Economic and cultural determinants of elite attitudes towards redistribution


Texto completo

(1)Three Studies on Inequality and Democracy: Elites, States, and Bounded Rationality in Latin America and Beyond. Matias López. Tesis de Doctorado en Ciencia Política. 2019. Instituto de Ciencia Política Pontifícia Universidad Católica de Chile Comité: Prof. Juan Pablo Luna (Tutor) Prof. David Altman Prof. Frances Hagopian.

(2) Study# 3: Economic and Cultural Determinants of Elite Attitudes Towards Redistribution (with Graziella Moraes Silva1, Chana Teeger2, Pedro Marques3). Abstract Previous studies posit that elites are willing to advance the redistribution of income and social goods when negative effects of inequality, such as crime and conflict, threaten their own safety. Nonetheless, elites’ willingness to tackle inequality seems to be low throughout the Global South in spite of such extreme negative consequences. Drawing on 56 in-depth interviews and closed-ended surveys with 543 political and economic elites, we document how elites’ perceptions of the poor mediate their response to threats resulting from inequality. In-depth interview data reveal that, in explaining the challenges of redistribution policies, respondents consistently characterize the poor as ignorant, uninformed, and irrational, in opposition to rational, maximizing elites. We theorize that these perceptions of the poor hinder elites’ willingness to support redistribution, even when redistribution suits protection needs. We test this relationship in random samples of elites in Brazil, South Africa, and Uruguay and find it to be robust. Keywords:​ Elites, Inequality, Perceptions, Social Policy, Redistribution JEL classification​: D71; ​D620; Z100. Introduction. 1. Graduate Institute of Geneva London School of Economics and Political Science 3 Instituto de Pesquisa Econômica Aplicada 2.

(3) Scholars of inequality have good reason to care about elites’ preferences. Elites have the ability to shape the politics of redistribution , owing to the power positions that they occupy (Higley and Burton, 2005; Reis and Moore, 2005; Hoffmann-Lange, 2007, 2018; Khan, 2012; López, 2013a; Yamakoshi and Dubrow, 2012). Existing scholarship suggests that elites will support more redistribution when their returns are affected by the negative consequences of inequality. For instance, numerous studies document the connection between inequality and criminal violence (Bourguignon et al., 2003; Choe, 2008; Fajnzylber et al., 2008), the latter of which can affect elites’ safety. Further, inequality poses threats to elites in the form of uprisings and political violence (Acemoglu and Robinson, 2005; Boix, 2003). Therefore, given the proper incentives, supporting redistribution can be an avenue for elites to protect themselves from the poor (de Swaan, 1988; Rueda and Stegmueller, 2016). Nevertheless, previous work in the Global South has found that, although elites acknowledge the negative consequences of inequality for themselves, on average, they have low commitment to redistribution initiatives (de Swaan et al., 2000; Reis and Moore, 2005; Hossain, 2005; Clarke and Sisson, 2005). This occurs even in contexts of relatively high state capacity, where elites acknowledge that tools with which to fight inequality either exist or could be developed (Reis and Moore, 2005). Further, it is true even for political elites who, in theory, should be responsive to the redistributive aspirations of the median voter (see Luna, 2014). What explains elites’ weak support for redistribution policies in contexts where negative consequences of inequality affect their own well-being? We address this question drawing on data from in-depth interviews and closed-ended surveys of Brazilian business, political, and civil service elites. Brazil is a highly unequal country where externalities of poverty and inequality are extreme, particularly in the form of urban violence. Studies have shown that Brazilian elites recognize inequality as being one of the most important national problems and identify urban violence as being one of its most important consequences (Reis and Moore, 2005; Reis, 2000; López, 2013b, 2014, 2016). Yet redistributive policies are still far from consensual (Paugam et al., 2017). By relying on a multi-method design, we answer Cousin et al.’s (2018) recent call for greater methodological diversity in elite research. Through the analysis of elite in-depth interviews in Brazil, we unpack elites’ attitudes toward inequality and its consequences, building an argument about why and how perceptions mediate the relationship between external incentives and redistributive outcomes. The argument is that elite support to redistributive policies is positively affected by their perception of externalities but negatively affected by their evaluation of the poor, which elites tend to see as ignorant and irrational, and thus politically and economically incompetent. We test such argument using OLS regressions in a data set drawn from a random.

(4) sample of Brazilian elites. Finally, to assess the external validity of our argument, we replicate the analysis with data gathered from random samples of elites in South Africa and Uruguay. In pointing to the impact of perceptions of the poor on elites’ attitudes toward redistributive policies, we identify cultural processes as the missing link between economic incentives, created by the negative effects of inequality, and subsequent policy outcomes. The idea of culture as a mediator has been promoted by both political economists (Alesina and Giuliano, 2009) and cultural sociologists (Lamont et al., 2014), but rarely addressed to account for elites’ action. A number of studies have addressed the connection between perceptions and different attitudes towards redistribution, both among the general population (Cavaillé and Trump, 2014; McCall and Kenworthy, 2009; Niemelä, 2008; Oorschot, 2006; Oorschot and Halman, 2000) and among elites (de Swaan et al., 2000; Hossain, 2005; Reis and Moore, 2005). Yet none, to the best of our knowledge, estimate the effect of both external incentives and individual perceptions on elites’ willingness to support redistributive social policies. Bridging political economy and cultural sociology, our study highlights how cultural processes mediate the relationship between material incentives and policy outputs. In what follows, we first review the literature on elites’ attitudes toward redistribution, including rival arguments, and present our theoretical contribution to understanding the economic and cultural constraints that shape these attitudes. Moving to our methods and data, we describe our criteria for case-selection, sampling, coding, and identification strategies. The results section we analyze the in-depth interviews through which we build our theoretical contribution and then proceed to model elites’ response with OLS regressions in Brazil, South Africa, and Uruguay. We close the paper by discussing the implications of our findings to the understanding of patterns of distribution in the Global South.. Explaining elites’ attitudes toward redistribution The bulk of the political economy literature on elites and redistribution assumes that elites will always try to prevent redistribution. Resource concentration is, after all, a defining characteristic of the elites and elites themselves are often held responsible for these inequalities, through their tax evasion, labor market, and state-capture practices. Under that assumption, scholars highlight how elite driven political processes (e.g. democratization) that should result in redistribution can be understood as undesired and limited concessions to the poor vis-à-vis threats from below (Acemoglu and Robinson 2005, Boix 2003). Following similar assumptions, others conclude that redistribution ultimately fails because it was never the elites’ actual intention to redistribute, but only to signal redistribution in order to pacify the poor. For instance, Soifer (2013) posits that elites support democratization when they anticipate that inefficient states will fail to effect.

