Who gets what and why? Government income redistribution in times of economic crisis

Transcription

1 Who gets what and why? Government income redistribution in times of economic crisis Ursula Dallinger, University of Trier, Germany Comments welcome: Abstract Over the last few decades government income redistribution cushioned the growing disparity in the distribution of market incomes in many highly industrialized countries. The redistributive efforts vary from country to country and with time, thus raising the question of what forces determine these market income adjustments. Whether redistributive efforts react to growing inequality and to demands by voting blocs harmed by economic change, or whether they are determined primarily by political factors has long been a subject of controversy. This controversy in comparative political economy is addressed in this paper. It focuses on distribution coalitions, groups constituted from lower and middle classes that have a great deal of electoral power and play a central role in decisions about who gets what and why. How political or socio-political institutions influence middle class coalition building has been analyzed, but the present paper also introduces the effect of income structures to ask if income differences favor certain kinds of distribution coalitions. The analysis used microdata from the Luxembourg Income Study to calculate indices both for income distribution by income groups (quintile shares) and for the effect of government redistribution on individual groups (share gains). These are supplemented with other theoretically relevant variables to yield a pooled cross-sectional time series of 81 observations in 19 countries for the period 1980 to The hypothesized influence of a changed income distribution and other attributes that vary over time is tested with fixed effects regression. The effects of time-invariant variables indicating political institution attributes are tested by means of OLS regressions for pooled data. 1 Introduction Various factors either drove or compensated for the growing gaps in income distribution during the past decades in most developed societies. Income distribution was variously influenced by skill-based technological change, changes in how wages are set or increasing participation by women in the workforce. This paper analyzes how income distribution between 1980 and 2005 was affected by government redistribution. Social transfers and taxes compensated at least in part for the growing disparity in market incomes over the past years (Kenworthy 2004; Ferrarini 2006; Pontusson 2006: 32). It will be demonstrated that while redistribution appears to have increased, it did not catch up with the dynamic of diverging market income distribution. What explains this trend? What explains the variation between countries of the extent to which they resort to welfare state redistribution? Page 1 of 25

2 Sociology has explored the determinants of the intertemporal and international scope of governmental redistribution to a much lesser degree than educational or wage inequality. This paper is intended to improve the understanding of governmental interventions in the distribution of factor or gross incomes, and it enquires into the social and political forces that determine who gets what and why? (Lasswell 1936; Lenski 1966). According to the classic, if controversial, median voter model for explaining redistribution, social policies can be expected to react to growing market income inequalities with more vigorous redistribution, because voters who benefit from redistributive programs demand equalizing policies and are in a position in democratic countries to impose them by majority vote. However, this type of relationship between inequality and redistribution could be established empirically only in part (Alesina et.al. 2001; Bowles/Gintis 2000; Kenworthy/Pontusson 2005; Perotti 1996). More recent research into redistribution has helped fill in the model s gaps. One avenue of research looked at how demands for redistribution per se came into being against a backdrop of cultural norms, expectations of mobility or social comparisons. In another, the focus was on the political transformation of voter demands into actual redistribution policies based on political power relationships and (socio) political institutions. Ultimately, it is through these political forces and institutions that voter demands are transformed into redistributive policies. That class coalitions determine the scope of redistribution is a matter of consensus. Those who benefit from corrections in the distribution of market incomes can impose their demands only by building coalitions with the middle class, which has the power of decisively influencing electoral outcomes. Specific class coalitions are encouraged by the political system s institutions (electoral rules) as well as welfare state institutions (Korpi/Palme 2008; Iversen/Soskice 2008). Political factors that tend to make it more likely for the middle class to coalesce with the lower or, as the case may be, with the higher social classes, explain the variations in redistribution scope accordingly. Current approaches thus contend that who gets what and why is determined by political coalitions of the middle class and political as well as socio-political institutions that stimulate these coalitions. Lupu and Pontusson (2011) gave primacy to the political orientation of the middle classes in conflicts over distribution and expanded it with the concept of social distances or affinities. They posit that the relative position of individual income classes within inequality structures influences the extent to which they will support redistribution. Social affinity with lower income classes would promote, while affinity with the upper classes would shrink, the likelihood of a common interest in redistribution ending in relevant political coalitions. Because it reformulates pre-political inequalities anew, we adopt the social distance concept. According to the median voter theorem is must be assumed that it was increasing inequality in the past that stimulated redistribution demands. In addition, in line with the debate over the effects of globalization (Breen 1997), insecurity and the risks attendant on open markets boost public demands for social protection and social equality that lead to redistribution. Crises are likely to result in the same response: citizens no longer look to the economy as the primary source for producing wellbeing and instead demand governmental interventions that are implemented according to redistribution theory. Hence, the working hypothesis is that the gaps between lower, middle and upper income tiers and how they change have either a positive or a negative effect on the scope Page 2 of 25

