the poor under PR than under majoritarian rule. To test this we ran the same model using the percentage reduction in the poverty rate instead of the reduction in the Gini coefficient as the dependent variable. Consistent with this proposition, it turns out that whereas the effect of partisanship is about the same, the direct effect of PR is now notably stronger.13 There may also be effects of electoral systems that we have not modeled. Torsten Persson and Guido Tabellini (2003), for example, argued that single-member plurality systems give politicians an incentive to target spending on geographically concentrated constituencies, whereas PR, with ideally only one electoral district, encourages politicians to spend more on universalistic benefit programs. Since universalistic programs are likely to be more redistributive than geographically targeted programs, this would mean that PR has a direct effect on redistribution. But our focus is on the effect of electoral systems on partisanship, to which we now turn.

Partisanship Although both government partisanship and the electoral system are important in explaining redistribution, partisanship itself is

110

Democracy, Inequality, and Representation

shaped by the distinct coalitional politics associated with different electoral systems. A key implication of our argument is that center-left governments tend to dominate over long periods of time under PR, whereas center-right governments tend to dominate under majoritarian institutions. Although the electoral system has a direct effect on redistribution, partisanship, we argue, is one of the key mechanisms through which it exerts an effect on redistribution. We use the partisan center of gravity (CoG) index as a dependent variable and indicators for party and electoral systems as independent variables. We have data for eighteen countries that have been democracies since the Second World War, beginning with the first democratic election after the war and ending in 1998. One country—Switzerland—has a collective executive that prevents coalition politics from having any influence on the composition of the government. We therefore exclude this case from the analysis, although every result reported in this section goes through with Switzerland included.14 Table 4.1 is a simple cross-tabulation of electoral systems and government partisanship using annual observations as the unit of analysis. Governments are coded as being left-of-center if their position on the composite left-right index is to the left of the overall mean. This is somewhat arbitrary since the mean may not correspond to a centrist position. An alternative would be to define the center as the middle of the scale. But in two of the three expert surveys, the middle of the scale is not explicitly defined as centrist in terms of an absolute standard, and experts may well equate it instead with the observed center of a party system, whether or not this center is shifted to the left or right. In practice, this choice has little effect on the results. Identifying a centrist position, however, is important for a different reason. If an LM leadership party in a majoritarian system is centrist, then the model implies that it stands a good chance of winning. Observing such a party in government is therefore consistent with the model. At the same time, it cannot be counted as confirmatory evidence, since we do not have any independent measure with which to determine whether the party platform is credible. The relative frequency of center and center-right governments therefore cannot be hypothesized a priori. Moreover, because our theory implies that the political space in majoritarian systems is tilted to the right (owing to strategic voting in a setting of incomplete platform commitment), if we include governments that are centrist in an absolute sense, these would be counted as center-left in terms of their relative position. Using a scale such as the composite CoG index, the results would therefore be biased against the theory, since the center on this scale is almost certainly affected by relative assessments. Our solution is to use one of the component measures in the

Electoral Institutions, Parties, and the Politics of Class

111

CoG index by Castles and Mair (1984) to exclude governments that are centrist in the absolute sense. The Castles-Mair measure is the only one that explicitly defines the middle value (3) as a party having a centrist left-right ideology.15 As pointed out at the beginning of this chapter, in a simple cross-tabulation of electoral system and government partisanship there is only one country, Germany, that does not conform to the predicted pattern. In this case, there were thirty-four years with center-right governments and only sixteen years with center-left governments. As we suggested earlier, a possible explanation is the role of the German Christian Democrats (CDU/CSU). This party is usually seen as a coalition of groups from different locations in the income distribution, where group differences are worked out through intra-party bargaining (as we would expect in a representative party). The Christian Democrats can therefore credibly claim to be closer to the center than a typical conservative party representing mainly high-income voters. This helps explain why the small, pivotal liberal party (FDP) chose to ally with CDU/CSU instead of the Social Democratic Party (SPD) during most of the postwar period. But note that even in this special case government policies are heavily influenced by PR, since the right has, in effect, gained access to government power only by accepting a compromise with lower-income groups that involves at least some redistribution. Germany aside, we can question the results in table 4.1 for the same reason that was pointed out when we used the government CoG measure to explain redistribution: those results could reflect differences in voter preferences rather than in coalitional party politics. Note, however, that strategic voting in majoritarian systems is expected to shift the legislative center to the right, and the distribution of seats in PR systems should not matter so long as coalitions can be formed that are either to the left or to the right of the center. So evidence on absolute differences in partisanship is clearly relevant to our theory. Still, using the relative measure of partisanship allows us to exclude explanations that emphasize the distribution of voter preferences, and it serves as a useful robustness test. In table 4.5, governments are therefore coded as center-left (center-right) only if they are to the left (right) of the legislative median (or the legislative mean in cases with single-party majority governments). This does not change the results very much, although they are (not surprisingly) slightly weaker. About two-thirds of governments under PR are now to the left of the legislative median, whereas two-thirds of governments under majoritarian institutions are to the right. As before, all of the countries conform to this pattern except one.16 What alternative explanations might there be for the pattern observed

112

Democracy, Inequality, and Representation

Table 4.5

Electoral Systems and the Number of Years with Governments Further to the Left or to the Right than the Median Legislator, 1945 to 1998 Government Partisanship Left

Electoral system Proportional Majoritarian

291 (9) 116 (1)

Right 171 (0) 226 (7)

Proportion of Right Governments 0.37 0.66

Source: electoral system, Lijphart (1994); government partisanship: Cusack and Fuchs (2002), Cusack and Engelhardt (2002). Note: Excludes governments that are classified as “centrist” on the Castles-Mair scale (Castles and Mair 1984).

in table 4.5? Because we use the difference between the position of the government and the median legislator, we have limited such alternatives to variables that affect the post-election partisan composition of governments. We thus implicitly “control” for all variables that may affect the distribution of preferences in the electorate. Although there are obviously a plethora of situationally specific factors that shape each instance of government formation, it is not easy to think of variables that would systematically bias the composition of governments in one ideological direction or the other. To our knowledge, there are only two candidates for such variables in the existing literature. The first goes back to Stein Rokkan’s (1970) wellknown explanation for the choice of electoral systems (see also Alesina and Glaeser 2004; Boix 1999). Rokkan argued that at the time of the extension of the franchise, when a united right faced a rising but divided left, the governing right chose to retain majoritarian institutions. Conversely, when a divided right faced a rising and united left, the response was to opt for PR. If this pattern of fractionalization persisted into the postwar period, the right would tend to have an advantage in majoritarian systems, while the left would tend to have an advantage under PR (in the latter case because the transaction costs of bargaining presumably rise with the number of parties). This would produce the pattern that our model predicts, but for different reasons. A simple test of this argument is to see whether there is a relationship between party fragmentation and electoral system in the expected direction. For this purpose, we use a variable in the Cusack-Engelhardt data

Electoral Institutions, Parties, and the Politics of Class

113

set, which is the difference between party fractionalization on the left and right, where fractionalization is defined in the usual way as one minus the sum of the squared seat shares held by parties to the left or to the right of the center (Rae 1968). There is in fact no significant correlation (r = −0.15), which could mean either that Rokkan was wrong or that the relationship between fractionalization and electoral systems has changed over time. Either way, fractionalization should not affect the relationship between electoral system and partisanship in the period on which we focus here. To confirm this, we ran a simple multiple regression, using partisanship as the dependent variable and electoral system and fractionalization as independent variables (see table 4.6).17 Note that the coefficient for electoral system variable is very similar whether fractionalization is included or not (compare columns 1 and 2). In substantive terms, going from a majoritarian to a PR system shifts the center of gravity of the government by a factor that is roughly equivalent to moving from an average Christian democratic government to a social democratic government, or from a conservative government to a Christian democratic government. Not surprisingly, greater fractionalization on the left than on the right does lead to more right-leaning governments on average. But this is not relevant to our story. The second argument is that vote-seat disproportionalities may favor the right under majoritarian institutions. The explanation would be that the boundaries of electoral districts in majoritarian democracies were drawn up before the full impact of the industrial revolution, which led to an underrepresentation of urban areas, where the left had the strongest support (Cox and Katz 2002; Monroe and Rose 2002; Rodden 2005). Although subsequent redistricting may have addressed some of these inequities, they could still play a role in explaining why the left is disadvantaged in majoritarian systems (PR being more unbiased by design). We tested this possibility using a variable that is simply the difference between the legislative seat share of right parties and these parties’ share of the vote. It is referred to as “right overrepresentation” in table 4.6. In contrast to left-right fragmentation, this variable does not register any significant effect, and the sign is in fact in the wrong direction (see column 3). This is somewhat puzzling, since there is a positive bivariate correlation between this variable and government partisanship (r = .37), as well as between this variable and electoral system (r = .51). The explanation is probably very simple, however. Remember that we are modeling government partisanship, not the governing party’s margin of victory. The latter does not matter in majoritarian “winner-take-all” systems.





0.37 17

Adusted R-squared Observations

Source: Authors’ calculations. Note: Standard errors are in parentheses. ** p < .05; *** p < .01 (two-tailed tests)



Female labor force participation 0.54 17









0.664*** (0.033) −0.147*** (0.047) 0.241** (0.094) —

(2) Government CoG Minus Legislative Median

0.653*** (0.039) –0.173*** (0.054) —

(1) Government CoG Minus Legislative Median

0.49 17





−0.036 (0.101) —

0.663*** (0.051) –0.184*** (0.063) —

(3) Government CoG Minus Legislative Median

Regression Results for Government Partisanship, 1950 to 1996

Electoral system (PR) Fragmentation (left minus right) Right overrepresentation Electoral participation Unionization

Constant

Table 4.6

0.55 17





0.501*** (0.046) –0.174*** (0.063) 0.201 (0.116) 0.077 (0.104) —

(4) Government CoG

0.49 17

0.001 (0.005) –0.004 (0.003) 0.004 (0.004)



0.375 (0.453) 0.176** (0.077) —

(5) Government CoG

Electoral Institutions, Parties, and the Politics of Class

115

Hence, the only scenario in which the vote-seat disproportionality would affect government partisanship is when the right loses the electoral vote but wins a majority of seats. As illustrated by the British case, such instances are rare. In only one postwar election (1951) did the Conservative Party win more seats than Labor despite losing the popular vote, and in another (1974) Labor in fact came out on top despite getting fewer votes.18 The last two columns use the absolute government CoG measure as the dependent variable, which maximizes the cross-national variance in partisanship. But when entered simultaneously, neither fragmentation nor overrepresentation registers a significant effect. The electoral system, on the other hand, continues to have roughly the same impact as before. Finally, the last column tests three variables that may reasonably be expected to affect the distribution of voter preferences, and hence the political center of a country. Predictably, high unionization rates are associated with more left-leaning governments, but the effect is weak and statistically insignificant. Electoral participation and female labor force participation (both of which might be expected to benefit the left) are also insignificant, and the signs are in the wrong direction.19 The electoral system remains the sole variable with a strong and statistically significant effect.

Conclusion Distribution and redistribution vary to a surprising extent across democracies that are roughly at similar levels of development. In this chapter, we have argued that much of this variance can be explained as the result of differences in electoral systems and the class coalitions they engender. To explain redistributive policies under democracy, it is essential to understand that policies are multidimensional and that groups have to form partisan coalitions to govern. Both features of redistributive politics are assumed away in standard political economy models, such as those that follow the setup in Meltzer and Richard (1981). If we permit allocation profiles across classes that do not have to conform to a linear tax, the incentives of the middle class to ally with the left and right will vary depending on the institutional conditions. There are two opposing incentives for the middle class: it has an incentive to ally with the poor to exploit the rich, but it also has an incentive to support the rich to avoid being exploited by the poor. In a majoritarian two-party system, the latter motive dominates because the middle class cannot be sure that the poor will not set policies in a governing center-left party. In a PR multiparty system, on the other hand, the first motive dominates because the

116

Democracy, Inequality, and Representation

middle class party can make sure that a coalition with the left party will not deviate from pursuing their common interest in taxing and redistributing from the rich. Center-right governments therefore tend to dominate in majoritarian systems, whereas center-left governments tend to dominate in PR systems. Insofar as governments also affect the pre-fisc distribution of income through educational spending, the argument also helps us understand why relatively egalitarian countries tend to be associated with high redistribution. But it also leaves a number of puzzles for future research. First, proportional representation appears to be less closely related to redistribution in Latin America and perhaps other new democracies. Is this because strong presidents, who are essentially majoritarian in nature, often coexist with PR electoral institutions? Is it because parties are less organized around class? Or is it because the poor have little collective action capacity in these countries while the rich can threaten coups? Future work will have to pay more attention to the nonelectoral aspects of the political system as well as to the organization of interests. On the latter point, it seems plausible that better measures of differences in the collective action capacity of the poor will also explain some of the residual variance in redistribution for rich democracies. Another puzzle is why some countries moved to PR electoral systems in the 1920s or earlier. In all these countries, at least one, and usually several, center-right parties supported the move. Even if this was initially a political mistake, there have been many opportunities for the center and right to move the system back to majoritarian elections since then, but they have not. Elsewhere we have suggested that the explanation may be related to the fact that the shift occurred in countries where employers and unions had made massive investments in training systems that essentially created cospecific assets, which in turn had to be protected politically and economically (Cusack, Iversen, and Soskice 2007). Political protection implied the representation of asset-owners in national decision-making bodies with authority to regulate these assets, either directly (say, through certification and standard setting) or indirectly (say, through the administration of unemployment benefits). Economic protection implied insurance against the risk of losing the value of the investment—a matter that was critical to skilled workers who could lose their jobs and have difficulty finding others if their skills were specific. PR systems and the corporatist institutions that tend to accompany them offered both representation (for example, through strong legislative committees with close ties to the bureaucracy) and a credible commitment to insurance against income loss as a result of a center-left bias of governing

Electoral Institutions, Parties, and the Politics of Class

117

coalitions. In our view, the relationship between the organization of production, political representation, and social insurance is a major area for future research. Finally, the argument may be expanded to explain changes in partisan advantage over time. Although we have abstracted from differences in the dispersion of the earnings distribution, it may be conjectured that as pre-fisc income inequality grows, middle class fears of being soaked by the poor grow in majoritarian systems, while their incentive to join the poor in soaking the rich intensifies under PR. Thus, contrary to Meltzer-Richard, rising inequality in majoritarian systems may be associated with a greater advantage for the right. This is in fact what Nolan McCarty, Keith Poole, and Howard Rosenthal (2006) found in a new study of polarization and partisanship in the United States. Whether the opposite is true in PR countries is an interesting question for future research.

118

Democracy, Inequality, and Representation

Appendix Table 4A.1

Country Means for the Variables Used in Regression Analysis Redistribution (Reduction in Inequality Partisanship Voter Gini) (Wages) (Right) Turnout Unionization

Australia Austria Belgium Canada Denmark Finland France Germany Ireland Italy Japan Netherlands New Zealand Norway Sweden United Kingdom United States

23.97 — 35.56 21.26 37.89 35.17 25.36 18.70 — 12.13 — 30.59 — 27.52 37.89 22.67 17.60

1.70 — 1.64 1.82 1.58 1.68 1.94 1.70 — 1.63 — 1.64 — 1.50 1.58 1.78 2.07

0.47 0.30 0.36 0.36 0.35 0.30 0.40 0.39 0.42 0.37 0.78 0.31 0.43 0.15 0.17 0.52 0.40

84 87 88 68 84 79 66 81 75 93 71 85 85 80 84 76 56

46 54 48 30 67 53 18 34 48 34 31 33 23 54 67 42 23

Electoral Institutions, Parties, and the Politics of Class

Veto Points 3 1 1 2 0 1 1 4 0 1 1 1 0 0 0 0 5

Electoral Right Per System Left OverCapita (PR) Fragmentation Representation Income 0 1 1 0 1 1 0 1 0 1 0 1 0 1 1 0 0

–0.39 –0.18 –0.34 0.18 –0.40 –0.18 0.10 –0.13 –0.33 0.20 0.22 0.18 –0.40 –0.02 –0.40 0.08 0.00

0.10 0.04 0.27 –0.11 0.07 0.09 0.09 0.15 0.70 0.08 0.28 –0.36 0.98 –0.32 –0.03 0.07 –0.17

10,909 8,311 8,949 11,670 9,982 8,661 9,485 9,729 5,807 7,777 7,918 9,269 — 9,863 9,982 9,282 13,651

119

Female Labor Force Participation Unemployment 46 51 43 48 63 66 51 51 37 38 56 35 47 52 63 54 53

4.63 2.76 7.89 6.91 6.83 4.48 4.57 4.86 9.09 8.12 1.77 4.62 — 2.28 6.83 5.01 5.74

Source: Luxembourg Income Study (LIS). Note: Time coverage is 1950 to 1996, except for redistribution and inequality, which are restricted to the available LIS observations.

Correlation Matrix

1 0.37 –0.38 –0.22 –0.01 –0.54 –0.09 0.66 –0.42 –0.45 0.55

0.80 –0.49

(2)

1 –0.38 –0.50 0.11 0.75 –0.44 0.34 –0.57 –0.13 0.12

(1)

(3)

–0.28 0.52

1 –0.24 –0.49 0.33 –0.66 0.14 0.46 –0.08

Source: Luxembourg Income Study (LIS). Note: Correlations are based on the period averages in table 4A.1.

(1) Redistribution (2) Inequality (3) Partisanship (4) Turnout (5) Unionization (6) Veto points (7) Electoral system (8) Left fragmentation (9) Right overrepresentation (10) Per capita income (11) Female labor force participation (12) Unemployment

Table 4A.2

–0.19 0.06

1 0.51 –0.43 0.71 –0.27 0.10 –0.51

(4)

0.48 –0.20

1 –0.56 0.49 –0.76 0.14 –0.18

(5)

–0.06 0.01

1 –0.27 0.14 –0.16 0.61

(6)

0.17 –0.20

1 –0.18 –0.24 –0.22

(7)

–0.37 0.02

1 –0.48 0.08

(8)

–0.168 0.63

1 –0.64

(9)

0.38 –0.41

1

(10)

1 –0.51

(11)

Electoral Institutions, Parties, and the Politics of Class

121

We thank the editors of this volume, the participants in the 2005 conference on democracy, inequality, and representation at the Maxwell School of Syracuse University, as well as Jim Alt, Klaus Armingeon, Neal Beck, David Brady, Geoffrey Brennan, Thomas Cusack, Jeff Frieden, Robert Goodin, Peter Hall, Peter Katzenstein, Robert Keohane, Herbert Kitschelt, Peter Lange, Gerard Roland, Fritz Scharpf, Ken Shepsle, and Michael Wallerstein for their many helpful comments on a previous version of this chapter.

Notes 1.

2.

3.

4.

5.

6.

7.

When the formateur is randomly chosen, the composition of the government does not necessarily reflect the preferences of M, but the incidence of such government is more frequent. The case where size determines the probability of being chosen as the formateur corresponds to the case where parties are chosen randomly. This does not alter the substantive results, as we proved in Iversen and David Soskice (2006). If M can make a take-it-or-leave-it offer, it can enforce M’s ideal point on either L or H. But this is not the reality of most coalition formations, in which counteroffers are invariably both made and considered. Branko Milanovic (2000) and Lars Osberg, Timothy Smeeding, and Jonathan Schwabish (2003) showed in detailed analyses of LIS data on redistribution that the poor always gain from democratic redistribution, the rich always lose, and the middle class does less well than the poor but better than the rich. Needless to say, there are very significant reputational costs of deviating from an M platform. Leaders as well as the rank and file understand this, and this is precisely why successful parties concentrate power in a moderate leader. But again, the ability of the leader to control the party is contingent on post-election circumstances, and the mere possibility of a deviation has important implications for our story. Indeed, this explains the focus of general elections in majoritarian systems on the perceived strength and moderation of the opposing leaders. Since the LM party is at an electoral disadvantage, it has a greater need and incentive to elect centrist leaders than the MH party. If this holds, the distribution of wins and losses will be more even, but the political spectrum will be shifted to the right. The contrast between the centrist Clinton and the rightist George W. Bush is a case in point. Though note too that this weakens the center-right bias in majoritarian systems, since a left deviation is less frightening for M.

122 8. 9.

10.

11.

Democracy, Inequality, and Representation We are grateful to these authors for letting us use their data. We did the same in a small number of cases where the government position was equivalent to the median legislator but the government was not a single-party majority government. Ireland is perhaps the most ambiguous case, but it is not part of the redistribution regression, and the results for partisanship are not sensitive to the particular electoral system measure we use or to whether Ireland is included or excluded. The effective number of parties is defined as one divided by the sum of the square root of the shares of seats held by different parties (or one divided by the Herfindahl index). N ⎡N ⎤ s s E ⎢∑ [(1 − λ ) u i , t − s ] ⋅ ∑ [1 − λ ] u i , t − ( N +1) − s ]⎥ = 0 , since the errors in the first = = s 1 1 s ⎣ ⎦ square bracket run from ui,t to ui,t−N1 and in the second from ui,t−(N1+1) to ui,t−(N1+1)−N2. The effect of going from a majoritarian system to a PR system is to increase redistribution to the poor by 0.7 standard deviations, whereas the effect on the Gini coefficient is 0.5 standard deviations. Because right parties cannot be excluded from government power in Switzerland, we should expect redistribution to be lower than in other PR countries. This is in fact the case, since the average pre- to post-tax-andtransfer reduction in the Gini is 9 percent in Switzerland, whereas it is 28 percent in other PR countries. We also excluded centrist governments from the PR cases because they neither confirm nor disconfirm the theory (although bias is less of a concern here). In total, 95 of 734 country-years were coded as centrist on the Castles-Mair scale. The “outlier” is no longer Germany since most governments in that country have in fact been to the left of the median, even as they have tended to be to the right compared to other PR systems. Instead, the deviant case is France, where slightly more than half (twenty-nine of fifty-two) of the observations are to the left of the legislative median. This is because the party with the median legislator tends to be very right wing, whereas French governments sometimes include representation from more moderate parties. The rightist orientation of French politics is also clearly evident in the fact that every president in the postwar period, except for François Mitterrand, has been from a right-of-center party. Since there is little meaningful variance in electoral systems over time, we simply ran a cross-section regression on the averages from 1950 to 1996 (for which we have complete data on several control variables). It is of course possible, indeed standard, to pool the country time series while correcting for serial correlation by adding a lagged dependent variable (PCTS). Our results hold up in such a specification—indeed, the levels of signifi1

12.

2

1

13.

14.

15.

16.

17.

Electoral Institutions, Parties, and the Politics of Class

18.

19.

123

cance improve notably—but as demonstrated in Goodrich (2004), it is misleading to use PCTS regressions when nearly all the evidence is cross-sectional. The right overrepresentation variable, defined as the difference between right seats and vote shares, overstates the right advantage in the case of Britain. The reason is that the Liberal Party is located between Labor and the Conservatives and always gets fewer seats than votes. As a result, both Labor and the Conservatives tend to get more seats than implied by their votes. However, we resisted the temptation to “finesse” the measure to reflect this and other unique national circumstances. The same is true for other potential variables that we tested, such as unemployment, the size of the industrial workforce, and income per capita.

References Alesina, Alberto, and Edward Glaeser. 2004. Fighting Poverty in the United States and Europe: A World of Difference. Oxford: Oxford University Press. Ansell, Ben. 2005. “From the Ballot to the Blackboard? Partisan and Institutional Effects on Human Capital Policy in the OECD.” Unpublished paper. Department of Government, Harvard University. Austen-Smith, David. 2000. “Redistributing Income Under Proportional Representation.” Journal of Political Economy 108(6): 1235–69. Beck, Nathaniel, and Jonathan Katz. 1995. “What to Do (and Not to Do) with Time-Series Cross-Section Data.” American Political Science Review 89(September): 634–48. Benabou, Roland. 1996. “Inequality and Growth.” In National Bureau of Economic Research Macro Annual, edited by Ben S. Bernanke and Julio J. Rotemberg. Cambridge, Mass.: MIT Press. Boix, Carles. 1998. Political Parties, Growth, and Equality. New York: Cambridge University Press. ———. 1999. “Setting the Rules of the Game: The Choice of Electoral Systems in Advanced Democracies.” American Political Science Review 93(3): 609–24. Bradley, David, Evelyne Huber, Stephanie Möller, François Nielsen, and John D. Stephens. 2003. “Distribution and Redistribution in Post-industrial Democracies.” World Politics 55(2): 193–228. Brady, David. 2003. “Rethinking the Sociological Measure of Poverty.” Social Forces 81(3): 715–52. Busemeyer, Marius. 2007. “Social Democrats and Education Spending.” Working paper 07/2. Leipzig: Max Planck Institute for the Study of Societies. Castles, Francis, and Peter Mair. 1984. “Left-Right Political Scales: Some Expert Judgments.” European Journal of Political Research 12(1): 73–88. Cox, Gary W., and Jonathan N. Katz. 2002. Elbridge Gerry’s Salamander: The Elec-

124

Democracy, Inequality, and Representation

toral Consequences of the Reappointment Revolution. Cambridge: Cambridge University Press. Cusack, Thomas R., and Lutz Engelhardt. 2002. “The PGL File Collection: File Structures and Procedures.” Wissenschaftszentrum Berlin für Sozialforschung. Cusack, Thomas R., and Susanne Fuchs. 2002. “Documentation Notes for Parties, Governments, and Legislatures Data Set.” Wissenschaftszentrum Berlin für Sozialforschung. Cusack, Thomas, Torben Iversen, and David Soskice. 2007. “Economic Interests and the Origins of Electoral Institutions.” American Political Science Review 101(3): 373–91. Goodrich, Benjamin. 2004. “Problems with and Solutions for Two-Dimensional Models of Continuous Dependent Variables.” Unpublished paper. Department of Government, Harvard University. Gross, Donald A., and Lee Sigelman. 1984. “Comparing Party Systems: A Multidimensional Approach.” Comparative Politics 16(4): 463–79. Heston, Alan, Robert Summers, and Bettina Aten. 2002. Penn World Table Version 6.1. Center for International Comparisons at the University of Pennsylvania (CICUP). Hicks, Alexander, and Duane Swank. 1992. “Politics, Institutions, and Welfare Spending in Industrialized Democracies, 1960–1982.” American Political Science Review 86(3): 649–74. Huber, Evelyne, Charles Ragin, and John Stephens. 1993. “Social Democracy, Christian Democracy, Constitutional Structure, and the Welfare State.” American Journal of Sociology 99(3): 711–49. Huber, Evelyne, and John D. Stephens. 2001. Development and Crisis of the Welfare State: Parties and Policies in Global Markets. Chicago, Ill.: University of Chicago Press. Huber, John D., and Ronald Inglehart. 1995. “Expert Interpretations of Party Space and Party Locations in Forty-two Societies.” Party Politics 1(1): 73–111. International Institute for Democracy and Electoral Assistance (IDEA). 1997. Voter Turnout from 1945 to 1997: A Global Report on Political Participation. Stockholm: IDEA Information Services. Iversen, Torben, and David Soskice. 2006. “Electoral Institutions and the Politics of Coalitions: Why Some Countries Redistribute More Than Others.” American Political Science Review 100(2): 165–81. Iversen, Torben, and John D. Stephens. 2008. “Partisan Politics, the Welfare State, and Three Worlds of Human Capital Formation.” Comparative Political Studies 41(4–5): 600–637. Korpi, Walter. 1983. The Democratic Class Struggle. London: Routledge & Kegan Paul. ———. 1989. “Power, Politics, and State Autonomy in the Development of Social Citizenship: Social Rights During Sickness in Eighteen OECD Countries Since 1930.” American Sociological Review 54(3): 309–28. Kwon, Hyeok Yong, and Jonas Pontusson. 2003. “The Zone of Partisanship: Par-

Electoral Institutions, Parties, and the Politics of Class

125

ties, Unions, and Welfare Spending in OECD Countries, 1962–1999.” Unpublished paper. Cornell University, Department of Political Science. Laasko, Markku, and Rein Taagepera. 1979. “Effective Number of Parties: A Measure with Applications to Western Europe.” Comparative Political Studies 12(3): 3–27. Laver, Michael, and W. Ben Hunt. 1992. Policy and Party Competition. New York: Routledge. Lijphart, Arend. 1994. Electoral Systems and Party Systems: A Study of Twenty-seven Democracies, 1945–1990. New York: Oxford University Press. ———. 1997. “Unequal Participation: Democracy’s Unresolved Dilemma.” American Political Science Review 91(1): 1–14. Mackie, Thomas T., and Richard Rose. 1991. The International Almanac of Electoral History, 3rd ed. London: Macmillan. McCarty, Nolan, Keith Poole, and Howard Rosenthal. 2006. Polarized America: The Dance of Ideology and Unequal Riches. Cambridge, Mass.: MIT Press. Meltzer, Allan H., and Scott F. Richard. 1981. “A Rational Theory of the Size of Government.” Journal of Political Economy 89(5): 914–27. Milanovic, Branko. 2000. “The Median-Voter Hypothesis, Income Inequality, and Income Redistribution: An Empirical Test with the Required Data.” European Journal of Political Economy 16(3): 367–410. Moene, Karl Ove, and Michael Wallerstein. 2001. “Inequality, Social Insurance, and Redistribution.” American Political Science Review 95(4): 859–74. Monroe, Burt L., and Amanda G. Rose. 2002. “Electoral Systems and Unimagined Consequences: Partisan Effects of Districted Proportional Representation.” American Journal of Political Science 46(1): 67–89. Organization for Economic Cooperation and Development (OECD). Various years. Labor Force Statistics. Paris: OECD. ———. 1995. “Income Distribution in OECD Countries: Evidence from the Luxembourg Income Study.” Social Policy Studies 18. Osberg, Lars, Timothy Smeeding, and Jonathan Schwabish. 2003. “Income Distribution and Public Social Expenditure: Theories, Effects, and Evidence.” Unpublished paper. Maxwell School, Syracuse University. Perotti, Roberto. 1996. “Growth, Income Distribution, and Democracy: What the Data Say.” Journal of Economic Growth 1(2): 149–87. Persson, Torsten, and Guido Tabellini. 1999. “The Size and Scope of Government: Comparative Politics with Rational Politicians.” European Economic Review 43(4–6): 699–735. ———. 2000. Political Economics: Explaining Economic Policy. Cambridge, Mass.: MIT Press. ———. 2003. The Economic Effects of Constitutions. Cambridge, Mass.: MIT Press. Powell, G. Bingham. 2002. “PR, the Median Voter, and Economic Policy: An Exploration.” Paper presented to the meetings of the American Political Science Association. Boston, Mass., August 29–September 1, 2002.

126

Democracy, Inequality, and Representation

Rae, Douglas. 1968. “A Note on Fractionalization of Some European Party Systems.” Comparative Political Studies 1(3): 413–18. Rodden, Jonathan. 2005. “Red States, Blue States, and the Welfare State: Political Geography, Representation, and Government Policy Around the World.” Unpublished paper. Department of Political Science, Massachusetts Institute of Technology. Roemer, John. 2004. “Does Democracy Engender Equality?” In Political Knowledge and the Public Interest, edited by Edward J. Mansfield and Richard Sisson. Columbus, Oh.: Ohio State University Press. Rogowski, Ronald, and Mark Andreas Kayser. 2002. “Majoritarian Electoral Systems and Consumer Power: Price-Level Evidence from the OECD Countries.” American Journal of Political Science 46(3): 526–39. Rokkan, Stein. 1970. Citizens, Elections, Parties: Approaches to the Comparative Study of the Processes of Development. Oslo: Universitetsforlaget. Ross, Michael. 2006. “Is Democracy Good Enough for the Poor?” American Journal of Political Science 50(4): 860–74. Visser, Jelle. 1989. European Trade Unions in Figures. Deventer, Netherlands: Kluwer Law and Taxation. ———. 1996. “Unionization Trends Revisited.” Unpublished paper. University of Amsterdam. Warwick, P. V. 1996. “Coalition Government Membership in West European Parliamentary Democracies.” British Journal of Political Science 26(4): 471–99. Wallerstein, Michael, and Karl Ove Moene. 2003. “Earnings Inequality and Welfare Spending: A Disaggregated Analysis.” World Politics 55(4): 485–516.

Chapter 5

Economic Institutions, Partisanship, and Inequality PABLO BERAMENDI AND THOMAS R. CUSACK

There is a “transatlantic consensus” on the recent developments in economic inequality (Atkinson 1999). This is the widely shared view that the waxing wage and income inequality seen in the principal AngloSaxon countries during the last decades is also reflected in similar rises within most other developed economies. Wages and salaries have grown ever more disparate as the skill premium ineluctably increases (Nickell and Bell 1996; see also Gottschalk and Smeeding 1997). In turn, as capital reaps ever greater rewards, those who depend on their own labor are losing out in both absolute and relative terms (Phillips 2002). And, finally, with the retreat of government—a widespread trend over the last two decades—the dampening effect of the welfare state has been weakened in terms of its ability to ameliorate the inequalities generated by labor, financial, and other markets (Korpi and Palme 2003). In the context of these processes, it is widely believed that within the industrialized countries inequality in income continues to increase. We join other scholars in questioning this “consensus” view; our principal criticism, which we substantiate here, is that it exaggerates the uniformity in trends toward inegalitarian societies (Bradley et al. 2003; Gottschalk and Smeeding 2000; Iversen and Soskice 2003; Kenworthy and Pontusson 2005). This lack of convergence ultimately stems from institutional and partisan differences across systems. Both the market and the state shape the distribution of income. However, the roles they play differ across the variety of income dimensions. The important question here is not so much whether politics makes a difference for inequality, but how political and institutional factors work their effects on inequality

128

Democracy, Inequality, and Representation

among these different income forms. In particular, what roles do they play in shaping the relatively large cross-national variation in wage and disposable income inequality and the narrower range in overall market income inequality? Although scholars in recent years have begun to pay greater attention to the question of income inequality, relatively little effort has yet been made to work systematically through the process by which inequalities in terms of different incomes are linked together (Atkinson and Brandolini 2003). Our analysis also explores how they are linked. We proceed in a series of steps. First, we examine the variation, both over time and across countries, in measures of the distribution of income in its different forms. Second, we lay out an argument about how these different income distributions are shaped by both economic and noneconomic forces. Particular attention is given to the role of government partisanship and economic institutions. The third step empirically evaluates the argument. Finally, we draw together our findings and their implications. A number of results emerge from this effort. First, the patterns of cross-national variation are quite dissimilar across different forms of income. OECD countries have been and continue to be much more diverse in the distributions of labor earnings and disposable income than they are in the distributions of market income. There is little support for the consensus view of convergence to greater inequality. Second, the larger cross-national variation in the distributions of labor earnings and disposable income can be attributed to the role of political actors (such as unions and, more importantly, political parties) and the economic institutions that allow actors to coordinate their activities. Third, the transmission of cross-national differences in wage inequality into marketbased inequality is muted. Fourth, the way in which political parties are able to pursue their goals varies across forms of income. While political parties are able to work their effect on the distribution of disposable income directly through their choices about fiscal redistribution, their capacity to shape the distribution of labor earnings is contingent on the degree of coordination provided by the institutional framework of the economy.

Patterns of Inequality Within OECD Countries We focus on the distributions of wages, market income, and disposable income. Wages are the monetary reward received in exchange for the labor an individual provides an employer. Market income, of which wages are a component, is the broadest measure of the income the individual

Economic Institutions, Partisanship, and Inequality

129

derives from the economic system exclusive of government transfers. Disposable income reflects the direct effects, after taxes and transfers, of government on the ultimate distribution of market income. There are useful data on the distribution of wages for thirteen countries that are aggregated into five-year averages. These are displayed using the 90/10 ratio (in other words, the ratio of the earnings of the top 10 percent of wage earners to those of the lowest 10 percent of wage earners) (OECD 1996).1 These can be seen in table 5.1. The pattern in over-time wage inequality is mixed across the OECD countries over the period from the late 1970s to the late 1990s (Gottschalk and Smeeding 1997). Some countries experienced greater inequality in wage dispersion, and some witnessed declines. In the United States and the United Kingdom, labor markets that were marked by already high levels of wage inequality saw that inequality surge upward through the 1980s and 1990s. Other countries, such as France, Finland, and Denmark, experienced very little change in levels of wage inequality over the time periods for which we have data. Moreover, in other countries, such as Germany and Belgium, low levels of wage inequality shrank even further. Wages are an important component of the income that individuals and households derive from the market. Still, they are only a part of total market-derived income.2 Figure 5.1 presents three-decade averages of dependent labor income as a share of total household market-based income and the income derived from other market sources (Mendoza, Razin, and Tesar 1994). Wage earnings generally constitute less than 70 percent of household income. Correspondingly, the average of 30 percent of this income derived from sources other than dependent employment makes up a significant part of market income. On the face of it, such flows are distributed differently than wages. The implications are clear: the overall market-based distribution is different from the wage distribution, and the forces shaping it are, at least in part, dissimilar. We can get a hint of the cross-national pattern in a more limited but detailed way by examining figure 5.2. We might anticipate that capital income flows are likely to be distributed in far different ways than are wage and salary earnings. Indeed, that is true. The average Gini coefficients (based on annual data for the early 1980s through 2001) for these two types of income in both Germany and the United States illustrate this point very well. Using annual data from the German Socio-Economic Panel (SOEP) and the American Panel Study of Income Dynamics (PSID), we can see that the labor income is distributed quite differently.3 At the same time, the distribution of capital income (highly unequal) is practically the same across two very different economic models: the famous “shareholder society” of the United States

Wage Inequality Across the OECD Countries (90/10 Ratio from OECD, 1996)

2.83 2.84 2.85 2.88 2.95

2.42 2.33 2.25

4.02 4.45 4.33 4.28

2.15 2.19 2.17

2.47 2.48 2.46 2.33 2.42

3.23 3.14 3.25 3.11 3.05 2.91 2.73 2.79

2.26 2.34 2.37

2.52 2.48 2.60 2.72

2.07 2.11 1.98

2.03 2.05 2.09 2.19 2.22

3.03 3.20 3.39 3.42 3.42

3.80 4.14 4.35 4.56 4.58

Source: Authors’ compilation. Note: Each period is five years in duration and follows the LIS wave dating convention: 1 = 1978 to 1982; 2 = 1983 to 1987; 3 = 1988 to 1992; 4 = 1993 to 1997; 5 = 1998 to 2002.

1 2 3 4 5

United United Wave Australia Belgium Canada Denmark Finland France Germany Italy Netherlands Norway Sweden Kingdom States

Table 5.1

Economic Institutions, Partisanship, and Inequality Figure 5.1

131

Sources of Household Sector Market Income Across OECD Countries, 1965 to 1995 Average

100

Share of Household Income

90 80 70 60 50 40 30 20 10 K

O D K SW E

N

U

ER SW T U SA

G

L FI CN

A

JP

U FR

TH

A

N

BE

IR

IT

0 Country Dependent Employment

Other Market Sources

Source: Authors’ calculations using formulas developed by Mendoza et al. (1994, n. 10). Note: IT = Italy; IR = Ireland; BE = Belgium; NTH = Netherlands; AU = Austria; FR = France; JP = Japan; AL = Australia; FI = Finland; CN = Canada; GER = Germany; SWT = Switzerland; USA = United States; UK = United Kingdom; NO = Norway; DK = Denmark; SWE = Sweden.

(Gini of .88) and the equally famous “stakeholder society” of Germany (Gini of .86).4 This surprising similarity carries over to the inequality measures on the total of market income, where these national sources show an average Gini of about .46 for Germany and .49 for the United States. In contrast to the mixed picture on cross-national developments in wage earnings distributions, and consistent with the pattern we have seen in the cases of Germany and the United States, the pattern in the distribution of market income is uniform across the eleven OECD countries for which data are available. Table 5.2 presents Gini indices of market household income per equivalent adult. This measure is obtained by weighting overall household income on the basis of household size de-

132

Democracy, Inequality, and Representation

Figure 5.2

Inequality in Different Measures of German and American Household Income: Labor, Capital, and Market

1.0 Germany United States

Gini Index

0.8

0.6

0.4

0.2

0.0 Labor

Capital

Market

Type of Income Source: Authors’ calculations using data from the German Socio-Economic Panel (SOEP) and the U.S. Panel Study of Income Dynamics (PSID).

fined by the LIS equivalence scale that takes into account the distributive importance of differences in terms of family structure. With the LIS equivalence scale, we give a weight of 0.5 to the first adult member of the household and a weight of 0.25 to each of the remaining members. Regardless of the wage-leveling forces in national economies and the redistributive character of national taxation and spending regimes, market income inequality has been high across these countries and has surged to even higher levels over the last two decades. Each of the Gini indices presented in this table represents the share of total market income generated within the economy that would have to be redistributed in order to achieve equality across all households in terms of the amount each household receives.5 We see, for example, that in the United States nearly half of all income would need to be redistributed to achieve equality in market outcomes. Even in an egalitarian society such as Sweden, the level of inequality in market outcomes is extremely high and has sometimes exceeded the levels of inequality seen in the United States and the United Kingdom.6

0.37 0.40 0.41 0.41

0.36 0.37 0.39 0.39 0.41

0.39 0.42 0.43

0.33 0.34 0.38 0.37

Finland 0.34 0.37 0.39 0.47

France 0.31 0.40 0.41 0.40 0.44

Germany 0.36 0.38 0.39

Netherlands 0.35 0.33 0.37 0.41 0.42

Norway

0.39 0.43 0.46 0.45 0.44

Sweden

0.37 0.42 0.44 0.45 0.46

United Kingdom

0.39 0.42 0.42 0.45 0.46

United States

Source: Authors’ calculations. Note: Income adjusted for household size using LIS equivalence scale. Each period is five years in duration and follows the LIS wave dating convention: 1 = 1978 to 1982; 2 = 1983 to 1987; 3 = 1988 to 1992; 4 = 1993 to 1997; 5 = 1998 to 2002.

1 2 3 4 5

Denmark

Market Income Inequality Across the OECD Countries (Gini Index Based on Luxembourg Income Study [LIS] Data)

Wave Australia Canada

Table 5.2

134

Democracy, Inequality, and Representation

In addition, we have data on the distribution of disposable income for thirteen OECD countries. These are displayed in table 5.3. Again, the LIS equivalence scale is employed and Gini indices of household income per equivalent adult are used to describe these distributions. Across the OECD countries, the Gini measures on disposable income are far lower than those for market income. Direct government intervention produces a far more equitable distribution of income. The scope of this intervention varies, and with that the breadth of the reduction in inequality. In the last period reported (1998 to 2000), the effective level of redistribution varied dramatically between states such as Sweden (18 percent of total income) and the United States (8 percent of total income). In terms of changes in the overall levels of inequality in disposable income, we observe that in most of the countries for which we have data the pattern over the last two decades has been one that involved a modest increase in the overall level of inequality or basically no change (as in France and the Netherlands). Three countries stand out in terms of increases in the degree of inequality: Sweden, the United Kingdom, and the United States. The direct workings of fiscal systems revealed by combining the figures in tables 5.2 and 5.3 are large even if they differ dramatically across nations. In states with modest welfare regimes, the net amount of total income being redistributed amounts anywhere from 8 to 12 percent of total income. In Sweden, nearly one-quarter of total income is redistributed. The redistribution amounts to huge sums in a relative sense, with on average around 20 percent in the United States and nearly 50 percent in Sweden.7 And even though in a number of countries, such as the United States and Sweden, the redistributive effects of state fiscal systems declined over time, they rose sharply in other countries, such as France and Germany. In sum, there are significant differences in the incidence of inequality across OECD countries, but the range of these differences varies across different income concepts. These differences have not changed dramatically over time. Table 5.4 displays the coefficients of variation in the level of inequality for the three forms of income during the first and last periods of the sample. These figures convey two main points. First, the OECD countries have been much more diverse in their distributions of labor earnings and disposable income than they were in their distributions of market income. Second, these patterns remained unaltered through the end of the century. If anything, cross-national differences in terms of disposable income inequality increased slightly over the last twenty-five years. In the next section, we lay out an argument to identify the factors at work behind these developments.

Disposable Income Inequality Across the OECD Countries (Gini Index Based on Luxembourg Income Study [LIS] Data)

0.28 0.29 0.30 0.31

0.23 0.23 0.25

0.28 0.28 0.28 0.29 0.31 0.25 0.24 0.26

0.21 0.21 0.23 0.25

0.29 0.30 0.29 0.29

0.24 0.26 0.25 0.26 0.26 0.31 0.29 0.34 0.35

0.26 0.27 0.26

0.22 0.23 0.23 0.24 0.26

0.20 0.22 0.23 0.22 0.26

0.27 0.30 0.34 0.35 0.35

0.30 0.34 0.34 0.36 0.38

Source: Authors’ calculations. Note: Income adjusted for household size using LIS equivalence scale. Each period is five years in duration and follows the LIS wave dating convention: 1 = 1978 to 1982; 2 = 1983 to 1987; 3 = 1988 to 1992; 4 = 1993 to 1997; 5 = 1998 to 2002.

1 2 3 4 5

United United Wave Australia Belgium Canada Denmark Finland France Germany Italy Netherlands Norway Sweden Kingdom States

Table 5.3

136

Democracy, Inequality, and Representation

Table 5.4

Cross-National Differences in Coefficients of Variation Across Measures of Inequality

Wave 1 2 3 4 5

Wages

Market

Disposable

0.24 0.25 0.26 0.25 0.24

0.07 0.09 0.08 0.07 0.07

0.15 0.15 0.15 0.16 0.17

Source: Authors’ calculations. Note: Each period is five years in duration and follows the LIS wave dating convention: 1 = 1978 to 1982; 2 = 1983 to 1987; 3 = 1988 to 1992; 4 = 1993 to 1997; 5 = 1998 to 2002.

The Argument: Parties, Institutions, and Inequalities A variety of factors shape the distributions of income within society, and not all of them are economic. Politics also plays a role. This role is not similar across different forms of income. Inherent to the well-documented existence of a structural dependency of the state on capital is the notion that the capacity of governments to shape the distribution of income is an inverse function of the number of exit options available to different income factors (capital, labor) (Przeworski and Wallerstein 1988; Wallerstein and Przeworski 1995; for related empirical analyses, see also Cusack and Beramendi 2006; Ganghof 2005). In line with this argument, we anticipate that government’s impact on prefiscal income should be primarily reflected in the distribution of wages. Government is constrained in its actions in other markets, such as finance. These volatile markets, whose members enjoy more exit options than wage earners, are sensitive to government intervention, and this sensitivity deters governments from attempting to influence directly the shape of the distribution of overall market income. Thus, we expect government’s role in shaping the distribution of market income to be very indirect once its influence on the distribution of earnings is accounted for. In addition, there is no gainsaying that governments play a central and immediate role in shaping the distribution of (final) disposable income. In government’s repertoire of policy instruments are tools that allow it to shape the distribution of labor market and disposable income. These instruments include regulations, taxes, and transfers. Regulations such as minimum-wage laws affect the distribution of earnings (see also Rueda, this volume). Taxes and transfers are sometimes powerful determinants of the distribution of disposable income. These instruments of government policy also affect prefiscal income through the anticipatory behav-

Economic Institutions, Partisanship, and Inequality

137

ioral responses on the parts of labor and capital (Beramendi 2001). Labor responses come in the form of labor supply decisions. Capital responds by adjusting investment decisions and labor demand. In the following, we analyze the role of political factors in shaping the distributions of earnings and disposable income. The role of politics can be understood in terms of institutions and the ideological preferences of governments (Hibbs 1977). Parties are seen as agents of different economic interests. Parties on the left are viewed as representing the interests of labor. Parties on the right are held to be agents of more affluent classes. Left wing governments are expected to tax, spend, and regulate more to achieve an equitable society (Bartels 2003; Bradley et al. 2003; Hibbs and Dennis 1988; Hicks and Swank 1984; Iversen and Soskice 2003). Analogously, parties on the right are expected to implement public policies that preserve inequitable outcomes deriving from the workings of the market. A second tradition in political science has highlighted the importance of labor market institutions in shaping the distribution of wage income (Wallerstein 1999; see also Iversen and Wren 1998). The effects of labor market institutions are both direct and indirect. The direct effects are seen in the constraints imposed on the behavior of labor and capital. Indirect effects are found in the way in which these institutions filter the impact of other determinants of inequality, most prominently government partisanship (Rueda and Pontusson 2000). Through a variety of ways, left wing policy aims at reducing inequality. In the case of wage equality, these paths include higher minimum wages, higher levels of benefit generosity, and higher labor tax rates. Higher minimum wages raise the wage floor directly. Increases in generosity raise the wage floor indirectly by increasing the reservation wage. Both compress the earnings distribution from the bottom. Higher tax rates on labor income reduce the incentives for wage increases in the upper half of the distribution, compressing it from the top. All three policies reduce wage inequality. Left wing parties also reduce disposable income inequality using higher levels of taxes and transfers. Non-left wing governments pursue a different type of policy. These governments can be from either a Christian Democratic or more liberal tradition. The former tradition is associated with entitlements based on the insurance principle, the maintenance of status differences, and the subsidiary correction of market outcomes (Esping-Andersen 1990; Huber and Stephens 2001; van Kersbergen 1995). In implementing these principles, Christian Democratic governments combine medium levels of taxes with a heavy reliance on social security contributions and public transfers. Although there is no reason to anticipate that this policy reduces wage inequality, a certain degree of income redistribution is to be

138

Democracy, Inequality, and Representation

expected. The egalitarian impact of this policy is likely to be smaller than that of a left wing government. Right wing liberal policy is anchored in the tenet that the market should be the dominating societal resource allocation mechanism. Taxes, transfers, and regulations are minimized. As a result, the expected redistributive effect of government policy is at its lowest level. Therefore, liberal policies are expected to lead to higher levels of both earnings and income inequality. Political parties are not alone in shaping the distribution of income. Trade unions and employers’ associations also matter. The impact of unions is to be seen in both wage inequality and redistribution. Unions have an aversion to wage inequality (Hibbs 1991). The stronger the union movement, the greater this aversion; to the extent that this greater strength rests on the inclusion of low-wage earners, the aversion is heightened (Freeman 1980). The power resources approach to the welfare state emphasizes the strength of the working class (Esping-Andersen 1990; Korpi 1983; Stephens 1979). The extent to which it is organized— for example, in unions—enhances its ability to influence government policy. With this influence, the working class is able to achieve a greater redistributive effort on the part of government, thereby reducing the level of inequality in disposable income. Accounts of redistribution and inequality based on the power of the working class tend to see employers as passive agents endorsing a unitary opposition to state intervention and redistribution. Yet employers’ preferences about state regulations and redistribution vary, among other things, according to the size, the sector of production, and the skill intensity of the firms (Mares 2003). Moreover, they are far from being mere spectators. The control by employers of the levels of private investment and labor demand gives employers’ associations a great deal of leverage over government policy. More specifically the potential reaction by employers may operate as a veto against particular forms of taxes, transfers, and regulations. This implies that both unions and employers’ associations have an input into the politics of inequality. As a result, an important part of the explanation of government policy and its distributive effects lies in the way in which the interplay between unions, employers, and the incumbent party is institutionalized. This brings us to the issue of coordination within the economy. The degree of wage coordination between capital and labor is conventionally regarded as a crucial aspect of the difference between liberal market economies (LMEs) and coordinated market economies (CMEs) (Hall and Soskice 2001). Let us consider briefly the nature of such differences and their implications for the politics of inequality. In LMEs, firms coordinate their activities through competitive market arrangements. Relations between capital and labor are organized by individuals, not by as-

Economic Institutions, Partisanship, and Inequality

139

sociations. Capitalists value their capacity to adjust to market fluctuations, and so too does labor by investing in portable, general skills. Neither has an incentive to coordinate outside the market. Markets are organized very differently in CMEs. Firms find incentives to coordinate with unions and the government around a fundamental “non-market-based” equilibrium between capital and labor. Such an equilibrium becomes politically effective through the wage coordination compromise between capital, labor, and the government. By virtue of this compromise, labor restrains wage demands, thereby contributing to lower inflation and better economic conditions, but most important, labor gains a degree of income insurance for workers (Goldthorpe 1984; Wallerstein, Golden, and Lange 1997). Government uses fiscal policy to compensate labor for its sacrifice, thereby reducing the costs of the compromise. This occurs through a large public insurance system that guarantees a good income level in periods of economic downturns and longer-term earnings (pensions). Furthermore, unions obtain higher leverage in wage negotiations and greater control over the implementation of some important public policies (Swenson and Pontusson 2000). This compromise between capital and labor is one facet of overall nonmarket-based coordination in CMEs that distinguishes them from LMEs. The institutional arrangements of corporate governance and their interplay with the workings of labor markets are also part of the picture. Firms within CMEs rely more on bank-based financing and display higher levels of cross-shareholding. By virtue of these arrangements, investors privilege long-term performance and firms pool risks. This institutional setup creates the conditions for long-term investments by both firms and employees (with specific skills). The resulting pressure on companies and workers to maintain continuous levels of profitability is reduced. This facilitates the sustainability of the compromise within the labor market (Hall and Gingerich 2001). This way of organizing the economy has distributive consequences. Within the labor market, the institutional position of unions is enhanced. Therefore, unions are better positioned to push for an egalitarian wage distribution, which should be reflected in lower levels of wage inequality (Wallerstein 1999). The distributive consequences of coordination go beyond the labor market.8 By making firms’ decisions less responsive to short-term profits, corporate governance arrangements in CMEs create the conditions for employers to accept a large welfare state. This acceptance facilitates income redistribution and should be reflected in lower levels of disposable income inequality. In addition, coordination matters because of its interaction with partisan politics. According to one view in the literature, high levels of coordi-

140

Democracy, Inequality, and Representation

nation between capital and labor constrain the impact of parties on public policy and therefore mute the impact of partisanship on the distribution of income.9 The intuition behind this view is that collective agreements generally incorporate all workers in a company or sector regardless of union membership status and that wage developments are at least indirectly tied to one another. Given these conditions, it is difficult to entertain the notion that government can significantly influence these autonomous bargaining agreements and thus influence wage distribution (Pontusson, Rueda, and Way 2002). In short, high levels of wagebargaining coordination have the effect of muting partisan effects on wage inequality. Our view on how the interplay between political parties and labor market institutions affects income inequality extends and qualifies this argument, in particular in the context of coordinated economies. High levels of wage-bargaining coordination, one of the most prominent features of a CME (Hall and Gingerich 2001), facilitate the implementation of left wing policy and constrain the implementation of policies favored by the right. In contrast, the absence of coordination between capital and labor facilitates the implementation of right wing preferences and constrains the egalitarian effects of left wing policy. Let us elaborate on why this position seems more plausible than that held by Pontusson and his colleagues. Coordination reduces the resistance of employers to a generous welfare state and constrains the economic costs of redistribution by ensuring the agreement of unions to wage moderation. Moreover, agreements between capital and labor facilitate the adoption of significantly higher taxes on labor (Cusack and Beramendi 2006). In such an institutional context, left wing parties are free to use the tools at their disposal to reduce wage inequality without incurring negative economic externalities. In this sense, left wing incumbency and wage-bargaining coordination reinforce each other’s egalitarian effects on wage distributions. As a result, wage inequality is at its lowest levels in those countries where coordination is high and left wing parties are in power. In contrast, strong coordination creates a hostile environment for the implementation of right wing policy. A number of mechanisms are at work in producing this outcome. First, the exchange of wage moderation for social protection implies a more complex distribution of capital and labor preferences concerning government policy. Although unions will certainly oppose any attempt at market “flexibilization” and welfare state reduction put forward by right wing parties, employers will not necessarily favor these attempts. Second, high levels of coordination imply that both employers and unions enjoy a degree of veto power over government policy. In highly coordinated environments, then, right wing par-

Economic Institutions, Partisanship, and Inequality

141

ties may choose to moderate their political platform in the first place. But even if they do not, right wing parties will ensure that their actual policy is much further away from their conventional positions than it would be in the case of a social democratic government. Thus, if right wing parties hold office in highly coordinated environments, earnings and disposable income inequality are likely to reach intermediate levels. Focusing on the policy choices of the two right wing governments in Swedish postwar history—the bourgeois and conservative coalitions of 1976 to 1982 and of 1991 to 1994—helps to illustrate the logic behind these expectations. In the earlier case, it is widely acknowledged that the fiscal and social policies of the first right wing government in forty years displayed a remarkable degree of continuity with previous cabinets (Huber and Stephens 2001). The lack of innovation in social policy was accompanied by the absence of significant rollbacks in publicly provided services or income transfers. There was not a clear policy alternative to the previous Social Democratic rule. Three reasons have been advanced to explain the absence of change in economic and fiscal policy during these years. First, there was what Mark Blyth (2001) referred to as a “cognitive lockout”: an apparent lack of a coherent policy alternative on the side of bourgeois parties. Second, the ability of LO (the most powerful trade union organization in the country) to call strikes, recently experienced during the 1970s, remained intact. Last but not least was the fact that employers had actually endorsed many of the Social Democratic postwar reforms, including the ones on pensions, sick pay, and active labor market policies (Swenson 2003; see also Hogfeldt 2003). A standard laissez-faire right wing policy was sure to lack extensive social support. In this context, both the policies and the discourse of the bourgeois parties could not part ways with a fundamental endorsement of the Swedish model’s commitment to egalitarian outcomes. After the return of Social Democrats to power in 1982, it took nine years for Swedish conservative parties to regain office. They were in power between 1991 and 1994, under very different economic and social circumstances, and equipped with a much more orthodox ideological and policy kit. The cleavage between export-oriented and domestic-oriented sectors had partially reduced the scope of wage coordination agreements. In this context, the offensive of the employers’ association (SAF) during the 1980s had successfully undermined the scope of coordination agreements, partially limiting the political leverage of the trade union (LO).10 In addition, the economic soundness of the Swedish model was questioned by the magnitude of the economic crisis and the fiscal imbalance. In this context, Social Democrats had to replace their campaign promises of a new wave of social programs for a series of austerity packages in the late 1980s (Huber and Stephens 2001, 243–57). In 1991,

142

Democracy, Inequality, and Representation

Social Democrats were defeated, and a new conservative bourgeois coalition came to power under the leadership of Carl Bildt. This time the right wing cabinet was much more willing to implement liberal policies to promote the interests of business. The new cabinet took two major initiatives. First, it pegged the krone to the ecu in the hope of reducing inflationary tendencies. Second, it passed a tax reform that mainly consisted in lowering the marginal tax rates on both personal and capital income while reducing tax loopholes and broadening the tax base (Steinmo 2002). The broadening of the tax base had a perverse effect in that it increased the incentives for saving as opposed to investing. By imposing deflationary measures in an already stagnant environment, these policies exacerbated the economic crisis. As a result, unemployment had gone up to about 10 percent by 1994, and the deficit reached 13 percent of GDP. In the context of a national employment and fiscal crisis, the Swedish bourgeois coalition had to turn to the Social Democrats and the labor movement to negotiate support for the much-needed austerity packages. In a nutshell, these packages consisted of selective cuts in entitlements and generosity in a number of social policy programs, as well as the partial introduction of complementary private alternatives in the provision of certain services (for example, elderly and health care). These changes did not constitute the paradigm shift in Swedish politics expected by parts of the business sector and the right. They were a coordinated response to the crisis. In fact, they continued to be implemented even after the Social Democrats returned to office in 1994. Overall, these policies should not be seen as the result of right wing parties implementing their agenda in an unconstrained manner. This would have been politically unfeasible. Rather, they reflect the outcome of a cross-class, cross-party compromise to make the system more fiscally sound and durable (Benner 2003; Huber and Stephens 2001). In conclusion, the two experiences of right wing government in Sweden both support and illustrate our contention that in highly coordinated environments the room for maneuver of right wing parties is much narrower than it is for left wing ones. The picture changes in the absence of coordination. Under such conditions, right wing parties receive the full support of employers and unions are left with far less institutional leverage. A right wing party’s capacity to let the market work as freely as possible is not constrained, and therefore we would expect inequality to be higher across all forms of income. The absence of coordination also has implications for the capacity of left wing parties to promote their distributional goals. Although these parties retain their capacity to compress the distribution of disposable income through fiscal redistribution, their leverage to affect wage inequality is much more limited. In coordinated contexts, the government uses fiscal policy as a means to influence the behavior of labor and capital

Economic Institutions, Partisanship, and Inequality

143

within wage agreements. In the absence of coordination, the signals sent by government through its fiscal policy are less effective in shaping union behavior. Unions have no guarantee that the government and employers will agree to the development of a large public insurance system. Hence, unions have no incentive to agree to wage moderation. Additionally, there is no enticement for them to accept the burden of the higher taxes on labor needed to sustain a generous welfare state (Cusack and Beramendi 2006). Rather, unions will demand large levels of redistribution while still pressing for nominal increases to sustain real wages. Under such conditions, governments lack the capacity to trade income insurance for wage moderation and high taxes on labor. As a result, there is no reason to expect that left wing parties will be able to compress significantly the shape of wage distribution. The inability of left wing incumbents to control the market responses of capital and labor is the crucial mechanism behind this expectation. Let us illustrate the way this mechanism operates by discussing in some detail the effects of minimum wages on employment and earnings. Minimum wages are the main tool in the hands of left wing cabinets to compress the wage distribution in low-coordination environments. Indeed, there is evidence of a significant relationship between left wing incumbency and the introduction of higher minimum wages (see Rueda, this volume). Yet the key issue from the point of view of the link between left wing partisanship and inequality is not simply whether leftist cabinets increase the minimum wage, but, more important, whether these increases actually have the effects intended by the governments enacting them. This brings us to the impact of minimum-wage regulations on the behavior of employers and employees. Standard economic theory suggests that an increase in the minimum wage leads perfectly competitive employers to reduce the demand for low-skilled workers. As a result, employment levels decline, and the bottom half of the distribution is compressed (Benjamin, Gunderson, and Riddell 1998). Were this disemployment effect to be the only one at work, left wing parties in low-coordination environments should undoubtedly be associated with less wage inequality. Yet the specialized literature has also identified an efficiency effect that works toward increasing (as opposed to decreasing) employment levels at the lower end of the earnings distribution. The basic intuition behind this second effect is as follows: a higher minimum wage increases the cost of losing the job. In addition, more people are willing to work if the salary is better. In other words, there is a higher supply of labor in the market. These two factors force low-skilled employees to work as efficiently as possible, and the monitoring costs of employers are thereby reduced. The firm’s resources are freed up, as a result of which it can hire more workers around the minimum-

144

Democracy, Inequality, and Representation

wage level.11 Finally, a higher minimum wage could also generate a spillover effect by way of which the earnings of workers in the upper parts of the distribution should be raised by an amount at least similar to the increase in the minimum wage. Otherwise, their incentives to continue to work would be undermined. Should the efficiency and spillover effects be the dominant ones, it is conceivable that an increase in the minimum wage generates higher employment levels for low-skilled workers without compressing at all the distribution of earnings. In fact, depending on the distributional assumptions made, higher minimum wages could even lead to higher levels of earnings inequality. Clearly, the ultimate direction of the sum of these three effects cannot be anticipated theoretically. It is an issue contingent upon many specific aspects of national labor markets. Hence, it is not surprising that the empirical evidence on the economic effects of minimum wages is mixed— both among and within nations. There are cases, such as France, in which left wing governments have apparently been able to compress the (lower half) of the wage distribution through minimum wages. JeanPaul Fitoussie (1994) saw the particular experience of the early years of the Mitterrand government (1981 to 1984) as an example of the preeminence of the disemployment effect. In contrast, the experiences of Canada and the United States offer examples in which higher minimum wages go hand in hand with higher levels of employment and earnings inequality. This would reflect a more complex set of effects of minimum wages on employment and earnings distribution, as documented by the econometric evidence.12 In conclusion, when left wing parties try to reduce wage inequality through the introduction of minimum wages, there is no guarantee that the final outcome will reflect the original goal. In fact, as suggested by several studies of Canada and the United States (see note 13), the employment effects of higher minimum wages may be either neutral or contrarian in that they might widen earnings inequality. The direction and magnitude of these effects is ultimately an empirical question. Thus, there is no reason to anticipate that, in the absence of coordination, left wing parties will be consistently and necessarily associated with more compressed wage distributions.