(5) redistribution. But that argument does not hold for many middle-income countries that have shown relatively high state capacity and infrastructural power, such as the ones we analyze in this paper. Other studies argue that elites purposefully promote such inefficiency by placing oversized bureaucracies, or by embedding institutional roadblocks to redistribution in the constitution prior to democratization (Acemoglu et al., 2011; Albertus and Menaldo, 2013, 2018). Such an argument also does not applies to our cases, which all have constitutions written during democratic times. In contrast, others argue that elites’ decision-making process is hardly influenced by threats from below, and that apparent concessions to the poor more regularly occur for reasons that are completely unrelated to redistributive demands. Those reasons include tension between urban economic elites and landed elites (Ansell and Samuels, 2014), constraints posed by international actors (Haggard and Kaufman, 2016), and splits within ruling coalitions (Haggard and Kaufman, 2016; O’Donnell et al., 2013). While compelling, these arguments largely ignore a growing body of empirical work that documents how elites systematically identify poverty and inequality as the main causes of crime and violence, which negatively affect their own lives (Reis and Moore, 2005; Rueda and Stegmueller, 2016). Moreover, elites frame inequality as a threat to democracy, and as fuel for political practices that they find undesirable, such as clientelism (López, 2014). One way of bridging these two perspectives is by suggesting that elites do not reject redistribution, but simply do not want to pay for it. Elites then are not avoiding redistributive social policies per se, but rather taxes. Nonetheless, previous studies show that elites have been successful in raising state revenue without paying for it, often relocating fiscal costs to the middle-class (Bogliaccini and Luna, 2016; Fairfield, 2013; Hacker and Pierson, 2010; Winters, 2011). If elites identify negative externalities of poverty and have the capacity to put forward effective social policies without necessarily bearing fiscal costs themselves, what prevents them from effectively coordinating around redistribution? Studies in political culture provide a different approach to this question. They contend that elites’ attitudes toward redistribution are determined by overarching values, norms, and political ideology (e.g. Putnam, 1971; Pye and Verba, 2015; Verba and Orren, 1985; Verba et al., 1987). Other studies highlight how elites’ attitudes toward redistribution are shaped by normative evaluations the poor (who can be seen as either deserving or undeserving) and of social policies (which can be seen as either fair or unfair) (Alesina and Angeletos, 2005; Hochschild, 1981; Katz, 2013). Largely based on the US case, this literature often implies that elites use moral justifications to endorse actions that meet their material interest, i.e. income concentration. Nevertheless, the lack of redistribution in many countries in the Global South does not seem to operate through perceptions of undeservingness or fairness. In Latin America (López 2016),.

(6) Africa (Kalanti and Manor, 2005; Kalebe-Nyamongo, 2012), and South Asia (Clarke and Sison, 2005), elites are generally aware of the unfairness of inequality as well as of its structural causes, and consequent deservingness of the poor. There too, however, elites do not fully embrace redistributive social policies. Focusing on the Global South, Reis and Moore’s (2005) influential study suggests that it is elites’ sense of responsibility that accounts for different distributive outcomes. They claim that incentives to redistribute are often present in the Global South, but that elites deposit responsibility to the state, do not engage personally in redistributive efforts, and consequently compromise the redistribution to which they often aspire. Other studies have shown that elites are unlikely to cooperate when they do not think that others will share the costs of redistribution (Feierherd et al., 2017) or when they question the reliability of the institutions charged with the task of overseeing redistribution (Berens and Schiller, 2017). Taken together, these studies suggest that elites’ actions are largely determined by their perceptions of what others might do. In sum, past research shows good reasons for why both economic and cultural processes should impact elites’ attitudes toward social policies, and consequently the shape of income distribution itself. Nonetheless, the two bodies of literature do not dialogue and, perhaps because of that, fail to predict and explain the patterns of elite action observed in middle-income countries. Brazil and South Africa, as many other cases in the Global South, are unequal democracies where elites suffer the effects of acute externalities of poverty, largely frame social policies as fair, and see the poor as deserving. Multiple theories and models predict that more redistribution should therefore take place, yet decades after democratization, income distribution continues to be highly skewed. This occurs, as we will show, because perceived incentives often go in different directions. The theoretical contribution of our paper lies in ​highlighting the cultural mechanisms that mediate the relationship between material (or economic) incentives and policy outputs. Based on the inductive analysis of in-depth interviews with Brazilian elites, we conceive that elites are indeed responsive to the negative externalities of poverty and inequality, which prompt them to consider redistributive measures as means to secure their own position. Yet in parallel, we identified that elites tend to see the poor as irrational and unable to make informed economic and political decisions, an evaluation that they associate with different narratives of why policies may ultimately fail to provide the security to which they aspire. Those narratives include skepticism regarding the poor’s capacity to lift themselves out of poverty, even with governmental assistance, and suspicion that political elites could take a vantage of the poor’s ignorance to build poverty traps for their own electoral benefit. We thus hypothesize that, as elites perceive the poor as ignorant and irrational, they will tend to be much less supportive of redistributive policies, regardless of the incentives offered by negative externalities..

(7) We acknowledge that there are other important structural and attitudinal variables that should impact upon elites’ attitudes towards redistribution. As documented above, other studies have identified tax avoidance, political culture (e.g. egalitarian vs liberal ideologies), and perceptions of personal or group responsibility as crucial in determining elites’ attitudes. We view these explanations as complementary to our own and we included covariates that account for them in our regression models. Our findings are also reconcilable with the large body of literature in political economy that points to the role of distrust in explaining negative attitudes towards social policies (Alesina and Giuliano, 2009; Axelrod, 1984; Berens and Schiller, 2017; Tabellini, 2009). In our account, perceptions of the poor lead elites to be suspicious of the actions of others, thus hindering coordination around redistribution. We argue that the perception of the poor as irrational triggers this distrust. Building on past research, we provide a theoretical contribution that solves these important contradictions in the literature. We do so based on the inductive analysis of what elites themselves have to say about the dilemmas of redistribution. We then test the main implications of this contribution using closed-ended surveys with elites.. Methods Our study draws on two different types of data (in-depth interviews and survey data) and three cases (Brazil, South Africa and Uruguay). Below, we introduce our cases and then detail our data collection and analysis procedures. Cases The bulk of our study focuses on Brazil, which represents a case of high inequality and concomitant strong negative consequences for elites, such as crime and political instability. We build on the case of Brazil because it provides a scenario in which we should observe more consistent support for redistributive policies, given the durable effects of negative externalities in recent decades. We then test our findings in two additional cases: South Africa and Uruguay. Brazil democratized in the late 1980s, opening to competitive politics in a context of extreme fiscal chaos in government, criminal violence in urban areas, land invasions in rural areas and the constant threat of a military coup (Hagopian, 2007). Redistribution was perceived to be in elites’ best interest as it would help to mitigate dramatic externalities (Reis, 2000). Indeed, Brazilian politics was dominated by an agenda of inequality and social justice (Weyland, 1996). In the early 2000s, the left-wing Workers’ Party (PT) came into power and, helped by favorable economic times, promoted massive policies of social inclusion and distribution (Campello and.

(8) Zucco 2016). At first, the business community and other elite sectors endorsed PT’s redistributive agenda and Brazil experienced a decline in income inequality. The expansion of cash grants and a real increase in the minimum wage played a significant role in this decline (Azevedo and Sanfelice 2013; Barros et al. 2010; Ferreira et al. 2014; Lusting et al. 2016), but it was dependent on a scenario of economic growth and high commodity prices (Carvalho 2018). When economic recession brought the need for a more substantial redistributive agenda, support from the business elite faded and inequality has been on the rise in recent years (Lavinas 2017). Decades after democratization, the country remains one of the most unequal in the globe and both poverty and its externalities remain unsolved. As elaborated below, we built our model of the relationship between perceptions of the poor and support for redistributive policies out of findings emerging from in-depth interviews with Brazilian elites and then tested in in a larger, randomly-drawn, sample. We moved beyond the Brazilian case by further testing the model in two other cases: South Africa and Uruguay. We view these as shadow cases, in the sense that these are not cases analyzed in-depth (see Gerring and Cojocaru, 2016).. In the comparative literature, South Africa is often positioned as similar to Brazil as a result of the importance of racial cleavages in both countries, their authoritarian histories, and their economic positions in their respective regions (Lieberman, 2003; Marx, 1998; Moraes Silva, 2012). In South Africa, as in Brazil, it is common to frame inequality and poverty as the main contributors to social problems that affect the non-poor, such as crime (Teeger, 2014). Nevertheless, in the past few decades, redistribution policies in South Africa have been limited and overall inequality, as measured by the Gini coefficient, has remained one of the highest in the world. Given the similar scenario, we expect patterns of elite perceptions in South Africa to be similar to those found in Brazil. Uruguayan elites, on the other hand, have also dealt with increasing crime rates and other problems that arose alongside income disparities after democratization, but coordination around redistribution seems to have been much more effective there than in the other two cases. The country is now among the most equal and safest in the developing world (Brinks, 2007; PNUD, 2013). We thus expect the patterns of elite perceptions to go in the opposite direction to those of their Brazilian and South African counterparts.. Interviews.

