INCOME INEQUALITY AND GOVERNMENT REDISTRIBUTION: A CROSS-NATIONAL STUDY

Abstract

This paper focuses on income inequality and government redistribution in 120 countries between 1980 and 2010. It begins by describing variation in inequality across countries and over time, distinguishing between income before and after government redistribution by way of taxes and social transfers. It then goes on to explore the sources of cross-national and over-time variation in inequality and redistribution with reference to a number of variables widely employed in the literature, including development level, economic globalization, ethnic fractionalization, political democracy and the partisan orientation of governing executives. We find that, other things being equal, per capita income is positively related to greater government redistribution and a more egalitarian distribution of post-government income, while ethnic fractionalization is related to less redistribution and greater inequality of disposable income. We find little evidence that economic globalization, democracy or partisan orientations are strongly related, in either direction, to the degree of government redistribution or post-government inequality in the countries we examine.

Headnote

This paper focuses on income inequality and government redistribution in 120 countries between 1980 and 2010. It begins by describing variation in inequality across countries and over time, distinguishing between income before and after government redistribution by way of taxes and social transfers. It then goes on to explore the sources of cross-national and over-time variation in inequality and redistribution with reference to a number of variables widely employed in the literature, including development level, economic globalization, ethnic fractionalization, political democracy and the partisan orientation of governing executives. We find that, other things being equal, per capita income is positively related to greater government redistribution and a more egalitarian distribution of post-government income, while ethnic fractionalization is related to less redistribution and greater inequality of disposable income. We find little evidence that economic globalization, democracy or partisan orientations are strongly related, in either direction, to the degree of government redistribution or post-government inequality in the countries we examine.

Key words: Income Inequality; Government Redistribution; Cross-National Analysis.

Over the last half-centuiy, few issues have attracted as much attention in the scholarly and policy communities-to say nothing of the popular imagination- as global income inequality. Although much attention has been directed to inequality across countries, inequality within countries is clearly the form of income distribution closest to ordinary people. While it may be of theoretical interest to slum dwellers in Mexico City that they are poor relative to average residents-or, for that matter, slum dwellers-in the United States, the fact that they are much worse off than residents of other parts of their own city is likely to be of more immediate interest. The aim of this paper is to focus on two dimensions of intra-country inequality in a large number of countries-120- over a relatively long period of time-31 years. The first task will be to describe cross-country and over-time variation in income distribution, distinguishing between income before and after government redistribution by way of taxes and transfers. The second will be to seek to explain variation in inequality and government redistribution with reference to a number of variables widely employed in the literature, including development level, economic globalization, ethnic fractionalization, democratization and partisan politics.

1 MEASURING INCOME INEQUALITY AND GOVERNMENT REDISTRIBUTION

The most important source of raw data on income inequality for a large number of countries is the World Income Inequality Database (WIID), which has been assembled by the United Nations University's World Institute for Development Economics Research. This paper will make use of a data set building upon the WIID that has been assembled by Frederick Soit (2013), a compilation that harmonizes WIID data, which vary considerably in income concept, unit of analysis, whether income is measured before or after taxes, etc. Solt's Standardized World Income Inequality Database (SWIID) is based on the most recent version of the WIID data set and follows the harmonization standards of the Luxembourg Income Study (LIS), whose data are a model of successful harmonization of income surveys conducted by national statistical authorities-but which covers only a relatively small number of countries, each for a only few points in time. Specifically, Soit uses what is known about the effect of definitional differences on inequality measures when data are available to estimate the effect when they are not As he puts it, "Because the factors that affect these ratios [between surveys employing different definitions] tend to change only slowly over time within a given country, the best prediction for a missing ratio will be based on data on the same ratio in the same country in proximate years" (2009, 236).2 Soit also fills in temporal gaps in survey availability by interpolating where appropriate, employing multiple imputation methods. Finally, he calculates standard errors for each measure in an effort to assess the stability of interpolated estimates, providing users with an empirically derived measure of their reliability.

Solt's data set measures within-country inequality of both pre-government market income and post-government disposable income. The basic unit of measurement is the Gini index, a summary indicator that ranges from 0 (all units receive the same income) to 1.000 (a single unit receives all income). In addition, Soit provides data for the relative difference between pre- and postgovernment inequality (calculated as [pre-government Gini - post-government Gini]/pre-government Gini) which taps the extent to which governments redistribute market income by way of taxes and transfers.

For purposes of this analysis we employed a fairly stringent criterion with respect to the stability of estimates, using only figures with standard errors of 3 or fewer Gini points on either pre- or post-government income.3 Even after these deletions, there remain 2407 country-years covering 120 countries, an average of 20.1 years per country (see the appendix for a complete list)

In describing global patterns of income inequality, we begin with some summary figures. Across all countries and years, the mean Gini index of pregovernment inequality is .424, that of post-government inequality is .366, and the mean reduction in pre-government values as a result of taxes and transfers is 14.1%. These summary figures are, however, averages of diverse national experiences. One way of focusing more closely on income distribution is to disaggregate countries by development level. In an effort to explore this dimension, we have sorted our 120 countries according to the World Bank's classification of countries as low-income, lower-middle income, middle-income or high-income economies. As can be seen in Figure 1, the highest pregovernment inequality is found in the lower-middle income group; the average Gini coefficient for this group is .457. Somewhat lower pre-government Ginis are in evidence in low- and upper-middle income economies, while the lowest Ginis (although not dramatically lower) are found in the high-income economies.

