INCOME INEQUALITY

Economic Disparities and the Middle Class in Affluent Countries

By Janet C. Gornick, Markus Jäntti

Stanford University Press

Copyright © 2013 Board of Trustees of the Leland Stanford Junior University
All rights reserved.
ISBN: 978-0-8047-7824-4

Contents

Foreword Anthony B. Atkinson……………………………………….xiiiAcknowledgments……………………………………………………xvContributors………………………………………………………xviiIntroduction Janet C. Gornick and Markus Jäntti………………………1PART I: Income: Trends in Household Income Inequality………………….CHAPTER ONE How Has Income Inequality Grown? The Reshaping of the Income
Distribution in LIS Countries Arthur S. Alderson and Kevin Doran……….51PART II: The Middle Class: The Middle Class in the Income Distribution…..CHAPTER TWO On the Identification of the Middle Class Anthony B.
Atkinson and Andrea Brandolini………………………………………77CHAPTER THREE Has Rising Inequality Reduced Middle-Class Income Growth?
Lane Kenworthy…………………………………………………….101CHAPTER FOUR Welfare Regimes, Cohorts, and the Middle Classes Louis
Chauvel…………………………………………………………..115PART III: Politics: Inequality, Political Behavior, and Public Opinion…..PART IV: Employment: Women’s Work, Inequality, and the Economic Status of
Families………………………………………………………….CHAPTER FIVE Political Sources of Government Redistribution in
High-Income Countries Vincent A. Mahler, David K. Jesuit, and Piotr R.
Paradowski………………………………………………………..145CHAPTER SIX Income Distribution, Inequality Perceptions, and
Redistributive Preferences in European Countries István György Tóth and
Tamás Keller………………………………………………………173CHAPTER SEVEN Women’s Employment and Household Income Inequality Susan
Harkness………………………………………………………….207CHAPTER EIGHT Women’s Employment, Unpaid Work, and Economic Inequality
Nancy Folbre, Janet C. Gornick, Helen Connolly, and Teresa Munzi………..234CHAPTER NINE Women’s Work, Family Earnings, and Public Policy Margarita
Estévez-Abe and Tanja Hethey-Maier…………………………………..261PART V: Wealth: The Distribution of Assets and Debt……………………CHAPTER TEN The Distribution of Assets and Debt Eva Sierminska, Timothy
M. Smeeding, and Serge Allegrezza……………………………………285CHAPTER ELEVEN The Joint Distribution of Income and Wealth Markus
Jäntti, Eva Sierminska, and Philippe Van Kerm…………………………312CHAPTER TWELVE The Fourth Retirement Pillar in Rich Countries Bruce
Bradbury………………………………………………………….334CHAPTER THIRTEEN Public Pension Entitlements and the Distribution of
Wealth Joachim R. Frick and Markus M. Grabka…………………………362PART VI: Country Case Studies: Inequality in Japan, Iceland, India, and
South Africa………………………………………………………CHAPTER FOURTEEN Income and Wealth Inequality in Japan Colin McKenzie….389CHAPTER FIFTEEN Income Inequality in Boom and Bust: A Tale from Iceland’s
Bubble Economy Stefán Ólafsson and Arnaldur Sölvi Kristjánsson…………416CHAPTER SIXTEEN Horizontal and Vertical Inequalities in India Reeve
Vanneman and Amaresh Dubey………………………………………….439CHAPTER SEVENTEEN Post-Apartheid Changes in South African Inequality
Murray Leibbrandt, Arden Finn, and Ingrid Woolard……………………..459Conclusion Janet C. Gornick and Markus Jäntti………………………..487Index…………………………………………………………….497

CHAPTER 1

How Has Income Inequality Grown?