(9) Theory-building follows the analysis of 56 in-depth interviews with Brazilian state and market elites. Of those, 42 were conducted in 1998 and 1999 by Reis and colleagues and have been made available to us for analysis (see Reis, 2000, 2005). We supplemented this dataset with an additional 14 interviews which we conducted between 2011 and 2013. In-depth interviews took place with business, political and civil service elites in the cities of São Paulo, Rio de Janeiro, Fortaleza and Salvador. For business elites, in-depth interviews were conducted with CEOs, CFOs, chairpersons of boards and leaders in organizations of business representation (e.g. a commerce chamber). For political elites, in-depth interviews were conducted with members of the lower chamber from the four largest parties in the country, as well as with local legislators and a few members of the executive branch in the states of Rio de Janeiro, São Paulo, Ceará and Bahia. Civil servants were selected from among those occupying top-tier positions in the federal government or occupying state and municipal secretaries in Rio de Janeiro, São Paulo, Ceará/Fortaleza and Bahia/Salvador. Interviews were semi-structured around issues of social and economic policy, lasted 50 minutes on average, and were recorded, transcribed, coded inductively and then recoded according to our main categories as described in Table 1. Because we combine data from 1998/1999 and 2012/13, a main concern centers around potential changes over time. While much did indeed change in Brazil between these two time periods, inequality and its externalities remain high. Aside from objective measures of inequality, data from the two time periods indicate that elites were consistently concerned about its negative effects on their own lives across time. In neither time period were interviewees primed to evaluate the poor. The finding regarding their perceptions of the poor as ignorant and uninformed emerged inductively when we analyzed the two sets of interviews together. Table 1 presents the frequencies of codes in the two time periods. Although we can see differences in the prevalence of these codes across time, the overall trend remains constant. In both time periods, elites viewed inequality as the source of externalities, redistribution as a desirable outcome, and the poor as deserving but ignorant in contrast to maximizing elites. Indeed, these perceptions seem to have solidified and become even more salient in the 2013 sample..

(10) Table 1: Rate of presence of codes in percentages over total amount of interviews per period. 1999. 2012-2013. (N=42). (N=14). Externality: violence. 62. 93. Externality: other. 62. 85.7. 5. 7. Poor as good/honest/want to make it. 21.4. 28.6. Poor as victims/passive/manipulated. 66.7. 57.1. Poor as irrational/ignorant. 64.3. 85.7. Elites are rational/more rational than the poor. 54.8. 57.1. 81. 100. Poor as capable. Redistribution is necessary. To test the main implication of the argument emerging inductively from our analysis of in-depth interviews, we ran OLS regressions on a dataset collected from surveys fielded in a random sample of Brazilian elites. The main implication is of course that perception of the poor as irrational will diminish elites’ support for redistributive policies. We then replicated this analysis in random samples of elites in South Africa and Uruguay.. Surveys Between 2013 and early 2015, we fielded 544 questionnaires in randomized samples of elites in the same sectors in Brazil, South Africa and Uruguay, with 60 respondents per sector in each.

(11) 4. country. Following a strategy that relates to Hoffman-Lange’s (2007) position method, elites were sampled based on their institutional positions. We focused on business leaders, elected officials at the national level and top-tier civil servants. We chose to focus on these elites because of their decision-making power and ability to shape policy. Elected officials have an obvious influence on policy, once it is in their hands to put forward legislation. Externalities have a double meaning for them, as they can also bear electoral costs as they may upset poor voters. Civil servants also hold significant power in terms of policy design and even in agenda setting. They are particularly powerful in the Latin American context (Dargent, 2014). The influence of economic elites in policymaking and agenda-setting processes is well documented (Gilens and Page, 2014). For the Brazilian survey we stratified our sample into three groups: (i) elected officials at the 5 federal level from each of Brazil’s four main parties (PMDB, PT, PSDB and PSD), (ii) top-tier civil servants within the federal government, and (iii) business persons (CEOs, CFOs or chairpersons of boards of the country’s top 300 companies). Civil servants were sampled from among those holding DAS 5 and 6 (which is the top status for civil servants in the country) positions in ministries related to economics, development, and social policy in the federal government. The list of the top 300 companies comes from a business publication in Brazil, Revista Exame,​ which is a ​Forbes​ type of publication widely used in the corporate world. In South Africa, we sampled members of the National Assembly from the ruling African National Congress (ANC) and the official opposition party, the Democratic Alliance (DA). We sampled in proportion to each party’s representation in the National Assembly, such that two-thirds of our respondents in the sector were from the ANC and one-third from the DA. In constructing our sample of civil servants, we strategically chose 20 government departments and sampled randomly from top-tier civil servants (Director Generals, Deputy Director Generals, Chief Operating Officers and Chief Directors). To construct our sample of business elites, we triangulated a list of the top 300 companies (by market capitalisation) listed on the Johannesburg Stock Exchange with the Africa Report Top 500 Companies in Africa (which was identified as a widely used list by a senior contact at the Competitions Commission). We randomly sampled companies from this list, allowing for respondents holding the following positions: CEO, CFO or Chairperson of the Board. In Uruguay, congresspersons were sampled from the country’s three main parties, the ruling Frente Amplio, and Partido Colorado and Partido Nacional in opposition. Civil servants were 4. In South Africa, the number of politicians was 63 and of business persons was 61 due to an error made by the surveying company, which continued to conduct interviews after the quota was filled. Since all interviewees were selected randomly, we decided to maintain the extra two cases in the dataset. 5 By main parties we mean those with more seats in parliament..

(12) sampled from among those holding the positions of directors in any of the ​Ministerios of the national government. Business persons were sampled from a list of the 250 biggest private companies, based on tax revenues and number of employees according to the national business census. Respondents were also CEOs, CFOs or chairpersons of boards. The surveys were administered through face-to-face interviews, except for a handful conducted over the phone. We did not a​ llow potential participants to nominate others to participate in their place. Generating probabilistic samples of elites is challenging due to a high expectation of non-response in an already small universe. Our strategy was to randomize substitutions for cases of non-response and continue to re-sample within the same sampling frame until we reached a quota of 60 respondents per sector. This means that all elites in the sample frame had a known and different from zero probability of being selected, but that probability was not necessarily the same given that new rounds of randomization increase the odds of selection. Still, those greater odds are distributed randomly and should not correlate with any meaningful attribute. The overall response rate was 32.3 percent for Brazil, 41 percent for South Africa, and 54.4 percent for Uruguay. Response rates per sector can be found in Table 2.. Table 2: Response rate per sector Elected officials. Civil servants. Businesspersons. Brazil. 60 (30%). 60 (51%). 60 (25%). South Africa. 63 (65%). 60 (35%). 61 (32%). Uruguay. 60 (86%). 60 (67%). 60 (30%). Our samples are particularly high-quality due to our probabilistic sampling, face-to-face instrument, and relatively high response rates. To the best of our knowledge, ours is the first elite survey following such standards. Estimating the effect of perceptions We run OLS regressions to estimate the effects of elites’ perceptions of the poor on their support for redistributive social policies and performed robustness checks using matching and permutation tests. Below, we provide further information on our variables and model specification..