As is evident, post-government Ginis are in every case lower than pregovernment Ginis. However, the degree to which taxes and public social transfers reduce inequality varies considerably across country groups. In lowand lower-middle income economies government taxes and transfers reduce market income only slightly, 3.9 and 4.0% respectively. For upper-middle income countries the reduction is nearly twice as great at 7.9%. Finally, for high-income economies the reduction is 23.8%, almost three times as great as the reduction in upper-middle income economies and six times that in the lowand lower-middle income groups. In short, the most important difference in the final distribution of income between countries at different levels of development is not the extent to which inequality is generated by the market but rather the extent to which market inequality is reduced by state action.

So far we have described average values over a number of years. The obvious next question is whether inequality has been increasing or decreasing over time. In the estimation of Firebaugh (2003, 160), "Within-nation income inequality grew from 1980 to 1995 for all regions except Africa (and recall that the African data are the least reliable)." Similarly, Addison and Cornia (2003) found that a majority of the countries they examined experienced rising inequality between the 1960s and the 1990s. Milanovic (2005, 144) observed that "while in the past, one's income depended much more on the class he belonged to than on the place (country) where he lived, by mid twentiethcentury, it was the country much more than the social class that mattered. In the second half of the 20th century, however, the situation reversed again: the importance of within-country inequality rose." Finally, Clark (2011) found that between 1990 and 2008 the share of total global inequality accounted for by within-country inequality rose from 21.1% to 31.9%.

In building upon these earlier analyses, we compare the average Gini index of both pre- and post-government inequality for all 120 of our countries between 1980 and 2010. The trends are reported in Figure 2. As can be seen, average global pre- and post-government inequality increased steadily between 1980 and 2000, as government redistribution kept pace with the growth of pregovernment inequality-but no more than that In about 2000, market inequality stabilized and government redistribution increased, with the result that post-government inequality declined slightly, although it remained well above its level in 1980.

In an effort to explore some of these global trends in more detail, and also to move beyond the still heterogeneous groupings based on development level that were employed earlier, we conclude by reporting trends in postgovernment inequality in a few representative individual countries. The graphs in Figure 3 represent the post-government Gini values for several groups of countries over the last three decades. One of the most dramatic comparisons is of India and China, which are the world's most populous countries by far and, thus, have a major effect on the entire world's income distribution-as has been amply demonstrated in comparisons of between-country inequality, which has been reduced by large gains in per capita income in these countries in recent years (Firebaugh 2003). However, both countries are characterized by very high levels of internal inequality, which substantially undermines the benefits to low-income groups of large increases in average income. China, for its part, began the period with a relatively egalitarian distribution in 1981 (the first year for which data are available), when market reforms were just beginning: its post-government Gini in that year was .289, well below the world average of .320. However, inequality grew very rapidly over the next three decades, to the point that China's post-government Gini in 2009 (the most recent year for which data are available) was .474, more than a hundred Gini points above that year's global average of .359. As can also be seen in Figure 3, India began the period for which data are available (1988-2007) with a considerably higher level of internal inequality than China, but inequality grew only modestly in that country over the subsequent two decades. In the end, the situation in the two countries was quite similar, but it is significant that inequality in China has grown very rapidly from a low base while inequality in India, historically very high, has changed very little over time.

Another interesting comparison is of four Latin American countries, Chile, Venezuela, Brazil and Argentina. Inequality in all four is high by world standards; the most inegalitarian is Brazil, followed closely by Chile, Argentina, and then Venezuela. As can be seen, however, there has been a decline in inequality in all four countries since about 2005, a trend different from much of the rest of the world.

Yet another comparison is of three countries from Eastern Europe, Poland, Hungary and the Russian Federation. Inequality in the first two of these countries began the 1980s at a relatively low level by global standards. It rose a good deal in the next decade and a half after the rapid institution of a market economy in the early 1990s, although even at the end of the period inequality was lower than in most other regions. Since about 2000, inequality in these countries has stabilized at levels comparable to those of the more egalitarian countries of the developed world. As to the Russian Federation, the first year data is available is 1992. As can be seen, inequality even at that time was high in comparison with other Eastern European countries. Since the 1990s, it has remained very high, more than a hundred Gini points above the other two countries and well above the world average.

Finally, we offer a comparison of Norway and the United States, which represent the low and high ends of the inequality spectrum in the developed world. As can be seen, post-government inequality has grown steadily in the United States since 1980, and from a base that was already high by the standards of the developed world. In Norway, on the other hand, inequality has changed little: its Gini index in 2010 was almost identical to that in 1980. As a result of these trends, the inequality gap between the two countries, already large in 1980, has nearly doubled over the subsequent three decades.

Before concluding this section, two qualifications are in order. First, it must be noted that income reflects only one aspect-albeit a very important one-of a broader concept of well-being as a whole. In particular, household income does not directly translate into good health, effective education, access to a clean environment, community solidarity, security from crime or the availability of leisure, among other aspects of broader well-being. That said, it is clear that income is a critical component of overall well-being, if only because of its strong correlation with numerous other desirable characteristics (Wilkinson and Pickett 2009). Second, income statistics obviously do not fully reflect transactions that occur outside the cash economy, which tend to be more prevalent in poor countries than in rich ones (Gërxhani 2004). As a result, inequality in the less developed world is likely somewhat lower than reported in the income inequality statistics that are the subject of this article. However that may be, the fact is that the informal economy is by its very nature impossible to capture fully. This is especially true in cross-national comparisons, in part because the inherent difficulty of determining the cash value of many in-kind goods and services, particularly those performed by family and friends, and in part because many such transactions are intentionally hidden by participants in an effort to avoid taxes or government regulative scrutiny.

