What s the Matter with Inequality? Evidence from Survey Experiments

Transcription

1 What s the Matter with Inequality? Evidence from Survey Experiments Glory Liu August 9, Introduction What s the matter with inequality? Despite mounting evidence demonstrating how economic inequality is detrimental to economic growth [Piketty and Saez, 2003, Piketty, 2014], or that it has bred an unequal democracy that is more responsive to the wealthy than to the masses [Bartels, 2008, Gilens, 2005, Hacker and Pierson, 2010, Gilens and Page, 2014], or that it undermines a democratic spirit of equal treatment [Reich and Satz, 2013, Anderson, 1999, 2008], there is a relative lack of public consensus about why inequality matters. At best, the public is aware of the severity of economic inequality and sensitive to di erent levels of information [Kuziemko et al., 2015], but the limited set of redistributive policy options fail to accurately capture their evaluations of the problem [McCall, 2013]. Furthermore, disagreement at the national level may merely be are ection of the highly-localized variation in experiences and beliefs about inequality and redistribution, rather than simply a lack of public awareness [Newman et al., 2015]. Seen in a less attering light, the public is deeply uninformed, ambivalent, or simply has erroneous views about inequality and redistribution [Kluegel and Smith, 1986, Bartels, 2008, Hacker and Pierson, 2010, Norton and Ariely, 2011]. Add to that the persistence of deeply-entrenched and often racialized anti-welfare sentiments, low faith in government, and it is perhaps unsurprising to nd that Americans largely oppose greater redistribution, resist changing their policy preferences, or become more conservative even as inequality worsens [Gilens, 2009, Benabou, 2000, Kelly and Enns, 2010, Kuziemko et al., 2015, Katz, 1990]. In short, it has become too convenient to blame the public s tolerance of inequality for the apparent political gridlock on the issue. What many of these studies assume or take for granted, however, is the notion that inequality is something objectionable; moreover, they con ate concerns about inequality 1

2 with other distinct moral values. As Starmans et al. [2017] have suggested, If it s true that inequality in itself isn t really what is bothering people, then we might be better o by more carefully pulling apart these concerns, and shifting the focus to the problems that matter to us more. But if people are not bothered by inequality, what else could they be concerned about? When we speak loosely about inequality, we are often referring to an unequal distribution of wealth: a composite concept with at least three distinct, but interrelated parts. An unequal distribution of income entails the position of those at the bottom of the distribution, the position of those at the top, and the distance between the top and bottom. 1 To each of these constitutive parts, then, the public can hold a multiplicity of evaluations that are not necessarily founded on purely egalitarian views that is, a belief that inequality is intrinsically bad or unjust. What this study contends, then, is that each of these components of an unequal distribution of economic resources invites a corresponding objection: people at the bottom have too little, people at the top have too much, or the gap between the top and bottom is too large. Few empirical studies to date, however, have attempted to capture the richness and complexity of these views in the public imagination. This study aims to ll this empirical lacuna by leveraging the conceptual work done by normative political theorists to inform the design of original survey experiments on public beliefs about inequality. I disentangle three overarching objections to inequality: a su cientarian objection (people at the bottom of a distribution have too little), a limitarian objection (people at the top have too much), and an egalitarian objection (the gap between the top and bottom is bad for other moral reasons). Separating these objections enables us to understand what constitutes American beliefs about fairness, equality, and redistribution. For example, if most Americans believe that what is unfair about inequality is not the gap between the rich and poor, but the fact that the people at the bottom have too little, the nature of that objection is entirely consistent with beliefs that economic inequality has increased from a factual standpoint. Moreover, we would be inclined to look for a di erent set of redistributive policies to match these beliefs about what is fair or unfair about inequality. To evaluate these claims empirically, I employ two di erent experimental designs to study public perceptions of inequality and fairness. Both experiments enable us to 1 Osberg and Smeeding [2006] similarly disaggregate the idea of inequality, which, they write, often melds together perceptions of income di erences between the top and the middle of the income distribution, attitudes twoards the gap between the middle classes and the poor, and preferences for a general leveling of pay. In their sociological study, the authors compare what people think speci c occupations should earn against what they think those occupations do earn, and they nd that in general, people underestimate inequality at the top end of the income distribution. 2