3 of redistribution. The effect of objective income gaps is mediated by political power relationships and institutions. 1 To describe who gets what and analyze which factors determine distribution structures requires a distribution metric that captures the distribution positions of individual income groups and how government redistribution changes them. The Gini Index, relied on in the majority of redistribution research, is not suited for this. Hence, this paper uses (quintile shares of total income). Redistribution is equal to the difference in shares before and after taxes and social transfers. These share gains were calculated using the Luxembourg Income Study (LIS) microdata. The pooled panel data set built with this data covers the period between 1980 and Fixed effects regressions are used to analyze the effect of temporal changes; the effects of time-invariant variables are analyzed with OLS regressions. Following a review of explanations for government income redistribution (Ch. 2.1) and the potential effects of crises (2.2), the hypotheses are stated precisely in (2.3). Ch. 3 presents the data sources and methods. The fourth chapter first describes how the positions of the income groups evolved over the past decades and how they were affected by social welfare redistribution. This is followed by multivariate analysis of influences on the welfare state s distributive effects. The final chapter summarizes the findings. 2. Welfare state redistribution explanations 2.1 Redistribution and democracy The median voter approach developed by Meltzer and Richards (1981) and Roemer (1975) represents the basic political economy model for explaining socio-political intervention in wage and salary distribution. It holds that democracies are able to correct market inequality in response to the expressed preferences of electoral majorities. When disparities in wage and income distribution and progressive taxation grow, those receiving incomes at or below the median can be expected to view redistribution as advantageous and to make appropriate political demands. As the arithmetic mean deviates increasingly from the median with increasing inequality, the group benefiting from redistribution will expand. In democracies, this group has the opportunity to decide election outcomes and to enforce redistributive demands. Should income inequality continue to increase, a growing majority would automatically demand redistribution measures to correct income disparities. These demands would abate when redistribution programs move the median voter closer to higher incomes but also expose him to an increased tax burden. This model was criticized on several grounds. First, because the relationship between inequality and the scope of redistribution efforts is far from being unequivocally validated empirically: there are countries with pronounced income disparities that have a less developed welfare state and less redistribution than countries that exhibit little inequality. Moreover, by no means do citizens of countries with higher wage disparities support government redistribution any more than do citizens of egalitarian countries. 3 The U.S., 1 In addition, job insecurity, mobility expectations and political orientation mediate demands favoring redistribution. The paper cannot address this subjective dimension, since there is not enough fit between survey data and the objective income data. Nonetheless, there is clear evidence that individual positioning within income structures, political orientation and electoral choices coincide, as do their socio-political preferences and support for redistribution (Svallfors 1998; 2004). 2 Not all countries have data available for the entire period. 3 Moene, K.O./ Wallerstein, M. (2001); Iversen, T./Soskice, D. (2006). Page 3 of 25

4 whose low redistribution compared with Europe is explained by American exceptionalism (Alesina et.al 2001; Alesina/Glaeser 2004), provides empirical evidence contradicting the median voter model. The reasons why American voters do not demand more redistribution and push it through in spite of the great income inequality is ascribed to the lack of unions, high immigration numbers and political institutions. It is true that country differences contradict the median voter theory. When using longitudinal data the postulated positive relationship between market inequality and redistribution clearly emerges. Redistribution s scope increases as wage inequality increases (Milanovic 2000; Kenworthy/Pontusson 2005; Kenworthy/McCall 2008). Furthermore, the relationship between increasing inequality and widespread public support for government redistribution clearly emerges if modeled as a non-linear effect (Dallinger 2010; Finseraas 2010; Jaeger 2011). But analysis using longitudinal data with fixed effects regressions uses only the temporal variation in the data while excluding the differing cross-national levels as well as the effect of country-specific institutions. Causal analysis using longitudinal data is more reliable 4, but at the same time, it is insensitive to the impact of time-invariant country attributes. 2.2 Politics and the middle class The median voter model was criticized for its simplistic modeling of the political process. In order to be transformed into organized politics and to have an effect, however, voter preferences do have to always filter through political decisions and institutions. According to power resource theory, leftist parties and unions speak for the interests of workers and low-income earners who benefit from redistribution. Social democratic parties make relevant demands in their electoral platforms and implement them to the extent their power resources permit made operative by the two means which let parties influence legislation: the number of seats held by parties of the left in parliament or in a government constituted from the party in question. The extent of unionization also affects wage distribution and forces governments to take into account workers interests with respect to income distribution. (Korpi 1983, 1989; Hicks/Swank 1992; Huber/Stephens 2001; Beramendi/Cusack 2008). Early socio-political research studies used the politically organized power of social groups to explain the various forms taken on by the welfare state and at what junctures in time it arose. Interest communities and ideological repertoires of different dominant social groups laid the cornerstones in the industrializing countries for liberal, social democratic or conservative types of welfare states or regimes. The thesis that the working class was dependent on forming coalitions with other social groups because of its limited power served to expand the power resource approach. The labor parties respective coalition partners required differing compromises concerning labor market regulation and social protection, thus accounting for the differences in the degree of stratification and decommodification of regimes (Goodin/LeGrand 1987; Esping-Andersen 1990; Baldwin 1990; Korpi/Palme 1998; Scruggs/Alan 2006; Svallfors 2004; Manow 2007). 5 Because of coalition building, left-wing parties by themselves do not set the implemented policy, even when they form part of the government. The leftist parties potential 4 In that they assess the intra-country variation and avoid the cross-section design s chronic problem of potential distortions of the results due to omitted variable bias. 5 However, the influence of leftist parties is subject to debate. It has been said that parties lost their meaning in the face of pressures to act that open economies and demographic changes bring (Swank 2002). Page 4 of 25

