Inequality aversion and demand for redistribution among citizens in developed and developing countries

(1)

Inequality Aversion and Demand for

Redistribution among Citizens in Developed

and Developing Countries

MSc Political Science – Political Economy

Student name: Rogier Daniël Boer

Student number: 11151714

Thesis Supervisor: Dr. S.J. Lim

Second Reader: Dr. S. Tanaka

Date: June 22, 2018

(2)

2

Abstract

Income inequality has increased across many nations worldwide due to sustained periods of economic growth. As a result, inequality and preferences for redistribution became a widely studied topic in the academic literature. However, less is known about the relationship between inequality aversion and public demand for redistributive policies implemented by the state. Especially citizens living in developing countries might be reluctant to actively pursue redistribution by the state, as these countries generally have weaker institutions and smaller governments. Adopting a global and cross-regional perspective, this thesis examines the nexus between inequality disapproval and support for redistribution in 59 countries.

Multiple linear mixed models are executed to determine under what conditions this relationship becomes stronger or weakens. Results show that institutional quality and social trust positively moderate the relationship between inequality averse citizens and their support for more generous redistributive policies. In particular, government effectiveness, democratization, control of corruption and trust in other people in the society conditions the nexus between inequality aversion and support for redistribution. Partial support has been found for trust in institutions. It appears that only trust in the civil service has a positive conditional effect on demand for redistribution. The results are robust with two indicators of support for redistribution: support for a progressive tax system and support for social security.

(4)

4

Acknowledgements

The writing and completion of this thesis would not have been possible without the assistance, support and guidance of several people. First and foremost, I owe an enormous debt of gratitude to my thesis supervisor Sijeong Lim, for her patience, motivation, enthusiasm, and immense knowledge about the research topic. Without her persistent support, expertise, and critical evaluations, the completion of thesis would not have been possible. Second, I would like to extend my gratitude towards my housemate Kevin Sutherland, and friends Coen van Oostrum and Remco de Wit, who have been a constant source of joy for me. Last but not least, I want to thank my college friends for the day by day assistance, company, dinners and fun at the university.

Rogier Daniël Boer Amsterdam June 22, 2018

(5)

5

1. Introduction

National income inequality has increased in many regions since the 1980s. Especially in low-income countries, the level of low-income inequality has remained persistently high over the past decades (Fabrizio et al., 2017). However, since the mid-1990s the picture of income inequality has become much more heterogeneous in developing countries. While, the distribution of income has become more equal in Latin American countries, inequality trends are inconsistent in African and Asian countries (Klasen, 2016). According to Klasen (2016) these trends in income inequality can to a large extent be attributed to the implementation of different government policies toward income redistribution.

But why do these government policies toward income redistribution differ so widely across countries? The academic literature on attitudes toward income inequality can broadly be divided into a domain for political economy researchers and sociologists. While standard political economy theory models the issue merely from a rational choice perspective, social researchers mainly tried to address attitudes toward income inequality through examining social justice norms and deservingness perceptions. In their seminal work, written from a political economy perspective, Meltzer and Richard (1981) suggest that rising levels of inequality lead to public disapproval of inequality, which in turn results in more welfare policies and less inequality. In contrast, researchers belonging to the social school of thought argue that inequality is much higher in the United States than in mainland Europe, while redistributive policies are more extensive in Europe. Hence, they claim that inequality aversion differs across individuals due to varying social norms about an adequate distribution of income. Moreover, these researchers argue that such social justice norms also vary across nations. Finally, there is also a group of researchers, who claim that in addition to perceptions on inequality and social justice norms, also economic circumstance may affect public demand for redistributive policies.

In short, it is known from existing bodies of literature that the extent to which inequality spurs popular demand for redistribution varies between societies, because people who are socialized in different environments perceive and assess inequality differently. However, it remains unclear in the academic literature if people that disapprove inequality also turn to their government to demand more generous redistributive policies. In fact, there can also be another source of variation in the nexus between inequality aversion and support for redistribution: even those

(6)

6

people who perceive the degree of inequality in their society as undesirable might be reluctant to turn to their government for redistributive policies. Two reasons may underlie this variation. First, people in developing countries might be reluctant to redistributive policies because they believe that their government is incapable of effectively redistributing public money due to institutional weaknesses (Tanzi, 1998a). Second, it has been argued that people in developed countries show higher levels of social trust, which makes that citizens in these countries have more deeply held norms of reciprocity and conditional obligations to other people in their society (Uslaner, 2016). As a result, I use both country-specific characteristics and individual-level perceptions in this study to test if such elements affect the nexus between inequality aversion and demand for redistributive policies. In particular, I check if objectively measured dimensions of institutional quality and subjective trust in institutions as well as in the society moderates the nexus between inequality disapproval and active support for redistribution. Various variables are incorporated into this research to measure these instruments in order to get more reliable results. Hence, the research question of this thesis is: why do citizens in developing countries that disapprove the level of inequality not actively turn to their government to demand more generous redistributive policies?

The present work contributes to the academic literature by examining a crucial issue that has been overlooked by many social and political economy researchers. First, a significant body of prior research proposes that attitudes toward inequality differ per country due to either varying social norm across groups, attitudes toward the actual level of inequality in a country, or economic circumstances (e.g. Lübker, 2004; Meltzer & Richard, 1981; Schmidt, 2012). In fact, only a few studies have empirically examined the nexus between inequality aversion and support for redistribution (e.g. Kuziemko et al., 2015, Svallfors, 2013). However, these pieces of research solely involve one-country studies that are prone to selection bias. Hence, evidence based on large sample sizes is sui generis in the academic literature. As a result, this work contributes to the literature by examining differences in the support for redistribution across a large number of developing and developed countries. Second, although several studies indicate that institutional quality, trust in institutions and social trust are associated with the implementation of redistributive measures or welfare policies (e.g. Algan, Cahuc and Sangier, 2011; Justman & Gradstein,1999; Gupta, Davoodi, and Alonso-Terme, 2002), it remains unclear whether such indicators also have a conditional effect on the relationship between inequality disapproval and

(7)

7

support for redistribution. Third, most researchers in the field focus exclusively on the effect of either objectively measured macro-level indicators or subjectively measured individual-level data. Therefore, this research contributes to the literature by comparing the effect of both macro- and micro-level data on support for redistribution

This thesis is organized as follows. Chapter 2 gives an overview of the literature on attitudes toward income inequality. Chapter 3 addresses the theory on support for redistribution and develops the research hypotheses. Chapter 4 discusses the data, variables and methodology used in this study. Chapter 5 provides the results of the statistical analyses. Chapter 6 presents various robustness checks. Chapter 7 provides the concluding remarks, implications, limitations, and directions for future research.