3 evaluate the independent e ects of three di erent aspects of inequality and their relation to fairness. My ndings suggest that the primary dimensions on which Americans form their views about inequality are less egalitarian in nature, and more su cientarian or mobilitarian in nature. In other words, what many Americans believe is wrong with inequality is not that it is unfair for some people to have more than others; rather, what is unfair about inequality is that some people have too little, or they don t have the opportunity to move up in society. Such ndings are consistent with studies that have found little evidence of a concern for economic inequality, challenging beliefs that people have either a strong aversion to inequality, or a strong preference for equality [Starmans et al., 2017]. In addition, the heterogeneity of these views fall along partisan lines. While both Democrats and Republicans weigh su cientarian concerns more than egalitarian ones, Democrats tend to view greater inequality as unfair, whereas Republicans actually have a tendency to view greater inequality as more fair. I trace these di erent partisan evaluations of inequality to di erent beliefs about the origins of individual wealth and poverty. By no means is this approach intended to capture the set of all possible objections to inequality either in theory or in practice. Nor does it make any claims about the normative status of certain theories solely on the basis of their validity in public opinion. Rather, this study strives to make a modest contribution to the set of tools used by social scientists to understand the multiplicity of views that the public may have about inequality and its discontents. By breaking down the very concept of inequality or more speci cally, an unequal distribution of income and wealth into its constitutive parts, I hope to show how American beliefs about inequality and fairness are far more nuanced and, indeed, consistent than previous studies have suggested. 2 Reexamining inequality aversion and the preference for equality By now, Americans are used to being bombarded with facts about inequality. The richest 20% of families own close to 90% of the nation s wealth, while the bottom 40% of people own less than 1% of the nation s wealth [Wol, 2014, Desilver, 2015]. Since 1979, the incomes of the top 1% have grown an enormous 200%, while the incomes of the middle class have grown a mere 40% (Congressional Budget O ce Data). And in 2015, an estimated 45 million Americans were still living below the poverty line, while the average pay ratio of CEOs was nearaly 204 times that of the median worker in America. 2 In light of such information, it is perhaps unsurprising to have witnessed the appeal of 2 This statistic comes from the Glassdoor Economic Research Blog. See Chamberlain [2015] 3

4 rhetoric of then-presidential hopeful, Senator Bernie Sanders, who once stated, A nation will not survive morally or economically when so few have so much, while so many have so little. Rhetorical power aside, Sanders claim should warrant further interrogation: what constitutes an objection to inequality? Is morally wrong, for example, for so few to have so much, independent of others having so little? The juxtaposition of the two ideas the rich and the poor results in another implicit claim: that the distance between the rich and poor is something that is objectionable as well. Even from these questions it is easy to see that people can care about inequality for a number of reasons, or have a diversity of objections to inequality. 3 These philosophical reasons have been the subject of extensive debate amongst political theorists and philosophers for decades. To the extent that people do care about the distribution of economic resources (such as income and wealth), I want to highlight three general frameworks for understanding why people might care about an unequal distribution. Borrowing from a broad range of arguments amongst normative theorists, I label these views the su cientarian view, the limitarian view, and egalitarian views. 2.1 The Su cientarian View One reason for objecting to inequality is the following: that people at the bottom of the distribution don t have enough. This view, known as a su cientarian view, argues that there is a moral ideal in people having enough (su ciency) rather than in people having the same amount (equality). If inequality were objctionable because there is some intrinsic value in equality, we would object to the gap between multi-millionaires and billionaires just as much as we object to the gap between multi-millionaires and people who have nothing. But the fact that we often don t suggests that what we nd deplorable about the gap between the multi-millionaires and the poor is not that the poor have less, but that they simply have too little [Frankfurt, 1987]. Moreover, another reason we would object to the insu ciency of those at the bottom is because we have good reason to believe that their insu ciency is avoidable. Thus, on the su cientarian view, equality and su ciency are logically and morally independent values. What is objectionable about an unequal distribution of wealth is not necessarily the di erence between the haves and have-nots, but that the have-nots live below a threshold of su ciency. Several empirical studies have captured these conceptual distinctions as attitudinal di erences in public opinion. Feldman and Steenbergen [2001] use humanitarianism and egalitarianism scales as predictors of di erent welfare policy preferences. Related 3 This phrase is borrowed from Scanlon [1996]. Scanlon presents a broad taxonomy of di erent objections to inequality, though much of his work focuses on the di erences in welfare and resources on egalitarian grounds. 4