(13) Outcome variable For the outcome variable we built an index of support for redistributive policies that is an interaction between respondents’ assessments of policies as desirable and viable. We asked respondents to evaluate a variety of social policies by telling us whether they thought each was a) desirable and b) viable. The policies mentioned were: universal healthcare, prioritizing the poor in higher education, provision of free higher education, quotas in higher education, free primary schooling, unemployment insurance, food programs, housing programs, decreasing income disparities between occupations, and a universal basic income program. These policies are specific enough to give a sense of what kind of redistributive effects they might advance, but general enough so that they could not be directly related to policy positions advanced by recognizable political groups, thus avoiding response bias caused by partisan preferences. Our index ranges from 0 (no policy is viable regardless of how many are desirable) to 100 (all policies are desirable and viable). In our measure, high scores imply that respondents not only view a high number of policies as desirable, but also that they believe that there are the means to implement such policies. By the same token, the index punishes “dreamers”, respondents who exhibit high support for policies, presumably because they consider them to be fair, but actually do not believe that they are feasible and thus would not engage time or resources to implement them. The distribution of the index can be seen in Figure 1.. Figure 1: Distribution of policy support.

(14) The division inside the box indicates the median. The lower and upper hinges correspond to the first and third quartiles. We believe that this measure better captures elites’ commitment to redistribution than would any single measure of support because of social desirability bias. As shown in Figure 1 and Table 3, on average, elites in the three countries indicated that almost eight out of ten policies are desirable. When we include the feasibility measure, however, we see mid-range scores in Brazil and Uruguay and even lower scores in South Africa. Explanatory variable The main independent variable, aimed at capturing elites’ views of poor, is a binary measure representing the respondent’s description of the average citizen as either wise (0) or uninformed and irrational (1) in response to the following question, which we posed:: “In your opinion, the majority of the Brazilian/South African/Uruguayan electorate: vote wisely OR are uninformed and irrational?” We are aware that this question does not mention the poor directly but rather asks about the average voter. In these middle-income and unequal countries, however, the median voter is understood to be poor. This was confirmed in the in-depth interviews, where participants used.

(15) these terms interchangeably. The measure we use focuses specifically on political behavior. This matches references to political manipulation, found in the in-depth interviews. The focus on average voters hones in on politics, rather than capturing perceptions of ignorance or irrational behavior in other dimensions of everyday life, such cultural taste or religious beliefs. In addition to this variable, we also used an ordinal measure for the poor’s perceived ignorance. The ordinal measure indicates participants’ agreement with the statement “​People do not know how to vote properly.”​ This measure allows for more variation and we use it to test the robustness of our findings. Covariates We included three types of control covariates in our models: 1) perceptions of externalities 2) biographical, ideological, and socioeconomic characteristics of respondents that might affect respondents’ predisposition to endorse social policies, 3) variables aimed at tapping alternative explanatory frameworks, as discussed above. We outline each of these measures below. Perceptions of externalities: Here we measure whether elites view poverty as a source of negative externalities for themselves. Respondents were asked to name the main consequence of poverty first in an open question, and then to choose two other consequences from a list. We coded perceptions of externalities to be present if elites named a consequence that could affect their own safety (e.g. violence, social conflict, criminality) in either the open- or close-ended question. The result is a binary measure where 1 represents the acknowledgement of externalities. Ideology: Previous studies have pointed to the effect of ideologies about equality on elites’ attitudes (Verba et al. 1987). We thus included a measure for ideology where we asked respondents what they thought was more important: liberty (0) or equality (1). Perception of responsibility: According to Reis and Moore (2005), perceptions of personal responsibility are a key in explaining elites’ attitudes toward redistribution. We asked respondents who they think is responsible for solving inequality and offered the option “people like yourself”. We added this covariate in the model as a binary measure where 1 means that respondents agree that people like them are responsible and 0 means that they point to others as responsible. Willingness to pay more taxes: studies often assume that elites oppose redistribution because of the tax implications for themselves (e.g. Acemoglu and Robinson 2005, Boix 2003). We asked elites whether they were willing to pay more taxes in order to implement the policies that they.

(16) identified as desirable and viable. The result is a binary variable where 1 represents respondents’ willingness to pay more taxes. Elite sectors: The source of elite power is itself an important factor determining the incentives to redistribute (López 2013a). We expect political and bureaucratic elites to be more favorable to redistributive policies if compared to business elites. We account for this by controlling for elite sector (political, economic, and civil service), with bureaucratic elites as the base line. Gender: coded as 0=male and 1=female. While women are a minority among elites and within our samples. We control for gender in order to discount a potential confounding effect. Age: Age could have an effect on political preferences as it implies differences in political socialization. We coded age based on interviewees’ year of birth. Race: Given the salience of racial inequality in Brazil and South Africa, it is possible that race could have a confounding effect on perceptions of the poor. We used respondents’ self-attributed racial identity. Although racial identification and categorization is a complex process, that varies over time and place, we tap into the historic and contemporary privileges of whiteness by constructing a binary measure where respondents who identified as “white” were assigned 0 and respondents identifying through any other racial category were assigned 1. Socioeconomic background: Respondents’ own socioeconomic background may affect their level of engagement with, and perceptions of, the poor. We use two variables to capture socioeconomic background: mother’s education and father’s education. Here, we rescalled responses into an ordinal measure ranging from 1 to 6, where 6 means college degree or higher, 5 means some college education, 4 means high school degree, 3 means some high school, 2 means primary, and 1 means less than primary. Table 3 presents the distribution of all referred measures in each country..

(17) Table 3: Distribution of covariates Covariate N policies desirable N policies viable Index of policy support Voters are uninformed and irrational People don’t know how to vote properly Age Brazil Race (non-white) Christian Freedom over equality Gender (female) Father’s education (on a 6-point scale) Mother’s education (on a 6-point scale) N policies desirable N policies viable Index of policy support Voters are uninformed and irrational People don’t know how to vote properly Age South Africa Race (non-white) Christian Freedom over equality Gender (female) Father’s education (on a 6-point scale) Mother’s education (on a 6-point scale) N policies desirable N policies viable Index of policy support Voters are uninformed and irrational People don’t know how to vote properly Uruguay Age Race (non-white) Christian Freedom over equality Gender (female) Father’s education (on a 6-point scale) Mother’s education (on a 6-point scale). Model specification The model specification goes as follows:. Mean 7.90 7.68 61.9 0.77. SD 1.73 1.79 22.10 0.47. Min 3 1 3 0. Max 10 10 100 1. 2.98. 1.52. 1. 5. 52.14 0.2 0.69 0.65 0.10 3.78 3.43 7.45 4.92 37.2 0.59. 10.29 0.4 0.46 0.47 0.31 2.17 1.94 1.70 2.17 20.57 1.93. 26 0 0 0 0 1 1 2 0 0 0. 82 1 1 1 1 6 6 10 10 100 1. 3.11. 1.52. 1. 5. 51.60 0.63 0.65 0.46 0.11 3.57 3.19 7.70 7.59 58.7 0.33. 9.67 0.48 0.48 0.5 0.31 1.69 1.53 1.39 1.16 17.86 0.47. 29 0 0 0 0 1 1 4 3 15 0. 80 1 1 1 1 6 6 10 10 100 1. 1.99. 1.37. 1. 5. 52.67 0.03 0.43 0.64 0.14 3.86 3.74. 10.45 0.18 0.49 0.48 0.35 1.61 1.63. 30 0 0 0 0 1 1. 79 1 1 1 1 6 6.