(8)

8

2. Literature Review

This chapter addresses the academic literature on inequality and public demand for redistributive policies. There exists a large body of research dedicated to investigating the determinants of income inequality First, a number of researchers argue that economic factors, such as economic development (Kuznets, 1955), openness to trade (Edwards, 1997) and foreign direct investment (Figini, 2011) are among the explanatory variables that impact income inequality. Second, also human capital and technological advancements influences a country’s income inequality (Acemoglu, 2002). Third, standard political economy theory claims that primarily government policies play an important role in affecting within-country inequality (e.g Atkinson, 1999; Meltzer & Richard, 1981). Finally, some scholars suggest it is crucial to have an understanding of what people perceive as social justice when studying governmental policies about income distribution (e.g. Alesina & Angeletos, 2005; Sen, 2000).

Within this research, especially the latter two explanations are of primary interest. Therefore, the next section assesses standard political economy theory on attitudes toward inequality. Hereafter, the effect of differing social norms on inequality perceptions is attended to. Additionally, a few other determinants that may affect public perceptions on inequality are addressed. Finally, the gap in the literature is addressed and the research question of the present work is given.

2.1 Standard Political Economy Theory on Attitudes to Income Inequality

Standard political economy scholars suggest that the level of income inequality depends largely upon the extent to which governments are directly involved in the national economy. According to Stack (1978) factors such as employment producing expenditures, state provided employment, and the introduction of fiscal and monetary policies have a considerable bearing on within-country income distributions. Furthermore, Stack (1978) emphasizes that the relationship between direct government involvement and income inequality is likely to be independent of the level of economic development and the rate of economic growth. However, the results found by Stack (1978) are highly controversial, as several researchers argue that these are prone to selection bias and do not hold under more complete model specifications (Firebaugh, 1980; Jackman 1980).

(9)

9

Yet, despite the criticisms towards the work of Stack (1978), the political economy literature is quite clear about the impact of some government policies on inequality. According to Atkinson (1999) levels of inequality are to a large extent subject to policy choices made by governments that influence redistribution of incomes and market incomes. This suggestion is supported by several empirical studies. Katz, Mahler, and Franz (1983) find that fiscal instruments, such as personal income taxes, can be effectively used to achieve smaller levels of income inequality. Furthermore, they suggest that in addition to fiscal policies, also social welfare policies and education affect income inequality. Indeed, Jao (2000) demonstrates that social welfare expenditure is positively associated with reduced income inequality in Taiwan, while Eckstein and Zilcha (1994) their findings demonstrate that government intervention in providing compulsory basic education leads to a more equal income distribution. Finally, it has been stated that also the level of corruption in the public sector affects the level of income inequality. According to Tanzi (1995) rising levels of corruption in the public sector are likely to accrue to high-income groups, as these groups are often the better connected to government officials in a country. Hence, he claims that levels of inequality are generally lower in countries that are better able to control corruption.

In short, according to standard political economy theory, governments are to a large extent able to fight rising levels of inequality. However, if national governments are indeed able to effectively shape an even distribution of income, how then is it possible that countries their levels of inequality differ so widely? According to scholars belonging to political economy theory, this is primarily due to the fact that citizens only turn to their government for redistributive policies when it has reached a certain threshold. One of the most frequently cited articles on this is written by Meltzer & Richard (1981). They argue that public disapproval of inequality moves back and forth with actual levels of inequality in a country. More specifically, they develop an economic model to explain how public demand for redistribution policies is shaped by the actual level of inequality. According to this model, a rising level of inequality leads to an increasing distance between the mean and median pretax income, as the median income gets further below the mean income. As long as the income of the median income person decreases further relative to the income of the mean income person, the more the median income person will benefit from redistributive policies, as the money he will receive will exceed his tax burden. Hence, when the actual level of income inequality in a country increases, this will naturally lead to more political

(10)

10

demand for redistributive policies. In turn, Meltzer and Richard (1981) argue that in a responsive democratic country this growing demand is likely to result in the implementation of redistributive policies, because politicians are not likely to withstand public demand for more redistribution as they want to maximize their chances to remain in power.

A large number of academic papers have tested the Meltzer-Richard hypothesis. Finseraas (2009) employed a model to assess the importance of income inequality on demand for redistribution in 22 European countries. His main finding is that increased levels of inequality are indeed positively associated with public demand for redistribution. Furthermore, he argues that if these calls for redistribution are directly transformed into redistributive policies as the Meltzer-Richard model assumes, cross-country differences are likely to reduce over time (Finseraas, 2009). Additionally, Mahler (2008) examines the sources of variation in redistributive policies across thirteen developed democracies from 1979 to 2000. He finds that pre-government inequality is indeed positively associated with the implementation of more generous redistributive policies. Finally, Moene & Wallerstein (2001) try to explain what kind of redistributive measures are primarily supported in times of higher levels of inequality. They suggest that a distinction must be made between welfare expenditures targeted to the employed and unemployed population. As such, greater inequality leads to more support for welfare expenditures targeted at employed people, but decreases when these benefits are targeted at the unemployed citizens (Moene & Wallerstein, 2001).

Due to the conventional wisdom that rising levels of inequality automatically lead to a growing public demand for redistributive policies, academic debate on income inequality has so far almost exclusively focused on self-interest. However, it has been claimed, primarily by social researchers, that the results of these quantitative political economy studies have several limitations. First, the electorate may have imperfect information about the actual degree of inequality. This can be due to cognitive limitations as well as media effects (Kenworthy & McCall, 2007). With regard to the former, according to cognitive dissonance theory, voting behavior might be influenced by historic votes for a certain candidate (Mullainathan and Washington, 2009). As regards the latter, it has been found that subscriptions to certain newspapers or television channels influence voting behavior as well as political opinions (e.g. Gerber, Karlan, and Bergan, 2009; DellaVigna and Kaplan, 2007). Consequently, citizens their

(11)

11

perceptions on income inequality might not fully reflect their opinion about the true level of inequality. As a result, it must be taken into consideration that attitudes toward income inequality in a country may not fully mirror people their opinion on the true level of inequality. Lastly, it has also been argued by social researchers that statistical analyses do not account for justice beliefs or distributional ethics, while preferences for redistribution of the electorate may be guided by social norms rather than purely self-interest though (e.g. Kahneman, Knetsch and Thaler, 1986; Schmidt-Catran, 2016). Therefore, the next section discusses the effect of social norms on the acceptance of income inequality.