5 to the su cientarian view but more capacious in scope, the trait of humanitarianism is characterized by a belief that the well being of other people is important and the belief that one bears a personal responsibility for this well being. Respondents who scored high on the humanitarianism scale were much more likely to advocate for more social spending on welfare programs; moreover, they were more likely to support only certain types of social spending, such as policies that provided direct aid. Respondents who scored high on the egalitarianism scale, on the other hand, were more likely to support more permanent interventions to address the structural problems of the welfare state. [Feldman and Steenbergen, 2001] How Americans, speci cally, weigh these competing normative values has also been captured empirically. A study by Osberg and Smeeding [2006] has suggested that Americans have become much more tolerant of inequality at the bottom of the distribution, while their attitudes towards inequality at the top end of the distribution have hardened; moreover, Americans seem to care less about social minimums compared to other developed nations. A more recent study by Pew Research, however, suggests the opposite: not only do Americans see poverty and inequality as distinct issues, more Americans seem to care about poverty and think that the government should do something about it than they care about inequality. A large majority of Republicans (64%) and Independents (82%) believe that the government should take some or a lot of action to address poverty, but only 45% of Republicans favor doing the same for inequality [Center, 2014]. Despite con icting conclusions, there is empirical precedent to suggest that we can parse objections to inequality on at least two dimensions, namely, su ciency and equality. Moreover, previous studies also suggest that peoples su cientarian reasons are just as strong if not stronger in some cases than egalitarian ones when it comes to in u- encing redistributive policy preferences. 2.2 The Limitarian View Another reason to object to inequality is a belief that having too much is morally wrong, what Ingrid Robeyns has coined limitarianism. In brief, the limitarian view argues that it is not morally permissible to have more resources than one needs to live a fully ourishing life. While some philosophers do o er arguments for why being rich is intrinsically bad, Robeyns defends a view of limitarianism that, at least on the surface, seems easier to grasp: that having too much is bad for reasons that have to do with political equality and the ability to meet unmet urgent needs [Robeyns]. For example, we have reasons to worry about political inequality stemming from massive economic inequality; because rich people have surplus money, they are both very able and very 5

6 likely to use that money to acquire political in uence and power, [Robeyns, 7]. The rich can wield disproportionate in uence in the political process whether through funding political campaigns, buying votes, agenda-setting, or even their ability to access media outlets to spread and in uence opinions more e ectively. As Gilens and Page [2014] have recently shown, for example, policymakers are much more responsive to the preferences of economic elites (those in the 90th percentile of the income distribution) than they are to the average citizen. The second thrust of the limitarian view suggests that having too much is morally deplorable because there are so many unmet urgent needs for which economic resources are required. The persistence of extreme global poverty, local disadvantages, and pressing collective action problems (such as addressing climate change) all could be signi cantly improved with the appropriate state-led actions and signi cant nancial resources. Thus, it seems plausible that given the moral urgency of such problems, we might object to a distribution of wealth in which some people have too much. The principle of limitarianism of course is much more complex than what I have laid out here. Further clarity is needed on the what the threshold of too much would be, and what the relationship between material resources and human ourishing is, for example. While important, these aspects of the limitarian view are discussed in works by other philosophers and are beyond the scope of this project here. At very least, the core intuition of the limitarian objection that it is morally objectionable for some people to have too much seems to have profound contemporary resonance. The rhetoric of the Occupy Movements, popular outcry against CEO compensations, and caricatures of the inequality debates as class warfare suggest that certain aspects of the limitarian view do shape perceptions and evaluations of inequality today. A number of recent studies have also revealed limitarian undertones in American views of inequality. Many Americans do favor reducing the pay of top earners, particularly CEOs of large national corporations [Page and Jacobs, 2009]. Yet despite the general progressivity of most American redistributive tax policy preferences, there is some con ict over taxing the top earners. Ballard-Rosa et al. [2017], for example, nd that some groups favor lower taxes on the wealthy, while others support raising taxes on the wealthy above a certain income. Osberg and Smeeding [2006] have found that Americans have become increasingly disapproving of income di erentials between top earners. In fact, despite the fact that Americans greatly underestimate the degree of actual inequality 4, there appears to be a wide social consensus around compressing wage di erentials. So not only does the American public largely agree on limiting what someone should earn, they also want that limit to be much lower than we would expect. 4 Americans estimated the gap between CEOs and the average worker to be about 8:1; in 2001, this ratio was 44:1 6