The Reshaping of the Income Distribution inLIS Countries

Arthur S. Alderson and Kevin Doran

Research on trends in inequality has cumulated to the point that manyscholars conclude that income inequality has been growing in the typicalsociety in recent decades (e.g., Cornia and Addison 2003). While this”U-turn” on inequality has received a great deal of social scientific attentionin and of itself, it has also spawned growing concern over the statusof the middle class. The idea that the middle class might be shrinking ordeclining began to receive popular attention in the United States in theearly 1980s. That is, in the face of rising income inequality, concern abouta particular pattern of change in the distribution of income—polarization,or the shift of individuals, households, and families from the middle of thedistribution to both the upper and lower tails—began to grow. Noting thetransition from an industrial to a service-based economy, Kuttner (1983),for instance, argued that the decline of well-paying manufacturing jobsand lack of unionization in the growing service sector were driving manyformerly “middle-class” families into the bottom tail of the income distributionand were polarizing the workforce. In a similar vein, Thurow (1984)reported that between 1967 and 1982, the percentage of U.S. families withincomes between 75 and 125 percent of the median declined from 28.2 percentto 23.7 percent and that the percentage of families falling above and belowthose bounds grew by roughly equal proportions. Beyond the empirics,concern with income polarization has also been heavily informed by theclassic vision of the middle class as a bulwark of political democracy andagainst social, political, and economic instability (Lipset 1963; see Thurow1984 and Pressman 2007).

As highlighted by Pressman (2007), academic interest in the fate of themiddle class grew across the 1980s and into the 1990s, before decliningduring the economic expansion of the latter 1990s. In the context of theeconomic downturn in 2008, the issue once again moved to the center ofscholarly and public debate. How should one approach the question of the”hollowing of the middle”? It is clear that conclusions are sensitive to definitionsand measurement (McMahon and Tschetter 1986; Horrigan andHaugen 1988). For instance, McMahon and Tschetter (1986), replicatingconflicting studies by Lawrence (1984) and Rosenthal (1985) in the UnitedStates, demonstrate that, when defined in terms of occupation, no hollowingof the middle is seen from 1973 to 1985, but when it is defined in termsof personal earnings, there is ample evidence of polarization.

More recently, a rough consensus appears to be emerging in studies ofincome around a “literal” definition that focuses on the fate of the populationfalling within 75 and 125 percent of the median income (Pressman2007; Ravallion 2010). However, as a number of commentators have noted(e.g., Wolfson 1994; Jenkins 1995; Foster and Wolfson 2010; Atkinson andBrandolini [see Chapter 2]), this definition often seems to raise as manyquestions as it answers. Critics have expressed concern regarding the rationalebehind the cut points, the nature of the social groups encompassed(and not encompassed) by this income range, the implications of the definitionfor our understanding of groups above and below 75 to 125 percent ofthe median, and, ultimately, the sensitivity of the conclusions drawn in suchresearch to the bounds applied.

While debate over conceptualization and operationalization continues,a sketch of recent work on the hollowing of the middle is instructive forthe research we conduct in this chapter. While the scholarly debate initiallyfocused nearly entirely on the case of the United States—with some findingempirical support for the hollowing of the middle (e.g., Lawrence 1984;McMahon and Tschetter 1986; Horrigan and Haugen 1988; Davis andHuston 1992) and others finding just the opposite (e.g., Rosenthal 1985;Levy 1987; Kosters and Ross 1988)—some scholars have turned, more recently,to examine the question in a broader, cross-national framework. Forexample, using data from the Luxembourg Income Study (LIS) Databaseand employing the 75 to 125 percent criterion, Pressman (2007) examinesshifts in the size of the middle-income population in 11 countries, employingavailable data from between 1980 and 2000. Pressman notes that severalcountries experienced notable reductions in the proportion of their populationsfalling in the middle of the income distribution (e.g., Sweden, -7.1 percent;United Kingdom and Taiwan, -4.5 percent; Spain, -3.2 percent; andUnited States, -2.4 percent) and that, where this occurred, it proceeded viaa process of polarization, with households shifting into both the upper andlower tails of the income distribution. Pooling data for all 11 countries andweighting by population size, he finds, overall, that downgrading—movementfrom the middle to the lower tail—exceeded upgrading—movementfrom the middle to the upper tail—by about a factor of 2 to 1.2

ACCOUNTS OF CONTEMPORARY TRENDSIN WITHIN-NATION INEQUALITY

In accounting for the hollowing of the middle, scholars draw on a rangeof explanations for trends in income inequality. In other cross-nationalresearch and research on trends in inequality in U.S. states and counties,we have worked to integrate three literatures that have emerged aroundthis debate. Our aim is to combine attention to factors affecting the distributionof wages and earnings—which have tended to be the concernof economists—with a focus on a range of institutional, demographic,and compositional factors that both shape the aggregation of wages andearnings into the distribution of household and family income and affectinequality in ways that may be “independent” of the distributionof earnings. Given that we have reviewed these literatures in detail elsewhere(e.g., Moller, Alderson, and Nielsen 2009), we touch here on onlythe broadest outlines of each.