2.2 Social Norms and the Acceptance of Income Inequality

Apart from the effect of actual levels of inequality on public attitudes toward inequality, social researchers argue that perceptions of fairness and deservingness also play an important role in public attitudes toward inequality. In particular, these researchers disagree with standard political economy theory due to the fact that in some countries, such as in mainland Europe, redistributive policies have been much more extensive than other countries, such as the United States, even though the distribution of income is much more uneven in the United States. Therefore, this school of thought argues that social justice, fairness and deservingness perceptions affect income inequality among countries and individuals. One of the main proponents of this stream of researchers is Amartya Sen (2000), who emphasizes that it is crucial to have an understanding of what people perceive as social justice when studying governmental policies on income inequality. He highlights that judgements of people on inequality are a function of both ideas about what is morally right and of their perception of the actual income distribution with which they compare these norms. Hence, individuals their attitudes toward actual income distributions are likely to be influenced by their ideas about what is normatively tolerable and what they actually see in the society around them.

Several empirical scholars document that human behavior is not necessarily driven only by the expectation of future material benefits. Fehr and Falk (2002) show by means of an experiment that there are powerful non-pecuniary which shape humans their perceptions about a fair distribution of income. Furthermore, as regards the distribution of income, Kahneman, Knetsch and Thaler (1986) demonstrate that individuals are willing to take a loss rather than accepting an income distribution they view as unfair. They suggest that these judgments of fairness are

(12)

12

influenced by framing and several other factors that are considered irrelevant by standard political economy theory. Additionally, Alesina and La Ferrara (2005) studied how people their perceptions on future income mobility shape their preferences for income redistribution. They find that preferences for redistributive policies depend on individual beliefs on what determines somebody his position on the social ladder. More specifically, they argue that individuals who believe that the society offers equal opportunities are more likely to disapprove redistributive policies. Furthermore, they show that individuals who believe that getting rich is not unduly influenced by factors other than hard work are also more averse to redistribution. Finally, one of the most frequently cited articles in this area is written by Fong (2001). She examines the relationship between self-interested individuals and their judgment of income inequality. Her results show that the self-interest model of standard political economy theory only has a little effect on citizens their judgments on the level of income inequality (Fong, 2001). Therefore, she suggests that negative attitudes toward income inequality are likely to be influenced by normative values, such as deservingness perceptions and believes about what is fair. As a result, it can be safely claimed that individuals are likely to have different opinions on inequality due to social norms.

However, does this mean that perceptions on inequality also vary across nations? A study conducted by Taylor-Gooby and Martin (2010) indeed reports that ideas about fairness and social provision differ widely between countries. For instance, German citizens are much more concerned about social welfare policies than British citizens. Furthermore, also the literature on cross-national cultures claims that some societies are much more egalitarian than others (e.g. House et al., 2004; Schwartz, 2008). In particular, Schwartz (2008) suggests that important values in egalitarian cultures are equality and social justice, whereas more hierarchical cultures rely on productive behavior and define unequal income distributions as legitimate. According to Elster (1989), such egalitarian believes can be found particularly in democratic societies. Finally, it has been argued that also levels of social trust within countries is associated with the degree to which social welfare policies are implemented. In particular, several scholars argue that in countries with higher levels of social trust, citizens are more likely to disapprove inequality and support social welfare policies (e.g. Rothstein & Uslaner, 2005; You, 2012). However, it remains unclear if social trust leads to less economic inequality or the other way around (You, 2012).

(13)

13

Apart from these studies, many researchers used data from large international social survey programs to study different attitudes towards income inequality across nations. Kelley and Evans (1993) comparatively examine public beliefs about the legitimacy of income inequality in nine countries. They find that almost no one believes in complete equality. However, they demonstrate that there exists no consensus on how large the differences between low-skilled and high-skilled labor should be. In particular, they suggest that this disagreement is mainly about the compensation of high-skilled jobs and less so about the compensation of low-skilled jobs. Moreover, they show that citizens living in different countries have contrasting opinions on the level of inequality. For instance, it turns out that citizens of ex-communist countries are much more egalitarian than people in capitalist countries.

Also Alesina and Angeletos (2005) examine why support for redistribution differs across nations. They find that varying attitudes toward income inequality are related to different social perceptions across states regarding the fairness of market outcomes rather than perceptions about the true level of inequality. More specifically, they explain that data of the World Values Survey indicate that 60 percent of Americans believe that poor people could become rich if they try hard enough, while this is only 29 percent of European citizens. Instead, citizens from European countries seem to believe that luck rather than hard work affects economic success (Alesina & Angeletos, 2005). More importantly, they also demonstrate that desires for social fairness also influence political outcomes, as there exists a strong positive correlation between social spending and the belief that luck determines income. As a result, Alesina and Angeletos (2005) conclude that differences in perceptions toward inequality across countries may be understood on the basis of different initial conditions as well as historical events. These results are consistent with the findings of Linos and West (2003), who find that people in English-speaking countries are more likely to view their societies as mobile than German and Norwegian people. They argue that due to these differences, support for more redistributive policies is much lower in English-speaking countries.

The view of Alesina & Angeletos (2005) and Linos & West (2003) that perceptions on social mobility influence citizens their attitudes on income inequality is supported by the literature on meritocracy. Meritocracy is a specific form of a normative consideration that may influence public opinions on the level of income inequality (Marks, 2010). According to Marks (2010),

(14)

14

proponents of meritocracy argue that ability is the dominant influence on socioeconomic outcomes in contemporary societies. Therefore, in meritocratic societies, people believe that poverty is merely related to poor self-management (Giddens, 1998). As a result, people living in meritocratic countries tend to prefer their governments to evenly redistribute possibilities rather than wealth (Gidden, 1998). According to Kaplan & Kaplan (1997) increasing meritocratic beliefs have been fed primarily by the rise of education as a qualification for employment, due to industrial and technological and advancements. Therefore, supporters of meritocracy can be particularly found in developed countries.

However, meritocratic beliefs seem to vary not only between developed and developing countries, but also across developed nations. According to McCoy and Major (2007) meritocratic beliefs are particularly ubiquitous in North-American societies, as people in these societies are regularly exposed to the message that individual advancement is possible for everyone through hard work. Also Oorschot and Halman (2000) find that relative to European countries, Americans view individual agency as much more responsible for economic success. Nevertheless, cross-country examinations on the effect of meritocratic beliefs on income inequality are very scarce in the academic literature. Thus, beliefs in meritocracy can be found particularly in developed nations, however, the extent to which people believe in meritocracy also varies largely across high-income countries.