7 These ndings ought to raise the following questions: to what extent can we separate an aversion to having too much from a desire for greater equality? It is plausible to imagine wanting to reduce the CEO-worker ratio signi cantly for the sake of greater equality on the one hand, and wanting to slash income gains at the top because having that much money is inherently morally objectionable? To the extent that these views are separable, however, this study aims to shed light on how limitarian reasons in u- ence perceptions of the fairnesss relative to other reasons. 2.3 Egalitarian views Importantly, the su cientarian view and the limitarian view are not strictly speaking egalitarian views; that is, they do not take the value of equality as their primary goal in rectifying a state of unequal a airs. Su cientarians, for example, would redistribute an unequal distribution of wealth to ameliorate absolute deprivation; greater equality may result, but equalizing material resources is not the primary goal. On the su cientarian or limitarian views, the relative position between individauls is morally irrelevant; for su cientarians, it doesn t matter where people in extreme poverty are relative to the middle class, what matters is that we alleviate their absolute deprivation. Purely egalitarian views, on the other hand, do take equality to be a fundamental value, and thus the relative position between and among individuals is morally relevant for determining what is fair or unfair, just or unjust about a given state of a airs. While most political theorists accept the notion of basic moral equality that everybody counts equally from a moral point of view, regardless of their race, gender, socioeconomic background, and other morally-arbitrary factors that arise from the accident of birth) egalitarian theories come in many stripes. 5 Some may value equality intrinsically, that is, equality for equality s sake. Others may value equality for instrumental reasons, such as equal social standing, or equal political voice and representation. 6 Some 5 This value of equal concern for human dignity is often taken to be the starting point of most egalitarian theories. Ronald Dworkin, for example, takes equal concern to be the foundational virtue of political community and legitimacy, and that a radically unequal distribution of wealth negatively re ects a government s capacity to show equal concern for all of its citizens. In Dworkin s words, No government is legitimate that does not show equal concern for the fate of all those citizens over whom it claims dominion and from whom it claims allegiance. Equal concern is the sovereign virtue of political community without it government is only tyranny and when a nation s wealth is very unequally distributed, as the wealth of even prosperous nations is now, then its equal concern is suspect. ([Dworkin, 2002, 1-2] 6 Derek Par t formalizes this distinction between telic egalitarians and deontic egalitarians: the former believe that it is bad in and of itself if some people have more than others; the latter reject this so-called principle of equality and instead believe that equality is good not just becaue equal outcomes are better, but equality is good for some other moral reason. However, Martin O Neill contends that this simple distinction between telic and deontic egalitarianism obscures a variety of some of the most attractive 7