2 SOURCES OF WITHIN-COUNTRY INCOME INEQUALITY

Now that we have offered an overview of recent trends in income inequality, we will conduct an empirical analysis of some of the main variables that have been employed in the scholarly literature seeking to explain the substantial crosscountry and over-time variation in within-country income inequality that is clearly in evidence. Our focus is on two aspects of inequality. One is the relative extent to which pre-government inequality has been ameliorated by government redistribution via taxes and transfers. The other is the resultant distribution of income that households receive after both market forces and government policies have had their effect.

The most basic explanation for the degree of income inequality within a country, as well as for the extent to which income is redistributed by the public sector, is its development level. The longest-standing hypothesis linking average income and inequality is that of Kuznets (1955), who proposed that inequality will be modest in pre-industrial societies, will rise substantially during the process of industrialization, and will then decline as workers are able to wrest wage concessions from employers and use their political power to construct welfare states that socialize some of the costs of health care, old age, disability and unemployment More recently, however, scholars have hypothesized a negative relationship between development level and inequality, given that few if any countries today are completely pre-industrial (Alderson et al. 2005, 406; Firebaugh 2003, 93-95). The previous section bears this out; as has been seen, inequality is highest in the less developed world. However, a more systematic analysis is in order, both for its own interest and as a control when examining the effect of other variables.

As to redistribution, a standard hypothesis in the literature on public finance is that, as per capita income increases, so too will the degree of government redistribution, a hypothesis sometimes called Wagner's Law. As described by Pampel and Williamson (1989, 26), "The public sector grows because the demand of households for services and their willingness to pay taxes are income elastic .... The [perceived] need of governments to meliorate the harmful effects of industrialization thus occurs simultaneously with increases in income." Our indicator of income level is real gross domestic product (GDP) per capita, which measures the average output of goods and services by residents of a country. The figures are in U.S. dollars, have been adjusted for purchasing power parities (PPPs) and are expressed in 2005 constant prices employing a chain index. The source is the Penn World Table (Heston et al. 2012), the most widely used data series on GDP.

Our next independent variable measures the extent to which a country is engaged in the global economy. Of all the hypotheses considered in this paper, perhaps the largest literature-and certainly the most vigorous debate-has focused on the relationship between countries' integration into the global economy and the degree of internal inequality they experience. The basic arguments are well-known. On one side, many commentators have claimed that the rapid growth of the global movement of goods, services and capital in recent decades has driven a wedge into countries' internal economies and polities, separating groups in a position to gain from globalization from those undermined by it (Hurrell and Woods 1995). Many other commentators, however, counter that global integration serves as a powerful engine of growth, benefitting all income groups but particularly those of low or moderate income (Bhagwati and Dehejia 1994).

More formally, a standard approach to the relationship between economic globalization and internal income distribution is the Stolper Samuelson (1941) theorem. Stolper and Samuelson argued that income groups controlling relatively abundant factors of production will gain from trade and other economic interactions, while those holding relatively scarce factors will suffer from them. Because unskilled labour is abundant and capital scarce in the less developed countries, the implication is that the unskilled will benefit from trade and that the opposite will be true in the developed world. Critics, however, question whether Stolper Samuelson dynamics are in practice overridden by gains from economies of scale, diversification and technological innovation that arise from trade, to the benefit of all income groups (Burtless 1995)-or perhaps especially to high income groups, who may be in the best position to gain from such trends.

As to redistribution, expected outcomes are similarly in dispute. On the one hand, it is often argued that integration into the global economy enmeshes governments in a ruthlessly competitive "race to the bottom" in social protection (Rodrik 1997). On the other hand, it is often claimed that economic liberalism is only politically viable if it is "embedded" in a broader mechanism to compensate vulnerable, but politically powerful, groups (Ruggie 1982)-the "domestic compensation" approach. In the words of Garrett (1998, 824), "The coupling of openness with domestic compensation remains a robust and desirable solution to the problem of reaping the efficiency benefits of capitalism while mitigating its costs in terms of social dislocations and inequality."

What is the empirical evidence concerning the effect of economic globalization on the relative standing of low income groups? As is noted in a recent review of work on the topic as it relates to the developing countries (Goldberg and Pavcnik 2007, 39), the consensus seems to be that "distributional changes went in the opposite direction from the one suggested by the [Stolper Samuelson] conventional wisdom: while globalization was expected to help the less skilled, who are presumed to be the locally relatively abundant factor in developing countries, there is overwhelming evidence that these are generally not better off, at least not relative to workers with higher skill or education levels." Goldberg and Pavcnik (2007, 40), however, "abstain from relying on crosscountry regressions to econometrically identify the effects" of globalization on inequality because of "inconsistencies in the measurement of inequality across countries." Instead, they offer detailed comparisons of a small number of countries. The aim of this paper is to offer a broader analysis using the better and more abundant data that have recently become available in the SWIID data set.

In measuring economic globalization we have made use of the KOF index of globalization (Dreher 2006; Dreher et al. 2008, updated in 2014), which offers data for all of our countiy-years. While the overall KOF index taps a number of aspects of globalization, we have employed only its index of economic globalization. Extending previous work along these lines, which tends to concentrate only on trade and foreign direct investment, the KOF index also measures the extent of national restrictions on international movement of goods, services and capital. Specifically, it is based on the following actual movements across national borders: trade, foreign direct investment flows, foreign direct investment stocks, portfolio investment and income payments to foreign nationals, all expressed as a percent of GDP. In addition, the KOF index measures the following restrictions to cross-border movements: tariffs, nontariff barriers, taxes on international trade and capital account restrictions. The resultant index ranges from 0 (least globalized) to 100 (most globalized).