The first literature is centered in economics and takes as its object thesimple fact of rising inequality. A central hypothesis is that wage inequalityhas risen in many countries because of skill-biased technological change.Technological advancements have increased the demand for highly educatedand skilled workers; this demand has outpaced supply and createdscarcity rents for the highly skilled (Levy and Murname 1992; Autor, Katz,and Krueger 1998).

The second literature, which is typically oriented toward cross-nationalcomparison, takes as its object the persistent differences in levels of inequalityacross countries and regions and the heterogeneous inequality experienceof different countries and regions. Here one finds a diversity ofarguments regarding the role of labor market institutions (e.g., centralizedwage-setting, unionization), the process of globalization (e.g., internationaltrade, investment, and migration), and the wave of domestic and internationalliberalization (e.g., Alderson and Nielsen 2002; Moller et al. 2003;Kenworthy and Pontusson 2005; Brady 2009). This research has increasedto such an extent that some scholars have begun to outline a “unified theory”that would explain recent trends in wage inequality, real wages, andunemployment across developed countries. This theory attributes recentinequality trends to the interplay of exogenous shocks—affecting labor supplyand demand and the stability of earnings—with the marked differencesin the institutional contexts of different countries and regions (Blank 1998;Blau and Kahn 2002; DiPrete et al. 2006). In this perspective, for instance,the effects of skill-biased technological change on inequality might varysubstantially depending on the institutional context.

The third literature takes as its object household and family income inequality.While sharing many of the same concerns of the second (e.g., labormarket institutions, globalization, etc.), this literature distinctly focuses,for instance, on how socio-demographic factors—the age distribution ofthe population, the composition of households, assortative mating, racialand ethnic cleavages—generate inequality among households and familiesthat is independent of the distribution of wages and earnings (e.g., Cancianand Reed 2001; McCall 2001; Moller, Alderson, and Nielsen 2009; Blau,Ferber, and Winkler 2010). In the case of the United States, for example,while household and family income inequality rose measurably across the1970s, the upswing in earnings inequality did not take off until the 1980s.And during the 1980s and early 1990s (when earnings inequality was rising),change in earnings inequality explains only about one-third of thechange in family/household income inequality (Burtless 1999).

While research on the issues raised in these literatures has grown ata remarkable pace over the last two decades, we must consider that theseexplanations often imply very different patterns of distributional changewhile predicting the same outcome in terms of the behavior of standardsummary measures of inequality (e.g., a rise in the Gini coefficient or inthe Theil index). Morris, Bernhardt, and Handcock (1994, 206) notedthis problem at the very outset of the revival of social scientific interest inincome inequality, suggesting that “empirical investigation … has beenhandicapped by methods that are insensitive to [patterns of distributionalchange].” What is at issue? Consider, for instance, now-standard accountsof the effects of globalization on income inequality in the Global North.They often suggest that globalization is producing an increasingly polarizedjob distribution, a growing upper tier with high wages and security, a growingbottom tier in low-wage and insecure service positions, and a “shrinking”middle (e.g., Wood 1994). Other accounts, however, imply a ratherdifferent pattern of distributional change. The skill-biased technologicalchange explanation we mentioned previously suggests that inequality is risingas a result of upgrading—that is, growth in the upper tail of the distributionthat has simply left less skilled workers behind. Similarly, scholarswho emphasize the effects of the growth of autocatalytic, “winner-takeall”markets describe a process in which various technological and institutionalchanges have combined to produce an expanding number of marketsin which rewards are concentrated in the hands of a small number of “winners”(Frank and Cook 1995). While implying very different patterns ofdistributional change—and while these patterns have distinct implicationsfor both policy and distributive justice—these accounts of rising inequalityeffectively point to the same increase in summary inequality measures asprima facie evidence in support of their premises. As such—and in a worldin which high-quality, comparable data on income are scarce—we believeit is useful to occasionally look “behind” the usual summary measures andclosely examine the actual pattern of distributional change, attending tochange at all points on the distribution and fully exploiting the availableinformation. In short, it is important to examine how income inequality hasgrown.