8 advocate for equality of resources [Dworkin, 1981], while others call for equality of welfare [Arneson, 1999], capabilities [Anderson, 1999, Sen, 1992], or other political or moral goods. For the purposes of this project, I highlight just two egalitarian objections to inequality that fall in the instrumental or non-intrinsic category, particularly in relation to recent empirical work. The rst of these objections is that inequality undermines relationships of equal standing in a democratic society. Underlying this statement is a key assumption that the philosopher Martin O Neill calls the deep social fact: that an improvement in the state of a airs with respect to the distribution of material goods is inextricably linked to the obtainment of other egalitarian goods, such as equal relations free of domination, or o ense to individual dignity [O neill, 2008]. In short, the gap between the rich and poor is objectionable because the widening gulf between the incomes of the top and bottom translate into di erences in access to political power and social standing. Not only do the rich wield more political in uence [Gilens, 2005, Gilens and Page, 2014, Hacker and Pierson, 2010, Bartels, 2008], they also have greater access to social goods. Particularly worrying is the increasing separation between socio-economic classes. As Reich and Satz [2013] write, As inequality has grown, the fate of the few has become unhinged from the fate of the many, as the wealthy nd security in cloistered neighborhoods, send their children to private schools, and attend sporting events in luxury skyboxes. In other words, economic inequality is morally objectionable because it accentuates and deepens di erences in social and political standing, both of which threaten the fabric of a democracy woven by a society of equals. The second view is that large disparities in income and wealth are objectionable because they constitute an impediment to fair equality of opportunity and social mobility. Shrinking opportunity for upward social mobility over time seems to undermine quintessentially American values of meritocracy: that people have an equal chance to get ahead, that overall growth bene ts everyone, and that individual hard work can yield success in life [Newman et al., 2015, McCall, 2013]. However, as recent studies have shown, social mobility overall is lower today than it was a generation ago, while at the same time highly contingent upon to factors such as one s parents income rank and geography [Chetty et al., 2014]. That some people particularly the rich and wellborn appear to be succeeding more than others undercuts American beliefs about equal opportunity: the idea that a rising tide lifts all boats, and that everyone has an equal opportunity to ascend the social ladder [McCall, 2013]. In other words, greater inequalversions of egalitarianism, which he calls non-intrinsic egalitarianism. Among some of these non-intrinsic egalitarian views include the idea that inequality is bad because it is often a condition of deprivation or su ereing, it creates stigmatizing di erences in status, or it leads to unacceptable forms of power and domination. See Par t [1997], O neill [2008], Scanlon [1996] 8

9 ity in wealth and income is symptomatic of a deeper problem: that some people have more opportunities than others, that hard work alone isn t enough to earn success in life. This perception of inequality as inequalities of opportunity, then, is a large input into how Americans assess the fairness or unfairness of inequality. The discussion above is by no means an exhaustive account of the myriad of theories in normative political theory; indeed, there are many more that have not been fully examined here. 7 However, the goal of this project is not to provide grounds for every possible reason that Americans could have for objecting to or tolerating inequality. Rather, it is simply to disentangle views that have, to date, been insu ciently separated in public opinion studies. Thus, in outlining both egalitarian and non-egalitarian reasons for objectiong to inequality, this project seeks to highlight the multidimensionality of views that inform American evaluations of the fairness of a distribution of income and wealth. In the next section, I outline the design of two original survey experiments that take up this task. 3 Study 1: Su ciency or Equality? 3.1 Design The rst study investigates how su cientarian versus egalitarian reasons in uence perceptions of fairness. In this study, respondents were presented with brief descriptions of four hypothetical societies and asked to evaluate how fair they thought the societies were. The two key independent variables featured in each description were the level of poverty (low or high), and the level of inequality (low or high); the table below summarizes the four scenarios. To reduce the cognitive load on survey respondents, descriptions of the inequality condition of each society were intentionally very basic and avoided the use of statistics comparing shares of income or wealth. High poverty was described as 15% of the population doesn t have enough to meet their basic needs, and the low poverty condition was described as almost everybody has enough to meet their basic needs. High inequality was described some people have a lot more than others, while low 7 For example, Segall has made a recent contribution forwarding the account of telic egalitarianism. Segall s argument argues that equality does have an intrinsic value precisely because of badness of inequalities that arise from luck or mutually consented-to inequalities (such as gifts and gambles.) Segall s work departs from other prominent luck egalitarian accounts of equality that presuppose that inequality is bad in an impersonal manner. Segall contends that inequality can be bad in non-instrumental, nonwelfare-a ecting to a speci c person. For a fuller account of the debate within telic egalitarian accounts of the badness of inequality, see also Temkin [2003] 9