The Methodological Approach The empirical evaluation of these predictions requires that we assess the impact of political and institutional factors as well as other conditioning factors on the distribution of income. The three different forms of income need to be treated as the objects of explanation. This is accomplished by specifying and estimating the following system of three equations:

Economic Institutions, Partisanship, and Inequality

145

WIit = α1 + β1MEit + β2TWIit + β3FPit + β4HCit + β5UDit + β6LGit + β7ECit + β8LGit * ECit + ε1

(5.1)

MIit = α2 + β9WIit + β10SMCit + β11OPit + ε2

(5.2)

DIit = α3 + β12MIit + β13ECit + β14UDit + β15LGit + ε3

(5.3)

Wage inequality (WI), market income inequality (MI), and disposable income inequality (DI) are the dependent variables in this system. Tables 5.1, 5.3, and 5.4 display the values of the dependent variables used in the analysis.13 The interdependence across the equations is restricted in that the dependent variable of the first equation, WI, is independent of the other two equations’ dependent variables, while the second, MI, depends on a set of exogenous variables plus the dependent variable of the first equation, WI, and the third dependent variable, DI, is a function of the second, MI, and another set of exogenous variables. Such a system of equations has a variety of labels, including recursive, triangular, and hierarchic. Whatever label is used, such a system can be consistently estimated with equation-by-equation ordinary least squares (OLS) (Green 2000). We have employed a number of different estimation techniques. One is the less conservative strategy of using singleequation techniques. Two alternative methods were OLS with robust standard errors and OLS with panel-corrected standard errors (PCSE). Results are practically identical, and so we report only the estimates based on the robust standard errors.14 In the second, more conservative, tack, we employed two-stage least squares (2SLS) to take into account the limited interdependence across the equations. Each observation represents a five-year average. The five-year periods conform to the LIS waves (see table 5.1) and, again, the series extend from the 1978 to 1982 period through 1998 to 2002. Limited data on income inequality restrict the number of observations available. Compounding this restriction is the loss of a number of cases owing to missing observations on the independent variables used. Using a panel design, there are forty-one cases where all the income inequality data plus data on the independent variables are available. The countries included in this restricted sample are Australia, Canada, Denmark, Finland, France, Germany, the Netherlands, Norway, Sweden, the United Kingdom, and the United States. For some countries we have as many as five observations. The sample is smaller for other countries, some of which have as few as two cases. Let us outline the reasoning that stands behind the forms specified. We first address the wage inequality equation (5.1). The first four terms included on the right-hand side of this equation represent a variety of

146

Democracy, Inequality, and Representation

factors that are meant to control for important transformations going on inside OECD labor markets that can be expected to have significant effects on the distribution of wages. First, ME is the number of manufacturing workers expressed as a percentage of the working-age population. This term is meant to represent the effects of deindustrialization and its inegalitarian impact as people lose jobs in the relatively high-paying manufacturing sector and need to take on lower-paying positions in services (Esping-Andersen 1990). Both the sectoral employment and age cohort variables come from various annual issues of the OECD’s Labor Force Statistics. We would anticipate that the sign on this variable’s parameter, β1, will be negative. In effect, as the manufacturing sector employs a greater share of the working-age population, the level of wage inequality declines. Alternatively, as the share of employment in this relatively high-paying sector declines, the level of wage inequality should increase. The second economic variable included is TWI, which stands for imports from the Third World. This variable is expressed as a percentage of GDP. The trade data derive from various annual issues of the International Monetary Fund’s (IMF) Directory of Trade Statistics. The GDP data come from the IMF’s Financial Statistics CD-ROM. TWI’s inclusion is justified by the need to control for the effects of Third World competition on wage levels in the manufacturing sector and their implications for the overall distribution of wages (Wood 1995). Our expectation is that the effect of this variable, captured in the parameter β2, is positive; in other words, higher levels of Third World imports raise the level of wage inequality. The third economic control is the female labor force participation rate, FP. This factor has been introduced in order to control for the inegalitarian consequence of high numbers of women being employed in the labor market. This distributional effect derives mainly from the wage discrimination practiced against women (Blau and Kahn 2000).15 In operational terms, FP is the number of women working expressed as a percentage of the female working-age population. The sources for these data are various annual issues of the OECD’s Labor Force Statistics. Note that the expectation is that the parameter β3 capturing the effect of this variable is positive; higher levels of female labor force participation increase the level of overall wage inequality in the distribution of labor market earnings. The last of the four economic controls is the proportion of collegeeducated in the adult population. This variable aims to capture the distribution of human capital in society, HC (Barro and Lee 2000). In the context of a process of skill-biased technological change, a more unequal distribution of human capital should be reflected in a more unequal distribution of wages. Thus, we anticipate the relationship between HC and earnings inequality to be positive (β4 > 0).

Economic Institutions, Partisanship, and Inequality

147

In addition to this set of economic controls, equation 5.1 includes a group of political and institutional terms reflecting the arguments presented in the last section. A compressing effect on the distribution of wage earnings can be seen in the level of union density measure, UDit. This term is meant to capture the strength of the labor movement and its capacity to achieve a valued goal of egalitarian wage structure.16 The anticipation here is that the parameter β5 on this term would take on a negative value. This brings us to the cluster of variables dealing with government partisanship (LG) and economic coordination (EC). The partisan term used is based on a long-term measure and is labeled left government inheritance. It represents the average of the last twenty years of a government ideology term.17 We assume that the effects of the ideological position of government are not all immediate, and indeed, many are likely to slowly work their effect over time. Recall that in the absence of wage-bargaining coordination we have identified several processes working in different directions. Thus, the net outcome cannot be anticipated theoretically, and the sign and significance of the parameter β6 for the partisan inheritance becomes an empirical question. A more straightforward effect for the economic coordination variable is expected. Thus, the parameter β7 on that variable is anticipated to be negative. That is, as the degree of coordination in the economy rises, the level of wage inequality declines.18 Finally, the parameter β8 on the interaction between these two terms, partisan inheritance and wage-bargaining coordination, is predicted to be negative, which is in keeping with our argument that coordination in the economy facilitates the egalitarian effect of left wing government policy. In the second equation, that for market income, there are three variables on the right-hand side. The first is wage inequality, WI. Since an appreciable amount of market income derives from dependent labor, it is clear that the level of inequality in the former is necessarily dependent on the degree of inequality in wages. Thus, we expect the parameter β9 on this variable to be positive. The second variable expected to influence the level of market income inequality directly is the degree of stock market capitalization, SMC, expressed as a percentage of GDP. (The data are from the World Bank’s Database on Financial Structure and Economic Development.) We anticipate that this variable has an inegalitarian effect in that only those with some degree of wealth or high income can afford to take advantage of the opportunity to earn even more income. Thus, we anticipate that β10 will have a positive and statistically significant sign. Finally, to capture the inequality-heightening effect on market income of a growing pension-age population, we include a demographic measure, OP, which has been operationalized as the percentage of the total population in retirement age (age sixty-five and older) (OECD, Labor Force Sta-

148

Democracy, Inequality, and Representation

tistics). Again, the expectation on the parameter β11 for this variable is that it will be positive.19 The final equation deals with the inequality in disposable income. It concentrates on the political and institutional determinants of fiscal redistribution. We focus on those political and institutional factors that, for a given distribution of market income β12, ultimately shape the allocation of income within society. By using the current level of market income inequality as a control variable, we isolate the effect of the variables of interest from all other determinants of the distribution of disposable income, including the feedback effects of previous redistributive policies (Beramendi 2001). In line with our argument, for a given level of market income inequality, the level of inequality in disposable income is specified as a function of union density (UD), the overall degree of coordination within the economy (EC), and government partisanship (LG). Here we employ the same indicator of the level of coordination within the economy (EC) that we employ in the wage inequality equation. Our expectations regarding the parameters in this equation are that, aside from that for the positive effect of the market income inequality variable, all of the others take on negative signs (β13, β14, β15 < 0). That is to say, the institutional and partisan terms act to reduce inequality. Finally, note that the estimation techniques employed are similar to the ones implemented in the case of market income inequality. We turn now to discuss our empirical findings. To facilitate the interpretation of the results we include an appendix at the end of the chapter with the descriptive statistics of all the variables included in the system of equations.

The Findings The estimation results for the wage inequality equation are based on OLS with robust standard errors. The results are reported in table 5.5. Let us first comment on the set of control variables, those meant to capture the social and economic transformations in OECD labor markets. The anticipated inegalitarian effect of deindustrialization (β1) on wage inequality is not observed. Nor is the impact of wage competition through the increase in imports from the Third World (β2) as expected. The estimated effect of the former is indistinguishable from zero, whereas the latter shows an effect working in the opposite direction to the one we anticipated. However, the anticipated distributive impact of increasing female participation in the labor force (β3 > 0) receives support. The levels of wage dispersion are significantly higher as more women enter the labor force in OECD countries. Fixing the value of all other independent variables, a significant in-

Economic Institutions, Partisanship, and Inequality Table 5.5

149

Estimation Results for Equation 5.1: Wage Inequality OLS (Robust SE)

Manufacturing employment Imports from Third World Female labor force participation rate Distribution of human capital Union density Left government inheritance Overall economic coordination Left government inheritance* overall economic coordination Constant R-squared Observations

.000 (.020) –.072** (.035) .023*** (.008) .018** (.007) –.019*** (.004) .710*** (.171) .306 (.248) –1.69*** (.302) 2.32*** (.360) .90 41

Source: Authors’ calculations. ** p < .05; *** p < .01

crease—that is, one standard deviation, or 8.6 percent—in the female labor force participation rates would lead wage inequality to grow by 14 percent. Similarly, the expected impact of the distribution of human capital in society (β4 > 0) is confirmed by the estimation results. Holding the value of all other variables at their mean, an increase by one standard deviation (equivalent to a 5.94 percent change) in the proportion of adults who have completed a college education makes wage inequality grow from 3.02 to 3.12 (equivalent to a 3.3 percent increase). The strength of the union movement displays the anticipated egalitarian effect on the wage distribution. The coefficient for union density (β5) is both negative and statistically significant. Again, the substantive effects are better illustrated with an example. In a country in which 42 percent of the labor force is unionized and all other variables in the model are set at their mean value, the predicted level of wage inequality would be 3.02. If

150

Democracy, Inequality, and Representation

in such a country the levels of unionization were to increase up to 52 percent, wage inequality would decline to a ratio of 2.84. Thus, countries with strong and encompassing unions are marked by much lower levels of wage inequality than societies where no such unions exist. Examining the estimates on the institutional and partisan variables (β6 to β8) allows us to portray how the interplay between government partisanship and economic coordination shapes wage inequality. A useful approach to analyze the impact of these two terms is to present their conditional coefficients within the upper and lower bounds of the 95 percent confidence interval. Figure 5.3 displays how the effect of overall economic coordination is contingent on different values of government partisanship. In turn, figure 5.4 presents the slope of government partisanship on wage inequality given different levels of coordination in the economy. The reader should focus on two central features of figure 5.3. The first is brought into relief by noting that the values on the vertical dimension are all negative. In essence, this is telling us that the effect of institutionalized economic coordination is to always reduce wage inequality. The second feature is the amplifying effect of leftist partisan inheritance. Eco-

Slope of Overall Economic Coordination

Figure 5.3

Effects of Economic Coordination Contingent on the Levels of Partisan Inheritance

0.5 Slope Upper Bound Lower Bound

0 −0.5 −1 −1.5 −2 −2.5 −3 −3.5 −4 0.27

0.47

0.67

0.87

1.07

1.27

1.47

Levels of Partisan Inheritance Source: Authors’ calculations.

1.67

1.87

Economic Institutions, Partisanship, and Inequality Figure 5.4

151

Effects of Partisan Inheritance Contingent on the Levels of Economic Coordination

Slope of Partisan Inheritance

1.5 Slope Upper Bound Lower Bound

1 0.5 0 −0.5 −1 −1.5 0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1

Levels of Economic Coordination Source: Authors’ calculations.

nomic coordination’s egalitarian impact rises as the level of left partisan inheritance increases. This result suggests the existence, highlighted by our argument, of a mutually reinforcing effect between high levels of coordination and a long history of government dominated by the left. Just as partisanship moderates the impact of coordination, so too is the impact of partisan inheritance conditional upon the levels of coordination in the economy (see figure 5.4). For example, where there is little or no wage coordination, a right wing partisan inheritance modestly elevates the levels of wage inequality, while a left wing partisan inheritance has the perverse effect of increasing that inequality to even greater levels. On the other hand, where there is a coordinated economic environment, the left finds itself in a favorable situation; the greater the level of leftist partisan inheritance, the higher the egalitarian effect on wage distribution. This point is conveyed by table 5.6, where the predicted levels of wage inequality under different partisan and institutional conditions are reported.20 In noncoordinated environments, left wing policy is not only incapable of creating wage equality (as anticipated by our argument) but

152

Democracy, Inequality, and Representation

Table 5.6

Predicted Values of Wage Inequality High WageBargaining Coordination

No WageBargaining Coordination

2.36 3.04

4.24 3.17

Left wing government Right wing government

Source: Authors’ calculations. Note: Predictions based on table 5.5. All other variables are at their mean values.

also seems to be the source of a number of responses by labor and capital that drive wage inequality in a direction opposite to that intended by the government. More specifically, in the context of our previous discussion on minimum-wage regulations, our results suggest that on average the efficiency and spillover effects offset the disemployment effects associated with minimum-wage policy. Alternatively, an institutional environment with high levels of coordination constrains the inegalitarian effects of right wing policies. In sum, our conception of the relationship between partisan politics, wage bargaining coordination, and wage inequality receives a good deal of empirical support. Turning now to the equation on market-based income inequality, we present four sets of estimation results in table 5.7. The first two columns report estimates of the market income equation as specified in equation Table 5.7

Estimation Results for Equation 5.2: Market-Based Income Inequality OLS (Robust SE)

Wage earnings inequality Stock market capitalization Pension-age population Partisanship inheritance Constant R-squared Observations Source: Authors’ calculations. * p < .10; ** p < .05; *** p < .01

TSLS (SE)

OLS (Robust SE)

TSLS (SE)

.015* (.008) .025*** (.007) .009*** (.002) —

.021* (.012) .022*** (.006) .011*** (.003) —

.014* (.008) .024** (.007) .009*** (.002) –.003 (.007)

.024* (.012) .022*** (.006) .011*** (.003) .007 (.01)

.250*** (.060)

.210*** (.080)

.25 (.06)

.189** (.085)

.58 41

.57 41

.58 41

.57 41

Economic Institutions, Partisanship, and Inequality

153

5.2. The goodness-of-fit measures are the same across the two techniques used, namely, OLS with robust standard errors and two-stage least squares. All the parameter estimates obtained (β8 to β10) display the expected signs and are statistically significant. Substantively, the estimated effects are of similar size, with the exception of the parameter (β9) on wage inequality. This parameter has a significant and positive effect in both estimations. However, it is significantly higher in the 2SLS results. Depending on the estimate used, if we keep the other variables in the equation at their mean and allow wage inequality to change from its minimum (2.0) to its maximum (4.6) value, then the relative increase in the size of the Gini index for market-based inequality would be 10 percent in the case of OLS estimates and 14 percent in the case of 2SLS estimates. The transmission of cross-national differences in wage inequality to market-based income inequality appears to be muted in comparison to other factors.21 Clearly, other elements not related to the labor market are also at work in shaping the distribution of market income.22 Recall that based on the notion of a structural dependence of the state on capital, our argument contends that, once the distribution of earnings is controlled for, parties are not able to affect the distribution of market income directly. With the addition of a control for partisan inheritance to the market income equation (equation 5.2), columns 3 and 4 in table 5.8 provide an empirical evaluation of this claim. The results strongly support our argument. Thus, the question remains as to what other determinants are shaping the distribution of market income. The degree of stock market capitalization is one of these. As anticipated, the estimated parameter (β10) is positive and statistically significant. This is consistent with the notion that in countries where participation in the stock market is heavily regulated by the government, the ability of wealthy families to increase their income is constrained. Alternatively, countries with unregulated stock markets provide wealthy families with profitable investment opportunities and they can thereby expand their share of market-based income. The scope of the government’s control of the workings of the stock market is a political decision. Thus, in light of our findings, the degree of stock market capitalization can be seen as a channel through which the distribution of market income becomes politicized, at least in an indirect way.23 In addition, the parameter (β10) capturing the effect of the share of the pension-age population displays the expected sign and is significantly different from zero. This result supports the expectation that as the share of population that has completed the transition from the labor market into retirement increases, market-based inequality also increases as these people lose their principal market-based income: wages and salaries from employment.

154

Democracy, Inequality, and Representation

Table 5.8

Estimation Results for Equation 5.3: Disposable Income Inequality OLS (Robust SE)

Market income inequality Coordinated market economy Union density Left government inheritance Constant R-squared Observations

.326*** (.069) –.042*** (.005) –.0009*** (.0001) –.019*** (.005) .242*** (.032) .87 41

TSLS (SE) .444*** (.093) –.039*** (.008) –.0008*** (.0004) –.017*** (.007) .190*** (.042) .87 41

Source: Authors’ calculations. * p < .10; ** p < .05; *** p < .01

Finally, table 5.8 reports parameter estimates on the determinants of disposable income inequality. As in the case of market-based inequality, OLS with robust standard errors and two-stage least squares are the two techniques being used. Both of the estimated equations show the same goodness of fit. Once again, all the parameter estimates display the expected signs and are statistically significant. They are also very similar in the magnitude of their effects, with the exception of the parameter (β11) on market-based income inequality. Its effect is the largest in the 2SLS estimation. In both equations, however, the expectation that a positive relationship will prevail between market-based and disposable income inequality is confirmed. We turn now to the factors outlined in our argument earlier that would alter a one-to-one duplication of market income distribution in the distribution of disposable income. These three factors included the overall levels of coordination, the degree of unionization of the labor force, and the government’s partisan inheritance.24 All three of these factors’ parameters (β12 to β14) take on the expected negative values and are statistically significant. What are the implications of these values? First, in those societies where there is little or no coordination, employers have no incentive to accept redistribution through the welfare state. Alternatively, in those societies where firms pool risks through cross-shareholding and coordinate with labor, employers concede higher levels of redistribution. In other words, economic coordination brings down the level

Economic Institutions, Partisanship, and Inequality Table 5.9

155

Disposable Income Inequality Low Overall Coordination

Left wing partisan inheritance Right wing partisan inheritance

High Overall Coordination

Low Union Density

High Union Density

Low Union Density

High Union Density

0.31 0.34

0.24 0.27

0.27 0.30

0.20 0.23

Source: Authors’ calculations. Note: Predicted values based on table 5.8, OLS

of inequality in disposable income. Our findings on the role of union density conform to results previously developed by other scholars (Huber and Stephens 2001; Bradley et al. 2003). In those societies where larger shares of workers are unionized, governments need to be more responsive to labor’s demand for insurance and redistribution. Therefore, ceteris paribus, larger unions imply more compressed distributions of disposable income. Finally, left wing partisan inheritance reduces inequality through the long-term institutionalization of higher levels of redistribution.25 This result is consistent with the recent findings by Torben Iversen and David Soskice (2003) about the cumulative impact of government partisanship on poverty reduction. The magnitudes of the effects of interest are analyzed in table 5.9, where the predicted values of the Gini coefficient for disposable income inequality based on the OLS model in table 5.8 are presented. The values in table 5.9 are the predicted levels of disposable income inequality under different combinations of partisanship, union density, and economic coordination.26 Market-based income inequality is set at its mean value (0.39). This exercise reveals that union density can reduce the Gini coefficient of disposable income inequality by up to seven points. This finding implies that in countries with strong unions the amount of redistribution necessary to achieve a perfectly egalitarian society is nearly seven points lower. In turn, overall economic coordination and partisanship inheritance reduce that amount by four and three points, respectively. An alternative way of reading the results reported in table 5.9 would be the following: holding all other variables constant, a change from the minimum to the maximum observed level of union density implies a 22 percent proportional reduction in the value of the Gini coefficient. With similar kinds of changes in the levels of overall economic coordination (from 0 to 0.95) and partisan inheritance (from 1.27 to 2.78), propor-

156

Democracy, Inequality, and Representation

tional reductions in the level of disposable income inequality equal to 13 percent and 9 percent, respectively, would be brought about. Even if independent from each other, the effects of union density, economic coordination, and partisan inheritance combine very differently in the real world. In some countries, such as the Scandinavian nations, all three factors are very high, thus reducing disposable income inequality to its lowest observed levels. In some other countries, the situation is reversed: coordination, union density, and partisan inheritance are very low, and as a result, disposable income inequality reaches its maximum levels. These patterns may lead the reader to wonder about the existence of complementarities between some of these elements— for instance, left wing parties and overall economic coordination. If this is the case, an interaction effect between these two factors should be observed. If left wing parties facilitate the existence of wage-bargaining agreements and depend on them to create an egalitarian wage distribution, should the capacity of left wing parties to shape the distribution of disposable income not be contingent as well upon the overall degree of coordination in the economy? The answer is that no such complementarity is in place, as confirmed by the reestimation, including interaction terms, of the models presented in table 5.8.27 And the reason for this lies in how directly governments are able to shape the distribution of disposable income inequality as opposed to the distribution of earnings. As argued earlier, government’s effects on wage inequality are indirect in that they are contingent on the agreement of unions to wage moderation and high taxes on labor. Such agreement takes place only under conditions of high wage-bargaining coordination, thereby producing the observed interaction effects. Alternatively, an increase in fiscal redistribution reduces inequality of disposable income directly, that is, without any other actors taking part in shaping the final outcome. Thus, for a given value of market income inequality, left wing governments can make use of fiscal redistribution to reduce inequality regardless of the institutional position of any other actor. As a result, no interaction effect is to be observed.

Conclusions This chapter has shown that among the OECD countries the patterns of cross-national variation across different forms of income distribution are quite dissimilar. Contrary to the conventional view, there is little evidence of a dramatic convergence toward greater inequality in terms of the distributions of labor earnings as well as the post-fisc distributions of

Economic Institutions, Partisanship, and Inequality

157

disposable household income. In turn, there are signs of widespread growth in inequality in overall market income. This can be seen as a manifestation of the structural dependence of the state on capital (see Przeworski and Wallerstein 1988; Wallerstein and Przeworski 1995). Although governments are able to shape the distributions of labor earnings and disposable income through taxes, transfers, and regulations, there are certain forms of pre-fisc income that are sheltered from direct government intervention. The larger cross-national variations in the distributions of earnings and disposable income can be attributed to the roles played by political actors (such as unions and, more important, political parties) and institutions that allow actors to coordinate their activities within the economy. The way in which political parties are able to pursue their distributional goals varies across forms of income. Through their legislative legacy, these parties are able to directly work their effect on the distribution of disposable income, in particular via their choices about fiscal redistribution.28 However, their capacity to do so with respect to wage earnings is highly contingent on the national economic institutional framework. Although high levels of economic coordination facilitate the left wing goal of a more egalitarian wage distribution, the pursuit of such goals in the absence of coordination generates perverse effects: the outcome obtained is the opposite of the one originally intended.29 In conjunction with the distinction between labor earnings and market income, our focus on economic coordination and its interplay with partisan traditions helps explain why there is more cross-national variation in terms of wages and disposable income inequality than there is in terms of market income inequality.30 We conclude now by discussing the implications of this chapter for the ongoing debates on the prospects for income inequality. How will inequality evolve in the future? The answer to this question depends in part on what we can expect in terms of the factors examined in this chapter. Consider first the demographic and economic trends. Despite significant cross-national differences, female labor force participation rates have uniformly expanded over the last few decades (Jaumotte 2003). There is not reason to anticipate that they will cease to expand further. However, a countervailing tendency needs to be considered. The discrepancy between male and female wage rates has begun to decrease recently across OECD countries (Blau and Kahn 2000). If this tendency continues, the inegalitarian effects associated with the increasing participation of women in the labor force will be muted. Should this not continue, we would expect a further increase in wage inequality and subse-

158

Democracy, Inequality, and Representation

quently, according to our model, a modest increase in the levels of market-based inequality. OECD population structures have steadily aged over the last few decades, and the prospects are that they will continue to do so. Recent projections for the United Kingdom indicate that the ratio of the elderly (population over age sixty-five) to the working-age population (between twenty and sixty-four years old) will rise from 24 percent in the year 2000 to 43 percent in 2040. Increases of similar or higher magnitude are projected for other OECD nations (Visco 2001). Our analysis suggests that if these forecasts materialize, further increases in market-based income inequality can be expected. Furthermore, the changing structure of financial markets adds new grounds for concern about the evolution of market-based inequality. A study by Thorsten Beck, Asli Demirgüç-Kunt, and Ross Levine (1999) reveals that levels of stock market capitalization have increased throughout the OECD and other countries during the last few decades. In the absence of any change in this widespread trend, we can only conclude that market income inequality is likely to grow even more. Whether these trends toward higher market inequalities carry over to final disposable income depends on the future leverage of political actors and economic institutions. The levels of union density are generally declining (Western 1997). If this continues, we can anticipate a reduction in unions’ capacity to promote egalitarian wage distributions and high levels of fiscal redistribution. However, as this very line of research has also pointed out, not all countries share in the decline. As a result, the levels of cross-national variation have actually increased. Thus, the varying fortunes of labor movements across advanced industrial societies will continue to foster cross-national variations in earnings and disposable income inequality. Contrary to the convergence school, a similar pattern of increasing variation is found in the evolution of economic institutions. Although the pillars of social partnership are said to be crumbling in some coordinated market economies, such as Germany, broader crossnational comparisons have failed to detect a consistent pattern of decline, either in the levels of wage-bargaining coordination or in the relations between organized labor and organized capital (Streeck and Hassel 2003; Wallerstein and Golden 2000). Finally, there seems to be no reason to expect partisan inheritance traditions to converge to the right in the medium run. In sum, the sources for divergence in distributive outcomes among advanced capitalist societies identified in this chapter seem likely to remain in place, keeping the realities of inequality as distant from the “transatlantic consensus” as they are today.

Economic Institutions, Partisanship, and Inequality

159

Appendix Table 5A.1

Descriptive Statistics of Variables Used in the Analysis (Forty-One Observations)

Variable

Mean

Std

Minimum

Maximum

Wage inequality (WI) Market income inequality (MI) Disposable income inequality (DI) Manufacturing employment (ME) Imports from the Third World (TWI) Female labor force participation rate (FP) Proportion of adults with a college degree (HC) Economic coordination (EC) Union density (UD) Partisan inheritance (LG) Stock market capitalization (SMC) Retirement-age population (OP)

3.02 .39 .27 19.40

.79 .04 .05 3.18

1.98 .33 .19 14.18

4.57 .47 .37 25.83

3.24

1.13

1.38

6.77

64.31

8.58

42.26

79.96

11.13 .46 42.84 .93 .48 13.8

5.94 .32 23.83 .42 .39 2.15

4.40 0.0 10.12 .27 .050 9.48

30.30 0.95 87.82 1.78 1.66 17.75

Source: Authors’ calculations.

Notes 1.

2.

3.

4. 5.

In line with the other two measures of income inequality used in this chapter relating to market and disposable income, we would have preferred to use data drawn from the Luxembourg Income Study (LIS). Comparability problems with the wage data in LIS, however, would greatly reduce the number of observations. In addition, the OECD wage data are only available in terms of inter-decile ratios and not in any other form, such as the Gini indices that we use in conjunction with the LIS-based data. A household’s market income includes earnings not only from dependent employment but those deriving from self-employment as well as interest, dividends, rents, and any other income from nonstate sources. The average Gini for Germany from 1984 to 2001 was .32, and that for the United States from 1981 to 2001 was .43. As shown in table 5.1, the 90/10 ratio of labor income in Germany was uniformly below three, while in the United States it was generally above four. Capital income is the difference between market and labor income. In providing a summary measure of the distribution of income, the Gini has the disadvantage of potentially obscuring processes taking place in different

160

6.

7.

8. 9.

10.

Democracy, Inequality, and Representation parts of the distribution (Atkinson and Brandolini 2003). Note, however, that the arguments throughout this chapter are concerned with the overall degree of dispersion across forms of income. John Stephens has suggested that our analysis is misleading because it includes pensioners. The concern is that pretax inequality in countries with “comprehensive” public pension systems (the Nordic countries) would be “artificially” high because pensioners in these countries make no provisions for retirement outside the public system. Since our concern is with societywide income inequality, it seems inappropriate to consider only “one variant of the working-age population.” We do not deny that the welfare state has as one of its primary clients those of retirement age. Indeed, the failure of some welfare states to provide adequate support for those of pensionable age is a major problem and should not be pushed aside nor relegated to the status of a nuisance. Stephens and his co-authors (Bradley et al. 2003, 224–25) “demonstrate that the assertion that the welfare state merely redistributes income across generations is wrong.” We question neither their contribution nor the fact that the welfare state does more than engage in intergenerational redistribution. However, in showing that redistribution is not simply across generations, we cannot help but observe that the intergenerational redistributive aspect of the welfare state is important to a significant proportion of the citizenry. In addition, a good deal of cross-class redistribution occurs within pension systems. Most pension systems are not guided exclusively by insurance principles. Finally, as we point out later, our findings hold regardless of the demographic base used for our income distribution measures. Note that Alberto Alesina and Edward Glaeser’s (2004) recent book on the differences between the United States and Europe in terms of both poverty and the efforts of the state to relieve it is somewhat misleading in its portrayal of transatlantic differences in market income distribution. It proceeds under the assumption that the gulf in the level of inequality between Americans and Europeans is very wide before taxes and redistribution (3, 56, 58)—indeed, far wider than it really is—and the authors thereby infer that the European welfare state systems are less redistributive than then actually are. A cursory examination of tables 5.2 and 5.3 shows how inaccurate Alesina and Glaeser’s portrayal is. Thus, we use a broader measure that captures the overall coordination within the economy developed by Peter Hall and Daniel Gingerich (2001). See Jonas Pontusson, David Rueda, and Christopher Way (2002); Rueda and Pontusson (2000, fn. 22); Rueda (chapter 6, this volume). The reader should be aware that our empirical analyses are not strictly comparable with those included in any of the other chapters in this volume. The transition toward a less coordinated and partially decentralized labor

Economic Institutions, Partisanship, and Inequality

11.

12.

13.

14.

15.

161

relations system in Sweden begs the question of whether, in the context of a globalized economy, employers and right wing parties have an incentive to pursue the retrenchment of the welfare state by defecting from or undermining wage coordination agreements. To the extent that this has been the case in Sweden, recent comparative analyses portray it as the exception, not the rule. To be sure, partial decentralization has occurred in other countries (Denmark), but the level of coordination among the decentralized units has actually increased. These changes would reflect functional adjustments to the internationalized economy endorsed by parties across the political spectrum as opposed to an institutional change promoted by right wing parties. See Franz Traxler (2003), Jen Blom-Hansen (2000), and Mette Anthonsen and Johannes Lindvall (2005). These effects have been identified primarily by scholars working within an efficiency wages framework as opposed to a mind-set articulated around the assumption of perfectly competitive labor markets (Rebitzer and Taylor 1995). In a nutshell, the efficiency effect is stronger in the short run (leading to positive employment effects and potentially wider earnings distributions), whereas the unemployment effects tend to become visible only in the long run. For an analysis of the employment effects of minimum-wage legislation across Canadian provinces, see Michael Baker, Dwayne Benjamin, and Shuchita Stanger (1997). The controversy around the different effects at work in the United States is particularly illuminating. David Newmark and William Washer (1992) found a negative effect of minimum wages on the employment of teens and young adults only in the long run. In contrast, focusing specifically on the fast-food industry in New Jersey, David Card and Alan Krueger (1994, 2000) found a positive employment effect of higher minimum wages. For a more general evaluation of the debate on minimum wages in the United States, see David Card and Alan Krueger (1995). The variable WI is the ratio of the ninetieth percentile to the tenth percentile of the distribution of earnings of full-time employees (see note 1 and table 5.1). MI is the Gini index for market income (table 5.3). DI is the Gini index for household disposable income (table 5.4). We do not report the results based on PCSE estimates because of the small number of time units. As Karl Moene and Michael Wallerstein (2003) pointed out, PCSE may be less accurate under such a condition. Nevertheless, the PCSE-based results are almost identical to those deriving from OLS with robust standard errors, as reported later in the chapter. Results are available from the authors. Clearly, there are other dimensions of female labor force participation that affect wage inequality, such as the greater likelihood of women taking part-

162

16. 17. 18.

19. 20.

21.

Democracy, Inequality, and Representation time jobs because of competing family burdens. The wage inequality data used in this chapter refer to full-time workers. Union density is the percentage of the labor force that holds membership in a union. This variable is a measure of the center of political gravity that characterizes the cabinet (Cusack and Engelhardt 2002). We use the indicator of coordination developed by Peter Hall and Daniel Gingerich (2001). The Hall and Gingerich measure captures not only the degree of wage-bargaining coordination but other dimensions, such as the structure of corporate governance and the existing complementarities between corporate governance and the labor relations system. The advantage of this indicator is that it takes into account the interdependencies between unions and employers in the context of coordination agreements. Unions accept wage restraint and high labor taxes only if employers accept the development of a large public insurance system. This agreement is more likely to exist if coordination among employers is in place. In any case, the results obtained for the wage inequality equation are not sensitive to substituting the Hall and Gingerich measure for the wage coordination index (WBC) developed by Lane Kenworthy. Two different estimation techniques are used for this equation: OLS with robust standard errors and two-stage least squares. “High coordination” refers to the maximum value of the Hall and Gingerich (2001) coordination index in our data set, which is 0.95. Lack of coordination (0) implies, among other things, fragmented wage bargaining and an absence of institutionalized corporate governance. The partisan inheritance values are the same values reported on the horizontal axis of figure 5.2. A value of 0.27 represents right wing partisan inheritance, whereas a value of 1.78 captures the far-left partisan inheritance. These results hold even when we adjust our analysis to take into account the suggestions by John Stephens (see note 6). As he points out, if his criticisms hold, the impact of the wage dispersion term on market-based income inequality should be much stronger than we find. We examined this conjecture. First, we calculated the Ginis of market-based inequality for the more restricted population group that David Bradley and his colleagues (2003) employ. Second, we replaced the variable measuring the share of the pension-age population with the unemployment rate. We estimated the equation using OLS with robust standard errors and 2SLS. The estimates obtained are similar to those in table 5.9. The parameter estimate using OLS on these data is .018, as opposed to .015. In turn, the parameter estimate using 2SLS is .041, as opposed to .021. Complete results for the two are available from the authors. We do not see the fact that these parameters are slightly larger as undermining our argument. We address this issue again

Economic Institutions, Partisanship, and Inequality

22.

23.

24.

25.

26.

163

when discussing the estimates for the disposable income inequality model in note 25. Jonas Pontusson has suggested that much of the gap that we identify between earnings and market income inequality may be an artifact of using different levels of analysis (see also Kenworthy and Pontusson 2005). Wages are measured at the individual level, whereas market income is expressed in terms of household income per equivalent adult. In doing so, we would be excluding from the analysis of market income inequality important aspects of the working of OECD labor markets that are ultimately reflected in different employment levels. We have reestimated the market income equations reported in table 5.8, adding a control for the level of total civilian employment as a percentage of the working-age population. The results are not affected at all by this change in the specification. Similarly, our findings are robust to the inclusion of controls for other labor market–related aspects, such as the level of unemployment and the percentage of GDP devoted to compensation of employees. Hall and Soskice (2001) argued that there is a complementarity between systems with a strong legislature (PR systems) and high levels of economic coordination. PR systems are better than majoritarian systems at providing actors with this monitoring and control capacity over government. As a result, government commitments are more credible in PR systems. This facilitates the long-term functioning of economic institutions and long-term financing of firms. In turn, the agreements reached within economic institutions directly shape government policy, including the degree of regulation in the stock market. Preliminary analyses (available from the authors) lend strong support to the position that national systems with dominant legislatures (see Cusack and Beramendi 2006) tend to have less-developed stock markets. Some have argued that any specification including both union density and partisan inheritance is inappropriate owing to the presence of high levels of multicollinearity (Bradley et al. 2003). In our view, even if these three factors partially covary, the causal logic linking each of the three factors to inequality is sufficiently independent to permit their inclusion in the model. From a statistical point of view, there are no compelling reasons to exclude any of these variables. None of the variables included in the model has a variance inflation factor higher than 1.3. Thus, there is no multicollinearity problem. Our results are independent of the use of either the total population or the working-age population employed by Bradley and his colleagues (2003). The parameter estimates and the associated statistics are practically the same as those found in table 5.9. Results are available from the authors. In calculating the predictions for table 5.9, “low” and “high” refer, respec-

164

27. 28.

29.

30.

Democracy, Inequality, and Representation tively, to the minimum (0) and the maximum (.95) values in the sample. A low value of union density means that only 10 percent of the labor force is unionized, whereas a high value implies that this percentage rises to 87.9 percent. Finally, a low value on the partisan term (0.27) represents right wing partisan inheritance, whereas a value of 1.78 captures the far left. These results are available from the authors. Our findings complement those of Bradley et al. (2003) in highlighting that the achievement of a more egalitarian distribution of income requires a long-term presence of Social Democratic parties in office. This finding is related to earlier contributions by Rueda and Pontusson (2000), Pontusson, Rueda, and Way (2002), and Rueda (this volume) on the relationship between partisanship, the varieties of capitalism, and wage inequality. Although these authors saw partisanship and coordination as substitutes for each other, our analysis suggests that left wing partisanship and economic coordination are better understood as complements of each other. These results directly complement Iversen and Soskice’s (chapter 4, this volume) analysis of how PR systems bias electoral outcomes toward the left, thereby shaping the politics of redistribution.

References Alesina, Alberto, and Edward Glaeser. 2004. Fighting Poverty in the United States and Europe: A World of Difference. Oxford: Oxford University Press. Anthonsen, Mette, and Johannes Lindvall. 2005. “Political Competition and the Resilience of Corporatism.” Paper presented to the meeting of the Midwest Political Science Association, Chicago, Ill., April 7–10, 2005. Atkinson, Anthony. 1999. “Is Rising Income Inequality Inevitable? A Critique of the Transatlantic Consensus.” WIDER annual lecture, University of Oslo, November 1, 1999. Oslo: UNU-WIDER (United Nations University, World Institute for Development Economics Research). Atkinson, Anthony B., and Andrea Brandolini. 2003. “The Panel-of-Countries Approach to Explaining Income Inequality: An Interdisciplinary Research Agenda.” Unpublished paper. Oxford University. Baker, Michael, Dwayne Benjamin, and Shuchita Stanger. 1997. “The Highs and Lows of the Minimum Wage Effect: A Time Series–Cross-Section Study of the Canadian Law.” Journal of Labor Economics 17(2): 318–50. Barro, Robert J., and Jong-Wha Lee. 2000. “International Data on Educational Attainment: Updates and Implications.” Working paper 42. Cambridge, Mass.: Harvard University, Center for International Development (August). Bartels, Larry. 2003. “Partisan Politics and the U.S. Income Distribution.” Unpub-

Economic Institutions, Partisanship, and Inequality

165

lished paper. Princeton University, Department of Political Science and Woodrow Wilson School of Public and International Affairs. Beck, Thorsten, Asli Demirgüç-Kunt, and Ross Levine. 1999. “A New Database on Financial Development and Structure.” Working paper 2146. Washington, D.C.: World Bank. Benjamin, Dwayne, Morley Gunderson, and W. Craig Riddell. 1998. Labor Market Economics. New York: McGraw-Hill. Benner, Mats. 2003. “The Scandinavian Challenge: The Future of Advanced Welfare States in the Knowledge Economy.” Acta Sociologica 46(2): 132–49. Beramendi, Pablo. 2001. “The Politics of Income Inequality: The Role of SecondOrder Effects.” Working paper 284. Luxembourg: Luxembourg Income Study (September). Blau, Francine, and Lawrence Kahn. 2000. “Gender Differences in Pay.” Journal of Economic Perspectives 14(4): 75–99. Blom-Hansen, Jens. 2000. “Still Corporatism in Scandinavia? A Survey of Recent Empirical Findings.” Scandinavian Political Studies 23(2): 157–81. Blyth, Mark. 2001. “The Transformation of the Swedish Model: Economic Ideas, Distributional Conflict, and Institutional Change.” World Politics 54(1): 1–26. Bradley, David, Evelyne Huber, Stephanie Möller, François Nielsen, and John D. Stephens. 2003. “Distribution and Redistribution in Postindustrial Democracies.” World Politics 55(2): 193–228. Card, David, and Alan B. Krueger. 1994. “Minimum Wages and Employment: A Case Study of the Fast-Food Industry in New Jersey and Pennsylvania.” American Economic Review 84(4): 772–93. ———. 1995. Myth and Measurement: The New Economics of Minimum Wage. Princeton, N.J.: Princeton University Press. ———. 2000. “Minimum Wages and Employment: A Case Study of the FastFood Industry in New Jersey and Pennsylvania: Reply.” American Economic Review 90(5): 1397–420. Cusack, Thomas R., and Pablo Beramendi. 2006. “Taxing Work.” European Journal of Political Research 45(1): 43–73. Cusack, Thomas R., and Lutz Engelhardt. 2002. “The PGL File Collection: File Structures and Procedures.” Paper. Berlin: Social Science Research Center Berlin (WZB). Esping-Andersen, Gøsta. 1990. The Three Worlds of Welfare Capitalism. Princeton, N.J.: Princeton University Press. Fitoussi, Jean-Paul. 1994. “Wage Distribution and Unemployment: The French Experience.” American Economic Review 84(2): 59–64. Freeman, Richard. 1980. “Unionism and the Dispersion of Wages.” Industrial and Labor Relations Review 34(1): 3–23. Ganghof, Steffen. 2005. “Progressive Income Taxation in Advanced OECD Coun-

166

Democracy, Inequality, and Representation

tries. Revisiting the Structural Dependence of the State on Capital.” Paper, Max Plank Institute, Cologne. Global Social Policy 5(1): 77–95. Goldthorpe, John, editor. 1984. Order and Conflict in Contemporary Capitalism. Oxford: Oxford University Press. Gottschalk, Peter, and Timothy Smeeding. 1997. “Cross-National Comparisons of Earnings and Income Inequality.” Journal of Economic Literature 35(2): 633–86. ———. 2000. “Empirical Evidence on Income Inequality in Industrial Countries.” In Handbook of Income Distribution, edited by Anthony B. Atkinson and François Bourguignon. New York: Elsevier. Green, William. 2000. Econometric Analysis. New York: Prentice-Hall. Hall, Peter, and Daniel Gingerich. 2001. “Varieties of Capitalism and Institutional Complementarities in the Macroeconomy.” Unpublished paper. Harvard University. Hall, Peter, and David Soskice. 2001. “Introduction.” In Varieties of Capitalism, edited by Peter Hall and David Soskice. New York: Oxford University Press. Hibbs, Douglas. 1977. “Political Parties and Macroeconomic Policy.” American Political Science Review 71(4): 1467–87. ———. 1991. “Market Forces, Trade Union Ideology, and Trends in Swedish Wage Dispersion.” Acta Sociologica 34(2): 89–102. Hibbs, Douglas, and Christopher Dennis. 1988. “Income Distribution in the United States.” American Political Science Review 82(2): 467–90. Hicks, Alexander, and Duane Swank. 1984. “Government Redistribution in Rich Capitalist Democracies.” Policy Studies Journal 13(2): 264–87. Hogfeldt, Peter. 2003. “The History and Politics of Corporate Ownership in Sweden.” Working paper 30/2003. Stockholm: Stockholm School of Economics. Huber, Evelyne, and John D. Stephens. 2001. Development and Crisis of the Welfare State. Chicago, Ill.: University of Chicago Press. Iversen, Torben, and David Soskice. 2003. “Electoral Systems and the Politics of Coalitions: Why Some Democracies Redistribute More Than Others.” Unpublished paper. Harvard University. Iversen, Torben, and Anne Wren. 1998. “Equality, Employment, and Budgetary Restraint: The Trilemma of the Service Economy.” World Politics 50(4): 507–56. Jaumotte, Florence. 2003. “Female Labor Force Participation: Past Trends and Main Determinants in OECD Countries.” Unpublished paper. Economics Department, OECD. Kenworthy, Lane, and Jonas Pontusson. 2005. “Rising Inequality and the Politics of Redistribution in Affluent Countries.” Perspectives on Politics 3(3): 449–71. Korpi, Walter. 1983. The Democratic Class Struggle. New York: Routledge. Korpi, Walter, and Joakim Palme. 2003. “New Politics and Class Politics in the Context of Austerity and Globalization: Welfare State Regress in Eighteen Countries, 1975–1995.” American Political Science Review 97(3): 425–46. Mares, Isabela. 2003. The Politics of Social Risks. Cambridge: Cambridge University Press.

Economic Institutions, Partisanship, and Inequality

167

Mendoza, Enrique, Assaf Razin, and Linda L. Tesar. 1994. “Effective Tax Rates in Macroeconomics: Cross-Country Estimates of Tax Rates on Factor Incomes and Consumption.” Journal of Monetary Economics 34(3): 297–323. Moene, Karl, and Michael Wallerstein. 2003. “Earnings Inequality and Welfare Spending: A Disaggregated Analysis.” World Politics 55(4): 485–517. Newmark, David, and William Washer. 1992. “Employment Effects of Minimum and Subminimum Wages: Panel Data on State Minimum Wage Laws.” Industrial and Labor Relations Review 46(1): 55–81. Nickell, Stephen, and Brian Bell. 1996. “Changes in the Distribution of Wages and Unemployment in OECD Countries.” American Economic Review 86(2): 302–8. Organization for Economic Cooperation and Development (OECD). 1996. Employment Outlook. Paris: OECD. Phillips, Kevin. 2002. Wealth and Democracy. New York: Broadway Books. Pontusson, Jonas, David Rueda, and Christopher Way. 2002. “Comparative Political Economy of Wage Distribution.” British Journal of Political Science 32(2): 281–308. Przeworski, Adam, and Michael Wallerstein. 1988. “The Structural Dependence of the State on Capital.” American Political Science Review 82(1): 11–29. Rebitzer, James B., and Lowell J. Taylor. 1995. “The Consequences of Minimum Wage Laws: Some New Theoretical Ideas.” Journal of Public Economics 56(2): 245–55. Rueda, David, and Jonas Pontusson. 2000. “Wage Inequality and the Varieties of Capitalism.” World Politics 52(3): 350–83. Steinmo, Sven. 2002. “Globalization and Taxation: Challenges to the Swedish Welfare State.” Comparative Political Studies 35(7): 839–62. Stephens, John D. 1979. The Transition from Capitalism to Socialism. New York: Macmillan. Streeck, Wolfgang, and Anke Hassel. 2003. “The Crumbling Pillars of Social Partnership.” West European Politics 26(4): 101–24. Swenson, Peter. 2003. Capitalist Against Markets. Oxford: Oxford University Press. Swenson, Peter, and Jonas Pontusson. 2000. “The Swedish Employer Offensive Against Centralized Bargaining.” In Unions, Employers, and Central Banks, edited by Torben Iversen, Jonas Pontusson, and David Soskice. Cambridge: Cambridge University Press. Traxler, Franz. 2003. “Bargaining, State Regulation, and the Trajectories of Industrial Relations.” European Journal of Industrial Relations 9(2): 141–61. Van Kersbergen, Kees. 1995. Social Capitalism. New York: Routledge. Visco, Ignazio. 2001. “The Fiscal Implications of Aging Populations in OECD Countries.” Paper presented to the meeting of the Oxford Center on Population Aging, Oxford, June 2001. Wallerstein, Michael. 1999. “Wage-Setting Institutions and Pay Inequality in Advanced Industrial Societies.” American Journal of Political Science 43(3): 649–80.

168

Democracy, Inequality, and Representation

Wallerstein, Michael, and Miriam Golden. 2000. “Postwar Wage Setting in the Nordic Countries.” In Unions, Employers, and Central Banks, edited by Torben Iversen, Jonas Pontusson, and David Soskice. Cambridge: Cambridge University Press. Wallerstein, Michael, Miriam Golden, and Peter Lange. 1997. “Unions, Employers’ Associations, and Wage-Setting Institutions in Northern and Central Europe, 1950–1992.” Industrial and Labor Relations Review 50(3): 379–401. Wallerstein, Michael, and Adam Przeworski. 1995. “Capital Taxation with Open Borders.” Review of International Political Economy 2: 425–45. Western, Bruce. 1997. Between Class and Market: Postwar Unionization in Capitalist Democracies. Princeton, N.J.: Princeton University Press. Wood, Adrian. 1995. North-South Trade, Employment, and Inequality. Oxford: Clarendon Press.

Chapter 6

Political Agency and Institutions: Explaining the Influence of Left Government and Corporatism on Inequality DAVID RUEDA

It is well known that wage inequality has increased dramatically in the United States over the last three decades. From 1973 to 1998, the hourly earnings of a full-time worker in the ninetieth percentile of the American distribution (someone whose earnings exceeded those of 90 percent of all workers) relative to a worker in the tenth percentile grew by 25 percent, and the corresponding figure for men only was nearly 40 percent. In the words of Paul Krugman (1996), today America is no longer a “middle-class nation.” Wage inequality has increased in most other OECD countries as well, but the extent of this phenomenon varies a great deal. Cross-national differences in levels of wage inequality, in fact, remain as great as they were in the 1970s. In the United States, the worker in the ninetieth percentile earned 4.63 times as much as the worker in the tenth percentile in 1996. In Sweden, on the other hand, a worker in the ninetieth percentile earned only 2.27 times as much as the worker in the tenth percentile. Inequality is frequently invoked as an explanation for a number of crucial issues in political science. It is often considered a determinant of processes as diverse as the decline of electoral turnout (Rosenstone and Hansen 1993; Verba, Nie, and Kim 1978), the increase in the support of extreme-right parties (Betz 1994), or the likelihood of political conflict (for a review, see Lichbach 1989). At the same time, recent work by labor economists demonstrates that supply and demand factors alone cannot

170

Democracy, Inequality, and Representation

account for cross-national variation in wage inequality (Blau and Kahn 1996; Freeman and Katz 1995; Gottschalk and Smeeding 1997). Most analysts would agree that political factors influence inequality in significant ways. Because inequality has political determinants and political consequences, it deserves to be a central concern of comparative political economy. The politics of inequality are fundamentally influenced by questions about political agency and institutional constraints. A large and influential literature in comparative politics has emphasized partisan differences as a determinant of political and economic outcomes (James Alt [1985] and Douglas Hibbs [1977, 1987] are commonly cited examples). According to this framework, political agency is indeed important, and different parties promote distinct economic outcomes (in terms of equality, unemployment, inflation, and so on). Other authors, however, have emphasized the role of institutions as a mediating force. Institutions, they argued, shape the ability of political actors to affect the economy (see, for example, Steinmo, Thelen, and Longstreth 1992). In the following pages, I argue that to understand the relationship between political agency and inequality we must do two essential things: separate the effects of government partisanship and policy on the economy and assess the influence of political agency once we account for the mediating role of institutions. A number of analysts of the political economy of industrialized democracies have argued that the partisan nature of governments should influence the levels of inequality in the economy (see, for example, Pontusson, Rueda, and Way 2002; Rueda and Pontusson 2000; Wallerstein 1999). While sharing the general partisan assumptions presented in this literature, I wish to emphasize in this chapter that governments do not possess the ability to transform the wage distribution directly. In other words, governments, whether conservative or liberal, cannot legislate a particular amount of inequality. They must rely on the design and implementation of policy to accomplish any degree of redistribution. To assess accurately the influence of government partisanship, therefore, we must explicitly separate the effects of political agency on policy and the effects of policy on inequality. The second element in the argument I present in this chapter is related to the role of institutions as factors affecting political agency. I argue that, in an analysis of inequality, the effects of government partisanship on policy and the effects of policy on economic outcomes are contingent on institutions. The starting point for my analysis is that partisan differences do affect the policies that governments are likely to promote. But I argue that these partisan differences are influential only when some institutions are in place. More specifically, I argue that even when they are

Political Agency and Institutions

171

committed to redistribution, leftist governments do not promote egalitarian policies unless they are convinced that the institutional context allows these policies to affect economic outcomes.

Wage Inequality in the OECD This chapter focuses on the effects of the relationship between political agency and institutions on the lower half of the wage distribution. As pointed out by Andrea Brandolini and Timothy Smeeding (this volume) and Pablo Beramendi and Thomas Cusack (this volume), there are significant differences in the incidence of inequality across OECD countries depending on what kind of income we look at. Nevertheless, wages are the most important component of individual and household income—and the wages of those in the lower half of the distribution are particularly important to left parties. An analysis of the influence that governments exert on inequality should therefore emphasize those with the lowest wages. It is reasonable to assume that if left government affects inequality, it does so by raising the wage levels of the most needy. Table 6.1 summarizes the wage inequality data that serve as the dependent variable for my analysis. For each country, the table provides the mean value for wage inequality at the lower half of the wage distribution (the 50/10 ratio) for the entire period 1973 to 1995, as well as the percentage change from the earliest to the most recent observation. It should be noted at the outset that these inequality measures refer to gross income from employment for individuals: they ignore other sources of income (government transfers, self-employment, income from capital, and so on) and exclude the distributive effects of taxation and income pooling within households. The data also are restricted to full-time employees, except in the case of Austria. Since part-time employees invariably earn less, on an hourly basis, than full-time employees, the figures in table 6.1 understate the extent of wage inequality in the other countries. And because the incidence of part-time employment has increased in most OECD countries since the early 1980s, these figures also understate the upward trend in wage inequality. Table 6.1 reveals important cross-national variation in wage inequality. In these sixteen countries, the average 50/10 ratio for the 1973 to 1995 period was 1.64. In other words, a person in the fiftieth percentile of the wage distribution (the wage median) earned, on average, 1.64 times as much as a person in the tenth percentile. Sweden, with an average 50/10 ratio of 1.33, stands out as the OECD country with the most compressed lower-half wage distribution. While the Scandinavian countries fall within a narrow range of very compressed lower-half wage dis-

172

Democracy, Inequality, and Representation

Table 6.1

Means and Percentage Changes in Inequality 50/10 Ratios

Country and Years Covered

Mean

Percentage Change

Australia, 1976 to 1995 Austria, 1980 to 1994 Belgium, 1986 to 1993 Canada, 1973 to 1994 Denmark, 1980 to 1994 Finland, 1977 to 1995 France, 1973 to 1995 Germany, 1984 to 1995 Italy, 1986 to 1995 Japan, 1975 to 1995 Netherlands, 1977 to 1995 Norway, 1980 to 1994 Sweden, 1975 to 1995 Switzerland, 1990 to 1995 United Kingdom, 1973 to 1995 United States, 1973 to 1995

1.66 1.96 1.45 2.30 1.40 1.46 1.66 1.63 1.42 1.70 1.56 1.39 1.33 1.61 1.78 2.00

3.1 0.0 –1.4 9.1 –2.8 –10.2 –5.7 –11.9 –3.4 –6.3 5.8 –6.4 0.0 0.0 1.5 11.0

Average Standard deviation

1.64 0.26

–1.1 6.38

Source: OECD (1996, 61–62) for all countries except the United States; for the United States, OECD (1993, 161; 1996, 103). Note: The percentage changes measure the variation from earliest to latest available observation in the country series.

tributions, the continental European countries included in this data set (France, Belgium, Germany, Italy, the Netherlands, and Switzerland) can be classified as a group with inequality levels slightly below the OECD average. The exception is, of course, Austria, which belongs at the opposite end of the spectrum with the United States, the United Kingdom, Japan, and Canada. All these countries exhibit considerably larger-thanaverage levels of inequality at the lower half of the wage distribution. Turning to change over time, the variation in the data is also very noticeable. From the earliest to the most recent observation available for each country, there are large increases of 50/10 inequality in the United States, Canada, and Australia. However, lower-half wage inequality fell quite significantly in Germany, Finland, Norway, and Japan. Table 6.1 shows the high degree of cross-country and over-time variation that exists in the sample. What accounts for these different patterns?

Political Agency and Institutions

173

In the following pages, I argue that the interplay of partisanship, corporatism, and policy is an important part of the story.

The Puzzle: Government Partisanship and Inequality at the Lower Half of the Wage Distribution The starting point for this chapter’s exploration of the determinants of wage inequality is the hypothesis that the partisan nature of governments influences wage inequality. Governments can influence a country’s wage distribution through a variety of policies (for example, those affecting minimum wages, social wages, taxes, and so on). The argument supporting the existence of a relationship between government partisanship and inequality at the lower half of the wage distribution can be explained in very simple terms. It hinges on the proposition that the policy preferences of left parties raise the wage floor for competition in the labor market. If there is a legislated minimum wage, for example, left governments are likely to set the minimum wage closer to the median wage than right governments. Another example would be the tendency of left governments to favor higher levels of social wages, therefore curtailing the inegalitarian effects of unemployment, and the general tendency of left governments to boost the relative bargaining power of unskilled workers. In one of the few existing political analyses of inequality at the lower end of the distribution, Jonas Pontusson, David Rueda, and Christopher Way (2002) analyzed the determinants of 50/10 ratios to test whether left governments do in fact raise the relative market power and wages of poorly paid workers. They did this through a set of regressions in which the relationship between Thomas Cusack’s (1997) measure of government partisanship and the levels of 50/10 inequality is explored. I reproduce their main results in the first column in table 6.2. The results in table 6.2 show that government partisanship does not significantly influence inequality at the lower half of the wage distribution.1 As Pontusson, Rueda, and Way recognized, this is a puzzling finding. In this regression, the coefficient for government partisanship is positive (as expected, social democratic governments would be associated with lower levels of wage inequality), but it does not come close to reaching statistical significance. These results therefore offer little support for the hypothesis that left parties promote relative wage gains for poorly paid workers by setting a floor for competition in the labor market. Since the lack of significance of the partisanship variable could be interpreted as a result of the presence of country and time dummies (the

174

Democracy, Inequality, and Representation

Table 6.2

The Effects of Government Partisanship on Inequality in the Lower Half of the Wage Distribution

Constant

Lagged dependent variable

Cabinet partisanship

Unemployment rate

Trade with less-developed countries

Female labor force participation

Private sector services

Union density

Wage bargaining centralization

Public sector employment

Observations Adjusted R-squared Fixed effects



.484 (.065) w0, the constraint γ ≤ 1 binds, all spending is insurance, and the second and now solely determinant first-order condition becomes u′(c E* ) ⎛ e ⎞ ⎛ α ⎞ ⎛ τ ′(t )w ⎞ =⎜ ⎟⎜ ⎟⎜ ⎟ u′(c N* ) ⎝ 1 − e ⎠ ⎝ β + r ⎠ ⎝ w L ⎠

[8.18],

which establishes a similarly positive, but flatter (the new term is less than 1), relation of wL to insurance-cum-total spending. M&W figure 3 (figure 8.1) illustrates these conclusions graphically: The figure reads as follows. All considerations are of mean-preserving movements in inequality, which M&W consider as movements of wL relative to fixed w ¯. As wL increases (skew decreases), the desired, unconstrained level of benefits to the unemployed strictly increases (the smoothly upward-sloping curve) and the level of total social spending strictly declines (the downward-sloping curve). Accordingly, the unconstrained, desired share of spending targeted to the unemployed (employed) also strictly increases (decreases) (the ratio of the preceding two curves). However, beyond some wage-cum-equality level, w0, unconstrained, desired insurance exceeds unconstrained, desired total social spending, so the constraint binds. Beyond this point, all spending targets the unemployed, and this insurance-cum-total spending remains upward

Inequality and Unemployment Figure 8.1

237

Preferred Policy of Employed Wage Earners

Spending

Total Benefits, or T(t*)

Total Benefits When All Benefits Are Targeted to Those Without Earnings, or T(t*⏐γ =0) Benefits for Those Without Earnings, or (1 − e)c*N Wage w0

w

Source: Moene and Wallerstein (2001, figure 3). Reprinted with permission.

sloping in wages-cum-equality, but with a desire to restrain total taxes dampening that slope. Thus, in equilibrium, welfare-targeted (insurance) spending strictly rises with wL (equality), although with a kink at w0 and more slowly thereafter; total social spending (insurance plus redistribution) declines with wL (equality), kinks at w0, and rises (more slowly) thereafter; and the share of total spending targeted to the unemployed rises weakly monotonically with wL (equality), reaching unity at w0 and staying at 100 percent thereafter.

Median-Preserving Increases in Income Skew in the Moene-Wallerstein Model Moene and Wallerstein (2001) discussed only mean-preserving increases in income skew, which they equated with falling median incomes and, implicitly, upper class incomes rising (to fix the mean). Median-preserving skew increases, conversely, would have mean income rising with the median fixed, implying that upper class incomes rise.11 The implications

238

Democracy, Inequality, and Representation

of such “yachts outpacing tugboats and rowboats”—which is by far the more common case empirically12—can be seen using figure 8.1 and the equilibrium conditions [8.14], [8.15], or [8.18]. In using figure 8.1, note that increases in w ¯ with wi fixed will be shifts in the curves and not moves along them. The T(t*) curve shifts outward as the base for the median to tax expands, while her own income stays fixed. Accordingly, she wants greater social spending at whatever wi she has stagnated. The larger tax base also enables the median to raise unemployment benefits, that is, to boost her social insurance consumption, and insurance being a normal good, she will. Thus, both curves shift outward: median-preserving skew-increases raise insurance and total spending. To see how redistribution and the tarw geting ratio are affected, consider [8.14] and [8.15]. In [8.14], as wi falls (skew rises), τ’ must fall, so t and thus total spending must rise, as just noted. In [8.15], the right-hand side is unaffected by an increase in w ¯, so the left-hand-side ratio of marginal utilities under (un)employment must not change either. However, with wi fixed and tax rates and base higher, c*E rises, and so the numerator marginal-utility falls. The denominator marginal-utility must therefore decline proportionately, and so c*N must rise too (that is, insurance spending rises, as we just saw). Furthermore, since c*N < c*E and utility is concave, denominator marginal-utilities exceed and decline faster than those in the numerator, so holding the ratio fixed requires a smaller increase in c*N than in c*E. Then, recall that c*E = (1 – t)wi + γτw ¯ and c N* = (1 − γ ) 1−e e τw . The τw ¯ term rises the same amount in each expression, of course, and the (1 – t)wi term in c*E declines because t rises and wi is fixed. Since 1−e e > 1 and is unchanged, and yet c*E must increase more than does c*N, we conclude that the share of spending on the employed, γ, and so its level, must increase. Finally, with both curves shifting upward, but the total spending curve more so, the kink point, w0, beyond which wage (equality) level all spending is insurance-targeted, must also shift outward (to a higher wage). Thus, in the equilibrium with γ* < 1, median-preserving skew increases—“yachts outpacing tugboats and rowboats”—raise total social policy spending, insurance (unemployed-targeted) spending, redistribution (employed-targeted) spending, and the share of redistribution in the total, and this unconstrained equilibrium applies through greater equality levels. In the constrained case of γ* = 1, “top-pulled” and “bottomdragged” increases in skew have similar effects,13 but the constraint binds in a narrower range (of lower skew: wi nearer w ¯) for top-pulled increases. These notable differences between the effects of top-pulled and bottom-dragged increases in skew, given the empirical prevalence of the former, imply that empirical results should differ accordingly depending on whether the average income level is controlled. With or without such

Inequality and Unemployment

239

control, we would expect to find effects of mean-/median-preserving increases in skew, respectively.

Continuous and Correlated Income Distributions and Unemployment Risks For our purposes, replacing M&W’s three discrete classes with continuous voter heterogeneity adds little beyond enhancing intrinsic realism and changing subscripts from the decisive median group, L, to the decisive median voter, i.14 Rather, the move serves more to facilitate exploring the correlation of unemployment risk to income, which proceeds by indexing αi and βi and assuming their ratio is negatively related to wi. To simplify, we ignore time discounting, which adds little of interest, so that the hire-fire ratios simultaneously determine the (un)employment rate, the share of life each worker spends (un)employed, and the relative weight in her intertemporal utility on consumption while (un)employed α as βi and α +iβ . With these ratios positively (negatively) related to αi + βi

(

i

i

)

income, mean-preserving increases in wage skew—that is, decreases in median voter income—now also entail increases in unemployment and in the median’s unemployment risk. The main implication is a flattening of the upward-sloping curve relating wi to insurance spending in figure 8.1 because the income effect, which had operated alone and had induced the median’s desired insurance to decline as she grew poorer, is now offset by a substitution effect as her unemployment risk rises in tandem. Assuming α β+i β > αi so that the median is employed the majori

i

αi + βi

ity of her life, this flattening does not switch the sign of the slope, nor does it change the fact that this curve will cross the (essentially unchanged) downward-sloping total spending curve at some w0 < w ¯ and continue upward, flatter still, thereafter. Thus, the equilibria as previously described remain qualitatively accurate, but the flattening of the curve probably also implies that w0 shifts rightward, expanding the range of income skews over which the simpler unconstrained results hold.

Incomplete Political Participation Correlated with Incomes and Unemployment Risks Next, consider that not everyone votes or, more generally, participates in politics equally or with equal effectiveness. Obviously, the relevant population with regard to democratic policy choice is the voting (more generally, the politically active) public, and in these models (see notes 9 and 14), the median voter (or effective participant) decides. Moreover, as is

240

Democracy, Inequality, and Representation

well established empirically (see, for example, Conway 1985; Harrop and Miller 1987; Verba, Nie, and Kim 1978; Wolfinger and Rosenstone 1980), the relatively wealthy have a higher propensity to vote than the relatively poor. Under rather general conditions (see Franzese 2002, chap. 2), these empirical regularities combine to imply that the median voter will be poorer (and closer to the median person) as voter participation increases.15 Referring back to figure 8.1, as participation declines, the median voter becomes a person with higher income than the population median, which is what is given on the figure’s x-axis. That is, the skew from median-voter to mean-person income is less than that from median-person to mean-person income. As participation rises, these ratios converge. Thus, both curves in figure 8.1 flatten as participation decreases. With unemployment risk also declining with income, the insurance spending curve also shifts downward as participation declines.16 As with income-correlated unemployment risk, declining participation also seems likely to shift w0, and the associated kinks in the relationships of skew to redistribution and insurance spending outward. We considered only voting, but other modes of participation—lobbying, campaign contributions, direct contact with representatives, and so on—also convey influence. Indeed, considering the minuscule probabilities that individual votes will alter election outcomes, these other modes are likely to be more influential than mere voting. Considering variation in effective political participation, however, only strengthens the empirical relevance of this discussion, for two reasons. First, as voting declines, the relative prevalence and influence of other participatory modes seems likely to increase. Second, socioeconomic status correlates even more strongly with extra-electoral participation—most obviously, sizably, and notoriously, in contributions—than voting: “Class differences in mobilization typically aggravate rather than mitigate the effects of class differences in political resources” (Rosenstone and Hansen 1993, 241; see also Verba, Nie, and Kim 1978; Verba, Schlozman, and Brady 1995). Therefore, not only does electoral representation of the poor and those at high risk of unemployment decline as turnout falls (see also Anderson and Beramendi, this volume), but the influence of extra-electoral participation rises and these disadvantaged groups are even less well represented there. Thus, voter turnout may adequately summarize effective political participation for present purposes. In sum, in the (democratic) polity, the relevant population for policymaking influence is the effectively politically active; in the economy—and specifically regarding the average median voter income to tax to fund social policy—the relevant population is instead the economically active (the employed); and regarding policy outlays, for instance, on social spending, the relevant population is the entire society of potential beneficiaries (divided into employed and unemployed camps by targeted policy tools if

Inequality and Unemployment

241

available). The society, economy, and polity together determine social policy outcomes, and this hints at the final and thorniest consideration.