Finally, while most of the studies mentioned before focused especially on perceptions on inequality, Lübker (2004) did a cross-country examination on both the acceptance of income inequality and support for redistribution. By using data from the International Social Survey Program, he presents evidence that in (i) organized market economies, (ii) countries that undergo a transition, and (iii) developing countries more people disapprove the level of income inequality if the Gini-Index becomes higher. Conversely, in English speaking countries people seem to care much less about inequality if the Gini-Index increases (see figure 1).

(15)

15

Figure 1: Perceptions of domestic inequality as “too large” and Gini-Index (Lübker, 2004)

Also with regard to calls for redistributive policies of governments, countries seem to differ widely. While in Eastern European countries support for the idea that states have a duty to reduce income differences was relatively high (80% or more), this was relatively low (less than 50%) in Switzerland and non-European English speaking countries (Lübker, 2004). Moreover, Lübker (2004) demonstrates that these different opinions were not related to actual levels of income inequality (see figure 2). For instance, in Sweden the support for redistributive policies increased despite of a declining level of inequality, whereas in Austria support for redistribution decreased unless a sharp increase in inequality.

Thus, in addition to self-interest, also social norms and beliefs about social justice influence citizens their attitudes toward income inequality. It appears that such egalitarian vis-à-vis meritocratic beliefs also vary across countries. As a result, these beliefs affect the support for redistributive policies across nations, which makes that countries differ widely in the extent to which they have implemented redistributive policies.

(16)

16

Figure 2: Synopsis of trends in inequality and support for redistribution (Lübker, 2004)

2.3 Economic Factors Affecting Attitudes toward Income Inequality

Although a large number of studies examined the relationship between either inequality or social norms on public attitudes toward inequality, some scholars claim that these studies suffer from an omitted variable bias. According to Schmidt (2012) much of the variation in the acceptance of inequality is indeed likely to rely on both the ideas of homo economicus and homo sociologicus. However, he argues that also country-level characteristics should be incorporated into research on the acceptance of inequality. In particular, Schmidt (2012) finds in a sample of 29 European countries that unemployment rates are positively associated with public demand for redistributive policies. This is in line with Blekesaune (2007), who shows that lower employment rates are related to public attitudes toward a more equal distribution of income. Hence, these researchers claim that worries among citizens to lose their jobs may lead to public attitudes toward less income inequality.

In addition to unemployment rates, also the level of wealth of a country may influence the acceptance of income inequality. According to Haller, Höllinger and Raubel (1990) it is the purpose of most citizens to provide a sufficient standard of living to the whole society by redistributing income. However, they emphasize that when national wealth increases a saturation effect is likely to emerge, as the basic needs of the net receivers are already met, which makes that citizens will no longer disapprove the level of inequality.

(17)

17

Thus, while most researchers focused on the relationship between either perceived inequality or social norms and attitudes toward inequality, also economic factors may impact individuals their ideas on income inequality.

2.4 Gap in the Literature

Previous work indicates that citizens from different societies may have different opinions on income distributions due to varying perceived levels of income inequality, varying social norms and values among individuals and societies, and economic circumstances. According to standard political economy theory rising levels of inequality lead to more negative public attitudes toward inequality. Yet, this relationship might be affected by cognitive limitations, misinformation, and media effects. Moreover, it has also been argued that in some countries citizens care much more about inequality than in other countries due to egalitarian vis-à-vis meritocratic beliefs and economic circumstances. The conceptual framework exhibited in figure 3 summarizes these factors and their link to one another.

However, less is known about the nexus between attitudes toward inequality and support for redistribution. In fact, most researchers in the academic literature overlooked the idea that inequality aversion may not necessarily lead to a turn to the national government to demand more generous redistributive policies. An exception is Lübker (2004) who examined why calls for redistribution differ across countries. He demonstrates that calls for redistribution are not exclusively related to perceptions on the actual level of inequality. However, his piece of research focused almost exclusively developed countries. Moreover, he did not investigate why inequality averse citizens do not always actively pursue redistributive policies.

Consequently, it remains unclear which factors moderate the nexus between inequality aversion and support for redistribution. Therefore, it is the purpose of this research to examine which country-specific and individual-level characteristics moderate the nexus between inequality aversion and support for redistribution. For instance, it might be harder for people in developing countries to turn to their government than for people living in developed countries, as people in developed countries may have more political rights. Moreover, according to Alesina (1999) and Tanzi (1998a) developing countries have a number of weak fiscal features, which may hinder them to successfully implement redistributive tax policies. Since citizens know about these

(18)

18

institutional weaknesses, these cast doubt on their confidence in the ability of their governments to implement tax policies or social welfare spending to redistribute income (Chu, Davoodi, and Gupta, 2000). In turn, when confidence in national institutions reduces, support for nongovernmental solutions, such as private charity, increases, while support for direct redistributive measures diminishes (Kuziemko et al., 2015). As a result, the research question of this thesis is: why do citizens that disapprove the level of inequality not always turn to their government to demand more generous redistributive policies?

(19)

19

3. Theoretical Framework

This chapter shows how this research fills the research gap and discusses its place in the literature. In particular, this chapter explores from a country- and individual-level perspective which mechanisms may moderate the nexus between inequality perceptions and public calls for redistribution.

3.1 Institutional Quality

A large amount of intellectual energy has been devoted to examining the nexus between institutional quality on the one hand and income inequality and support for redistribution on the other hand. According to Rothstein, Samanni and Teorell (2012) historically inherited government institutions have played a key role in preferences to give the state responsibility for implementing redistributive policies. In particular, Rothstein and Uslaner (2005) argue that Scandinavian countries already made major changes in their government institutions during the first half of the 20th century which swept away corruption and led to the implementation of sound public policies. According to Rothstein and Uslaner (2005) these changes resulted in improved perceptions among citizens on their governments capability in handling public policies. These suggestions are supported by various empirical studies. First, Rothstein, Samanni and Teorell (2012) demonstrate that there is a positive effect of quality of government on a country’s expenditure on social insurance and social protection. Second, Pitlik and Kouba (2013) present evidence that the perceived level of institutional quality is also positively associated with preferences for government intervention. As a result, it can be safely claimed that people are only likely to support redistribution when they perceive their government institutions as effective and impartial (Rothstein, 2009).