10 inequality was described as Nobody has much more than anybody else. Bullet points, rather than full sentences, were used in order to reduce response bias induced from question wording (such as high inequality and low poverty or low inequality but high poverty ). Table 1: Survey Scenarios Inequality Poverty Society 1 High Low Society 2 High High Society 3 Low High Society 4 Low Low The dependent variable was a rating of fairness. Respondents evaluated the fairness of each society on a scale of 1-7 (1 being very unfair, 7 being very fair). Fairness was intentionally left unde ned; what is of interest in this study is not how each respondent de ned fairness, but how certain inputs inequality or poverty di erentially in uence a respondent s conception of what is fair. Respondents were also asked a series of agree/disagree questions on their general attitudes about the rich and poor. The statements were crafted in such a way to capture a respondent s beliefs about the agency: who are what is responsible for a person s success in life. For example, the idea that People are rich because they were born under the right circumstances attributes wealth to luck; on the other hand, the idea that people are rich because they work hard to get to where they are, attributes wealth to individual work ethic. Finally, the idea that people are rich because the political and economic system gives them more advantages attributes wealth to exogenous factors such as political power and economic forces. An analogous set of questions asked respondents to indicate their views about the poor. The full list of these questions appears in the Appendix. The study was elded in July of 2014 through Survey Sampling International (SSI) as part of an omnibus survey that covered three di erent topics: the ethics of weapons development, attitudes towards socio-economic inequality and fairness, and international law. Part II attitudes about socio-economic inequality and fairness is the relevant portion for this study. The sample was composed of 1350 American adults balanced across gender, age, education, race, and political party according to US census targets. 273 of the observations were dropped from the sample for speeding (spending less than half the median time on the survey, or not answering the relevant section of the survey), which brought the total number of observations for this study to

11 3.2 Key Findings The average ratings for each society are reported in Table 1 below and the difference in means (i.e., the e ects of poverty and inequality) are plotted in Plot 1. On average, respondents gave the highest rating to the society with low poverty and low inequality (average score of 4.50), while they gave the lowest rating to the society with high inequality and high poverty (average score of 2.96). Respondents rated the society with low poverty and high inequality approximately 1 point higher than the society with high poverty and low inequality (4.15 versus 3.17). The di erences in mean ratings illuminate the di erential e ects of poverty and inequality on a respondent s evaluation of fairness. On average, the e ect of poverty is signi cantly greater than the e ect of inequality on a respondent s evaluation of fairness. Holding inequality constant, the e ect of poverty is at least 3 times greater than the e ect of inequality when poverty is held constant. In addition, all 1077 respondents ranked societies with low poverty higher than the societies with high poverty. Such ndings suggest that at very least, su cientarian evaluations are not only distinct from egalitarian reasoning in view about fairness, but also that the su cientarian reaons matter more. In other words, when it comes to fairness, Americans appear to express neither a strong inequality aversion or a preference for equality, but rather a strong preference for su ciency. Table 2: Sample Means and Di erences in Means Low Poverty High Poverty Low Inequality E ect of poverty when inequality is low High Inequality E ect of poverty when inequality is high E ect of inequality when E ect of inequality poverty is low: when poverty is high.36**.21** [.22,.50] [.11,.31] 1.33** [1.22, 1.44] 1.18** [1.09, 1.28] Partisanship and Evaluations of Fairness One of the most central issues that distinguishes the two major parties in the US is di erent views about the appropriate role of government in redistributing resources and guiding economic growth. Previous studies have demonstrated the strong link between partisanship and redistributive policy preferences [Campbell et al., 1960, Bartels, 2002, 11

12 Figure 1: 2008, Lenz, 2009, Ballard-Rosa et al., 2017]. Following this line of work, this study also nds that partisanship has strong links to evaluations of inequality and fairness. Average ratings of fairness by party ID are reported in Table 3 and Figure 2 below. On average, both Democrats and Republicans rated societies that met the su ciency condition higher than the societies that did not; the average e ects of poverty were substantially higher than the average e ects of inequality across both groups (between about 1.3 and close to 4 times larger than the e ect of inequality). The di erential effects of inequality, however, are notable. Democrats rated societies with high inequality almost a full point lower (less fair) than their Republican counterparts. Holding poverty constant, moving from a low- to high-inequality society drove fairness ratings down amongst Democrats, but the opposite was true for Republicans. In fact, societies with high levels of inequality were rated more fair by Republicans. Put another way, these results suggest that respondents who identi ed as Democrats show much more inequality aversion while those who identi ed as Republicans appear to have a preference for inequality. It is not necessarily party, per se, that explains these di erences. Rather, partisanship may simply be capturing a set of more fundamental worldviews about what determines success in life, which in turn, shape beliefs about what s fair. For example, while a large majority (62% of Democrats, 77% of Republicans) agree that The rich work hard to get to where they are, there is signi cantly more disagreement over whether 12