The Endogeneity of Economy, Policy, and Politics Finally we come to the previously ignored gorilla in the room: endogeneity. The distributional and employment outcomes that are the key explanators in these political-economic models of social policy, w skew ≡ wi and u ≡ αα+β , are themselves affected by the redistribution and social insurance policies aimed at ameliorating them. Regarding incomes and income skew measured post-tax-and-transfer, obviously these social policies affect them as well as the other way around; such effects are the policies’ raisons d’être. Even with pre-tax-and-transfer measures, and even regarding (un)employment outcomes and risks, these social policies have important disincentive and distortionary effects. Indeed, these effects are the subject of much of welfare economics. Likewise, political participation should condition these relationships of economic outcomes to policies, yet it too is endogenous to the economic conditions whose effects on social policies we argue it moderates (see also Anderson and Beramendi, this volume). Empirical exploration of the earlier theoretical propositions therefore must somehow confront the endogeneity of economy, policy, and politics. With five endogenous outcomes across these three spheres in the present case, we follow a simultaneous system-ofequations approach.

The Empirical Model: A System of Equations for Inequality and Unemployment, Redistribution and Social Insurance, and Political Participation The theoretically suggested system of equations involves two economic conditions—U = unemployment and S = skew—two social policies—I = social insurance and R = redistribution—and P = political participation. In general, to identify a system of M simultaneous equations—here, M = 5: S = s(U,R,I,P,·,εS); U = u(S,R,I,P, ·,εS); R = r(S,U,I,P, ·,εR); I = i(S,U,R,P, ·,εI); P = p(S,U,R,I, ·,εP)

[8.19]

—we must “tie down” M(M − 1) terms—here, 5 × 4 = 20—by sufficient restrictions on the equations given by some extra-empirical information (Greene 2003, 378–95). We follow the most common strategy for provid-

242

Democracy, Inequality, and Representation

ing such identifying information: that of imposing exclusions, that is, making some variables from among these five endogenous ones or the other regressors (represented by · in [8.19]) excludable from some equations.17 Each excluded right-hand-side variable reduces the number of parameters to estimate by one per variable per equation. In words, exclusion assumptions or arguments are statement like: “Variable z affects one or some of the endogenous variables but does not affect others except insofar as it affects (causes) the first one or set.” If, for example, we could find five variables, one per equation, that enter, in this sense, only their one equation, then each would give four restrictions (namely, that the coefficient on each variable in the other four equations is zero), yielding the minimum 5 × 4 = 20 needed and just identifying the system. Finding more than the minimum additional outside information—that is, over-identifying the system—adds efficiency (“ties down the system more firmly”) and opens the possibility of testing over-identifying restrictions.18

Identification by Exclusions Among the Endogenous Variables We begin by considering which endogenous variables, S (skew), U (unemployment), R (redistribution), I (insurance), and P (participation), enter which others’ equations, starting with the economic outcomes, skew and unemployment. First, [8.14] would indicate that skew, not unemployment, enters the redistribution equation, and our elaborations modified this conclusion only by extending the empirically applicable range of [8.14]. We argue that unemployment does not enter our skew equation either, at least not strongly directly. The unemployed have zero wages, so U directly affects mean wages, skew’s theoretical denominator. This effect, however, is probably small: 10 percent unemployment, for example, lowers mean wages by just 0.1 times the (likely low) wages when working of the jobless. Furthermore, because percentiles like the median (fiftieth) are not directly affected by extremes above and below them, and because the jobless come mostly from the lower end of the wage and income distribution, we can evade much of even this small direct simultaneity by using percentile ratios instead of median-to-mean ratios to measure skew, especially if we use higher percentiles. That is, 90/50 ratios provide a stronger basis for some of the exclusion restrictions we intend to impose than the 90/10 ratios more commonly used. Since either summarizes the distribution with equal effectiveness, we use 90/50.19 Unemployment also adds labor-supply competitors and so affects wages at all percentiles, but this too should mostly affect lower percentiles that compete more directly with the jobless, giving further argument for the higher-percentile 90/50 ratio. U does, however, enter the remaining equations: participation because the unemployed tend to drop out not only from being economically

Inequality and Unemployment

243

active but also from being politically active; and social insurance, obviously, because the unemployed are its target. Just as obviously and centrally, skew enters both the redistribution and insurance equations, as seen in [8.14] and [8.15], and the participation equation, as discussed later in the chapter. Finally, we argue that the income distribution, and so skew, would not affect unemployment except through its effects on policies. Turning to the policy variables, redistribution clearly enters the skew equation, especially insofar as our S measure reflects the post-tax-andtransfer income distribution, such impact being the policies’ intent (and effect: see Atkinson, Rainwater, and Smeeding 1995; Danziger and Gottschalk 1995; Gottschalk and Smeeding 1997; Smeeding et al. 1990). Redistribution indirectly affects even pre-tax-and-transfer skew, however, and also unemployment, since it alters labor-market functioning, for example, by raising reservation wages. The same argument places insurance in the unemployment equation, but it would not enter our 90/50 skew since social insurance directly aids only the jobless at the lowest (wagezero) end of the distribution. As for the policy equations, redistribution and insurance correlate in [8.14] and [8.15] only because they both relate to skew,20 yet they should enter each other’s equation anyway owing to policy substitute or complement effects. Lastly, we might argue that redistribution and insurance should affect participation only as they affect recipients’ socioeconomic status (SES), that is, only via U and S here, but Sara Binzer Hobolt and Robert Klemmensen (2006) argue and find instead that social spending recipients—perhaps responding to a sense that policy regards, and so politics involve, them—do have a higher propensity to vote, even controlling for their post-tax-and-transfer SES. (Their supportive evidence does not address the endogeneity issues raised here, however.) The expected positive feedback in equality, unemployment, redistribution, social insurance, and participation raises the possibility of multiple political-economic equilibria, with two basins of attraction—one of high and one of low political participation, social insurance and redistributive spending, unemployment, and equality—consonant with Alberto Alesina and George-Marios Angeletos’s (2005a, 2005b) social policy multiple equilibria, but derived from political participation rather than societal preference. Political participation, finally, only directly affects the policy variables, R and W; an effect on economic outcomes that did not work through policy is hard to imagine. The remaining system is thus the following: S = s(R,·, εS); U = u(I,R, ·,εU); R = r(S,I,P, ·,εR); I = wi(U,S,R,P, ·,εI); P = p(U,S,R,I, ·,εP)

[8.19a]

244

Democracy, Inequality, and Representation

Identification by Exclusions Among the Exogenous Variables By [8.19a], we have reduced the twenty coefficients to identify by six exclusions to fourteen. We introduce next some potentially exogenous regressors and discuss their inclusions or exclusions before proceeding to specify how the regressors enter the estimation equations and then to consider estimation methods. These other regressors are variables related to demographics (D), for example, the age distribution; socioeconomic institutional and interest structure (SIS), for example, trade exposure and structure; domestic political institutions (DPI), for example, governmental and electoral systems; and current political contexts (CPC), for example, government partisanship and electoral competitiveness. Additionally, we might find further identification leverage in the international (spatial) interdependence of the economic- and policy-outcome dependent variables—that is, in economic conditions and policies abroad— which we write as U~i, S~i, R~i, and I~i. This further elaborates the system of equations to: S = s(R,S~i,D,SIS,DPI,CPC, εS); U = u(I,R,U~i,D,SIS,DPI,CPC, εU); R = r(S,I,P,R~i,D,SIS,DPI,CPC, εR); I = i(U,S,R,P,I~i,D,SIS,DPI,CPC, εI); P = p(U,S,R,I,D,SIS,DPI,CPC, εP)

[8.19b]

Spatial Interdependence Insofar as economic conditions are diffused across borders by trade and investment flows and, more broadly, international competition depends on economic conditions abroad, S~i and U~i can enter as regressors for the skew and unemployment equations that would enter other equations only through these domestic economic outcomes. Analogously, the economic policies of a nation’s competitors and partners affect the costs and benefits of its domestic policies (see, for example, Franzese and Hays 2006b, 2007b; Basinger and Hallerberg 2004), so R~i and I~i may enter the redistribution and insurance equations. (Cross-national dependence of mass participation, however, seems rather unlikely; see, for example, Kayser 2007.) As Robert Franzese and Jude Hays (2004, 2006a, 2007a, 2007b) explain and explore, such spatial-lag regressors entail their own endogeneity issues: if, for example, France affects Germany and Germany affects France, then the spatial lag, a weighted average of the dependent variable in the other (~i) units, is endogenous. This spatial-

Inequality and Unemployment

245

simultaneity bias may be small enough or redressed effectively enough by our time-lagging of it (an imperfect stratagem), and the identification leverage that these spatial lags offer upon the simultaneity of central interest here (outcome and regressor simultaneity within a spatial unit) may be large enough, to render usage of spatial lags as quasi-instruments advantageous (see Bartels 1991). Each quasi-instrumental spatial-lag enters one equation and so brings four exclusions—sixteen more in total, two more than the fourteen remaining to fulfill the necessary rank condition for identification of our system. However, the order condition—which is necessary and sufficient with the rank condition and which requires that the exclusions equaling or exceeding M(M – 1) (here, 20) are distributed across the equations such that each is tied down by at least one unique exogenous aspect of its specification—is not satisfied yet. The participation equation as yet lacks such a unique exogenous component and so is unidentified, whereas the insurance equation is just-identified, basically by its quasi-instrumental spatial lag, and the skew, unemployment, and redistribution equations are all over-identified, having both their own unique quasi-instrumental spatial lags and three, two, or one further exclusions, respectively, from among the endogenous variables. Furthermore, we would not want to rest identification of the system solely on quasi-instrumental variables, and we can find further leverage in some of the other exogenous regressors, to which we turn now.

Demographics Demographic variables (D), especially the age distribution, can provide some regressors of a more certain exogeneity.21 Unfortunately, however, most demographics relevant to one of the outcome variables would also affect most or all of the others. The over-sixty-five share of the population, Pop65, for instance, should impinge upon redistribution or insurance (insofar as those measures comprise public pensions and other agedependent spending, like health and child care), but age-demographics like these certainly affect employment and income distribution outcomes directly also (see, for example, Smeeding and Sullivan 1998). Age also has among the most robust and sizable known effects on voter participation, so Pop65 probably enters all the equations and thus, while exogenous, provides no identification leverage for any of them (unless we could determine that it enters them differently, which we cannot). The under-fourteen population share, Pop14, also causally relates to economic outcomes, surely unemployment and possibly skew. Again, redistributive or insurance spending clearly depend on Pop14 too—for example, education and related spending programs—but participation as a

246

Democracy, Inequality, and Representation

share of the eligible-age (over-fifteen) population might not. Our system is now: S = s ( R , S~i , Pop65, Pop14, SIS , DPI , CPC , ε S ) ; U = u ( I , R ,U ~i , Pop65, Pop14, SIS , DPI , CPC , εU) R = r ( S , I , P , R~i , Pop65, Pop14, SIS , DPI , CPC , ε R ) ; I = i (U , S , R , P , I ~i , Pop65, Pop14, SIS , DPI , CPC , ε I ) P = p (U , S , R , I , Pop65, SIS , DPI , CPC , ε P )

[8.19c]

Excluding Pop14 from the participation equation provides the missing unique exogenous aspect to that equation. Our system is now identified if we credit the quasi-instrumentality of the spatial lags. In [8.19c], we have begun to indicate our situation vis-à-vis identification by placing arcs over endogenous variables, double-underlining regressors that appear in only one equation—which suffice to identify that equation’s left-hand-side variable for inclusion on the right of other equations—and single-underlining regressors that do not appear in all equations, which provide leverage on those outcome variables from which they are excluded.

Socioeconomic Institutional and Interest Structures Next we consider socioeconomic institutions and interest structures, like unionization and corporatism, trade exposure and structure, stock market capitalization and outcomes, and female labor force participation. Unions work to enhance and protect members’ wages, and evidence that they affect the wage and income distributions is rife (see, for example, Freeman 1991). Union density, UDen, thus enters the skew equation (directly or indirectly as unionized sectors array at or near the percentiles used in the 90/50 skew measure). Then, largely as a consequence of this aim and effect, unionization exacerbates insider-outsider conflicts that can spur unemployment, so UDen enters the unemployment equation too. Strong unionization is also likely to represents effective political influence that favors both redistribution and insurance spending, and union members, in strongly empirically supported theory, have a greater propensity to participate politically. Thus, unfortunately, UDen provides no identification leverage (without further information or expectations about the shape of these relationships, which we do not have). Likewise corporatism, Corp, or the coordination or centralization of labor organization, centrally affects wage and employment outcomes, including skew

Inequality and Unemployment

247

and unemployment, and the balance of political influence on redistribution and insurance. However, Corp seems unlikely to affect voter participation beyond the positive effect already accounted for by unionization, except insofar as it affects these policies and outcomes, so we may perhaps exclude Corp from the participation equation. By the Stolper-Samuelson theorem and related international trade theories, trade exposure, TExp, and especially exposure to trade with developing (labor-rich, capital-scarce) economies, TExpD, should increase some combination of skew and unemployment.22 The unemployment impact arises if and insofar as some real-wage or -price inflexibility and unemployment exists (for example, due to monopoly union or firm power) rather than the Stolper-Samuelson assumed perfect competition and full employment.23 By similar reasoning, openness and, especially, trade with developing countries would also shape redistribution and insurance demand (see, for example, Cameron 1984; Katzenstein 1985; Rodrik 1998). However, insofar as these policy effects of trade occur because of the Stolper-Samuelson and related employment and wage effects, we can exclude these trade-structural variables from the policy outcomes. Likewise, trade structure may affect participation, if at all, only through these economic outcomes and policies. Most theory would cast international financial exposure, FinExp, in an identical role to trade exposure in these regards. Evelyne Huber and John Stephens (2001) argued that historical social democratic (SDG), Christian democratic (CDG), or secular conservative (SCG) governance—an aspect of socioeconomic structure (as opposed to incumbent-government partisanship, which is a “current political context”)—shapes the generosity and structure of social policy. Accordingly, these factors enter the policy equations, but not the outcome or participation equations, wherein their effects, if any, should arise only through the policies. In particular, Huber and Stephens argued, essentially, that SDG and CDG, but not SCG, policy legacies involve strong social insurance generosity. Conversely, neither the SCG nor the CDG legacies involve the generous general redistribution of SDG. Accordingly, we can capture these hypotheses succinctly by including just SCG in the insurance equation, expecting a negative coefficient, and just SDG in the redistribution equation, expecting a positive one.24 Next, female labor force participation, FLFP, may operate analogously to demography in shaping skew and unemployment and also may spur both demand for and supply of insurance and redistribution. Indeed, the type of spending response to FLFP probably depends on whether a historical legacy of SDG predominates, because both the CDG and SCG legacies have supported FLFP far less (Huber and Stephens 2001).25 We suspect also that FLFP—or more exactly, its societal and other roots—may

248

Democracy, Inequality, and Representation

affect participation as well; in other words, what favors female labor force participation probably favors female electoral participation as well. Its interaction with SDG, however, is unlikely to affect participation except through its effects on policy. Finally, stock market capitalization (SMC), stock market returns, or percentage increases in indices (SMR), and their interaction (SMC·SMR) seem likely to affect income skew in a “yachts outpacing tugboats and rowboats” phenomenon. Such outpacing is likely to be proportionate to stock market (and other investment) returns and should have a greater effect on income skew the greater the capitalization is (actually, ideally, domestic stock ownership prevalence). Less directly, and more speculatively, stock market capitalization, reflecting an emphasis on a particular form of corporate finance, may have implications for wages and employment (Hall and Soskice 2001). This probably has the obvious implications for interest and political-influence distributions as well, so we suspect that stock market capitalization enters the policy equations as well. Current returns and the interaction should be less relevant (directly) to these policy variables, however, and such financial market terms seem unlikely to affect participation. This brings the specification of the five-equation system to this state: S = s ( R , S~i , Pop65, Pop14,UDen, Corp, FinExp, TExp, TExpD , FLFP , SMC , SMR , SMC ⋅ SMR , DPI , CPC , ε s ) ; U = u ( I , R ,U ~i , Pop65, Pop14,UDen, Corp, FinExp, TExp, TExpD , FLFP , SMC , SMR , SMC ⋅ SMR , DPI , CPC , εU ) ; R = r ( S , I , P , R~i , Pop65, Pop14,UDen, Corp, FLFP , SDG , FLFP ⋅ SDG , SMC , DPI , CPC , ε R ) ; I = i (U , S , R , P , I ~i , Pop65, Pop14,UDen, Corp, FLFP , SCG , FLFP ⋅ SDG , SMC , DPI , CPC , ε I ) ; P = p (U , S , R , I , Pop65,UDen, FLFP , DPI , CPC , ε P )

[8.19d]

Domestic Political Institutions Two important domestic political institutions here are (the natural log of) district magnitude, DMag, and presidentialism, Pres, which a long line of political science research and some recent, influential political economy formalizations (for textbook exposition, see Persson and Tabellini 2000) have connected to redistributive and distributive policies. These theories expect greater redistributive and less targeted spending in more proportional (larger district-magnitude) systems and more total public spending

Inequality and Unemployment

249

in parliamentary than in presidential systems. Further, district magnitude and other electoral-law features, notably the onerousness of registration requirements, RegReq, and mandatory voting (abstention fines), MandVote, play theoretically long-noted and empirically well-established large roles in determining participation. Presidentialism may also affect participation—negatively if presidential and other elections stagger, thereby diffusing relevant policymaking authority across multiple elections and reducing the importance of each, and positively or with no effect otherwise. Other institutions that diffuse authority across elections (EleDiff), like bicameralism, federalism, or frequent referenda, should have similar effects. These institutions are likely also to affect policy but to affect unemployment or skew solely through policy. Likewise, the degree of intraparty competition, IPC, which plurality, majority, and especially transferable-vote systems strengthen, may affect participation, and policy is unlikely to affect economic outcomes except in that manner. Specifically, IPC may foster participation if that competition inspires voters, but IPC is more likely to dampen turnout because it weakens party discipline. (Party discipline eases voters’ electoral information burdens, so its weakening would tend to depress turnout.) Regarding policy, IPC favors targeted (insurance) over broader (redistribution) tools, as with DMag (see, for example, Ariga 2007; Cox and Rosenbluth 1995; Shugart and Carey 1992). Mandatory voting and registration burdens, finally, should affect policy only by affecting participation. These DPI, especially the electorallaw features, provide crucial identification leverage on participation, which had been relatively lacking heretofore: S = ds ( R , S~i , Pop65, Pop14,UDen, Corp, FinExp, TExp, TExpD , FLFP , SMC , SMR , SMC ⋅ SMR , CPC , ε S ) ; U = u ( I , R ,U ~i , Pop65, Pop14,UDen, Corp, FinExp, TExp, TExpD , FLFP , SMC , SMR , SMC ⋅ SMR , CPC , εU ) ; R = r ( S , I , P , R~i , Pop65, Pop14,UDen, Corp, FLFP , SDG , FLFP ⋅ SDG, SMC , Pres, DMag, IPC , EleDiff ,CPC , ε R ) ; I = i (U , S , R , P , I ~i , Pop65, Pop14,UDen, Corp, FLFP , SCG , FLFP ⋅ SDG, SMC , Pres, DMag, IPC , EleDiff , CPC , ε I ) ; P = p (U , S , R , I , Pop65,UDen, FLFP , Pres, DMag, IPC , EleDiff , MandVote, RegReq, CPC , ε P )

[8.19e]

Current Political Contexts Lastly, under current political contexts, CPC, we consider incumbentgovernment characteristics—majority status (GMaj), fragmentation

250

Democracy, Inequality, and Representation

(GFrag), polarization (GPol), and partisanship along a left-right axis (GPart) or by SDG, CDG, or SCG classifications—and current-electoral conditions, like election indicators, E, and competitiveness, Comp. We gauge currentgovernment ideology by CDG and GPart, which allows policy to relate to ideology linearly, roughly curvilinearly, or linearly but with Christian democracy lying off the direct line from left (SDG) to right (SCG) ideologically (see, for example, Swank 2002). All of these CPC factors should affect the policy variables directly; only competitiveness should affect participation directly, and none should affect the economic outcomes except through policy and participation effects.26 This gives us the following (penultimate) specification of our system: S = s ( R , S~i , Pop65, Pop14,UDen, Corp, FinExp, TExp, TExpD , FLFP , SMC , SMR , SMC ⋅ SMR , ε S ) ; U = u ( I , R ,U ~i , Pop65, Pop14,UDen, Corp, FinExp, TExp, TExpD , FLFP , SMC , SMR , SMC ⋅ SMR , εU ) ; R = r ( S , I , P , R~i , Pop65, Pop14,UDen, Corp, FLFP , SDG , FLFP ⋅ SDG , SMC , Pres, DMag, IPC , EleDiff , GMaj , GFrag, GPol , GPart , CDG , E , Comp, E ⋅ Comp, ε R ) ; I = i (U , S , R , P , I ~i , Pop65, Pop14,UDen, Corp, FLFP , SCG , FLFP ⋅ SDG , SMC , Pres, DMag, IPC , EleDiff , GMaj , GFrag, GPol , GPart , CDG , E , Comp, E ⋅ Comp, ε I ) ; P = p (U , S , R , I , Pop65,UDen, FLFP , Pres, DMag, IPC , EleDiff , MandVote, RegReq, Comp, ε P )

[8.19f]

Notice that theory and substance allow us to offer empirical models that strongly distinguish (identify) the three outcome types: economy, policy, and politics. Skew and unemployment alone have the trade- and financialexposure and stock-return regressors, which should be strong exogenous explanators of them. The economic outcomes also exclude participation, whereas the policy ones do not. Similarly, redistribution and insurance have unique exogenous regressors in the CPC factors (with the exception of competitiveness), which should strongly predict policies, and participation uniquely responds to two electoral-law features known to predict it well. However, the two economic and two policy outcomes seem less sharply distinguished from each other. The economic outcomes are distinguished by their unique spatial lags and by the fact that insurance spending affects unemployment but not skew. Redistribution and insurance policies are distin-

Inequality and Unemployment

251

guished by their unique spatial lags, in that unemployment affects I but not R spending, and by the fact that SCG and CDG legacies similarly (negatively) affect R whereas CDG and SDG legacies similarly (positively) affect I. These perhaps weaker distinctions may nonetheless offer sufficient empirical leverage, because the spatial lags probably have strong explanatory bite for the economic outcomes, while the importance of partisan historical legacy may adequately compensate for what might prove weaker interdependence among nations’ social policies. Moreover, further useful distinctions arise in specifying precisely how each factor enters each function.27

Empirical-Model Specification, Data, Estimation, and Results We now describe these functions more fully. Our theories generally lack the precision to suggest specific functional forms, so we assume the usual linear additivity here, accepting “best linear approximations” to what are likely nonlinear relationships.28 Earlier, however, we have suggested certain interactions (we assume them to be linear interactions) among some of the regressors: between stock market capitalization and returns in the economic-outcome equations, and between FLFP and SDG and between elections and competitiveness in the policy equations. As we convert [8.19] to specific regression models and add dynamics to those models, we add two more interactions. As argued earlier (see also Franzese 2002, ch. 2), participation interacts with skew to shape the effective political demand for social policies, and then government fragmentation, polarization, and majority status interact with the lagged dependent variables in the policy equations to reflect veto-actor arguments of policy adjustment retardation (see Franzese 2002, chap. 3). These interactions and the dynamics add several further over-identifying exclusions to the system. Thus, in the end, the specific equations with which we would ideally like to begin our empirical explorations and evaluations are these:

S = α i0 + α1 St −1 + α 2 S~i + α 3 R + α 4 Pop65 + α 5 Pop14 + α 6UDen + α 7Corp + α 8 FinExp + α 9 TExp + α10 TExpD + α11FLFP + α12 SMC + α13 SMR [8.20a] + α14 SMC ⋅ SMR + ε S

U = βi0 + β1U t −1 _ β2U ~1 + β 3I + β 4 R + β5Pop65 + β6 Pop14 + β 7UDen + β8Corp + β9 FinExp + β10 TExp + β11TExpD + β12 FLFP + β13 SMC + β14 SMR + β15 SMC ⋅ SMR + εU

[8.20b]

252

Democracy, Inequality, and Representation

R = γ i0 + γ 1 Rt −1 + γ 2 GFrag ⋅ Rt −1 + γ 3GPol ⋅ Rt −1 + γ 4 R~i + γ 5 S + γ 6 P + γ 7 S ⋅ P + γ 8 I + γ 9Pop65 + γ 10 Pop14 + γ 11UDen + γ 12 Corp + γ 13FLFP + γ 14 SDG + γ 15 FLFP ⋅ SDG + γ 16 SMC + γ 17 Pres + γ 18 DMag + γ 19 IPC + γ 20 EleDiff + γ 21GFrag + γ 22 GPol + γ 23GPart + γ 24 CDG + γ 25 E + γ 26Comp + γ 27 E ⋅ Comp + ε R

[8.20c]

I = φi0 + φ1 I t −1 + φ2 GFrag ⋅ I t −1 + φ3GPol ⋅ I t −1 + φ 4 I ~i + φ5U + φ6 S + φ 7 P + φ8 S ⋅ P + φ9 R + φ10Pop65 + φ11 Pop14 + φ12UDen + φ13Corp + φ14 FLFP + φ15 SDG + φ16 FLFP ⋅ SDG + φ17 SCG + φ18 SMC + φ19 Pres + φ20 DMag + φ21 IPC + φ22 EleDiff + φ23GFrag + φ24 GPol + φ25GPart + φ26 CDG + φ27 E + φ28Comp + φ29 E ⋅ Comp + ε I

[8.20d]

P = ω 0 + ω1 Pt −1 + ω 2U + ω 3 S + ω 4 R + ω 5 I + ω 6Pop65 + ω 7UDen + ω 8FLFP + ω 9 Pres + ω10 PProx + ω11DMag + ω12 IPC + ω13 EleDiff + ω14 MandVote + ω15 RegReq + ω16Comp + ε P

[8.20e]

Notice first that we gain additional identification leverage on each equation if we assume or can establish theoretically or substantively that the time-predetermined nature of the lags modeling the temporal dynamics suffices to ensure exogeneity. We do assume so, although without full confidence, especially for the very slow-moving skew, because the set of estimated coefficients and the stationarity for the outcomes that they imply seem much more plausible with the assumption.29 Note next how the policy equations draw from veto-actor theory (Tsebelis 2002) to derive more specification precision and identification leverage by allowing current-government fragmentation and polarization to modify policy adjustment rates (Franzese 2002, chap. 3), with the expectation being retardation. (We also measure fragmentation and polarization to reflect how the current government’s majority status relates GFrag and GPol to policy retardation, and so GMaj need not enter directly as a regressor.30) These interactions in the dynamics provide further identification leverage for the policy equations.31 Next, notice the i superscripts on

Inequality and Unemployment

253

the constants of the economic and policy equations but not the participation equations, indicating the country fixed effects used or omitted. Again, coefficient estimates and the implied stationarity of the outcomes seemed more empirically sensible allowing these fixed effects. The participation equation excludes them, however, because several key explanators in that model—including some of core substantive interest on which we rely heavily for identification, like MandVote and RegReq—hardly vary over time.32 (We do, however, allow the country dummies as instruments in all models, as discussed later.) Finally, notice that regressors in [8.20] involving trade exposure to developing countries, TExpD, or electoral competitiveness, Comp, are grayed; unfortunately, this signifies that we have not yet found measures of sufficient cross-country-time coverage to include them in our current estimation models.33

Data and Measurement For empirical estimation of the system [8.20], we assembled (building from the work, and thanks to the generosity of others) a database covering twenty-three developed democracies over forty-four years, 1960 to 2003, although much less than that has ultimately proven usable (without extensive imputation).34 Skew is usually the limiting reagent, as inequality data are notoriously spotty, despite that the Luxembourg Income Study (LIS) (Atkinson, Rainwater, and Smeeding 1995; Smeeding et al. 1990) and OECD efforts having improved matters greatly. We use data on earnings by population decile to construct ratios of the ninetieth to the fiftieth deciles’ incomes, as discussed earlier.35 By linear interpolation of a few missing country-years (28 of the 360 total that we assemble),36 we obtain unbroken annual series of at least some years for nineteen of the twenty-three countries.37 St – 1 is the one-year time lag, and S~i, as with all the spatial lags, is the unweighted average of that variable, that year, in the other data-set countries. Standardized unemployment rates, U, are from the OECD via the “Comparative Political Data Set, 1960–2003,” of Klaus Armingeon and his colleagues (2005; henceforth ALMP). Where possible, we expand coverage of these data (from 660 to 939 country-years) by country-specific linear regression on the unstandardized rates. The R-squared of these fitting models always exceeds .9 (implying that standardized rates relate to unstandardized ones within a country by a relatively fixed factor). Following M&W, insurance spending, I, is OECD social benefits, excluding the health and pension categories. (This leaves primarily unemployment and welfare benefits.) Redistributive spending, R, is total spending minus social benefits and military spending. (R and I thus both exclude health and pensions.) We extend these data (as generously pro-

254

Democracy, Inequality, and Representation

vided to us by Moene and Wallerstein) from 1980 to 1999 in eighteen countries to cover twenty countries38 and the period 1960 to 2002 in most of them (from about 330 observations to about 775 observations) by regressing M&W’s original data on that from the current OECD social benefits and public health data sets, plus current disbursements (total spending). This effort to recreate M&W’s procedure for the current, expanded-coverage data set yielded R-squareds of the fitting models usually exceeding .9 and often approaching 1.0, indicating near-perfect replication. We measure participation, P, as voters’ percentage of the eligible-age population, smoothing ALMP’s vturn measure of that percentage, which holds participation constant at the last election’s rate until the next, by averaging the current, previous, and next two years (capturing exactly one election cycle per window in most cases). The age demographics, Pop65 and Pop14, are from the OECD via ALMP. Union density, UDen, is active members (excluding the unemployed and retired) as a share of employment, taken from Bernhard Ebbinghaus and Jelle Visser (2000) via ALMP and from Miriam Golden, Peter Lange, and Micahel Wallerstein (1997) and extended by regression on gross membership and the age demographics. The corporatism index, Corp, is Lane Kenworthy’s (2001). Trade exposure, TExp, is from OECD sources via ALMP, and international financial exposure, FinExp, is the sum of current and capital account openness from Dennis Quinn and Carla Inclan (1997). Female labor force participation rates, FLFP, are taken from Huber et al. (2004). We take our measures for stock market capitalization, SMC, and returns, SMR, from the Global Financial Database,39 using all-market December 31 closing values divided by nominal GDP in domestic currency for SMC and its year-on-year percentage change for SMR. Our indices for government fragmentation, GFrag, polarization, GPol, and partisanship, GPart, derive from Thomas Cusack’s rich, thorough, and usefully designed “Parties, Governments, and Legislatures Data Set.”40 Using GSppt, the percentage legislative seat-share of the governing (cabinet) parties,41 we obtain GFrag for majority governments (GSppt > 1⁄2) as the raw number of governing parties (counting nonpartisans as half a party). Raw numbers are a more appropriate representation of Tsebelis’s veto-actor conception of fragmentation than effective (sizeweighted) numbers (see Franzese 2002, chap. 3) because, in that theory, any governing party, regardless of size, can veto policy change since its presence in government indicates its necessity to that coalition. If GSppt < 1/2 (minority government), GFrag is a GSppt-weighted count of the raw number of governing parties and the effective number of opposition parties. A minority coalition need not add all other parties to build a major-

Inequality and Unemployment

255

ity to change policy, so using raw numbers of opposition parties would lead to exaggeration. Short of analyzing each parliamentary context at length, we construct a convenient proxy for the number of veto-acting opposition parties by weighing their counts by size (using effective numbers), reflecting the notion that larger parties are more likely to be necessary partners in building legislative majorities. GPol likewise adopts a veto-actor conception of polarization, using party ideological ranges (size-unweighted) rather than standard deviations or variances (sizeweighted). GPol measures leftmost to rightmost governing party if GSppt > 1⁄2 and across the whole legislature in the case of minority government.42 To generate GPart, we use Cusack’s processing of the Comparative Manifesto Project (CMP) data set into left-right scores for parties and then of those party left-right scores into average left-right scores for parliaments and for cabinets. For GSppt > 1⁄2, we use the cabinet’s score directly, and for GSppt < 1⁄2 we use the GSppt-weighted average of the cabinet and the legislature. We derive our measures of cumulative social democratic and secular conservative government, SDG and SCG, and of current Christian democratic government (which sums center and right, Christian and Catholic), CDG, from Huber and colleagues’ (2004) data on cabinet and legislative seat-shares, using Cusack’s GSppt to enhance those measures (that is, for minority cabinets the parties’ government seat-shares are the GSpptweighted average of the cabinet and legislative shares). “Cumulative” refers to the sum from 1960. Pres, from Matt Golder (2005), equals 1 in presidential systems (Switzerland and the United States), 0.5 in semi-presidential systems (Finland, France, Iceland, and Portugal), and 0 in parliamentary systems. Unfortunately, Pres does not vary over time within-country in our sample. The natural log of (average) district magnitude (the number of representatives divided by the number of districts), DMag, is from Golder too, and it varies a little within-country. Our intra-party competition index, IPC, is crude, as it merely sums indicators for plurality, majority, and transferable-vote electoral systems, again from Golder (2005). We code the German mixed system and the similar new ones in Italy, Japan, and New Zealand as 0.5, reflecting their part-plurality nature (although in Japan the other part is transferable-vote, so IPC is 1). Our measure of authority diffusion across elections, EleDiff, adds “effective federalism” and “provincial-election importance” measures from the World Bank Database of Political Institutions (Beck et al. 2001) to Arend Lijphart’s (1999) prevalent-referenda (Switzerland) and effective-bicameralism indicators. EleDiff varies only slightly within countries. Our pre-electionyear indicator, E, allocates sums of 1 to the 365 days before lower-house elections (from ALMP).43 The participation model also includes a mea-

256

Democracy, Inequality, and Representation

sure of proximity of presidential to parliamentary elections, Prox1 (from Golder 2005). Finally, MandVote and RegReq, our measures of compulsory voting and registration-requirement burdens, derive from our own analysis of electoral-system data from the International Institute for Democracy and Electoral Assistance.44 MandVote varies from 0 to 1, according to the degree of enforcement indicated (none = 0, weak = .5, strong = 1) times the severity of punishments (none = 0, nominal fine or other weak sanction = .5, “appreciable fine” = 1) times the share of the country’s provinces in which the law is in force. RegReq simply indicates (0,1) whether a national voter registry exists or whether voters must self-register.45 Neither index varies over time within-country.46

Estimation Strategies Estimation strategies for systems of equations are numerous and variegated, as are those for time-series cross-sections (TSCS), so the number and variety of combinations that may be appropriate to estimate our system (of five equations, from data in eighteen countries over, on average, seventeen to eighteen years) are multiplicatively great. We must consider whether to allow unexplained cross-country differences in conditional means as fixed (or random) effects, recognizing that failing to do so when heterogeneity (conditional on the rest of the model) exists can bias estimation (if these omitted conditional-mean differences correlate with included regressors) and induce inefficiency. Conversely, regardless of whether heterogeneity exists, fixed effects prevent the direct recovery of the effects of any time-invariant explanators and can severely compromise estimation and complicate or obfuscate interpretation of the effects of slowly or rarely moving regressors (see Plümper and Tröger 2007), and random effects rely on questionable assumptions in aggregate TSCS contexts. We must consider also whether to add any other variables from among the regressors to our list of endogenous variables. Failing to acknowledge the endogeneity of some regressors will bias results, but treating variables as endogenous that are not (or that are not too importantly so) adds to the empirical identification burden of the remaining exogenous variables and to the researcher’s difficulties finding viable instruments. Then we should consider also whether the TSCS data structure might add other exogenous factors beyond the current set to the instruments, thereby gaining further crucial identification and estimation leverage if the additional conditions are true but inducing otherwise avoidable bias if they are not. Lastly, we should consider whether and how to use information in the data about cross-equation relations (like error covariances) or instead to estimate the equations separately. Esti-

Inequality and Unemployment

257

mating jointly can enhance efficiency notably; estimating separately forsakes these gains but isolates each equation’s estimation from any specification or other problems in the other equations. All these considerations are in addition to alternative plausible theoretical specifications. We have explored a range of options so far:



Estimating the system’s equations jointly (3SLS), separately (2SLS), or exogenously (SUR)



Including or excluding fixed effects; deciding to include country dummies for the first four but not for the fifth equation (that is, for policies and outcomes but not for participation, which contains several substantively interesting regressors that move very slowly or rarely)



Including or excluding among the system’s instruments these country dummies, a full set of year dummies, or both; choosing to include both country and year dummies as instruments



Deciding which regressors besides our five outcomes to consider endogenous; settling upon the interaction of skew and participation as the only one, its endogeneity being most crucial substantively47



Exploring and reconsidering several theory-derived specification choices discussed earlier: 1.

Adding insurance spending to the skew equation

2.

Adding to the policy equations interactions of participation times unemployment

3.

Adding to the policy equations interactions of GMaj or GSppt times the lagged dependent variable

4.

Deciding whether to treat the temporal and spatial lags as exogenous

5.

Deciding whether to add a control for real GDP growth and, if so, whether to treat it as endogenous48

With one major exception, coefficient estimates were remarkably consistent across all combinations of these considerations and options. Without the country dummies, the coefficients on skew, participation, and their product in the policy equations reverse signs to negative, negative, positive. (Sven Wilson and Daniel Butler [2007] noted a similar reversal of M&W results with fixed effects.) Either pattern only partially supports

258

Democracy, Inequality, and Representation

the M&W model, regardless of whether skew increases are bottomdragged or top-led; the extension regarding varying participation likewise receives mixed support. None of the other estimation-strategy options affect any of the coefficient estimates nearly so much. Accordingly, since systems estimation with heterogeneous intercepts and more instruments is far more efficient, we report the model estimated by (iterated) three-stage least squares (3SLS), with country dummies on the righthand side of [8.20a] to [8.20d] and both country and year dummies in the instrument list.

Estimation Results Table 8.1 presents these estimated coefficients (in bold) and standard errors (beneath them), with entries significant or nearly so in italics. We omit the country fixed-effect estimates to conserve space (but they are available on request, as are all replication codes and data). The results contain strong support for some aspects of previous theory, our own additions, or conventional wisdom, but also many notable surprises. We first briefly survey the estimated relationships of the exogenous explanators to the outcomes of our system before turning to our central interest in the estimated endogenous relationships among the outcomes. In confirmatory results, we find slow temporal adjustment rates for all five outcomes and strong spatial interdependence for economic policies and outcomes. We find that corporatism reduces income skew; union density may do so too, but it also, more clearly, supports turnout. Smaller youth and pensioner populations (that is, a larger working-age population) boost unemployment and inequality, while Pop65 also reduces turnout. Financial (but not trade) exposure seems to increase inequality, trade (but perhaps not financial) exposure seems to raise unemployment, and greater stock market capitalization induces less redistribution and social insurance. We find that historically having a social democratic government interacts positively with female labor force participation to expand both social policies and the strong, intuitive effects on it of current-government partisanship, with little sign that Christian democrats lie somewhere off that left-right line. Most impressive, however, is how well political science theories can explain turnout variation. Authority diffusion across elections, electoral systems that foster intra-party competition (SMD and Limited Vote), nonconcurrent presidential elections, and onerous registration requirements all depress turnout, and of course, mandatory-voting laws increase it. In more mixed results, we find support for veto-actor retardation of social insurance policy adjustment rates via government fragmentation, but our polarization measure does less well, and redistribution policy ad-

Inequality and Unemployment

259

justment rates seem impervious to both. The relations of corporatism and union density to unemployment have their expected signs, but significance is marginal or lacking. The negative relation of district magnitude to both social insurance and redistributive spending is unexpected given the emerging consensus that these electoral systems favor broadly targeted over narrowly targeted public spending, but these results probably also imply a positive association with the excluded health and pension spending and so could be read as refining and confirming that consensus. Finally, just as policy under Christian democratic governments is not significantly distinguishable from that of secular parties similarly positioned on the left-right axis, cumulative secular conservative government is not strongly distinct from historical CDG in social insurance policies. We find several null or contradictory findings also. Stock market capitalization seems not to affect 90/50 skew and may even reduce unemployment, and market returns fail to register at all with unemployment or skew. We suspect, however, that the signal-to-noise ratio in our returns measure, being a closing price for a single day (and an odd one, December 31), may be low and that capitalization is proxying for GDP here. That age demographics and UDen fail to affect social policies, and that corporatism has a marginally significant negative association with them, are surprising findings but perhaps are best seen as plain null results. Similar null findings emerge for FLFP and DMag in the participation equation (but note that IPC already draws the PR/SMD distinction) and electoral cycles in the social policy equations. Electoral diffusion and intra-party competition have very little and almost zero within-country variation, respectively, so their failure to register strongly in fixed-effect policy regressions is unsurprising. Finally, the insignificant positive relation of FLFP to skew is surprising, and its highly significant negative relation to unemployment quite so, given the intuitive relationships of the working-age population share with S and U. However, we might best credit chance for this one variable of the seventy-six just discussed being significantly opposite of expectations. Our central interests surround the causal relationships among the endogenous variables of our system. Figure 8.2 summarizes the statistically significant of these effects. To assist comparisons, it gives standardized (beta) coefficients, with one (two, zero) asterisk(s) indicating significance at the .10 (.05, .15) level. We start with the determinants of income skew and unemployment. In specifying the system, we stressed that these social policies could affect these economic outcomes, redistribution the former and both the latter. Empirically, we find causal effects of both social policies on unemployment. As critics allege, social policy generosity does seem to undermine labor market performance, by raising unemployment. Sensibly, the size and statistical significance of the effect from social insur-

260 Table 8.1

Democracy, Inequality, and Representation Empirical System of Skew, Unemployment, Redistribution, Social Insurance, and Participation: Estimation Results

Time lag GFrag × time lag GPol × time lag Spatial lag

Skew

Unemployment

Redistribution

0.8196 0.0361 —

0.6364 0.0318 —

0.9151 0.0322 0.0031 0.0119 –0.0002 0.0005 0.1275 0.0575 —





U

0.0809 0.0308 —

0.2069 0.0450 —

S





P





S×P





R I Pop65 Pop14 UDen Corp FinExp TExp SMC SMR SMC × SMR

0.0002 0.0012 — –0.0048 0.0032 –0.0039 0.0020 –0.0005 0.0005 –0.0040 0.0019 0.0027 0.0016 –0.0003 0.0003 0.0000 0.0001 –0.0001 0.0001 0.0000 0.0000

0.0495 0.0331 0.3950 0.0470 –0.1111 0.0835 –0.0907 0.0551 0.0025 0.0131 –0.0371 0.0502 0.0323 0.0409 0.0381 0.0094 –0.0152 0.0025 0.0007 0.0026 0.0001 0.0001

12.1539 3.3458 0.2145 0.0871 –0.1201 0.0461 — –0.1077 0.0469 0.0060 0.0706 0.0078 0.0470 0.0049 0.0133 –0.0774 0.0441 —

Social Insurance

Political Participation

0.8062 0.0484 0.0212 0.0110 –0.0002 0.0006 0.2232 0.0595 –0.0577 0.0377 8.4677 3.2131 0.2080 0.0842 –0.1069 0.0433 0.0143 0.0230 —

0.9269 0.0208 —

–0.0022 0.0685 –0.0340 0.0395 0.0074 0.0115 –0.0508 0.0347 —





–0.0092 0.0019 —

–0.0084 0.0017 —





— — — –0.0421 0.0562 1.9459 1.2057 — — 0.0918 0.0468 0.0571 0.0796 –0.1585 0.0747 — 0.0301 0.0108 — — — — — —

Inequality and Unemployment Table 8.1

261

Continued Social Insurance

Political Participation –0.0147 0.0133 —



–0.0676 0.0438 –0.1801 0.3157 –0.0037 0.0998 0.0643 0.1746 0.0012 0.0069 –0.0096 0.0028 0.2439 0.4105 –0.0016 0.0821 —

–0.0320 0.0133 –0.1172 0.0761 0.0026 0.0011 0.0115 0.0171 –0.1092 0.0328 –0.1561 0.2442 –0.0471 0.0863 –0.1366 0.1059 0.0014 0.0049 –0.0056 0.0024 0.0084 0.3122 0.0255 0.0620 —









MandVote









RegReq









311 – 30 = 281 .9508

311 – 40 = 271 .9835

Skew

Unemployment

Redistribution

0.0002 0.0005 —

–0.1001 0.0166 —

CumSDG × FLFP





CumSCG





0.0011 0.0153 –0.1348 0.0888 0.0026 0.0013 —

ln(DMag)





IPC





EleDiff





GFrag





GPol





GPart





CurrCDG





E





Pres



PresProx

FLFP CumSDG

Number of observations – number of coefficients = 0Free 311 – 29 = 282 R-squared .9797

— — –0.0072 0.0763 –1.1866 0.5831 –0.1322 0.0575 — — — — — –1.5602 0.6080 2.3025 0.4866 1.8808 0.5580 –1.3230 0.6651

311 – 311 – 42 = 26915 = 296 .9860 .9891

Source: Authors’ calculations. Note: Equations estimated with country fixed effects (omitted) simultaneously by iterated 3SLS, with S × P in addition to the five dependent variables treated as endogenous and with year and country fixed effects in addition to all other regressors treated as instruments. Estimated coefficients are in bold, with standard errors underneath. Entries significant or nearly so in italics.

262 Figure 8.2

Democracy, Inequality, and Representation Causal Relationships Among the Endogenous Variables .31** Redistribution

Skew

−.09**

.13* .03**

.03

Participation .06 .22**

.20**

.39** Social Insurance

Unemployment −.06

Source: Authors’ calculations. Note: The numbers are standardized coefficients. The conditional coefficients for skew and participation are calculated assuming low participation (39.4 percent) and low skew (1.44), respectively. *p < .10; **p < .05; coefficients without asterisks are marginally significant at p < .15

ance to unemployment are much greater than those of redistribution; benefits targeted to those lacking income affect individuals’ decision to work more strongly than do benefits not contingent on employment status. We find scant evidence that redistribution affects pre-tax-and-transfer 90/50 ratios. Recall, however, that we intentionally avoided the direct effects on post-tax-and-transfer skew by using pretax measures and the indirect effects on pretax income for individuals at the lower end of the distribution by using the fiftieth percentile denominator. We turn next to the policy variables. The policy effects of income skew and participation are conditional, as expected, but the nature of that interaction is unexpected. We find that higher participation attenuates the relationship between skew and policy generosity, and concomitantly that the relationship of participation to social policy generosity flattens with greater skew.49 Nonetheless, at low political participation and income skew, either variable has a significantly positive effect on social policy generosity, consistent with our theoretical expectations for median-preserving increases in income skew. The standardized coefficients of figure 8.2, which assume low participation (39.4 percent) and skew (1.44), suggest that participation matters more for social insurance than for broad redistribution, whereas a one standard deviation increase in skew explains a greater proportion of the sample variance in redistribution than

Inequality and Unemployment

263

in welfare and insurance policy. We also find some evidence of policy substitution from social insurance to redistribution, but not in the other direction. This suggests that increases in insurance are funded to some extent by cuts in redistribution, but increases in redistribution do not induce the converse reductions in social insurance, implying that funding derives to a greater extent from cuts made elsewhere or from revenue increases. Surprisingly, we also find that increases in unemployment may decrease social insurance spending; the substantive and statistical significance is only marginal. Finally, we consider the determinants of political participation. With respect to economic conditions, we find no evidence of negative relationships of unemployment or skew with participation, but we do find that income skew has a small, positive, and marginally significant effect on participation. This might reflect the high returns from redistribution for those at the lower end of the income distribution, but again, these effects are only of marginal substantive and statistical significance. The evidence for direct policy effects on political participation is stronger, supporting Sarah Hobolt and Robert Klemmensen (2006) in that both redistribution and insurance spending bolster participation, although the insurance spending effect is not significant. To calculate and present the implications of the estimated system in terms of the outcomes’ responses to substantively interesting counterfactuals is difficult; nevertheless, we can offer two sorts of counterfactuals.50 In the first, we set the levels of S, U, R, I, and P and of the explanatory variable of interest in our example, FinExp, to hypothetical starting values (their 1967 sample averages in our case) and then compute the response of S, U, R, I, and P to those values based on the estimated coefficients from table 8.1.51 Then we repeat the calculation except with a different value for the variable of interest. The difference between these two calculations (each a vector of five outcomes) is approximately the effect on those five outcomes of the hypothetical.52 We then approximate the long-run impact of a permanent change simply by multiplying the result for each outcome by its long-run multiplier, 1/(1 – ρ), with ρ being the time-lag coefficient. Thus we calculate the response of skew, unemployment, redistribution, social insurance, and participation to an increase in FinExp of 4.5 points, its sample-average trend from 1960 to 2000, with starting values corresponding to sample averages in 1967 (the nearest available to that starting point) of S ≈ 1.77, U ≈ 5.15, R ≈ 11.2, I ≈ 7.64, P ≈ 79.5. Table 8.2 reports these estimates. In the long run, this increase in financial exposure would directly and indirectly cause both inequality and unemployment to rise, the ninetieth percentile outpacing the fiftieth by an appreciable nearly 7 percent (which is also about 7 percent of the sample range of about 1.5 to 2.5) and unemployment by a

264 Table 8.2

Democracy, Inequality, and Representation Empirical System of Skew, Unemployment, Redistribution, Social Insurance, and Participation: Estimation Results

Δ(FinExp) = 4.5

Skew

Unemployment

Redistribution

Insurance

Participation

0.06739

0.39580

0.38477

–0.03988

0.27571

Source: Authors’ calculations.

noticeable nearly half-point (+.4, although this is far less of its sample range of about 18.6). The +.4 percent of GDP in redistribution induced indirectly by FinExp’s effect on these economic outcomes is non-negligible (and about the same proportion of its sample range of about 25). The –.04 percent of GDP induced change in insurance spending is negligible absolutely and proportionately, as is the +.25 percent indirect effect on voter participation. Our other sort of hypothetical begins similarly by choosing starting values for S, U, R, I, and P and the explanator of interest, in this case government partisanship, GPart. We then allow S, U, R, I, and P to respond to these starting values according to their estimated mutual endogeneity and temporal dynamics, ignoring as complicating the spatial dynamics as orthogonal to our central interests here, until converging to a steady state. From those steady-state values, we then perturb the variable of interest, here GPart, by some hypothetical series of values the response to which we wish to estimate. In this case, we take three well-known empirical cases, Germany, Sweden, and the United Kingdom, and set as starting values their actual S, U, R, I, and P values in the first year we have all five variables. We then let the estimated system converge from these and the 1960 value of GPart before tracking the innovations from the steady state as the estimated responses of S, U, R, I, and P to the actual historical path of GPart, as seen in figures 8.3 to 8.5.53 In figure 8.3, we see the last two years of center-right (CDP-FDP) government in Germany in 1960 and 1961, followed by the twenty-one years of center-left (SDP-FDP) government, 1962 to 1982, and then the cycle and rightward drift of the next eighteen years, all plotted against the right axis. The estimated responses to this are, from the top, the (mostly direct) effect on social policies in the rather steady upward trend of redistribution, to +3.55 percent of GDP at maximum, and then some oscillation with a slight downward trend, and the similarly patterned rise and then cycling drift in social insurance to +1.72 percent of GDP, down to +1.24 percent, and back again. The responses this induced in pre-taxand-transfer inequality are negligible, as we estimated almost no feedback from policies to inequality, but the impact on unemployment is quite appreciable, owing to the sizable distortionary effects of social pol-

Inequality and Unemployment

4.0

60

3.5

50

3.0

40

2.5

30

2.0

20

1.5

10

1.0

0

0.5

−10

0.0

−20

−0.5

−30

Government Partisanship

Estimated Responses to the Actual Historical Path in Germany of Government Partisanship

19 6 19 0 1962 6 19 4 6 19 6 6 19 8 7 19 0 1972 7 19 4 7 19 6 7 19 8 8 19 0 1982 8 19 4 8 19 6 8 19 8 9 19 0 1992 9 19 4 9 19 6 98

Responses

Figure 8.3

265

Year Skew Unemployment Redistribution

Social Insurance Political Participation Government Partisanship

Source: Authors’ calculations.

icy we estimated, following that now-familiar pattern: trending to +2.22 percent before falling back to +1.81 percent, and then almost back again before declining anew. The feedback from social policies to participation, meanwhile, induced a steady trend once more, this time to +3.5 percent, and with only slight oscillation visible before finally abating. Figure 8.4 plots the analogous consideration for Sweden, which started with center-left government in 1960 and then experienced, first leftgovernment, then center-left, then center-right government, through 1998. Again, the responses in redistribution are most dramatic, rising +1.41 percent of GDP in the early 1970s before declining hyperbolically to +.44 percent of GDP at the start of the 1990s, and then plummeting to –.91 percent by the end of that decade. The patterns in social insurance spending and in unemployment are similar but flatter, declining to +.68

Democracy, Inequality, and Representation

Figure 8.4

Estimated Responses to the Actual Historical Path in Sweden of Government Partisanship

2.0

30

1.5

20 10

Responses

1.0

0 0.5 −10 0.0

−20

−0.5

−30 −40

−1.5

−50

19 6 19 0 1962 6 19 4 6 19 6 6 19 8 7 19 0 1972 7 19 4 7 19 6 7 19 8 8 19 0 1982 8 19 4 8 19 6 8 19 8 9 19 0 1992 9 19 4 9 19 6 98

−1.0

Government Partisanship

266

Year Skew Unemployment Redistribution

Social Insurance Political Participation Government Partisanship

Source: Authors’ calculations.

percent and +.85 percent, +.26 percent and +.33 percent, and –.45 percent and –.41 percent, respectively. Participation follows a smoother response path, again, in Sweden, where it is vaguely sloping-hill-shaped, declining to almost +1 percent and between +.9 percent and +.7 percent over the plateau. Finally, figure 8.5 shows the analogous responses in the United Kingdom. There, the brief return to the left in the early 1970s against a background of otherwise strongly rightward-trending government places a hiccup in the otherwise steadily and sharply declining response paths. And of course, in neither the United Kingdom nor Sweden is any response in skew noticeable because, again, little feedback was found (in good part, by construction) from policy or outcomes to the (pretax, 90/50 ratio) measure.

Inequality and Unemployment

267

40

0.0

30

−0.5

20

−1.0

10

−1.5

0

−2.0

−10

−2.5

−20

−3.0

−30

−3.5

−40

Government Partisanship

0.5

19 6 19 0 6 19 2 6 19 4 6 19 6 6 19 8 7 19 0 7 19 2 7 19 4 7 19 6 7 19 8 8 19 0 8 19 2 8 19 4 8 19 6 8 19 8 9 19 0 9 19 2 9 19 4 9 19 6 9 20 8 00

Responses

Figure 8.5 Estimated Responses to the Actual Historical Path in the United Kingdom of Government Partisanship

Year Skew Unemployment Redistribution

Social Insurance Political Participation Government Partisanship

Source: Authors’ calculations.

Conclusions Conflicts of interest over the generosity and structure of social policy include that between the relatively poor and wealthy and that between the unemployed or precariously employed and the safely employed. The former conflict underlies the famous median voter result that democratic demand for broad redistribution increases in the income skew, and the latter yields a different theoretical conclusion: that inequality reduces median voter demand for social insurance. In each case, the generosity and structure of social policy may affect simultaneously the efficiency of the labor market and the political participation of the less fortunate, thus shifting the income and job security status of the median voter. These considerations imply endogenous relationships between economic per-

268

Democracy, Inequality, and Representation

formance (employment or income level and distribution), social policy (redistribution and social insurance), and politics (political participation). They also raise the theoretical possibility of multiple political-economic equilibria, with two basins of attraction: one with high equality and unemployment, redistributive, and social insurance spending and high political participation and another with the opposite pattern. This chapter has elaborated the theoretical expectations regarding these endogenous relationships, suggested identification conditions that derive from the theory and substance, and offered empirical estimates of the resulting system of equations. Our empirical analysis thus improves upon extant studies that ignore the endogenous relationships among these political, economic, and policy variables. Substantively, our results suggest that income inequality and political participation are important causes of social policy generosity in the developed democracies. However, our empirical results also revealed some puzzling surprises. Clearly, much work remains to refine the empirical specification and analysis and perhaps also to reconsider and advance our current theoretical understandings of this endogenous system of employment risk and income inequality, redistribution and insurance policies, and effective citizen input.

Robert Franzese acknowledges support from National Science Foundation grant 0340195. We thank Carrie Steele at Illinois and Kenichi Ariga, Nam Kyu Kim, and Joel Simmons at Michigan for invaluable research assistance in assembling the data and Chris Anderson, Pablo Beramendi, Sara Hobolt, Lane Kenworthy, Yves L’Horty, Xiaobo Lu, and Stephanie Rickard for helpful comments on earlier drafts.

Notes 1.

2. 3.

In earlier versions of this work, we first showed that the classical MeltzerRichards model obtains as a special case of the M&W model with full employment (and also therefore no targeting decision to make) and then that, with unemployment but no targeting of benefits, some insurance motivation for the universal benefits arises but the equilibrium remains very similar to Meltzer-Richards. Thus, the combination of unemployment and the ability to target benefits to unemployment is necessary to the M&W results. All notation and equation numbering given here exactly follows M&W to facilitate comparison. As the authors noted (and as illustrated later), high-income earners can face unemployment risk without qualitative (but with quantitative) change to the conclusions, provided unemployment risk remains weakly negatively correlated with income.

Inequality and Unemployment 4.

5.

6.

269

The job-loss and job-finding rates imply a system of differential equations, which M&W solve for these steady-state equilibria and from which they derive their comparative statics. We discuss only the steady states and so skip explicit derivation, expressing the model instead in simpler, static terms (without further loss of content or generality). Empirical estimates generally suggest coefficients of relative risk aversion of μ = 1. Log utility, which has constant relative risk aversion μ = 1, would not satisfy this, but other functions in the class of constant relative risk aversion could. . . . and assuming infinitely lived actors, but relaxation of this assumption adds only notational (and actuarial) complexity, requiring replacement of r with 1−expr , with H the actor’s life expectancy, in the asset equation that − rH

7.

8.

9.

10. 11.

12. 13. 14.

produces [8.6]. This would add little relevant substantive content, although it could possibly introduce a social life insurance motivation. A few other implications surrounding the first, second, and cross (with w ¯) derivatives of τ(t) emerge as well (see Franzese 2002, ch. 2). For example, the greater the deadweight losses (the more concave is τ(t)), the less redistribution the median voter seeks by this redistributive motivation (or, indeed, by the insurance motivation also). Moene and Wallerstein consider three classes: the permanently unemployed; low-wage, at-risk (L) workers; and high-wage, permanently employed (H) workers. L is the median. Thus, we follow them in replacing subscript i with L in this section. Median voter equilibria do not generally obtain in more than one dimension: here, t and γ. Moene and Wallerstein show that if the policy choices are sequential (Shepsle’s SIE) or if the party system prevents coalitions of rich and poor versus middle, then the middle is median in both dimensions and remains determinant. However, the single-crossing property would seem to hold, voter heterogeneity having just one dimension that identically orders voter preferences on both policies, so direct median voter equilibria should also exist. We have rewritten these two first-order conditions slightly to isolate further the implicit optimum choices on t and γ. Technically, lower class income growth can also fix or move the mean, contribute to doing so, or move contrarily and so require greater upper class income growth (all subject to remaining below median of course). We heard the catchy phrase from Tim Smeeding, whose work is among that establishing it as more common. In other words, in both cases all spending is insurance, and it rises with skew increases at a lesser rate than in the unconstrained case. The three-class model facilitated M&W’s proof that their party-coalitional equilibrium was the median voter equilibria. They also showed the equivalence of a structure-induced equilibrium, and we noted that the single-

270

15. 16.

17.

18.

19.

20.

21.

Democracy, Inequality, and Representation crossing property holds, and so direct median voter equilibria would obtain too (see note 9). We allow an additional dimension of heterogeneity here, unemployment risk, but we assume it to correlate perfectly with income, and it still orders both policy preferences, so all three options remain. Indeed, Jack Nagel (1987, 117–19) shows that American voters, at least, are generally wealthier than nonvoters. Intuitively, in the limit, with no voters or effective political participants— that is, in pure autocracy—both curves are flat (at zero), and policy is completely insensitive to societal interests. Other sources of information include any identities that are known to hold in or across equations (and thus are unnecessary to estimate); other restrictions, besides exclusions, on the coefficients in or across equations (for example, that some are equal or proportionate); knowledge of differing functional forms for the equations; and restrictions on the variance-covariance of residuals across equations. This list is not exhaustive, exclusive, or disjoint. Bayesian priors on parameters, for example, can also add information useful for identification. The validity of any empirical strategy of identifying an endogenous system, and so the credibility and creditability of estimation results, ultimately rests on the strength of the theoretical and substantive arguments that produced the identifying restrictions. Therefore, identification—that is, just-identifying assumptions (endogeneity versus exogeneity)—cannot be directly tested empirically. Over-identifying restrictions, however, can be so tested; given some assumptions that suffice to identify a system, we can test empirically whether additional restrictions seem to enhance efficiency (if true) or render estimates inconsistent (if false). Such tests of over-identifying restrictions, furthermore, are only asymptotic and are usually weak in small samples—that is, in practice. The phrasing “test for endo-/exogeneity” should be avoided, being almost oxymoronic. Indeed, the median-to-mean ratios of the theory, which are less widely available empirically, relate more tightly to 90/50 than to 50/10 or 90/10 ratios, since the ninetieth-percentile numerator more heavily influences the mean than does the fiftieth and since the denominator is the desired measure exactly. R and I relate directly only past some critical equality level, w0, at which I exhausts revenue, binding R to zero. No developed democracy has ever produced such a policy mix, and not remotely in our sample, so we might safely ignore this complication, experience having proven that the degree of equality that produces it, even as refracted by unequal participation, has never arisen. Even demographics, however, are not entirely unproblematic: for example, retirement age, insofar as it enters pensioner-targeted spending, is politically determined and may be set in response to the amount and costs of insurance spending.

Inequality and Unemployment 22.

23.

24. 25. 26. 27.

28.

29.

30.

31.

32. 33.

271

Specifically, Stolper-Samuelson holds that (unskilled) labor in developed (human-and-physical-capital-rich, labor-scarce) countries, which presumably occupy income percentiles around the denominator of most skew measures, loses by trade, more so as trade increases with countries that have greater endowment-ratios of the opposite type. In economic theory, insofar as quantities as well as prices adjust to clear markets (here employment and wages), the price (wage) implications of equilibria that assume market clearing solely by prices (wages) are shared between prices (wages) and quantities (employment) once rigidities enter. However, due caution regarding the terms involved in interactions would lead one to insert SDG in both. Policy programs in response to FLFP may also bolster FLFP, which suggests a further endogeneity not explored here. We smooth participation rates across election and non-election years, so the election date does not affect our measure. Future work may benefit further from using these first empirical estimations of the resulting system—which we consider preliminary and suggestive only—to discover refinements that would allow more exclusions or otherwise enhance the specification. We do know that district magnitude enters reciprocally (or in logs) because its effects arise through proportionate reduction in the effective threshold for entry to parliament, which is approximately 1/(2 × DMag). In essence, time precedence will fail to ensure exogeneity given instantaneous (within the observational period) endogeneity or some failure of the empirical specification to capture the dynamics fully (such as when expectations are important and not modeled, or inadequately modeled, empirically). In this case, the extremely smooth and slow time adjustment of skew leaves serious doubts as to whether our measurement precision suffices for the deviations of skew from the modeled AR(1) to have adequate signal-tonoise ratios for us even to judge how well we might have met this condition. We did explore the possibility, however, finding some indications, weak and not robust across specifications, that majority status speeds policy adjustment or reduces welfare and insurance spending levels beyond the role our GFrag and GPol allow it. These terms provide two more unique (and exogenous) regressors to the policy equations if we continue to assert the “poor man’s exogeneity”— time—and even without that assumption, the GFrag and GPol modification of the simple, uniform-across-all-countries AR(1) dynamics would be unique to the policy equations and so help identify them. Thomas Plümper and Vera Tröger (2007) offered an alternative strategy that could have been fruitfully applied here. Even to define Comp comparably across our heterogeneous sample of democracies is a very daunting task (but see Ariga 2007).