(18) Y=α+β1PERCEPTION+ β 2SEC+ β3EXT+ β 4Xi+ε. Where Y = ∑ (desired policies) * ∑ (f easible policies), α represents the intercept, PERCEPTION stands for respondents’ agreement with the idea that average voters are uninformed and irrational, SEC represents membership of one of the three elite sectors, EXT represents respondents’ perceptions of the negative externalities of inequality, and ​Xi represents the control covariates. We also conducted propensity score matching as well as a permutation test with the full data set (which brings together data from all three countries) to provide a general and more robust estimation of the effect size of the perception of the poor. Results can be found in the Appendix. For effects on each particular policy we ran logistic regressions. Results can be found in the appendix. Before presenting findings from the survey data, we first outline the findings emerging from the in-depth interviews with Brazilian elites and show how they informed our theorization about the role of perceptions of the poor.. Results from in-depth interviews. As shown in the introduction, there are competing explanations in the literature regarding elites’ attitudes toward inequality and redistributive policies. Some studies assume elites’ willingness to redistribute as a function of the costs imposed by the poor (e.g. Acemoglu and Robinson, 2005; Boix, 2003; de Swaan, 1988), while others assume it as a function of normative evaluations and overarching values (e.g. Hochschild, 1981; Katz, 2013; Verba et al., 1987). One way to unpack the incentives and motivations guiding elites’ behavior is to focus on what they have to say about inequality, and how their response relates to those previous explanations. According to the first proposition, elites should be compelled to solve inequality if income skewness generates costs and relevant threats to them. In economic terms, if inequality affects elites’ own utility function, they should perceive an incentive to improve the utility function of others as a means to maximize their own.. Elites’ framing of inequality in Brazil supports that idea. Almost all of the elites interviewed perceived inequality as being a problem affecting not only the poor but also themselves. When discussing the consequences of inequality, they most frequently mentioned violence, crime, and.

(19) safety—although they also mentioned other issues, such as the lack of qualified workforce. As documented , in Table 1 (in the Methods section), 49 out of 56 interviewees described some kind of externality of poverty, with 39 of them focusing on violence and criminality as a major consequence of inequality. These trends were clear in data from both time periods. Inequality, and not simply poverty, is the main source of such externalities because elites’ realize that it is the contrast between their affluence and the extreme life conditions of the poor that generates conflict. In 2013, for instance, a congressman lamented how “there is no fun in having money if you are in a restaurant and there are people hitting the glass window asking for food outside. You don’t enjoy your money as much as you could.” He then quickly made the link to safety: Not to mention public safety! What is the fun in living inside a bulletproof car, in a gated community? All those lights, security staff… I mean, this is no quality of life ... What sense does it make to be someone who can only feel safe when vacationing abroad? Others spoke explicitly about the growth of elites’ concern over inequality being conditional on inequality’s effect on their personal well-being. In 2013, a civil servant exemplified this idea quite succinctly by stating that “it is when the rich feel their safety threatened that they say: ‘wait a minute, this is affecting my personal safety’.” Others contrasted their lives in Brazil to their experiences overseas, again highlighting their concerns over safety. For instance, businessman interviewed in 2012 stated: Look, I have an apartment in New York. Over there, I walk on the street, I feel great, living without worries in New York, but not in São Paulo. I do not walk on the street in São Paulo. Literally! I walk on a treadmill. I don’t walk on a treadmill in New York, I walk on the street. What does that imply in terms of wellbeing? I mean, the saddest thing is that there is no need for that, [this is] a country with an extraordinary economic potential. Given the Brazilian elites’ acknowledgement of inequality as being a problem not only for the poor but also for themselves, it should come as no surprise that they have, overall, a positive attitude towards redistribution. Nearly all interviewees argued that redistribution policies are necessary. More importantly, they posited that less inequality is in their own interest, as they would be able to enjoy their wealth, as well as enjoy economic opportunities with a better trained and more satisfied workforce. By improving the poor’s well-being, they argued, they could improve their own. The rationale assumed by cited economic models thus holds. Nonetheless, based on those same elements we should observe a more consistent support of particular redistributive policies, which should then significantly ameliorate inequality, mitigate conflict,.

(20) and bring elites and the poor to a high level equilibrium. This is not what we observe in Brazil though. In other words, the mechanism is present but the outcome is not. Therefore, the story of distributive conflict may not be entirely about economics. Another venue for understanding elites’ reaction to inequality has to do with their cultural interpretation of poverty and moral assessment of the poor. While previous work has documented how moral distinctions made between the “deserving” and “undeserving” poor help explain support for redistribution (Katz, 2013), the Brazilian elites whom we interviewed did not make such distinctions, nor did they blame the poor for their own poverty. Instead, they largely perceived poverty as structural and viewed the poor as deserving of relief, expressing admiration for their hard work and pity for their hardships. For example, in 1999 a politician argued that “most of them [poor people] are honest, dignified people,” while in 2013 a businessman argued that the poor are “absolutely normal people” who just “want to get ahead.” They also acknowledge that the difference between outcomes has to do with differences in opportunities, suggesting that policies that aim at leveling the field are fair. In the words of a politician in 1999: There has never been a commitment to give the working class its fair share. The reason that these social classes are marginal is not that they don’t want to work. They are marginalized because they are not given opportunities. Studies have argued that cultural understandings about the nature of inequality (e.g. of opportunity) and about fairness are key in shaping elites’ attitudes toward inequality (Hoschild, 1981; Verbat et al., 1987). In Verba et al.’s (1987) work, for instance, similar ideas about inequality are portrayed as the cultural pillars of advanced social democracies, such as Sweden, not of highly conflictive unequal democracies such as Brazil. In other words, cultural traits linking elites to more redistribution are also present, but the predicted output of more robust social policies is not. A closer look at the descriptions of the poor and poverty in the in-depth interviews reveals that Brazilian elites also draw strong distinctions between themselves and the poor; distinctions that are not based on merit, deservingness, or ideals about equality. When discussing the poor’s capacity to lift themselves out of poverty, interviewees recurrently described them as uninformed, ignorant, uneducated, and even irrational. Such perceptions prompted the elites to question the poor’s ability to act in their own self-interest. For example, in 1999, a politician explained that “improvements will never come from the excluded masses.” Conceptualizing the poor as incapable of reason, he argued that little can expected of them:.

(21) As I said, we cannot expect anything from them. We cannot expect anything from a person who is not capable of reason, at the most basic level, of [making sense of] what is going on in the world.. Similarly, in 2013 a businessman suggested an inextricable link between poverty and “lack of knowledge”: The main consequence of poverty is the lack of knowledge. The lack of knowledge. And then health problems, malnutrition, all the educational problem. We don’t have the basic conditions for people to have a minimum amount of knowledge, basic notions so that they can have a better life. Across sectors and across time, interviewees described poor people as limited in their capacity for rational action due to ignorance and lack of education, caused by poverty itself. These types of descriptions appeared spontaneously in 39 of the 56 interviews and were highly salient across both time periods. In short, although interviewees did not blame the poor for their poverty, they described the poor almost as victims who are —most of the time irreversibly—unable to act in their own material interest because of the structural constraints of poverty, which are seen as imposing strong limitations on rationality. Poverty itself was seen as causing irrational behavior. A businessman in 1999 articulated this view succinctly by stating: “Extreme poverty makes one not see anything. It is a black veil over people’s heads. It blinds them.” Because of the perceived effect of poverty on people’s discernment, our interviewees were less convinced of the effectiveness of redistributive social policies and often expressed a fatalistic view regarding the challenges of poverty and inequality. In 1999, one businessman explained these challenges to be insurmountable by using the metaphor of poverty as a virus: The problem is much more serious than it seems. Because it is kind of like an infestation. You need to fight the virus first… If you’ve got a multiplying factor, and limit resources, then you can fight it a bit. But, with the multiplication factor, the problem gets bigger.. Although the strong language of poverty as a “virus” was not particularly recurrent in our interviews, it illustrated two points we found repeatedly. First, poverty was often presented as a threat that is difficult to deal with—in this case, a virus that can cause problems not only for the infected. Second, interviewees did not blame the poor for their poverty—in this case the interviewee does not perceive the poor as having chosen to be infected by a “virus.”.