However, less is known about the effect of government quality on the relationship between inequality aversion and demand for redistribution. An exception is Svallfors (2013) who examines if government quality can also affect the nexus between egalitarian beliefs and support for social welfare spending. In a sample of 18 European countries he finds support for his hypothesis that government quality moderates the relationship between egalitarian attitudes and support for more extensive welfare policies and higher taxes. Moreover, Svallfors (2013) highlights that especially Nordic countries, which are among the countries with the highest

(20)

20

institutional quality worldwide, can be taken as an example of countries in which citizens that disapprove inequality also show high levels of support for higher taxes and spending on social welfare. Conversely, in poorer countries with a low quality of government, the nexus between inequality aversion and support for social welfare programs is much weaker. Although these results clearly suggest that government quality affects the nexus between inequality aversion and support for redistribution, it must be mentioned that Svallfors (2013) his sample only comprised European countries and thus did not include a large number of developing countries.

Apart from a high quality of government, Rothstein (2009) also suggests that citizens are more likely to support redistributive policies when they view their government as impartial. Indeed, it appears from the literature that also the level of corruption, which is in fact an element of institutional quality, may affect the relationship between inequality aversion and support for redistribution (Kaufmann, Kraay, and Mastruzzi, 2010). In particular, Lambsdorff (1999) suggests that corruption is correlated with distortions in the execution of political policies. This view is shared by Mauro (1995), who finds that countries which are more corrupt have weaker institutions. Conversely, other researchers indicate that corruption does not necessarily weaken institutional quality, but rather takes place in societies with less well-established institutions. Rose-Ackerman (2008) emphasizes that corruption primarily takes place in countries with weak institutions where formal laws are not well-known and cannot be enforced. Additionally, Tanzi (1998b) suggests that effective supervisors and good auditing offices discourages government officials to engage in corrupt activities.

Thus, in countries with objectively weaker institutions, citizens fear that their financial resources won’t be distributed by the state in an effective and efficient way. As a result, inequality averse citizens in countries with weak institutions are less likely to turn to their government for redistributive policies. Therefore, I expect that:

H1: The nexus between inequality aversion and support for redistributive policies is stronger in countries with high-quality institutions than in countries with low-quality institutions.

A number of researchers claim that the quality of institutions in a country is associated with the level of trust that individual citizens have in their government. Thus, due to a weak objective institutional quality, individuals may come to distrust their government. For instance, Edlund

(21)

21

(1999) argues that Americans show lower levels of confidence in government because it is not functioning very well. This view is in line with Tanzi (1998a) and Chu, Davoodi, and Gupta (2000), who claim that institutional weaknesses such as weak fiscal policies and welfare programs are associated with less confidence in the working of their institutions. Further, Algan, Cahuc and Sangnier (2011) demonstrate that confidence in institutions among individual citizens is associated with more positive perceptions about the working of the welfare state.

Hence, in countries with weaker institutions individuals are likely to lose their faith in the capability of their government to effectively redistribute income. In turn, when citizens their confidence in institutions diminishes, they might also be less likely to demand redistributive policies. According to Edlund and Lindh (2013), confidence in public institutions among individuals is indeed positively associated with support for redistribution. However, less is known about a conditional effect of confidence in institutions on the nexus between income inequality aversion and support for redistribution. An exception on this is Kuziemko et al. (2015) who demonstrate, by means of an experiment within an American firm, that even when individuals their concern for inequality increases citizens who distrust their government remain reluctant to demand redistributive policies. Hence, they argue that low levels of confidence in the government among citizens impedes them from translating a concern for inequality into real support for redistribution.

Summing up, the academic literature is clear about a link between the institutional quality of a country and subjective trust that individual citizens have in their institutions. More specifically, citizens with lower levels of confidence in their government are less likely to put their financial resources in the hand of government officials for the purpose of redistributive policies. Hence, this research does not only focus on the objective institutional quality measured from a country-level perspective, but also incorporates subjective country-levels of confidence among citizens measured from an individual-level perspective. Therefore, I hypothesize that:

H2: The nexus between inequality aversion and support for redistributive policies is stronger among individuals showing trust to government institutions than among citizens showing distrust to government institutions.

(22)

22

3.2 Social Trust

In addition to trust in institutions, another mechanism that may influence the nexus between income inequality aversion and support for redistribution is the level of interpersonal trust in a country. In fact, a number of studies suggest that basic values relating to social justice have a strong effect on support for redistributive policies (e.g. Kulin, 2009; Kulin & Svallfors, 2013). With regard to trust, Rothstein & Uslaner (2005) emphasize that individuals who score high on social trust believe that they have a responsibility to provide possibilities for poorer citizens. They argue that citizens who believe that most people in their society will not harm their interest are also more likely to participate in politics and civic organizations and to be more tolerant to minorities in their society (Rothstein & Uslaner 2005). Similarly, Knack & Zak (2003) argue that interpersonal trust is negatively associated with social distance. They suggest that when trust increases also the cooperative norms of citizens extend to a wider radius of social groups which may also encompass citizens from different social classes. This view is in line with a number of scholars which argue that people with higher levels of social trust are more likely to get involved in their society (Putnam, 1993; Brehm & Rahn, 1997). More importantly, according to Uslaner (2016) social trust also affects activities in which people engage without having the expectation of getting reciprocal returns. As a result, he argues that social trust affects citizens their attitudes toward giving to charity and supporting policies that aim to redistribute income from richer people to poorer citizens (Uslaner, 2016).

A few scholars examine direct causality between interpersonal trust on the one hand and support for redistribution on the other hand. Berg and Bjørnskov (2011) demonstrate that countries with higher historical levels of trust are better able to finance higher government spending, as these countries face fewer problems with collecting tax. Moreover, Daniele and Geys (2015) present evidence, based on the European Social Survey, that citizens showing high levels of interpersonal trust are more likely to support redistributive policies when controlling for trust in public institutions. Moreover, by using a sub-sample of second generation immigrants, they confirm a causal nature from interpersonal trust to support for income redistribution. These findings are consistent with the findings of Uslaner (2008) who presents evidence that immigrants from Nordic countries living in the United States show higher levels of interpersonal trust than immigrants from low-trust countries.

(23)

23

Thus, in countries where citizens disapprove inequality but do not trust others in their society, they are less likely to support redistributive policies, because a norm of reciprocity is absent. Indeed, when looking at data from the World Values Survey (2014), it appears that in many countries with low levels of social trust, citizens who disapprove inequality are less likely to demand redistributive policies from their government compared to citizens living in countries with high levels of social trust (see table 1). For instance, in Brazil, Rwanda, and Cyprus the scores on inequality aversion are relatively high compared to the average score among 59 countries worldwide. However, social trust is below average in these three countries. In such a scenario people are less likely to demand redistributive policies, because they do not want to share with people in their society at personal cost. Indeed, in countries with similar levels of inequality aversion but much higher levels of social trust – such as Sweden, China and Estonia – support for redistribution is much higher. Hence, in countries that score high on social trust the nexus between inequality aversion and support for redistribution might be much stronger due to more deeply held norms of reciprocity and conditional obligations to other people in their society, which makes them more involved in their society. As a result, I expect that:

H3: The nexus between inequality aversion and support for redistributive policies is stronger among individuals showing trust to their society than among citizens showing distrust to their society.