272 34.

35.

36.

37. 38. 39. 40. 41.

42.

43.

44. 45. 46.

47.

Democracy, Inequality, and Representation The twenty-three countries are Australia, Austria, Belgium, Canada, Denmark, Finland, France, Germany, Greece, Iceland, Ireland, Italy, Japan, Luxembourg, the Netherlands, New Zealand, Norway, Portugal, Spain, Sweden, Switzerland, the United Kingdom, and the United States. These data were generously given to us by David Rueda, and given to him by the OECD Education, Employment, Labor, and Social Affairs Directorate (Pontusson and Rueda 2000). We gather and extend available data aggressively, merging multiple data sources on identical or related variables and, rarely, by linear interpolation and trend autoregression, but all this adds almost no usable data here beyond these few added to skew. Greece, Iceland, and Luxembourg were dropped, and Spain remained with just one observation (1995). Greece, Iceland, Luxembourg, and Spain were dropped. See the Global Financial Data website at http://www.globalfinancialdata.com. See the web page for Cusack’s “Parties, Governments, and Legislatures Data Set” at http://www.wz-berlin.de/mp/ism/people/misc/cusack/d_sets.en.htm. The president’s party represents the cabinet party in all the following discussion for the U.S. case. “Legislature” here always refers to the lower (more powerful) chamber in cases of bicameralism. GPol exaggerates by thus implicitly assuming all legislative parties are vetoactors. As with GFrag, GPol would do better to find some convenient generalization to reflect that larger opposition parties are more likely to be in veto-acting positions than smaller ones. Cusack’s data provide several useful indicators of governing and opposition fragmentation and key-party ideological locations that could improve our GPol measure and also enhance our GFrag simplification in future work. E sums these unit allocations in the rare cases of multiple elections within one year. The U.S. case is an exception, allocating 7/9 over the 365 days before an on-year election (reflecting the president, the House, and one-third of the Senate being elected) and 4/9 to the 365 days before off-year elections (House plus one-third of the Senate). See the institute’s website at http://www.idea.int. Information on the site may allow finer granularity measures of registration ease based on the assistance provided. As noted earlier, we do not have TExpD and Comp at this time. Pres, which has no within-country variation, is also omitted from the equations with fixed effects. All other variables have at least some within-country variance, but several (as mentioned) have little or very little. Accordingly, their coefficients’ estimates should be interpreted with extra caution (see Plümper and Tröger 2007). This option then raised the question of whether we should expand the instrument set to some or all of the products of some or all of the exogenous

Inequality and Unemployment

48.

49.

50.

273

regressors in those equations. Fearing overfitting in the instrumentation stage and thereby reintroducing endogeneity (especially since we already risk doing so in using full sets of year and country dummies and time lags as instruments), we chose not to do so. Some counterintuitive results regarding the effects of stock market capitalization led us to wonder whether it was proxying GDP growth. Adding GDP growth as an endogenous or exogenous regressor, however, altered these and other results only marginally. Note that these results, which account for endogeneity, use much new data, apply several different theories, and apply different methodologies and specifications, are strongly at variance with Franzese (2002, ch. 2). If we had just the five endogenous variables, yi(5 × 1), with coefficients θ(5 × 5) in each others’ equations (θpq is variable q’s coefficient in p’s equation and θpp = 0) and the matrix of exogenous variables with associated coefficients, Xβ (5 × k)(k × 1), then y = θy + Xβ + ε, which solves for y in terms of X and the coefficients as y = (I – θ)–1(Xβ + ε). With this expression, we could interpret responses in all five outcomes according to whatever counterfactual shocks in ε or X interested us. However, two of the endogenous variables, participation and skew, interact in determining two of the outcomes, redistribution and insurance, so the θ here include conditional coefficients that depend on y. That is, the coefficients on skew and participation in the insurance and redistribution equations depend, respectively, on the level of participation and skew; thus, the responses of the outcomes depend on the level of the outcomes from which our counterfactual begins. However, finally, for any given set of coefficients, only one or some set(s) of outcome values are consistent with themselves—that is, only one or some set(s) of outcome values generate policy and other outcome responses to skew and participation consistent with those outcome values. Such sets are called steady-state, or equilibrium, values. No guarantee exists, however, that these steady-state values would lie within substantively plausible ranges, because no assurance exists that the sample ever came to or near the steady state(s). In sum, stable estimates of substantive effects require consideration of changes from steady states, yet the steady states implied by the coefficient estimates may lie at substantively implausible values of y and thus give conditional coefficients for the interactive terms that are implausibly, or even illogically, large or small. Such is our case, at least for the counterfactuals we considered. We eliminated from consideration the spatial interdependence, the country fixed effects, and all exogenous variables except the one that we perturb counterfactually. This meant that we did not have to enter the entirety of all five equations—just θ, the time lag, and the x and β in question—but this also virtually assured that the equilibrium values from which we would calculate changes in response to hypotheticals would be substantively meaningless. This is unimportant if we report only

274

51. 52.

53.

Democracy, Inequality, and Representation changes from that steady state, wherever it may be, as the effect—only to some extent, however, because the aforementioned conditional coefficients would not be entirely valid. This, finally, is why we adopt the compromise expedients described in the text. We omit all aspects of the equations except those involving S, U, R, I, P, their time lags, or the exogenous variable of the hypothetical (see note 50). This only approximates because some proportion of each calculation’s update is its proportionate step from the initial values toward the steady state and only the remainder is the desired response. By subtracting the two calculations, we largely eliminate the steps toward the steady state, but some remain as the difference in these proportionate steps. Since the steady states are very far from the initial values relative to the response to the hypothetical, the proportionate steps are of almost identical size. These too are only approximations because the steady-state conditional coefficients described in note 50 are not realistic—implausibly large in this case—as explained there.

References Alesina, Alberto, and George-Marios Angeletos. 2005a. “Fairness and Redistribution: United States vs. Europe.” American Economic Review 95(3): 913–35. ———. 2005b. “Redistribution, Corruption, and Fairness.” Journal of Monetary Economics 52(7): 1227–44. Ariga, Kenichi. 2007. “Electoral Value of Party Label and Public Policymaking in Developed Democracies.” Unpublished manuscript. Department of Political Science, University of Michigan, Ann Arbor. Armingeon, Klaus, Philipp Leimgruber, Michelle Beyeler, and Sarah Menegale. 2005. “Comparative Political Data Set, 1960–2003.” Berne: Institute of Political Science, University of Berne. Atkinson, Anthony, Lee Rainwater, and Timothy Smeeding. 1995. Income Distribution in OECD Countries: The Evidence from the Luxembourg Income Study (LIS). Social Policy Studies 18. Paris: OECD. Bartels, Larry. 1991. “Instrumental and ‘Quasi-Instrumental’ Variables.” American Journal of Political Science 35(3): 777–800. Basinger, Scott J., and Mark Hallerberg. 2004. “Remodeling the Competition for Capital: How Domestic Politics Erases the Race to the Bottom.” American Political Science Review 98(2): 261–76. Beck, Thorsten, George Clarke, Alberto Groff, Philip Keefer, and Patrick Walsh. 2001. “New Tools in Comparative Political Economy: The Database of Political Institutions.” World Bank Economic Review 15(1): 165–76. Cameron, David. 1984. “Social Democracy, Corporatism, Labor Quiescence, and the Representation of Economic Interest in Advanced Capitalist Society.” In

Inequality and Unemployment

275

Order and Conflict in Contemporary Capitalism, edited by John H. Goldthorpe. New York: Oxford University Press. Conway, M. Margaret. 1985. Political Participation in the United States. Washington: Congressional Quarterly Press. Cox, Gary W., and Frances Rosenbluth. 1995. “Anatomy of a Split: The Liberal Democrats of Japan.” Electoral Studies 14(4): 355–76. Danziger, Sheldon H., and Peter Gottschalk. 1995. America Unequal. Cambridge, Mass.: Harvard University Press. Ebbinghaus, Bernhard, and Jelle Visser. 2000. Trade Unions in Western Europe Since 1945. London: Palgrave Macmillan. Franzese, Robert. 2002. Macroeconomic Policies of Developed Democracies. Cambridge: Cambridge University Press. Franzese, Robert, and Jude C. Hays. 2004. “Empirical Modeling Strategies for Spatial Interdependence: Omitted-Variable vs. Simultaneity Biases.” Paper presented to the summer meetings of the Political Methodology Society, July 2004, Palo Alto, Calif. ———. 2006a. “Spatio-Temporal Models for Political Science Panel and Time-Series Cross-Section Data.” Paper presented to the summer meetings of the Political Methodology Society, July 2006, Palo Alto, Calif. ———. 2006b. “Strategic Interaction Among EU Governments in Active-LaborMarket Policymaking: Subsidiarity and Policy Coordination Under the European Employment Strategy.” European Union Politics 7(2): 167–89. ———. 2007a. “Empirical Models of Spatial Interdependence.” In The Oxford Handbook of Political Methodology, edited by Janet M. Box-Steffensmeier, Henry E. Brady, and David Collier. Oxford: Oxford University Press. ———. 2007b. “Spatial-Econometric Models of Cross-Sectional Interdependence in Political Science Panel and Time-Series Cross-Section Data.” Political Analysis 15(2): 140–64. Freeman, Richard B. 1991. “How Much Has De-Unionization Contributed to the Rise in Male Earnings Inequality?” Working paper 3826. Cambridge, Mass.: National Bureau of Economic Research (August). Golden, Miriam, Peter Lange, and Michael Wallerstein. 1997. “Union Centralization Among Advanced Industrial Societies: An Empirical Study.” Version dated June 16, 2006. Dataset available at http://www.shelley.polisci.ucla.edu/ data/index.html. Golder, Matt. 2005. “Democratic Electoral Systems Around the World, 1946– 2000.” Electoral Studies 24(1): 103–21. Gottschalk, Peter, and Timothy Smeeding. 1997. “Cross-National Comparisons of Earnings and Income Inequality.” Journal of Economic Literature 35(2): 633–87. Greene, William H. 2003. Econometric Analysis, 5th ed. Upper Saddle River, N.J.: Prentice Hall. Hall, Peter, and David Soskice, editors. 2001. Varieties of Capitalism: The Institutional Foundations of Comparative Advantage. New York: Oxford University Press.

276

Democracy, Inequality, and Representation

Harrop, Martin, and William L. Miller. 1987. Elections and Voters: A Comparative Introduction. London: Macmillan. Hobolt, Sara Binzer, and Robert Klemmensen. 2006. “Welfare to Vote: The Effect of Government Spending on Turnout.” Paper presented to annual meeting of the Midwest Political Science Association, April 2006, Chicago, Ill. Huber, Evelyne, Charles Ragin, John D. Stephens, David Brady, and Jason Beckfield. 2004. “Comparative Welfare States Data Set.” Northwestern University, University of North Carolina, Duke University, and Indian University. Accessed at http://www.lisproject.org/publications/welfaredata/welfareaccess.htm. Huber, Evelyne, and John D. Stephens. 2001. Developmetn and Crisis of the Welfare State: Parties and Policies in Global Markets. Chicago.: University of Chicago Press. Katzenstein, Peter J. 1985. Small States in World Markets: Industrial Policy in Europe. Ithaca, N.Y.: Cornell University Press. Kayser, Mark Andreas. 2007. “How Domestic Is Domestic Politics? Globalization and Elections.” Annual Review of Political Science 10: 341–62. Kenworthy, Lane. 2001 “Wage-Setting Coordination Scores” (June 17). Accessed at http://www.u.arizona.edu/~lkenwor/WageCoorScores.pdf. Lijphart, Arend. 1999. Patterns of Democracy: Government Forms and Performance in Thirty-Six Countries. New Haven, Conn.: Yale University Press. Moene, Karl Ove, and Michael Wallerstein. 2001. “Inequality, Social Insurance, and Redistribution.” American Political Science Review 95(4): 859–74. Nagel, Jack H. 1987. Participation. Englewood Cliffs, N.J.: Prentice-Hall. Persson, Torsten, and Guido Tabellini. 2000. Political Economics. Cambridge, Mass.: MIT Press Plümper, Thomas, and Vera Tröger. 2007. “Efficient Estimation of Time-Variant and Rarely Changing Variables in Finite Sample Panel Analyses with Unit Fixed Effects.” Political Analysis 15(2): 124–39. Pontusson, Jonas, and David Rueda. 2000. “Wage Inequality and Varieties of Capitalism.” World Politics 52(3): 350–83. Quinn, Dennis P., and Carla Inclan. 1997. “The Origins of Financial Openness: A Study of Current and Capital Account Liberalization.” American Journal of Political Science 41(3): 771–813. Rodrik, Dani. 1998. “Why Do More Open Economies Have Bigger Governments?” Journal of Political Economy 106(5): 997–1032. Rosenstone, Steven J., and John Mark Hansen. 1993. Mobilization, Participation, and Democracy in America. New York: Macmillan. Shugart, Matthew Soberg, and John M. Carey. 1992. Presidents and Assemblies: Constitutional Design and Electoral Dynamics. Cambridge: Cambridge University Press. Smeeding, Timothy, Michael O’Higgins, Lee Rainwater, and Anthony B. Atkinson, editors. 1990. Poverty, Inequality, and Income Distribution in Comparative Perspective: The Luxembourg Income Study (LIS). New York: Harvester Wheatsheaf.

Inequality and Unemployment

277

Smeeding, Timothy, and Dennis H. Sullivan. 1998. “Generations and the Distribution of Economic Well-being: A Cross-National View.” American Economic Review 88(2): 254–58. Swank, Duane. 2002. Global Capital, Political Institutions, and Policy Change in Developed Welfare States. Cambridge: Cambridge University Press. Tsebelis, George. 2002. Veto Players: How Political Institutions Work. Princeton, N.J.: Princeton University Press. Verba, Sidney, Norman H. Nie, and Jae-On Kim. 1978. Participation and Political Equality: A Seven-Nation Comparison. Chicago, Ill.: University of Chicago Press. Verba, Sidney, Kay Lehman Schlozman, and Henry E. Brady. 1995. Voice and Equality: Civic Voluntarism in American Politics. Cambridge, Mass.: Harvard University Press. Wilson, Sven E., and Daniel M. Butler. 2007. “A Lot More to Do: The Sensitivity of Time-Series Cross-Section Analyses to Simple Alternative Specifications.” Political Analysis 15(2): 101–23. Wolfinger, Raymond E., and Steven J. Rosenstone. 1980. Who Votes? New Haven, Conn.: Yale University Press.

Chapter 9

Income, Inequality, and Electoral Participation CHRISTOPHER J. ANDERSON AND PABLO BERAMENDI

The supposition that material welfare influences whether and how citizens participate in democratic politics has a long and rich tradition in the social sciences. Moreover, the notion that income and income inequality matter to democratic processes and the quality of democratic outcomes is widely accepted. Yet, while scholars have vigorously investigated the impact on participation of citizens’ resources in terms of social status, education, and levels of income, the issue of how relative income at the individual level and income inequality at the macro level affect civic participation has received relatively little attention from social scientists. This relative lack of attention by scholars to the question of how income inequality cross-nationally and differences in income at the level of individual citizens affect civic life is surprising given the normative importance attached to inequality’s effects on democracy. After all, it has long been argued that economic inequality leads to inferior democratic outcomes because it concentrates power among a smaller group of people and increases politicians’ responsiveness to an ever-smaller group of advantaged citizens (Bartels 2002; Dahl 1971; Pateman 1971; Schattschneider 1960). To help develop a better understanding of the connection between economic and political inequality we therefore examine the impact of individuals’ location within a country’s income distribution (relative income) and cross-national differences in income distributions (inequality) on people’s propensity to vote. Based on individual and macrolevel data collected in eighteen OECD democracies, we find that income significantly affects electoral participation. At the level of individual citizens, we find that the effects of income

Income, Inequality, and Electoral Participation

279

differentials are essentially linear, such that individuals who are below the median income in society are less likely to participate in elections, while those above the median income are more likely to do so. Moreover, our results show that the effect of income on electoral participation increases monotonically. We also find that income inequality increases electoral abstention across advanced democracies. Moreover, our analyses reveal not only that overall income inequality affects people at different ends of the income distribution similarly, but that greater inequality at either the upper or lower end of the income distribution has equally deleterious effects on electoral participation. Thus, our analyses show that income inequality— regardless of its specific shape—reduces electoral participation and that it does so for all kinds of individuals. However, we also argue that the mechanism by which inequality affects participation is likely to differ between high- and low-income individuals: while inequality at the lower end reduces participation among low-income individuals because it deprives them of resources, higher inequalities at the upper end reduce participation by creating disincentives for the very rich to get involved. In this chapter, we first explore the logic of how income and income inequality affect democratic representation and civic behavior. We then develop a model of income and political action that we subsequently test with data from eighteen OECD countries. We conclude with a discussion of a possible research agenda linking income inequality and the quality of civic involvement that would bring insights from political economy and political psychology to bear on the study of citizen behavior.

Absolute Income and Electoral Participation Students of mass political behavior have long argued that a country’s economic conditions affect people’s beliefs about whether government is legitimate and effective as well as citizens’ propensity to be involved in the political process. In fact, a sizable literature has investigated the link between economic conditions and turnout at the macro level (for an excellent review, see Radcliff 1992). Painting with a broad brush, we note that the empirical findings from this research point in several directions. First, they indicate that the impact of macroeconomic fluctuations on turnout, in terms of both size of impact and direction, differs between the developed and the developing world. While a bad economy depresses turnout in the industrialized societies, it increases turnout in the less industrialized ones (Aguilar and Pacek 2000; Bahry and Lipsmeyer 2001; Radcliff 1992). Second, the impact of the business cycle on turnout is more pronounced in countries with less generous welfare states (Pacek and Radcliff 1995; Radcliff 1992).

280

Democracy, Inequality, and Representation

Related but much less extensive strands of research have examined the impact of economic conditions on other forms of participation and citizen involvement, finding that a good economy generally fosters more involvement (van Deth and Elff 2004). However, this literature is limited in scope and depth, and the most plausible inference to be drawn from it is that the question of how the macroeconomy is related to different kinds of participation—at the aggregate level—remains open. Although these studies have produced valuable insights, their focus has consistently been on examining general macroeconomic performance indicators such as GDP growth or unemployment rather than income. This has created a bit of a separation between macro and micro levels of analysis. Specifically, while studies relating macroeconomic outcomes to citizen participation tend to focus on broad measures of the national economy and cross-national evidence, they tend to be rooted in individual-level theories about citizen participation. Yet empirical tests of these individual-level theories tend to focus on individuals’ specific circumstances rather than the national economy. And while it is assumed that general economic conditions filter down to the individual level, the connection between the two is not always obvious. Speaking generally, individual-level theories about the connection between material well being and participation typically view citizens’ economic situation either as a resource or an incentive for participation. Specifically, to the extent that they theorize about the individual-level foundations of how individual well being affects participation, studies that view well being as a resource focus on income as one of several resources that enhance the potential for participation. Conversely, a smaller set of alternative theories view income as producing incentives to participate or abstain. And while the resource-based view sees income as mattering in an absolute sense—the more resources one has, the more participation one will produce—the incentive-based view sees income as mattering in a relative sense—that is, in terms of individual income relative to others. Of the two perspectives on how income matters, the study of income as a political resource has received considerably more scholarly attention. Scholars exploring what has traditionally been referred to as the “SES model”—or, more recently, the “resource” model of participation—have theorized and found that individuals of higher socioeconomic status and, concomitantly, more material, cognitive, and other resources (such as time) to be involved in politics participate at higher rates in a variety of participatory acts (Leighley 1995; Verba, Schlozman, and Brady 1995). In this context, it is important to note that income has been treated as a critical ingredient in, as well as a prominent proxy for, socioeconomic status, typically alongside education and social class (Verba, Schlozman,

Income, Inequality, and Electoral Participation

281

and Brady 1995). A substantial literature has documented that higher levels of education and social status (measured by income or class or a combination of the three) are the most consistently significant predictors of political action across a variety of countries (Almond and Verba 1963; Barnes and Kaase 1979; Teorell, Torcal, and Montero 2007). In The Civic Culture, for instance, Gabriel Almond and Sidney Verba (1963) reported that respondents with higher levels of education, income, and work skills were more likely to express pride in their country’s political institutions and participate in politics. Yet, while research on the demographic and attitudinal correlates of political participation has long assumed that higher socioeconomic status goes along with greater involvement in the existing political system, the resource-based view has seldom (if ever) theorized that the role that income plays in individuals’ decisions to participate is different from education or social status.1 An exception was the pathbreaking study by Sidney Verba, Kay Schlozman, and Henry Brady (1995; see also Brady, Verba, and Schlozman 1995), which sought to specify more explicitly the role that resources (including income) play in shaping participation. In particular, this study hypothesized that time, money, and civic skills are critical skills for participation. Perhaps surprisingly, despite this more explicit focus on resources, the evidence for the effect of income on participation remains somewhat ambiguous, even in this revised SES model. For one, income is not theorized to matter explicitly, but only as part of a socioeconomic hierarchy model. Moreover, education and life circumstances (types of jobs, having children, and so on) seem to matter more for engendering free time to participate or for developing critical civic skills. This ambiguity is both theoretical and empirical. For one, the theoretical prediction that income can be used to purchase time (the so-called leisure effect) can be countered by the argument that high wages raise the opportunity costs of free time (Mincer 1962; Sharp 1981; see also Brady, Verba, and Schlozman 1995, 274). Moreover, on an empirical level, Brady and his colleagues (1995) actually found no evidence of a relationship between SES and whether respondents had time to engage in political action, and levels of income (measured as family income) did not matter for political action once the possible endogeneity among critical variables was accounted for. What is more, at least one study has found a statistically significant and negative coefficient for income on voter turnout once education is controlled for (Chapman and Palda 1983). While the SES effect has been replicated in a variety of contexts, it refers to some absolute level of resources that matters in a linear fashion, with the understanding that more is better than less and that an incre-

282

Democracy, Inequality, and Representation

mental increase in resources results in an equally sized increase in participation. Moreover, with the exception of studies that have sought to measure resources directly (time, money, and skills, for example), the vast majority of research in the area usually has measured socioeconomic status broadly and as an amalgamation by combining education and income measures for individuals (sometimes along with measures of class). Taken together, then, the resource-based view of income is that it should matter—and perhaps does matter—but the findings of the SES literature suggest that it may be more proximate variables, such as civic skills and life circumstances, that are driven more by education than income and drive the effect between status and participation. Put another way, absolute levels of income do not seem to matter consistently to participation, but income obviously should not be counted out a priori as a potentially important factor. In sum, while research focused at the macro level does not usually consider the impact of income on participation, it does suggest that a good economy goes with more participation, at least in the industrialized countries. Moreover, the preponderance of research focused on individual behavior focuses on income as a resource, suggesting that more income (or well being) produces ever more participation. Yet these studies leave open a number of questions regarding the role that income plays in shaping civic behavior. Specifically, it remains unclear whether income really does matter in some absolute sense—as a resource, where more is better—or in a relative sense—where having less than others matters. In this chapter, we examine these issues in more detail.

Relative Income: Competing Logics of Political Action The notion that people who have greater resources (higher socioeconomic status, more time, more money, and so on) or are doing well materially can be expected to participate more in politics is based on the idea that income means having the power and resources to participate. The logic of this argument is essentially that economic resources (such as income) readily translate into politically relevant resources (time, for example) and therefore into more political action; such an effect is essentially linear; and doing poorly materially makes people dissatisfied with representative democracy, and less likely to be involved in politics as a result. On its face, the existing evidence linking income inequality and participation at the aggregate level points in a similar direction. On those rare occasions when scholars have considered the impact of macrolevel income inequality on citizen participation, they have found that inequality

Income, Inequality, and Electoral Participation

283

is associated with lower levels of involvement. Among the few who have investigated this relationship, Robert Goodin and John Dryzek (1980), for example, found that income inequality had a negative effect on turnout in elections held in the late 1950s across thirty-eight democracies. Similarly, Carles Boix (2003, 118–29) and Frederick Solt (2004) found that income inequality depressed participation at the subnational level in the United States early in the twentieth century and in Italian regions during the 1970s and 1980s, respectively. And Eliana La Ferrara (2002) found that income inequality is associated with lower group participation in rural Tanzania. Although this evidence is suggestive, it does not qualify as dispositive. For one, some of it, such as Goodin and Dryzek’s (1980), is based on simple bivariate correlations, which makes it difficult to draw unambiguous inferences absent a more fully specified estimation model. Moreover, and perhaps more importantly, it is not entirely clear why exactly income inequality at the macro level should translate into individual decisions to participate (or not). So far as we can tell, the logic underlying this argument is based on an assumption about citizens’ understanding of the political consequences of inequality. Specifically, it assumes that (relatively) poorer citizens view greater inequality as favoring the policy interests of the (relatively) rich, and that they view a more unequal distribution of income as distorting the representational process in favor of the rich more than a more equal distribution. Thus, a more unequal world leads to perceptions of the political system as unresponsive and a lower incentive to spend scarce resources on participating when its benefits are less than assured. All this is plausible, but there also are good theoretical reasons to believe that relative income can affect participation in different, and perhaps contradictory, ways. That is, based on available theory and evidence, relative income can be predicted to either stimulate or diminish participation, and both of these predictions are based on viewing income as providing citizens with incentives to participate or abstain.

Income as an Incentive to Participate and Abstain For one, the prediction that relative income stimulates participation can be derived from an understanding of income inequality as a proxy for heterogeneous preferences. If inequality in incomes reflects a diverse set of preferences (about, say, defending or changing the status quo with regard to redistributive policy), we would predict that distance from the median income produces incentives to participate to either defend the status quo (at the upper end of the income distribution) or to change the status quo (among those at the lower end). If this is the case, inequality

284

Democracy, Inequality, and Representation

as an indicator of heterogeneous preferences would imply more political conflict among a greater number of people and greater incentives to become politically involved. In this way, then, inequality would be viewed as an engine of politics at both ends of the income distribution, where more inequality begets more political activity (incentives to change versus preserving the status quo) and very little inequality begets apathy, at least among both the relatively well-off and the less well-off. An alternative prediction about the impact of relative income differences on participation can be based on work that has sought to understand people’s incentives to provide public goods and engage in collective action. Specifically, Mancur Olson (1965) argued that the effect of increased inequality on collective action may be positive when the extent to which individuals benefit from the good depends on their initial endowments (see also Bergstrom, Blume, and Varian 1986). That is, for any given group size, when the share of the benefits is positively related to individual wealth, richer members have more incentives to contribute resources or to monitor others. In this case, higher inequality would alleviate the free-rider problem and lead to increased public good provision (see also Baland and Platteau 1997). Recent theoretical work calls these conclusions into question, however, or at least qualifies them in important ways. Specifically, if members of a group are allowed to stop participating—that is, if free-riding is possible and cheap—increased inequality worsens the free-rider problem, and the set of contributors may shrink substantially (La Ferrara 2002).2 Thus, in contrast to Olson (1965) and Ted Bergstrom, Larry Blume, and Hal Varian (1986), who predicted a positive and a neutral effect of income inequality on public good provision, respectively, recent studies have argued that the impact of inequality on collective action may indeed be negative (La Ferrara 2002). Thus, increased inequality may lead to less collective action when the free-rider problem gets worse (Baland and Platteau 1997), and higher disparities in income may lead to less efficiency if there is high complementarity among the inputs of rich and poor members of the group (Baland and Ray 1997). To complicate matters further, the effects of inequality on participation may not fall equally on people at different rungs of the income distribution. La Ferrara (2002), for example, found that the effect of inequality on collective action is negative if redistribution leads to more inequality in the bottom part of the distribution, or, alternatively, that the effect can be positive if the redistribution benefits a few rich people at the expense of a large mass of poor people so that there is relatively more equality at the bottom of the distribution. In all instances, La Ferrara argued, if participation does decrease, it is the relatively richer members who choose to drop out (La Ferrara 2002, 247–48). She found evidence consistent

Income, Inequality, and Electoral Participation

285

with this in rural villages in Tanzania; Goodin and Dryzek (1980) similarly found that turnout is lower among the relatively less wealthy and in places with higher income inequalities. We suspect that there are a number of reasons for these quite divergent theoretical expectations about the impact of income on participation, as well as the possibly divergent empirical findings regarding the relationship between income inequality and participation. Perhaps the most straightforward way to summarize these would be to suggest that income can and should be viewed as both a resource and an incentive to participate, and that viewing income as a resource versus an incentive can lead to competing predictions at different ends of the income distribution. That is, incentives and resources can either complement or compete with one another. For example, resources and incentives may complement one another at the upper end of the income distribution—with both predicting that participation will be greater or that individuals at the upper end of the distribution will be better able to overcome any collective action problem—with resources thus dominating incentives. Alternatively, at very high levels of income, incentives to be involved may diminish, thus counteracting the impact of resources. Moreover, existing work suggests that resources and incentives clash at the lower end of the income distribution—with income as a resource predicting less participation and income as an incentive predicting more—but that resources to participate dominate incentives to participate.

The Issues of Causality and Disparate Effects Most empirical and theoretical studies of the relationship between inequality and electoral participation do not address the issue of causality. Specifically, while there is good reason to believe that inequality drives participation, there also is reason to believe that it could work in the reverse. That is, participation may drive (in)equality by creating demands for social protection (Mueller and Stratmann 2003). Specifically, it is commonly argued that inequality is shaped, at least in part, by income policies and policies of social protection, since more extensive levels of redistribution and income maintenance tend to lower levels of income inequality (for recent arguments along these lines, see Iversen and Soskice, this volume; see also Beramendi and Cusack 2004; Kenworthy and Pontusson 2005; Pontusson, Rueda, and Way 2002; Rueda and Pontusson 2000; Schwabish, Smeeding, and Osberg 2003). As a result, the effects that inequality may have on participation can be traced back in the causal chain to measures aimed at providing different kinds and levels of income protection (Franzese 1998). In a way, then, inequality effects can also be seen as the mirror image of welfare states, and as a result, in-

286

Democracy, Inequality, and Representation

equality may well be endogenous to participation (Mahler 2002; see also Franzese and Hays, this volume).

Analysis To get an empirical handle on these competing claims we need to address several issues. First, what is the impact of individual income on participation? Second, what is the impact of income inequality on participation? And, third, what is the causal relationship between income inequality and participation? To answer these questions we must investigate the impact of individual differences and variation in national contexts on the behavior of individuals. Because such an investigation requires a cross-national research design that combines levels of analysis, the data analyzed here include both individual-level and aggregate-level information. Our individuallevel data come from surveys collected as part of the World Values Surveys (WVS) from 1999 to 2001 (ICPSR study 3975). The following eighteen countries—drawn from the OECD countries and varying with regard to the relevant institutional and contextual characteristics (see the appendix)—provided the most important survey items and had a sufficient number of cases for multivariate analysis: Australia, Austria, Belgium, Canada, Denmark, Finland, France, Germany, Greece, Ireland, Italy, the Netherlands, Norway, Portugal, Spain, Sweden, the United Kingdom, and the United States.

The Dependent Variables The individual-level dependent variable investigated in this study is vote intention (turnout). Turnout is a variable that has been used widely in research on political behavior for a variety of purposes. We coded this variable such that a higher value indicated the intention to abstain from participation in the next election. For details on question wording and coding, see the appendix.

The Main Independent Variables: Inequality and Income Our key independent variables are inequality and individual income. Our inequality variable is the Gini coefficient of the disposable household income per equivalent adult. In calculating this index, we use waves 4 (around 1995) and 5 (around 2000, if available) of the Luxembourg Income Study (LIS) data set and calculate the average value between the

Income, Inequality, and Electoral Participation

287

two. We adjust the data using the LIS equivalence scale.3 The measure ranges from 0 for countries with no inequality to a theoretical maximum of 1 when one household is in possession of the total income of society. The actual range of scores in our sample of countries is from .24 (Sweden) to .37 (United States), with a mean score of 0.29. To assess whether the impact of inequalities in the lower half of the income distribution are similar to those of inequalities in the upper half, we use two alternative measurements, also based on the disposable household income per equivalent adult. The first is the ratio between the fiftieth and tenth percentiles of the distribution. This measure, which captures the degree of dispersion in the lower half of the distribution, ranges from 1.18 to 2.66, with an average value of 2.05 among the countries in our sample. The second is the ratio between the ninety-ninth and fiftieth percentiles of the distribution, which in turn captures the degree of the dispersion in the upper half of the distribution. This measure ranges from 2.35 to 7.03, with an average value of 3.86 among the countries in our sample.4 At the individual level, income is measured with the help of a tenpoint scale that places individuals in deciles, which are derived from a question about total household income, including wages, salaries, pensions, and other income. We then recode the income scale to identify individuals by whether their income falls below or above the median income in society and by how much. Thus, our income distance variable (or relative income) ranges from –4 (individuals in the fourth decile below the median income) to +5 (individuals in the fifth decile above the median income category). To examine whether the effect of this variable on the odds of participation is linear, we also include a variable squaring the income distance variable.5

The Control Variables We also control for a host of factors at the level of individuals and countries that may affect our dependent variable. For example, demographic and attitudinal factors at the individual level also affect political involvement and must therefore be controlled for in our models. A rich literature has examined political participation at the level of individuals (Verba and Nie 1972; Verba, Schlozman, and Brady 1995). Generally speaking, citizens are more likely to participate if they have the necessary resources and motivations to get involved. The most prominent proxy for resources is socioeconomic resource level (Verba, Schlozman, and Brady 1995). A substantial literature has documented that higher levels of education and social status (measured by income or social class) are the most consis-

288

Democracy, Inequality, and Representation

tently significant predictors of political action across a variety of countries. We sought to capture the effects of a respondent’s level of education with the help of a survey question about the highest educational level attained. Scholars have also found age and sex to be important predictors of political participation. Both can be categorized under the rubric of the (socioeconomic) resource model of participation (Leighley 1995; Verba, Schlozman, and Brady 1995). This research suggests that increased social responsibilities tend to go along with increased motivations to become politically active (Nie, Verba, and Kim 1974). Some have argued that the relationship between age and participation is curvilinear, with the youngest and the oldest respondents the least likely to be involved. We also include a variable measuring respondents’ marital status, with the expectation that married individuals are more likely to be politically active. Aside from identifying older citizens as more frequent participants in the political process, researchers also have found across a number of democracies that men are more likely than women to have the resources needed to engage in political acts (Dalton 2002). Related research suggests that gender stereotypes contribute to a greater proclivity of men to engage in political action (Hansen 1997; Jennings 1983). We therefore include a control variable for gender in our models of behavior. To properly isolate the impact of relative income and inequality on participatory behaviors, we also include control variables measuring individuals’ occupation (manual workers and skilled manual workers are coded 1, and 0 otherwise), as well as a dummy variable to indicate whether the individual is currently unemployed (coded 1, 0 otherwise). In addition, our estimation models include a number of attitudinal controls. There is considerable evidence that part of the decision to engage in conventional political action has to do with affirming the political system’s legitimacy or expressing one’s personal proclivities as a socially active person. Thus, including such variables helps us capture the actual effect of income on the vote by eliminating that part of the variation that is unrelated to income and income inequality. The attitudinal variables include distrust toward the political system, subjective well being (happiness), and interpersonal trust. Our expectation is that trusting and happy respondents are significantly more likely to participate. Finally, existing literature suggests controlling for a country’s macrolevel characteristics. In our case, this includes most importantly the level of economic development (measured by the GDP per capita) as an indicator of absolute level of wealth in a society (and thus also as a proxy for the average voter’s absolute level of income). In addition, we included a measure of economic growth (annual percentage of GDP growth)

Income, Inequality, and Electoral Participation

289

(Brehm and Rahn 1997; Curtis, Baer, and Grabb 2001). Finally, our model includes measures for institutional characteristics that have been found to affect electoral participation, namely, characteristics of the electoral system in the form of disproportionality and compulsory voting rules (Jackman 1987; Norris 2004). Coding procedures and descriptive statistics for all variables are listed in the appendix.

Multivariate Analysis and Results To tackle the theoretical issues of endogeneity and disparate effects requires several econometric adjustments to a basic estimation model estimating the effects of inequality on electoral turnout (see Franzese and Hays, this volume). For one, it requires that we account for the possible endogeneity of income inequality in our models of participation. Moreover, it requires that we consider whether there is a difference in the impact of inequality on participation, depending on whether we consider the overall distribution of income in a country or that part of the income distribution in which an individual respondent is located. Finally, it requires that we investigate the interactive impact of inequality and relative income on participation. To account for the potential endogeneity of inequality, we rely on an instrumental variable approach to estimate two-way simultaneous equation models. To help identify the impact of inequality on participation we estimate a first equation in which inequality is a function of the degree of decommodification (Scruggs Index) and the dependency ratio (the share of dependent to labor force population). The latter is our instrument, that is, a variable that is directly related to the existing levels of inequality (see, for instance, Beramendi and Cusack 2004) but not to electoral participation. The R-squared for this first-stage equation is 0.70. We then use the predicted values of the dependent variable of the first-stage equation to estimate the impact of aggregate inequality on the probability of abstention. A similar procedure is used to instrument the 50/10 and 99/50 ratios.

Relative Income and Participation Our investigation begins by considering the impact of individual income on participation. Table 9.1 shows the results of the two-way instrumented variable models estimating the impact of individuals’ income on voter abstention. To ensure the robustness of our results we estimate two types of two-way models: a basic logistic regression model as well as a hierarchical linear model with random intercepts (using generalized linear

Table 9.1

Individual Income and Electoral Abstention

Distance to median income Distance to median income-squared Age Age-squared Education Female Marital status Life satisfaction Interpersonal trust Distrust in institutions Unskilled manual Skilled or semiskilled manual Unemployed GDP per capita Growth 1995 to 1999 Index of disproportionality Australia Belgium Constant Pseudo R-squared Observations

Logit

Multilevel

–0.069** (0.010) 0.006* (–0.003) –0.026** (0.008) 0.0002* (0.000) –0.043** (0.011) 0.219** (0.041) 0.145** (0.048) –0.022 (0.011) –0.203** (0.044) 0.438** (0.039) 0.077 (0.071) –0.041 (0.052) –0.071 (0.107) –0.0002* (0.000) –0.188** (0.020) 0.000 (0.003) –1.616** (0.098) –0.101** (0.082)

–0.012** (0.002) 0.001* (0.0005) –0.004** (0.001) 0.00003* (0.000) –0.006** (0.002) 0.034** (0.007) 0.024** (0.008) –0.004* (0.002) –0.033** (0.007) 0.075** (0.006) 0.013 (0.012) –0.008 (0.009) –0.005 (0.018) –0.00003* (0.000) –0.029** (0.003) 0.000 (0.001) –0.202** (0.011) –0.016** (0.014)

1.489 (0.301) 0.07 15,088

0.659 (0.048) 0.07 15,088

Source: Authors’ calculations based on 1990–2001 World Values Survey and the Luxembourg Income Study (LIS) data set. Note: Two-way instrumented variable models. Multilevel models are hierarchical linear models with random intercepts estimated with generalized linear latent and mixed models (GLLAMM). Robust standard errors are in parentheses. * p < .05; ** p < .01

Income, Inequality, and Electoral Participation

291

latent and mixed models [GLLAMM]; see Rabe-Hesketh and Skrondal 2005) to account for the multilevel nature of the data (with individuals nested within countries). The results show that individuals whose income falls below the median are less likely to turn out to vote. In contrast, citizens above the median income are less likely to abstain from electoral participation. These results are squarely consistent with the notion that individual income matters and that being among the relatively low earners breeds political abstention. Our analyses also reveal that this effect is mostly linear. Although the variable measuring squared distance achieves statistical significance, our calculations of the substantive effects reveal, to the extent that there is curvilinearity, it is moderate. Figures 9.1 and 9.2 show the substantive effects.6 The results show that the effects of income are substantively significant. Note that the variable measuring individuals’ distance from the median income category ranges from –4 (below the median) to +5 (above the median) in our data. Moving from the lowest income category (the income category with the greatest distance below the median income category at –4), where individuals have a .27 probability of abstention, to the highest income category (the income category with the greatest distance below the median income category at +5), where individuals have a .17 probability of abstention, means reducing the odds of not turning out by 10 percent (figure 9.1). Moreover, changes in the odds of abstention due to changes in relative income are most pronounced at the lower end of the income distribution. Moving from the lowest income decile below the median to the median decile decreases the odds of abstention by 7 percent, while a move from the median decile to the top (or even the fourth) decile decreases the probability of abstention by about three percentage points. Figure 9.2 reveals that any curvilinearity, to the extent that it exists, occurs most visibly at very high levels of income—in fact, at levels of income outside the bounds of our analyses. In figure 9.2, we simulated the effects of very high levels of income on the odds of abstention, showing that the odds of abstention increase again after we pass a hypothetical (and technically impossible) mark of +6. The control variables all behave as expected or turn out to be statistically insignificant. Thus, among the individual-level controls, age, education, and gender all have the predicted effects, as do the attitudinal variables. In contrast, variables measuring occupation or employment status fail to achieve conventional levels of significance. At the macro level, higher levels of economic development and a good economy increase turnout, as does compulsory voting (in the case of Australia, one of the

292 Figure 9.1

Democracy, Inequality, and Representation Individual Income and the Probability of Abstention 0.30 0.25 0.20 0.15 0.10

Abstention Upper Confidence Level 95 Percent Lower Confidence Level 95 Percent −4 −3 −2 −1

0

1

2

4

3

5

Distance to Median Income Category (Median = 0) Source: Authors’ calculations based on 1990–2001 World Values Survey and the Luxembourg Income Study (LIS) data set.

Figure 9.2

Individual Income and the Probability of Abstention (Simulated Distance Less than 5) 0.3 0.3 0.3 0.2 0.2 0.2 0.2 0.2 0.1 0.1 0.1

Abstention Upper Confidence Level 95 Percent Lower Confidence Level 95 Percent

−4 −3 −2 −1 0 1 2

3 4 5

6

7

8 9 10

Distance to Median Income Category (Median = 0) Source: Authors’ calculations based on 1990–2001 World Values Survey and the Luxembourg Income Study (LIS) data set.

Income, Inequality, and Electoral Participation

293

two countries with strictly enforced compulsory voting laws). In contrast, the measure of electoral system proportionality fails to achieve conventional levels of significance. Because they do not constitute the primary focus of the analysis, we will not comment further on the effects of the control variables.

The Effect of Income Inequality on Electoral Participation This argument and the individual-level findings we present have important macrolevel implications—namely, that countries with the most income inequality should have the lowest electoral participation, and that countries where inequality increases over time should experience drops in electoral participation. Although the focus of this chapter is not on systematically investigating these macrolevel dynamics (for such an investigation, see Anderson and Beramendi 2007), a cursory examination of cross-national patterns for the OECD countries and cross-temporal data for the United States supports the contention that greater inequality is indeed associated with lower turnout at the macro level. Data for the United States between 1960 and 2000, for example, illustrate that turnout is negatively related to the overall level of income inequality in the country at the time of the election (figure 9.4).7 This negative relationship appears at the cross-national level as well: figure 9.3 presents average levels of turnout in elections and average levels of inequality throughout the period 1980 to 2002 for sixteen advanced industrialized democracies.8 Both figures reveal that there is in fact a negative relationship between income inequality and electoral turnout, even though in both cases there are several meaningful residuals. While these data are suggestive of a link between inequality and lower participation, we analyzed models of turnout that control for a host of macro- and microlevel factors that are known to shape electoral participation in order to examine more accurately and systematically whether and how aggregate income inequality matters. Specifically, we estimated two-way logistic and multilevel models identical to the models in table 9.1, with the one exception that we now add the instrumented overall inequality variables in the estimation model. To examine the effect of inequality, we conducted two specific tests. First, to see whether inequality influences participation, as suggested in the literature, we examined the impact of overall inequality on electoral participation. Second, we examined whether different kinds of inequality matter differently to participation by substituting inequality in the upper and lower halves of the income distribution for the overall inequality measure. This test was designed to investigate whether inequality that produces ever-richer or ever-poorer individuals in the income distribu-

294

Democracy, Inequality, and Representation

Figure 9.3

Inequality and Turnout in OECD Democracies, 1980 to 2002

100

90

80

70

Italy Belgium Denmark Sweden Australia Austria Netherlands France Ireland Norway Germany United Kingdom Finland Canada

60 United States 50

Switzerland

40

30 0.2

0.25

0.3

0.35

Inequality (Gini Coefficient for Disposable Income) Source: Authors’ calculations based on 1990–2001 World Values Survey and the Luxembourg Income Study (LIS) data set.

tion leads to different participatory outcomes. Put another way, if the results shown in table 9.1 about the linearity of income’s effects are roughly correct, then it should not matter for electoral participation whether inequality is greater at the upper end or the lower end of the income distribution—both should lead to lower turnout to a roughly equal degree. The results of these estimations are shown in table 9.2. They show that inequality does indeed matter. Specifically, more inequality is associated with lower levels of electoral participation. These results are consistent with previous work on the deleterious effects of income inequality on the quality of citizen participation. Moreover, the results show that different kinds of inequality have similar effects: inequality at the upper end of the distribution and inequality at the lower end of the income distribution both enhance abstention.

Income, Inequality, and Electoral Participation Figure 9.4

295

Income Inequality and Turnout in the United States, 1960 to 2000

65

60

55

50

45 0.35

0.37

0.39

0.40

0.43

Inequality (Gini Coefficient) Source: Authors’ calculations based on 1990–2001 World Values Survey and the Luxembourg Income Study (LIS) data set.

To see how much inequality matters we again calculated the substantive effects, which are shown in figures 9.5 to 9.7. Figure 9.5 shows the substantive impact of overall inequality on the probability of abstention, while figures 9.6 and 9.7 show the impact of inequality at the higher and upper ends of the distribution on the odds of abstention, respectively. As inequality goes up, so do the odds of abstention from electoral participation. These effects are unambiguous, and the inference does not depend on what kind of inequality we consider. Moving from the most equal countries in our sample (Sweden and Finland) to the most unequal (the United States) means increasing the odds of abstention by about 14 percent: while the probability of abstention is only about 12 percent in the most equal country, it is roughly 26 percent in the country with the most unequal income distribution. And while the different scales for inequality at the top versus inequality at the bottom make a direct comparison of the direct effects haz-

Interpersonal trust

Life satisfaction

Marital status

Female

Education

Age-squared

Age

P99TOP50

P50TOP10

Overall Gini

Distance to median income-squared

–0.022** (0.008) 0.000 (0.000) –0.063** (0.013) 0.287** (0.045) 0.145** (0.052) –0.012 (0.012) –0.194** (0.049)

–0.064** (0.011) 0.003 (0.003) 16.903** (1.243)

Logit

–0.003* (0.001) 0.000 (0.000) –0.01** (0.002) 0.041** (0.007) 0.023** (0.008) –0.002 (0.002) –0.027** (0.007)

–0.010** (0.002) 0.001 (0.001) 2.97** (0.193)

Multilevel

Effects of Overall Inequality on Electoral Abstention

Distance to median income

Table 9.2

–0.024** (0.008) 0.000 (0.000) –0.054** (0.013) 0.253** (0.045) 0.168** (0.052) –0.024* (0.012) –0.238** (0.049)

0.993** (0.292)

–0.068** (0.011) 0.005 (0.003)

Logit

–0.003** (0.001) 0.000 (0.000) –0.008** (0.002) 0.037** (0.007) 0.027** (0.008) –0.004* (0.002) –0.036** (0.007)

0.198** (0.044)

–0.011** (0.002) 0.001* (0.001)

Multilevel

0.931** (0.100) –0.023** (0.008) 0.000 (0.000) –0.062** (0.013) 0.269** (0.045) 0.163** (0.052) –0.017 (0.012) –0.23** (0.049)

–0.068** (0.011) 0.004 (0.003)

Logit

0.173** (0.016) –0.003* (0.001) 0.000 (0.000) –0.01** (0.002) 0.039** (0.007) 0.026** (0.008) –0.003 (0.002) –0.034** (0.007)

–0.011** (0.002) 0.001 (0.001)

Multilevel

–0.889* (0.406) 13,384 0.09

0.422** (0.042) 0.01 (0.08) –0.024 (0.056) 0.198 (0.113) 0.0004** (0.000) –0.161** (0.022) –0.036** (0.004) –2.29** (0.108) –0.288** (0.083)

0.451** (0.042) 0.033 (0.078) –0.057 (0.056) 0.114 (0.114) 0.0004** (0.000) –0.161** (0.022) –0.013* (0.005) –2.02** (0.150) –0.057 (0.082)

0.074** (0.007) 0.007 (0.013) –0.011 (0.009) 0.029 (0.019) 0.0001** (0.000) –0.026** (0.003) –0.003** (0.001) –0.28** (0.020) –0.011 (0.014)

0.442** (0.042) –0.007 (0.079) –0.05 (0.056) 0.168 (0.113) 0.0004** (0.000) –0.171** (0.021) –0.035** (0.005) –2.117** (0.110) –0.292** (0.085)

0.219** 1.831** 0.614** 0.883* (0.064) (0.420) (0.061) (0.381) 13,384 13,384 13,384 13,384 0.09 0.07 0.07 0.08

0.07** (0.007) 0.007 (0.013) –0.007 (0.009) 0.042* (0.019) 0.0001** (0.000) –0.024** (0.003) –0.007** (0.001) –0.324** (0.013) –0.058** (0.014) 0.499** (0.058) 13,384 0.08

0.073** (0.007) 0.002 (0.013) –0.01 (0.009) 0.037 (0.019) 0.0001** (0.000) –0.027** (0.003) –0.007** (0.001) –0.296** (0.014) –0.057** (0.014)

Source: Authors’ calculations based on 1990–2001 World Values Survey and the Luxembourg Income Study (LIS) data set. Note: Two-way instrumented variable models. Multilevel models are hierarchical linear models with random intercepts estimated with generalized linear latent and mixed models (GLLAMM). Robust standard errors are in parentheses. P50TOP10 = ratio between the fiftieth and tenth income percentiles (equivalized disposable household income). P99TOP50 = the ratio between the ninety-ninth and fiftieth income percentiles (equivalized disposable household income). * p < .05; ** p < .01

Observations Pseudo-R-squared

Constant

Belgium

Australia

Index of disproportionality

Growth 1995 to 1999

GDP per capita

Unemployed

Skilled or semiskilled manual

Unskilled manual

Distrust in institutions

298

Democracy, Inequality, and Representation

Figure 9.5

Aggregate Inequality (Gini) and the Probability of Abstention

0.30 0.25 0.20 0.15 0.10 Abstention Upper Confidence Level 95 Percent Lower Confidence Level 95 Percent

0.05

0. 24 0. 25 0. 26 0. 27 0. 28 0. 29 0. 3 0. 31 0. 32 0. 33 0. 34 0. 35

0

Gini Coefficient (Disposable Income) Source: Authors’ calculations based on 1990–2001 World Values Survey and the Luxembourg Income Study (LIS) data set.

Figure 9.6

Inequality in the Lower Half and the Probability of Abstention

0.5

Abstention Upper Confidence Level 95 Percent Lower Confidence Level 95 Percent

0.4 0.3 0.2 0.1 0

1.2

1.4

1.6

1.8

2

2.2

2.4

2.6

Ratio P50/P10 (Disposable Income) Source: Authors’ calculations based on 1990–2001 World Values Survey and the Luxembourg Income Study (LIS) data set.

Income, Inequality, and Electoral Participation Figure 9.7

299

Inequality in the Upper Half and the Probability of Abstention 0.7

Abstention Upper Confidence Level 95 Percent Lower Confidence Level 95 Percent

0.6 0.5 0.4 0.3 0.2 0.1 0

2.3

3

3.7

4.4

5.1

5.8

6.5

Ratio P99/P50 (Disposable Income) Source: Authors’ calculations based on 1990–2001 World Values Survey and the Luxembourg Income Study (LIS) data set.

ardous, it is clear that both types of inequality reduce electoral participation and that they do so in significant ways. The difference in the odds of abstention between the country with the least unequal income distribution at the lower end of income (France) and the countries with the most unequally distributed income among low-income individuals (Portugal, the United States, and Spain) is 23 percent. Similarly, moving from the country with the most equal distribution among high-income earners (France) to the most unequal one (Portugal) increases the odds of abstention by roughly 35 percent.

The Effects of Income Inequality on Individuals Above and Below the Median So far we have established several key facts. First, the effect of income on electoral participation at the individual level is basically linear and positive. That is, individuals with relatively higher incomes participate more and those with lower incomes relative to the median income participate less in elections. Moreover, we have established that inequality affects participation negatively—and it does so regardless of whether income is more concentrated at the upper or lower end of the income distribution. Thus, inequality increases abstention because it increases the number of

300

Democracy, Inequality, and Representation

relatively less well-to-do people at the high end of income, the low end, or both. If this is true, then income inequality should also have similar effects on individuals, regardless of whether they are among the relatively high or low earners in society. To examine whether income inequality has such an equal effect we split our sample of respondents into those individuals whose income falls below the median income and those whose income falls above the median income in society. We then examined the impact of inequality on the electoral intentions of these two groups. The results, shown in table 9.3 and figures 9.8 and 9.9, reveal that inequality increases both groups’ propensity to abstain and that it does so in a similar fashion, substantively speaking. The results showing the impact of inequality on participation reported so far are quite robust. They show that higher levels of income inequality lead to lower levels of voter participation in elections. The coefficient for inequality is significant and positive, indicating that greater inequality produces more electoral abstention. These results are robust regardless of which estimation technique is employed. To further examine these relations we calculated the substantive effects (shown in figures 9.8 and 9.9) of inequality on electoral participation in the two groups. They reveal that, unsurprisingly, individuals above and below the median income have differential odds of participating overall. However, they also show that inequality has a similar substantive effect of increasing abstention. Moving from the most equal to the least equal country increases the odds of abstention among individuals below the median by 13 percent (from .16 to .29), while it boosts the odds of abstention among individuals with above-median incomes by an almost identical 14 percent (from .09 to .23). Taken together, then, the results show that inequality affects participation patterns in similar ways among those above and below the median income in society in the sense that, although they start from different initial odds of participating, greater inequality reduces voter turnout among individuals in both groups.

Conclusions In this study, we sought to explore both the theoretical and empirical foundations of the relationship between economic and political inequality by focusing on the relationship between individual income, income inequality, and electoral participation. Based on individual and macrolevel data collected in eighteen OECD

Table 9.3

Inequality and Electoral Abstention Above and Below the Median Logit

Overall Gini Age Age-squared Education Female Marital status Life satisfaction Interpersonal trust Distrust in institutions Unskilled manual Skilled or semiskilled manual Unemployed GDP per capita Growth 1995 to 1999 Index of disproportionality Australia Belgium Constant Observations Pseudo-R-squared

Multilevel

Below

Above

Below

Above

7.518** (2.057) –0.007 (0.011) 0.000 (0.000) –0.066** (0.020) 0.274** (0.068) 0.19** (0.070) –0.015 (0.016) –0.202** (0.073) 0.318** (0.058) 0.101 (0.100) –0.035 (0.077) 0.313* (0.137) –0.003** (0.000) –0.182** (0.037) –0.019** (0.006) –1.825** (0.159) –0.161 (0.127)

9.643** (1.793) –0.037* (0.016) 0.000 (0.000) –0.071** (0.019) 0.353** (0.070) 0.023 (0.086) –0.023 (0.022) –0.255** (0.074) 0.612** (0.071) –0.145 (0.160) 0.017 (0.098) 0.312 (0.247) –0.003** (0.000) –0.153** (0.029) –0.02** (0.007) –2.114** (0.174) –0.307* (0.126)

1.606** (0.338) –0.001 (0.002) 0.000 (0.000) –0.011** (0.003) 0.046** (0.012) 0.032** (0.012) –0.003 (0.003) –0.032** (0.012) 0.057** (0.010) 0.018 (0.018) –0.007 (0.013) 0.065* (0.025) –0.001** (0.000) –0.031** (0.006) –0.004** (0.001) –0.274** (0.022) –0.031 (0.024)

1.738** (0.256) –0.005* (0.002) 0.000 (0.000) –0.01** (0.003) 0.046** (0.010) 0.004 (0.012) –0.004 (0.003) –0.034** (0.010) 0.087** (0.010) –0.015 (0.023) 0.002 (0.014) 0.051 (0.037) 0.004** (0.000) –0.022** (0.004) –0.004** (0.001) –0.26** (0.018) –0.056** (0.019)

1.231 (0.631) 5,683 0.07

0.503 0.55** (0.689) (0.104) 6,123 5,683 0.08 0.07

0.369** (0.092) 6,123 0.08

Source: Authors’ calculations based on 1990–2001 World Values Survey and the Luxembourg Income Study (LIS) data set. Note: Two-way instrumented variable models. Multilevel models are hierarchical linear models with random intercepts estimated with generalized linear latent and mixed models (GLLAMM). Robust standard errors are in parentheses. * p < .05; ** p < .01

302

Democracy, Inequality, and Representation

Figure 9.8

The Impact of Aggregate Inequality on the Probability of Abstention (Individuals Below the Median) 0.4 0.3 0.2 0.1

35

34

0.

33

0.

32

0.

0.

31

3

0.

0.

28

0.

27

0.

0.

26

25

0.

0.

0.

24

0

29

Abstention Upper Confidence Level 95 Percent Lower Confidence Level 95 Percent

Aggregate Inequality (Gini Coefficient for Disposable Income) Source: Authors’ calculations based on 1990–2001 World Values Survey and the Luxembourg Income Study (LIS) data set.

Figure 9.9

The Impact of Aggregate Inequality on the Probability of Abstention (Individuals Above the Median)

0.3

0.2

0.1 Abstention Upper Confidence Level 95 Percent Lower Confidence Level 95 Percent 35 0.

34 0.

33 0.

32 0.

31 0.

3 0.

29 0.

28 0.

27 0.

26 0.

25 0.

0.

24

0 Aggregate Inequality (Gini Coefficient for Disposable Income) Source: Authors’ calculations based on 1990–2001 World Values Survey and the Luxembourg Income Study (LIS) data set.

Income, Inequality, and Electoral Participation

303

democracies, we find that income inequality at the macro level diminishes electoral participation. At the level of individual citizens, we find that the effects of income differentials are basically linear: individuals in the lower half of the income distribution are less likely to participate in elections, while those in the upper half are more likely to do so. We also find that income inequality decreases electoral turnout, and that it does so regardless of whether we consider overall inequality or inequality that is concentrated at the upper or lower end of society. Thus, income inequality reduces electoral participation, and this effect is of similar magnitude for differently situated individuals. At the same time, we would argue that it is critical to differentiate the impact of income as a resource versus the role that income plays as an incentive to get involved. Specifically, we argue that the mechanism by which inequality affects participation differs between high- and low-income individuals: while inequality at the lower end reduces participation among low-income individuals because it deprives them of resources, higher inequalities at the upper end reduce participation by creating disincentives for very rich people to get involved. In this chapter, we do not directly examine these mechanisms, but our inferences are consistent with the evidence we present. We suspect that these mechanisms help to produce the results we see, in large measure because of the connection between egalitarian income distributions and welfare state regimes, and in particular the role of decommodification (Esping-Andersen 1990) in shaping the connection between income and participation. Decommodification pertains to access and eligibility rules, income replacement levels, and the range of protection against social risks (Scruggs and Allan 2003). In countries with high levels of decommodification, individuals are less dependent on the market and the cash nexus. Decommodification means that the state guarantees social rights, that the commodity status of individuals is reduced, and that individuals receive benefits as a matter of need. Decommodification helps overcome the resources problem at the bottom end of the income distribution and creates incentives to mobilize at the very top. Thus, decommodification has comparable effects on people all across the income spectrum, but for different reasons. As a result, countries that decouple the connection between work and income have higher rates of political participation regardless of citizens’ level of income. Taken together, then, our results suggest that low- and middle-income individuals’ decision to participate is driven by resources. Moreover, as income goes up, participation becomes affordable—and provided that there are incentives, higher-income individuals are therefore more likely to participate. Yet when income is extremely high, there emerge disincentives to participate given that there are too many ways for individuals

304

Democracy, Inequality, and Representation

in the highest income categories to avoid the consequences of policies to care sufficiently to be involved in the electoral process. Although we would expect this reaction only from individuals at the very top of the income distribution, this expectation is consistent with Jan van Deth’s (2000) finding that politics is much less salient for high-income individuals (see, however, Rogowski and MacRae, this volume, on high-income individuals’ incentives to change a country’s electoral rules). It also is clear that inequality matters for participation and that it matters similarly across all levels of income. Thus, even if the operating mechanisms are different, the direction of the effects is similar across individuals with different levels of income. Put simply, unequal resources promote abstention among those in the lower half of the income distribution, whereas disincentives to engage politically foster abstention among those in the upper half. If this is true, an increase in the levels of any kind of inequality should be reflected in lower levels of electoral participation—and this is exactly what we find. In the end, then, welfare states affect the quality of democracy by shaping the connection between income and the quantity of democratic participation, and they do so for people at all ends of the income distribution.

Appendix Measures and Coding Voting abstention: “If there were a national election tomorrow, for which party on this list would you vote?” 1: Would not vote; 0: all others. Female: 1: male; 2: female. Age: Actual age. Age-squared: Actual age squared. Education: “What is the highest educational level that you have attained?” Eight-point scale. Unemployed: “Are you employed now or not?” 1: not employed; 0: all others. Married: 1: married; 0: all others. Income distance: “Here is a scale of incomes. We would like to know in what group your household is, counting all wages, salaries, pensions, and other incomes that come in. Just give the letter of the

Income, Inequality, and Electoral Participation

305

group your household falls into, before taxes and other deductions.” 1: lowest; 10: highest. Income categories were coded by deciles for each society and then recoded to reflect the individual’s distance from the country’s median income category. Distrust institutions: “I am going to name a number of organizations. For each one, could you tell me how much confidence you have in it: is it a great deal of confidence (1), quite a lot of confidence (2), not very much confidence (3), or none at all (4)?” Index based on average ratings of confidence in parliament, police, justice system, civil service, and army. Cronbach’s alpha: .72. Life satisfaction: “All things considered, how satisfied are you with your life as a whole these days? Please use this card to help with your answer.” On a 1-to-10 scale, 10 = completely satisfied. Interpersonal trust: “Generally speaking, would you say that most people can be trusted or that you can’t be too careful in dealing with people?” 1: “most people can be trusted”; 0: “can’t be too careful” (translation: “have to be very careful”). Disproportionality: Gallagher least index of electoral system disproportionality measured by the square root of half the sum of the squares of the difference between percentage of vote and percentage of seats for each party. For two most recent legislative elections. Source: Michael McDonald, Binghamton University. Level of economic development: GDP per capita in current international dollars (in thousands), purchase power parities at the time of the survey. Source: World Bank (2003). Economic growth: Annual percentage of the GDP per capita growth at the time of the survey. Source: World Bank (2003). Income inequality: Gini coefficient of the disposable household income per equivalent adult. Average of country scores based on waves 4 (around 1995) and 5 (around 2000, if available) of the LIS data set. Data are adjusted using the LIS equivalence scale. Range: 0 to 1. Decommodification: Source of country scores: Scruggs (2004). Compulsory voting: Indicates countries with compulsory voting laws that are strictly enforced. 1: Australia and Belgium; 0: all others. Source: Gratschew (2001).

306

Democracy, Inequality, and Representation

Table 9A.1

Inequality of Disposable Income

Country Australia Austria Belgium Canada Denmark Finland France Germany Greece Ireland Italy Netherlands Norway Portugal Spain Sweden United States United Kingdom

Gini

50/10

99/50

0.310 0.293 0.284 0.301 0.266 0.240 0.310 0.260 0.320 0.331 0.345 0.264 0.250 0.320 0.305 0.239 0.371 0.350

2.157 1.849 1.816 2.152 2.011 1.630 1.186 1.825 2.483 2.137 2.371 1.821 1.691 2.668 2.567 1.656 2.617 2.142

3.252 3.169 3.219 3.420 3.700 2.992 2.357 3.219 5.282 4.369 4.645 2.928 3.019 7.031 4.971 2.841 5.511 4.552

Source: Authors’ calculations based on the Luxembourg Income Study (LIS) data set.

We are grateful to the participants in the Conference on Democracy, Inequality, and Representation held at the Maxwell School of Syracuse University, as well as the anonymous reviewers for their many thoughtful comments on the earlier draft. Many thanks also to Michael McDonald and Lyle Scruggs for help with obtaining some of the data. The survey data are available as ICPSR study 3975. The original collector of the data, the Interuniversity Consortium for Political and Social Research (ICPSR), and the relevant funding agency bear no responsibility for uses of this collection or for interpretations or inferences based on such uses.

Notes 1.

Some researchers have found that the positive effect of education on political support is less consistent than portrayed by Almond and Verba (1963). Thus, Ola Listhaug and Matti Wiberg (1995) found, for example, that the effects of education on support for “order” institutions (church, armed forces, police) are frequently negative, while the relationship with support for parliament is

Income, Inequality, and Electoral Participation

2.