(22) In the narratives of our interviewees, the perception of the poor as being uneducated, unwise, and incapable of rational decision-making seemed to be particularly consequential in the case of political behavior. ​Ignorant, poor voters were viewed as posing a danger due to their susceptibility to manipulation by other elites. Political elites, in particular, were accused of subverting desirable redistributive policies to satisfy their immediate clientelistic needs, while voters watched passively. Contrary to their views of the poor, elites understood politicians as being rational and selfish maximizers. The link between ignorance and political manipulation was made by interviewees in the 1999 and 2013 datasets. One businessman interviewed in 1999 distinguished between the poor and the “thinking and property-holding classes,” who—in his view—are “not ashamed to keep in place this unjust system.” Another businessman claimed, in 1999, that “the government wastes money in all it does” because policies end up serving the “individual gains” of other elites, while a businessman interviewed in 2013 explained how there is “obviously a political agenda” behind cash transfer policies—a project to “garner votes to remain in power.” This understanding seems to lead some of the elite interviewees to blame the poor for their poverty—not due to their lack of deservingness or laziness, but because of their political behavior. The quote below, from a businessman interviewed in 2013, illustrates this reasoning: The unpreparedness of the population during the electoral process is so serious… it is a process of vote buying, a process of using policies for marketing, and this way the prospect of actual change becomes very limited. Politicians and civil servants, who were accused by business elites of biasing policies to benefit themselves, expressed surprisingly similar sentiments. When asked about the unintended consequences of social policies, a civil servant, in 2013, stated that he saw “it all with great suspicion”, implying that political elites could fraud policies in order to channel benefits to loyalists. When discussing his own policy agenda, a congressman similarly argued, in 1999, that politicians “are used to making politics out of the basic needs of the population.” In their descriptions, the elites we interviewed constructed a bind: poverty was the cause of the poor’s ignorance, but it could not be tackled effectively precisely because of poor people’s lack of political sophistication and strategic reasoning. Although our interviewees generally viewed redistribution as fair and desirable, they remained skeptical regarding the odds of success of social policies, claiming that such policies were often used to garner political support and had no real effect on social mobility. T​he resilience of these perceptions over time is striking. In the period between the two sets of interviews, Brazil went from a center-right to a center-left government coalition, and from a period of economic crisis to a one of resource abundance. Yet, the way in which elites spoke about poor voters showed little change..

(23) Our findings suggest that there is a strong link between perceptions of the poor and distrust, both toward the poor and toward other elites. Distrust is a strong feature of the civic culture in Brazil and Latin America in general, and an important cultural variable in explaining attitudes toward redistribution (Alesina and Giuliano, 2009; Blofield 2011a; Reis 1998). In the case of elites, results suggest that elites often feel like they cannot trust poor because of their irrationality, but they cannot fully trust other elites either as the latter might explore the poor instead of lifting them out of poverty. The narratives of distrust are thus closely connected with the distinction that interviewees make between irrational poor and selfish maximizer elites. ​First, elites express that the “ignorant” poor will be incapable of acting in their own interests, and thus will miss the opportunities granted to them by social policies. Second, elites express fear that other utility-maximizing elites will use redistributive policies to strengthen clientelist ties, putting immediate private gains over future collective goods. A venture that is, again, made possible by the ignorance of the poor. It follows that the perception of the poor as irrational should counterweight the effect of negative externalities, and thus explain elites’ inconsistent support for redistributive policies. In other words, economic and cultural determinants go in opposite directions. Figure 2 summarizes this process. Figure 2: Causal argument.

(24) The implication of findings emerging from the interview data is that perceptions of the poor as ignorant and irrational should reduce elites’ support for redistributive measures, all things being equal. We test this hypothesis in the next section. Results from surveys. We test the implications of our theory regarding the relationship between elites’ perceptions of the poor and their willingness to redistribute in survey data collected from random samples of elites in Brazil, South Africa, and Uruguay. Brazil Survey data reveal similar pattern to those found in the interview data. Figure 3 shows that findings in the two samples converge. When asked about the main consequences of poverty and inequality, over 60 percent of elites surveyed and an equal proportion of interview participants pointed to violence or conflict. When asked whether they believed that redistribution was necessary, 95 percent of survey respondents answered in the affirmative. This too was similar to findings from the interviews, where 86 percent of participants spoke globally about redistribution as being necessary. Finally, like the Brazilian elites who participated in the in-depth interviews in both time periods, those who took part in the closed-ended survey also characterized the poor as being ignorant and irrational. When asked about “average voters” (our proxy for elites’ perceptions of the poor as political actors), 77 percent of respondents in the survey chose to describe them as ignorant and irrational. This was similar to the rate of spontaneous mentions of the poor’s ignorance in the interviews. Findings from the larger and probabilistic sample of Brazilian elites thus converges with findings from the in-depth interview data: elites see inequality as the sources of threats and view redistribution as desirable, but many also perceive the poor as ignorant and irrational. If our hypothesis holds, then we should observe a reduction in elites’ willingness to support redistributive in the presence that perception, regardless of their acknowledgement of externalities.. Figure 3: Convergence between interview findings and survey results.

(25) As elaborated in the methods section, we constructed an index to measure elites’ support for redistributive policies. This allowed us to capture not just an abstract sentiment of global support for redistribution, but a more specific assessment of elites’ views on both the desirability and feasibility of a range of redistributive policies. Table 4 shows that, all else being constant, elites’ assessment of the poor as being uninformed and irrational reduces their support for redistributive policies by between 11 and 27 points, within a 0 to 100 range of our index of policy support. The effect is robust to the inclusion of several covariates, such as race, age, gender, and political ideology among others. The impact of perceptions of externalities on elites’ endorsement of social policies is positive, as we predict, yet rather weak and thus disappointing from the perspective of the larger story told here, which follows upon the political economy literature. It is possible that our binary measure of perceived externalities is generating a ceiling effect because it does not allow respondents to grade their concern. For our purposes, what is more important is that perceptions of the poor and perceptions of externalities go in opposite directions and are not themselves correlated (see correlation matrix in the appendix), which supports the idea that the effect of perceptions of the poor is independent of perceptions of externalities..