One of the longest-standing preoccupations of social science is the extent to which ethnicity serves as a basis for income inequality and complicates government redistribution. Literally thousands of studies have considered this topic, with reference to nearly every ethnically heterogeneous country in the world. The available cross-national evidence suggests that there is a positive relationship between ethnic fractionalization and income inequality. Milanovic (2003, 30), for example, found that, across a wide range of countries, "each 10 percent increase in fractionalization raises inequality by 3.3 Gini points." Similarly, Alesina et al. (1999) and Easterly and Levine (1997, 1205-1206) found that higher levels of ethnic fractionalization generally are associated with higher levels of post-government inequality, because "polarized societies will be both prone to competitive rent-seeking by the different groups and have difficulty agreeing on public goods like infrastructure, education, and good policies."

As to the expected effect of ethnic fractionalization on government redistribution, one would expect ethnic differences to complicate redistributive politics by encouraging social cleavages on the basis of ethnicity rather than income (Baldwin and Huber 2010; Jensen and Skaaning 2014). On the other hand, it is possible that the opposite will be the case if ethnic attachments trump income, such that redistributive policies benefitting low-income members of an ethnic group receive support even from high-income members of that group who do not gain directly from them.

In measuring ethnic fractionalization, we employ an indicator developed by Fearon (2003) that has been widely used in the recent literature on the topic. Although ethnic groups are ultimately determined by objective physical, linguistic or religious characteristics, Fearon has made an effort to define ethnicity in terms of groups that consider themselves to be distinct in politically significant ways. Fearon's data set covers all of our countries. However, his data are not available over time, so we have coded all time points for a given country at the same value. Although this is not ideal, ethnic attachments undoubtedly change at a much slower rate than any of our other variables. In Fearon's measure a higher value represents greater fractionalization.4

Our final two variables measure aspects of national political institutions, processes and policies. The first of these taps the extent to which a country practices political democracy. The most basic hypothesis linking democracy with income inequality and government redistribution suggests that democratic regimes, which (theoretically, at least) provide each member of society an equal voice in determining the regulations and policies that affect income inequality, will result in a more egalitarian distribution than authoritarian regimes, where decision-making is more concentrated. As put by Burkhart (1997, 149), "Democracy, due to its spreading political power, tends to spread economic power as well."

While this hypothesis is certainly plausible, it is not at all clear whether it actually operates across a wide range of countries and time periods. It is possible, for example, that democracies will reflect and reinforce existing inegalitarian market outcomes rather than serve as a vehicle to change them. Certainly, there is plenty of evidence that democratic political forms can coexist with a good deal of inequality; indeed, in the longstanding democracies, the trend in recent decades has been in an inegalitarian direction. Similarly, the democratic transition in most of Latin America in the last three decades has not been associated with an equally dramatic reduction in that region's historically high level of inequality (Kaufman 2009)-although, as has been seen, inequality has been gradually declining in recent years. Conversely, authoritarian regimes tend to have more power than democracies to alter market incomes if they wish to do so, since they are less constrained by checks on their power. The communist regimes of the twentieth century are a case in point.

What does the cross-national empirical literature reveal about the relationship between democracy and income inequality? As it happens, empirical analyses exploring the link between democratic political forms and an egalitarian distribution of income have often failed to find a direct relationship. Lee (2005, 175), for example, found that the degree of democracy in a country was "not directly associated with inequality," although it was possible that it had some effect through public sector expenditures. Similarly, while Boix (2003, 37) found that inequality in democratic systems "increases the redistributive demands of the population," he also notes that, as inequality moderates, resistance to further redistribution tends to grow, and that elections "have only a marginal impact" on inequality beyond a certain point (see also Acemoglu and Robinson 2006). As one recent review of the literature on the topic (Bermeo 2009, 24) concludes, "Overall, the long-term trends [in democracies] appear disappointing from an egalitarian perspective." Still, as she goes on to say, "the term 'appear' deserves emphasis because figures measuring economic inequality are notoriously problematic," and she highlights some of the measurement problems that plagued earlier cross-country measures of inequality. Our hope is that the more complete and comparable data employed here will help to address these concerns.

A great deal of effort has been devoted by social scientists to measuring democratic political forms in a way that can be meaningfully compared across a wide range of countries and time periods. In the view of many observers (see, e.g., Coppedge 2002, 35), the most sophisticated measure is the Polity IV database (Marshall and Jaggers 2009), which offers a variable ranging from -10 (least democratic) to 10 (most democratic) based on six component measures that record key qualities of executive recruitment, constraints on executive authority, and political competition.

The presence of democratic political forms is obviously not the only political variable of interest. Just as important, arguably, is the ideological orientation of the governing party or coalition. In fact, one of the perennial questions asked by political scientists is whether the partisan composition of governments helps to explain the degree of income inequality in a country or the scope of government policies aimed at reducing it One possibility is that leftist governments will be more inclined to support generous social benefits than conservative governments, the so-called "power resources" approach (Korpi and Palme 2003; Bradley et al. 2003; Huber et al. 2006). On the other hand, it is possible that left and right governments alike will find it difficult to accomplish major changes in redistributive policies because many of the key players, including business interests and social benefit recipients, are so entrenched that they are impervious to efforts to make more than marginal alterations to longstanding redistributive taxes and social transfers, whatever the intentions of political leaders.