5 influence is further weakened by party polarization and other issues that in classic fashion deter the voters addressed by the left from voting in their own interest. The voting public can be split to such an extent that groups satisfied with the prevailing unequal distribution vote for parties that guard their economic interest, while those who benefit from higher taxes instead cast their votes based on non-economic issues (Burtless/Jencks 2003; Finseraas 2010). Moreover, certain veto barriers define power resources and legislative influence in a political system (Iversen/Soskice 2008). Voter participation determines how much influence parties with redistributive programs will exert. The less the lower strata participate in election, the lower this influence will be; it lets politicians pay less attention to the concerns of income redistribution beneficiaries on whose votes they depend. The extent to which redistribution and election turnouts correlate has been empirically verified. (Pontusson/Kenworthy 2005; Mahler/Jesuit 2006, Mahler et.al. 2010). Beyond that, the welfare state itself influences voter participation. While most European countries encourage social participation by granting citizens wide-ranging social rights, the U.S. implements social policies offering marginal security, which reduces political participation. Voters in the U.S., come from higher strata than in European countries because voter participation varies by income and social strata (Alber/Kohler 2007). The party coalition hypothesis was expanded by institutional theory. It holds that political institutions incentivize coalition building. (Iversen/Soskice 2006 und 2008; Beramendi/Anderson 2008). Put simply, individual parties represent the interests of the lower, middle and upper classes. The middle class (party) can form coalitions with both the upper and the lower class. The middle class is interested in coalition building with the upper class or rightist parties out of fear of being controlled by the lower class. At the same time, however, the middle has an incentive for coalition building with the poor or left wing parties out of a mutual desire to exploit the upper class. Which of the two coalition types the middle opts for will determine if a redistribution-friendly or -limiting government forms. Where a country s constitution provides for a system of proportional representation (PR) it favors middle-left coalitions. Since all parties participate in the government based on their share of the vote, parties are apt to keep their promises to the group that they represent. Since middle class voters accordingly can be sure that their party will implement their votes in parliament, they then support coalitions with leftist parties. Together with them, the middle class can push through policies offering more resources for redistribution at the expense of the upper class. In a majority system, by contrast, the party that gets the most votes forms the government. Parties offer broad compromise platforms oriented either to the middle-left or middle-right in order to win a majority. For the middle, this carries the risk that parties with a middle-left platform will drift left after the election, raise taxes and disproportionately skew collective goods toward the bottom. The middle class would prefer to choose a middle-right platform in order to avoid this. The majority system therefore creates incentives for center-right coalitions because it means lower taxes for the middle class. Countries with christian-democratic, class-spanning organized parties, however, do not fit into this categorization. Comparative socio-political research also explains coalitions of lower and middle classes through institutionally created incentives, which here are socio-political in na- Page 5 of 25