3. 4.

5.

6. 7.

8.

307

mostly positive. In addition, there is a new politics argument that high levels of education can in fact lead to critical attitudes and political dissatisfaction (Dalton 2002). The traditional literature on public goods has looked at the impact of economic heterogeneity on aggregate contributions, hence on efficiency in public good provision. The so-called neutrality theorem by Ted Bergstrom and his colleagues (1986) shows that under the assumption that everyone consumes the public good in the same amount and that individual contributions are perfect substitutes in its production (that is, in the case of a “pure” public good), and provided that the set of people contributing to the good does not change, small redistributions of income should not affect the aggregate provision of public goods. The reason for this “neutrality” of redistributive policies is that individuals should change their contributions accordingly so that aggregate provision is unchanged. For details, see the Luxembourg Income Study website, http://www.lis project.org. We did not top-code the income distributions (trim the distribution at the upper extreme to eliminate extreme outliers), since the samples had only small numbers of extreme outliers. We therefore had no reason to assume that the results would be sensitive to this. Note that the income variable does not measure income (y) but rank in income (F(y)), and rank in income is not necessarily as linear as income (expressed in currency units), though we expect the correlation to be very high. These were calculated holding all other variables at their mean. Turnout is measured in presidential election years. Income inequality is measured by the Gini coefficient for family income as reported by the March Current Population Survey (CPS). The Pearson correlation between turnout and inequality is –.66. As measured by the Gini coefficient for disposable income inequality reported by LIS (key figures: http://www.lisproject.org). The Pearson correlation for the relationship between turnout and inequality is −.43. The most notable outlier is Switzerland, where turnout is lower than its level of inequality would suggest. According to Mark Franklin (2004, 95–98), turnout rates in Switzerland dropped significantly after the 1963 introduction of an executive “cartel” where parties rotate to control policy in the executive. As policy became more independent from the composition of the legislature and executive responsiveness declined, voters’ incentives to vote in order to choose members of parliament were significantly reduced.

References Aguilar, Edwin E., and Alexander C. Pacek. 2000. “Macroeconomic Conditions, Voter Turnout, and the Working-Class/Economically Disadvantaged

308

Democracy, Inequality, and Representation

Party Vote in Developing Countries.” Comparative Political Studies 33(8): 995– 1017. Almond, Gabriel A., and Sidney Verba. 1963. The Civic Culture. Boston, Mass.: Little, Brown. Anderson, Christopher J., and Pablo Beramendi. 2007. “Income Inequality and Voter Turnout in Comparative Perspective.” Working paper. Ithaca, N.Y.: Department of Government, Cornell University. Bahry, Donna, and Christine Lipsmeyer. 2001. “Economic Adversity and Public Mobilization in Russia.” Electoral Studies 20(3): 371–98. Baland, Jean-Marie, and Jean-Philippe Platteau. 1997. “Wealth Inequality and Efficiency in the Commons; Part I: The Unregulated Case.” Oxford Economic Papers 49(4): 451–82. Baland, Jean-Marie, and Debraj Ray. 1997. “Technical Complementarities and the Impact of Inequality on Collective Action: An Application to the Commons.” Unpublished paper, University of Namur. Barnes, Samuel, and Max Kaase, with Klaus Allerbeck, Barbara Farah, Felix Heunks, Ronald Inglehart, M. Kent Jennings, Hans-Dieter Klingemann, Alan Marsh, and Leopold Rosenmayr. 1979. Political Action: Mass Participation in Five Western Democracies. Beverly Hills, Calif.: Sage Publications. Bartels, Larry. 2002. “Economic Inequality and Political Representation.” Paper presented to the annual meeting of the American Political Science Association. Boston, Mass., September 2002. Beramendi, Pablo, and Thomas R. Cusack. 2004. “Diverse Disparities: The Politics and Economics of Wage, Market, and Disposable Income Inequalities.” Discussion paper 2004-08. Berlin: Social Science Research Center Berlin (WZB). Bergstrom, Ted, Larry Blume, and Hal Varian. 1986. “On the Private Provision of Public Goods.” Journal of Public Economics 29(1): 25–49. Boix, Carles. 2003. Democracy and Redistribution. Cambridge: Cambridge University Press. Brady, Henry, Sidney Verba, and Kay Schlozman. 1995. “Beyond SES: A Resource Model of Political Participation.” American Political Science Review 89(2): 271–94. Brehm, John, and Wendy Rahn. 1997. “Individual-Level Evidence for the Causes and Consequences of Social Capital.” American Journal of Political Science 41(3): 999–1023. Chapman, Randall G., and Kristian S. Palda. 1983. “Electoral Turnout in Rational Voting and Consumption Perspectives.” Journal of Consumer Research 9(4): 337–46. Curtis, James E., Douglas E. Baer, and Edward G. Grabb. 2001. “Nations of Joiners: Explaining Voluntary Association Membership in Democratic Societies.” American Sociological Review 66(6): 783–805.

Income, Inequality, and Electoral Participation

309

Dahl, Robert Alan. 1971. Polyarchy, Participation, and Opposition. New Haven, Conn.: Yale University Press. Dalton, Russell J. 2002. Citizen Politics: Public Opinion and Political Parties in Advanced Western Democracies. 3rd ed. Chatham, N.J.: Chatham House. Esping-Andersen, Gøsta. 1990. Three Worlds of Welfare Capitalism. Princeton, N.J.: Princeton University Press. Franklin, Mark N. 2004. Voter Turnout and the Dynamics of Electoral Competition in Established Democracies Since 1945. New York: Cambridge University Press. Franzese, Robert. 1998. “Political Participation, Income Distribution, and Public Transfers in Developed Democracies.” Paper presented to the annual meeting of the American Political Science Association, Boston, Mass., September 3–6, 1998. Goodin, Robert, and John Dryzek. 1980. “Rational Participation: The Politics of Relative Power.” British Journal of Political Science 10(3): 273–92. Gratschew, Maria. 2001. “What Is Compulsory Voting?” (April). Accessed at International Institute for Democracy and Electoral Assistance (IDEA), http:// www.idea.int/vt/compulsory_voting.cfm. Hansen, Susan B. 1997. “Talking About Politics: Gender and Contextual Effects on Political Proselytizing.” Journal of Politics 59(1): 73–103. Jackman, Robert W. 1987. “Political Institutions and Voter Turnout in the Industrial Democracies.” American Political Science Review 81(2): 405–23. Jennings, M. Kent. 1983. “Gender Roles and Inequalities in Political Participation: Results from an Eight-Nation Study.” Western Political Quarterly 36(3): 364–85. Kenworthy, Lane, and Jonas Pontusson. 2005. “Rising Inequality and the Politics of Redistribution in Affluent Countries.” Perspectives on Politics 3(3): 449–71. La Ferrara, Eliana. 2002. “Inequality and Group Participation: Theory and Evidence from Rural Tanzania.” Journal of Public Economics 85(2): 235–73. Leighley, Jan. 1995. “Attitudes, Opportunities, and Incentives: A Field Essay on Political Participation.” Political Research Quarterly 48(1): 181–209. Listhaug, Ola, and Matti Wiberg. 1995. “Confidence in Political and Private Institutions.” In Citizens and the State, edited by Hans-Dieter Klingemann and Dieter Fuchs. New York: Oxford University Press. Mahler, Vincent. 2002. “Exploring the Subnational Dimension of Income Inequality: An Analysis of the Relationship Between Inequality and Electoral Turnout in the Developed Countries.” International Studies Quarterly 46(1): 117–42. Mincer, Jacob. 1962. “Labor Force Participation of Married Women: A Study of Labor Supply.” In Aspects of Labor Economics, edited by National Bureau of Economic Research. Princeton, N.J.: Princeton University Press. Mueller, Dennis C., and Thomas Stratmann. 2003. “The Economic Effects of Democratic Participation.” Journal of Public Economics 87(9–10): 2129–55.

310

Democracy, Inequality, and Representation

Nie, Norman H., Sidney Verba, and Jae-on Kim. 1974. “Political Participation and the Life Cycle.” Comparative Politics 6(3): 319–40. Norris, Pippa. 2004. Electoral Engineering: Voting Rules and Political Behavior. New York: Cambridge University Press. Olson, Mancur. 1965. The Logic of Collective Action. Cambridge, Mass.: Harvard University Press. Pacek, Alexander, and Benjamin Radcliff. 1995. “Economic Voting and the Welfare State: A Cross-National Analysis.” Journal of Politics 57(1): 44–61. Pateman, Carole. 1971. “Political Culture, Political Structure, and Political Change.” British Journal of Political Science 1(3): 291–305. Pontusson, Jonas, David Rueda, and Christopher Way. 2002. “Comparative Political Economy of Wage Distribution.” British Journal of Political Science 32(2): 281–308. Rabe-Hesketh, Sophia, and Anders Skrondal. 2005. Multilevel and Longitudinal Modeling Using Stata. College Station, Tex.: Stata Press. Radcliff, Benjamin. 1992. “The Welfare State, Turnout, and the Economy: A Comparative Analysis.” American Political Science Review 86(2): 444–56. Rueda, David, and Jonas Pontusson. 2000. “Wage Inequality and Varieties of Capitalism.” World Politics 52(3): 350–83. Schattschneider, E. E. 1960. The Semisovereign People: A Realist’s View of Democracy in America. Hinsdale, Ill.: Dryden Press. Schwabish, Jonathan, Timothy Smeeding, and Lars Osberg. 2003. “Income Distribution and Social Expenditures.” Luxembourg Income Study working paper 350. Syracuse, N.Y.: Syracuse University, Maxwell School. Scruggs, Lyle. 2004. Welfare State Entitlements Data Set: A Comparative Institutional Analysis of Eighteen Welfare States. Version 1.0. Accessed at http://sp.uconn.edu/ ~scruggs/wp.htm. Scruggs, Lyle, and James Allan. 2003. “Trends in Welfare State Decommodification in Eighteen Advanced Industrial Democracies, 1972–2000.” Paper presented to the annual meeting of the American Political Science Association. Philadelphia, Pa., August 29–31, 2003. Sharp, Clifford. 1981. The Economics of Time. Oxford: Martin Robinson. Solt, Frederick. 2004. “Civics or Structure? Revisiting the Origins of Democratic Quality in the Italian Regions.” British Journal of Political Science 34(1): 123–35. Teorell, Jan, Mariano Torcal, and Joseì Ramoìn Montero. 2007. “Political Participation: Mapping the Terrain.” In Citizenship and Involvement in European Democracies, edited by Jan W. van Deth, José Ramón Montero, and Anders Westholm. London: Routledge. Van Deth, Jan W. 2000. “Interesting but Irrelevant: Social Capital and the Saliency of Politics in Western Europe.” European Journal of Political Research 37(2): 115–47. Van Deth, Jan W., and Martin Elff. 2004. “Politicization, Economic Development,

Income, Inequality, and Electoral Participation

311

and Political Interest in Europe.” European Journal of Political Research 43(3): 477–508. Verba, Sidney, and Norman H. Nie. 1972. Participation in America. New York: Harper & Row. Verba, Sidney, Kay L. Schlozman, and Henry Brady. 1995. Voice and Equality: Civic Voluntarism in American Politics. Cambridge: Cambridge University Press. World Bank. 2003. World Development Indicators (CD-ROM).

Chapter 10

Inequality as a Source of Political Polarization: A Comparative Analysis of Twelve OECD Countries JONAS PONTUSSON AND DAVID RUEDA

This chapter focuses on the effects of income inequality on party politics in industrialized democracies. Having devoted a great deal of attention to the political determinants of income distribution in the 1990s, students of comparative political economy have recently begun to address how the distribution of income affects politics and, in particular, government policy (see, for example, Bradley et al. 2003; Kenworthy and Pontusson 2005; Mahler 2006; Moene and Wallerstein 2001, 2003). To date, virtually all the comparative literature on this topic takes the Meltzer-Richard model as its point of departure and investigates the association between inequality and various measures of redistributive government spending (Meltzer and Richard 1981). A common conclusion in the literature is that the core proposition of the Meltzer-Richard model—that inequality generates more redistributive government—provides precious little leverage, if any at all, on the problem of explaining why some countries have more redistributive welfare states than others. Theoretically, we seek to break new ground by elaborating a partisan model of the political effects of inequality that abandons the MeltzerRichard premise that the preferences of the median voter determine party policy. In our analytical framework, parties of the left and the right draw their core constituencies from different segments of the income distribution, and inequality affects the policy preferences of these constituencies differently. In its simplest version, our model predicts that core left voters want more redistribution and core right voters want less redistribution as inequality rises. This main argument, however, is

Inequality as a Source of Political Polarization

313

greatly affected by two significant factors: the kind of inequality in question and the degree of mobilization among low-income workers. Our empirical analysis seeks to explain party positions in electoral campaigns, as measured by the Comparative Manifesto Project (CMP), rather than policy outputs. To some significant extent, using election manifestos to measure party positions allows us to bracket the economic and bureaucratic constraints that parties inevitably face in government and thus to focus more directly on party responses to (changes in) voter preferences. In contrast to Meltzer and Richard, we do not assume that voting alone determines government policy. The motivation behind this chapter partly derives from Nolan McCarty, Keith Poole, and Howard Rosenthal’s (2006) analysis of the recent polarization of American politics. These authors document that partisanship in congressional roll-call voting declined in the 1950s, held steady through most of the 1960s and 1970s, and then increased sharply from the late 1970s onwards. They demonstrate that this pattern parallels trends in income distribution in a very striking manner (and also that income has become a better predictor of individual party choice as inequality has increased over the last three decades). Polarization can take several different forms. If left parties move to the left and right parties move to the right, we observe what we here refer to as “symmetric polarization.” If right parties move to the right while left parties stay put, or if both parties move to the right but right parties move farther to the right than left parties, we observe “right-skewed polarization.” Conversely, “left-skewed polarization” represents a third potential scenario. To distinguish among these alternative scenarios we estimate the effects of inequality on left-right positions adopted by the main parties of the left and the right in each of the twelve countries included in our analysis.1 Theoretically and empirically, we distinguish between the partisan effects of wage inequality and those of other forms of income inequality. The core constituencies of left and right parties are distinguished from each other not only by where they fall in the overall income distribution but also by the sources of their income. We argue that left parties are particularly responsive to wage inequality because their core constituencies consist of voters who derive the lion’s share of their income from dependent employment. As wages account for a considerably smaller share of total income among the core constituencies of right parties, these parties should be more responsive to other manifestations of inequality. We argue further that political mobilization of low-income groups, measured by aggregate voter turnout and union density, conditions partisan responses to inequality. To anticipate, the results reported here indicate that wage inequality is

314

Democracy, Inequality, and Representation

associated with more leftist left parties at medium and high levels of lowincome mobilization, while there is no significant association between wage inequality and the policy positions held by right parties. By contrast, household disposable income inequality is associated with more rightist right parties at low and medium levels of low-income mobilization, while there is no significant association between household income inequality and the policy positions held by left parties. It should be noted at the outset that our regression analysis controls for the center of political gravity. It is commonplace to observe that the entire political spectrum is farther to the left or, alternatively, that redistributive policies are more generally accepted in some countries (say, Sweden) than in others (say, the United States). It is also commonplace to observe that politics in most industrialized countries shifted to the right and that redistributive policies became more contested in the 1980s and 1990s. For reasons that we elaborate later, we do not believe that these broad cross-national differences and trends can be explained in terms of contemporary income distribution patterns. To discern the common and significant political effects of inequality, we must control for the center of political gravity in different countries and different years. The next section articulates our theoretical framework. We then discuss the data set we have constructed to test our hypotheses and specify how the variables employed in our regression analysis are measured. The next section presents and discusses the results, and the following section explores patterns in the data that pertain to changes in inequality and partisan politics in specific countries over the 1980s and 1990s. We conclude with some thoughts about the implications of our analysis and directions for future research.

The Theoretical Framework We begin by recapitulating the core elements of the well-known MeltzerRichard model and then transform it to integrate some partisan considerations. We then introduce the idea that different forms of inequality have different implications for parties of the left and right, and we develop the argument that the political mobilization of low-income voters, as measured by voter turnout and unionization, conditions partisan responses to inequality. Finally, we restrict the scope of our model of redistributive politics by arguing that preferences for redistribution shape the spread of party positions around a median position that is determined by a complex combination of historical factors and cannot be derived from contemporary preferences for redistribution.

Inequality as a Source of Political Polarization

315

Meltzer-Richard with Core Constituencies The Meltzer-Richard model is inspired by Downsian median voter theory. Like other median voter models, it assumes that parties are motivated by winning elections and have no enduring commitments to particular policies. In a two-party system, winning elections requires the support of the median voter, and as a result, parties converge on her preferences in their actual behavior in government as well as their election promises. In multiparty systems, the influence of the median voter on government policy is mediated by interparty bargaining, but the party that represents her can be expected to determine the composition and policies of coalition governments (Powell 2000). The contribution of the Meltzer-Richard model is its focus on redistribution and the way it conceives of the median voter’s preferences. The model assumes that government redistribution takes the form of a flatrate (lump-sum) benefit received by all citizens and financed by a proportional (linear) income tax (see Romer 1975). At 100 percent taxation, all citizens are brought to the mean income. Citizens with market incomes below the mean income would favor 100 percent taxation if it were not for the fact that taxation entails a disincentive effect that reduces the mean income. As a result of this disincentive effect, there is a middle group of income earners for whom the deadweight costs of taxation exceed the value of the benefits provided by the government, even though their market income is below the mean income. Holding the deadweight costs of taxation constant, the amount of redistribution preferred by the median voter in the Meltzer-Richard framework becomes a function of the distance between her market income and the average income. Because a few individuals have very large incomes, the distribution of income in capitalist societies is invariably skewed such that the average is higher than the median, but the degree of skew, and therefore the distance between the median and the mean, varies. Figure 10.1 illustrates this point with reference to two hypothetical countries with the same mean income. Country B has a more inegalitarian income distribution than country A, and as a result, the distance between the mean and the median incomes is greater (d2 > d1). By the logic of the Meltzer-Richard model, we would expect the median income earner to want more redistribution in country B than in country A, and this preference should translate into government policy. Our own theoretical framework shares some of the core assumptions (and limitations) of the Meltzer-Richard model. In particular, we incorporate the idea that voters’ preferences for redistribution are determined,

316

Democracy, Inequality, and Representation

Figure 10.1

Illustration of the Metzer-Richard Model

f(x) Country A

d1

Income Median (B)

Mean

f(x) Country B

d2

Income Median (A) Source: Authors’ compilation.

Mean

Inequality as a Source of Political Polarization

317

at least in part, by the distance between their income and the mean income. At the same time, we introduce a number of considerations that make for a model of redistributive politics that is certainly more complex but also, we believe, more realistic and more interesting than the rather barren model proposed by Meltzer and Richard. To begin with, we depart from the Meltzer-Richard model by positing that parties of the left and the right have core constituencies to which they are historically and ideologically committed as well as organizationally tied. In emphasizing core constituencies and enduring policy commitments, we draw on an extensive literature in comparative political economy that identifies partisan effects on macroeconomic policy and social spending (see, for example, Garrett 1998; Hibbs 1987).2 We also draw on the literature on political behavior and electoral competition that suggests that it is more accurate to conceive of parties as programmatic organizations with well-developed ties to particular social groups. In Bingham Powell’s words, the existence of a relationship between “strong, continuing expectations about parties and the interests of social groups not only creates easily identifiable choices for citizens, it also makes it easier for parties to seek out their probable supporters and mobilize them at election time” (1982, 116). Though we are not aware of any systematic comparative study of this question, a great deal of country-specific evidence indicates that left parties draw more of their support from the lower half of the income distribution than right parties do. McCarty, Poole, and Rosenthal (2006) showed that this is the case for the United States, where class and income have arguably played a less important role in structuring party politics than in any other advanced capitalist country. There is good reason to suppose, then, that in the countries included in our analysis the income of the median left voter is typically lower than the mean income overall, while the income of the median right voter is higher than the mean income. If these conditions hold, a partisan version of the Meltzer-Richard model readily suggests itself, with the preferences for redistribution of median left and right voters being determined by the distance between their income and the mean income. The further the income of the median left voter is from the mean, the more she stands to gain from redistribution. On the other hand, the further the income of the median right voter is from the mean, the more she stands to lose from redistribution. Thus, we might expect that greater inequality, illustrated by the shift from country A to country B in figure 10.2, generates partisan polarization over redistributive policy. The proposition that the median right voter wants less redistribution as inequality rises may seem odd, for in the Meltzer-Richard framework someone with an income above the mean always wants zero redistribu-

318

Democracy, Inequality, and Representation

Figure 10.2

Illustration of Our Model

f(x) Country A

Left A

Income Right A

Mean f(x) Country B

Income Left B Mean Source: Authors’ compilation.

Right B

Inequality as a Source of Political Polarization

319

tion. Still, it should be evident that the amount of income loss that a given redistributive scheme entails for such a person increases as the distance to the mean increases. Within the Meltzer-Richard framework, we might say that the intensity of the preference for zero redistribution increases with inequality. The willingness of someone in, say, the seventieth percentile of the income distribution to devote more money or effort to defeating redistributive proposals should increase with inequality. Put differently, the importance that such a person assigns to zero redistribution, relative to other policy preferences, should increase with inequality. We do not mean to suggest that parties are oblivious to the preferences of the median voter. Following Kaare Strom (1990), among others, we assume that parties are motivated by winning elections and, at the same time, by serving the interests of their core constituencies. These objectives are inextricably linked, though they may well pull parties in opposite directions at any given juncture. On the one hand, parties that never win elections or influence government are of little use to their core constituencies. On the other hand, the enthusiasm of party activists and the support of interest organizations matter greatly to voter mobilization. The bottom line here is that the “preferences of the median voter” are hardly exogenous to the dynamics of electoral competition and mobilization: who the median voter is depends on the success of parties in mobilizing citizens to vote. In our conceptualization, parties are constantly engaged in balancing the preferences of core voters against the preferences of swing voters (Aldrich 1995).

Different Forms of Inequality In the Meltzer-Richard model and in the literature that it has inspired, income inequality is conceived as an essentially homogenous phenomenon that can be captured by a single parameter—such as, for example, the Gini coefficient. By contrast, we hypothesize that different forms of income inequality have different political effects. This idea is closely related to that of parties of the left and the right being tied to different social groups. In our conceptualization, the core constituencies of these parties are distinguished from each other not only by where they fall in the overall income distribution, as indicated earlier, but also by the sources of their income. Quite simply, we postulate that the core constituencies of left parties consist of voters who derive the lion’s share of their income from dependent employment. By comparison, wages account for a considerably smaller share of total income among core constituencies of right parties, which include the self-employed and individuals with substantial real and financial assets. (Note that we do not consider the poor, who derive

320

Democracy, Inequality, and Representation

much of their income from government transfers, to be a core constituency of either the left or the right). As a result of these differences in the makeup of their core constituencies, it seems reasonable to suppose that left parties are particularly responsive to wage inequality while right parties are more responsive to other manifestations of income inequality. Empirically, we explore this intuition by estimating models that include measures of both wage inequality among full-time employees and disposable household income inequality. To be more specific, we expect wage inequality to be associated with left parties that advocate more strongly for redistribution, and we expect household income inequality to be associated with stronger opposition to redistribution from right parties. Controlling for wage inequality, we do not expect to observe any effects of household income inequality on the positions adopted by left parties. Similarly, we do not expect to observe any effects of wage inequality on the positions of right parties so long as we control for household income inequality. Wage inequality and household income inequality tend to move in tandem, but the extent to which this is so varies across countries. In some countries, growing wage inequality has been the principal source of increasing household income inequality over the last two or three decades. In other countries, however, household inequality has grown while wage inequality has remained relatively stable (see Kenworthy and Pontusson 2005). Everything else being equal, the argument sketched so far leads us to expect rising inequality to be associated with left-skewed polarization where (and when) it has primarily occurred through wage dispersion and to be associated with right-skewed polarization where (and when) it has primarily occurred through other mechanisms.

Low-Income Mobilization Our model posits further that partisan responses to wage and household income inequality are conditioned by income differentials in political participation. As Meltzer and Richard (1981) recognized, their prediction that inequality will be associated with more redistribution rests on the unrealistic assumption that all income earners vote. Under any other circumstance, testing the Meltzer-Richard model requires us to distinguish between the income of the median voter and the median income (Barnes 2006; Nelson 1999). The discrepancy between the two is particularly pronounced in the United States, not only because of low voter turnout but also because many low-income earners are not citizens (McCarty, Poole, and Rosenthal 2006, chap. 4). With reference to figure 10.1, the point here is the following: the Meltzer-Richard model predicts that a

Inequality as a Source of Political Polarization

321

shift from the income distribution of country A to that of country B will generate more redistribution, but it could well be the case that this rise in income inequality is associated with an increase in the inequality of voting. If citizens with low income disproportionately drop out of the political process, increased income inequality will not necessarily translate into an increase in the distance between the median voter and the mean income (see also Anderson and Beramendi, this volume). Income skew in voting is bound to diminish as aggregate voter turnout approaches 100 percent; as Vincent Mahler (2006) demonstrated, these two factors are indeed closely correlated on a cross-national basis. Like much of the existing literature, we conceive of aggregate voter turnout as a proxy measure for income skew in voting. However, we do not believe that aggregate voter turnout alone suffices to explain variation in the extent to which parties pay attention to the preferences of potential low-income voters. In the comparative political economy literature, organized labor is commonly considered a political force that promotes redistribution by mobilizing workers who stand to benefit from it. The extent to which unions organize and represent low-income workers varies across countries and over time. As more encompassing union movements reach into the upper half of the wage distribution, the political effects of their increased mobilizing capacity may well be offset by the rising heterogeneity of the interests they represent. Still, it seems reasonable to assume that the income of the median union member falls below the income of the median voter in the electorate as a whole and that unionization makes low-income voters more aware of their relative income and more likely to participate in politics. As Jonas Pontusson and Heyok Yong Kwon’s (2006) analysis of individual-level survey data demonstrates, union membership is associated with stronger preferences for redistribution (see also Kumlin and Svallfors 2007). There are two alternatives regarding the mechanisms whereby lowincome mobilization conditions partisan responses to inequality. One way in which low-income mobilization may affect partisan electoral strategies is through the composition and preferences of the core constituencies of left parties. The other is through a direct effect on the median voter, shifting her to a more pro-redistributive position and thereby affecting the strategic behavior of both right and left parties. We treat the choice between these alternatives as an essentially empirical matter. If the first mechanism is more influential than the second, we should observe that the association between wage inequality and pro-redistributive left-party positions becomes stronger at higher levels of low-income mobilization, while the association between household income inequality and antiredistributive right-party positions is unaffected by the level of low-

322

Democracy, Inequality, and Representation

income mobilization. If the second mechanism dominates the first, we should observe not only that the association between wage inequality and pro-redistributive left-party positions becomes stronger at higher levels of low-income mobilization, but also that the association between household income inequality and anti-redistributive right-party positions becomes weaker at higher levels of low-income mobilization. Low-income mobilization affects our hypotheses about the kind of polarization that inequality produces. To illustrate, suppose that wage inequality and disposable household income inequality both rise significantly in a setting characterized by high mobilization. If the effects of mobilization are specific to left parties, we would expect this scenario to translate into symmetric polarization, with left parties responding to wage inequality by adopting more leftist positions and right parties responding to household income inequality by adopting more rightist positions. If high mobilization instead makes both parties more “leftist” in their response to inequality, we would expect this scenario to be associated with left-skewed polarization.

The Center of Political Gravity By all accounts, what we might call the center of gravity in party politics varies across countries and over time. For instance, the position of the most right wing of the five main parties contesting the Dutch general election of 1998 was, according to the Comparative Manifesto Project, more leftist than the position of Bill Clinton in the presidential election of 1996. While the Netherlands is clearly a more egalitarian country than the United States, we do not believe that contemporary differences in the distribution of income explain why the center of gravity in Dutch politics is further to the left than the center of gravity in American politics. If there is a causal relationship between income distribution and the center of political gravity, it is at least as likely to run in the opposite direction. More leftist government policies must surely play a role in any account of why the distribution of wages and disposable household income is more compressed in the Netherlands than in the United States. (We return to the question of reverse causality in due course.) There is also a great deal of evidence suggesting that the center of political gravity moved to the right in many OECD countries during the 1980s and 1990s. This trend appears to have been quite pervasive and, for this very reason, cannot be explained simply in terms of trends in the distribution of income. As we shall document later, rising inequality is by no means a universal trend among the countries included in our analysis. A number of other plausible explanations for the apparent shift to the right should be noted. One line of argument holds that this shift reflects

Inequality as a Source of Political Polarization

323

the “growth to limits” of redistributive welfare states. Tax fatigue certainly became a prevalent feature of electoral politics in the 1980s and 1990s, and many voters as well as politicians seem to have become convinced that redistributive policies had reached a point of diminishing returns. In a different vein, the rightward shift of party politics might be attributed to the erosion of the sociological foundations of traditional left politics with the decline of the industrial working class, the decline of unions, and the decline of class voting. Finally, it also seems quite plausible to attribute this rightward shift to pressures associated with globalization, that is, the international integration of financial markets and the intensification of international competition in product markets. We believe that all of these arguments are relevant to the evolution of party politics since the mid-1970s and that the forces they identify cannot be straightforwardly captured by a few quantitative variables. Our data set is too small to evaluate the relative merits of these arguments in any systematic fashion. Still, our theoretical model makes predictions about the effects of inequality on relative party positions—not about inequality’s effects on the center of political gravity. To estimate these effects of inequality we control for the center of political gravity by including a measure of the position of the median voter developed by HeeMin Kim and Richard Fording (1998, 2003) on the right-hand side of our regression equations. As we explain below, Kim and Fording estimate the position of the median voter based on left-right scores of party election manifestos and the distribution of votes among parties. The Kim-Fording measure confirms that the center of political gravity did indeed shift to the right in most OECD countries in the 1980s and 1990s (see figure 10.4 in the next section). As we shall see, their measure of the position of the median voter turns out to be a strong predictor of the positions adopted by both left and right parties. In itself, this is a somewhat trivial finding, since party positions are used to estimate the median position. More interestingly, however, the results we report here indicate that the rightward shift of party politics in the 1980s and 1990s was skewed. Main left parties generally shifted their positions to the right to a more significant extent than main right parties. Again, our goal here is not to explain either the rightward shift or the convergence associated with this shift. Rather, our analysis explores the effects of inequality on party positions while holding these trends constant.

The Data and Measures This section describes the data set we constructed to explore the effects of wage and household income inequality on party politics and discusses our measurements for dependent and independent variables.3 The units

324

Democracy, Inequality, and Representation

of observation in our data set are country-election-years. For each election from the late 1940s onwards, the Comparative Manifesto Project provides measures of party positions on the left-right dimension, and these measures serve as our dependent variables. Recently published CMP data (Klingemann et al. 2006) enable us to include elections through 2003, but the availability of relevant measures of inequality restricts the number of countries and election-years included in our data set.

Inequality We draw on two sources for our measures of inequality: the OECD data set on relative wages and the Luxembourg Income Study (LIS). Commonly used in the existing literature, these are the best available data sets providing wage and income measures that are comparable across countries. Pertaining to gross (pretax) earnings for full-time employees, the OECD data set enables us to calculate various decile ratios. Our measure of wage inequality is the 90/10 ratio: the ratio of earnings of someone in the ninetieth percentile (the bottom of the top 10 percent of the wage distribution) to the earnings of someone in the tenth percentile (the top of the bottom 10 percent). The inequality measure that we derive from the LIS database is the Gini coefficient for disposable household income among working-age households. This measure encompasses all kinds of income—government transfers and returns on financial assets as well as income from employment—and takes into account the (re)distributive effects of taxation and income pooling within households. The Gini coefficient is commonly interpreted as the percentage of total income that would have to be redistributed in order to achieve perfect equality. Like the 90/10 wage ratio, this is a broad summary measure of inequality. There is certainly a lot more that we might want to know about the shape of the income distribution, but for our purposes these inequality measures would seem to be quite sufficient. We measure household income inequality in terms of disposable income (post-tax-and-transfer income) rather than market income (pretax-and-transfer income) because our theoretical framework posits that voters form policy and party preferences based on their position in the income distribution.4 We assume that voters have some knowledge, however imperfect, about their relative income.5 This assumption seems less reasonable for the market income of households than for the disposable income of households or the gross wages of individuals. In addition, cross-national comparisons of market income inequality are highly misleading unless we exclude elderly households (see Kenworthy and Pon-

Inequality as a Source of Political Polarization

325

tusson 2005). In countries with generous public pension systems, many households headed by retired people have no market income at all, but this does not mean that they are poor. Given that the elderly constitute a large segment of the electorate, we do not wish to exclude them from our analysis. For eight countries, the most recent version of the OECD data set on relative wages (OECD 2004) contains more or less complete time series of annual observations from the mid-1970s (or late 1970s) to the early 2000s (or late 1990s). However, a number of countries do not enter the OECD data set until the 1980s, the early 1990s, or even the late 1990s, and for some countries the time series ends at some point in the 1990s. The LIS data set is organized on the basis of five-year waves, with observations in each wave pertaining to different years for different countries. For the early waves (the mid-1970s and early 1980s), the LIS data set covers only a small number of countries. In constructing our own data set, we have proceeded as follows. We include as a case any country-election-year for which we have at least one observation of both wage inequality and household disposable income inequality for the year in question or any of the preceding four years. When we have multiple observations of inequality over the five years, which is typically the case for wage inequality, we average these observations. To maximize the number of countries included in our analysis, we use wage inequality data from an earlier version of the OECD data set (OECD 1999) for Belgium and Norway.6 On the other hand, we decided to drop five observations for Austria, Canada, and Switzerland. For Switzerland, we could only generate a single electionyear observation, and the post-1997 time series for Canada in OECD (2004) is strikingly more erratic than the time series for other countries. Austria was eliminated because it was the only remaining country with only two election-year observations. As shown in table 10.1, the upshot of these procedures is a data set that includes twelve countries, for a total of sixty-eight country-electionyear observations. For Denmark and Norway, the data set includes three observations. At the other end of the spectrum, the data set includes nine observations for Sweden and eight observations for Australia and the United Kingdom. On average, we have 5.7 observations per country. While fifty-eight of the observations for household inequality are singleyear observations and five of these are contemporaneous with our observations of party positions, only five of our observations for wage inequality are single observations (none contemporaneous), and fully fifty-five of these observations are based on averaging across four or five years. Before we proceed, it should be noted that our data, as summarized in table 10.1, do not bear out the common notion of an OECD-wide in-

326 Table 10.1

Democracy, Inequality, and Representation Country-Election-Years Covered and Descriptive Inequality Data Wage Inequality Election Years

Australia

Belgium Britain

Denmark Finland France Germany Italy Netherlands Norway Sweden

United States

1983, 1984, 1987, 1990, 1993, 1996, 1998, 2001 1987, 1991, 1995, 1999 1974 (February), 1974 (October), 1979, 1983, 1987, 1992, 1997, 2001 1988, 1990, 1994 1987, 1991, 1995, 1999, 2003 1981, 1986, 1988, 1993, 1997, 2002 1987, 1990, 1994, 1998, 2002 1987, 1992, 1994, 1996 1986, 1989, 1994, 1998, 2002, 2003 1993, 1997, 2001 1976, 1979, 1982, 1985, 1988, 1991, 1994, 1998, 2002 1976, 1980, 1984, 1988, 1992, 1996, 2000

Household Inequality

Most Recent

Changea

Most Recent

Changea

2.998 1.96

+6.0% —a

.317 .258

+12.8% +13.7

3.45 2.155

+17.3 –1.7

.343 .236

+28.0 –7.1

2.417

+2.5

.247

+18.2

3.106

–5.1

.278

–5.8

3.036 2.372

+9.4 +5.0

.275 .339

+7.0 +14.1

2.92 1.99

+18.5 –1.5

.248 .251

–4.6 +8.7

2.28

+12.6

.252

+27.9

4.592

+24.3

.370

+22.9

Source: wage inequality: OECD (1999, 2004); household inequality: Luxembourg Income Study (LIS), “Income Inequality Measures,” accessed April 15, 2007 at http://www.lisproject.org/keyfigures/ineqbble.html. a Change is measured as the change from the minimum to the most recent observation unless the most recent observation is also the minimum observation; in the latter cases, change is measured as the change from the maximum observation to the most recent observation. A break in the series does not allow us to calculate change for Belgium.

crease in inequality since the early 1980s. The United Kingdom, Sweden, and the United States stand out as the OECD countries in which wage inequality and household income inequality have both increased quite dramatically. However, wage inequality declined in Denmark, France, and Norway and increased only modestly in Australia, Finland, and Italy over the (variable) time periods for which data are available. The tendency for household income inequality to increase is more pervasive, but Den-

Inequality as a Source of Political Polarization

327

mark, France, and the Netherlands bucked this trend, and we observe fairly modest increases in Germany and Norway.

Party Positions The dependent variables of the empirical models reported here are based on data from the Comparative Manifesto Project and refer to party positions on the left-right dimensions, as measured by Michael Laver and Ian Budge (1992) and subsequent CMP publications (Budge et al. 2001; Klingemann et al. 2006). Briefly, the CMP identifies fifty-four policy areas (or categories) and reports the percentage of “quasi sentences” in election manifestos that fall into each of these areas. Laver and Budge (1992) used factor analysis to identify two groups of thirteen categories that load at the opposite ends of an underlying dimension and calculate left-right scores for individual parties by summing the percentages of manifesto statements that fall into each of the opposing groups and subtracting the percentage of left statements from the percentage of right statements. This yields a left-right index that ranges from −100 (extreme left) to +100 (extreme right).7 It is commonplace to argue that the CMP data tell us more about the salience of particular issues than about party positions on these issues. As Kenneth Benoit and Michael Laver (2006) pointed out, however, virtually all of the CMP coding categories are in fact explicitly or implicitly positional (see also McDonald and Mendes 2001). For Benoit and Laver, the more important limitations of CMP-derived left-right scores have to do with the absence of any estimates of measurement error and the fact that they fail to capture variation in the meaning of the left-right divide across countries and over time. With regard to the latter issue, Benoit and Laver emphasized that the left-right dimension was inductively derived from an analysis of party manifestos between 1945 and 1985 (and therefore does not include, for example, party positions on environmental issues). Our analysis depends on being able to track changes in party positions over time. The expert surveys that Benoit and Laver favor as an alternative to the CMP approach provide, at best, two observations of party positions per country. The absence of any estimates of measurement error in the CMP database is simply the price that we have to pay to obtain a more time-sensitive set of left-right scores. As for the meaning of the leftright divide in politics changing over time, this is arguably not such a serious problem since our theoretical framework pertains to the representation of voter preferences for redistribution. For us, the problem with the CMP left-right dimension is that it contains too many policy items rather than too few. A left-right index focusing more strictly on policies

328

Democracy, Inequality, and Representation

with a redistributive impact would be desirable, but the so-called welfare dimension in the CMP data set does not fit the bill. As Gøsta EspingAndersen (1990) and others have long argued, there are many political forces in Europe, most notably Christian democracy, that favor social protection without favoring redistribution. Several studies (for example, Powell 2000) have shown that the standard CMP left-right scores provide a reasonably good summary of what parties stand for in elections and that the left-right dimension is in fact a meaningful factor for voters. There is also evidence in the existing literature suggesting that these scores can be used to predict what parties actually do when they come to power (see, for example, Budge and Hofferbert 1990). Furthermore, the CMP’s left-right index correlates reasonably well with various party classification schemes based on expert surveys (see Gabel and Huber 2000; McDonald and Kim, n.d.). For the main parties of the left and right combined, the correlation between the most recent left-right scores in our data set and the expert scores on the general left-right dimension reported by Benoit and Laver (2006) is .71. Even more noteworthy, the correlation between our most recent left-right scores (for main parties) and Benoit and Laver’s expert scores on their “taxes-versus-spending” dimension is .77. The fact that the left-right dimension, as measured here, encompasses issues that do not pertain directly to redistribution arguably militates against finding effects of inequality on party positions. There is certainly no reason to believe that measuring the positions of parties in this manner biases the exercise in favor of our theoretical expectations. It should also be noted that there is a great deal of election-to-election volatility in left-right scores (for the same party) in the CMP data. This volatility reflects not only measurement error but also, we believe, strategic signaling by parties. For instance, a left party that has decided to move to the center may exaggerate the extent of its move to offset its reputation. Smoothing party scores over several elections might yield more accurate measures of party positions (McDonald and Mendes 2001), but it would also introduce an obvious endogeneity problem into our analysis. To avoid invoking inequality in year t as an explanation of party positions in some previous year, we stick with single-year (current) observations of party positions. Again, this approach may generate noise that militates against finding statistically significant effects of inequality. The dependent variable of the empirical models that we report below is the left-right score of either the main party of the left or the main party of the right, with higher scores representing more rightist positions in both cases. We code as “main party of the left” the party that won the largest share of the left vote in the most elections included in our data

Inequality as a Source of Political Polarization Table 10.2

329

Main Parties of the Left and Right

Australia Belgium Britain Denmark Finland France Germany Italy Netherlands Norway Sweden United States

Left

Right

Labour Socialists (SP+PS) Labor Social Democrats (SD) Social Democrats (SSDP) Socialists (PS) Social Democrats (SPD) PCI/PDS Labor (PvdA) Labor (DNA) Social Democrats (SAP) Democrats

Liberals Christian Democrats (CVP+PSC) Conservatives Conservatives (KF) Center Party (SK) Gaullists (RPR, UMP) Christian Democrats (CDU/CSU) Christian Democrats (DC) Christian Democrats (CDA) Conservatives (H) Moderates Republicans

Source: Authors’ compilation.

set, and similarly, we code as “main party of the right” the party that won the largest share of the non-left vote in the most elections (see table 10.2). While party positions change, our analysis thus holds the identity of main left parties and main right parties constant.8 Figure 10.3 graphs annual averages for the left-right scores of the left and right main parties over the period 1975 to 1998 in the twelve countries included in the data set. Values for non-election years have been interpolated linearly, so that all twelve countries are included in most of the annual averages.9 In marked contrast to the American case (McCarty, Poole, and Rosenthal 2006), we do not observe any secular OECD-wide trend toward the polarization of party politics over this period. If wage inequality and household inequality had uniformly increased across the OECD countries over the same period, this would be a most damning picture for this chapter’s partisan Meltzer-Richard model. As noted already, however, inequality increased significantly only in some of the countries included in our data set (see table 10.1). Also, we hasten to stress that our framework posits that trends in wage inequality and household income inequality have different political effects and that other variables must be taken into account. By focusing on trends over time and pooling data across twelve countries, figure 10.3 hides much of the interesting variation in our data set. In short, it is necessary to engage in multivariate analysis to estimate the effects of different forms of inequality on party politics.

330

Democracy, Inequality, and Representation

Figure 10.3

Positions of Main Left and Main Right Parties on the Left-Right Dimension: Yearly Means for Twelve Countries, 1975 to 1998

40

Left-Right Score

20

0

–20

Mean Right Party Position Mean Left Party Position

–40 1975

1980

1985

1990

1995

2000

Year Source: Authors’ calculations based on data in Klingemann et al. (2006).

Other Variables As indicated earlier, we believe that the political mobilization of lowincome groups matters to party responses to inequality. We hypothesize that this variable either renders left parties alone more leftist when they react to inequality or, alternatively, has a similar effect on left and right parties. In principle, it would be desirable to estimate separately how voter turnout and unionization condition partisan responses to inequality, but our data set is quite limited, and these variables are correlated with each other. To simplify matters, and to avoid multicollinearity problems, we combine turnout and unionization into a single variable, which we refer to as “low-income mobilization.” We generate this single measure of low-income mobilization by summing standardized scores for voter turnout and union density and lag the impact of this variable by averaging observations over five years, including the election year in question.10 Table 10.3 reports mobilization scores by country. Based on the most recent observations in our data set as well as average scores, Sweden,

Inequality as a Source of Political Polarization Table 10.3

331

Mobilization Scores by Country

Sweden Denmark Belgium Australia Finland Norway Italy Britain Germany Netherlands France United States

Average

Mid-1980s

Most Recent

2.241 1.839 1.561 .956 .772 .359 .287 –.398 –.423 –.863 –1.83 –3.458

2.411 2.023 1.557 1.219 .831 .586a .458 .033 .096 –.536 –.968 –3.328

1.725 1.655 1.49 .468 .472 .311 .060 –1.207 –.600 –1.106 –2.058 –3.454

Source: Sum of standardized scores for voter turnout and net union density (union members as a percentage of the employed labor force). Turnout data from Armingeon et al. (2004), supplemented by Internet sources for 2003. Union density data from Ebbinghaus and Visser (2000) except for Australia, Japan, the United Kingdom, and the United States: pre-1990 figures for these countries from Visser (1996) and post-1990 figures provided by Ebbinghaus. The following observations were extrapolated: all countries for 2001, Switzerland for 2002 and 2003, Sweden for 2002, Finland for 2002 and 2003, the Netherlands for 2002 and 2003, France for 2002, and Germany for 2002. a The Norwegian “mid-1980s” figure refers to 1993.

Denmark, and Belgium stand out as the countries with the highest levels of low-income mobilization. At the other end of the spectrum, the United States stands out as the country in which low-income groups are by far the least mobilized as participants in the political process. France and the Netherlands also fall into the low-mobilization camp, as does the United Kingdom in the more recent period. The decline of low-income mobilization is most striking in the British case but emerges very clearly as a general trend in our data. In every single country included in our analysis, the most recent mobilization scores are also the lowest. With total observations of only sixty-eight, we want to keep the number of control variables to a minimum. However, it is clearly necessary to control for the center of political gravity. Our theoretical framework generates predictions about the effects of inequality on relative party positions, not about its effects on the center of political gravity. As mentioned earlier, we control for the effects of the center of political gravity by including a measure of the position of the median voter developed by Kim and Fording (1998, 2003) in our analysis. Using CMP data, Kim and Fording identify the midpoints between parties that have been ranked on

332

Democracy, Inequality, and Representation

the left-right dimension and assume that the policy preferences of those who voted for a particular party are evenly distributed across the interval between the two midpoints that separate this party from the parties to its immediate right and immediate left. They then estimate the position of the median voter in the electorate. It deserves to be underscored that Kim and Fording assess the position of the median voter based on policy positions articulated by parties and do not rely on any direct evidence on voter opinions or preferences. Their measure might more appropriately be conceived as a measure of “the center of gravity in electoral competition.” On the other hand, it seems quite accurate to think of the position of the median voter as being constructed by parties in competition with each other. Mindful of the complex issues involved here, we stick with the variable label used by Kim and Fording. We have rescaled Kim and Fording’s measure of the position of the median voter so that it conforms to the standard CMP measure of party positions, ranging from –100 to +100, with higher numbers representing more rightist positions. The actual variable included in our regression models is the average value of the median voter’s position for the election year in question and the preceding four years. Following Kim and Fording, our five-year averages are based on linearly interpolated values for non-election years. This setup captures the idea that shifts in the center of gravity are not simply an unanticipated outcome of elections. We assume that parties observe shifts in voter opinions and the policy positions of their competitors between elections and take such shifts into account when they prepare their election programs. At the same time, we expect that it takes parties some time to respond to changes in the position of the median voter. Tracking the evolution of the average median voter position on the left-right dimension in our twelve countries, figure 10.4 strongly confirms that the time period covered by our analysis is characterized by a rightward trend in electoral politics.11 To reiterate, our goal in this chapter is not to explain the rightward shift illustrated by figure 10.4 but rather to explore the effects of inequality on party strategies while controlling for this shift. We expect the rightward shift of the median voter to be associated with more rightist positions held by the main parties of the left and the right alike. Our regression models include one final control variable: the effective number of parties, as measured by Markku Laakso and Rein Taagepera (1979). This variable is also measured as a five-year average. The motivation for including it is simply to control for the effects of party-system dynamics. The most obvious hypothesis along these lines is that multiparty competition is a source of political polarization, pushing main left parties to the left and main right parties to the right (see Cox 1990).

Inequality as a Source of Political Polarization Figure 10.4

333

Median Voter Position on the Left-Right Dimension: Yearly Means for Twelve Countries, 1975 to 1998

Mean Median Voter Position

5

0

–5

–10

–15

–20 1975

1980

1985

1990

1995

2000

Year Source: Transformed Kim-Fording measure, based on data downloaded from HeeMin Kim’s home page, accessed April 15, 2007.

Empirical Results The results reported here were obtained by estimating a series of models with the following specification: Yit = β0 + β1X1it + . . . + βnXnit + εit

(10.1)

where Yit represents the positions on the left-right dimension of either left or right parties, β0 represents a general intercept, X1 to Xn are the explanatory variables (wage inequality, household income inequality, lowincome mobilization, the position of the median voter, and the effective number of parties),12 β1 to βn are the slopes of the explanatory variables, and εit denotes the errors. We recognize that there may be a number of country-specific effects that we cannot estimate directly (specific historical circumstances, institutional complexities, and so on) and that the existence of countryspecific omitted variables could affect the accuracy of our estimation. To

334

Democracy, Inequality, and Representation

mitigate this potential problem we estimate random effects.13 The results presented here therefore were obtained through a set of standard generalized least squares random-effects models. We also estimate standard errors that are robust to correlation within countries. All our results report robust variance estimates (more specifically, the Huber/White/sandwich estimate of variance). All models reported in table 10.4 estimate the effects of both wage inequality and household income inequality. The first two models estimate only the direct effects of these and the other variables identified earlier. The four interaction models explore the effects of low-income mobilization on the relationship between inequality and party positions. Because of the potential problem posed by multicollinearity, we estimate the effects of interacting mobilization with wage inequality and household inequality separately. Setting the effects of inequality aside for the time being, our results show that the median voter position is associated with those held by left and right parties alike. In all three models with left party positions as the dependent variable, this variable is significant at better than the 99 percent confidence level. Once we control for interaction effects, the median voter also becomes a statistically significant predictor of right party positions. Given that party positions are used to estimate the position of the median voter, it is hardly surprising that parties of the left and the right move in the same direction as the median voter. A far more interesting finding is that the size of the coefficient for this variable is much larger (invariably more than three times as large) in the models with left party positions as the dependent variable. It appears that left parties are more vulnerable to shifts to the right by the median voter. This makes sense, since, for right parties, the median voter has moved in the same direction as their core constituencies. For the left, this has not been the case, and left parties have had to make larger strategic adjustments than right parties in order to remain competitive as the center of gravity has shifted in a rightward direction. Our results do not support the proposition that multiparty competition is a source of polarization. According to our findings, the effective number of parties has no effect on the position of left parties, but it has a strong negative effect on the position of right parties. Consistent with Torben Iversen and David Soskice’s (2006) thesis that proportional representation favors the left, this finding suggests that right parties move to the left when they are faced with multiparty competition or, alternatively, that more centrist parties tend to dominate more rightist parties when the right is fragmented. When we do not control for interaction effects, wage inequality is weakly associated with more leftist left parties and appears to have no ef-

Inequality as a Source of Political Polarization Table 10.4

335

Determinants of Party Positions on the Left-Right Dimension Main Effects Left

Right

WI*MOB Left

Right

Constant

9.419 2.819 14.768 17.795 (21.208) 23.040 (21.411) (20.612) .657 .903 .490 .388 Wage inequality –11.425 1.239 –16.093 –5.912 (7.117) (7.006) (6.148) (4.840) .108 .860 .009 .222 Household inequality 53.295 111.193 72.658 124.163 (76.687) (43.506) (92.074) (39.130) .487 .011 .431 .002 Low-income mobilization –1.116 5.236 6.296 16.847 (1.488) (.2.405) (3.805) (4.627) .454 .029 .098 .000 WI*mobilization –2.658 –4.137 (1.031) (1.135) .010 .000 HI*mobilization

Median voter position

Number of parties

R-squared overall Observations

.535 (.054) .000 –.461 (.935) .622 .432 68

.134 (.086) .119 –4.314 (1.240) .001 .401 68

.549 (.076) .000 –.430 (.982) .661 .472 68

HI*MOB Left

Right

16.219 18.138 (17.315) (21.756) .349 .404 –17.997 –7.025 (6.339) (5.554) .005 .206 86.709 136.948 (90.073) (42.926) .336 .001 13.122 23.462 (6.056) (5.666) .030 .000

–55.216 –70.375 (20.980) (17.217) .008 .000 .145 .571 .169 (.074) (.071) (.074) .049 .000 .023 –4.779 –.453 –4.817 (1.023) (.923) (.995) .000 .624 .000 .485 68

.489 68

.482 68

Source: Authors’ calculations. Note: Results are from generalized least squares random-effects models. Numbers are estimated coefficients; numbers in parentheses are robust variance standard errors that adjust for within-country correlation; numbers in italics are p-values from two-sided t-tests.

fect whatsoever on the position of right parties. By contrast, household income inequality is strongly associated with more rightist right parties, but we do not observe any relationship between household inequality and the position of left parties.14 The coefficient of our mobilization variable is negative but not statistically significant for left parties, while positive and significant for right parties. Rather surprisingly, high voter

336

Democracy, Inequality, and Representation

turnout and union density appear to be associated with more right-leaning right parties. When we interact mobilization with either measure of inequality, the direct effect of mobilization is positive for both left parties and right parties. For our purposes, however, the key point is that all interaction terms have negative coefficients and are significant at the 99 percent level. As mobilization increases, left and right parties alike move to the left in response to either form of inequality. As is the case with all interactive models, the results in table 10.4 are not easy to interpret. Testing this chapter’s hypotheses requires assessing the effects of wage and household income inequality at different levels of mobilization. Using the estimates from the interaction models in table 10.4, figures 10.5 and 10.6 graph the conditional coefficients of wage and household inequality at different levels of mobilization (and the 95 percent confidence intervals around these estimates). These figures provide very strong confirmation of the hypotheses in our theoretical framework. The association between wage inequality and left parties is significant only at medium and high levels of mobilization, and the coefficient of wage inequality is always negative, increasing in size with mobilization. For right parties, the coefficient for wage inequality is insignificant at most levels of mobilization. Only at very high levels of mobilization do we observe a statistically significant association between wage inequality and more leftist (or less rightist) right parties. In figure 10.6, we observe a strong and very significant association between household income inequality and more right wing right parties at low levels of mobilization. As mobilization increases, this association disappears. The point estimates for the impact of household inequality on left parties follow a very similar trajectory, but these estimates never satisfy conventional criteria of statistical significance. Figure 10.5 makes clear that increasing wage inequality pushes left parties to the left when mobilization is high (at the level of the mean or higher), but it is difficult to assess the substantive significance of these results. To understand what these estimates mean we can compare two countries. The United States is a country with a very low level of mobilization. In 1980, for example, the value for our five-year average of union density was 21.5, and the value for our five-year average of voter turnout was 45.44. After we standardize these two measures and add them up, we obtain a measure of mobilization equal to –3.33. This is not the lowest of the mobilization observations in our sample, but as indicated in table 10.3, it is within the range of very low values. In 1980 the five-year average for the 90/10 ratio in the United States was already a pretty high 3.76. By the year 2000, however, the five-year average for the 90/10 ratio in the United States had reached a whopping 4.59. Our results suggest that because of the low level of mobilization (a level that in fact decreased further from 1980 to 2000), an increase in wage in-

Figure 10.5

Effects of Wage Inequality on Left and Right Party Positions, Conditional on Levels of Mobilization

20

Coefficient of Variable

Upper Confidence Bound Effects on Right Lower Confidence Bound

Effects on Left Lower Confidence Bound Upper Confidence Bound

30

10 0 –10 –20 –30 –40 –50 –3.7

2.4

0 Mobilization

Source: Authors’ calculations based on regression results presented in table 10.4.

Figure 10.6

Effects of Household Income Inequality on Left and Right Party Positions, Conditional on Levels of Mobilization

700

Effects on Left Lower Confidence Bound Upper Confidence Bound Upper Confidence Bound Effects on Right Lower Confidence Bound

Coefficient of Variable

600 500 400 300 200 100 0 –100 –200 –3.7

0 Mobilization

Source: Authors’ calculations based on regression results presented in table 10.4.

2.4

338

Democracy, Inequality, and Representation

equality in the United States would have no significant effect on the position of the Democratic Party. Sweden, on the other hand, has the highest level of mobilization in our sample. In 1988 the value for our five-year average of union density was 82.76, and the value for our five-year average of voter turnout was 89.42. After we standardize these two measures and add them up, we obtain a measure of mobilization equal to 2.4. What would be the effect of the increase in inequality we have described in the previous paragraph if the United States had the mobilization level of Sweden? Our interaction results show that an increase in the 90/10 ratio from 3.76 to 4.59 would have been associated with a move equal to around nineteen points to the left by the Democratic Party. To put this in context, the Democratic Party had a score of –21.2 on the left-right dimension in 1980. Our results suggest that if mobilization had been as high in the United States as in Sweden, this increase in wage inequality would have pushed the Democratic Party’s position to the left by nineteen points (to –40.2), ceteris paribus. Instead, the position of the Democratic Party in 2000 (–3.6) was much more centrist. Similarly, figure 10.6 makes clear that increasing household income inequality pushes right parties to the left (that is, it makes them less conservative) as mobilization grows. We can again assess the substantive significance of these results by comparing two countries. In 1992 the value for our five-year average of union density in the United States was 15.58, and the value for our five-year average of voter turnout was 43.78. After we standardize these two measures and add them up, we obtain a measure of mobilization equal to –3.7. This is in fact the lowest value for mobilization in our sample. In 1992 the corresponding household income Gini value for the United States was .338.15 By the year 2000, however, the value of the household income Gini had increased to .370.16 Figure 10.6 shows that, because of the low level of mobilization, an increase in household income inequality in the United States would have a big effect on the position of the Republican Party. An increase from .338 to .370 in the Gini is associated with a move to the right by the Republican Party equal to thirteen points on the left-right dimension. The United Kingdom in 1979, on the other hand, had a level of mobilization quite close to the mean in our sample. In 1979 the value for our five-year average of union density was 51.9, and the value for our fiveyear average of voter turnout was 73.58. After we standardize these two measures and add them up, we obtain a measure of mobilization equal to –0.03 (close to the mean, which is 0). What would be the effect of the increase in household income inequality we have described in the previous paragraph if the United States had the mobilization level of the United Kingdom? Our interaction results show that an increase in the Gini from

Inequality as a Source of Political Polarization

339

.338 to .370 when mobilization is at the mean17 is associated with a move equal to around four points to the right by the Republican Party. To put these numbers in context, the Republican Party had a score of 30.42 on the left-right dimension in 1992. Our results suggest that with the American level of mobilization, the increase in household inequality in the United States from 1992 to 2000 would have pushed the score to 43.42. However, if mobilization had been as high in the United States as in the United Kingdom in 1979, the increase in household income inequality would have only moved the Republican Party’s position to a score of 34.42. Our theoretical model implies that causality runs from the distribution of income to party positions via the policy preferences of core constituencies, as well as the policy preferences of the median voter in the electorate as a whole. We readily admit that causality might also run in the opposite direction: from party politics to the distribution of income. For the United States, Larry Bartels (2008, chap. 2) argued persuasively that the policies pursued by Republican administrations have been a major source of the growth of inequality in disposable household income since the 1970s. However, we do not believe that reverse causality can adequately account for the results presented here. There are several reasons for this. Two of those reasons have already been mentioned. To reiterate, first, our analysis is based on measures of inequality that are temporally prior to our measures of party positions. Second, there is a significant amount of intertemporal volatility in our measures of party positions. It should again be noted that while we do observe a secular and quite pervasive rightward shift of the center of political gravity across the countries included in our analysis (figures 10.3 and 10.4), rising inequality is not a secular and pervasive trend in our data set (table 10.1). There are some additional reasons for our belief that this chapter’s arguments model causality correctly. The first one is that the reversecausality objection pertains primarily to the effects of household income inequality, since our measure of household income inequality refers to disposable income and thus takes into account the effects of taxation and government transfers. Government partisanship certainly affects the distribution of wages through minimum-wage legislation and the indirect, second-order effects of taxation and social benefits. But (as shown by Rueda, this volume), the connection between partisanship and policy, on the one hand, and policy and wage inequality, on the other, is not completely straightforward, being highly dependent on the institutional context. Reverse causality simply does not provide a plausible account of why we observe a strong association between wage inequality and more leftist (redistributive) positions held by left parties. For a subset of nine countries (fifty-four observations), the association between household income inequality and the positions held by right

340

Democracy, Inequality, and Representation

parties at low levels of mobilization still obtains when we replicate our interaction model with household inequality measured in terms of market income (before taxes and transfers) rather than disposable income. The finding that increasing household inequality is associated with more rightist right parties when mobilization is low is therefore less vulnerable to the reverse-causality objection than might at first appear to be the case. Finally, the conditioning effects of low-income mobilization surely make more sense if we think of causality as running from the income distribution to party politics rather than the other way around.