While measuring ideological orientation is an inherently imperfect enterprise, a particularly careful effort is that of the World Bank's Database of Political Institutions (Beck et al. 2001; Keefer 2013), in which governing executives are characterized as either "left," "center" or "right" The primary focus of this measure is on economic issues, especially the distributive issues that are the main concerns of this paper. The variable is constructed such that governments classified as "left" are coded 3, "middle" are coded 2 and "right" are coded 1. Executives that could not be unambiguously classified (e.g., those controlled by ethnic or regional parties) are coded as missing.

In concluding this section, it is worth noting that, although most of the variables described above have been extensively examined in earlier cross-national empirical work, the data employed in those studies have, of necessity, been much less extensive and definitionally consistent than those analyzed here. For example, Reuveny and Li (2003) used Gini indexes that took no account of definitional differences in income concept, unit of analysis, coverage of government transfers, etc. Moreover, they addressed the problem posed by extensive missing data by computing average Gini values over an entire decade; as a result, temporal trends were neglected. Rudra (2004) used similarly limited data, and her analysis was limited to only 46 countries for an average of about 4 years each while Ha's (2012) cross-national study included 59 countries for an average of fewer than 3 years each. Finally, while Lee et al. (2007) did make an effort to control for definitional inconsistencies by including them on the right side of their equations, their data set included an unbalanced panel of only 311 observations, in contrast to the 2407 examined here.

In addition, as has already been noted, few broad cross-national studies explicitly consider the role of government in reducing inequality generated by the market As a result, in accounting for the final distribution of income it has been difficult to disentangle the effect of market inequality, which is to some extent a "given" that is largely determined by deep-seated historical forces, from that of public sector taxes and transfers which are presumably more susceptible to short and medium term political dynamics.5 This study, in contrast, measures both government redistribution and the resultant postgovernment distribution of income.

3 Results

Now that our variables have been introduced, it is time to report the results of our empirical analysis. As has been indicated, we will employ two dependent variables: a measure of the percentage reduction of pre-government income inequality as a result of taxes and government transfers and the resultant Gini index of post-government (disposable) income. Each equation includes the five independent variables introduced earlier: real PPP-adjusted GDP per capita; the KOF index of economic globalization; Fearon's measure of ethnic fractionalization; the Polity IV measure of democracy; and the World Bank's measure of the ideological orientation of a country's governing executive.

As to regression techniques, our data set constitutes an unbalanced pooled cross-sectional time series in which data vary both across countries and over time within countries, but there are gaps such that not every country has data for every year between 1980 and 2010. In analyzing unbalanced pools of this sort, the standard method is to employ a Huber-White "sandwich" robust estimator that clusters observations by country.6

We begin with our entire data set. The first dependent variable taps the relative difference between pre- and post-government inequality, that is, the redistributive impact of the state. As can be seen in Table 1, as per capita income rises, government redistribution, on average, also increases, a relationship that is statistically significant at the p<.001 level. (2-tailed tests are used here and throughout) This is consistent with Wagner's Law; it suggests that governments with higher income at their disposal have not only the means but also the will to reduce pre-government income inequality. While previous work would lead us to expect development level to be associated with redistribution, it is useful to know that the expected relationship continues to hold when we use the more extensive and comparable data of the SWIID data set.

Also statistically significant (at the p<.01 level) and in the expected direction, in this case negative, is the relationship between government redistribution and our measure of ethnic fractionalization: as fractionalization rises, redistribution, other things being equal, decreases. The widespread expectation (see, e.g., Easterly and Levine 1997 and Alesina et al. 1999) that ethnic divisions complicate redistributive politics appears to be confirmed.

Finally, we find a positive and statistically significant relationship between a country's openness to the global economy and the extent to which its government redistributes income by way of taxes and transfers. This offers confirmation of the "domestic compensation" perspective on redistribution, which suggests that governments seek to compensate groups affected by international competition.

As interesting as the results that are in evidence are those that are not Specifically, there is not a significant relationship between the extent to which a government redistributes market income and its score on the Polity IV democracy scale. While this is perhaps surprising, it is worth observing that the "Third Wave" of democratization between the 1970s and the early 2000s in Latin America, Eastern Europe and parts of Sub-Saharan Africa and East Asia, has not been accompanied by a corresponding decrease in income inequality. Quite the contrary. Indeed, even the very longstanding democracies of North America and Western Europe have experienced a steady increase in inequality since the early 1980s. On the other hand, some of the most inegalitarian and least redistributive regimes in our study are highly authoritarian; examples include Brazil in the early 1980s or present-day China or Haiti. In short, there does not seem to be a systematic relationship-in either direction-between democratization and state redistribution. Nor is there a statistically significant relationship between the ideological orientation of a country's executive and the extent to which its state redistributes market income by way of taxes and social transfers, which fails to confirm the power resources perspective that partisan politics plays a central role in redistribution.

What of our measure of post-government inequality, which represents the final distribution of income after both market forces and government redistribution have had their effect? To start, GDP per capita is significantly negatively related to the Gini index of post-government inequality, confirming that development level is indeed associated with a more egalitarian distribution of postgovernment income. Similarly, ethnic fractionalization is positively and significantly related to post-government inequality: across all of the countries and years we examine, ethnically heterogeneous societies are, on average, less egalitarian than more homogenous societies.

As to our other independent variables, none is strongly related, in either direction, to income inequality across the countries and years we examine. In particular, our measure of economic globalization, which was positively related to government redistribution, is unrelated to post-government inequality. Nor are the democratization score or our measure of the ideological orientation of political executives.