6 ture, however. They are said to shape the perception of interests 6 and the nature of social relationships. The varying scope of redistribution is explained by the different incentives that the marginal or universal type of welfare state provides. The marginal type, which directs minimum security selectively at poorer persons, divides the welfare clientele from those who finance the benefits. By contrast, the universal type s social insurance garners widespread acceptance by insuring against risks to everyone s standard of living posed by old age, illness and job loss. In this system, where everyone is both financier and potential benefit recipient, the resistance to deductions is low and tax base is broad. In addition, social insurance schemes like these create better-performing risk pools, since they aggregate heterogeneous groups with different risk profiles. 7 Finally, the scope of governmental redistribution is explained by the middle class security interests. Institutions that avoid a gulf between recipients and those who finance them, promote coalitions between workers and middle class and stabilize the universal welfare state (Rothstein 1998; Korpi/Palme 1998; Brooks/Manza 2006; Palme 2006). 2.3 Inequality structures and social affinity Redistribution coalitions not only can be traced back to political institutional incentives, but also to the proximity of social groups in the income hierarchy. Lower, middle and upper classes 8 exhibited similar political behavior and preferences based on the burdens and benefits of redistribution. Here, the social distance between the haves and havenots varying internationally and inter-temporally is of central importance for redistribution demands and coalitions. (Shapiro 2002; Schwabish et.al. 2004; Osberg et.al. 2004). A large gap between middle and upper class lets social expenditures (non-elderly only, proxy for redistribution) decline demonstrably; large income gaps between lower and middle classes see demands for redistribution and hence social expenditures rise. Growing economic distance caused commitment by the upper income group to weaken for social security systems perceived as lacking adequate intrinsic utility (Schwabish et.al. 2004: 5). The greater the gap between higher incomes and the rest of society becomes the more those who are well off distance themselves from the financing of social collective goods. Important insights emerge from the political economy of voter behavior about the specific position of the middle-income groups and their orientation in redistribution coalitions. Here we find an explanation for participation in pressure groups for or against public spending (Kristov et.al. 1992), which may transmute into support for redistributive parties. Taxpayers on the one hand, and transfer payment recipients on the other, would be expected to assign themselves to distinctly different pressure groups. It is critical how the intermediate group, which neither depends to any significant degree on transfers nor is burdened to the same extent by taxes as the highest income groups, orients itself. The decision to join a pressure group depends on the costs incurred by 6 This refers to the interest perceptions of the actors that specific institutional arrangements promote and that call for government redistribution in case of growing inequality, not to a median voter s autonomous self-interest. 7 In this way, despite lapses of focus on the needy, universal social insurance achieved a more egalitarian distribution than countries with targeted social programs for the poor. The latter exhibit higher poverty and inequality rates, which the concept of the redistribution paradox grapples with (Korpi/Palme 1998; Iversen/ Soskice 2004). 8 The division into the poor, intermediate group and the rich, or transfer payees and taxpayers simplifies social structures into idealized types in order to develop assumptions that can be tested empirically. Page 6 of 25

7 political participation. The poor lack resources for taking action, while middle class and higher actors can become active. Voters who have diffuse expectations concerning the costs vs. the utility of redistribution, i.e. those belonging to the middle class, occupy a neutral position that will only shift under additional conditions: the middle class will choose a pressure group favoring the lower class if it itself expects insecurity and decline during crises. Exchanging places with the poor becomes that much more likely, the more distinct the gap with the wealthy is. Should prosperity instead offer prospects for their own economic ascent, the middle class expects to change places with the upper class. Therefore, expectations for one s own social mobility create affinity with groups that have a clear political position. Economic crises ought to boost the middle s motivation to protect itself; growth should have the opposite effect and diminish it. 9 Lupu and Pontussan (2011) empirically verified a similar social affinity hypothesis: government redistribution scope is the greater the farther apart the middle and upper income decile are and the closer the lower half s groups line up. When there is a small distance between middle and lower incomes, it therefore appears that the electiondeciding, middle-income tier will support parties advocating redistributionist programs Government redistribution, globalization and crises Current prevailing opinion has it that globalization raises competitive pressure on national economies and especially on wage costs, which are not competitive due to the high social security contributions. A comprehensive welfare state is thus thought to impair the economy s capacity for meeting the challenges posed by an open globalized economy. Deregulation and liberalization of free markets are also said to compel dismantling of the welfare state and of redistribution to the poor. The social sciences objected to this view early on. To the contrary, they argue, globalization causes a rise in the demand for social security by a public that is progressively exposed to global risks as enterprises increasingly pass them on to their employees (Breen 1997). Globalization and the welfare state go hand in hand, for social programs bear the costs of economic change and unemployment and compensate for economic risks (Garrett 1998). Under conditions of open economies, social interventions into a risky free market gain even greater political legitimacy. This also explains why social expenditures in the previous decades marked by growing globalization did not decrease, but were instead stable or even increased. Contrary to expectations of cuts in social security expenditures, a resurgence of the welfare state took place, with the share of GDP that governments devoted to social expenditures actually expanding during The welfare effort in open, post-industrial economies is robust because of the need for a social safety net. The literature on the effect that globalization exerts on the welfare state underlines the significance of governmental reactions to economic pressures. These are similar everywhere, but the adaption to international market forces obviously varies from country to country. Thus, globalization is channeled by specific domestic responses by govern- 9 The more others reflect one s own traits with regard to income, race, religion etc., the stronger affinity is moreover. Interest in others wellbeing develops most easily toward persons with whom we identify. 10 Lupu/Pontusson also investigate if redistribution preference mediate between inequality and redistribution, for they determine governmental redistribution. A positive, significant bivariate correlation between voter preferences and redistribution appears to confirm the causal mechanism. Since survey data on preferences and income redistribution intersect little, only small samples can be generated, unfortunately. Page 7 of 25