Patterns of Change over Time Since most of the variation in our sample is cross-sectional, our empirical models do a much better job of explaining variation across countries than they do of explaining over-time variation within countries. This should not come as a huge surprise given the volatility of left-right scores from one election to the next and the limited number of election years per country in our data set. Our goal in this section is to demonstrate, in an admittedly less rigorous fashion, that the theoretical framework elaborated earlier sheds light on within-country changes over time as well as on between-country differences. For a subset of our twelve countries, we explore the extent to which trends in wage and household income inequality might be invoked to explain patterns of partisan polarization or convergence over the 1980s and 1990s. To begin with, it should be noted that the results presented earlier indicate quite definitely that the conditioning effects of low-income mobilization are not specific to left parties. As we have seen, right parties as well as left parties become more leftist in their response to inequality, be it wage or household income inequality, as low-income mobilization rises. This finding allows us to articulate more precise expectations as to how different inequality trends and levels of mobilization jointly give rise to different patterns of partisan polarization. We present these expectations in table 10.5. Our theoretical model and regression results lead us to expect that rising wage inequality in the absence of rising household income inequality will generate left-skewed polarization at medium and high levels of mobilization. Conversely, rising household income inequality in the absence of rising wage inequality will generate right-skewed polarization at low and medium levels of mobilization. Finally, the joint occurrence of these inequality trends will generate right-skewed mobilization at low levels of mobilization, but left-skewed polarization at high levels of mobilization and symmetric polarization at medium levels of polarization. We hasten to point out that these expectations are based on holding levels of mobi-

Inequality as a Source of Political Polarization Table 10.5

341

Expected Polarization Patterns Low-Income Mobilization

Wage inequality rising Household income inequality rising Both wage inequality and household income inequality rising

Low

Medium

High

No polarization

Left-skewed polarization Right-skewed polarization Symmetric polarization

Left-skewed polarization No polarization

Right-skewed polarization Right-skewed polarization

Left-skewed polarization

Source: Authors’ compilation.

lization constant. In many countries, mobilization levels have fallen at the same time as either or both forms of income inequality have increased, rendering the predictions of our model more ambiguous. For seven countries, table 10.6 reports on the positions held by the main parties of the left and right at the beginning of the 1980s and 2000s. To mitigate the problem of election-to-election volatility, we have averaged the scores for two consecutive elections in the 1970s (or the late 1970s and very early 1980s) and the scores for the two last elections included in the CMP database. In addition, table 10.6 records the absolute difference in left-right scores between the main left and main right parties and the midpoint between their positions (the latter measure being akin to the center-of-gravity variable used in the previous analysis). Considering changes in these two measures jointly enables us to distinguish between right-skewed and left-skewed polarization or convergence. To keep our discussion relatively simple, table 10.6 includes the three countries with the highest average mobilization scores and the three countries with the lowest average mobilization scores (see table 10.3). In addition, we include the United Kingdom as an ambiguous case of special interest. For our purposes, comparing the trajectories of partisan politics in Sweden, the United States, and the United Kingdom is particularly germane because all three countries experienced very large increases in both wage and household income inequality in the 1980s and 1990s (see table 10.1), and yet they vary dramatically in terms of our conditioning variable—political mobilization of low-income groups. In our data set, Sweden is the country with the highest mobilization scores, while the United States is the country with the lowest mobilization scores. In Sweden, voter turnout dropped from 90.7 percent in 1979 to 80.1 percent in 2002, but union density essentially held steady over this period. In the United States, by contrast, voter turnout held steady while union den-

342 Table 10.6

Democracy, Inequality, and Representation Left-Right Scores of the Main Left and Right Parties Circa 1980 and 2000, Selected Countries

United States 1976, 1980 1996, 2000 Change United Kingdom 1974 (October), 1979 1997, 2002 Change Sweden 1976, 1979 1998, 2002 Change France 1978, 1981 1997, 2002 Change Denmark 1977, 1979 1998, 2001 Change Belgium 1977, 1978 1995, 1999 Change Netherlands 1977, 1981 2002, 2003 Change Twelve-country average Early Recent Change

Left

Right

Left-Right Difference

Midpoint

–20.5 2.6 23.1

14.5 18.7 14.3

34.5 26.1 –8.4

–3.3 15.7 19.0

–27.1 6.8 33.9

17.9 20.3 2.4

45.0 13.5 –31.5

–4.6 13.6 18.2

–13.4 –10.9 2.5

12.7 37.7 25.0

16.4 48.6 22.5

.4 13.4 13.0

–33.5 –14.7 18.8

17.3 –6.1 –23.4

50.8 8.6 –5.1

–8.1 –10.4 –2.3

–12.1 –4.2 7.9

29.0 19.8 –9.2

41.1 24.0 –17.1

8.5 7.8 –.7

–20.5 –19.2 1.2

–1.5 –5.4 –3.9

19.0 13.9 –5.1

–8.1 –10.4 –2.3

–37.1 –5.2 31.9

–15.5 2.5 18.0

21.6 7.7 –13.9

–26.3 –1.3 18.6

–22.1 –5.3 16.8

7.9 18.6 10.7

30.0 23.8 –6.2

–7.1 6.7 13.8

Source: Authors’ calculations based on data in Klingemann et al. (2006).

sity continued to decline. However, the changes in mobilization scores recorded for these countries are minor by comparison to those in the United Kingdom, which was closer to the mean mobilization score than any other country at the onset of the 1980s but had become a low-mobilization country by the late 1990s. By the logic set out here, we should observe right-skewed polarization

Inequality as a Source of Political Polarization

343

in the United States and left-skewed polarization in Sweden. Considering the United Kingdom to be a medium-mobilization country, we would expect to observe more or less symmetric polarization there, but the decline of mobilization renders this expectation more ambiguous. The data for the United States, the United Kingdom, and Sweden presented in table 10.6 do not immediately confirm our expectations, but a more careful look suggests that our theoretical framework does shed some light on these cases. Let us start with the American case. According to the CMP data presented in table 10.6, the Republicans did indeed move to the right from the late 1970s to the late 1990s, but their rightward shift was not nearly as large as the rightward shift of the Democrats. As a result, we observe right-skewed convergence rather than right-skewed polarization. This brings out an important limitation of the CMP data for the United States, namely, that the data are based exclusively on coding party platforms in presidential elections. Analyzing congressional behavior, McCarty, Poole, and Rosenthal (2006) demonstrated conclusively that the Democrats and the Republicans actually moved apart in this period, and their evidence suggests that the widening gap was primarily due to the Republicans moving to the right. Based on qualitative evidence, Jacob Hacker and Paul Pierson (2005) also made a compelling case that it was the Republicans who moved “off center” in the 1980s and 1990s. The CMP data notwithstanding, the United States can readily be characterized as a case of right-skewed polarization and thus fits very well with the predictions of our model. Like the United States, the United Kingdom emerges from table 10.6 as a case of right-skewed convergence, with the Labour Party moving sharply to the right and the Conservatives essentially staying put. In this case, we have no particular reason to doubt the CMP data. The problem is rather that the figures presented in this table cover two very distinctive phases in the development of British politics—the Thatcher era and the Blair era. According to the CMP data, the Labour Party’s position in the election of 1992 was actually to the left of the position that it had held in the second (October) election of 1974 (–30.4 as compared to –27.1), while the Conservatives were much further to the right in 1992 than they had been in the second election of 1974 (27.9 as compared to 11.4). Thus, the United Kingdom in the 1970s and the 1980s might indeed be seen as a case of symmetric polarization generated by rising wage and household inequality under conditions of medium mobilization. Our theoretical framework highlights two conditions that help explain Labour’s sharp turn to the right under Tony Blair and the Conservatives’ subsequent move toward the center. In part, Blair’s move to the right can be interpreted as a consequence of sharp declines in voter turnout and unionization over the 1980s and early 1990s. Perhaps more important,

344

Democracy, Inequality, and Representation

and less commonly recognized, the transition from polarization to rightskewed convergence in British party politics coincided with a marked deceleration of inequality growth. Wage inequality in the United Kingdom increased by 15.3 percent from 1978 (an all-time low) to 1992, and the Gini coefficient for disposable household income increased by 24.4 percent from 1979 to 1991. By contrast, wage inequality increased by only 4.7 percent from 1992 to 2002, and household income inequality increased by 2.1 percent from 1991 to 1999 according to our data. Sweden is indeed a case of polarization, as our theory predicts, but the polarization that we observe in table 10.6 is right-skewed rather than left-skewed. Despite rising wage inequality and high levels of mobilization, the Swedish Social Democrats moved to the right rather than the left over the 1980s and 1990s. Three points deserve to be made regarding this apparent puzzle. The first one is that all the growth in wage inequality reported in table 10.1 actually occurred in the 1990s (from 1980 to 1990, the Swedish 90/10 wage ratio dropped from 2.03 to 2.01). The sharp right turn taken by the Swedish Social Democrats in the early 1990s predated the rise of wage inequality. Coinciding with a deep economic crisis, this right turn is not particularly difficult to explain. By the election of 2002, however, the Social Democrats had essentially moved back to the position they had held in the early 1980s.18 To some significant extent, this leftist course correction can be seen as a direct response to political pressures on the party leadership associated with rising wage inequality under conditions of high mobilization. The second point is that the decline of voter turnout in Sweden, from 90 percent in the mid-1980s to 80 percent in 2002, mitigated the Social Democrats’ responsiveness to wage inequality. Third, and perhaps most importantly, the contrast with the British Labour Party and the American Democrats is striking. In a fundamental sense, the Swedish Social Democrats hardly moved at all, neither left nor right, during the 1980s and 1990s. Considering that many other left parties followed the median voter in a move sharply to the right during this period, this observation seems quite consistent with our theoretical model. From our perspective, what is truly puzzling about the Swedish case is not the fact that the Social Democrats did not turn left in response to rising wage inequality, but rather the (increasingly) rightist orientation of Swedish Conservatives, despite persistently high levels of low-income mobilization. We do not have a tidy solution to this puzzle. Suffice it to note here that the Conservatives did move sharply toward the center in their successful election campaign of 2006. Let us now consider, more briefly, the countries included in the lower panel of table 10.6. France is a very interesting case because it is the only OECD country in which wage inequality and household income inequal-

Inequality as a Source of Political Polarization

345

ity both declined considerably from 1980 to 2000 (see table 10.1). Our model predicts that in these circumstances there would be limited incentives for party polarization, regardless of the level of mobilization. France, of course, is a case of very low mobilization, primarily on account of the weakness of French unions. Though we have wage data only for the 1980s, Denmark seems to be a high-mobilization case in which wage inequality was essentially stable while household income inequality declined over the 1980s and 1990s. Here too we expect to see no party polarization, in the absence of pressures for or against redistribution from the core constituencies of the left and the right. According to our data, wage inequality in Belgium dropped sharply in the first half of the 1990s. Although this drop may be due to a series break in the data, it seems safe to assume that wage inequality did not increase in Belgium over the 1980s and 1990s. On this assumption, Belgium is a case of stable or falling wage inequality and rising household inequality. Given that Belgium is also a case of high mobilization, constraining right parties’ response to household income inequality, our model suggests that we should observe no party polarization in Belgium. With respect to inequality, the Netherlands represents the mirror image of Belgium: rising wage inequality combined with falling household income inequality. If low-income groups are politically mobilized to the same extent as they are in Belgium, we would expect this scenario to generate left-skewed polarization. However, the Netherlands is unambiguously a case of low mobilization, which reduces the left’s responsiveness to wage inequality. Thus, our model predicts no polarization in the Netherlands as well. Of the twelve countries included in our analysis, France, Denmark, and Belgium are the only cases in which we observe depolarization without the center of political gravity shifting to the right over the 1980s and 1990s. This observation represents, we think, a striking confirmation of core elements of our theoretical framework. On the other hand, the Dutch case clearly does not conform to our expectations. The fact that the Dutch Labor Party moved sharply to the right despite rising wage inequality can partly be explained by falling mobilization of low-income groups, but the rightist shift of the Dutch Christian Democrats in the absence of rising household income inequality cannot be explained within our theoretical framework. It should again be noted that the effects of wage and household income inequality identified by our regression analysis depend critically on controlling for the center of gravity and that this variable (the position of the median voter) is associated with right-skewed convergence. Because of more conservative median voters, left parties have moved to the right more than right parties. Clearly, inequality trends, alone or in conjunc-

346

Democracy, Inequality, and Representation

tion with low-income mobilization, do not provide a complete account of the dynamics of party politics in industrialized democracies. Nonetheless, the preceding discussion indicates that the framework proposed in this chapter not only generates accurate predictions about cross-national differences but also yields insights that are useful for understanding the trajectories of party politics in many countries.

Conclusions The main message of this chapter is that different forms of inequality have different consequences for partisan politics. The conclusions from our analysis can be summarized briefly. First, wage inequality tends to be associated with left-skewed polarization, and household income inequality tends to be associated with right-skewed polarization. Second, the former association holds at medium and high levels of mobilization of low-income groups, while the latter holds at low and medium levels of mobilization. Our explanation of the differential effects of wage inequality and household inequality rests on two basic claims. First, the core constituencies of left parties care primarily about wage inequality and do not necessarily become more supportive of redistribution as household income inequality rises. The effect of wage inequality on left parties, however, is present only when low-income mobilization is high. Second, the core constituencies of right parties care primarily about household income inequality, but high levels of low-income mobilization make right parties less likely to respond to inequality in accordance with the preferences of their core constituencies (that is, opposing redistribution as inequality rises). In concluding, let us again stress that between-country differences drive a large part of our empirical results. In future research, we plan to explore inequality as a determinant of change over time (within countries) in a more focused and systematic manner. Empirically, this requires longer time series. Theoretically, such an analysis would seem to call for several modifications of the model that we have proposed. In particular, we believe that it becomes essential to take into account cross-national differences in perceptions of legitimate income differentials (Svallfors 2006). There are good reasons to believe that a given increase in the amount of inequality will have different effects in a more egalitarian country, like Sweden, than in the United States. As noted earlier, the Meltzer-Richard model and the literature it has inspired conceive the politics of redistribution in terms of individual voters calculating the costs and benefits of redistribution. From this perspective, we would not expect to find that different forms of inequality have different political effects. The fact that we do find differential effects of

Inequality as a Source of Political Polarization

347

wage inequality and household inequality suggests that voters and other political actors (party activists, trade unionists, and so on) care about relative income. At the same time, it seems clear that voters operate with only limited, sometimes very distorted, information about what the distribution of income looks like and where they themselves fall in the distribution of income. This represents another important topic for future research, based on survey data. From a comparative perspective, the obvious question is whether the salience of different forms of inequality varies across countries—or, in other words, across different macro-institutional configurations. For instance, it seems plausible to suppose that wage inequality matters more in countries with encompassing unions and more institutionalized, economywide wage bargaining. We conclude by pointing out that while our chapter aims to bring one of the classical themes in politics (the relationship between inequality and democratic representation) back to current debates in comparative political economy, it is possible to look at our findings with a certain sense of pessimism. Most OECD countries have experienced significant declines in both voter turnout and union density since the early 1970s. Our argument implies that increasing levels of inequality are bound to affect left parties less and less, while they are bound to make right parties more and more opposed to redistribution. In this sense, low-income workers seem to be caught in a vicious circle. Increasing inequality makes their preferences for redistribution stronger, but decreasing mobilization makes their demands less relevant to left parties, which in turn makes these parties less redistributive when they get to power and so inequality continues to grow. Decreasing mobilization, moreover, makes right parties more likely to respond to inequality in accordance with the preferences of their core constituencies (that is, by opposing redistribution as inequality rises). This again makes these parties less redistributive when they get to power and so inequality grows even more. A more optimistic interpretation is possible. Although we treat it as such in the previous analysis, working class mobilization is not entirely exogenous to the behavior of left parties. It is up to left politicians, after all, to dedicate resources to increasing the political participation of lowincome voters. As argued by Anderson and Beramendi (this volume), voter turnout should be understood as the product of people’s incentives to vote as well as parties’ incentives to mobilize specific groups of voters. Although the effectiveness of efforts by left parties to mobilize low-income workers is far from automatic, increasing political participation surely is a way to escape the vicious circle described here. It is therefore in the hands of left parties, at least partly, to promote the participation of those most vulnerable to increases in inequality and, in the process, to make politics more responsive to their demands.

348

Democracy, Inequality, and Representation

Appendix Data Sources and Specifications Party positions: Data from Klingemann et al. (2006); see text for explanation. Wage inequality: 90/10 wage ratios from OECD (2004), supplemented by data from OECD (1999) for Belgium and Norway. Household income inequality: Gini coefficients for disposable household income available at LIS, “Income Inequality Measures,” accessed April 15, 2007 at http://www.lisproject.org/keyfigures/ineqtable.htm. Low-income mobilization: Sum of standardized scores for voter turnout and net union density (union members as a percentage of the employed labor force). Turnout data from Armingeon et al. (2004), supplemented by internet sources for 2003. Union density data from Bernhard Ebbinghaus and Jelle Visser (2000) except for Australia, Japan, the United Kingdom, and the United States: pre1990 figures for these countries from Visser (1996) and post-1990 figures provided by Ebbinghaus. The following observations were extrapolated: all countries for 2001, Switzerland for 2002 and 2003, Sweden for 2002, Finland for 2002 and 2003, the Netherlands for 2002 and 2003, France for 2002, and Germany for 2002. Median position: Transformed Kim-Fording measure (see text for explanation), based on data downloaded from HeeMin Kim’s home page, accessed April 15, 2007 at http://garnet.acns.fsu.edu/ %7Ehkim/. Effective number of parties: Based on a measure developed by Marku Laakso and Rein Taagepera (1979); data from Klaus Armingeon et al. (2004); updated for 2003 based on CMP data in Hans-Dieter Klingemann et al. (2006).

Earlier versions of this chapter were presented at the 2005 and 2006 annual meetings of the American Political Science Association, at the 2006 International Conference of Europeanists, and in workshops at Princeton University, Korea University, the University of Essex, Syracuse University, Cornell University, and the Social Science Research Center Berlin. We greatly benefited from comments in all these meetings. We would particularly like to thank Chris Anderson, Larry Bartels, Pablo Beramendi, Nigel Bowles, Matt Cleary, Daniel Gingerich, Torben Iversen, Staffan Kumlin, Stephen Nelson, Thomas Romer, Ken Scheve, David Soskice, Stefan Svallfors, Chris Way, and Bruce Western.

Inequality as a Source of Political Polarization Table 10A.1

349

Summary Statistics

Variable Main left position Main right position Wage inequality (90/10 ratio) Household inequality (Gini coefficient) Low-income mobilization Median position Effective number of parties

Standard Deviation

Minimum

Maximum

–11.507 17.593

15.698 17.065

–48.5 –10.55

29.26 59.8

2.796

.635

1.96

.271 0 –2.6836 4.333

.042 1.689 20.51432 1.760

Mean

.197 –3.697 –47.04074 2.020

4.592 .370 2.413 41.77728 9.776

Source: party positions: Klingemann et al. (2006); wage inequality: OECD (1999, 2004); household inequality: Luxembourg Income Study (LIS), “Income Inequality Measures,” accessed April 15, 2007 at http://www.lisproject.org/keyfigures/ineqtable.htm; low-income mobilization: sum of standardized scores for voter turnout and net union density (union members as a percentage of the employed labor force); turnout data from Armingeon et al. (2004), supplemented by internet sources for 2003; union density data from Bernhard Ebbinghaus and Jelle Visser (2000) except for Australia, Japan, the United Kingdom, and the United States: pre-1990 figures for these countries from Visser (1996) and post-1990 figures provided by Ebbinghaus; the following observations were extrapolated: all countries for 2001, Switzerland for 2002 and 2003, Sweden for 2002, Finland for 2002 and 2003, the Netherlands for 2002 and 2003, France for 2002, and Germany for 2002; median position: transformed Kim-Fording measure, based on data downloaded from HeeMin Kim’s home page, accessed April 15, 2007 at http://www .garnet.acns.fsu.edu%7Ehkim/; effective number of parties: based on a measure developed by Laakso and Taagepera (1979); data from Armingeon et al. (2004); updated for 2003 based on CMP data in Klingemann et al. (2006).

Notes 1.

2.

3. 4.

As we explain later, data availability determined the countries included in our analysis. The twelve countries included are Australia, Belgium, Britain, Denmark, Finland, France, Germany, Italy, the Netherlands, Norway, Sweden, and the United States. Altogether, our analysis encompasses sixtyeight election-years over the period 1974 to 2003. In this respect, our main claim to novelty is that we apply partisan theory to the question of how income distribution affects politics. Most existing alternatives to the Meltzer-Richard model (for example, Iversen and Soskice 2001; Moene and Wallerstein 2001) share—or at least do not challenge— the assumption that the median voter determines government policy. The notable exception represented by Woojin Lee and John Roemer (2005) informs our own discussion. See the appendix for a list of our data sources and tables 10.1, 10.2, 10.3, and 10A.1 for summary statistics. Another reason for measuring household inequality in terms of disposable income is that it enables us to include Belgium, France, and Italy in our

350

5.

6.

7.

8.

9.

10. 11. 12. 13.

14.

Democracy, Inequality, and Representation analysis. The LIS database does not allow for the calculation of household market income for these countries. Note also that the measure of household income inequality used here adjusts for household size based on the conventional LIS formula. There is a good deal of evidence to suggest that perceptions of “legitimate income differentials” vary across countries (see Svallfors 2006, ch. 4). We plan to explore the relevance of such perceptions in future work. In the new OECD data set, Belgium and Norway stand out as the two countries with the most compressed distribution of wages in the late 1990s and early 2000s (90/10 ratios of 1.96 and 2.00, respectively, in 2000). In earlier OECD data sets, Norway had the lowest 90/10 ratio (1.99) and Belgium the third-lowest (2.24), with Sweden in second place, in 1993. In our view, the two data sets are sufficiently in agreement to justify using the old measures for these two countries. The same does not hold for Canada. See David Armstrong and Ryan Bakker (2006) for a review of alternative methods for extracting a left-right dimension from CMP data. As the authors pointed out, the measures generated by these techniques are highly correlated with the conventional CMP left-right index. For left parties, the coding scheme presented in table 10.2 is unproblematic, because the same party won the largest share of left votes in every election included in our data set. For most countries, the coding of main right parties is also straightforward, but the Italian case is problematic, since Forza Italia displaced the Christian Democrats as the main party of the right in the election of 1994. Recoding “main right” for Italy in 1994 and 1996 does not significantly alter the findings reported here. Note also that the left-right scores for Belgian socialists and Christian Democrats used here are the average for Flemish and French-speaking parties. The time series are of different duration for each country. For some countries in our sample, the last available election falls as early as 1996 (this is the case with Italy), while for a few others we have data after 2000. The composition of the cross-country mean should be kept in mind when looking at the observations after 1996, since they may reflect the countries included in the measure rather than a general pattern. For non-election years, our source on voter turnout (Armingeon et al. 2004) records the turnout figure for the previous election. As was the case with party data, the time series are of different duration for each country. See note 9 for details. These include the interaction between inequality and mobilization in some regressions. An alternative would be to estimate models with fixed effects, but our need to include (almost) time-invariant explanatory variables in the analysis, like the effective number of parties, makes this impossible. For details on estimating random effects with panel data, see Cheng Hsiao (1986). Needless to say, perhaps, the size of the coefficients for wage inequality and

Inequality as a Source of Political Polarization

15. 16. 17. 18.

351

household inequality should not be compared with each other, since the metrics for these variables are very different (see table 10A.1). This is from a LIS survey conducted in 1991. This is the average from LIS surveys conducted in 1997 and 2000. We are rounding the value for Britain in 1979 to 0. The Swedish Social Democrats moved from –23.9 on the left-right scale in 1988 to –3.52 in 1998. It should be noted that the CMP data set reports a position of (+) 23.79 for 1994. This is surely a measurement error, but the 1994 observation is nonetheless included in our regression analysis, and it is undoubtedly an outlier that works against us.

References Aldrich, John. 1995. Why Parties? Chicago, Ill.: University of Chicago Press. Armingeon, Klaus, Philipp Leimgruber, Michelle Beyeler, and Sarah Menegale. 2004. “Comparative Political Data Set, 1960–2000.” Berne, Switzerland: University of Berne, Institute of Political Science. Armstrong, David, and Ryan Bakker. 2006. “Take That, You Lousy Dimension.” Unpublished paper, University of Maryland and University of North Carolina. Barnes, Lucy. 2006. “The Income Distribution of Voters.” Unpublished paper. Harvard University. Bartels, Larry. 2008. Unequal Democracy: The Political Economy of the New Gilded Age. New York: Russell Sage Foundation. Benoit, Kenneth, and Michael Laver. 2006. Party Policy in Modern Democracies. New York: Oxford University Press. Bradley, David, Evelyne Huber, Stephanie Möller, François Nielsen, and John D. Stephens. 2003. “Distribution and Redistribution in Postindustrial Democracies.” World Politics 55(2): 193–228. Budge, Ian, and Richard Hofferbert. 1990. “Mandates and Policy Outputs.” American Political Science Review 84(1): 111–31. Budge, Ian, Hans-Dieter Klingemann, Andrea Volkens, Judith Bara, and Eric Tanenbaum. 2001. Mapping Policy Preferences. Oxford: Oxford University Press. Cox, Gary. 1990. “Centripetal and Centrifugal Incentives in Electoral Systems.” American Journal of Political Science 34(4): 903–35. Ebbinghaus, Bernhard, and Jelle Visser. 2000. Trade Unions in Western Europe Since 1945. London: Macmillan. Esping-Andersen, Gøsta. 1990. The Three Worlds of Welfare Capitalism. Princeton, N.J.: Princeton University Press. Gabel, Matthew, and John Huber. 2000. “Putting Parties in Their Place.” American Journal of Political Science 44(1): 94–103. Garrett, Geoffrey. 1998. Partisan Politics in the Global Economy. New York: Cambridge University Press. Hacker, Jacob, and Paul Pierson. 2005. Off Center: The Republican Revolution and the Erosion of American Democracy. New Haven, Conn.: Yale University Press.

352

Democracy, Inequality, and Representation

Hibbs, Douglas. 1987. The Political Economy of Industrial Democracies. Cambridge, Mass.: Harvard University Press. Hsiao, Cheng. 1986. Analysis of Panel Data. New York: Cambridge University Press. Iversen, Torben, and David Soskice. 2001. “An Asset Theory of Social Policy Preferences.” American Political Science Review 95(4): 875–93. ———. 2006. “Electoral Systems and the Politics of Coalitions.” American Political Science Review 100(2): 165–81. Kenworthy, Lane, and Jonas Pontusson. 2005. “Rising Inequality and the Politics of Redistribution in Affluent Countries.” Perspectives on Politics 3(3): 449–71. Kim, HeeMin, and Richard Fording. 1998. “Voter Ideology in Western Democracies, 1946–1989.” European Journal of Political Research 33(1): 73–97. ———. 2003. “Voter Ideology in Western Democracies: An Update.” European Journal of Political Research 42(1): 95–105. Klingemann, Hans-Dieter, Andrea Volkens, Judith Bara, and Ian Budge. 2006. Mapping Policy Preferences II. Oxford: Oxford University Press. Kumlin, Staffan, and Stefan Svallfors. 2007. “Social Stratification and Political Articulation.” In Social Justice, Legitimacy, and the Welfare State, edited by Steffen Mau and Benjamin Veghte. Aldershot, UK: Ashgate. Laakso, Markku, and Rein Taagepera. 1979. “Effective Number of Parties: A Measure with Application to West Europe.” Comparative Political Studies 12(1): 3–27. Laver, Michael, and Ian Budge. 1992. Party Policy and Government Coalitions. New York: St. Martin’s Press. Lee, Woojin, and John Roemer. 2005. “The Rise and Fall of Unionized Labor Markets.” Economic Journal 115(500): 28–67. Mahler, Vincent. 2006. “Income Redistribution by the State.” Paper presented to the annual meeting of the American Political Science Association. Philadelphia, Pa., August 30–September 3, 2006. McCarty, Nolan, Keith Poole, and Howard Rosenthal. 2006. Polarized America: The Dance of Ideology and Unequal Riches. Cambridge, Mass.: MIT Press. McDonald, Michael, and Myunghee Kim. n.d. “Cross-National Comparisons of Party Left-Right Positions.” Unpublished paper. Binghamton University. McDonald, Michael, and Silvia Mendes. 2001. “The Policy Space of Party Manifestos.” In Estimating the Policy Positions of Political Actors, edited by Michael Laver. London: Routledge. Meltzer, Allan, and Scott Richard. 1981. “A Rational Theory of the Size of Government.” Journal of Political Economy 89(5): 914–27. Moene, Karl Ove, and Michael Wallerstein. 2001. “Inequality, Social Insurance, and Redistribution.” American Political Science Review 95(4): 859–74. ———. 2003. “Earnings Inequality and Welfare Spending.” World Politics 55(4): 485–516. Nelson, Phillip. 1999. “Redistribution and the Income of the Median Voter.” Public Choice 98(1–2): 187–94.

Inequality as a Source of Political Polarization

353

Organization for Economic Cooperation and Development (OECD). 1999. Decile Earnings Database. Paris: OECD. ———. 2004. Decile Earnings Database. Paris: OECD. Pontusson, Jonas, and Heyok Yong Kwon. 2006. “Power Resource Theory Revisited and Revised.” Paper presented to the annual meeting of the American Political Science Association. Philadelphia, Pa., August 30–September 3, 2006. Powell, Bingham. 1982. Contemporary Democracies. Cambridge, Mass.: Harvard University Press. ———. 2000. Elections as Instruments of Democracy. New Haven, Conn.: Yale University Press. Romer, Thomas. 1975. “Individual Welfare, Majority Voting, and the Properties of a Linear Income Tax.” Journal of Public Economics 4(2): 163–85. Strom, Kaare. 1990. “A Behavioral Theory of Competitive Political Parties.” American Journal of Political Science 34(2): 565–98. Svallfors, Stefan. 2006. The Moral Economy of Class. Stanford, Calif.: Stanford University Press. Visser, Jelle. 1996. “Unionization Trends Revisited.” Unpublished paper. Center for Research of European Societies and Industrial Relations, Amsterdam.

Chapter 11

Inequality and Institutions: What Theory, History, and (Some) Data Tell Us RONALD ROGOWSKI AND DUNCAN C. MACRAE

That institutions covary with political and economic inequality seems obvious. Societies with feudal or clientelistic politics are characterized by extreme economic inequality, and democracies are associated (despite some notable exceptions) with greater economic equality than autocracies. Even within the set of democracies, institutions and inequality seem to move together. Countries with proportional methods of election, for example, display greater economic equality than countries like the United States that have majoritarian electoral institutions. But do the institutions cause the inequality, does inequality constrain institutions, or is this link caused by some more fundamental source of change, such as technology or trade? In this chapter, we attempt to place some of the findings of this book in historical context. In the process, we emphasize that changes in economic and military technology, trade, and factor endowments influence the evolution of political institutions. We note that changes in technology, trade, or factor endowments can dramatically increase or decrease social and economic inequality. Where these exogenous changes increase inequality, we argue that political entrepreneurs have incentives to adopt less representative political institutions—or to do away with democratic institutions altogether. By contrast, decreasing inequality creates incentives for political entrepreneurs to broaden political participation. To support this argument, we present several historical case studies where substantial changes in factor endowments or technology quickly led to more inclusive or exclusive political institutions. We also present quantitative

Inequality and Institutions

355

evidence that increasing labor force participation and the demand for labor created by the two world wars encouraged European countries to expand the right to vote during the late nineteenth and early twentieth centuries. Previous chapters in this book have addressed some of the outstanding challenges faced by the literature on the link between inequality and political institutions and policy. In chapter 2, Andrea Brandolini and Timothy Smeeding tackle the first of these challenges: measuring inequality trends. In the process, they point out that patterns in inequality can vary depending on the portion of the income distribution that is measured, on whether the measure tracks earnings or income, and on whether income measures include in-kind government transfers like education or health care. On the policy side of the equation, Lyle Scruggs (chapter 3) describes trends in welfare state benefits in a number of developed countries since 1970. Better understanding of these measurement issues and efforts to develop more complete indicators of long-term historical trends are critical to advancing our understanding of the basic empirical facts that support inferences about how inequality and political institutions are related. A second challenge, noted by a number of authors in this volume, is the debate about the direction of the causal relationship between inequality and political institutions such as the electoral system, centralized wage-bargaining institutions, and the ideological composition of government and policy. Many political scientists and economists who have addressed the issue argue that political institutions affect inequality. In chapter 4, for example, Torben Iversen and David Soskice present a model that shows that center-left governments dominate under proportional representation (PR) electoral systems, while center-right governments dominate under majoritarian systems, and that PR systems redistribute more than majoritarian systems and lead to lower inequality. G. Bingham Powell (2002) also demonstrated a similar connection between the electoral system, the ideological composition of governments, and redistribution.1 Centralized wage-bargaining institutions, present since the 1930s in Northern European countries with strong welfare states, are also frequently assigned a causal role in producing more equal societies. Using these institutions, strong unions can negotiate social protection in return for more moderate wage demands. Pablo Beramendi and Thomas Cusack take up this theme in chapter 5, arguing that centralized wage-bargaining institutions allow left wing political parties to choose policies that lead to a more equal distribution of income and constrain right wing parties. In chapter 6, David Rueda agrees that centralized wage-bargaining institutions can play a key role in affecting the level of inequality. In his

356

Democracy, Inequality, and Representation

view, however, political parties can take up the slack if these institutions are weak. A number of recent papers, including some of the chapters in this volume, flip this causal relationship between inequality and political institutions on its head. In a pioneering paper, Davide Ticchi and Andrea Vindigni (2003) argue that economic inequality is likely to constrain the choice of electoral system made by democracies. Based both on a cunning model and on an anecdotal survey of twentieth-century history, they contend that high inequality leads the median voter to choose a majoritarian electoral system, while low inequality leads to a rational preference for proportional representation.2 Robert Franzese and Jude Hays (chapter 8) find that inequality is an important cause of social policy generosity. In chapter 9, Christopher Anderson and Pablo Beramendi argue that increasing income inequality depresses electoral participation. Recent work by Kenneth Scheve and David Stasavage (2007) also found little evidence for an effect of either partisanship or wage-bargaining institutions on inequality prior to the 1970s, based on a new long-term inequality measure derived from tax data. They argue that we must look to some underlying economic or political process that might have affected both inequality and political institutions in the early part of the twentieth century and suggest as possibilities both wartime taxation and technological change.3 Finally, economic historians have argued that inequality influences the long-run development of democratic institutions. In their study of the uneven history of franchise extension in the countries of the New World, Stanley Engerman and Kenneth Sokoloff (2002) found that colonial-era inequality of wealth—particularly large grants of land to privileged elites in many Latin American countries—led directly to narrow participation and continued political inequality, extending down to the present day. Similarly, they argue that the franchise was extended earliest in U.S. states with high land-labor ratios— and hence, in their view, with high wages and low social inequality (Engerman and Sokoloff 2001). Carles Boix (2003, especially chaps. 2 and 3) shows a strong link between rising income equality, as measured by the Gini index, and the emergence and survival of democracy after 1950 (see also Boix and Garicano 2002). Boix also found that prior to 1850 reasonable proxies for social equality predict the probability of transition to, and the survival of, democracy.4 This result holds even when he controls for wealth, as measured by per capita GDP. Boix’s game-theoretic model of political transitions also specifies the expected direction of causation: “increasing levels of economic equality bolster the chances of democracy” (2003, 10). In Democracy and Development, Adam Przeworski and his colleagues (2000) found that higher GDP per capita made democracy likelier to sur-

Inequality and Institutions

357

vive in the period 1950 to 1990. But an important subfinding, somewhat downplayed because the data were sparse, reinforces the impression that inequality affects institutions. According to these authors, greater equality produced higher odds of democratic survival, while an increase in equality made it likelier that dictatorship would yield to democracy. Specifically, they find that democracies with Gini indices above the median were over four times as likely to fail as those with below-median inequality.5 In addition, dictatorships that experienced a decrease in inequality were more than twice as likely to yield to a democracy as dictatorships that experienced an increase in inequality.6 We incline, on logical and historical grounds, to the possibility that both equality and institutions are affected by exogenous social change, usually occasioned by innovations in technology, trade, demographics, or some combination of these three. In our view, the causal sequence is almost always 1.

Social change, which leads to . . .

2.

Change in inequality, followed (usually in short order) by . . .

3.

Change in institutions, which, in turn, may occasion . . .

4.

Further change in inequality.

To support our contention, we advance a rudimentary model and two kinds of evidence. On the modeling front, we argue that standard production functions and utilitarian welfare maximization lead logically from equality of skills and endowments to economic, social, and political equality. Empirically, we first present a series of nine historical cases, which range from the rise of democracy in ancient Greece to the effects of the two world wars. For each, we summarize the mainstream opinion among historians. In every case, with but slight nuance, we observe changes in technology or trade leading to changes in economic inequality and political institutions. Second, we survey previous work on democracy and franchise expansion and supplement those analyses with an empirical analysis of the expansion of the right to vote in nine European countries between 1840 and 1944. We conclude, in line with previous historical work, that the expansion of the franchise around World War I and World War II in Europe was probably brought on by two main factors: changes in technology and trade that diminished social and economic inequality; and the war-driven entry of new groups, including women, into the labor force.

358

Democracy, Inequality, and Representation

A Simple Model of Political Inequality We break little new theoretical ground here but chiefly emphasize two points that follow from existing models: exogenous shifts in demography, investment, cross-border trade, or technology can profoundly affect economic inequality; and the greater the economic inequality that prevails in a society, the greater are the welfare losses from democracy.

Exogenous Changes in Inequality The ratio of labor wages (w) to the rent of capital (r) serves as a plausible measure of equality that has been used empirically to good effect in recent economic history (O’Rourke and Williamson 1999, especially chap. 4). In the simple Cobb-Douglas production function shown in equation 11.1, we see that the ratio of the wages of labor to the rent of capital (w/r) rises linearly with the capital-labor ratio (K/L). If, in the standard notation, Y = AKαL1– á, 0 < α < 1, then w/r, the wage-rental ratio, is just7 w (1 − α ) ⎛ K ⎞ = ⎜ ⎟ α ⎝ L⎠ r

(11.1)

Any exogenous event that decreases the supply of labor while holding the supply of capital constant, such as the Black Death (discussed later in the chapter), raises the wage-rental ratio and makes society more equal. On the other hand, a major war that destroys capital but leaves most labor intact—and this is what happened during World War II in most belligerent countries—(all else being equal) increases inequality. The standard theory of international trade, embodied in the HeckscherOhlin and Stolper-Samuelson theorems, simply puts the same point in a context of cross-border exchange. When trade opens between a capitalabundant country and a labor-abundant country, inequality increases in the former (which, by importing labor-intensive goods, has tapped into a larger pool of labor) and diminishes in the latter (which, by importing capital-intensive goods and services, has effectively increased its supply of capital). Similarly, any technological change that increases the relative importance of capital in production—that raises the value of α—lowers the wage-rental ratio and increases inequality. By contrast, any change that makes labor relatively more productive—decreases α—would make for greater equality, that is, an increased wage-rental ratio. We expand this last point to consider what seems particularly crucial within and between states, namely, the production of military power.8 If

Inequality and Institutions

359

some shift of technology, like the introduction of better infantry tactics, increases the marginal productivity of labor in the production of military force, labor becomes more valued relative to capital. The wage of this labor, whether in monetary or political terms, rises and inequality decreases. Similarly, any technological shift that increases the relative marginal productivity of capital increases inequality. As we shall see later, the rise of armored knights at the advent of feudalism seems to have been exactly such a shift. It is standard to note that any change in the return to a factor (for example, wage of labor or rent of capital) also affects the value of endowments of that factor, unless the change is assumed to be transitory. When imports from the New World lowered European grain prices to a fraction of their former levels in the latter half of the nineteenth century, for example, both land rents and land prices collapsed in all countries that remained open to trade (O’Rourke and Williamson 1999, chaps. 3 and 4). In short, returns to factors affect factor prices—the value of individuals’ endowments.

Welfare Consequences of Changes in Inequality We begin our model by recalling a general and important proposition, namely, that social welfare is usually maximized by adopting the policy preferred by the average, rather than the median, citizen. Based on this proposition, we show that increasing the distance between the average income and the median income—increasing social inequality—also increases the welfare loss from representative democratic institutions, which, we assume, enact the preferences of the median citizen. Following Torsten Persson and Guido Tabellini (2000, 48–49) in both notation and substance, let citizens be of different types indexed by i. Each citizen has quasi-linear preferences expressed in equation 11.2. In this setup, c i is the private consumption of the ith individual, g is either a “pure” public good or a publicly provided private good that must be provided in exactly the same non-negative amount to every citizen, and H(·) is a continuous and concave function (Hg > 0, Hgg < 0), identical among all citizens. w i = c i + H(g)

(11.2)

For simplicity, we assume that government supplies g by imposing a flat-rate tax τ on each individual’s income yi, so that c i = (1 – τ)yi. We further assume that the population is of size (mass) unity, so that the government budget constraint (letting y denote average income) is simply

360

Democracy, Inequality, and Representation

τy = g (whence it of course follows that τ = g/y). Finally, we assume that yi is linear in some endowment (for example, land or skill) di, so that yi = adi and y = ad, where a is some positive constant and d is average endowment. Thus, we can transform our original statement of citizen welfare to: g⎞ ⎛ w i = ⎜1 − ⎟ ad i + H ( g) ⎝ ad ⎠

(11.3)

This is maximized where ∂wi/∂g = – di/d + Hg(g) = 0, implying that individual i’s optimal level of supply of the publicly provided good g is given as: ⎛ di ⎞ gi = H −1 g ⎜ ⎟ ⎝d⎠

(11.4)

Since H is stipulated as concave, Hg is positive but decreasing in g; this implies that Hg–1 is also positive and decreasing. Thus, we know that the greater a citizen’s endowment (and hence income), the less g she prefers. By the standard utilitarian criterion, social welfare is taken as identical to the summed welfare of all individual citizens, but this is maximized when (and only when) the welfare of the average citizen—that is, the citizen of average endowment—is maximized.9 Thus, the socially optimal level of g is given by: H −g 1

d = H −g 1(1) d

(11.5)

In a democracy, however, the level of g provided under majority rule is that of the citizen of median endowment—the one at exactly the fiftieth percentile of endowment—which we denote dm. In virtually all known democracies, the income of the median voter is less than that of the average voter (ym < y), implying (in this setup) that the endowment of the median voter is also less than that of the average voter (dm < d). Thus, it follows that dm/d < 1, and hence that normally democracy overprovides the public good from a social welfare perspective. In a simpleminded partial-equilibrium sense, this insight leads to the prediction that, since under democracy the median voter prevails, greater inequality (and consequently greater distance between the median and the average voter) entails more redistribution (Meltzer and Richard 1981). If instead we seek to endogenize institutional choice, we gain a quite

Inequality and Institutions Figure 11.1

361

Relative Positions of Median and Average Incomes for Different Institutional Choices

ym

yPR

yMAJ

yLIM

yABS

Source: Authors’ compilation. Note: Figure assumes fixed median income ( y m) and displays the ordinal location of average income under proportional representation ( y, PR), majoritarian democracy ( y, MAJ), democracy with limited franchise ( y, LIM), and absolutism ( y ABS).

different insight: the greater the difference between the median and the average voter, the greater the welfare gains from making institutions less democratic. Less democratic institutions move policy toward the position of the richer and less redistributive average voter (see figure 11.1). As a result, greater inequality of endowments increases the incentives of, and the potential resources available to, political entrepreneurs to bring about institutional change.10 In a highly equal society, where by definition the median and the average voter are almost identical, there is little or no welfare loss from adopting a PR democracy; the system that we now know empirically most reliably yields the policy desired by the median voter (Huber and Powell 1994). As society becomes more unequal, the gap between the mean and median incomes grows. As a practical matter, the income distribution can change shape in any number of different ways. Greater inequality, for example, may come about because of gains at the top of the income distribution, or losses at the bottom, or both. The point we wish to emphasize, however, is that an increasing gap between the median and mean incomes moves the average income (and endowment) “up” the distribution to a higher percentile in every case. Growing inequality thus makes the average voter less favorable to redistribution. If a highly equal society were to become less equal, there could be welfare gains from adopting (or retaining) a majoritarian democracy. Majoritarian institutions have been shown empirically to produce policies that, on average, are to the (less redistributional) “right” of the median voter.11 As inequality rises further, and the policy distance between the average and median citizen increases again, there can be welfare gains from a restricted, typically property-based, franchise. Again, this restricted franchise would yield a policy that is closer to the ideal point of the average citizen. At the extreme of inequality, where the citi-

362

Democracy, Inequality, and Representation

zen of “average” endowments might be literally in the top percentile, the welfare-maximizing franchise might become indistinguishable from aristocracy or monarchy. Suppose that, as in the case of predemocratic ancient Greece (discussed in the historical case section), all political power is vested in some small fraction—say, the top twentieth—of the income distribution. If initially this group faithfully represents the person of average income in the society, welfare is maximized. Now suppose, as the historians believe happened in the Greek case, exogenous technological and commercial changes so devalue traditional endowments—like land and military skills—that the average citizen is suddenly to be found in the seventieth percentile of the income distribution rather than in the top twentieth. Under this scenario, continuing a policy that embodies the bliss point of the top twentieth entails severe welfare losses. Presumably society as a whole will be better off moving to a system of governance that brings policy closer to what the new average citizen wants. In our picture, political entrepreneurs, and sometimes professional revolutionaries, do exactly that—essentially speculating on their share of the welfare improvement that will result. If, as at the rise of feudalism, the average endowment suddenly moves from something not too distant from the median—say, even the seventieth percentile—to the ninety-ninth percentile, then again there will be welfare improvements from an institutional change that guarantees a policy closer to that of the new average citizen. As before, political entrepreneurs can be expected to work toward such a change, producing a less democratic society.

What Historians Tell Us We might expect that historians, including economic historians, would simply remain agnostic on the question of cause and effect, but at least as we read them, scholars in these disciplines accept almost unanimously that social changes cause changes in inequality and institutions. More precisely, historians generally argue that major changes to institutions are usually explained by exogenous shocks that increase or decrease inequality of endowments. We set these stories of rapid change in institutions against explanations that invoke “destiny,” or perhaps “original sin,” to explain enduring patterns of institutions, policy, and (often) economic stagnation or growth. A leading example of the latter is the brilliant and hugely influential paper by Daren Acemoglu, Simon Johnson, and James Robinson, “Reversal of Fortune,” which argues that greater precolonial wealth, population density, and inequality of conquered territories gave colonial powers incentives to

Inequality and Institutions

363

impose extractive institutions (2002). The evil legacies of these institutions—continued inequality, governmental predation, and bad policy—allegedly hinder economic growth down to the present day. In some sense, all seems to have been determined by conditions at the moment of European conquest. Here we consider a series of historical accounts in which traditional, and seemingly unchangeable, political institutions in fact were transformed rapidly. We proceed roughly in chronological order.

Ancient Greek Democracy Over about a century and a half, many of the ancient Greek city-states, starting probably with Corinth and culminating in Athens, shifted from the kind of aristocratic exclusiveness and arrogance so un-self-consciously portrayed in the Iliad to a form of democracy12 (albeit limited to free males) far more extreme than anything known to the modern world.13 Historians almost unanimously, and in our opinion quite persuasively, explain this rapid change of institutions as a result of two factors. First, a rapidly growing population density induced an economic change from self-sufficient agriculture to extensive trading and manufacture for export. This economic change reduced the value of land and increased that of labor and human capital, giving rise to both a prosperous middle class and vastly better-off wage earners. Second, military technology shifted from the use of armored knights—which only the rich could become because training and armor were privately supplied—to the use of “hoplites,” or lightly armored infantry working in close formation. Naval power—with ships financed from state revenues or by forced “contributions” from the rich—also began to play a larger role, especially in Athens. These changes in military technology made the middle class and wage earners crucial to successful warfare. The considerable investments of the elite in training and armor became almost insignificant militarily. Technological and economic changes, in short, greatly reduced the previous inequality of endowments and of income, and institutional change quickly followed.14

The Rise and Fall of the Roman Republic Pre-Republican Rome had an aristocratic governance structure similar to that of predemocratic Greece. Rome also initially experienced a military revolution similar to the one in Greece, but witnessed very little economic change. As a result, Roman institutions democratized to a lesser extent than Greek ones. In the Roman Republic, the powers of the plebeian comitia tributa and comitia centuriata expanded. Commoners—

364

Democracy, Inequality, and Representation

mostly smallholding farmers producing for urban markets—gained tribunes and plebiscites. And the Senate ceased to be purely aristocratic but retained far more power than the analogous Athenian Areopagus. Traditional elite endowments in land became militarily less valuable but retained their economic worth. Political institutions changed, but only within limits. This quasi-democratic, or “mixed,” constitution persisted only until the Second Punic War. This conflict ruined the smallholding peasantry by conscripting the peasant infantrymen for years at a stretch and assuring Roman access to cheap, slave-produced grain from North Africa. The war also proved the military superiority of professional legions, paid and equipped by the state, over conscript citizen-soldiers. Smallholders’ endowments lost both economic and military value. Valiant efforts to combat the growing inequality politically, for example, through radical land reform, failed.15 Increased inequality ultimately led to Caesarism, in which the only institutions that mattered were the plutocratic Senate and, increasingly, the professional army. In the Roman case, endowments first became more equal and then, after the Second Punic War, radically less equal. These changes in turn altered income inequality and caused political institutions to first become more democratic and then more autocratic.

The Rise of Feudalism Contrary to our usual impressions, the long decline of the Roman Empire was not paralleled by a slow evolution toward feudalism. Rather, the main institutions of feudalism—benefice, commendation, and vassalage—arose with startling rapidity between about 700 and 750 CE from what had been a much more egalitarian society (Ganshof 1952; Pirenne 1936). Moreover, these feudal institutions emerged virtually full-blown in the Merovingian Frankish kingdom governed by Charles Martel.16 After much dispute about the causes of this sudden change, the eminent medieval historian Lynn White Jr. (1962, chap. 1) offered a bold hypothesis: that the direct source of the rapid economic and political change had been a new revolution in the technology of warfare, which made armored knights once again—and for the first time since the rise of the hoplites— the overwhelmingly dominant type of warrior. The crucial technological change, White argued, was the introduction (probably from the Near East) of the stirrup. This innovation permitted mounted warriors to lunge against targets with the full momentum of horse and rider without running the risk of being thrust off over the horse’s tail. But knights capable of exploiting this new technology were extremely expensive to train and equip, just as they had been in ancient Greece.

Inequality and Institutions

365

Martel solved the problem by reallocating endowments: giving huge estates, mostly carved out of church lands, to his principal followers in return for their provision of fixed quantities of properly equipped knights. This reduced the formerly independent peasants to the status of tenants or forced laborers on the newly created estates. Martel demonstrated the great superiority of the new technology and his new institutions in a long series of successful conflicts, culminating in the (Huntingtonian) Battle of Poitiers. Virtually all of the rest of Europe quickly emulated him. Again, the short version is: technological change drives changes in inequality and in institutions.

The “Democratizing” Black Death Between 1347 and 1351, bubonic plague killed at least one-third of Europe’s total population. It killed an even higher fraction of the skilled craftsmen and merchants who lived in the nascent cities. In simple economic terms, labor suddenly became scarce relative to land and capital— which, of course, the plague did not destroy—and skilled urban labor became particularly scarce. The result, even in a still rather tradition-bound economy, was a sharp increase in real wages, and particularly in the returns to skill. Servile peasants soon fled their feudal lords for the higher wages of the cities. Often lords found that they could retain their tenants only by commutation of traditional service dues—so many days of labor each week on the lord’s lands—to cash rents on generous terms. The lords’ political control over their tenants also diminished. Within the cities, workers’ and artisans’ relative scarcity, and their consequently more favorable bargaining position and higher wages, led to increasing demands for a greater share of political power. An exogenous disaster led to increasing social and economic equality, which led in turn to greater political equality and participation.17

The Reformation In the traditional iconography, the Reformation—that great democratizing movement within Christianity, characterized above all by the doctrines of individual salvation, individual access to the Scriptures, and the “priesthood of all believers”—is captured in images of individual reasoning, courage, and revolt: Wycliffe translating the Bible into the vernacular; Huss dying at the stake; Luther nailing his Ninety-Five Theses (in Latin, but quickly translated into German and broadcast throughout Germany) to the church door. In fact, most historians now accept that the Reformation grew directly out of three interrelated sociotechnical changes that radically equalized societies (Eisenstein 1979).

366

Democracy, Inequality, and Representation

The first of these changes was growing lay literacy from about 1350. Literacy increased particularly among merchants and skilled artisans, probably as an outgrowth of the Black Death and the growing returns to skill. For obvious commercial reasons, literacy expanded chiefly in vernacular tongues and much less so in Latin. Demand for written works in the vernacular grew rapidly. The supply of written works, however, was constrained by labor-intensive hand-copying and had been traditionally focused on the reproduction of works in Latin. Book prices, of course, skyrocketed, leading directly to Gutenberg’s invention of movable type, and hence of modern printing and mass production of written works, around 1450. Printing drastically cheapened written works and made books more available to the public, which further increased literacy. Finally, the rapid translation and printing of the Bible into most leading European vernaculars permitted the increasingly literate lay population direct access to the sacred texts. It also revealed to many of them how ignorant of Scripture many priests and bishops actually were. As Elizabeth Eisenstein (1979) and others have shown, the geographical link between capitalism and Protestantism is tenuous (pace Weber), but the links of place and timing between printing and Protestantism are extremely strong. The social and technological changes that created greater equality between clergy and laity led directly to institutional changes that recognized that equality.

The Rise of Absolutism In a very short interval between about 1625 and 1660, many of the weak and decentralized states typical of European feudalism rapidly gave way to a highly centralized and powerful monarchy. This occurred most notably in France under Richelieu, in Prussia under the Great Elector, and in Sweden under Gustavus Adolphus. Traditionally powerful councils or parliaments of feudal elites were tamed or abolished. What quickly emerged became known, with some justice, as “absolutist” monarchy.18 Virtually every student of the period, including the estimable Samuel Finer (1975), has linked this sudden institutional change to yet another major change in military technology, the “military revolution of the seventeenth century” (Downing 1992; Eltis 1995).19 The combination of longbows, crossbows, and firearms had earlier broken the military monopoly of the mounted knights, but now the new tactics (pioneered by the Swiss) of the “tercio” proved superior to all rivals, particularly in the incessant combat of the Thirty Years’ War (1618 to 1648). Like the Roman legions of the Imperial period, the new military technique depended on well-drilled infantry. It also permitted—and com-

Inequality and Institutions

367

petitive pressure soon required—much larger armies.20 Both the numbers and the more extensive training of the newer troops required vast new revenues that traditional bodies would never approve—indeed, which could only be imposed by force. John Ulric Nef (1940, 126–29, cited in Finer 1975, 128), for example, estimated that the total burden of taxation in France increased by more than an order of magnitude between 1540 and 1640, and grew two and a half times between about 1600 and 1640. But once the monarch had gained those revenues and suppressed the traditional bodies, he possessed a standing army that could, and normally did, make him absolute. Here the change in military technology increased inequality of endowments and income. The monarch became not just the richest and most powerful of nobles but, by his control of the new state apparatus, orders of magnitude greater than his strongest potential rival. Political influence, consequently, was restricted to a far smaller circle. This change, of course, did not come without resistance. In the bloody “frondes” of the midseventeenth century, both noble and popular resistance almost overthrew the French monarchy.

The Age of Democratic Revolution From about the 1780s, at least in the economically most advanced societies, political participation began to broaden. Parliaments again became more powerful, and the parliamentary franchise was extended, leading ultimately—and certainly with authoritarian reversals along the way—to the representative democracies that now prevail almost universally in the wealthy industrial economies. The most perceptive contemporary observers of this “first wave” of democratization, including above all Alexis de Tocqueville (1969, 9–20), saw its source unambiguously in the gradual and progressive development of social equality. The more individuals grew to be equal in such endowments as capital, literacy, military contribution, and knowledge, the less able they were to tolerate inequality of political participation. Subsequent historians have confirmed Tocqueville’s claim of rising equality, at least for prerevolutionary French rural society where peasants’ ownership of land was expanding (Tocqueville 1955). Across Europe more broadly, historians have pointed to three major sources of rising equality, two of which closely parallel those of ancient Greece. The first of these sources of equality was the “Napoleonic” military revolution. The armies of both the French Republic after 1792 and of Napoleon proved the military superiority of the levée en masse. These were the enormous armies of citizen-soldiers that forced rival powers to turn large numbers of their own subjects into willing—or at least not rebellious—

368

Democracy, Inequality, and Representation

recruits. The second source of equality was the huge expansion of maritime commerce in the eighteenth century,21 which worked to the benefit of commercial elites and, in labor-abundant Europe, of workers.22 Trade increased both within Europe and overseas because of greatly improved shipbuilding and navigation. Finally, the possibility of emigration to the high-wage New World, even if the trip was hard and risky, put a floor under European wages and thus limited the political oppression that European masses found tolerable.23 Within the New World itself, high wages and the open frontier reduced inequality. These high wages also increased pressure for broader political participation, as first argued, at least in traditional accounts, by Frederick Jackson Turner.24 The trend toward greater equality of endowments received a powerful new impetus from the further revolution in shipping and communication of the nineteenth century, particularly during the period from 1870 to 1914. Steamships, railways, canals, the telegraph, and the telephone drastically reduced both the cost of transporting goods and people and the cost of transmitting information. As Kevin O’Rourke and Jeffrey Williamson (1999) showed, one strong and rapid effect was a rise in European wages and a reduction in European inequality, as measured, for example, by wage-rental ratios. And as we show later in the chapter, the expansion of the European franchise seems to have gone hand in hand with these economic changes.

The Two World Wars of the Twentieth Century World Wars I and II were “total” in a way that no wars of the eighteenth or nineteenth century had been—except, perhaps, for the U.S. Civil War. Vast civilian populations were mobilized. Industries were created virtually overnight. New populations were brought into the workforce and the military. In virtually all of the warring powers, for example, women were mobilized into industrial workforces to replace men called away to the front. And in many countries, previously isolated or disadvantaged groups, like African Americans in the United States and peasants in continental Europe, were drawn into urban industry and the military.25 The overwhelming need for workers drastically raised real wages and forced concessions to organized labor—even, notably, in Imperial Germany during World War I (Feldman 1966). Elites, including the traditionally powerful European landed elites, submitted to crushing new tax burdens, capital levies, rent controls, and combat-related destruction of capital; some of these measures and outcomes would have long-lasting effects on overall measures of income inequality.26 In short, military and industrial needs made traditionally inferior endowments worth more, made traditionally superior ones worth less, and

Inequality and Institutions

369

broke down “cultural” barriers to mobility. African Americans, ill-educated peasants, and women, who had been traditionally confined to the home, were suddenly needed for the war effort. The overall result was a great equalization of society that played out politically as a “second wave” of democratization. This second wave included, but was by no means limited to, substantial extensions of the franchise. We analyze the statistical evidence later on. Table 11.1 presents a synopsis of this sprint through history in tabular form. Two major messages emerge from this table, at least for us. First, equality of endowments usually changes with, or even before, changes occur in political institutions in the standard historical accounts. And second, the main exogenous sources of change in inequality and institutions have been changes in military technology, trade, migration, disease, and information technology. Now we may, of course, have gone astray in either of two ways. We may simply have read selectively among historians, embracing those who take the “inequality determines institutions” perspective. Possible alternatives to our view include the older “independence of ideas” perspective that people come to believe in greater equality—used to explain the abolition of slavery or the acceptance of human rights—and that social or economic changes follow (Hunt 2007). Current work by Gregory Clark (2007) also argued that the industrial revolution—and the greater equality that came with it—was at least in part a result of a diffusion of entrepreneurial ideas throughout England. Alternatively, historians may still be burdened with a watered-down Marxist perspective in which change in “conditions and relations of production” conditions all else, and institutions are simply assumed to be “epiphenomenal.” We do not, however, think either of these two possibilities is true. Although we are willing to debate the point, we believe that we have accurately summarized, and offered examples of, the standard thinking among historians about major institutional change, including such emphatically nonMarxist historians as Lynn White Jr. (1962).

The Expansion of the Franchise in Europe So we entertain a third possibility: historians may be taking the long view while political scientists, sociologists, and economists take an appropriately more myopic one. Historians are right about huge sweeping changes like the Black Death or the invention of printing, but perhaps the view that “institutions determine inequality” is right for the shorter and more realistic term that we normally face. This alternate picture of recent history would be consistent with some of the earlier chapters of this book, which give top billing to political institutions as drivers of

370 Table 11.1

Democracy, Inequality, and Representation Some Major Historical Changes in Institutions and Inequality

Institutional Change

Antecedent Change in Inequality

Rise of ancient Greek democracy

Decreased inequality between aristocrats and commoners

Rise of Roman Republic

Decreased inequality between Senatorial and other classes Increased inequality between rich and poor, military and nonmilitary

Fall of Roman Republic, rise of Caesarism

Rise of feudalism Decline of feudalism, democratization of Renaissance cities Reformation: democratization of Church Rise of absolutism

Age of democratic revolution in Europe and North America Second wave of democratization broadening of franchise and participation in Europe and the United States

Increased inequality between lords and peasants Decreased inequality between lords and peasants, urban elites and masses Decreased inequality between laity and clergy Increased inequality between state elites and subjects Decreased inequality between traditional elites (especially landed elites) and masses Further decrease in inequality between owners and workers, males and females, majorities and minorities

Exogenous Shock(s) that Changed Inequality Military technology (rise of hoplites), trade (especially with Magna Graecia, Black Sea regions, Egypt) Military technology (rise of infantry; parallels with Greece) Trade (imports of slaveproduced grain), military technology (superiority of professional soldiers) Military technology: stirrup, mounted knights Disease (Black Death): sudden rise in capital-labor and land-labor ratios Information technology: literacy, printing, vernacular literature Military technology (“military revolution”), crushing burden of taxation Military technology (levée en masse), trade (within Europe and overseas), possibility of emigration Two world wars: insatiable demand for soldiers and workers

Source: Authors’ compilation.

trends in inequality. Again, we think, not so. For one thing, exogenous changes in technology and trade that shift the value of endowments, and hence raise or lower inequality, are hardly a thing of the past. These changes have occurred in recent decades and are occurring today. As we shall see later in the chapter, changes in inequality appear to

Inequality and Institutions

371

have had a significant effect on political institutions in the nineteenth and twentieth centuries. Specifically, we argue that increased labor force participation—an indicator of changing inequality of endowments and of the bargaining power of workers—was strongly associated with an expansion of the right to vote in nine European countries between 1840 and 1944.27 Some impressive studies have begun to entertain the previously heretical opinion that inequality may determine institutions, or at least ideology and policy. As discussed earlier, Przeworski and his colleagues (2000) and Boix and Luis Garicano (2002) showed that reasonable proxies for social equality predict the probability of transition to or survival of democracy. Other scholars have examined the link between inequality and franchise expansion (a more specific indicator of democratization), both across countries and among the U.S. states. Daron Acemoglu and James Robinson (2000, 2006), with typical brilliance and technical finesse, developed a nice model that links franchise expansion to elites’ fear of revolution. The upshot of this is that the likelihood of a democratic transition bears an inverse-U relationship to economic inequality. Democratization is completely unlikely in the most equal societies and in the most unequal, where elites will repress demands, but likeliest at the middle ranges of inequality. Again, none of this settles the question of causal priority empirically, and neither, at least at this point, do we.28 The model and case studies presented earlier argued that increasing equality of endowments in a society made democracy more likely in late nineteenth-century and early twentieth-century Europe. Increasing industrialization raised the level of capital per worker and increased the returns to skill. As in the Reformation, the diffusion of skills and education among the population created a more equal society where workers had the ability to bargain for political power.29 This increase in bargaining power resulted in significant increases in political enfranchisement as measured by the fraction of the adult population eligible to vote. Political enfranchisement accelerated during World Wars I and II. These “total” wars, as is evident even in American history, dramatically affected the demand for labor, raising wages and drawing minorities and women into the labor force in large numbers. The exogenous shock of war thus decreased inequality of endowments and—at least in our interpretation—led to increased political participation. The effect was not limited to the belligerent countries. Neutral Sweden, for example, was a major exporter of high-quality ores and steel and also experienced a wartime boom that greatly increased the demand for labor. Prior to this period of increasing industrialization, the right to vote, if it existed at all, was quite exclusive in most of Europe. Usually the franchise was restricted to men who met particular criteria: property owner-

372

Democracy, Inequality, and Representation

ship, payment of a specified minimum in taxes, literacy, or some combination of these. But between roughly 1850 and the end of World War I, the right to vote was extended—gradually in most cases, rapidly in some—to all adult males. In the Netherlands, for example, the right to vote was restricted between 1849 and 1887 to male citizens age twentythree and over who exceeded a relatively high tax threshold. As a result, only about 5 percent of the adult population—about 11.3 percent of men age twenty and over—were allowed to vote in 1870. The restrictions on male suffrage, however, were gradually reduced in the Netherlands until all men age twenty-five and over were given the right to vote in 1918. With a few exceptions, European countries granted women the right to vote shortly after men had achieved universal suffrage, generally between World War I and the end of World War II. The Netherlands is again fairly typical of the pattern, women age twenty-five and over having achieved the franchise in 1922.30 To assess the extent of the franchise consistently across countries and time, we adopt the measure found in Peter Flora (1983), namely, the fraction of the population age twenty or over who are legally entitled to vote. Figure 11.2 shows the relationship between this measure of political enfranchisement and the fraction of the population in the labor force.31 The graphs in figure 11.2 confirm both of our intuitions about the relationship between labor force participation outlined earlier. First, we see a fairly strong positive correlation (r = 0.62) between time trends in labor force participation and enfranchisement. In Belgium, for example, labor force participation increased from 60.2 percent of the population to 68.6 percent of the population between the period 1885 to 1889 and the period 1940 to 1944. At the same time, enfranchisement increased from 3.9 percent of the population age twenty and over to 45.4 percent of the population. Figure 11.2 also provides further evidence that the two world wars, and especially World War I, had a dramatic effect on the level of political enfranchisement in Europe. Germany is one of the most impressive examples of this effect. Labor force participation and the fraction of the population able to vote both jumped substantially during World War I in Germany. Between the period just before World War I (1910 to 1914) and the period just after the war (1920 to 1924), labor force participation increased from 61.4 percent of the population to 67.5 percent. And the fraction of the population able to vote increased from 38.7 percent to 96.5 percent during the same decade.32 Other countries—among them the United Kingdom, the Netherlands, and Denmark—also significantly expanded the franchise around World War I.33 Labor force participation has obviously continued to evolve since World War II. We argue, however, that changes in labor force participation in the late 1800s and early 1900s were particularly salient because

Figure 11.2

Labor Force Participation and Enfranchisement in Nine Countries, 1830 to 1975

.7

100 80 60

.6 40 20 0

.5 1820

1900

1860

1940

1980

Percentage Enfranchisement

Labor Force Participation

Belgium

Five-Year Average

.7

100 80 60

.6 40 20 0

.5 1820

1900

1860

1940

1980

Percentage Enfranchisement

Labor Force Participation

Denmark

Five-Year Average

.7

100 80 60

.6 40 20 0

.5 1820

1900

1860

1940

1980

Five-Year Average Labor Force Participation Percent Enfranchised

Smoothed Participation Smoothed Enfranchisement

Percentage Enfranchisement

Labor Force Participation

France

Figure 11.2

Continued

100 80 60

.6 40 20 0

.5 1820

1900

1860

1940

1980

Percentage Enfranchisement

Labor Force Participation

Germany .7

Five-Year Average

.7

100 80 60

.6 40 20 0

.5 1820

1900

1860

1940

1980

Percentage Enfranchisement

Labor Force Participation

Italy

Five-Year Average

.7

100 80 60

.6 40 20 0

.5 1820

1900

1860

1940

1980

Five-Year Average Labor Force Participation Percent Enfranchised

Smoothed Participation Smoothed Enfranchisement

Percentage Enfranchisement

Labor Force Participation

Netherlands

Figure 11.2

Continued

.7

100 80 60

.6 40 20 0

.5 1820

1900

1860

1940

1980

Percentage Enfranchisement

Labor Force Participation

Norway

Five-Year Average

100 80 60

.6 40 20 0

.5 1820

1900

1860

1940

1980

Percentage Enfranchisement

Labor Force Participation

Sweden .7

Five-Year Average

.7

100 80 60

.6 40 20 0

.5 1820

1900

1860

1940

1980

Five-Year Average Labor Force Participation Percent Enfranchised Source: Flora (1983), Williamson (n.d.). Note: Vertical bars denote World Wars I and II.