The preceding results are, of course, averages across a wide range of countries. In exploring these relationships further, it is useful to distinguish between the very different types of countries represented by the World Bank's classification of countries into low, lower-middle, upper-middle and high income groups, as described earlier. The results of such an analysis are reported in Table 2. We begin with the low-income group. With respect to government redistribution, three statistically significant results are in evidence: ethnic fractionalization, integration into the global economy, and the partisan orientation of a country's governing executive are all positively related to the relative extent of redistribution. As to ethnic fractionalization, the positive relationship in evidence is somewhat surprising, since fractionalization is expected to complicate state redistribution (and appears to do so across the entire dataset). One possible explanation is that in very poor countries ethnicity trumps class such that ethnic groups engage in inter-group political bargaining that extends benefits to a wider range of income levels within their ethnic group than would otherwise be the case (Baldwin and Huber 2010). Second, there is a statistically significant positive relationship between a country's globalization score and redistribution, which offers confirmation of the domestic compensation approach even in the context of very poor countries (although, of course, the absolute level of compensation is smaller than in other regions). Finally, there is a significant relationship between the ideological orientation of a country's governing executive and the degree to which its state redistributes market income, a relationship that was not in evidence for the entire data set but does seem to exist for this subset of countries.

What of the post-government Gini? Three statistically significant relationships are also in evidence. The first is a negative relationship between ethnic fractionalization and post-government inequality. Together with the previous result on redistribution, this can be interpreted as a situation in which even a measure of redistribution on ethnic grounds cannot overcome an initially high level of ethnically-based inequality. In this regard, it is notable that low-income countries are more heterogeneous than those in other income categories: their average fractionalization score is .67, as opposed to .48 for the lower-middle income group, .42 for the upper-middle income group and .28 for the high income group.

In addition, we find a positive relationship between post-government inequality and a negative relationship with a country's partisan orientation. As to the former, it appears that very poor democracies are actually slightly less egalitarian than very poor authoritarian regimes, although one would not want to make too much of this, given the small number of democracies in this income group. As to the latter relationship, it does appear that leftist governments in very poor countries-whether democratic or not-are somewhat more egalitarian than conservative regimes.

It is now time to move on to the countries in our lower-middle income category. This is, as one might imagine, a diverse group, including several of the poorer countries in Latin America such as El Salvador, Paraguay, Nicaragua and Bolivia; several of the richer countries in Sub-Saharan Africa, such as Ghana and Lesotho; and several populous countries of South Asia, including India and Pakistan. As it happens, this diversity appears to be reflected in our results: not a single statistically significant result is in evidence linking either government redistribution or post-government inequality to any of our independent variables.

What of the upper-middle income countries? As can be seen, for this income group GDP per capita is positively related to government redistribution, once again confirming Wagner's Law relating these variables. No other significant relationships are, however, in evidence. As to post-government inequality we find that, even for this narrower range of income, as a country's average income rises, internal inequality declines. In addition, ethnic fractionalization is once again positively related to post-government inequality, as is the extent of a country's integration into the global economic system. The latter relationship is in evidence only for this group, which includes countries that are typically experiencing rapid economic growth-growth that is clearly unevenly experienced by income groups.

Finally we turn to our high-income countries. As can be seen, ethnic fractionalization once again appears to complicate redistributive politics. Similarly, a country's economic globalization score is positively related to the extent to which its state redistributes income by way of taxes and transfers, confirming the domestic compensation hypothesis. Both of these results are consistent with a good deal of previous quantitative research-which, of course, is more extensive for this group than for any of the others. In addition, there is a modest positive relationship between a country's democratization score and government redistribution, the first and only time we have seen such a relationship. As to post-government inequality only one significant relationship is in evidence: as was the case for the upper-middle income country group, ethnic fractionalization appears to serve as a basis for income inequality.

As to non-findings, GDP per capita is not significantly related, in either direction, to either government redistribution or post-government inequality in the developed countries we examine, which is not surprising given the limited variation in per capita income for these countries. Similarly, income inequality is not significantly related to the KOF index of economic globalization. This is consistent with a good deal of empirical evidence that, while globalization harms some workers in the developed world, it helps others, such that there is not a strong overall impact on inequality (Pontusson 2005,199).

In sum, a number of our expectations about the sources of cross-national variation in inequality and government redistribution are borne out by our empirical analysis. The fact that these expectations hold up for data that are much more extensive, recent and comparable than those used in previous work helps to place our understanding of the sources of inequality and government redistribution on a much firmer base.

One non-finding of particular interest is the general lack of a relationship, in either direction, between democratic political forms and government redistribution or post-government inequality. This is in contrast to the expectation-or perhaps hope-of many observers that democracies will be more egalitarian than non-democracies, reflecting the broader political participation democracy embodies (Przeworski 2009). As has been noted, empirical evidence bearing out this expectation has been scarce (Kaufman 2009; Bermeo 2009). Still, given the importance of the topic, it is worth exploring it further. One way in which this can be done is to focus on countries that experienced a democratic transition between 1980 and 2010, the period covered in our analysis, operationally defining transition as movement from below the Polity IV threshold for considering a country a full democracy (+6) to above the threshold.7 Of our countries, 24 experienced such a transition (we do not include countries that experienced more than one transition, with periods of authoritarianism between.) For these cases we have constructed a series of country-by-country difference of means tests, comparing the average Gini index of post-government inequality before and after the transition. Of the 24 countries, 10 experienced a statistically significant (at the p < .05 level) decrease in inequality; in 10 there was no significant change; and in 4 there was a significant increase in inequality. This further confirms our earlier finding that there is little evidence of a strong and systematic relationship between democratization and income inequality, controlling for other variables.