8 ments. Political elites choose a variety of economic and social policies that prevent the costs of change from causing political eruptions. (Katzenstein 1985: 24) Some maintain that the overall size of the welfare state and the extent of redistribution remain a matter of political choice even under economic constraints (Rodrik 1997; Mahler 2004; Brady et.al. 2005). Especially the small European states adapted their economies to open dynamic markets and at the same time pursued a strategy of domestic compensation. Economic openness and dependence established a compelling need for consensus, which through complex and delicate political arrangements has transformed conflict among the main social forces in small European states. (Katzenstein 1985: 34) Social policy programs compensate for the individual and social costs of open markets that affect the individual (Rieger/Leibfried 1998). Domestic compensation approaches expect national factors like political forces or the aging population to drive the expansion of government redistribution. Internal prowelfare constituencies are expected to push governments into pursuing courses of action preferred by voters. The political power of the voters certainly depends on political actors carrying the demands of the public into the political arena, on institutional characteristics of the political system and on electoral turnout (Kenworthy/Pontusson 2005). What was ascribed to globalization also applies to developments during economic crises: national institutions filter them as well. The voting public is particularly impacted by crisis through job losses or negative employment trends. Under such circumstances, the public s welfare expectations turn from the economy to more emphasis on the government (Dallinger 2009). The public s political demands for a redistributionist policy grow. Politicians who want to stay in office have to respond with relevant policies. (Pierson 2007; Cusack et al. 2008). Domestic political factors are evident in the proportion of left-wing voters, in how many seats leftist parties hold in legislative bodies or which parties form governments. Domestic characteristics, such as the extent of union membership or the strength of labor market regulation, were also used to explain the variation in social expenditures and redistribution scope (Cameron 1978; Garrett 1998 a, b; Mahler 2004). 11 Globalization and interest in redistribution should be mediated via the uncertainty of the individual s unemployment and future income along with it. Persons in occupations with a high prevailing unemployment rate anticipate a commensurately high risk of losing their jobs. They support a social safety net more strongly than those who have jobs in which unemployment is low. (Rehm 2009, Kitschelt/Rehm 2006). 12 Hypotheses First, this paper examines the hypothesis that the scope of redistribution is determined by the relative positions of specific groups in the income distribution. This takes up the median voter hypothesis, which implies that politics will react to growing inequality and to the demands and interests of those negatively affected by it. It calls for studying the relevance of inequality structures or income gaps. However, the hypothesis is ambiguous: a) according to Lupu/Pontusson, a slight distance between middle and down below 11 Whether left and right partisanship still is important for changes in social policies is an ongoing debate. While some authors find that today both left and right governments support social expenditures to gain favor with the public (Mahler 2004), others have found that right-leaning parties are still negative for social security whereas left-leaning ones are still positively correlated with government welfare efforts. 12 The paper foregoes further analysis of the insurance motive and the different ways it expresses itself in the context of various production regimes, since findings on the subject exist (Iversen/Soskice 2001). Page 8 of 25