Smoothed Participation Smoothed Enfranchisement

Percentage Enfranchisement

Labor Force Participation

United Kingdom

–3.713*** (0.698) 153 0.83

–4.667*** (0.450) 153 0.82

0.317** (0.095) 3.831*** (1.031) 0.149** (0.063) —

(2)

–5.879*** (0.638) 153 0.80



0.407*** (0.114) 4.642*** (0.986) —

(3)

–2.204*** (0.643) 153 0.80

0.334*** (0.085) 0.374*** (0.096)

0.303*** (0.078) —

(4)

–4.092*** (0.768) 153 0.76

0.263** (0.101) 3.572* (1.681) 0.199* (0.106) 0.168 (0.141)

(5)

–5.116*** (0.729) 153 0.74

0.321** (0.123) 4.470*** (1.170) 0.138* (0.063)

(6)

0.429** (0.152) 5.662*** (1.255)

(7)

–6.737*** (1.153) 153 0.71

GLS

–2.272** (0.869) 153 0.74

0.327*** (0.084) 0.358** (0.107)

0.311** (0.105)

(8)

Source: Authors’ compilation. Note: Robust standard errors clustered on country are in parentheses. Countries in regressions include Belgium, Denmark, France, Germany, Italy, the Netherlands, Norway, Sweden, and the United Kingdom. Observations are five year averages. All models include a full set of country dummies. * p < .10; ** p < .05; ***p < .01

Observations R-squared

Constant

World War II

Inter-war period

L.F. participation

0.265** (0.082) 2.953* (1.493) 0.221* (0.112) 0.198 (0.138)

(1)

OLS

Ordinary Least Squares (OLS) and Generalized Least Squares (GLS) Estimates of Labor Force Participation and Wealth on Political Enfranchisement in Nine Countries, 1840 to 1944

Log(GDP/capita)

Table 11.2

Inequality and Institutions

377

these changes mostly reflected movement out of households and away from farms, where productivity was far lower, and hence tended to improve the lot of workers and reduce inequality in society. Today an increase in labor force participation may reflect movement into part-time employment with poor benefits and often represents the employment of second earners in a family. We might expect that these more modern trends would have less of an impact on equality of endowments than in earlier periods. To model expanding political rights in Europe better we present the regressions in table 11.2. In addition to labor force participation, the regressions include controls for per capita gross domestic product (GDP).34 A number of studies going back to Seymour Lipset (1959) find a strong correlation between wealth and democratic institutions, although the exact cause of this relationship has been debated in the literature (Przeworski and Limongi 1993). We also include dummy variables for the interwar and World War II periods and a full set of country fixed effects.35 One technical issue, of course, is the fact that the fraction of the population who can be allowed to vote is bounded between 0 and 100 percent.36 Ordinary least squares (OLS), however, does not necessarily obey this restriction and so may produce unreasonable predicted values. Also, we expect that the effect of the independent variables changes depending on the level of enfranchisement. It seems unlikely, for example, that growing labor force participation would produce as much of an increase in enfranchisement when close to 100 percent of the population can already vote as when only 30 percent of the population can vote. As a result, OLS should underestimate the effect of the causal variables on enfranchisement. In practice, OLS does a reasonably good job of predicting levels of enfranchisement in this sample. But to compensate for this potential problem we also produce estimates using generalized least squares (GLS). To do this we first produce predicted values for the regression using OLS. We then censor the predicted values to be between zero and one and use generalized least squares to estimate the coefficients. We weight the results by p*(1 – p), where p is the censored predicted level of enfranchisement from the OLS regression. This procedure places more weight on intermediate levels of political enfranchisement. It places almost no weight on observations that approach 0 and 100 percent, which compensates for the problems outlined earlier and should more accurately estimate the effect of the independent variables on enfranchisement (Hanushek and Jackson 1977, 194). We draw the following conclusions from these regressions. First, increasing labor force participation led to significantly higher levels of political enfranchisement in these nine European countries, even control-

378

Democracy, Inequality, and Representation

ling for increasing wealth and the effect of the two world wars. In both the OLS and GLS models, labor force participation remains a significant predictor of enfranchisement at above the 90 percent level in every version of the model. Second, increasing per capita wealth also led to increases in the franchise, even prior to World War I. This effect is consistently significant at better than the .05 level. Finally, and independent of these other effects, countries expanded the franchise after World War I. As is often the case, the small sample size and relatively large confidence intervals make it difficult to draw meaningful conclusions about the relative importance of these variables. But it is still interesting to illustrate how large an effect the point estimates suggest. Between the 1885 to 1889 period and the 1925 to 1929 period, the average level of labor force participation in the sample increased from 59.4 percent of the population to 64.8 percent. The point estimate in model 5 predicts that this change produced a 19.3 percent increase in the fraction of the population allowed to vote. Put another way, the estimates suggest that growing labor force participation caused slightly more than 35 percent of the total increase in the extent of the voter franchise during this period. Real per capita GDP increased substantially, from an average of $2,510 per person in 1885 to 1889 to $4,238 per person in 1925 to 1929.37 According to the model, this increase in wealth expanded the franchise by 14.3 percent, or more than one-quarter of the total change in the proportion of people who were allowed to vote. Even controlling for these two factors, the interwar dummy in model 5 picks up an almost 20 percent increase in enfranchisement between the period prior to World War I and the period directly afterwards. Presumably, this variable picks up the leveling effects of the wartime experience that are not captured by either growing wealth or growing labor force participation. But rapid changes in inequality are not limited to the late nineteenth and early twentieth centuries. Previous chapters have reviewed these trends for recent decades in detail. A stylized summary might be that while there are still significant cross-sectional differences in levels of inequality across countries, the 1980s and 1990s saw a sizable, and by all accounts mostly exogenous, increase in income inequality in a number of the advanced economies. This has been especially true of the Anglo countries, like the United States and the United Kingdom. But increases in inequality have not been limited to these historically liberal economies. Brandolini and Smeeding (chapter 2) point out that even historically equal countries like Sweden and Finland saw increases in earnings inequality in the 1980s and 1990s, although in some cases welfare policies moderated the impact or delayed adjustments in household income that resulted from these trends.

Inequality and Institutions

379

In this chapter, we have emphasized that exogenous changes in factor endowments, trade, and technology affect democratic political institutions. We argue that these exogenous changes can alter overall levels of income inequality by, among other things, raising or lowering the wage of labor relative to capital. If technological, social, or economic factors create growing income inequality, then countries will quickly adopt less representative, and less democratic, political institutions. Changes in factor endowments that produce greater equality, by contrast, foster more inclusive political institutions that bring policy closer to the position of the median citizen. During the twentieth century, we find, the two world wars and growing labor force participation helped expand the franchise in Europe. Consistent with our theoretical expectations about the effect of growing inequality during recent decades, Pontusson and Rueda (chapter 10) argue that growing inequality in the 1980s and 1990s led to increasing political polarization and caused parties of the right to become less in favor of redistribution. We also suspect, exactly as do Ticchi and Vindigni (2003), that increasing equality in many countries during the twentieth century was associated not only with enfranchisement but with the shift to PR; increasing inequality in recent years, by contrast, may encourage a change to majoritarian methods of election, including the wider use of single-member districts. In our view, the mechanism that does the work is technological, demographic, and trade-pattern change that alters the relative value of endowments. Speculating further, we predict strong democratizing effects in laborabundant countries, such as China and India, that open themselves to trade. Both the theory and the empirical evidence on labor force participation suggest that as expanded trade increases in labor-abundant countries, the returns to less-skilled labor will also increase. And as more people join the labor force, institutions in these countries will become more democratic. Conversely, the theory implies that highly representative democratic political institutions are a product of relatively equal societies. They may not be introduced, or may not succeed, when these conditions are absent.

We are grateful to Jeff Toolan and Grace Demos for very helpful research assistance, and we thank Jeff Williamson for kindly providing some of the data used in this project. The views expressed in this chapter are solely the personal opinions of the authors and do not reflect the views of any organization or employer.

380

Democracy, Inequality, and Representation

Notes 1.

2.

3.

4.

5. 6.

7.

8.

9.

10.

11.

Alberto Alesina, Edward Glaeser, and Bruce Sacerdote (2001) also noted that PR independently increases both redistribution and equality. Ronald Rogowski and Mark Kayser (2002) claimed, with at least a whiff of supporting evidence, that PR electoral systems benefit producers and disadvantage consumers. On a narrower but important front, a number of authors have shown, however counterintuitively, that higher income inequality leads to less demand for some types of redistributive spending (Moene and Wallerstein 2003; Moffitt, Ribar, and Wilhelm 1998). This finding is contrary to the traditional Meltzer-Richard model, in which demand for redistribution rises with the gap between the median voter and the average voter—that is, precisely with income inequality (1981). For more on the data series used for this analysis and long-term trends in the share of income going to top earners, see Anthony Atkinson and Thomas Piketty (2007) as well as Piketty and Emmanuel Saez (2006). These proxies are the percentage of land held by family farms, an index of educational attainment, and the average of the urban and non-agricultural population percentage. The Gini coefficient is a standard measure of income inequality. For unequal democracies, the probability of a transition to authoritarian rule was .0131. For more equal democracies, the probability of failure was only .0028. For dictatorships, the respective transition probabilities (this time to democratic rule) were .0542 for those that experienced a decrease in inequality and .0221 for those that experienced an increase. See Przeworski et al. (2000, tables 2.15.B1 and 2.15.C1). Labor’s wage is its marginal productivity, that is, w = (1– α) A (K/L)α. The rent of capital is capital’s marginal productivity, or r = α A (K/L)α– 1. The ratio of these is self-evidently as stated earlier. Recall Max Weber’s defining characteristic of the modern state: that it possesses a monopoly, within its territory, of the legitimate use of force. That military power is crucial as between states is the standard perspective of socalled realist theories of international relations. Average welfare is simply w = Σwi/n = 1/n Σwi; obviously, what maximizes w maximizes Σwi, and vice versa. In the simplified notation adopted here, n = 1, making the point even more obvious. For more on the role of political entrepreneurs, see Becker (1983). In the case of extremely unequal societies, resources might also be devoted to military suppression of popular demands. See Bingham Powell (2002) on the tendency of PR to produce more left-leaning policy than majoritarian systems and the tendency of majoritarian systems to form governments that are even more conservative than the median voter.

Inequality and Institutions 12. 13.

14.

15.

16. 17.

18. 19. 20.

21.

22.

23.

381

Often there was, as at Athens, an intervening period of “tyranny”—actually more akin to “boss rule.” For one thing, what we would regard as the legislature was chosen by lot, like a modern jury, not by election. This form of government still finds some advocates. William F. Buckley Jr. (1963), for example, once wrote that he would rather be governed by the first two thousand names drawn at random from the Boston telephone directory than by two thousand faculty members of Harvard University. At the same time, most authorities emphasize that the political changes further diminished inequality, particularly via provision of public goods like roads and clean water. The efforts at land reform were led by the Gracchi, two aristocratic but radical brothers who would probably have won the requisite majorities for their proposals but were successively assassinated. Fortunately, nothing of this kind ever happens in modern democracies. Martel was nominally mayor of the palace and also the grandfather of Charlemagne. Some authorities, most notable among them Douglass North, have argued cogently that Europe became vulnerable to plague only because of growing food shortages, a result of population growth having pushed the “Malthusian limits” of the then-available agricultural technology (North and Thomas 1973). That said, recent scholarship advances evidence that Douglass North’s version of “Malthusian limits“ was wrong: population continued to expand right up to 1348 (Clark 2007; Herlihy 1997). Many historians interpret the reign of Charles I, ultimately deposed by Parliament and executed, as a failed attempt to establish absolutism in England. Probably equally important were the preceding changes in tactics in the sixteenth century; see, for example, David Eltis (1995). Previously never more than 75,000, total French troop strength rose to 150,000 in 1635, and to over 400,000 by the end of the seventeenth century (Lynn 1994). Swedish troop strength tripled, from 15,000 in the 1590s to 45,000 in the 1630s, and rose again to 70,000 in the 1650s (Maddison 2001, 81). Angus Maddison (2001, 95) estimated that world carrying capacity of sailing ships nearly tripled between the 1670s and the 1780s, rising from 1.45 million tons to 3.95 million tons. These trends, of course, accelerated in the nineteenth century, particularly with the advent of railroads and steamships (O’Rourke and Williamson 1999). Real wages increased also by virtue of cheaper foodstuffs, food being a much larger share of the budgets of the poor. Grain was imported into Northern Europe from Eastern Europe, as well as sugar from the Caribbean, and, perhaps most important in this period, new and more productive crops

382

24.

25.

26. 27.

28.

29. 30.

31.

32.

33.

Democracy, Inequality, and Representation from the New World, above all the potato and maize, were introduced into European agriculture (Braudel 1973; Langer 1963). “The frontier individualism has from the beginning promoted democracy. The frontier states that came into the Union in the first quarter-century of its existence came in with democratic suffrage provisions, and had reactive effects of the highest importance on the older States whose peoples were being attracted there” (Turner 1893). In the United States, African American males were drafted into the armed forces in large numbers in both world wars but remained in segregated units under white officers. Thomas Piketty (2003) presented perhaps the best-documented case of this in his discussion of French income inequality. The nine countries included in this sample are Belgium, Denmark, France, Germany, Italy, the Netherlands, Norway, Sweden, and the United Kingdom. One of the present authors, however, has established a strong link between changes in inequality and changes in government ideology (MacRae 2004). In the advanced economies for which data are available over the last thirty years, rising inequality appears to have led consistently to more right wing (less redistributive) governments. In addition, changes in inequality levels tend to occur prior to changes in ideology. This leads to the conclusion that, at least over the short term, it is inequality that causes changes in government ideology and not the other way around. For a similar argument based on the organizational ability of social classes, see Evelyne Huber, Dietrich Rueschemeyer, and John Stephens (1993). Switzerland is the obvious exception. It is not, however, included in the sample because we lacked labor force participation data. Men in Switzerland achieved universal suffrage in 1848, women only in 1971. This may have had something to do with the fact that the country remained neutral in both world wars and so was not exposed to quite the same pressures as other European countries. Certainly it is the case that Switzerland did not experience the same decline in inequality as other European states in this period; see Fabien Dell, Piketty, and Saez (2007). Each of the points in the graphs in figure 11.2 represents a five-year average value. The enfranchisement data series is reported in Peter Flora (1983). The labor force participation variable is an unpublished dataset that was kindly provided by Jeffrey Williamson. From 1871 to 1919, only men age twenty-five and over could vote in Reichstag elections; in the Weimar Republic, women were enfranchised and the voting age was lowered to twenty. See, for example, Brandenburgische landeszentrale für politische bildung, http://www.politische-bildung-bran denburg.de/links/wahlen/landtagswahlen5.htm. See Pikkety and Saez (2006) for additional taxation-based data showing de-

Inequality and Institutions

34. 35.

36.

37.

383

clines in inequality in a number of developed countries during the periods between World War I and World War II, although they suggest that the largest impact of the wars was on capital owners. Data on GDP per capita are from Angus Maddison (2003). The interwar period variable is equal to one for the years between 1915 and 1939. The World War II variable is equal to one between 1940 and the last year in the sample, 1944. Actually, in the Flora (1983) measure, the enfranchised fraction of the population can exceed 100 percent because the divisor is the population age twenty and above. If, for example, all people eighteen and older were allowed to vote, this population would exceed 100 percent of those twenty and older. In practice, however, the value never exceeds 100 percent by a large amount, and in our sample it does not happen at all. The fraction enfranchised in the sample used in table 11.2 varies between 2.5 percent and 98.1 percent of the population. These are 1990 International Geary-Khamis dollars (see Maddison 2003).

References Acemoglu, Daron, and James A. Robinson. 2000. “Why Did the West Extend the Franchise? Democracy, Inequality, and Growth in Historical Perspective.” Quarterly Journal of Economics 115(4): 1167–99. ———. 2006. Economic Origins of Dictatorship and Democracy. New York: Cambridge University Press. Acemoglu, Daron, Simon Johnson, and James Robinson. 2002. “Reversal of Fortune: Geography and Institutions in the Making of the Modern World Income Distribution.” Quarterly Journal of Economics 117(4): 1231–94. Alesina, Alberto, Edward Glaeser, and Bruce Sacerdote. 2001. “Why Doesn’t the United States Have a European-Style Welfare State?” Brookings Papers on Economic Activity 2(Fall): 187–278. Atkinson, Anthony B., and Thomas Piketty, eds. 2007. Top Incomes over the Twentieth Century: A Contrast Between Continental European and English-Speaking Countries. Oxford: Oxford University Press. Becker, Gary S. 1983. “A Theory of Competition Among Pressure Groups for Political Influence.” Quarterly Journal of Economics 98(3): 371–400. Boix, Carles. 2003. Democracy and Redistribution. Cambridge: Cambridge University Press. Boix, Carles, and Luis Garicano. 2002. “Democracy, Inequality, and Country-Specific Wealth.” Unpublished paper. Yale University. Accessed at http://www .yale.edu/leitner/pdf/PEW-Boix.pdf. Braudel, Fernand. 1973. Capitalism and Material Life, 1400–1800. Translated by Miriam Kochan. New York: Harper Colophon Books.

384

Democracy, Inequality, and Representation

Buckley, William F., Jr. 1963. Rumbles Left and Right: A Book About Troublesome People and Their Ideas. New York: G. P. Putnam’s Sons. Clark, Gregory. 2007. A Farewell to Alms: A Brief Economic History of the World. Princeton, N.J.: Princeton University Press. De Tocqueville, Alexis. 1955. The Old Régime and the French Revolution. Translated by Stuart Gilbert. Garden City, N.Y.: Doubleday. ———. 1969. Democracy in America. Translated by George Lawrence. Edited by J. P. Mayer. New York: Harper & Row. Dell, Fabien, Thomas Piketty, and Emmanuel Saez. 2007. “Income and Wealth Concentration in Switzerland over the Twentieth Century.” In Top Incomes over the Twentieth Century: A Contrast Between Continental European and English-Speaking Countries, edited by Anthony B. Atkinson and Thomas Piketty. Oxford: Oxford University Press. Downing, Brian M. 1992. The Military Revolution and Political Change: Origins of Democracy and Autocracy in Early Modern Europe. Princeton, N.J.: Princeton University Press. Eisenstein, Elizabeth L. 1979. The Printing Press as an Agent of Change: Communications and Cultural Transformations in Early Modern Europe. New York: Cambridge University Press. Eltis, David. 1995. The Military Revolution in Sixteenth-Century Europe. New York: Tauris Academic Studies. Engerman, Stanley L., and Kenneth L. Sokoloff. 2001. “The Evolution of Suffrage Institutions in the New World.” NBER working paper W8512, October 2001. Cambridge, Mass.: National Bureau of Economic Research. Accessed at SSRN, http://ssrn.com/abstract=285650. ———. 2002. “Factor Endowments, Inequality, and Paths of Development Among New World Economies.” Working paper 9259. Cambridge, Mass.: National Bureau of Economic Research. Feldman, Gerald D. 1966. Army, Industry, and Labor in Germany, 1914–1918. Princeton, N.J.: Princeton University Press. Finer, Samuel E. 1975. “State- and Nation-Building in Europe: The Role of the Military.” In The Formation of National States in Western Europe, edited by Charles Tilly. Princeton, N.J.: Princeton University Press. Flora, Peter. 1983. State, Economy, and Society in Western Europe, 1815–1975: A Data Handbook in Two Volumes. Vol. 1. Chicago, Ill.: St. James Press. Ganshof, François Louis. 1952. Feudalism. Translated by Phillip Grierson. New York: Longmans, Green. Hanushek, Eric A., and John E. Jackson. 1977. Statistical Methods for Social Scientists. New York: Academic Press. Herlihy, David. 1997. The Black Death and the Transformation of the West. Edited and with an introduction by Samuel K. Cohn Jr. Cambridge, Mass.: Harvard University Press. Huber, John D., and G. Bingham Powell. 1994. “Congruence Between Citizens and Policymakers in Two Visions of Democracy.” World Politics 46(3): 291–326.

Inequality and Institutions

385

Huber, Evelyne, Dietrich Rueschemeyer, and John D. Stephens. 1993. “The Impact of Economic Development on Democracy.” Journal of Economic Perspectives 7(3): 71–86. Hunt, Lynn. 2007. Inventing Human Rights (A History). New York: W. W. Norton. Langer, William L. 1963. “Europe’s Initial Population Explosion.” American Historical Review 69(1): 1–17. Lipset, Seymour Martin. 1959. “Some Social Requisites of Democracy: Economic Development and Political Legitimacy.” American Political Science Review 53(1): 69–105. Lynn, John. 1994. “Recalculating French Army Strength During the Grand Siècle, 1610–1715.” French Historical Studies 18(4): 881–906. MacRae, Duncan C. 2004. “Earnings Inequality, Conservative Government, and the Demand for Redistribution in Developed Countries.” Ph.D. dissertation, University of California at Los Angeles, Department of Political Science. Maddison, Angus. 2001. The World Economy: A Millennial Perspective, Development Center Studies. Paris: OECD. ———. 2003. The World Economy: Historical Statistics. Computer file. Paris: OECD. Meltzer, Allan H., and Scott F. Richard. 1981. “A Rational Theory of the Size of Government.” Journal of Political Economy 89(5): 914–27. Moene, Karl Ove, and Michael Wallerstein. 2003. “Earnings Inequality and Welfare Spending: A Disaggregated Analysis.” World Politics 55(4): 485–516. Moffitt, Robert, David Ribar, and Mark Wilhelm. 1998. “The Decline of Welfare Benefits in the U.S.: The Role of Wage Inequality.” Journal of Public Economics 68(3): 421–52. Nef, John Ulric. 1940. Industry and Government in France and England, 1540–1640. Ithaca, N.Y.: Cornell University Press. North, Douglass C., and Robert P. Thomas. 1973. The Rise of the Western World: A New Economic History. Cambridge: Cambridge University Press. O’Rourke, Kevin, and Jeffrey Williamson. 1999. Globalization and History: The Evolution of a Nineteenth-Century Atlantic Economy. Cambridge, Mass.: MIT Press. Persson, Torsten, and Guido Tabellini. 2000. Political Economics: Explaining Economic Policy. Cambridge, Mass.: MIT Press. Piketty, Thomas. 2003. “Income Inequality in France, 1901–1998.” Journal of Political Economy 111(5): 1004–42. Piketty, Thomas, and Emmanuel Saez. 2006. “The Evolution of Top Incomes: A Historical and International Perspective.” AEA Papers and Proceedings 96(2): 200–205. Pirenne, Henri. 1936. Economic and Social History of Medieval Europe. New York: Harcourt, Brace, and Co. Powell, G. Bingham. 2002. “PR, the Median Voter, and Economic Policy: An Exploration.” Paper presented to the annual meeting of the American Political Science Association. Boston, Mass., August 29–September 1, 2002.

386

Democracy, Inequality, and Representation

Przeworski, Adam, and Fernando Limongi. 1993. “Political Regimes and Economic Growth.” Journal of Economic Perspectives 7(3): 51–69. Przeworksi, Adam, Michael E. Alvarez, Jose A. Cheibub, and Fernando Limongi. 2000. Democracy and Development: Political Institutions and Well-being in the World, 1950–1990. Cambridge: Cambridge University Press. Rogowski, Ronald, and Mark A. Kayser. 2002. “Majoritarian Electoral Systems and Consumer Power: Price-Level Evidence from the OECD Countries.” American Journal of Political Science 46(3): 526–39. Scheve, Kenneth, and David Stasavage. 2007. “Political Institutions, Partisanship, and Inequality in the Long Run.” Accessed at http://pantheon.yale.edu/ ~ks298/index_files/inequality_ss_1007.pdf. Ticchi, Davide, and Andrea Vindigni. 2003. “Endogenous Constitutions.” Seminar paper 726. Stockholm: Stockholm University, Institute for International Economic Studies. Turner, Frederick Jackson. 1893. “The Significance of the Frontier in American History.” Paper presented to the meeting of the American Historical Association. Chicago, Ill. (July 12, 1893). Published in Report of the American Historical Association (1893): 199–227. Accessed at http://history.sandiego.edu/GEN/ text/civ/turner.html. White, Lynn T. Jr., 1962. Medieval Technology and Social Change. Oxford: Clarendon Press. Williamson, Jeffrey. n.d. Unpublished data on labor force participation.

Chapter 12

Inequality and Democratic Representation: The Road Traveled and the Path Ahead PABLO BERAMENDI AND CHRISTOPHER J. ANDERSON

The chapters in this book have examined the relationship between income inequality and processes of democratic representation in the advanced democracies of the West. They have traced the dimensions, evolution, and differences in income inequality across most if not all of the rich countries, including the United States and much of Europe. But aside from documenting trends and differences in inequality across countries, they have also aimed to understand the political causes and consequences of inequality—that is, the impact that inequality has on the workings of democracy, as well as how democratic politics helps shape levels of inequality in society. The volume is fundamentally driven by a concern with inequality because too much of it is thought to be hazardous for the maintenance of democratic political systems and because of normative concerns with equality as a bedrock of democracy, albeit one that is under siege. As a result, understanding what drives inequality and how it, in turn, shapes what voters, parties, and governments do is essential for understanding what makes democratic systems and market economies function the way they do. To understand the contours, dynamics, and structure of inequality and its connection to the processes of democratic representation, our approach has been comparative in nature. That is, we have argued that we can better understand the interplay of democracy and inequality by systematically comparing the ways in which different countries organize political competition and policymaking processes to produce redistributive and macroeconomic outcomes. Aside from being comparative, our ap-

388

Democracy, Inequality, and Representation

proach has been both theoretical and empirical. Starting from standard models of representation and redistribution, we have taken apart the black box of voter demands, party government, electoral institutions, and welfare state politics. The chapters have painted a picture of inequality and democracy that shows that high levels of inequality are far from automatic and, conversely, that more equal income distributions are anything but unattainable. Instead, greater income equality requires the fortuitous interplay of different combinations of voter demands and mobilization, the right kinds of electoral rules, and friendly parties in government; inequality, in turn, shapes the existence, power, and behavior of these players and institutions. In thinking about and examining income inequality’s connection to democratic representation, we thus followed two major avenues of inquiry. The first revolved around the role that political processes and institutions play in shaping aggregate distributions of income in society. The second avenue was focused on the impact of inequality on politics—that is, the ways in which inequality shapes the preferences and behavior of voters, parties, and governments. This chapter reports the main insights we offer on the different nuts and bolts of the process that generates and sustains inequality and links them to some of the most pressing, but yet to be answered, questions about the political economy of inequality in advanced industrialized democracies.

The Road Traveled This volume was motivated by our identification of three issues as most in need of scholarly attention to fully grasp the forces that shape inequality and inequality’s consequences for democratic politics: the multidimensionality of income inequality, the need to move beyond the standard median voter model of redistribution and democratic politics and address the interplay between partisanship and representation as an alternative way to understand the observable patterns of variation in inequality across OECD countries, and the need to recognize and tackle the joint endogeneity between the decision to get politically involved, the selection of political institutions, and inequality. The chapters in this volume significantly advance our understanding of the politics of inequality with respect to each of these three issues. Andrea Brandolini and Timothy Smeeding’s extensive review of existing cross-national data on income equality shows not only that there has not been a common trend toward higher levels of income inequality across all of the industrialized countries, but also that the observed trends vary across different types of income. If we take each type of income to be a dimension of inequality that jointly represents overall levels of inequality

Inequality and Democratic Representation

389

in society—a point also highlighted by Pablo Beramendi and Thomas Cusack in their chapter—we start to get a better conceptual and empirical handle on the multidimensional nature of inequality (see also Kenworthy and Pontusson 2005; Pontusson 2005). Obviously, distinguishing among different forms of income and their political origins does not exhaust the sources of multidimensionality in income inequality, but it constitutes a major step forward in recognizing that inequality manifests itself in different ways. Brandolini and Smeeding also offer estimates of the effects of government policies and social spending efforts on inequality, revealing that much of the cross-country variation in different kinds of income is driven by redistribution. By tracing the evolution of welfare generosity in OECD countries over the last three decades, Lyle Scruggs’s chapter offers a detailed analysis of the dynamics of public policy interventions. Based on a novel measure of generosity, Scruggs compares the effects of program generosity with conventional spending measures and argues that the benefit generosity index is a better empirical predictor of income redistribution in large part because, conceptually, it is a more valid approach to defining the welfare state. Empirically, Scruggs establishes that those countries traditionally considered the most generous are the ones facing higher levels of retrenchment, which are thus triggering changes in income inequality. Conceptually, Scruggs moves from an approach based on the size of the welfare budget to an approach that sees social security regimes as commitments to insure risks (Esping-Andersen 1990; Korpi 1983). In his own words, “this approach assumes that the conditions stipulated in national social insurance programs—which are the most important sources of non-life insurance for the bulk of the population in all industrial countries—better encompass the extent of welfare state generosity.” Like income inequality, these commitments are multidimensional; while some of them operate only within the labor market, others involve the entire society. Both between and within countries (across fiscal programs), these commitments vary in terms of scope, dependency on previous earnings, duration, eligibility, and tax implications, reflecting a myriad of choices over multiple issues. The next hurdle we tackle in the volume is the systematic identification of the political and institutional foundations that underlie these choices. In undertaking this task, several contributors to the project set out to overcome the limitations of existing models—most notably the one-dimensional median voter framework pioneered by Allan Meltzer and Scott Richard—by addressing the interplay between partisan preferences and institutional configurations as a way to understand the choices underpinning the variation in commitments identified by Scruggs. Several important contributions to the field emerge from these efforts.

390

Democracy, Inequality, and Representation

First, the analyses reveal that the electoral system shapes the composition of governing coalitions, and thereby the levels and nature of redistribution. As Torben Iversen and David Soskice argue in chapter 4, centerleft governments dominate under proportional representation systems (PR), while center-right governments dominate under majoritarian systems. As a result, countries with PR electoral systems redistribute more than those with majoritarian electoral systems. By tracing the origins of redistribution to partisanship first, and then from partisanship to the electoral system, Iversen and Soskice illuminate the process by which redistributive coalitions form, and they do so in ways that would be simply impossible from conceptualizations of politics built around the median voter model. The second contribution to the connection between government partisanship and inequality has to do with the way in which political parties are able to pursue their goals. Specifically, partisan governments’ ability to affect income inequality varies across forms of income and, more importantly, depends systematically on the level of economic coordination in society. Parties define their policy in anticipation of what market actors will do. This explains the importance of economic policy coordination as a mechanism that reduces uncertainties about the nature and direction of these reactions across different realms of the economy. As a result, we should observe the effects of economic coordination especially in those areas where the behavioral reactions to fiscal policy are more likely to occur—in particular, the labor market. Consistent with this, Beramendi and Cusack find in chapter 5 that, while political parties directly affect the distribution of disposable income through their choices about fiscal redistribution, their capacity to shape the distribution of earnings is contingent on the degree of wage-bargaining coordination. In contrast, governments driven by partisan concerns appear to have no leverage to shape the nonwage share of market income. Third, to further illuminate the nature of the interplay between parties and institutions, on the one hand, and to assess as precisely as possible the relative influence of actors and institutions, on the other, it is important to distinguish the partisan origins of policy from the translation of government policies into actual policy outcomes—in our case, inequality. By effectively developing this approach in a study of the lower half of the wage distribution, David Rueda’s chapter makes a significant contribution by showing why and how institutions (corporatism) mediate the influence of political agency (parties) on the distribution of income. His findings on wage inequality concur with Beramendi and Cusack’s regarding the need to overcome the limitations of previous institution-free approaches, and he also concludes that centralized wage-bargaining institutions play a central role in shaping the level of inequality. In con-

Inequality and Democratic Representation

391

trast, however, Rueda offers a more optimistic take on the ability of political parties to pursue equality in the absence of strong economic institutions. Finally, governments’ responses to external shocks are mediated by existing institutional configurations. Institutions matter because they affect the patterns of interest aggregation—that is, they shape the extent to which the voice of workers exposed to different kinds of risks is heard via processes of democratic representation. In their analysis in chapter 7, Thomas Cusack, Torben Iversen, and Philipp Rehm pay particular attention to the role of two institutional factors in shaping how governments react to economic shocks: the presence (absence) of national training systems and the nature of the electoral system. Training systems shape the composition of skills in the labor force, which, in turn, affects the level of demand for social insurance. Electoral systems condition the partisan composition of the government and mediate its ability to make longterm policy commitments (as elaborated in detail by Iversen and Soskice in chapter 4). In sum, these chapters, individually and collectively, take our understanding of the origins of inequality well beyond the limits set by approaches exclusively based on the partisan or median voter’s preferences by recognizing the multidimensionality of inequality and the importance of partisanship. But the story does not end here. The second half of the volume engages the third issue motivating this project, namely, the endogeneity among political factors, political processes, and inequality. To start, Robert Franzese and Jude Hays highlight the complex nature and empirical salience of several of these relationships in chapter 8. Building on and extending the insurance model developed by Karl Ove Moene and Michael Wallerstein (2001), Franzese and Hays argue that the generosity and structure of social policy affect simultaneously the efficiency of the labor market and the political participation of the less fortunate. This alters the relative position of voters and the nature of political conflict in the immediate future. At the same time, there are good theoretical reasons to expect unemployment and participation to shape the nature of public policy interventions. Franzese and Hays’s results suggest that income inequality causes redistributive efforts to increase, that participation increases both social insurance and redistribution, and that redistribution at the same time increases citizens’ political input. By analytically exploring and taking on the challenge of empirically estimating multiple endogenous relationships, Franzese and Hays pave the way toward a better grasp of the causal relationships between citizens’ involvement, policy choices, and economic outcomes. With this very goal in mind, the remaining chapters focus on different aspects of the influence of inequality on processes of democratic repre-

392

Democracy, Inequality, and Representation

sentation. Building on an original dataset that measures risk objectively (for details, see Rehm 2005), Cusack, Iversen, and Rehm also show that citizens’ demand for redistribution is a function of actual or potential unemployment. First, actual job loss reduces income and moves individuals closer to the bottom end of the income distribution—that is, the group of those with a self-interest in redistribution is expanded. Second, job losses may raise the demand for redistribution among workers currently employed but worried about their job because redistributive spending serves as an insurance against the risk of future income loss. As a result, exposure to risk—that is, the likelihood of future income loss—combined with relative income is a very strong predictor of redistributive preferences. While inequality conditions citizens’ preferences for redistribution, it also affects the likelihood that they will become involved in politics. By focusing in detail on the relationship between income inequality and political participation, Christopher Anderson and Pablo Beramendi return to one of the endogenous relationships identified by Franzese and Hays in their more general chapter. Specifically, Anderson and Beramendi examine the conditions under which income inequality affects citizens’ decision to participate in political action. The results are intriguing: income inequality at the macro level depresses electoral participation. And these effects of income differentials are manifest at the level of individual citizens as well: individuals below the median income in society are less likely to participate in elections, while those above the median income are more likely to do so, and at roughly equal rates. These findings suggest that the decision of low- and middle-income individuals to participate is driven to a large extent by resources. As income goes up, participation becomes affordable—and provided that there are incentives, wealthier individuals are therefore more likely to participate. Paradoxically, when income is extremely high, there also emerge disincentives to participate because there are too many ways for individuals in the highest income categories to avoid the consequences of policies to care sufficiently about being involved in the electoral process. An important implication of these results is that, consistent with the findings reported by Franzese and Hays, redistribution plays a prominent role in mediating the relationship between the skew of the distribution of income and the involvement of citizens, particularly those with relatively lower income. The effect of inequality on the democratic process goes beyond individual decisions to participate. It also shapes how political parties define their positions on distributive issues. In particular, inequality can lead to political polarization. And it does so in ways that vary across forms of income, as reflected in the results reported by Jonas Pontusson and David Rueda. Whereas wage inequality tends to be associated with polarization

Inequality and Democratic Representation

393

skewed toward the left, inequality in household disposable incomes tends to be associated with right-skewed polarization. Pontusson and Rueda also find that the impact of inequality on polarization is conditional on the level of mobilization. The association between wage inequality and left wing polarization holds only when the mobilization of low-income groups is considerable, while the link between right wing polarization and disposable income inequality rests on low levels of mobilization among lower-income voters. To explain these regularities, Pontusson and Rueda offer a theory of differential preferences. The core constituencies of left parties care primarily about wage inequality and do not necessarily become more supportive of redistribution as household income inequality rises. In contrast, the core constituencies of right parties care primarily about household income inequality and are threatened by higher levels of redistribution brought on by rising mobilization of low-income groups. By exploring the way in which the skew of the income distribution shapes party politics, this chapter illuminates a largely overlooked mechanism through which inequality conditions the democratic process. Finally, the shadow of endogeneity also extends to the very institutions whose direct and mediating effects have been analyzed in previous chapters. As Ronald Rogowski and Duncan MacRae put it, “that institutions covary with political and economic inequality seems obvious.” The challenge is to tap the process by way of which changes in inequality lead to institutional change and innovation. To tackle this question, Rogowski and MacRae argue that changes in economic and military technology, trade, and factor endowments influence the evolution of political institutions through their effect on inequality. That is, both inequality and institutional choices respond to previous structural transformations. In developing this argument, Rogowski and MacRae make a number of important points. First, the chapter puts forward a general logic, according to which social and structural changes alter the relative value of assets, producing changes in inequality, which, in turn, lead to institutional changes that may bring about further changes in inequality. Second, Rogowski and MacRae illustrate the plausibility of this logic with a rich set of historical case studies (from ancient Greece to the interwar period) and quantitative evidence that increasing labor force participation and the demand for labor created by the two world wars encouraged European countries to expand the right to vote during the late nineteenth and early twentieth centuries. This evidence leads Rogowski and MacRae to conclude that political entrepreneurs will push for the adoption of less representative political institutions, if not the sheer elimination of democracy, when exogenous changes increase income inequality. In contrast, social transformations that bring about equality create incen-

394

Democracy, Inequality, and Representation

tives to increase political participation through institutional change. Finally, by providing a broad historical perspective on the connection between the nature of representative institutions and inequality, this chapter points to many open questions that deserve systematic attention in future research. In what follows, we address some of these issues in the context of a broader discussion of the challenges ahead in inequality research.

The Road Ahead We intended the collective contribution embodied in this volume to be an important first step toward overcoming the challenges posed by the multidimensionality of income inequality and the need to disentangle the endogenous nature of democratic political processes and institutions, on the one hand, and inequality, on the other. In addressing these challenges, the chapters reveal a good deal about what we know, but they also put in sharper relief what it is we need to know more about if we are to understand the full story of inequality and democratic representation. In this section, we therefore address some of the limitations of our analyses and suggest several ways to advance the agenda of research into the causes and consequences of inequality. Chief among the limits is the fact that, even though the process of generating and sustaining inequality is essentially dynamic, the bulk of the analyses developed in this volume focus more on the cross-national (static) variation of policies, institutions, and outcomes and less on the dynamics of their interplay over time. Although some of the contributions directly address the question of changes over time in different dimensions of inequality and redistribution (most notably the chapters by Brandolini and Smeeding and by Scruggs), a fully dynamic story is often missing from causal analyses of the relationships examined, both in this volume and elsewhere. This predominantly static approach comes at a price: it imposes, often implicitly, heroic assumptions about the incentives and strategies of citizens and other political actors, about our understanding of the distributional effects of policy regimes, and about the time horizons involved in the relationships among market and political actors. Thus, in our view, the next hurdle in inequality research is the articulation and empirical evaluation of a genuinely dynamic political and economic theory of inequality. Here we outline what we see as the key challenges in undertaking this effort and suggest ways to integrate new research in order to advance the frontiers of inequality research beyond the boundaries of this volume. Chief among the challenges ahead of us is the need to take time more seriously, both by expanding the existing his-

Inequality and Democratic Representation

395

torical and geographical breadth at the aggregate level and by pursuing a systematic comparative political economy of the life course at the individual level. These goals, in turn, bring about two additional issues: the need to expand the politics of inequality by integrating multiple dimensions that directly condition people’s distributional trade-offs, and, more generally, the task of bridging the micro-macro link in the context of a theory of individual decisionmaking. On this latter question, the task of further exploring the connection between actual and perceived inequalities and its implications for political action strike us as particularly salient issues worthy of investigation. Finally, we emphasize the need for a better understanding of the structural limits to the political pursuit of equality and the evolution of those limits over time. We elaborate on each of these points in turn.

The Macro Level: Historical and Geographical Breadth This volume has highlighted the endogenous relationships between democratic politics and inequality in significant ways, but we have covered only a small length of a much longer path yet to be explored. One might wonder why there has been relatively little effort so far to unpack the bidirectional relationship between politics and the distribution of assets, opportunities, and income in society over the historical long run. In fairness, much of the explanation for this lies in the nature of the data generally available to the field, this volume included. The OECD wage data change only very slowly over time, and together with the Luxembourg Income Study (LIS) data, existing data sources offer at best five to six data points over the period 1970 to 2000 for most countries.1 While a new LIS project will eventually produce comparable measures of inequality for up to twenty more middle-income nations, including China and India, we remain to a large extent constrained by data that are limited in quantity and comparability.2 Existing data constraints reduce our historical and geographical breadth and limit our universe of reference to, at best, twenty advanced industrial societies for the last three decades. What is more, it is important to keep in mind that our observations for the post-1973 period do not constitute a random sample. Rather, they represent a specific subperiod of a longer process in which causality might work in very different ways. This is a point made forcefully in a recent paper by David Stasavage and Kenneth Scheve (2007), whose analysis of inequality data based on the share of the top 1 percent of the income distribution (as documented from available tax records in Atkinson and Piketty 2006) casts doubt on the al-

396

Democracy, Inequality, and Representation

leged egalitarian effects of certain economic and political institutions. Stasavage and Scheve point out that the downward trend in the top income share started before these institutions were adopted, thus opening new questions on the direction of causality and the actual role of institutions such as wage coordination agreements in promoting or securing equality. These findings also speak to some of the findings of this book, most notably the chapters by Beramendi and Cusack, by Rueda, and by Iversen and Soskice; though these findings do not necessarily invalidate the central theses of these authors, they imply a more general point, namely, that the actual causal origins of the associations established in these chapters lie in political choices about the design of institutions that took place much earlier in historical time. This brings us back to the endemic issues of endogeneity and to the need to extend our historical breadth to be able to fully establish the nature of the origins of, and causal links between, institutions, political choices, and distribution. The framework developed by Rogowski and MacRae offers a useful starting point. It casts a wide net that helps delimit the terrain for which more precise analyses are needed. The field is already responding to the task: a new set of research projects are conceptualizing the origins of representative institutions as the outcome of distributive conflicts and trade-offs among actors interested in keeping as tight a grip as possible on future decisions about the allocation of resources. As part of this stream of research, inequality is gaining prominence as an important cause behind the choice of institutions such as electoral rules, decentralized fiscal arrangements, and corporatism. In a recent paper, Davide Ticchi and Andrea Vindigni (2003) argued, for instance, that inequality shapes the choice of electoral rules. According to their model, high inequality drives the median voter’s preferences toward a majoritarian system, whereas low inequality leads to a preference for proportional representation systems. This argument leads to new questions about the way inequality shapes the relevant actors’ preferences and strategies: Is low inequality a sufficient condition for proportional representation to be adopted? Is the effect of inequality independent of or conditional on other factors such as the threat from the left (Boix 1999)? Or does it reflect preexisting patterns of coordination among asset-specific interests at the local and regional levels (Cusack, Iversen, and Soskice 2007; Iversen and Soskice 2007)? Similarly, does inequality foster fiscal decentralization under institutional conditions that prioritize particularistic interests (Beramendi 2007)? How does this relationship work in the presence of alternative institutional arrangements? Finally, is a certain level of equality a necessary condition for the adoption of corporatist institutions, as implied by Woojin Lee and John Roemer (2005), or is this relationship

Inequality and Democratic Representation

397

mediated by other institutional variables? Cathie Jo Martin and Duane Swank (forthcoming) suggest the latter when they model the level of economic coordination as a function of proportional representation and the presence or absence of federal institutions.3 Finally, how did inequality interact with employers’ incentives to centralize labor markets (Swenson 1991)? Subsequently, as the importance of inequality throughout the history of institutional choices and evolution becomes clearer, it is only natural to ask whether the changes in inequality can be attributed to exogenous changes. Where do these changes in inequality actually originate? Along with the study of institutional origins, the first link in the chain delineated by Rogowski and MacRae calls out for additional theoretical and empirical scrutiny.4 By increasing the historical breadth of their analyses, comparative political economists will be better able to distinguish the effects of institutions from the conditions under which institutions themselves were selected (Przeworski 2007), to identify the marginal effects of political agents and institutions over time, and therefore to establish empirically the direction of causality between inequality and political factors at different points in time. This volume has contributed a great deal to addressing these issues, but there remains a long way ahead. Along with the need to go back in time to disentangle causal relationships, the comparative study of redistribution and inequality is in much need of expanding across space as well. The amount, depth, and sophistication of the work on advanced industrial societies is not matched outside of that small handful of countries. In other words, we know very little about the political economy of insurance and redistribution in well over 90 percent of the countries of the world (for an overview, see Wibbels and Ahlquist 2008), and there are very few attempts to theorize redistribution systematically on a global basis (Mares 2005). Keeping these two worlds apart reveals two shortcomings. First, too many researchers of the OECD countries seem blissfully prepared to assume that their theories and findings derived from a handful of cases are truly general. Second, much of the mostly case-based research on social policy in the developing world has almost entirely eschewed engagement with cutting-edge theoretical and empirical developments in the literature on the OECD. Yet, in keeping the two worlds apart, the field builds on the implicit and a priori assumption that the politics of inequality is contingent on the level of development. But why would we expect different causal logics in developed and developing countries? Erik Wibbels and John Ahlquist (2008) did find, in fact, that closed economies in the developing world have larger insurance systems, which happens to be just the opposite of what we find in the advanced industrialized countries (Cameron 1978; Rodrik 1998), but the underlying cause for this distinc-

398

Democracy, Inequality, and Representation

tion is unclear. Addressing the reasons behind this difference, determining whether they indeed apply to other aspects of redistribution, and, more generally, ensuring that “general” theoretical claims about the role of parties, electoral systems, or federalism are not region-specific are therefore major hurdles that currently stand in the way of fully understanding the patterns of inequality and redistribution around the world.5

The Micro Level: Inequality, Redistribution, and the Life Course So far we have concentrated on issues of variation across space and time at the macro level of nation-states. Yet a country’s level of inequality at any given point simply reflects the combination of a large number of individual experiences. As a result, to fully develop a dynamic theory of inequality also requires a better theory of individual decisionmaking—in particular, a better theory about the choices that market actors make in the face of distributional trade-offs over the course of their lives. The notion of risk plays an important role here—it implies that the likelihood of a future loss of income weighs heavily in how people define their preferences with regard to insurance and redistribution. As the literature on labor market risks, policy preferences, and policy regimes (Atkinson 1995; Iversen and Soskice 2001; Mares 2003; Moene and Wallerstein 2001; Varian 1980) shows, scholars continue to devote significant attention to the origins of these risks and their role in shaping preferences and public policies (Cusack, Iversen, and Rehm, this volume; Rehm 2005). In contrast to the more popular versions of the median voter model, much of the analytical leverage inherent in the notion of risk lies in the fact that it provides an avenue to introduce intertemporal considerations into the political economy of redistribution and inequality. That is to say, risks matter because they reflect current expectations about the likelihood of incurring an income loss at some point in the future. Theories abound as to what determines the odds of such an event, as do the possible measurements of risks. But the more general intuition remains, namely, that individuals’ concerns about future income losses shape their current public policy choices. Efforts to unpack the origins and implications of different types of risks, despite their intuitive and undeniable value, do not exhaust intertemporal considerations over the life course by market and political actors. Defined as the probability of a future income loss, risks constitute only one aspect, albeit a very important one, of a more general phenomenon, namely, that of income mobility. In turn, risk aversion captures how concerned actors are about such probabilities. Yet there are many

Inequality and Democratic Representation

399

other dimensions of mobility whose implications for the political economy of inequality and redistribution remain to be fully explored. Consider first the relationship between inequality and mobility, at both the individual and aggregate level. In static analyses, where individuals are myopically concerned about their current position within the income distribution, preferences for redistribution become a direct function of individual income. However, we know that individuals vary in the extent to which they are stuck in a certain position in the income scale. This is likely to condition the way they approach institutional and policy choices that could influence their lives beyond the short term. The existing literature on risks and preferences (for an overview, see Rehm 2007; Cusack, Iversen, and Rehm, this volume) focuses on the implications of downward mobility and its relationship to public insurance systems. But we know very little about the reverse of the coin, namely, the consequences of upward mobility. Indeed, the few existing studies analyzing the political and institutional consequences of income mobility (Acemoglu and Robinson 2006; Benabou and Ok 2001; Leventoglu 2005; Piketty 1995) suggest that there is vast theoretical and empirical territory to be explored here. For instance, Bahar Leventoglu (2005) first, and Daron Acemoglu and James Robinson (2006) thereafter, show that the expected patterns of social mobility affect the very choice of political regime. Indeed, if the choice of political regime is based on the expectation of future welfare under different tax structures, people’s expectations about whether they will be stuck in low income categories or move up the ladder to become richer should carry significant weight in their calculations. Of equal or perhaps even larger importance is the question of people’s views about the relationship between the nature of the political regime and levels of mobility. In sum, if the choice of political regimes is theorized as a redistributive conflict with important temporal elements, the anticipated levels of aggregate social fluidity and individuals’ fortunes within such flows are crucial factors that require theoretical and empirical attention. Similarly, more work is needed to shed light on the links between social mobility, different dimensions of inequality, and the nature of politico-economic regimes. Here again, a few recently published papers point to critical issues for comparative political economy. Roland Benabou and Efe Ok (2001) offered a theoretical analysis of how the “prospect of upward mobility” shapes the demand for redistribution, offering a formalization of the oft-cited idea that lower-income individuals will not support redistribution if they expect to become high-income earners in the medium run. Benabou and Ok showed “that there exist a range of incomes below the mean where agents oppose lasting redistri-

400

Democracy, Inequality, and Representation

bution if (and in a sense, only if) tomorrow’s expected income is an increasing and concave function of today’s income” (449). A number of important implications follow from this analysis. First, the connection between income mobility and preferences for insurance and redistribution deserves further scrutiny. Benabou and Ok offer preliminary evidence on the basis of the Panel Study of Income Dynamics (PSID), but their operationalization and analysis of public policy preferences are very preliminary. Moreover, since their analysis is circumscribed to one country, it offers very little leverage for comparative analysis. Is the relationship between income mobility and preferences for insurance and redistribution universal, or is it contingent upon specific institutional features of politico-economic regimes? Although taking on this question comparatively is a demanding task that requires the combination of panel income data and household surveys with political questions, it is no longer an impossible undertaking. Once the political salience of income mobility has been established, political economists ought to turn their attention to the determinants of income mobility—in particular, its socioeconomic, political, and institutional foundations. Analytically, and in slightly more technical terms, this requires that we endogenize individuals’ income mobility not only with respect to their demographic characteristics but also with respect to other dimensions of inequality (class, occupational characteristics, wealth) and, more importantly, with respect to those political and institutional features that condition the behavior of market actors. Some specific issues within this more general agenda are already the subject of recent research—for instance, the links between asset specificity, risk, mobility, and preferences for redistribution (Iversen and Soskice 2001). But this work is also essentially cross-sectional in nature and covers only a partial aspect of the dynamics between a market actor’s individual choices and the political and economic implications of those choices. Indeed, for reasons related mostly to data availability, the comparative political economy of the relationship between different politicoeconomic regimes and individuals’ labor market and investment choices over the life course (including human capital) remains an underdeveloped field. To be sure, while numerous sociological studies have examined specific aspects of the life course in a number of countries (DiPrete 2002; DiPrete and McManus 2000, 2001), very few have systematically analyzed how political and economic differences at the macro level shape life-course dynamics at the micro level. This process, in turn, feeds back into the cross-national patterns of long-term inequality. Thus, a number of questions remain: How do different economic institutions, labor market arrangements, levels of economic coordination, public policy choices, and regulatory frameworks condition individual choices—and thereby

Inequality and Democratic Representation

401

the levels and trends in income mobility of different nations? By which mechanism does this process feed back into the process shaping individuals’ policy preferences? And how does it alter the distribution of longterm income? Addressing these questions systematically will expand our understanding of the politics of inequality by shedding much-needed light on a number of related issues that have remained outside the core interests of comparative political economy so far. The first among these is the problem of incidence (Atkinson and Stiglitz 1980; Beramendi 2001). By incidence we refer to the fact that political and institutional factors shape the distribution of income not only directly, through public policy interventions, but also indirectly, through the behavioral responses of market actors (mainly via investments and labor supply decisions). Some behavioral responses are reactions to policy and legislative changes, whereas others occur in anticipation of forthcoming changes in the political and institutional environment. For example, capital owners may vote with their feet in response to unanticipated changes in the regulatory framework of their area of activity (such as the introduction of more restricted environmental protection legislation by a new government). In turn, some employees might decide to stop working if, as a result of a change in the tax code, they anticipate being better off under new welfare arrangements. These connections have been the object of inquiry in disciplines such as economic sociology and applied labor economics, but they have not been incorporated into the comparative politics of distribution. For the most part, the field (this volume included) still restricts its efforts to a first-order incidence framework—that is, to the study of the direct effects of policy choices and institutional changes. As a result, our understanding of the politics of inequality remains biased toward those effects that are directly observable through aggregate data and cross-sectional surveys, and blind toward indirect effects that can only be captured through a combination of macrolevel data and longitudinal data sets. Moreover, the identification of direct effects using aggregate data poses significant challenges in the presence of indirect effects. This is often referred to as the “counterfactual problem.” One of the most extensively used indicators of redistribution—the proportional reduction in the value of the Gini coefficient for market income after taxes and transfers are brought into the picture—captures all types of effects, direct and indirect. Substantively, the presence of several orders of incidence begs the question of whether the egalitarian impact of the welfare state is undermined by its indirect effects, an issue on which there is no definite answer yet. As the quality and quantity of data available to scholars improve, the need and opportunity to identify and disentangle these two

402

Democracy, Inequality, and Representation

sets of effects becomes more pressing as a way to improve our grasp of the political origins of different distributions of income, and their continuation over time.

Different Kinds of Inequality Focusing on the dynamics of individual decisionmaking over the life course also opens up the possibility of considering the political foundations of two other dimensions of inequality—namely, inequalities of wealth and opportunities—and their interplay with the distribution of income and its evolution over time. Wealth matters for people’s ability to undertake long-term endeavors. It determines their ability to cope with business start-up costs, to enter imperfect credit markets, to assume risks, to invest in human capital while waiting for deferred returns, or to sail through the downward slides of the business cycle (Keister 2000; Morillas 2007; Shapiro and Wolff 2001; Sherraden 1991). By determining economic opportunities through these and other mechanisms, wealth inequality produces an uneven playing field for economic actors, thereby preconditioning many decisions that will lead to varying distributions of income. As with many of the less explored areas of inequality, the importance of wealth runs parallel to the difficulty of measuring it, and it is only now that the first systematic efforts to offer cross-national comparisons of wealth inequality are under way. Figure 12.1, which presents a comparative overview of the levels of wealth and income inequality in a number of advanced industrial societies, provides a number of interesting insights, some of them more in line with conventional wisdom than others.6 Consistent with received wisdom, wealth inequality is extremely high in the United States (Keister 2000). A Gini coefficient of 0.75 implies that a government willing to achieve a perfectly egalitarian society would have to redistribute three-quarters of society’s available wealth. The level of wealth inequality found in Finland also seems consistent with the patterns of income inequality identified by the contributors to this volume. Yet the extremely inegalitarian distributions of wealth in the other Scandinavian countries (Norway, Sweden, and Denmark) and the relatively more egalitarian profile of the United Kingdom and Italy appear to go against the established understanding of the distributive profiles in these societies. The gap between the distributions of income and wealth in Scandinavia is particularly interesting, as it seems to confirm recent accounts that redistribution in the most egalitarian societies is mostly a matter of conflict within labor rather than between capital and labor (Beramendi and Rueda 2007; Cusack and Beramendi 2006; Lindert 2004). The great variation in wealth inequality among the Scandinavian

Inequality and Democratic Representation

Wealth

Net Worth

It 20 aly 04 Ca 19 nad 99 a U n St ite 20 ate d 01 s

Diposable

It 20 aly 02 K Uni in te 20 gdo d 00 m

0.8 0.7 0.6 0.5 0.4 0.3 0.2 0.1 0

Gini Coefficients for Different Concepts of Income and Wealth

Fi n 19 lan 94 d Sw 20 ede 02 n Fi n 19 lan 98 d G er 20 ma 02 ny N o 20 rwa 02 y

Gini

Figure 12.1

403

Source: Luxembourg Income Survey (LIS).

countries, and between these countries and the rest of the world, constitutes a puzzle. Why do some countries exhibit higher levels of wealth inequality than others? What are the political and institutional foundations of these differences? Why do some countries (Norway and Sweden) exhibit relatively lower levels of income inequality and very high levels of wealth inequality, whereas others (the United States) are consistently inegalitarian along different dimensions of inequality? Wealth plays a central role in determining people’s life chances and their ability to fare well during adverse tides, and it conditions the impact that income may have at any given point in time. Thus, a deeper comparative knowledge of the political foundations of wealth inequality ought to rank very high among future priorities in the field. The third area where a life-course perspective and a micro-macro approach could be fruitfully combined to advance existing understandings of inequality is in understanding the link between politics, institutions, and equality of opportunity (Dworkin 2000; Roemer 1998). As we noted earlier, wealth, inequality, and opportunities are intimately related to one another, but wealth does not fully determine opportunities (and vice versa). There are many ways in which the constraints on equality of opportunity imposed by unequal distributions of wealth can be overcome through politics. In a recent study, John Roemer and his colleagues (2003) asked to what extent different fiscal systems help equalize the ability of different social groups to achieve similar levels of income. Their approach builds on the idea that differences in outcomes can be broken into two compo-

404

Democracy, Inequality, and Representation

nents, namely, differences due to effort and differences beyond people’s control (see Roemer 1998). Normatively, they argue, only the latter are to be equalized. And so they identify a set of circumstances that are clearly prior or exogenous to individual effort and attribute the rest of the variation in income to differences in effort. With all the necessary cautions, their results suggest that “given the restriction of policies to ones of affine income taxation in which all citizens face the same tax policy, the northern European countries do very well in regard to equalizing opportunities for income acquisition” (Roemer et al. 2003, 563). This important finding begs the question of the politics behind equality of opportunity (or lack thereof). What makes northern countries equalize opportunities better? What leads other countries not to pursue such a route—or perhaps to pursue that route and fail? What combination of policies and institutions works better to equalize life chances, not just current incomes? Does the equalization of income through the fiscal system bring about equalization of opportunities? The research of Roemer and his colleagues and a good deal of comparative sociological work relating low levels of income inequality to improved levels of equalization of educational and occupational opportunities (Eriksson and Goldthorpe 1992) seem to suggest so. However, we are far from being able to claim that we have managed to establish the political and institutional mechanisms behind these associations in any systematic way. Taken together, the study of the politics behind higher orders of incidence, the analysis of the links between political regimes, wealth dispersion, and individual opportunities, and the focus on the political and institutional processes through which multiple dimensions of inequality feed back on each other require that we bridge a number of factors that operate at several levels of analysis, ranging from individual life-course choices to national policy and institutional choices. While building these bridges poses formidable analytical and methodological challenges, a better understanding of the micro-macro links in the politics of income mobility will provide more solid foundations to revisit the old question of whether inequality is offset by mobility at the aggregate level. Indeed, there is another elephant in the room, but we are not the first to appreciate its contours. This honor belongs to an economist. Friedman’s razor would immediately regard most of current comparative political economy (this volume included) as a biased and ultimately futile ideological endeavor to demonstrate the shortcomings of the American political economy. He would cogently argue that relating political and institutional factors to measures of inequality capturing the distribution of income at a particular point in time is the wrong way of ranking and evaluating societies.

Inequality and Democratic Representation

405

Although such data cast a dark blanket over the image of the American dream, they neglect the fact that people may have different fortunes over time. That America is more unequal than Europe, he would argue, is far from worrying; instead, it actually is a good sign in the sense that inequality is a necessary correlate of more open, fluid, and dynamic economies. In such countries, the sky is the limit for people with the right work ethic and intuitions. European societies are more equal, but only because the arteries of their economic system are clogged with taxes and unfair transfers to undeserving people. Short-term equality would therefore be a bad indicator, one that captures the negative, short-term externalities of a dynamic economy and hides the long-term distributional consequences of any intervention in the economy. If inequalities in the short run can be offset by mobility in the medium run, why worry so much about them? From very different ideological trenches, Sara Jarvis and Stephen Jenkins make a similar claim: “To some people, greater inequality at a point in time is more tolerable if accompanied by significant mobility” (1998, 428). Ideological tastes aside, there is a point to this line of reasoning that has been neglected for far too long and that connects directly with the need to better understand the actual incidence of redistributive policies and the structure and evolution of the distribution of opportunities in society. Although preliminary comparative analyses suggest that income mobility is positively associated with income equality (Beramendi and Morillas 2008; Gangl 2005; Gangl, Palme, and Kenworthy 2008), additional efforts are warranted, not only because an empirical evaluation of the neoliberal objection is interesting in its own right, but more importantly, because it introduces a new way of evaluating political economies.

A Missing Link: Individual Winners and Losers and Subjective Inequality If inequality research is to move forward at the aggregate and individual levels, we need a map of the political process connecting both ends. Such a process, as depicted in figure 1.1, starts with the very definition of preferences at the individual level and ends with a diverse array of outcomes, largely mediated by the way different institutions aggregate preferences. Most of the literature is trapped in a sort of oversimplification of this process that keeps important aspects of the politics of inequality hidden from plain view. One of these aspects is the characterization, in objective terms, of how the distribution of winners and losers of different political economy regimes evolves through time, and why.

406

Democracy, Inequality, and Representation

Countries vary in how they respond to similar structural transformations, thereby generating different distributions of winners and losers. Although we have a general sense of who some of these winners and losers are (Esping-Andersen 1993, 1999; Iversen and Cusack 2000; Iversen and Wren 1998), existing characterizations of the fate of different social groups in a context of changing inequalities are both too general and too static. Yet, if policy compromises change in combination with exogenous transformations of production, certainly who wins and who loses under different regimes is not static, as partially reflected by the literature on the insider-outsider divide (Esping-Andersen 1999; King and Rueda, forthcoming; Rueda 2005, 2007). Alas, the sociology of stratification remains apolitical while comparative political economy remains largely asociological. To the best of our knowledge, there are no systematic efforts to map out, on the basis of microdata, how the distribution of winners and losers in different nations has changed over time as a result of changes in fiscal and economy policies and, more importantly, how such changes have affected the democratic process. For instance, has the balance of welfare-state efforts really tilted toward outsiders in recent decades (Mares 2006), or do insiders retain the condition of primary beneficiaries of redistribution? If the former, do all countries behave alike? With what consequences? The actual evolution of winners and losers is one key element for understanding changes in the coalitions in support of different levels of redistribution, yet it is not the only one. As reflected by the contributions to this volume, the comparative political economy of inequality and redistribution is surprisingly devoid of psychology, on the one hand, and collective action, on the other. As such, the field remains ruthlessly structuralist in two ways. First, the issue of the perception of inequality and its political implications is hardly part of the research agenda. The dominant assumption is that people perceive both the overall levels of inequality in society and their relative position within it accurately. Second, for the most part, comparative students of inequality and redistribution take collective action as given, despite the voluminous evidence in political science and sociology that not everyone is equally likely to engage in politics. Obviously, there are good theoretical and empirical reasons to cast doubts on these assumptions. It is also important to note that they inhibit the development of a much-needed, and still lacking, individual-level model of voter decisionmaking. How accurately do people perceive inequality and redistribution? Do citizens perceive changes in inequality over time? Which dimension of inequality do they perceive? Moreover, how do citizens evaluate these perceptions? Who thinks inequality is bad for society, and with what consequences? Although there is evidence that people are averse to too

Inequality and Democratic Representation

407

much income inequality (Alesina, Di Tella, and MacCulloch 2004), it is not at all clear that they perceive income inequality accurately (Kenworthy and McCall 2008), nor do we know which dimension of income inequality they focus on when forming perceptions about the distribution of income in society. A recent and lone study by Lars Osberg and Timothy Smeeding (2006) examined the accuracy of citizens’ perceptions of income inequality across OECD countries, including the United States. They found that subjective estimates of inequality in pay diverge substantially from actual data. Moreover, they reported that Americans are less aware of inequality at the top of the income distribution than citizens in other OECD countries and less concerned about reducing differentials at the bottom of the distribution. Alongside the findings of Alberto Alesina and his colleagues (2004) that Europeans are more sensitive to income inequality than Americans, this suggests that income distributions are not necessarily perceived accurately and that people in different countries think about them differently. There is also good reason to believe that there is significant withincountry variation in how individuals evaluate income inequalities—that is, whether they believe it to be good or bad. A recent paper by Christopher Anderson and Matthew Singer (forthcoming) tapped into this question, at least in part. These authors found that citizens in countries with higher levels of income inequality express more negative attitudes toward public institutions. Furthermore, they showed that the effect of inequality on attitudes toward the political system is conditional on the individual’s partisan position. As opposed to its effects on more conservative citizens, inequality leads left wing individuals to reach much more negative attitudes toward the system. Consistent with this, researchers also have found that individuals on the left favor equality to a greater extent than those on the right (Listhaug and Aalberg 1999) and that individuals on the political left are much more sensitive to inequality (Alesina, Di Tella, and MacCullough 2004). These findings have important implications. If politics mediates reactions to inequality, does it mediate perceptions and evaluations as well? Are perceptions and evaluations set in stone in the early stages of the process of political socialization, or are they subject to short-term political manipulation by political parties? How are they affected by the way people experience the actual implementation of public policies? If people misperceive their actual position in the income distribution, how does this affect the likelihood of them being politically engaged? What are the steps in between people’s actual relative position, their perceptions, their evaluations, and their political reactions? Given the importance of this chain, it is critical to expand the analysis of the intervening mechanisms beyond the standard association between structural factors,

408

Democracy, Inequality, and Representation

social class, and support for redistribution (for an overview, see Kumlin 2007). As we move up in the process of democratic representation, the connection between individuals’ perceptions and evaluations and their organized political action needs to be made (Bartels 2005). We suspect that inequality’s pernicious effects are likely to be particularly pronounced and detrimental to the quality of democratic life in the context of developing countries. This is consistent with several cross-national quantitative studies that have found that income inequality reduces the longevity of democratic regimes and increases the odds of revolution (Hibbs 1973; Muller 1988). And even though income inequality is unlikely to lead to revolution in the affluent democracies anytime soon, research on citizen participation in these countries has found that it begets inequalities in civic participation, interpersonal trust, and the quality of democratic representation (Anderson and Beramendi, this volume; Brady 2004; Campbell 2005; Franzese and Hays, this volume; Gilens 2005; Goodin and Dryzek 1980; Rothstein and Uslaner 2005; Uslaner and Brown 2005; Mettler and Soss 2004). On the whole, however, the relationship between income inequality and collective action is either unexplored (as in the case of social protest) or inconclusive (as reflected by the literature on peasant rebellions) (see Brockett 1992; Muller and Seligson 1987; Wang et al. 1993). Finally, while there are some suggestive intuitions, the field lacks a systematic theory of the connection between inequality and collective action.7 Overcoming this gap strikes us as critical if we are to gain a more dynamic understanding of how the combination of structural and policy changes shapes the political mobilization of contending interest and if we are to fully establish the connection between individual citizens, political processes, and aggregate outcomes. Although it seems natural to expect a connection between inequality and different forms of collective action, this link remains, in many ways, a mystery.