4 CONCLUSION

This paper has had two primary aims. The first has been to explore global patterns of within-country income inequality and government redistribution, making use of a major new data set on income inequality. The second has been to explain cross-national variation in inequality and government redistribution with reference to five variables that have been widely employed in the literature: GDP per capita, economic globalization, ethnic fractionalization, democratization, and the partisan orientation of governing parties or coalitions.

With respect to the first aim, our most obvious finding is that income inequality, both before and after government redistribution, varies widely across countries of the world. So too does the extent to which governments redistribute market income by way of taxes and social transfers. In fact, the substantially lower post-government inequality found in high-income countries is more a product of extensive government redistribution than of an egalitarian distribution of pre-government income. In addition to considering cross-country patterns, we also charted the steady growth of within-country inequality over the last three decades, a trend that, while proceeding at different speeds, has been evident in the vast majority of the countries we have considered.

Our second aim has been to explore the main sources of cross-country variation in income inequality and government redistribution. Across all countries and years, and also for groupings of countries at different levels of development, we found that per capita income was positively related to a more egalitarian distribution of income and more government redistribution. In addition, ethnic fractionalization, other things being equal, tends to be associated with greater inequality and to complicate government redistribution. Economic globalization, for its part, tends to be associated with greater redistribution, although generally not a more egalitarian distribution of post-government income, suggesting that any increase in market inequality associated with economic globalization is to some degree counteracted by government redistribution. As to democratization, we did not find a strong relationship in either direction between the presence of democratic political forms and income inequality or government redistribution, a finding that was reinforced by a supplementary analysis of countries that experienced a democratic transition during the period under consideration.

Few issues of our time are as important as the growth of income inequality in much of the world over the last three decades. However, as was suggested at the beginning of the paper, much our understanding of the sources of crossnational variation in income inequality and government redistribution to date has been based on a limited and less than fully comparable database. This paper has sought to address this gap in the literature by examining a carefully harmonized data set covering a large number of countries over three decades, in an effort to better understand global patterns of income inequality and the forces that shape it.

Footnote

2 The SWIID extends earlier work by Babones and Alvarez-Rivadulla (2007), who apply constant adjustments across all countries.

3 We have also dropped any countries that no longer exist as unified states, such as Yugoslavia, the Soviet Union or Czechoslovakia. "Germany" consistently refers to the Federal Republic.

Footnote

4 Baldwin and Huber (2010) have developed a measure of ethnic fractionalization that takes into account economic differences between groups, which would obviously be desirable for our purposes. However, their measure covers only 46 countries, in contrast to the full coverage provided by Fearon's measure.

Footnote

5 One exception is an article on Latin American and the Caribbean by Morgan and Kelly [2013), which seeks to explain cross-national variation in both market and disposable income, on the assumption that government policies can influence the final distribution of income in ways other than direct redistribution.

6 The "robust cluster" regression option in Stata 13.1 was used. Listwise deletion was employed, with the result that regressions are based on 1774 of our original 2407 cases. In an effort to consider whether significant collinearity was present, VIF values were calculated. The highest for any equation was 2.07, well below the conventional criterion of 4.00.

Footnote

7 A recent empirical analysis in which distributive conflict was related to transitions to democracy found only a limited relationship between the two (Haggard and Kaufman 2012).

References

REFERENCES

Acemoglu, Daron and james A. Robinson. 2006. Economic Origins of Dictatorship and Democracy. New York: Cambridge University Press.

Addison, Tony and Giovanni Andrea Cornia. 2003. Income Distribution Changes and their Impact in the Post-World War II Period. UNU, WIDER, Discussion Paper No. 2003/28.

Alderson, Arthur S., Jason Beckfield and Francois Nielsen. 2005. Exactly How Has Income Inequality Changed? Patterns of Distributional Change in Core Societies. International Journal of Comparative Sociology 46 (5-6): 405-423.

Alesina, Alberto, Reza Baqir and William Easterly. 1999. Public Goods and Ethnic Divisions. Quarterly Journal of Economics 114 (4): 1243-1284.

Babones, Salvatore J. and Maria Jose Alvarez-Rivadulla. 2007. Standardized Income Inequality Data for Use in Cross-National Research. Sociological Inquiry 77 (1): 3-22.

Baldwin, Kate and John D. Huber. 2010. Economic versus Cultural Differences: Forms of Ethnic Diversity and Public Goods Provision. American Political Science Review 104 (4): 644-662.

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): 165-175.

Bermeo, Nancy. 2009. Does Electoral Democracy Boost Economic Inequality? Journal of Democracy 20 (4): 21-35.

Bhagwati, Jagdish and Vivek H. Dehejia. 1994. Freer Trade and Wages of the Unskilled: Is Marx Striking Again? In Bhagwati and Martin Kosters eds. Trade and Wages: Leveling Down? Washington, DC: American Enterprise Institute.

Boix, Carles. 2003. Democracy and Redistribution. Cambridge: Cambridge University Press.

Bradley, David, Evelyene Huber, Stephanie Möller, Francois Nielsen and John D. Stephens. 2003. Distribution and Redistribution in Postindustrial Democracies. World Politics 55 (2): 93-128.

Burkhart, Ross E. 1997. Comparative Democracy and Income Distribution: Shape and Direction of the Causal Arrow .Journal of Politics 59 (1): 148-164.

Burtless, Gary. 1995. International Trade and the Rise of Earnings Inequality. Journal of Economic Literature 33 (2): 800-816.

Clark, Rob. 2011. World Income Inequality in the Global Era: New Estimates, 19902008. Social Problems 58 (4): 565-592.

Coppedge, Michael. 2002. Democracy and Dimensions: Comments on Munck and Verkuilen. Comparative Political Studies 35 (1): 35-39.