9 produces the social affinity necessary for joint redistribution coalitions; the middle would have to decline as well in that case. The smaller the distance between the first and the third quintile is the more Q1 benefits from redistribution. b) when the lowest quintile declines, its demands lead to government redistribution that prevent the decline. Political power relationships and institutional incentives, not inequality trends, are decisive for explaining varying kinds of redistribution. Hypothesis 2 a) states that political power determines what the results of redistribution will be. The influences of leftist parties and unions, as well as higher voter participation, are determinants of redistribution downward. Hypothesis 2 b) similarly targets political factors, but focuses instead on the institutions: the proportional electoral system and widely diverging types of welfare states that hold appeal for the middle, too, exert a positive influence on downward redistribution. 2 c) The assumption of inequality-induced redistribution in the first hypothesis loses its effect in view of the real power exerted by political factors. The domestic demand approach also assumes that political filters primarily decide what the reaction to globalization and crisis-ridden processes ought to be. Analogous to the above hypotheses, a third one tests, with respect to globalization effects, if continuing (longitudinal perspective) or greater economic openness (country comparison) alone suffice to increase redistribution or if relevant political forces and institutions are the primary influencers. Risk heightened by economic openness is felt by the public in the form of job losses. Hypothesis 4 examines if high unemployment promotes the desire to selfinsure as a precursor to government redistribution. 3 Methods Variables for income distribution and redistribution were calculated using the microdata of the Luxembourg Income Study (LIS) (see below). 13 Available dataset waves I through VI allow observing the period from 1980 to The time intervals between measurements not being identical for all countries, and varying numbers of waves being entered per country, the result is an unbalanced, pooled cross-sectional time series data set consisting of 88 observations for 19 countries, at 2 to 6 observations per country. Other variables come from various data sources (see below). Where values are not available in an unbalanced panel, they are treated as missing data. These gaps are acceptable so long as they do not systematically affect the redistribution variables to be studied. The missing data points must be random. There is no indication of systematic losses in the listing of available waves and countries shown in Table 1 of the Appendix. Income distribution is stated as the percentile share of total income received in each case by a specific percentage of the population. 14 Measuring the income distribution in this manner relies on the fact that deviations of the percentile share from the population group s size here we have chosen fifths, i.e. quintiles express inequality. Proceeding in this way depicts distribution structures and the position of individual groups, which the Gini Index does not do. 13 LIS sources the data from representative, national microdata sets, then recodes them using consistent standards to makes them comparable internationally. 14 Households are sorted according to income size into five equally-large groups, so that each quintile represents 20% of the survey subjects. The share of total income received by each group can then be determined. Page 9 of 25

10 It is possible, moreover, to determine government redistribution differentiated for individual income groups by comparing the quintile shares once on the basis of gross income, and then once more on the basis of household disposable net income after social transfers and taxes. Redistribution effects can be positive (share gain) or negative (share loss). Redistribution down to the first and second quintile is used later on as the dependent variable. Redistribution is calculated for the working-age population. People younger than 20 and older than 60 years are excluded to avoid distortions (Iversen/Soskice 2006; Bradley et.al. 2003; Lupu/Pontusson 2011). They emerge because elderly are retirees with little or no market income, unless they have income from rentals, leases, self-employment or capital markets. When computing redistribution as the difference between market and disposable income, in the case of the elderly, it results in a jump from zero or little market income to net income from government service transfers. To define this jump as redistribution just would capture a methodological artifact and the degree to which old age security depends on private or public provision. Analyses that include the elderly risk to interpret a pensioner effect as redistribution instead (see Pamp 2011). Predictor variables Quintile shares allow building ratio metrics, such as the ratio showing the relationship between the income shares of the tenth and first deciles (D9/1). This paper relies on the development (and cross-national variation) of ratios that portray the income disparities between income quintiles at the market income level. For instance, the distribution of market-generated inequality is a predictive variable for redistribution. Unemployment is kept as a control variable in all models. This prevents concluding that there is more redistribution ( automatic compensation, Kenworthy/Pontusson 2005) because of the higher number of persons without market income. Unemployment is often treated in the literature as an indicator of the impact on the public of economic crises. As we will see, however, unemployment is barely significant, so that we can concentrate more on variables that mediate domestic demands. Political power is taken into account with variables that depict partisan influence in parliament since that is where legislation is crafted. The Govparty variable from the Comparative Political Data Set (Armingeon et al. 2010) uses the following scheme to represent the cabinet make up (Schmidt Index): 1: hegemony of right and center parties; 2: domination by right and center parties; 3: balance between right and left parties; 4: domination by social-democratic and other leftist parties; 5: hegemony of socialdemocratic and other leftist parties. Percentile voter participation during the most recent parliamentary election gives an indicator of how political influence is brought to bear. Union power is captured by the percentage of union members in the workforce. Since unions also formulate socio-political objectives and do not just agitate in the area of wage setting, it makes sense to consider their influence. Election participation is The influence of electoral systems and social welfare regimes on coalitions between middle and lower classes, assumed in institution-theoretical models, is controlled for by dummy variables for countries with proportional electoral systems and for a single universal welfare state model (Sweden, Norway, Finland and Denmark). The impact of income disparities on the scope of redistribution and other independent, time-variant variables is tested by means of fixed effects regressions. These only use the temporal variation in the data and exclude the differences between countries. They supply consistent estimators, even when the individual effects correlate with other (un- Page 10 of 25