(26) Table 4 —​Effect of perceptions of the poor on elites’ support for redistribution in Brazil (5). (6). -8.310 (5.209). -12.844* (5.913). -5.834 (16.369). -7.724 (23.160). Poor are ignorant Agrees partially vs disagrees. -8.473 (4.946). -11.690* (5.592). Poor are ignorant Agrees vs disagrees. -18.252*** (5.390). -19.153** (6.015). Poor are uninformed and irrational. (1). (2). (3). (4). -19.376*** (4.120). -15.831*** (4.041). -11.926** (4.135). -13.144** (4.492). Poor are ignorant Disagrees partially vs disagrees. Poor are ignorant Neither agrees or disagrees. disagrees. vs. -15.426*** (3.955). -7.765 (4.454). -3.754 (5.149). -9.245* (4.384). -6.451 (5.105). Political elite. 15.381** (4.54). 14.398* (5.595). 15.818*** (4.505). 14.221** (5.281). Externalities. 1.755 (4.015). 3.776 (4.270). 0.287 (3.850). 2.192 (4.133). Business elite. yes. Controls. yes. Intercept. 75.980*** (3.358). 79.307*** (3.321). 67.740*** (5.117). 79.848*** (14.211). 69.792*** (5.437). 79.092*** (14.272). R​2. 0.129. 0.209. 0.359. 0.302. 0.283. 0.304. n. 152. 152. 151. 137. 161. 144. *p<.010 ** p<.05 ***p<.001 Results are presented in this fashion: coefficient, (standard error). Results also show that political elites tend to be more supportive of redistributive measures when compared to the other elites in the sample. This is expected given politicians’ ties with poor voters. Still, the effect of perceptions of the poor remains strong even when controlling for elite sectors, suggesting that average effects are not induced by partisan dynamics or by other.

(27) electoral incentives. Overall, the regression models support the hypothesis emerging from the in-depth interviews. South Africa In South Africa, almost 60% of respondents chose to describe the average voter as irrational and uninformed. Table 5 presents results for the South African data. Table 5—​Effect of perceptions of the poor on elites’ support for redistribution in South Africa. Poor are uninformed and irrational. (1). (2). (3). (4). -15.207*** (3.397). -15.124*** (3.633). -13.055*** (3.566). -9.910** (3.587). (5). (6). Poor are ignorant Disagrees partially vs disagrees. -12.112* (5.183). -17.166** (5.668). Poor are ignorant Neither agrees or disagrees vs disagrees. -15.906 (14.153). -15.620 (14.192). Poor are ignorant Agrees partially vs disagrees. -19.665*** (5.049). -19.444*** (5.477). Poor are ignorant Agrees vs disagrees. -16.053** (5.190). -14.505** (5.342). -0.239 (3.633). 5.656 (3.903). 8.382* (4.118). 2.727 (3.898). 7.266 (4.076). Political elite. 15.381** (4.54). 14.398* (5.595). 12.359** (4.087). 15.062*** (4.439). Externalities. 1.755 (4.015). 3.776 (4.270). -0.395 (4.457). 1.235 (4.369). Business elite. yes. Controls Intercept. 46.922*** (2.719). 46.954*** (2.774). 39.970*** (5.134). 32.028* (13.596). R​2. 0.125. 0.125. 0.196. 0.357. N. 142. 142. 142. 132. yes 46.574*** (6.267). 46.022** (14.452). 0.212. 0.380. 147. 137. *p<.010 ** p<.05 ***p<.001 Results are presented in this fashion: coefficient, (standard error).

(28) As predicted, results are very similar to those found in Brazil, with a robust negative effect of elites’ perceptions of the poor as uninformed and irrational on their support for redistributive policies. As in Brazil, results hold even when we include other covariates. Importantly, the effect is robust when we include our variable measuring perceptions of externalities. The South African data adds support to the hypothesis that perceptions of the poor as ignorant, uninformed, and irrational results in reduced support for redistributive policies in a case other than Brazil but with similar distributive dilemmas. Again, differences between elite sectors are reflected in politicians’ higher levels of support for redistributive policies, but the effect of perceptions of the poor is again robust to sectoral differences and to the inclusion of other covariates. The results from South Africa show impressive similarity with those from Brazil.. Uruguay Uruguay is a negative case in our study, from the viewpoint that the country has shown a better and more consistent performance in reducing poverty and inequality in recent times. Only 33% of interviewees chose to describe the average voter as irrational. Among politicians and civil servants, agreement was less than 15%. Table 6 presents results for the Uruguayan sample. Consistent with our prediction, we observe that the average impact of perceptions of the poor tends to be much smaller and potentially null in Uruguay when accounting for other covariates. Uruguayan elites’ commitment to redistributive policies seems more evenly distributed. This result suits our theory: in a context where elites, for the most part, do not see the poor as ignorant and irrational, support to redistribution tends to be more evenly distributed. Yet it is also true that what is absorbing much of the effect of perceptions is the effect of elite sectors, because business elites tend to agree with the idea of average voters being irrational much more than their peers in other sectors. This creates a confounding problem in our data for Uruguay. We thus confirm Uruguay as a negative case, where the general absence of the causal variable allows elites to better coordinate around redistribution..

(29) Table 6—​Effect of perceptions of the poor on elites’ support for redistribution in Uruguay (1). (5). (6). Poor are ignorant Disagrees partially vs disagrees. -4.053 (4.485). -2.046 (5.042). Poor are ignorant Neither agrees or disagrees vs disagrees. -3.465 (8.465). -3.898 (8.436). Poor are ignorant Agrees partially vs disagrees. -8.846 (5.025). -3.040 (6.139). Poor are ignorant Agrees vs disagrees. -11.136 (6.827). -4.534 (8.060). Poor are uninformed and irrational. -11.682*** (1.909). (2). (3). (4). -6.585 (3.859). -6590 (3.877). -4.138 (4.176). -8.902* (3.827). -8.500* (4.098). -7.527 (4.698). -7.284 (4.454). -7.640 (5.249). Political elite. 1.766 (3.517). -0.008 (3.953). 2.968 (3.569). 0.216 (4.053). Externalities. 2.290 (3.251). 0.667 (3.738). 2.324 (3.299). 0.995 (3.848). no. yes. no. yes. Business elite. no. Controls. no. Intercept. 60.663*** (1.909). 62.117*** (1.983). 59.850*** (3.396). 69.165*** (12.438). 59.185*** (3.468). 69.646*** (13.105). R​2. 0.081. 0.113. 0.118. 0.190. 0.130. 0.189. N. 151. 151. 151. 126. 153. 124. *p<.010 ** p<.05 ***p<.001 Results are presented in this fashion: coefficient, (standard error). All in all, results show that elites who portray the poor as irrational tend to be less supportive of redistributive policies in Brazil, which is the case that inspired our argument, and in the two other cases. Alternative non-parametric estimations using merged data from the three countries also portray a negative effect of over 10 points in our scale of support of social policies (see.

(30) appendix). Such an effect is much more substantial than the positive effect of externalities, which fits the story that we developed in the former section: elites observe strong economic incentives to support redistributive measures, but ultimately hold their endorsement because of processes triggered by their perception of the poor as irrational. This means that economic and cultural determinants work in opposite directions, which explains why coordination around social policies is so difficult. Discussion and Conclusions Our results go against the common assumption in political economy that elites will always view rampant inequality as being in their best interest, and will thus look for ways to block redistribution (Acemoglu et al. 2011; Albertus and Menaldo 2018; Boix 2003; Soifer 2013). Focusing on the cases of Brazil, South Africa, and Uruguay, our study demonstrates that externalities of inequality prompt positive attitudes among elites regarding the need of redistribution, but that perceptions of the poor are key intervening variables in shaping their support for redistribution policies. We find that elites systematically identify poverty and inequality as being the main causes of crime and violence, which negatively affect their own lives. All things considered, elites envision a scenario in which some redistribution would solve many of their own problems at a reasonable cost—a cost that they claim to be willing to pay if they have some assurance that such policies will work. Our findings point to the role of elites’ perceptions of the poor in hindering their willingness to pursue such goals. Their views of the poor as ignorant, irrational, and susceptible to manipulation by other elites, ultimately lead them to reject social policies which they would otherwise find in their own best interest. These findings point to the importance not only of economic, but also of cultural variables in explaining the reproduction of inequality. Would it be possible, however, that our data reflect nothing but a discursive cover-up of a crude rational calculus? In terms of the interview data, while we do not necessarily take everything we were told in the interviews at face value, we do not believe that the interviewees were deliberately deceiving themselves and us. Indeed, why would elites not prefer a scenario in which they still hold power and wealth, but without suffering the negative externalities of poverty? After all, political and market elites have the tools for allowing some redistribution while allocating costs elsewhere (e.g. the middle class). In such a scenario, elites benefit from policy results but also avoid bearing their cost. Our findings suggest that perceptions of the poor play a key role in explaining why elites did not propose these types of solutions. In terms of the effect of perceptions on policy support, we also do not think that it reflects respondents’ justifications to us of their pre-existing policy preferences. Our output variable is composed of the interaction between two series of questions. It would be virtually impossible for.