Limits to Equality An explicit vindication of political factors as a key—if not the key—force underlying the origins and evolution of different forms of inequality (as well as of different redistributive equilibria) runs through all the chapters in this volume. This brings us to the issue of whether there are limits to equality: What is the scope of influence of the democratic politics of distribution? Or put more simply, is perfect equality achievable politically? The answer is most likely no, which, in turn, raises the question of why. Classic arguments regarding the economic effects of redistribution pro-

Inequality and Democratic Representation

409

vide one possible reason: too much redistribution limits investment and growth (for critical overviews, see Esping-Andersen 1994; Pontusson 2005). Therefore, if governments want to have a pie to share, the distribution of the pieces cannot be too egalitarian. In fact, trying to impose too much equality via politics may even jeopardize democracy itself (Acemoglu and Robinson 2006; Boix 2003). Moreover, heterogeneity among social groups and individuals and the increasing diversification of economic factors are bound to yield distributions of preferences that make perfect equality a political chimera. Unfortunately, such arguments do not provide much guidance as to how much redistribution is too much redistribution. Nor do they explain why this hypothetical threshold may vary across countries. This issue is central to the literature on inequality and economic growth—a literature that is rich in theoretical contentions and very inconclusive in terms of empirical results (Banerjee and Duflo 2003; Benabou 1996). Standard median voter accounts imply that inequality is harmful for growth: as the median voter gets poorer, she demands more redistribution; redistribution, in turn, generates economic inefficiencies. On the basis of a multiple generation general equilibrium model, Torsten Persson and Guido Tabellini came to similar conclusions: “Income inequality is harmful for growth because it leads to policies that do not protect property rights and do not allow full private appropriation of returns from investment” (1994, 617). According to this logic, then, any amount of redistribution is bound to be too much redistribution. This line of reasoning is subject to two major caveats. First, it assumes that changes in income are a linear function of the income skew. Second, it builds on the notions, very dear to the median voter model of redistribution, that the demand for redistribution leads to increases in pretax income inequality and that the purpose of fiscal systems is exclusively to satisfy the extractive hunger of the poor, thereby damming investment rates. On the first point, recent empirical estimations have cast serious doubts about the linearity of the relationship between changes in inequality and the growth rate (Banerjee and Duflo 2003; see also Perotti 1993; Voitchovsky 2005). On the second point, this volume brings to bear ample evidence that neither of the two tenets that anchor the median voter approach holds. First, countries with more egalitarian pretax distributions tend to redistribute more (Beramendi and Cusack, this volume; Iversen and Soskice, this volume). And, second, it is simply not the case that countries making larger public insurance efforts necessarily show less investment, more unemployment, and less growth (Atkinson 1999; Daveri and Tabellini 2000; Hall and Soskice 2001; Kenworthy 2008). Taken together, then, in light of the limitations of the dominant theo-

410

Democracy, Inequality, and Representation

retical approaches, there is much to gain from revisiting the relationship between inequality and economic growth from the more refined understanding of the political economy of inequality and redistribution developed in this volume. The evidence suggests that both excessive inequality and excessive equality harm growth. The next hurdle is then to theorize and empirically identify those thresholds and the political and institutional constellations that will ensure that nations fall on the right side of the curve. This is a complex task that will require that we travel quite a way down the path outlined in this chapter. But we foresee no better way to keep on answering the call for a “political and social economy” launched by Simon Kuznets over fifty years ago.

We thank Rafael Morillas, Karen Remmer, David Rueda, Tim Smeeding, Guillermo Trejo, and Erik Wibbels for their comments and assistance on previous versions of this chapter.

Notes 1.

2.

3.

4.

5.

Soon LIS will add another data point for the early 2000s, and it has already produced some new comparative asset data (the Luxembourg Wealth Study), albeit for only ten nations at this time. In making this statement, we are aware of the existence of more ambitious databases in terms of coverage across space and time (for example, the World Bank database compiled by Deininger and Squire, or the UNU-Wider data). We believe, however, that the concerns about their quality and comparability are large enough to cast serious doubts on any finding that relies exclusively on them. For additional discussion of these issues, see Anthony Atkinson and Andrea Brandolini (2001). On a more positive note, the new LIS project on middle-income countries will provide additional checks on their comparability with LIS. For an example of the advantages of long-term historical analysis of labor market institutions and their interplay with the welfare state, see Peter Swenson (2002) and Kathleen Thelen (2004). For instance, David Stasavage and Kenneth Scheve (2007) pointed to the role of war as a possible explanatory factor for the declining trend in inequality they observe in the first half of the twentieth century. Yet another example of the urgency to take on this task is provided by Anirudh Krishna (forthcoming). This body of work shows that, contrary to what is observed in the developed world (Anderson and Beramendi, chapter 9, this volume), poor people in developing countries participate as much (often even more) than their richer counterparts. The identification of the reasons

Inequality and Democratic Representation

6.

7.

411

behind these divergent patters requires a common theoretical framework that speaks to both the developed and the developing world. Data source: Luxembourg Income Survey (LIS, http://www.lisproject.org). All incomes are adjusted by the square root of the number of family members. Lis_dpi: disposable income inequality as defined by the Luxembourg Income Study. Nw1: net worth, that is to say, the sum of total financial (TFA1) and nonfinancial (TNF1) assets minus total debt. Finally, wealth is defined as the sum of Lis_dpi and Nw1. The countries included are Finland, Sweden, Norway, Denmark, Italy, Canada, the United Kingdom, and the United States. These calculations were made by Philipp Rehm on the basis of the microdata. Our thanks to him for sharing the data and figures with us. For instance, on the basis of a rational actor model, Mark Lichbach (1990) develops an argument as to why there should be no link between inequality and rebellions. In contrast, Roger Gould (2003) contends that even though existing levels of inequality do not generate social conflict, this is not the case when the poor or, more generally, the members of large social strata experience sudden and unexpected changes in inequalities.

References Acemoglu, Daron, and James Robinson. 2006. Economic Origins of Democracy and Dictatorship. Cambridge: Cambridge University Press. Alesina, Alberto, Rafael Di Tella, and Robert MacCulloch. 2004. “Inequality and Happiness: Are Europeans and Americans Different?” Journal of Public Economics 88(9–10): 2009–42. Anderson, Christopher J., and Matthew M. Singer. Forthcoming. “The Sensitive Left and the Impervious Right: Multilevel Models and the Politics of Inequality, Ideology, and Legitimacy in Europe.” Comparative Political Studies. Atkinson, Anthony B. 1995. Incomes and the Welfare State. Cambridge: Cambridge University Press. ———. 1999. The Economic Consequences of Rolling Back the Welfare State. Cambridge, Mass.: MIT Press and University of Munich Center for Economic Studies (CES). Atkinson, Anthony B., and Andrea Brandolini. 2001. “Promises and Pitfalls in the Use of Secondary Data Sets: Income Inequality in OECD Countries as a Case Study.” Journal of Economic Literature 39(3): 771–80. Atkinson, Anthony B., and Thomas Piketty. 2006. Top Incomes over the Twentieth Century: A Contrast Between European and English-Speaking Countries. Oxford: Oxford University Press. Atkinson, Anthony B., and Joseph Stiglitz. 1980. Lectures in Public Economics. New York: McGraw-Hill. Banerjee, Abhijit V., and Esther Duflo. 2003. “Inequality and Growth: What Can the Data Say?” Journal of Economic Growth 8(3): 267–99.

412

Democracy, Inequality, and Representation

Bartels, Larry. 2005. “Economic Inequality and Political Representation.” Unpublished paper. Princeton University. Benabou, Roland. 1996. “Inequality and Growth.” Working paper 96-22. New York: C. V. Starr Center for Applied Economics, New York University. Benabou, Roland, and Efe A. Ok. 2001. “Social Mobility and the Demand for Redistribution: The POUM Hypothesis.” Quarterly Journal of Economics 116(2): 447–87. Beramendi, Pablo. 2001. “The Politics of Income Inequality in the OECD: The Role of Second-Order Effects.” Working paper 284. Luxembourg: Luxembourg Income Study. ———. 2007. “Inequality and the Territorial Fragmentation of Solidarity.” International Organization 61(4): 783–820. Beramendi, Pablo, and Juan Rafael Morillas. 2008. “The Political Economy of Persistent Inequality.” Unpublished paper. Duke University and Durham University. Beramendi, Pablo, and David Rueda. 2007. “Social Democracy Constrained: Indirect Taxation in Industrialized Democracies.” British Journal of Political Science 37(4): 619–41. Boix, Carles. 1999. “Setting the Rules of the Game: The Choice of Electoral Systems in Advanced Democracies.” American Political Science Review 93(3): 609–24. ———. 2003. Democracy and Redistribution. Cambridge: Cambridge University Press. Brady, Henry E. 2004. “An Analytical Perspective on Participatory Inequality and Income Inequality.” In Social Inequality, edited by Kathryn M. Neckerman. New York: Russell Sage Foundation. Brockett, Charles. 1992. “Measuring Political Violence and Land Inequality in Central America.” American Political Science Review 86(1): 169–76. Cameron, David. 1978. “The Expansion of the Public Economy: A Comparative Analysis.” American Political Science Review 72(4): 1343–61. Campbell, Andrea. 2005. How Policies Make Citizens: Senior Political Activism and the American Welfare State. Princeton, N.J.: Princeton University Press. Cusack, Thomas R., and Pablo Beramendi. 2006. “Taxing Work.” European Journal of Political Research 45(1): 43–73. Cusack, Thomas R., Torben Iversen, and David Soskice. 2007. “Economic Interests and the Origins of Electoral Systems.” American Political Science Review 101(3): 373–93. Daveri, Francesco, and Guido Tabellini. 2000. “Unemployment and Taxes.” Economic Policy 30(1): 49–89. DiPrete, Thomas A. 2002. “Life-Course Risks, Mobility Regimes, and Mobility Consequences: A Comparison of Sweden, Germany, and the United States.” American Journal of Sociology 108(2): 267–309. DiPrete, Thomas, and Patricia A. McManus. 2000. “Family Change, Employment

Inequality and Democratic Representation

413

Transitions, and the Welfare State: A Comparison of Household Income Dynamics in the U.S. and Germany.” American Sociological Review 65(3): 343–70. ———. 2001. “The Welfare State, the Labor Market, and the Stability of Family Income: A Comparison of the U.S. and Germany.” EurAmerica 30(4): 1–46. Dworkin, Ronald. 2000. Sovereign Virtue. Cambridge, Mass.: Harvard University Press. Eriksson, Robert, and John Goldthorpe. 1992. The Constant Flux. Oxford: Clarendon Press. Esping-Andersen, Gøsta. 1990. The Three Worlds of Welfare Capitalism. Princeton, N.J.: Princeton University Press. ———. 1993. Changing Classes. London: Sage Publications. ———. 1994. “The Welfare State and the Economy.” In The Handbook of Economic Sociology, edited by Neil Smelser and Richard Swedberg. Princeton, N.J.: Princeton University Press. ———. 1999. “Politics Without Class?” In Continuity and Change in Contemporary Capitalism, edited by Herbert Kitschelt, Peter Lange, Gary Marks, and John Stephens. Cambridge: Cambridge University Press. Gangl, Markus. 2005. “Income Inequality, Permanent Incomes, and Income Dynamics: Comparing Europe to the United States.” Work and Occupations 32(2): 140–62. Gangl, Markus, Joakim Palme, and Lane Kenworthy. 2008. “Is High Inequality Offset by Mobility?” Unpublished paper. Department of Sociology, University of Arizona. Gilens, Martin. 2005. “Inequality and Democratic Responsiveness.” Public Opinion Quarterly 69(5): 778–96. Goodin, Robert, and John Dryzek. 1980. “Rational Participation: The Politics of Relative Power.” British Journal of Political Science 10(3): 273–92. Gould, Roger. 2003. Collision of Wills. Chicago, Ill.: University of Chicago Press. Hall, Peter H., and David Soskice. 2001. “Introduction.” In Varieties of Capitalism, edited by Peter H. Hall and David Soskice. Oxford: Oxford University Press. Hibbs, Douglas. 1973. Mass Political Violence. New York: John Wiley & Sons. Iversen, Torben, and Thomas R. Cusack. 2000. “The Causes of Welfare State Expansion: Deindustrialization or Globalization?” World Politics 52(3): 313–49. Iversen, Torben, and David Soskice. 2001. “An Asset Theory of Social Policy Preferences.” American Political Science Review 95(4): 875–93. ———. 2007. “Distribution and Redistribution: The Shadow of the Nineteenth Century.” Unpublished paper. Harvard University and Duke University. Iversen, Torben, and Anne Wren. 1998. “Equality, Employment, and Budgetary Restraint: The Trilemma of the Service Economy.” World Politics 50(4): 507–56. Jarvis, Sara, and Stephen Jenkins. 1998. “How Much Income Mobility Is There in Britain?” Economic Journal 108(447): 428–43. Keister, Lisa. 2000. Wealth in America: Trends in Wealth Inequality. Cambridge: Cambridge University Press.

414

Democracy, Inequality, and Representation

Kenworthy, Lane. 2008. Jobs with Equality. Oxford: Oxford University Press. Kenworthy, Lane, and Leslie McCall. 2008. “Inequality, Public Opinion, and Redistribution.” Socio-Economic Review 6(1): 35–68. Kenworthy, Lane, and Jonas Pontusson. 2005. “Rising Inequality and the Politics of Redistribution in Affluent Countries.” Perspectives on Politics 3(3): 449–74. King, Desmond, and David Rueda. Forthcoming. “Cheap Labor: The New Politics of Bread and Roses in Advanced Industrial Democracies?” Perspectives on Politics. Korpi, Walter 1983. The Democratic Class Struggle. London: Routledge. Krishna, Anirudh, editor. Forthcoming. Poverty, Participation, and Democracy: A Global Perspective. Cambridge: Cambridge University Press. Kumlin, Steffan. 2007. “The Welfare State: Values, Policy Preferences, and Performance Evaluations.” In The Oxford Handbook of Political Behavior, edited by Russell J. Dalton and Hans-Dieter Klingemann. Oxford: Oxford University Press. Lee, Woojin, and John E. Roemer. 2005. “The Rise and Fall of Unionized Labor Markets: A Political Economy Approach.” Economic Journal 115(500): 28–67. Leventoglu, Bahar. 2005. “Social Mobility and Political Transitions.” Journal of Theoretical Politics 17(4): 465–96. Lichbach, Mark I. 1990. “Will Rational People Rebel Against Inequality? Samson’s Choice.” American Journal of Political Science 34(4): 1049–76. Lindert, Peter. 2004. Growing Public. Cambridge: Cambridge University Press. Listhaug, Ola, and Toril Aalberg. 1999 “Comparative Public Opinion on Distributive Justice: A Study of Equality Ideals and Attitudes Toward Current Policies.” International Journal of Comparative Sociology 40(1): 117–40. Mares, Isabela. 2003. The Politics of Social Risk. Cambridge: Cambridge University Press. ———. 2005. “Social Protection Around the World: External Insecurity, State Capacity, and Domestic Political Cleavages.” Comparative Political Studies 38(8): 623–51. ———. 2006. Taxation, Wage Bargaining, and Unemployment. Cambridge: Cambridge University Press. Martin, Cathie Jo, and Duane Swank. Forthcoming. “The Political Origins of Coordinated Capitalism: Party Systems and State Structure in the Age of Innocence.” American Political Science Review. Mettler, Suzanne, and Joe Soss. 2004. “The Consequences of Public Policy for Democratic Citizenship: Bridging Policy Studies and Mass Politics.” Perspectives on Politics 2(1): 55–73. Moene, Karl O., and Michael Wallerstein. 2001. “Inequality, Social Insurance, and Redistribution.” American Political Science Review 95(4): 859–75. Morillas, Rafael. 2007. “Assets, Earnings Mobility, and the Black/White Gap.” Social Science Research 36(2): 808–33.

Inequality and Democratic Representation

415

Muller, Edward N. 1988. “Democracy, Economic Development, and Income Inequality.” American Sociological Review 53(1): 50–68. Muller, Edward N., and Mitchell A. Seligson. 1987. “Inequality and Insurgency.” American Political Science Review 81(2): 425–51. Osberg, Lars, and Timothy Smeeding. 2006. “‘Fair’ Inequality? Attitudes Toward Pay Differentials: The United States in Comparative Perspective.” American Sociological Review 71(3): 450–73. Perotti, Roberto. 1993. “Political Equilibrium, Income Distribution, and Growth.” Review of Economic Studies 60(4): 755–76. Persson, Torsten, and Guido Tabellini. 1994. “Is Inequality Harmful for Growth?” American Economic Review 84(3): 600–621. Piketty, Thomas. 1995. “Social Mobility and Redistributive Politics.” Quarterly Journal of Economics 110(3): 551–84. Pontusson, Jonas. 2005. Inequality and Prosperity: Social Europe vs. Liberal America. Ithaca, N.Y.: Cornell University Press. Przeworski, Adam. 2007. “Is the Science of Comparative Politics Possible?” In Handbook of Comparative Politics, edited by Carles Boix and Susan Stokes. Oxford: Oxford University Press. Rehm, Philipp. 2005. “Citizen Support for the Welfare State: Determinants of Preferences for Income Redistribution.” Discussion paper 2005-02. Berlin: Social Science Research Center Berlin (WZB). ———. 2007. “Risky Reforms? Determinants of Unemployment Benefits in OECD Countries.” Paper presented to the annual meeting of the American Political Science Association. Chicago, Ill., August 30–September 2, 2007. Rodrik, Dani. 1998. “Why Do More Open Economies Have Bigger Governments?” Journal of Political Economy 106(5): 997–1032. Roemer, John. 1998. Equality of Opportunity. Cambridge, Mass.: Harvard University Press. Roemer, John, Rolf Aaberge, Ugo Colombino, Johan Fritzell, Stephen P. Jenkins, Arnaud Lefranc, Ive Marx, Marianne Page, Evert Pommer, Javier RuizCastillo, Maria Jesus San Segundo, Torben Tranaes, Alain Trannoy, Gert G. Wagner, and Ignacio Zubiri. 2003. “To What Extent Do Fiscal Regimes Equalize Opportunities for Income Acquisition Among Citizens?” Journal of Public Economics 87(3–4): 539–65. Rothstein, Bo, and Eric M. Uslaner. 2005. “All for All: Equality, Corruption, and Social Trust.” World Politics 58(1): 41–72. Rueda, David. 2005. “Insider-Outsider Politics in Industrialized Democracies: The Challenge to Industrialized Democracies.” American Political Science Review 99(1): 61–74. ———. 2007. Social Democracy Inside-Out. Oxford: Oxford University Press. Shapiro, Thomas M., and Edward N. Wolff, editors. 2001. Assets for the Poor. New York: Russell Sage Foundation.

416

Democracy, Inequality, and Representation

Sherraden, Michael. 1991. Assets and the Poor: A New American Welfare Policy. New York: M. E. Sharpe. Stasavage, David, and Kenneth Scheve. 2007. “Institutions, Partisanship, and Inequality in the Long Run.” Unpublished paper. New York University and Yale University. Swenson, Peter. 1991. “Bringing Capital Back In, or Social Democracy Reconsidered.” World Politics 43(4): 513–44. ———. 2002. Capitalists Against Markets. Oxford: Oxford University Press. Thelen, Kathleen. 2004. How Institutions Evolve. Cambridge: Cambridge University Press. Ticchi, Davide, and Andrea Vindigni. 2003. “Endogenous Constitutions.” Working paper 726. Stockholm: Institute for International Economic Studies, Stockholm University. Uslaner, Ric, and Mitchell Brown. 2005. “Inequality, Trust, and Civic Engagement.” American Politics Research 33(6): 868–94. Varian, Hal R. 1980. “Redistributive Taxation as Social Insurance.” Journal of Public Economics 14(1): 49–67. Voitchovsky, Sarah. 2005 “Does the Profile of Economic Inequality Matter for Economic Growth?” Journal of Economic Growth 10(3): 273–96. Wang, T. Y., William J. Dixon, Edward N. Muller, and Mitchell A. Seligson. 1993. “Inequality and Political Violence Revisited.” American Political Science Review 87(4): 977–94. Wibbels, Erik, and John Ahlquist. 2008. “Development, Autarky, and Social Insurance.” Unpublished paper. Duke University and University of Washington.

Index

Boldface numbers refer to figures and tables. Aaron, H. Joseph, 40 absolute income, and electoral participation, 279–82 absolutism, 366–67 abstention, electoral, 279, 290, 291, 292, 295–99, 301–2 Acemoglu, D., 26, 362, 371, 399 Adema, W., 65 African Americans, 368–69 age analysis: political participation, 288; redistribution preference, 212 Ahlquist, J., 397 Alesina, A., 160n7, 380n1, 407 Allan, J., 86n5, 7 Almond, G., 281, 306n1 Alt, J., 6 American Political Science Review, 96 Anderson, C., 407 Ansell, B., 99 Armingeon, K., 253, 348 Atkinson, A., 26, 30, 54n1 Atkinson index, 30 attitudinal variables, 288 Austen-Smith, D., 9 Australia: disposable income inequality, 28, 135, 306; household income inequality, 326; income inequality

and voter turnout, 294; market income inequality, 133; party and electoral systems, 104; political mobilization, 331; political parties, 329; social spending trends, 72; wage inequality, 130, 172, 326; welfare benefit generosity, 70, 72 Austria: disposable income inequality, 28, 306; electoral system, 103; income inequality and voter turnout, 294; party and electoral systems, 104; social spending trends, 72; wage inequality, 172; welfare benefit generosity, 70, 72 Bank of Italy, 51 Banks, J., 9 Bartels, L., 339 BEA. See Bureau of Economic Analysis Beck, N., 182 Beck, T., 158 Becker, I., 49 Belgium: CD-left party coalitions, 99; disposable income inequality, 28, 135, 306; household income inequality, 326; income inequality and voter turnout, 294; labor force par-

418

Index

Belgium (continued) ticipation, 373; party and electoral systems, 104; political mobilization, 331; political parties, 329; political polarization, 342, 345; social spending trends, 72; voting rights, 373; wage distribution, 350n6; wage inequality, 130, 172, 326; welfare benefit generosity, 70, 72 Benabou, R., 399–400 Benoit, K., 327 Beramendi, P., 177 Bergstrom, T., 284, 307n2 bicameralism, 105, 249 bivariate correlation, 77–78, 283 Black Death, 365 Blais, A., 212–13 Blake, D., 212–13 Blanchard, O., 221, 230 Blue Book (BB), 42 Blume, L., 284 Blyth, M., 141 Boix, C., 26, 99, 283, 356, 371 Bos, W., 48 Bradley, D., 7, 78, 81, 87n16, 87n17, 92, 162n21, 163n25, 164n28 Brady, D., 7 Brady, H., 281 Brandolini, A., 26, 41, 43, 51, 54n1 Britain. See United Kingdom British Household Panel Survey, 42 bubonic plague, 365 Buckley, W., 381n13 Budge, I., 191, 327 Bureau of Economic Analysis (BEA), 40, 41, 55n8 Busemeyer, M., 99 business cycle, 279 Canada: disposable income inequality, 28, 135, 306; income inequality and voter turnout, 294; income inequality trends, 42–45; market in-

come inequality, 133; minimum wage, 144, 161n12; net worth inequality, 403; party and electoral systems, 104; social spending trends, 72; wage inequality, 130, 172; wealth inequality, 403; welfare benefit generosity, 70, 72 capital income, 129, 131, 132, 159n4 capital market openness, definition of, 79 Card, D., 161n12 Castles, F., 111 causality, 285, 339, 355–56 CBO. See Congressional Budget Office CD (Christian Democratic) parties. See Christian Democratic parties center of gravity (CoG), political, 110, 222, 314, 322–23, 331–32 Central Bureau of Statistics (CBS), 48 centrist governments, 110–11 Charles Martel, 364–65 Chevalier, P., 50 Christian Democratic (CD) parties: and female labor force participation, 258; in Germany, 111; and income redistribution, 137–38; and left-centered parties’ influence, 98–99; measurement of, 255; and welfare benefit generosity, 247 The Civic Culture (Almond and Verba), 281 Clark, G., 369 class, 26, 281, 287–88 CMEs. See coordinated market economies CMP. See Comparative Manifesto Project coalition formation, model of, 96–101 Cobb-Douglas production function, 358 collective action, 284, 408 college education, 146, 149

Index Comparative Manifesto Project (CMP), 313, 322, 324, 327–28 Comparative Welfare Entitlements Dataset (CWED), 62 Concialdi, P., 50 Congressional Budget Office (CBO), 42 “consensus view” of economic inequality, 127–28 constitutional veto points, 79, 105, 109 consumption, government, 179, 188, 190, 195n15, 17 coordinated market economies (CMEs), 138–40 coordination, economic. See economic coordination corporate income, taxes on, 180–82, 191 corporatism: definition of, 190; economic characteristics, 194 n.6; equality requirement, 396–97; and income skew, 258; measurement of, 254; and partisanship, 183, 186, 188, 189; and policy, 176–77, 183, 186, 188; traits of, 176 counterfactual problem, 401–2 CPS. See Current Population Survey cross-national comparisons: disposable income differences, 27–36; electoral institutions and systems, 104; household income inequality, 326; income inequality trends, 3–4, 40–51; in-kind benefits, 39–40; market income inequality, 38, 131–33, 324–25; methodology, 26; 90/10 ratio, 28, 129, 130, 324, 326; redistribution, 37–39, 48–50; research considerations, 26–27; wage inequality, 130, 171–73, 326 Current Population Survey (CPS), 40, 41–42, 55n8, 307n7 Cusack, T., 103, 173, 177, 179, 222, 227, 254, 255

419

CWED. See Comparative Welfare Entitlements Dataset Czech Republic, disposable income inequality, 28 Database on Financial Structure and Economic Development, 147 data sources: disposable income, 87n18; disposable income inequality, 306, 348; electoral institutions, 102; endogeneity, 253; female labor force participation, 146; household income inequality, 348; income distribution, 54–55n7; income inequality, 324–27; manufacturing sector employment, 146; partisanship, 178, 191; political polarization, 313, 323–27, 348; public opinion, 210–11; public spending, 63–65; redistribution, 102, 203, 210–11; relative wages, 324, 325; social insurance programs, 62; stock market capitalization, 254; trade, 146, 191, 254; union density, 348; voter turnout, 348; wage inequality, 348; welfare spending, 62, 63–65 debt, government, 184, 190 decile ratio, 29–30, 36 decommodification index, 66, 68, 303 De Deken, J., 85n2 deindustrialization, 148, 227–28 Deininger, K., 26 Demirgüç-Kunt, A., 158 democracy and democracies: and equality, 367–68, 379; and inequality, 25–26, 278, 356–57, 371, 387–88; political and economic differences among, 93–94; and trade, 379; transitions to, 371; and wealth, 377 Democracy and Development (Przeworski), 356–57

420

Index

Democracy and Redistribution (Boix), 26 Democratic Party (U.S.), 338 demographics, 245–46, 259 Denmark: disposable income inequality, 28, 135, 306; household income inequality, 326; income inequality and voter turnout, 294; labor force participation, 373; market income inequality, 133; party and electoral systems, 104; political mobilization, 331; political parties, 329; political polarization, 342, 345; social spending trends, 72; unemployment insurance, 71; voting rights, 373; wage inequality, 130, 172, 326; welfare benefit generosity, 70, 72 dependency ratio, 223, 289 Dion, S., 212–13 Directory of Trade Statistics, 146 disemployment effects, 143, 144 disparate effects, 285–86 disposable income: absolute differences, 31, 35–36; conclusions, 51–53; cross-national comparison of differences, 27–36; data sources, 87n18; definition of, 27; distribution in thirty-two countries, 28; government’s role in distribution, 136–37; in household income inequality measurement, 324–25; and in-kind benefits, 39–40; monetary redistribution levels, 37–39, 48–50; relative differences, 29–31 disposable income inequality: data sources, 306, 348; estimation, 145, 148, 153, 154–55; in high-income and middle-income countries, 28; and market income inequality, 148; in OECD countries, 134, 135, 306; and partisanship, 148, 155, 156; in rich countries, 50–51; and unions,

148, 155, 156; and wage coordination policies, 156 distribution of income, 11, 12, 54–55n7, 136–37. See also redistribution Dolado, J., 179, 191 Downsian model of political competition, 6 Dryzek, J., 283, 285 Earned Income Tax Credit (EITC), 65 Ebbinghaus, B., 254, 348 economic conditions, 279–80 economic coordination: coordinated market economies, 138–40; and disposable income inequality, 148, 154–55, 156; measurement of, 162n18; and partisanship, 150–51; and PR governments, 163n23; and wage inequality, 152 economic growth and development, 11, 25, 64–65, 288, 409 economic inequality: “consensus view” of, 127–28; and electoral systems, 356; factors in, 136–44; future of, 157–58; measurement of, 356; methodology, 144–48; patterns within OECD countries, 128–36; research findings, 148–57; wealth, 402–3. See also income inequality; wage inequality economic institutions, 11–14, 127–28 Economic Origins of Dictatorship and Democracy (Acemoglu and Robinson), 26 Economic Report of the President (1994), 26 Economic Trends (ET) series, 42 The Economist, 3 educational attainment, 281, 287–88 efficiency effects, 143 efficiency wages, 161n11

Index EFS. See Expenditure and Food Survey Eisenstein, E., 366 EITC. See Earned Income Tax Credit elderly population, 158 elections, and welfare spending, 9 electoral abstention, 279, 290, 291, 292, 295–99, 301–2 electoral institutions and systems: coalition formation model, 96–101; cross-national comparison, 104; data sources, 102; and economic inequality, 356; and inequality, 396–97; measurement of, 103–4, 255–56; partisanship, 94–96, 101, 115; and party fragmentation, 112–13; and redistribution, 8–9, 101–2, 107–9, 208–9, 222, 228; research considerations, 95–96. See also majoritarian systems; political parties; proportional representation (PR) electoral participation: and absolute income, 279–82; conclusions, 300–304; and income, 239–41; and income inequality, 282–87, 288, 293–300, 304; methodology, 286–89, 304–5; multivariate analysis, 289–300; and relative income, 282–86, 288, 289–93; research considerations, 278–79; and unemployment risks, 239–41. See also voter turnout employers, 138 Employment Outlook, 182 endogeneity: data, 253; estimation results, 258–67; estimation strategies, 256–58; importance in research, 14; instrumental variable approach, 289; measurement of, 253–56; overview, 232, 241; system-ofequations approach, 241–53 endowments, 359, 379

421

Engerman, S., 356 entitlement approach, to welfare, 66 equality, 25, 367–68, 379, 396–97, 408–10 Esping-Andersen, G., 63–64, 66, 69, 179, 328 Estonia, disposable income inequality, 28 EUROMOD, 38–39 European Community Household Panel, 27, 36 Expenditure and Food Survey (EFS), 43 factor prices, 359 Family Expenditure Survey (FES), 43 fast-food industry, 161n12 federalism, 105, 249 FES. See Family Expenditure Survey feudalism, 364–65 50/10 ratios, 172, 173, 306 financial exposure, 258 financial openness, 184, 190 Financial Statistics, 146 Finer, S., 366 Finland: disposable income inequality, 28, 135, 306; household income inequality, 326; income inequality and voter turnout, 294; income inequality trends, 46; market income inequality, 133; net worth inequality, 403; party and electoral systems, 104; political mobilization, 331; political parties, 329; social spending trends, 72; wage inequality, 130, 172, 326; wealth inequality, 403; welfare benefit generosity, 70, 72 fiscal policy, 6, 403–4 Fitoussie, P., 144 fixed effects, 182–83 Flora, P., 372, 383n36 Fording, R., 323, 331–32, 348

422

Index

France: democratization, 367–68; disposable income inequality, 28, 135, 306; electoral system, 103; household income inequality, 326; income inequality and voter turnout, 294; income inequality trends, 47–48; labor force participation, 373; market income inequality, 133; minimum wage, 144; party and electoral systems, 104; political mobilization, 331; political parties, 329; political polarization, 342, 344–45; social spending trends, 72; voting rights, 373; wage inequality, 130, 172, 326; welfare benefit generosity, 72 Frankish kingdom, 364–65 Franzese, R., 244 free-rider problem, 284 Frenette, M., 44 Friedman, M., 404–5 future research agenda: collective action, 408; different types of inequality, 402–5; at individual level, 398–402; at macro level, 394–98; overview, 394–95; political polarization, 346; proportional representation, 116–17 Gallagher, M., 104 game theory, 356 Garfinkel, I., 39 Garicano, L., 371 Garrett, G., 11, 177 GDP. See gross domestic product gender analysis: political participation, 288; redistribution preference, 212. See also women generosity index. See welfare benefit generosity index German Christian Democrats (CDU/CSU), 111 Germany: capital income vs. U.S.,

129, 131; disposable income inequality, 28, 135, 306; endogenous relationships analysis, 264–65; household income inequality, 326; income inequality and voter turnout, 294; income inequality trends, 46–47; labor force participation, 372, 374; market income inequality, 133; net worth inequality, 403; party and electoral systems, 104, 111; political mobilization, 331; political parties, 329; social spending trends, 72; voting rights, 372, 374; wage inequality, 130, 172, 326; wealth inequality, 403; welfare benefit generosity, 70, 72 Gingerich, D., 160n8, 162n18 Gini index: Canada, 44; cross-national comparison, 28, 38; definition of, 324; disadvantage of, 159–60n5; electoral participation analysis, 286–87; Finland, 47; France, 50; Germany, 132; Italy, 51; of market income, 38, 76; Netherlands, 48; redistribution analysis, 37, 102; Sweden, 45; U.K., 43; U.S., 40–42, 132, 338; welfare benefit generosity and redistribution outcomes, 76–83; West Germany, 49 Glaeser, E., 160n7, 380n1 globalization, 3, 203, 228 Golden, M., 254, 255 Goodin, R., 283, 285 Gottschalk, P., 28, 36 Gould, R., 411n7 government, role in distribution, 136–37 government center of gravity (CoG), 110, 222, 314, 322–23, 331–32 government consumption, 179, 188, 190, 195n15, 17 government debt, 184, 190

Index government employment, 186, 191, 212–13 government partisanship. See partisanship government regulation, 136 government transfers, 49–50, 52, 93, 222 Gratschew, M., 305 Great Britain. See United Kingdom Greece, ancient, 362, 363 Greece, modern, disposable income inequality, 28, 306 Green, D., 44 Gross, D., 103 gross domestic product (GDP): and democratization, 356–57; growth, 184, 190; U.S. vs. Italy and Finland, 31; and voter enfranchisement, 378 Gustafsson, B., 45, 45 Hacker, J., 343 Hall, P., 160n8, 162n18, 163n23 Hasse diagram, 30–31, 34 Hays, J., 244 HBS. See Household Budget Survey Hibbs, D., 6, 11–12 Hicks, A., 190 hierarchical linear model, 289 historical analysis: absolutism, 366–67; ancient Greece, 362, 363; Black Death, 365; conclusions, 378–79; democratization, 367–68; feudalism, 364–65; overview, 356, 362–63; Reformation, 365–66; Roman Republic, 363–64; voting rights, 369–78; World Wars I and II, 368–69 Hobolt, S., 263 Household Budget Survey (HBS), 47 household equivalency, 87n13 household income, wages as percentage of, 129

423

household income inequality: crossnational comparison, 326; data sources, 348; measurement of, 324–25; and political polarization, 334–40, 341, 346; and right-oriented political parties, 320, 336; in U.K., 344 Huber, E., 102, 105, 247, 254, 255 human capital, 6, 146, 149 Hungary, disposable income inequality, 28 ideological differences. See partisanship IDS. See Income Distribution Survey IFS. See Institute for Fiscal Studies ILO. See International labor Office IMF. See International Monetary Fund imports, 146, 148, 185 incentive model of political participation, 280–81, 285, 303 incidence, 401 Inclan, C., 254 income: absolute income, 279–82; and electoral abstention, 290, 291, 292; and electoral participation, 240, 278–87, 289–93, 303–4; measurement of, 55n7, 287; as political resource, 280–81; real income, 31, 35–36; and redistribution preference, 212, 218–21; relative income, 282–86, 288, 289–93; as resource vs. incentive, 303; and risk exposure, 220–21; and unemployment risk, 239; wages as percentage of, 129. See also disposable income; market income; wages Income and Consumption Survey (EVS), 49 income distribution, 11, 12, 54–55n7, 136–37. See also redistribution Income Distribution Survey (IDS), 45, 46, 47

424

Index

income inequality: citizens’ perceptions of, 407; data sources, 324–27; and electoral abstention, 279, 295–99, 301–2; and electoral participation, 282–87, 288, 293–300, 304; endogeneity of, 289; forms of, 319–20; measurement of, 324–27; median-preserving increases, 237–39; Moene-Wallerstein model, 233–37; and redistribution, 380n2; and right-oriented political parties, 320; social welfare consequences, 359–62; and voter turnout, 293. See also disposable income inequality; household income inequality; market income inequality income mobility, 399, 405 Income Panel Survey (IPS), 48 income replacement rates, 67, 175 income skew: endogenous relationships analysis, 258–67; in Germany, 265; Meltzer-Richard model, 315; in Sweden, 266; in U.K., 267; in voting, 321 individual winners and losers, 405–8 industrial revolution, 369 inequality: different forms of, 402–5; political inequality, 358–62. See also economic inequality in-kind benefits, 39–40 Institute for Fiscal Studies (IFS), 42 institutions: and inequality, 354; median voter model’s neglect of, 7; and political agency, 170–71, 176; role in politics of distribution, 10–14. See also political institutions International Institute for Democracy and Electoral Assistance, 105 International labor Office (ILO), 211 International Monetary Fund (IMF), 35, 146 international openness, 184, 191

International Social Survey Program (ISSP), 210–11 International Standard Classification of Occupations (1988) (ISCO), 211 intra-party competition, 249, 255 IPS. See Income Panel Survey Ireland: disposable income inequality, 28, 306; income inequality and voter turnout, 294; party and electoral systems, 104; social spending trends, 72; welfare benefit generosity, 72; welfare spending, 64–65 ISCO. See International Standard Classification of Occupations Ishikawa, T., 27, 28, 36 Israel, disposable income inequality, 28 ISSP. See International Social Survey Program Italy: disposable income inequality, 28, 135, 306; household income inequality, 326; income inequality and voter turnout, 294; income inequality trends, 48; labor force participation, 374; net worth inequality, 403; party and electoral systems, 104; political mobilization, 331; political parties, 329; social spending trends, 72; voting rights, 374; wage inequality, 130, 172, 326; wealth inequality, 403; welfare benefit generosity, 70, 72 Iversen, T., 10, 12–13, 121n2, 155, 179, 185, 190, 211, 227, 334 Jäntti, M., 46 Japan: disposable income inequality, 28; minimum wage, 191; partisanship, 191; party and electoral systems, 104; social spending trends, 72; wage inequality, 172; welfare benefit generosity, 70, 72 Jarvis, S., 405

Index Jenkins, S., 42, 405 Johnson, S., 362 Katz, J., 182 Katzenstein, P., 176 Kayser, M., 380 Keman, H., 191 Kennedy, J. F., 3 Kenworthy, L., 190, 254 Kim, H., 323, 331–32, 348 Kittel, B., 85n2 Klemmensen, R., 263 Klingemann, H., 348 Korpi, W., 66, 85n4 Krishna, A., 410n5 Krueger, A., 161n12 Krugman, P., 169 Kuznets, S., 5, 25, 40, 410 Kwon, H. Yong, 321 Laasko, M., 103, 332, 348 labor, in coordinated market economies, 139 labor force participation: and voting rights, 372–77; of women, 105, 107, 146, 148–49, 157–58, 185, 247–48, 254 Labor Force Statistics, 146, 223 labor force surveys, 211 labor market institutions and policies, 6, 137 labor’s wage, 358 Labour Party (Britain), 343 Ladaique, M., 65 La Ferrara, E., 283, 284–85 Lange, P., 11, 254 Latin America, 116 Laver, M., 327 Lee, W., 396–97 left-oriented governments: and disposable income inequality, 155, 156; education spending, 99, 108–9; by electoral system type, 95; government consumption, 179; policy

425

preferences, 188; and redistribution, 94–95, 224–25, 227–28; tax policy, 180; and unionization, 115; wage coordination, 142–43, 177; wage floor hypothesis, 173, 175–76 left-oriented political parties: core constituencies, 317, 319–20; equality preference, 407; and household income inequality, 335; inequality impact, 319–20; median voters, 334; policy preferences, 137, 312; positions of, 334; and wage inequality, 334, 336 Lenski, G., 25 Level-of-Living Survey (LLS), 45 Leventoglu, B., 399 Levine, R., 158 liberal market economies (LMEs), 138–39 Lichbach, M., 411n7 Lijphart, A., 103–4, 105, 222 Lipset, S., 377 LIS. See Luxembourg Income Study Listhaug, O., 306n2 literacy, 99, 366 LLS. See Level-of-Living Survey LMEs. See liberal market economies logistic regression model, 289 Lorenz dominance, 30, 31, 32–33, 46 low-income voters, 320–22, 330–31, 347, 348 Luther, M., 365 Luxembourg, disposable income inequality, 28 Luxembourg Income Study (LIS): disposable income, 36, 38, 76, 87n18; disposable income inequality, 306, 348; electoral participation analysis, 286–87; income inequality, 27–29, 145, 324; limitations, 159n1, 395; market income, 38, 76; poverty rate, 93; redistribution analysis, 37, 77, 102

426

Index

Mackie, T., 105 MacRae, D., 393, 396, 397 Maddison, A., 381n21 Mahler, V., 321 Mair, P., 111 majoritarian systems: coalitional politics under, 97–98, 110; and inequality, 361, 396; key indicators of, 103, 104; and partisanship, 94, 95; payoffs from support of different parties or coalitions, 100; and redistribution, 224; right-oriented governments under, 113; social insurance provision, 209 manufacturing workers, as percentage of population (ME), 146 market income: cross-national comparison, 131–33; definition of, 128–29, 159n2; government’s role in distribution, 136; measurement of, 147; mismeasurement of, 76–77; sources of, 131. See also wages market income inequality: cross-national comparison, 38, 131–33, 324–25; and disposable income inequality, 148; estimation, 145, 152–54; future of, 158 married people, 288 Martel, C., 364–65 Martin, C., 397 McCarty, N., 117, 313, 317, 343 McDonald, M., 191, 305 measurement. See methodology median voter model: assumptions of, 5–6; limitations of, 6–9; positions of, 323, 332, 333, 334, 348; redistribution preference, 317, 319. See also Meltzer-Richard model Meltzer, A., 93, 98, 99, 105, 115, 205, 212, 218, 320 Meltzer-Richard model: assumptions of, 315; with core constituencies,

315–19; country differences in redistribution, 312; illustration of, 316, 360; income inequality, 319; voting inequality, 320–21 men, electoral participation, 288 Mendes, S., 191 methodology: Christian Democratic parties, 255; corporatism, 254; disposable income inequality, 145, 148, 153, 154–55; economic coordination, 162n18; economic inequality, 356; electoral institutions and systems, 103–4, 255–56; electoral participation, 286–89; endogeneity, 253–56; household income inequality, 324–25; income, 55n7, 287; income inequality, 324–27; market income, 76–77, 147; minimum wage, 179; partisanship, 103, 147, 177–78, 182–83, 186, 254–55; political participation, 254; political polarization, 323–34; redistribution, 210–13, 221–23; unemployment, 191; union density, 254; wage inequality, 324; welfare benefit generosity index, 66–69, 74 Mexico, disposable income inequality, 28 middle class, 26, 115–16 Milanovic, B., 121n4 military, 358–59, 366–67, 368–69 minimum wage: definition of, 191; and earnings distribution, 136; effects of, 143–44, 152, 161n12, 178–79; measurement of, 179; setting by left vs. right-oriented governments, 173 mobility, 399–400, 405 mobilization, political, 320–22, 330–31, 335–40, 347, 348. See also voter turnout Moene, K., 161n14, 233–37, 391 Möller, S., 78, 81, 84, 87n16

Index multiparty competition, 332, 334 multivariate analysis, 78–83, 289–300 Nagel, J., 270n15 Napoleon, 367 National Accounts, 222 National Statistical Office, 43 Nef, J. Ulric, 367 neoliberal regimes, 13 Netherlands: CD-left party coalitions, 99; center of political gravity, 322; disposable income inequality, 28, 135, 306; household income inequality, 326; income inequality and voter turnout, 294; income inequality trends, 46; labor force participation, 374; market income inequality, 133; party and electoral systems, 104; political mobilization, 331; political parties, 329; political polarization, 342, 345; social spending trends, 72; voting rights, 372, 374; wage inequality, 130, 172, 326; welfare benefit generosity, 70, 72 net income replacement rates, 67 net worth inequality, 403 Neumark, D., 191 Newmark, D., 161n12 New Zealand: party and electoral systems, 104; social spending trends, 72; welfare benefit generosity, 72 90/10 ratio: cross-national comparison, 28, 129, 130, 324, 326; limitations and alternatives, 242; Sweden, 338; U.S., 336 90/50 ratio, 242, 259, 262 99/50 ratio, 306 Niskanen, W., 212 noncash redistribution, 39–40 nonlinear least squares, 223 North, D., 381n17 Norway: disposable income inequal-

427

ity, 28, 135, 306; household income inequality, 326; income inequality and voter turnout, 294; labor force participation, 375; market income inequality, 133; net worth inequality, 403; party and electoral systems, 104; political mobilization, 331; political parties, 329; social spending trends, 72; voting rights, 375; wage distribution, 350n6; wage inequality, 130, 172, 326; wealth inequality, 403; welfare benefit generosity, 72 OECD. See Organization for Economic Cooperation and Development Ok, E., 399–400 OLS. See ordinary least squares Olson, M., 284 openness, international and financial, 184, 190, 191 ordinary least squares (OLS), 145, 376, 377 Organization for Economic Cooperation and Development (OECD): Comparative Welfare Entitlements Dataset (CWED), 62; Employment Outlook, 182; inequality effect on electoral participation, 293–300; inequality patterns, 128–36; Labor Force Statistics, 146, 223; National Accounts, 222; purchasing power parity indices, 54n5; relative wage data set, 324, 325; social expenditure database, 63; Structure of Earnings Database, 54n1; wage inequality, 130, 171–73 O’Rourke, K., 368 Osberg, L., 30, 39, 121n4, 407 P10, 28, 29, 36 P90, 28, 29, 36 Palmer, E., 45

428

Index

panel-corrected standard errors (PCSE), 182–83 Panel Study of Income Dynamics (PSID), 129, 400 partisan center of gravity (CoG), 110, 222, 314, 322–23, 331–32 partisanship: coalitional dynamics, 94–95; and corporatism, 176–77, 183, 186, 188, 189; data sources, 178, 191; and disposable income inequality, 148, 155, 156; and economic coordination, 150–51; future of, 158; in Germany, 265; and inequality, 382n28; limitations of, 10–11; methodology, 103, 147, 177–78, 182–83, 186, 254–55; model, 106–7; and policy preferences, 178–88; and redistribution, 6–7, 94, 109–15, 208–9, 222, 226–27, 321–22; in Sweden, 266; in U.K., 267; variables, 105–6, 183–86; and wage distribution, 339; and wage inequality, 151–52, 170, 173–75. See also left-oriented governments; polarization, political; right-oriented governments PCSE. See panel-corrected standard errors Pearson correlation, 307n7, 307n8 Penn World Table, 78 pensions, 81 perceptions, 407–8 Persson, T., 25, 109, 359, 409 Picot, G., 44 Pierson, P., 63, 213, 222, 343 Piketty, T., 382n26 Plümper, T., 271n32 Poland, disposable income inequality, 28 polarization, political: center of political gravity, 322–23, 331–32; change over time, 340–46; conclusions, 346–47; data sources, 313, 323–27,

348; forms of, 313; and inequality, 319–20, 333–40; low-income voter mobilization, 320–22, 330–31, 335–40, 347, 348; Meltzer-Richard model with core constituencies, 315–19; methodology, 323–34; party positions, 327–30, 334; research considerations, 312–14, 346; theoretical framework, 314–23 policy instruments, 136–37 policy issues and preferences: corporatism, 176–77, 183, 186, 188; of left-oriented governments and political parties, 137, 188, 312; and partisanship, 178–88; of right-oriented political parties, 312; and unemployment, 184; variables for analysis of, 183–84; wage inequality, 188 political agency, research considerations, 170 political center of gravity (CoG), 110, 222, 314, 322–24, 331–32 political contexts, 249–51 political ideology. See partisanship political inequality, 358–62 political institutions: endogenous relationships analysis, 248–49; and inequality, 354–57; and political participation, 289; role in politics of distribution, 11–14. See also electoral institutions and systems political mobilization, 320–22, 330–31, 335–40, 347, 348. See also voter turnout political participation: endogenous relationships analyses, 241–43, 258–67; in Germany, 265; incentive model of, 280–81, 285; and income, 239–41; measurement of, 254; and political institutions, 249; resource model of, 280–82, 285; in Sweden, 266; in U.K., 267; and un-

Index employment risks, 239–41. See also electoral participation political parties: in coordinated market economies, 139–40; and economic outcomes, 6, 170; effective number of, 332, 334, 348; fragmentation of electoral system, 112–13; goals of, 11; intra-party competition and political participation, 249, 255; platforms and policy preferences, 194n10; positions of, 327–30, 334, 348; role in politics of distribution, 11–14, 137–38; weight of, 194n12. See also polarization, political political polarization. See polarization, political political science, 25 Pontusson, J., 12–13, 140, 163n22, 164n29, 173, 175, 189–90, 321 Poole, K., 117, 313, 317, 343 population, 158 Portugal, disposable income inequality, 28, 306 post-tax income, 76–77 poverty rates, 76, 77, 93 Powell, G. Bingham, 94, 355, 380n11 power resource theory, 7, 105, 138 PPP. See purchasing power parity presidentialism, 248–49, 255 pre-tax-and-transfer inequality, 105, 107 printing, 366 proportional representation (PR): coalitional politics under, 96–97, 100–101, 110; and economic coordination, 163n23; and electoral systems, 209; future research, 116–17; and inequality, 356, 396; key indicators of, 103, 104; and partisanship, 94, 95; payoffs from support of different parties or coalitions, 100; and redistribution, 109, 204, 209–10, 224, 228, 380n1

429

Protestants, 365–66 PR. See proportional representation Przeworski, A., 356–57, 371 PSID. See Panel Study of Income Dynamics public education, 99, 108 public goods, 284 public institutions, 407 public opinion data, 210–11 public pensions, 71, 73, 160n6 public sector employment, 186, 191, 212–13, 227 public spending, 62, 63–65 purchasing power parity (PPP), 31 Quinn, D., 79, 254 Ragin, C., 105 Rainwater, L., 39 rational actor model, 411n4 real income, 31, 35–36 realized risk, 212 real per capita income, 105, 107 redistribution: coalition formation model, 96–101; cross-national comparison, 37–39, 48–50; data sources, 102, 203, 210–11; demand for, 203–7, 228; electoral system impact, 107–9; endogenous relationships analyses, 241–43, 258–67; in Germany, 265; indicators of, 76; limits to, 408–9; measurement of, 37; median voter model, 5–9, 25; models of, 106–7, 204–10, 233–37; partisanship impact, 109–15; research considerations, 203–4; and risk exposure, 210–21; and shocks, 221–28; spending, 253; supply of, 208–10, 228; in Sweden, 266; in U.K., 267; variables, 102–6, 212–13, 222–23; and welfare benefit generosity, 76–83. See also Meltzer-Richard model

430

Index

redistricting, 113 referenda, 105, 249 Reformation, 365–66 Rehm, P., 211 relative income, 282–86, 288, 289–93 relative wages, 324, 325 rent of capital, 358 Republican Party (U.S.), 338, 339 research considerations: cross-national comparison, 26–27; economic inequality, 148–57; electoral institutions and systems, 95–96; electoral participation, 278–79; individual winners and losers, 405–7; overview, 388–94; political agency, 170; redistribution, 203–4; subjective inequality, 407–8; welfare benefit generosity index, 62; welfare spending, 62–63. See also future research agenda research methodology. See methodology resource model of political participation, 280–82, 285, 303 retiree ratio, 64, 65 retirement age population, 147, 154, 212, 258 “Reversal of Fortune” (Acemoglu, Johnson, and Robinson), 362–63 Richard, S., 93, 98, 99, 105, 115, 205, 212, 218, 320 right-oriented governments: and disposable income inequality, 155; education spending, 99; by electoral system type, 95; and redistribution, 94–95, 109, 224–25; in Sweden, 141–42; trend toward, 322–23; vote-seat disproportionalities, 113 right-oriented political parties: core constituencies, 317, 319–20; and economic coordination, 140–41; and household income inequality,

335, 336, 340; inequality impact, 319–20; median voters, 334; policy preferences as inequality rises, 312; positions of, 334; and wage inequality, 334–35 risk, 205–6, 208, 210–21, 398–99 Ritakallio, V., 46 Robinson, J., 26, 362, 371, 399 Roemer, J., 396–97, 403–4 Rogowski, R., 380n1, 393, 396, 397 Rokkan, S., 112, 113 Romania, disposable income inequality, 28 Roman Republic, 363–64 Rose, R., 105 Rosenthal, H., 117, 313, 317, 343 Royal Commission on the Distribution of Income and Wealth, 43 Rueda, D., 12–13, 164n29, 173, 175, 189–90, 272n35 Russia, disposable income inequality, 28 Sacerdote, B., 380n1 Scandinavia, 13, 156, 402–3 SCF. See Survey of Consumer Finances Scheve, K., 356, 395–96, 410n4 Schlozman, K., 281 Schwabish, J., 30, 39, 121n4 Scruggs, L., 86n5, 7 Scuggs Index, 289 self-employed workers, redistribution preference, 212 service industry, 185, 191 SHIW. See Survey of Household Income and Wealth shocks, and redistribution, 221–28, 228 sick pay, 86n8 Sigelman, L., 103 Singer, M., 407 skill levels, of workers, 205–6, 211

Index SLID. See Survey of Labor and Income Dynamics Slovak Republic, disposable income inequality, 28 Slovenia, disposable income inequality, 28 Smeeding, T., 28, 30, 36, 39, 121n4, 407 Social Citizenship Indicator Project, 85n4 social class, 26, 281, 287–88 social democratic parties, 247, 255, 258 Social Democrats (Sweden), 141–42, 344 social insurance: endogenous relationships analyses, 241–43, 258–67; in Germany, 265; Moene-Wallerstein model, 233–37; spending, 253; in Sweden, 266; in U.K., 267 social mobility, 399–400 social rights approach, to welfare, 66 social security, 37 social spending, 72. See also welfare spending social welfare, 359–62. See also welfare spending Socio-Economic Panel (SOEP), 47, 49, 129 socioeconomic status, and political participation, 280–82 SOEP. See Socio-Economic Panel Sokoloff, K., 356 Solow, R., 40 Solt, F., 283 Soskice, D., 121n2, 155, 163n23, 211, 334 Spain, disposable income inequality, 28, 306 spatial interdependence, 244–45, 258 spending ratios, 64, 74 spillover effects, 144 Squire, L., 26

431

standard of living, 31, 35 Stark, T., 44 Stasavage, D., 356, 395–96, 410n4 Statistics Canada, 44 Statistics Finland, 47 statutory minimum wage, 191 Stephens, J., 102, 105, 160n6, 162n21, 247, 254, 255 stock market capitalization: data source, 254; and income skew, 259; and market income inequality, 147, 154; and redistribution, 258; and social insurance, 258; trends, 158; wage and employment effects, 248, 259 stock market returns, 248, 254 Stolper-Samuelson theorem, 247 Strom, K., 319 students, redistribution preference, 212 subjective inequality, 407–8 Supplementary Security Insurance, 71 Survey of Consumer Finances (SCF), 44 Survey of Household Income and Wealth (SHIW), 51 Survey of Labor and Income Dynamics (SLID), 44 Swank, D., 397 Sweden: disposable income inequality, 28, 135, 306; endogenous relationships analysis, 265–66; household income inequality, 326; income inequality and voter turnout, 294; income inequality trends, 45–46; labor force participation, 375; market income inequality, 133; net worth inequality, 403; party and electoral systems, 104; political mobilization, 331, 338; political parties, 329; political polarization, 342, 344; poverty rate

432

Index

Sweden (continued) reduction from taxation and transfers, 93; redistribution in, 134; right wing governments, 141–42; social spending trends, 72; union density, 341; unions, 141; voter turnout, 341, 344; voting rights, 375; wage inequality, 130, 169, 172, 326; wealth inequality, 403; welfare benefit generosity, 70, 72 Switzerland: disposable income inequality, 28; electoral system, 110; income inequality and voter turnout, 294; social spending trends, 72; unemployment insurance, 71; voting rights, 382n30; wage inequality, 172; welfare benefit generosity, 70, 72 Taagepera, R., 103, 332, 348 Tabellini, G., 25, 109, 359, 409 Taiwan, disposable income inequality, 28 Tanzania, 283, 285 tax-and-transfer systems, 49–50, 52 taxation, 25, 65, 180–82, 188, 191 tax policy, 6 Tax Revenue Survey (France), 47–48, 50 technological change, 358 teens, minimum wage and employment, 161n12 Theil index, 30 third world imports (TWI), 146, 148 Thirty Years’ War, 366–67 three-stage least squares, 258 Ticchi, D., 356, 379, 396 time-series cross-sections (TSCS), 256 Tocqueville, A. de, 367 trade: data sources, 146, 191, 254; and democratization, 379; in eighteenth century, 368; and inequality, 358; third world imports, 146, 148;

wage and employment effects, 148, 247, 258; and wage inequality, 185 training systems, 204, 208, 222–23, 225, 228 Tröger, V., 271n32 unemployment: endogenous relationships analyses, 241–43, 258–67; in Germany, 265; and income, 239; measurement of, 191; MoeneWallerstein model, 233–37; and policy preferences, 184; and redistribution, 105, 107, 203–4, 205, 206, 213, 215, 217; in Sweden, 266; in U.K., 267; and wage inequality, 184–85; and welfare benefit generosity, 64 unemployment insurance, 71, 86n8 UNESCO, 223 union density: data sources, 348; definition of, 191; and disposable income inequality, 148, 155, 158; endogenous relationships analysis, 246–47; and income skew, 258; measurement of, 254; and partisanship, 183; and political polarization, 336; and social policies, 259; in Sweden, 341; in U.K., 338; in U.S., 338; and wage distribution, 147, 185; and wage inequality, 149–50 unions and unionization: in coordinated market economies, 139; and disposable income inequality, 148, 155, 156; members’ redistribution preference, 213; and partisanship, 115; and redistribution, 105, 321; role in politics of distribution, 138; in Sweden, 141; and wage coordination, 8, 143 United Kingdom: disposable income inequality, 28, 135, 306; elderly population, 158; endogenous rela-

Index tionships analysis, 266–67; household income inequality, 326, 344; income inequality and voter turnout, 294; income inequality trends, 42; labor force participation, 375; market income inequality, 133; net worth inequality, 403; party and electoral systems, 104; political mobilization, 331, 338; political parties, 329; political polarization, 342, 343–44; redistribution trends, 49–50; social spending trends, 72; union density, 338; voter turnout, 338; voting rights, 375; wage inequality, 130, 172, 326, 344; wealth inequality, 403; welfare benefit generosity, 70, 72; welfare spending, 64–65 United Nations, 50 United States: capital income vs. Germany, 129, 131; disposable income inequality, 28, 135, 306; household income inequality, 326; income inequality and voter turnout, 294, 295; income inequality trends, 40–42; market income inequality, 133; minimum wage, 144, 161n12; net worth inequality, 403; party and electoral systems, 104; political mobilization, 331, 336; political parties, 329; political polarization, 342, 343; poverty rate reduction from taxation and transfers, 93; redistribution in, 134; social spending trends, 72; Supplementary Security Insurance, 71; union density, 338; voter turnout, 293, 338; wage inequality, 130, 172, 326; wealth inequality, 403; welfare benefit generosity, 70, 72 U.S. Census Bureau, 41 Uusitalo, H., 45

433

van Deth, J., 304 Varian, H., 284 Verba, S., 281, 306n1 veto points, 79, 105, 109 Vindigni, A., 356, 379, 396 Visser, J., 105, 254, 348 vocational training systems, 204, 208, 222–23, 225, 228 Voitchovsky, S., 30 voter mobilization, 320–22, 330–31, 335–40, 347, 348. See also voter turnout voter turnout: and business cycle, 279; data sources, 348; and economic conditions, 279; factors in, 258; and income, 286; and income inequality, 283, 293, 321; and political polarization, 335–36; and redistribution, 105; in Sweden, 341, 344; in U.K., 338; in U.S., 293, 338 vote-seat disproportionalities, 113 voting participation. See electoral participation voting rights, 369–78, 379 wage-bargaining centralization, 177, 185–86, 190, 355, 356 wage coordination policies: absence of, 142–43; definition of, 79; and disposable income inequality, 156; example of, 7–8; of LMEs vs. CMEs, 138–39; in Sweden, 141, 142–43 wage dispersion, definition of, 79 wage floor hypothesis, 173, 175–76 wage inequality: cross-national comparison, 130, 171–73, 326; data sources, 348; estimation, 145–46, 148–52; factors in, 184–86; future of, 157–58, 189–90; and left-oriented political parties, 320; and market income, 147; measurement of, 324; and partisanship, 151–52,

434

Index

wage inequality (continued) 170, 173–75; and policy, 188; political issues, 169–70; and political polarization, 313–14, 334–40, 341, 346; tax impact, 180; trends, 169; in U.K., 344; and unions, 138; in U.S., 169 wage-rental ratio, 358 wages: definition of, 128; distribution for OECD countries, 129; efficiency wages, 161n11; government’s role in distribution, 136–37; labor market institutions’ role in distribution, 137; percentage of household income, 129; relative wages, 324, 325. See also minimum wage Wagner’s Law, 105, 107 Wallerstein, M., 161n14, 185–86, 233–37, 254, 391 Wascher, W., 191 Washer, W., 161n12 Way, C., 164n29, 173, 175, 189–90 wealth, and democratic institutions, 377, 378 wealth inequality, 402–3 Weber, M., 380 welfare benefit generosity index: limitations, 69; methodology, 66–69, 74; and redistributive outcomes, 76–83, 85; research considerations, 62; trends, 69–73; vs. welfare spending, 74–76 welfare programs, 179 welfare reform, 62

welfare spending: data sources and limitations, 62, 63–65; definition of, 79; and political parties, 6; research considerations, 62–63; trends, 72, 73–74, 85; vs. welfare benefit generosity, 74–76 West Germany, income inequality trends, 46–47 WFTC. See Working Family Tax Credit White, L., 364, 369 Wibbels, E., 397 Wiberg, M., 306n1 Williamson, J., 368, 382n31 winners and losers, 405–8 Woldendorp, J., 191 Wolfers, J., 221, 230 Wolfson, M., 43–44 women: labor force participation, 105, 107, 146, 148–49, 157–58, 185, 247–48, 254; redistribution preference, 212; voting rights, 372 Wood, A., 185 working-age households, 87n15 working class mobilization, 94, 138, 347 Working Family Tax Credit (WFTC), 65 World Bank, 27, 54n5, 147, 305 World Wars I and II, 368–69, 371, 372 Wren, A., 12–13, 185 youth, 258