11 observed) variables. The risk inherent in cross-section designs that unobserved political or cultural country attributes would influence the coefficients (omitted variables bias) can therefore not materialize. Fixed effects models with longitudinal data prevent erroneously interpreting the effects stemming from cross-national variation as temporal causal effects (Kenworthy 2009; Meier Jaeger 2011). Explanations of redistribution also rest on time-constant institutional attributes of countries, which fixed effects procedures do not take into account. Country differences are tested with OLS regression of pooled data. This allows analyzing institutional characteristics and accounting for the fact that cross-national variation is greater than the intertemporal variation within individual countries Findings 4.1 Who gets what? This section examines the distribution of both market income and disposable income for the population quintiles, the effect of redistribution and its development over time. Who gets what; and who benefits from government redistribution? We look first at the shares each quintile obtains from market income, calculated as an average across all waves. As Figure 1 shows, there is a similar structure in every country a broad gap between the highest quintile and all other quintiles and smaller differences between the bottom and the middle. If we take the example of Germany, the lower middle class gets about 12% of the market income, the middle middle class about 16% and the upper middle class about 22%. The bottom quintile must make do with only 7% of the market income, while the top quintile gets 43%. This structure varies between countries, as we can see from the differing horizontal positions of dots on the individual country lines. Figure 1, which sorts countries by size of the Q3 market income share, indicates that the middle in Scandinavian countries enjoys the best income position, with the highest quintile not pulling away from the middle as strongly as in other countries. However, the narrow gap between the shares of market income that the middle and the top receive does not necessarily benefit the lowest quintile; in Denmark and Sweden, in particular, this quintile does not profit from the closeness between middle and top. From a comparative point of view, the market income share of the middle middle is worst in France, the Netherlands, Germany and the U.S. Among continental European countries, the unfavorable position of this group contrasts with a better position of the lower middle and the lowest quintiles. In some liberal countries, notably Canada and Australia, the middle middle income group occupies a relatively good position when compared to continental European countries. However, in all the liberal countries examined here, the lowest quintile occupies a more unfavorable position. Hence, the lower middle and the lowest quintile are relatively better off in continental European countries. Since these results are based on market income, this is likely due to the regulation of labor relations and coordinated wage bargaining. By comparison, the nature of labor relations in liberal countries leads to a distribution that tends to benefit the middle middle and upper middle classes. The upper middle class obtains a larger share of the market income in these countries, while the lower middle class has a relatively smaller share. 15 According to an ANOVA test, 67% of the variance is at the country level, 32% of variance on the time level. Page 11 of 25

12 Figure 1: Quintile shares of market income, average over all waves (countries sorted by size of Q3) About here Figure 2: Quintile shares of disposable income, average over all waves (countries sorted by size of Q3) About here When the same measure is calculated based on net disposable household income, the influence of the welfare state becomes apparent. As shown in Figure 2, government redistribution results in a gain for the lower income quintiles and a loss to the top quintile. The degree to which this takes place varies country by country, of course. Regarding the group this article focuses on, it is remarkable how little the income share of the middle classes change when we move from market income to disposable income. In terms of disposable income, the Scandinavian countries remain the most advantageous for the middle middle class; after redistribution, the lowest quintile moves considerably closer to the middle. Germany is among the bottom third in terms of the income position of the middle middle, though the lowest quintile does relatively better. In terms of the income distribution that reflects welfare state measures, Germany, along with France, the Netherlands and Spain, range even behind the U.S. and the UK with respect to the position of the middle middle quintile. However, as was true of the market income distribution, the lower middle class and the lowest class quintiles are better off in continental European countries than in liberal countries. Here, the middle holds a more advantageous position, though the lowest quintile receives a relatively disadvantaged income share compared to other countries. Nevertheless, we cannot generalize and there are exceptions among both liberal and continental groups, especially among the former. For example, both Canada and Australia provide more equality for the middle meaning less distance from the top than do the countries of continental Europe. Which income class benefits from government redistribution? Redistribution s impact can be seen in Fig. 3, which shows the difference in income shares received by the individual quintiles before and after taxes. For each country, the five bars show the scope of redistribution in each of the quintiles, for the mid-1980s in the first set, for the mid-2000s in the second set, in order to assess a changing influence of the welfare state on the income distribution over time. In all countries, and for both periods, government redistribution increases the income share of the first quintile and decreases the share of the fifth quintile. The share of wealth the quintiles in the middle receive changes little by comparison, when comparing pre-tax to post tax/transfer income. Relative to market income, the share of the lower middle and the middle middle classes increase slightly in most of the countries, while the share of the upper middle class decreases. To sum up, welfare state interventions provide a larger volume of income for the lowest class, and reduce the volume at the top. Figure 3: Change in redistribution by the welfare state. About here The variation over time in how strongly the welfare state influences the income distribution is instructive. Between 1985 and 2005, the income share of the lowest income class expanded through redistribution, except in the UK. The amount by which government intervention reduced the income share of the top quintile grew during the same period again, except for the UK. In short, over these two decades, the effect of welfare state- Page 12 of 25