Table 1: Impact of Social Trust on Support for Redistribution

1_{The total sample covers 59 countries worldwide (see table 2).}

Countries Social Trust Inequality Aversion Support for Progressive Tax Brazil Rwanda Cyprus 7.1% 16.6% 7.5% 4.91 5.73 6.52 4.51 4.97 5.68 Sweden China Estonia 60.1% 60.3% 39.0% 5.12 5.55 6.21 6.42 7.29 7.15

(24)

24

4. Methods

This study follows an inductive approach. The analysis started with collecting data that has been relevant to the research topic: perceptions on inequality and support for redistribution in developing and developed countries. Data on this has been derived from the World Values Survey 2010-2014 (WVS, 2014). As expected, initial insights into this data made sure that the nexus between inequality aversion and support for redistribution is indeed stronger in developed countries than in developing countries. Therefore, it is the purpose of this thesis to explore why these differences among citizens in developed and developing countries exist. The main research method for this comprises an analysis of quantitative data derived from several sources which are discussed in the sections below. Hereafter, the methodology of the statistical analyses is addressed.

4.1 Data

The main data source for this research is the World Values Survey 2010-2014, which is produced by the Institute for Social Research at the University of Michigan. This survey covers more than 250 questions about what individuals around the world value in life. More than 90,000 respondents across 59 nations from all continents filled in this survey. Therefore, the issue of selection bias is likely to be avoided. Answers to various questions within this survey are used to measure the independent variable, the individual-level moderating variables, the dependent variable and the individual-level control variables. Within the sample, 24 countries are considered by the World Bank (2018a) as “high-income countries”, hence these countries are considered as developed countries, while the remaining countries are treated as developing countries (see table 2). In addition to data extracted from the World Values Survey, data from a variety of sources is used to measure the country-level variables, which the Freedom House, Transparency International and the World Bank. All variables are further discussed in the next section.

(25)

25

Table 2:Developed and developing countries covered in this research

Developed Countries Developing Countries

1. Argentina 1. Algeria 2. Australia 2. Armenia 3. Chile 3. Azerbaijan 4. Cyprus 4. Belarus 5. Estonia 5. Brazil 6. Germany 6. China

7. Hong Kong 7. Colombia

8. Japan 8. Ecuador

9. South Korea 9. Egypt 10. Kuwait 10. Georgia 11. Netherlands 11. Ghana 12. New Zealand 12. India

13. Poland 13. Iraq 14. Qatar 14. Jordan 15. Romania 15. Kazakhstan 16. Russia 16. Kyrgyzstan 17. Singapore 17. Lebanon 18. Slovenia 18. Libya 19. Spain 19. Malaysia 20. Sweden 20. Mexico 21. Taiwan 21. Morocco 22. Trinidad & Tobago 22. Nigeria 23. Uruguay 23. Pakistan 24. United States 24. Palestine

25. Peru 26. Philippines 27. Rwanda 28. South Africa 29. Thailand 30. Tunisia 31. Turkey 32. Ukraine 33. Uzbekistan 34. Yemen 35. Zimbabwe

(26)

26

4.2 Variables

The main dependent variable in this research is support for redistribution. The mean of answers to two statements of the World Values Survey are used to measure support for redistribution. These statements are: (i) “Governments should tax the rich and subsidize the poor”, and (ii) “People should receive state aid for unemployment” (WVS, 2014 p. 9). Respondents had to state to what extent they agree with the given statements on a 10-point scale. These two statements are chosen because both a more progressive tax system and more social security can be used by national governments as effective measures to redistribute income. In addition, to assess the robustness of the results of the main statistical analyses and to check if the results are different for each of these statements, the answers to these questions are also used as separate variables in chapter 6. I call these separate variables subsequently (i) support for progressive tax and (ii) support for social security.

With regard to the independent variable, one question within the World Values Survey asked respondents to indicate to what extent they believe that income inequality should be accepted. More precisely, respondents were asked to indicate on a 10-point whether they agree with the statement “incomes should be made more equal” on the one hand or “we need larger income differences as incentives for individual effort” on the other hand (WVS, 2014, p. 7). Hence, I use the answers given to this question to measure to what extent respondents disapprove income inequality and call this variable inequality aversion. For ease of interpretation, I reversed the scale, so that a 10 indicates that a respondent is extremely inequality averse. After reversing the scale, it turns out that 53% of the respondents filled in a number between 6 and 10, which implies that a small majority of respondents within this sample is inequality averse (WVS, 2014).

As mentioned in the theoretical framework three key hypotheses are tested in this research. All of them involve a conditional effect from either a country- or individual-level perspective. To measure the conditional effect mentioned in hypothesis 1 and 2, multiple variables are used in order to achieve more reliable results. With regard to my hypothesis 1, three country-level variables are used to estimate the institutional quality. According to the World Bank, six broad dimensions can be used to measure institutional quality, which are voice and accountability, political stability and absence of violence/terrorism, government effectiveness, regulatory quality, rule of law, and control of corruption, which they call the Worldwide Governance Indicators

(27)

27

(Kaufmann, Kraay and Mastruzzi, 2010). However, Kaufmann, Kraay and Mastruzzi (2010) indicate that each of these indicators provide highly specific and disaggregated information that might be of great interest for researchers. Therefore, I use only three indicators which, according to the literature, are likely to have an effect on the nexus between inequality aversion and demand for redistribution.