In earlier research, we used the available high-quality data from the LISDatabase to examine the experiences of 16 high-income countries acrossthe period from the late 1960s to the late 1990s (Alderson, Beckfield, andNielsen 2005). We next expanded our investigation to include seven transitionaland middle-income countries (Alderson and Doran 2010). Most recently,we updated our earlier analyses, taking advantage of the latest waveof data from LIS centered around 2004, and began to look at households ininteresting social locations (e.g., female-headed households). In investigatingthe inequality experience of these countries, we seek to understand howinequality grew and to what extent the observed patterns of distributionalchange are heterogeneous or homogeneous. In this chapter, we address thequestion “Has the middle hollowed out?”

RELATIVE DISTRIBUTION METHODS AND DATA

To address these questions, we use relative distribution methods. Developedby Handcock and Morris (1999), methods based on the relative distributionpowerfully assist in the description of distributional change and enablecounterfactual comparison of compositionally adjusted distributions. Thebasic idea underlying the relative distribution is to take the values of onedistribution (the comparison distribution) and express them as positions inanother (the reference distribution). Consider two distributions of householdincome, one measured at t – 1 and one at t. Treat the t – 1 distributionas the reference distribution and t as the comparison distribution. Whenthere are no differences between the comparison and reference distributions,the relative distribution of the grade-transformed data will be uniformor “flat”—that is, the proportion of households falling within a givenquantile’s cut points of the reference distribution t – 1 at t is the same as att – 1. When there are differences between comparison and reference distributions,the relative distribution will “rise” or “fall”—that is, the proportionof households falling within a given quantile’s cut points of thereference distribution t – 1 at t will be larger or smaller than at t – 1. Inthis fashion, one can distinguish among growth, stability, and decline at allpoints on the distribution.

Another advantageous feature of these methods is that one can usethe relative data to develop summary measures to characterize any patternof change that one might be interested in exploring. Handcock and Morris(1999) developed a measure of polarization that captures the degree towhich there is divergence from, or convergence toward, the center of thedistribution and is thus ideally suited to addressing the question of the “hollowingof the middle.” For quantile data Q, the median relative polarizationindex (MRP) takes the following form (Morris, Bernhardt, and Handcock1994, 217):

MRP_t(Q) = 4/[Q – 2]|[i – 1/2]/Q – [1/2]| × (g_t(i) – Q/[Q – 2]),

where g_t(i) is the relative distribution—the proportion of year t‘s householdswhose median-adjusted incomes fall between each pair of quantilecut points, divided by the proportion in the reference year i = 1, 2, …, Q,and the adjustment by 1/2 establishes the mid-point for each quantile. Theindex varies between -1 and 1. It takes the value of 0 when there has beenno change in the distribution of household income relative to the referenceyear. Positive values signify relative polarization (i.e., growth in the tails ofthe distribution), and negative values signify relative convergence towardthe center of the distribution (i.e., less polarization).

The median relative polarization index can be decomposed into thecontributions to distributional change made by the segments of the distributionabove and below the median (Handcock and Morris 1999), enablingone to distinguish “upgrading” from “downgrading.” For quantile data,the lower relative polarization index (LRP) and the upper relative polarizationindex (URP) are calculated as follows:

LRP_t/URP_t(Q) = 8/[Q – 2]|[i – 1/2]/Q – [1/2]|×( g_t(i) – Q/[Q – 2]).

They have the same theoretical range as the MRP and decompose theoverall polarization index (Morris, Bernhardt, and Handcock 1994, 209):MRP_t = LRP_t/2 + URP_t/2.

We apply these techniques to data drawn from the LIS Database(2010). For each country/year used in the analysis, we generate quantileboundaries for the distribution of household income (equivalent netdisposable household income), adjusting for household size using a standardequivalence scale (i.e., the square root of the number of persons inthe household).
(Continues…)Excerpted from INCOME INEQUALITY by Janet C. Gornick, Markus Jäntti. Copyright © 2013 Board of Trustees of the Leland Stanford Junior University. Excerpted by permission of Stanford University Press.
All rights reserved. No part of this excerpt may be reproduced or reprinted without permission in writing from the publisher.
Excerpts are provided by Dial-A-Book Inc. solely for the personal use of visitors to this web site.