13 induced redistribution increased in advanced industrial nations (see also Mahler/Jesuit 2006). Its impact on the middle classes is much less significant than for groups at the top and the bottom of a given society. It is true that redistribution to the lower middle as well as to the lowest class is of growing importance in most countries. Only in a few countries (U.S., Sweden, Poland, and Luxembourg) did the extent of redistribution become more significant for the middle middle between 1985 and Government redistribution without compensation of market income inequality That the welfare state increasingly influences the distribution of income does not necessarily mean improvement for the lower class or a disadvantage for the upper class. Although the welfare state tries to improve the welfare share of the lowest income group(s) more than that of other groups, the former s position does not improve due to the strong dynamics of market income. Quite the contrary; despite increasing redistribution, the position of the lower income group worsened. This can be seen in table 1, which compares the income shares of individual quintiles in terms of both market income and disposable income and shows how important the changes to these shares between the mid-1980s and the mid-2000s were. The losses in market income share of the lower income groups were moderated, but not fully compensated, at the level of disposable income. Instead, the top income group shows an increase in market income share, which persists despite an increase in redistribution (a pattern again with crossnational variations). Net growth took place at the top of the income hierarchy. According to the recent OECD report on income distribution (2011: 234), this gain at the top mainly results from capital market income. The share of income accruing to the bottom has become smaller, despite compensation provided by the welfare state. The consequences for the three groups in the middle differ significantly with respect to changes to market and welfare state income distribution. In many countries, the upper middle quintile benefited from the gains to the upper income quintile. The lower middle and middle middle by contrast had to accept a net worsening of their positions. Still, they saw their shrinking market income share almost completely compensated for. Thus, the lower middle shares the destiny of the lowest quintile, but only has to absorb small losses in market income and disposable income. How much of the losses at the bottom are compensated and gains at the top moderated varies by country. In Germany, for example, the bottom quintile s share of market income declined between 1985 and 2005 by 2.35 percentage points, but disposable income dropped only by 0.5 percentage points. The market income share of the topmost quintile rose by 2.46 percentage points, while disposable income grew by a mere 0.76 percentage points. However, in most countries, including in the Scandinavian countries, even after welfare state intervention losses at the bottom and gains at the top, are more pronounced than in Germany. Only in Denmark and Switzerland do the quintiles below the top show gains (data for 1990 to 2005 only). Redistribution partially compensated for growing market disparities, but failed to compensate for the far more serious worsening in the distribution of market income and why? According to the models discussed above, voter preferences that vary with income distribution on the one hand and political factors on the other should determine the trend of redistribution by the government. Does it in fact react to losses to market distribution positions and thereby accede to the (not directly measured) demands of the affected Page 13 of 25

14 groups? Alternatively, does the political power of various interest groups control redistributive government interventions? We will now test this for time-variant variables using fixed effect regressions. The results in Table 1 (model 1) show that government redistribution does indeed expand when the gap between middle and lower market incomes grows. As expected, the coefficient of the gap variable is positive and highly significant. The deterioration of the relative income positions experienced by households at the bottom of the hierarchy meant more downward redistribution. Notably, the changed gap between lower and middle market incomes has a stronger effect than the distance from the highest quintile (model 2). The pulling away by higher market incomes hence is less relevant for redistribution downward than closeness to the income positions of the middle class. Social comparison processes and acquiescence to higher gains at the highest income levels may explain this. With regard to hypothesis 1, we can state that it is not the narrow gap between poor and middle households that triggers more redistribution, but the decline of the lower income tier instead. The globalization variable taken into model 3 exhibits a significant positive effect. It confirms the assumption that the growing volatility in an open, internationally integrated economy will boost voter demands for regulatory interventions in the resulting marketbased distribution. This result is valid even independent of unemployment, which is normally considered in the literature as the reason why globalization arouses preferences for and ultimately causes growth in government redistribution. Table 1: Fixed effects regressions on redistribution (share gain bottom quintile) Ratio Q3-Q1 mi.798***.778***.874*** (5.45) (5.28) (4.6) Ratio Q5-Q1 mi 2.52** (3.90) Economic openness.001*** n.s. Domestic response Union density (5.04) n.s. n.s. Voter turnout Left parties in government (1 5)* n.s. n.s. n.s. n.s. Unemployment (Control).102* (2.15).129* (2.37).103* (2.17).087+ (1.74) Constants.160 (.41).083* (2.25) 1.58 (.41) 7.08*** (4.01) (.72) R Square 54% 44% 55% 6% 55% N Notes: T-statistics in brackets; R-Square within variance. Significance: *** <.001, ** <.01, * <.05. * Left parties: 1 = dominance conservative parties, 5 = dominance left parties. Surprisingly, all the variables used to measure the effects of political forces (models 4 and 5) turn out to be insignificant (even when the remaining variables for income struc- Page 14 of 25