The first indicator to measure institutional quality is government effectiveness, which measures the ability of government to effectively formulate and implement sound policies (Kaufmann, Kraay and Mastruzzi, 2010). More specifically, this indicator captures the perceived quality of the civil service, public service and the credibility of the government’s commitment to implement such policies. As it has been argued that particularly the government’s capability in handling public policies is likely to affect demand for redistribution, this indicator is of great interest for this research (e.g. Rothstein & Uslaner, 2005; see section 3.1). Hence, I use the most recent government effectiveness scores of 2016 obtained from the Worldwide Governance Indicators dataset as a first construct to measure institutional quality. Data within this dataset is provided by surveys distributed among a large number of firms and citizens as well as expert assessments and covers all countries within my sample (Kaufmann and Kraay, 2017). Scores on government effectiveness are scaled from 0 (weak) to 5 (strong) and cover all countries within my sample. In addition, it has been argued that also the level of corruption may play a role in affecting citizens their perceptions on the implementation of redistributive policies. According to Rothstein (2009), citizens who view their government as impartial are more likely to prefer redistributive policies. Moreover, the World Bank as well as several researchers suggest that control of corruption is an important element of institutional quality, as this captures the respect of citizens for the institutions that govern economic interactions (e.g. Kaufmann, Kraay, Mastruzzi, 2010; Mauro, 1995).Therefore, I use the level of corruption as a second construct to measure institutional quality. Data on this is collected from Transparency International (2018), because their corruption indicator entails a secondary rating system that measures solely the perceived levels of corruption in the public sector. More precisely, Transparency International measures, by surveying experts and business executives, the extent to which public officials are believed to take bribes, engage in fraudulent practices and try to misuse public money (Sung, 2004; Transparency International, 2018). The corruption scores of Transparency International are

(28)

28

scaled from 0 to 10 and cover all countries within my sample. For ease of interpretation the scores on corruption are reversed, meaning that a 10 refers to extremely corrupt and a 1 to extremely non-corrupt.

Further, Kaufmann, Kraay and Mastruzzi (2010) also suggest that voice and accountability is an important element of institutional quality. According to them, voice and accountability refers to perceptions on the extent to which citizens are able to participate in selecting their government. Hence, this indicator in fact measures the degree to which a country is democratized. Apart from the World Bank, the level of democratization is also measured by the Freedom House (2018a). In contrast to the World Bank, the Freedom House takes a more holistic approach in measuring democracy by using questions about both political rights and civil liberties. In particular, the Freedom House uses questions about citizens’ rights to vote, the exercise of sovereignty within national borders, the existence of a multi-party system with meaningful opposition, and the efficacy of elected politicians amongst others to measure democracy (Karatnycky, 2001). The democracy scores are produced each year by more than one hundred analysts consisting of academics, think tanks, and human right communities (Freedom House, 2018b). Democracy scores are measured by an ordinal variable from 1 (lowest) to 7 (highest) and cover again all countries within my sample.

Also the conditional effect of hypothesis 2 is measured by various variables. However, in contrast to my hypothesis 1, hypothesis 2 takes a micro-level perspective. Therefore, only respondent specific scores obtained from the World Values Survey 2010-2014 can be used to measure trust in institutions among individuals. In the World Values Survey, respondents are asked to tell how much confidence they have in a number of organizations on a scale from 1 (a great deal) to 4 (none at all). In fact, four of these organizations refer to government institutions that may have an impact on either the formulation or execution of redistributive policies, which are: (i) political parties, (ii) the parliament, and (iii) the civil service. Scores on all these three questions are reversed for ease of interpretations, so that the score increases if confidence rises.

With regard to hypothesis 3, one variable is used to measure the degree of social trust. Since also the interaction effect of hypothesis 3 is taken from a micro-perspective, again data from the World Values Survey (2014) is used. In particular, responses to the question “Generally speaking, would you say that most people can be trusted or that you need to be very careful in

(29)

29

dealing with people?” are used to measure social trust. In fact, answers to this question are used by many frequently cited scholars in the social literature to measure social trust, which makes this construct likely to have a high validity in measuring social trust (e.g. Brehm & Rahn, 1997; Kawachi, Kennedy and Glass, 1999). Answers to this question are reversed for readability. This way social trust becomes a binary variable that equals 1 if respondents say that they need to be very careful and 2 if respondents say that most people can be trusted.

With regard to the control variables, a set of country- and individual-level indicators are included in this research for each regression model to avoid omitted variable bias. Two macro-level control variables are incorporated into this research, which are GDP per capita and government expenditure. According to Dallinger (2010), GDP per capita may negatively affect support for redistribution, because a saturation effect emerges when the basic needs of receivers of public redistribution policies are met. Conversely, government expenditure may have a positive effect on support for redistribution, as support for more generous redistributive policies is generally stronger in countries were government spending on social welfare programs is heaviest (Esping-Andersen, 1980). For both variables data is collected from datasets of the World Bank (2018b; 2018c). GDP per capita refers to the gross domestic product in US dollars divided by the population of a country (World Bank, 2018b). Government expenditure refers to payments for operating activities of the government in providing goods and services measured as a percentage of a country’s GDP (World Bank, 2018c). It includes the compensation of employees and social benefits among others. For both variables, the most recent estimate of the past five years is taken for each country within my sample. GDP per capita is divided by 1,000 for ease of presentation. Finally, the micro-level control variables are again extracted from data of the World Values Survey. These include the respondent’s gender, and responses to questions about the highest level of education attained by the respondent and the income group of the respondent’s household. Table 3 provides a description of the variables and their sources considered in this study. Gender is chosen because it has been argued that women are generally more supportive of redistributive policies than men (Estevez-Abe, Iversen, and Soskice, 1999). Income group and education are chosen because these elements may improve social security of respondents and thus reduce their support for redistribution (e.g. Meltzer & Richard, 1981; Linos & West, 2004).

(30)

30

Table 3. Definition of variables.

Variable Definition Source

Dependent Variables

Support for Redistribution Support for Progressive Tax

Support for Social Security

Independent Variable Inequality Aversion Moderating Variables Government effectiveness Corruption Democracy Confidence in Political Parties Confidence in the Parliament

Confidence in the Civil Service Social Trust Control Variables GDP per capita Government expense Gender Income Group Education

Ordinal variable that reflects the mean score of Support for Progressive Tax and Support for Social Security 10-category ordinal variable that comprises answers to “Governments should tax the rich and subsidize the poor” 10-category ordinal variable that comprises answers to “People should receive state aid for unemployment”. 10-category ordinal variable that comprises answers to the statement of “incomes should be made more equal” vs. “we need larger income differences as incentives for individual effort”

Interval variable ranging from -2.5 (weak) to 2.5 (strong) to estimate governance performance

Interval variable ranging from 1 (low) to 10 (high) to estimate the level of corruption in the public sector 7-category ordinal variable ranging from 1 (low) to 7 (high) to estimate the level of democracy

4-category ordinal variable that comprises answers to “How much confidence you have in political parties” 4-category ordinal variable that comprises answers to “How much confidence you have in the parliament” 4-category ordinal variable that comprises answers to “How much confidence you have in the parliament” Binary variable that equals 1 if1 if respondents say that they need to be very careful and 2 if respondents say that most people can be trusted.