(31) respondents to anticipate their score. Still, on the subject of reverse causation, one might argue that elites’ perceptions of the poor are caused by their observation that redistribution policies tend to fail. This is indeed possible but not incompatible with our argument. It could suggest that past experiences of failed redistribution may lead elites to construct explanations that blame the poor, and thus hinder future support for redistributive policies. Our findings also suggest that elites’ limited support to redistribution is not the result of an assessment that poverty and inequality pose minimal threats (Ansell and Samuels, 2014; Haggard and Kaufman, 2012, 2016) or that existing institutions shield them from such threats (Albertus and Menaldo, 2018). Even if elites rely on predatory state practices, low-paid labor, and other institutions that generated the vicious cycles of impoverishment and inequality, our results show that elites worry about the consequences of such inequality for their own safety and well-being, as indeed predicted by existing literature (see de Swaan 1988, Reis and Moore 2005, Rueda and Stegmueller, 2016). Yet, this assessment does not automatically translate into more support for redistribution. Instead elites’ views of the poor curtail any such commitment. In documenting these trends, we have argued that cultural variables--in the form of perceptions of others--are crucial for our understanding of why redistributive policies have garnered limited elite support. Our interpretation of the data is as follows: because elites view the poor as ignorant and irrational, they do not trust them to make informed decisions and instead view them as susceptible to manipulation by other, utility-maximizing, elites. Perceptions of the poor, in other words, disrupt the effect of elites’ perceptions of negative externalities on their commitment to redistributive social policies. The theoretical contribution of our study then is to highlight how culture mediates the effect of material interests on policy preferences. Our study points to the role of perceptions in the reproduction of inequality, but it does not allow us to unpack how these perceptions develop and take hold in the first place. It is possible that overall levels of inequality may themselves foster a degree of social distance that underpins these perceptions, creating a feedback loop. This may be why we find these perceptions to be more salient in Brazil and South Africa than in Uruguay. Although our case selection is limited, the comparative trends offer important avenues for future research. In addition to the effects of overall levels of inequality, researchers may wish to examine the role of other macro-level variables in explaining micro-level perceptions and preferences. These might include the salience of racial boundaries, the legacies of authoritarian rule, and type of welfare regime to name but a few. Doing so will continue to advance our knowledge of both the cultural and economic determinants of redistribution regimes.. References.

(32) Acemoglu, Daron and James A. Robinson (2005) ​Economic origins of dictatorship and democracy​. Cambridge University Press. Acemoglu, Daron, Davide Ticchi, and Andrea Vindigni (2011) "Emergence and persistence of inefficient states." ​Journal of the European Economic Association​, 9.2, 177-208. Albertus, Michael and Gay, Victor (2017) "Unlikely Democrats: Economic Elite Uncertainty Under Dictatorship and Support for Democratization." ​American Journal of Political Science,​ 61.3, 624-641. Albertus, Michael and Menaldo, Victor (2014) "Gaming democracy: elite dominance during transition and the prospects for redistribution." ​British Journal of Political Science​ 44.3, 575-603. Alesina, Alberto F. and Giuliano, Paola (2009). Preferences for redistribution. No. w14825. National Bureau of Economic Research. Alesina, A., & Angeletos, G. M. (2005). Fairness and redistribution. ​The American Economic Review​, ​95​(4), 960-980. Almond, Gabriel, and Verba, Sidney (1963) ​The Civic Culture: Political Attitudes and Democracy in Five Nations​. Boston. MA: Little, Brown and Company. Ansell, Ben W., and David J. Samuels (2014). ​Inequality and Democratization.​ Cambridge University Press. Axelson, Robert (1984) ​The evolution of cooperation​. Basic Books. Berens, Sara, and von Schiller, Armin (2017) "Taxing Higher Incomes: What Makes the High-Income Earners Consent to More Progressive Taxation in Latin America?" ​Political Behavior​ 39.3, 703-729. Boltanski, Luc and Thévenot, Laurent (1991) ​De la justification : les économies de la grandeur.​ Gallimard. Bourguignon, François, Nuñez, Jairo, and Sanchez, Fabio (2003) "A structural model of crime and inequality in Colombia." ​Journal of the European Economic Association​ 1.2​-​3, 440-449. Blofield, Merike (2011a) “Desigualdad y política en América Latina”. ​Journal of Democracy en Español​, 3, 58-74. Blofield, Merike (2011b) ​The​ ​great gap: Inequality and the politics of redistribution in Latin America​. Penn State Press..

(33) Buquet, Daniel, and Piñeiro, Rafael (2016) "Uruguay's Shift from Clientelism." ​Journal of Democracy,​ 27.1, 139-151. Cavaillé, Charlotte, and Trump, Kris-Stella (2014) "The two facets of social policy preferences." The Journal of Politics,​ 77.1, 146-160. Choe, Jongmook (2008) "Income inequality and crime in the United States." ​Economics Letters 101.1, 31-33. Clarke, Gerard, and Sison, Marites (2003) "Voices from the Top of the Pile: Elite Perceptions of Poverty and the Poor in the Philippines." ​Development and Change​ 34.2, 215-242. Cousin, Bruno, Khan, Shamus and Mears, Ashley (2018) “Theoretical and methodological pathways for research on elites”. ​Socio-Economic Review.​ 16.2, 225–249 de Swaan, Abram (1998) ​In care of the state: Health care, education, and welfare in Europe and the USA in the modern era​. Oxford University Press, USA. de Swaan, Abram, et al (2000) "Elite perceptions of the poor: reflections for a comparative research project." ​Current sociology​ 48.1, 43-54. de Swaan, Abram (2005) "Elite perceptions of the poor: reflections on a comparative research project." In: Reis, Elisa and Moore, Mick (eds.) ​Elite perceptions of poverty and inequality​, University of Chicago Press , pp. 182-194. Fairfield, Tasha (2010) "Business power and tax reform: Taxing income and profits in Chile and Argentina."​ Latin American Politics and Society​ 52.2, 37-71. Fajnzylber, Pablo, Lederman, Daniel and Loayza, Norman (2002) "What causes violent crime?" European Economic Review,​ 46.7, 1323-1357. Feierherd, German, Schiumerini, Luis and Stokes, Susan (2017) "When Do the Wealthy Support Redistribution? Inequality Aversion in Buenos Aires." ​British Journal of Political Science,​ 1-13. Gerring, John and Cojocaru, Lee (2016) "Selecting cases for intensive analysis: a diversity of goals and methods." ​Sociological Methods & Research​ 45.3, 392-423. Gilens, Martin and Page, Benjamin I. (2014) "Testing theories of American politics: Elites, interest groups, and average citizens." ​Perspectives on politics​ 12.3, 564-581. Hacker, Jacob S. and Pierson,Paul (2010) ​Winner-take-all politics: How Washington made the rich richer--and turned its back on the middle class.​ Simon and Schuster, 2010. Hagopian, Frances (2007) ​Traditional politics and regime change in Brazil​. Cambridge University Press..