Dreher, Axel. 2006. Does Globalization Affect Growth? Evidence from a New Index of Globalization. Applied Economics 38 (10): 1091-1110.

Dreher, Axel, Noel Gaston and Pirn Martens. 2008, updated in 2014. Measuring Globalization - Gauging its Consequences. New York: Springer.

Easterly, William and Ross Levine. 1997. Africa's Growth Tragedy: Policies and Ethnic Divisions. Quarterly Journal of Economics 112 (4): 1203-1250.

Fearon, James D. 2003. Ethnic and Cultural Diversity by Country. Journal of Economic Growth 8 (2): 195-222.

Firebaugh, Glenn. 2003. The New Geography of Global Income Inequality. Cambridge, MA: Harvard University Press.

Garrett, Geoffrey. 1998. Partisan Politics in the Global Economy. New York: Cambridge University Press.

Gërxhani, Klarita. 2004. The Informal Sector in Developed and Less Developed Countries: A Literature Survey. Public Choice 120 (3-4): 267-300.

Goldberg, Pinepoli Koujianou and Nina Pavcnik. 2007. Distributional Effects of Globalization in Developing Countries. Journal of Economic Literature 45 (1): 39-82.

Ha, Eunyoung. 2012. Globalization, Government Ideology, and Income Inequality in Developing Countries. The Journal of Politics 74 (2): 541-557.

Haggard, Stephan and Robert R. Kaufman. 2012. Inequality and Regime Change: Democratic Transitions and the Stability of Democratic Rule. American Political Science Review 106 (3): 495-516.

Heston, Alan, Robert Summers and Bettina Aten. 2012. Penn World Table Version 7.1. Center for International Comparisons of Production, Income and Prices at the University of Pennsylvania. Available at http://pwtecon.upenn.edu/php_site/pwt_ index.php.

Huber, Evelyne, Francois Nielsen, Jenny Pribble and John D. Stephens. 2006. Politics and Inequality in Latin America and the Caribbean. American Sociological Review 71 (6): 943-963.

Hurrell, Andrew and Ngaire Woods. 1995. Globalisation and Inequality. Millennium: Journal of International Affairs 24 (3): 447-470.

Jensen, Carsten and Svend-Erik Skaaning. 2014. Democracy, Ethnic Fractionalisation, and the Politics of Social Spending: Disentangling a Conditional Relationship. International Political Science Review, May 2014, available on-line.

Kaufman, Robert R. 2009. The Political Effects of Inequality in Latin America: Some Inconvenient Facts. Comparative Politics 41 (3): 359-379.

Keefer, Philip E. 2013. Database of Political Institutions. Washington, DC: World Bank.

Korpi, Walter and Joakim Palme. 2003. New Politics and Class Politics in the Context of Austerity and Globalization: Welfare State Regress in 18 Countries, 1975-95. American Political Science Review 97 (3): 425-446.

Kuznets, Simon. 1955. Economic Growth and Income Inequality. American Economic Review 65 (1): 1-28.

Lee, Cheol-Sung. 2005. Income Inequality, Democracy and Public Sector Size. American Sociological Review 70 (1): 158-181.

Lee, Cheol-Sung, Francois Nielsen and Arthur S. Alderson. 2007. Income Inequality, Global Economy and the State. Social Forces 86 (1): 77-111.

Marshall, Monty G. and Keith Jaggers. 2009. Polity IV Project: Political Regime Characteristics and Transitions, 1800-2007. Severn, MD: Center for Systematic Peace, available atwww.systemicpeace.org/polity/polity4.htm.

Milanovic, Branko. 2005. Worlds Apart: Measuring International and Global Inequality. Princeton, NJ: Princeton University Press.

Milanovic, Branko. 2003. Is Inequality in Africa Really Different? Washington, DC: World Bank Working Paper WPS3169.

Morgan, Jana and Nathan J. Kelly. 2013. Market Inequality and Redistribution in Latin America and the Caribbean. The Journal of Politics 75 (3): 672-685.

Pampel, Fred C. and John B. Williamson. 1989. Age, Class, Politics and the Welfare State. New York: Cambridge University Press.

Pontusson, Jonas. 2005. Inequality and Prosperity: Social Europe vs. Liberal America. Ithaca, NY: Cornell University Press.

Przeworski, Adam. 2009. Democracy, Equality and Redistribution. In Political Judgement: Essays for John Dunn, eds. Bourke, Richard and Raymond Geuss. Cambridge: Cambridge University Press.

Reuveny, Rafael and Qi Li. 2003. Economic Openness, Democracy and Income Inequality. Comparative Political Studies 36 (5): 575-601.

Rodrik, Dani. 1997. Has Globalization Gone Too Far? Washington, DC: Institute for International Economics.

Rudra, Nita. 2004. Openness, Welfare Spending and Inequality in the Developing World. International Studies Quarterly 48 (3): 683-709.

Ruggie, John Gerard. 1982. International Regimes, Transactions and Change: Embedded Liberalism in the Postwar Economic Order. International Organization 36 (2): 379415.

Soit, Frederick. 2009. Standardizing the World Income Inequality Database. Social Science Quarterly 90 (2), 231-242

Soit, Frederick. 2013. The Standardized World Income Inequality Database. Available at http://handle.net/1902.1/11992 Frederick Soit [Distributor] vlO.

Stolper, Wolfgang F. and Paul A. Samuelson. 1941. Protection and Real Wages. Review of Economic Studies 9 (1): 58-73.

Wilkinson, Richard and Kate Pickett 2010. The Spirit Level: Why Greater Equality Makes Societies Stronger. New York: Bloomsbury Press.