Continuous variable that reflects the GDP per capita in 1.000s of US$

Interval variable that reflects the Government Expense as a percentage of a country’s GDP

Binary variable that equals 1 if the respondent is male and equals 2 if the respondent is female

10-category ordinal variable that comprises answers to “We would like to know in what group your household is” 9-category ordinal variable that comprises answers to “What is the highest educational level that you have attained?”

World Values Survey (2014) World Values Survey (2014) World Values Survey (2014)

World Values Survey (2014)

Kaufmann and Kraay (2017) Transparency International (2018) Freedom House (2018)

World Values Survey (2014) World Values Survey (2014) World Values Survey (2014) World Values Survey (2014)

World Bank (2018b) World Bank (2018c) World Values Survey (2014) World Values Survey (2014) World Values Survey (2014)

(31)

31

4.3 Methodology

In this thesis, two linear mixed effect models are performed for each moderating variable. This statistical model has the advantage that it can deal with repeated measurements that are made on the same statistical units. Hence, with regard to this study, respondents could be sampled from within countries (Smith, 2012). The first model tests the direct relationship of the independent variable and each specific moderating variable on the dependent variable support or redistribution, while including the country random effect as well as the control variables. The second model adds the interaction effect of each specific moderating variable.

(32)

32

5. Results Statistical Analyses

This chapter presents the results of the statistical analyses. In the next section, the descriptive statistics of each variable are given. Hereafter, the results of each hypothesis proposed in chapter 3 are discussed.

5.1 Descriptive Statistics

The descriptive statistics of all variables are exhibited in table 6. The number of observations of the independent variable, dependent variable, and individual-level moderating and control variables differs per indicator. This is due the fact that these variables are derived from questions of the World Values Survey (2014) where respondents could indicate that they either did not know an answer to the question or did not want to answer the question. Additionally, also the number of observations of the macro-level control variables differ. This is due to the fact that data on GDP per capita and government expenditure have not been available for each country within my sample. For instance, the World Bank does not provide data for Taiwan, as they do not consider Taiwan as an independent country.

The descriptive statistics indicate that on average the respondents support the introduction of more redistributive policies. In particular, support for social security seems to be stronger than support for a progressive tax system. In line with these scores, also the majority of people disapprove actual levels of inequality. However, the disapproval of inequality appears to be weaker than support for redistributive policies. Further, most countries considered in this research have relatively good scores for government effectiveness and the level of democracy. Surprisingly, the mean score for corruption is above five, meaning that countries within this sample score relatively high on the level of corruption. Average scores for confidence in government, the parliament, political parties, and the civil service appear to be somewhat in between “not so much confidence” and “quite a lot of confidence”. However, confidence in the civil service is much lower than confidence in political parties and the parliament. Finally, the mean of social trust is almost exactly between 1 and 2, meaning that respondents have a moderate level of trust in other people in their society. With regard to the control variables it is particularly remarkable to notice that the majority of respondents were female.

(33)

33

Table 4: Descriptive Statistics

Variable Mean S.D. Min Max Skewness Kurtosis N

1. Support for Redistribution

2. Support for a Progressive Tax System 3. Support for Social Security

4. Inequality Aversion 5. Government Effectiveness 6. Corruption

7. Democracy

8. Confidence in Political Parties 9. Confidence in the Parliament 10. Confidence in the Civil Service 11. Social Trust

12. GDP per Capita (in 1.000s of US$) 13. Government Expense (as % of GDP) 14. Gender 15. Education 16. Income Group 6.65 6.29 7.02 5.68 2.75 6.36 4.50 2.91 2.72 2.44 1.52 15.02 24.87 1.75 5.65 4.86 2.37 2.98 2.74 2.97 0.95 1.91 2.15 0.94 0.89 0.89 0.50 17.32 8.54 0.43 2.43 2.11 1 1 1 1 0.61 2.1 1 1 1 1 1 0.7 2.4 1 1 1 10 10 10 10 4.71 9.4 7 4 4 4 2 59.32 40.3 2 9 10 -0.40 -0.38 -0.68 -0.02 0.17 -0.63 -0.37 -0.20 -0.42 -0.03 -0.08 -1.32 -0.36 -1.16 -0.24 0.01 -0.54 -1.05 -0.56 -1.18 -0.36 -0.69 -1.35 -0.88 -0.60 -0.77 -1.99 0.30 -0.30 -0.66 -0.98 -0.49 84505 85408 85835 86311 89150 89150 89150 82843 84706 83544 89059 85781 75124 86747 88313 86255

(34)

34

5.2 Institutional Quality

The results of the direct and conditional effect of institutional quality on support for redistribution are shown in table 5. It turns out that all three macro-level variables do not have a statistically meaningful direct effect on support for redistribution. This result is not surprising given the fact that the scores of these variables only vary between countries and thus not among respondents. This substantially increases the standard error of these effects, which in turn leads to higher p-values. Conversely, the estimate of the direct effect of inequality aversion on support for redistribution is positive in each model and appears to be around 0.065 (statistically significant at 1%). Moreover, it turns out that all three moderating variables affect the nexus between inequality aversion and support for redistribution.

In particular, government effectiveness has a significant positive moderating effect on the nexus between inequality disapproval and demand for redistribution. The estimate of the direct effect of inequality aversion on support for redistribution is 0.065, while the interaction effect of inequality aversion and government effectiveness is 0.107. Both effects are significant at a 1 % confidence level. This implies that when government effectiveness is at the theoretically lowest level (0), the effect of inequality aversion on support for redistribution is 0.065. But when government effectiveness is at the highest level (5), the effect of inequality aversion on support for redistribution becomes 0.065 plus 0.107 multiplied with 5, which is exactly 0.6. This means that inequality averse people are much more likely to turn to their government for redistributive policies when their governments are more effective in implementing public policies. This result is consistent with the findings of Svallfors (2013), who investigated if quality of government moderates the relationship between egalitarian beliefs and support for welfare policies. While Svallfors (2013) focused only on European countries and concentrated on preferences for welfare state spending, this result demonstrates from a much larger sample that government effectiveness also has a conditional effect on support for redistribution.

With regard to the level of corruption, the interaction effect appears to be negative. More specifically, the estimate of the interaction effect is -0.115 (statistically significant at 1%). This implies that citizens who want incomes to be distributed more equally are less likely to turn to their government for more generous redistributive policies when the level of corruption in the public sector is high. An explanation for this finding is that high levels of corruption leads to a

Inequality aversion and demand for redistribution among citizens in developed and developing countries