13 Table 3: Di erence in Means: Party ID E ect Low Poverty High Poverty Low Inequality High Inequality E ect of inequality when poverty is low: -.52** [-.83, -.21].88 ** [.68, 1.08] E ect of inequality when poverty is high: -.30 [-.546, -.06].46 ** [.32,.60] E ect of poverty when inequality is low E ect of poverty when inequality is high.90 ** [.66, 1.13] 1.55 ** [1.37, 1.74] 1.11 ** [.92, 1.31] 1.14 ** [.97, 1.30] poverty is the result of an individual failure to work hard. 45% of Republicans agree that Some people are poor because they don t work hard enough, while just 23% of Democrats agree with that idea. A larger proportion of Democrats (approximately 74%, compared to 67% of Republicans) agree that Some people are rich because they were born under more fortunate circumstances. That di erence widens when asked about whether wealth might be the result of the political and economic advantage. 64% of Democrats agree that Some people are rich because the political and economic system works to their advantage, while just 47% of Republicans agreed. Similarly, 68% of Democrats agreed with the idea that poverty is the result of the political and economic system, while close to 60% of Republicans rejected it. In light of these ndings, we can better understand what it is that partisanship captures when we analyze the di erent evaluations of fairness. Republicans, for example, might seem much more tolerant of inequality and to an extent, prefer inequality because they share a belief about just deserts. A society with high inequality, then, is one that justly rewards people for their di erential e orts. Such a view is not inconsistent with other egalitarian concerns about political and moral equality, however; indeed, as scholars have noted, most Americans hold pluralistic views about fairness in which people are di erentially rewarded in the economic domain, while equally regarded in the political and social domains [Hochschild, 1981]. Thus, at least amongst Republicans in the sample, a rming a positive belief in the moral idea of su ciency is not inconsistent with a certain degree of inegalitarianism: above a threshold of su ciency, that 13

14 Figure 2: some people get more than others is not only fair, but more fair than if everyone were equal in terms of material resources. However, the nature of this survey design yields a potential observational equivalence problem. For example, take the case of the society with high inequality and low poverty. Two respondents can give this society the same rating for radically di erent reasons: one, with the view that su ciency is what s fair, the other, with the view that inequality is fair. Because the questionnaire did not provide information on how the distribution of the hypothetical societies had come about, we cannot determine the extent to which respondent s evaluations of fairness are motivated more by a strong su cientarian commitment, or a conception of fairness that accepts inequality as an inherently fair if not desirable feature of society. In spite of these indeterminacies, this rst study does shed light on the di erential inputs in American beliefs about inequality and fairness. The ndings of this study suggest that broadly speaking, fairness does not necessarily imply an aversion to inequality, but more importantly, aversion to insu ciency. However, su ciency and equality are not the only two relevant dimensions of fairness worth investigating. 14

15 4 Study 2: Su ciency, Equality, and Mobility 4.1 Design A second study expands on the rst by investigating how Americans evaluate the fairness of a distribution along three dimensions: su ciency, equality, and mobility. This study leveraged a choice-based conjoint survey experiment design to evaluate how these di erent attributes in uenced peoples perceptions of fairness. Respondents were presented with three pairs of pro les of hypothetical countries. Each country prole provided information on three attributes: the incomes of the top 10%, the incomes of the bottom 10%, and the chances that someone born into poverty becomes rich over time. While only these three attributes were explicitly de ned, an additional attribute is implicit in this design, namely, the level of inequality, captured in the di erence between the incomes of the top and bottom. The main task of each respondent was to choose which country was more fair. The values of the attributes were randomly assigned for each country pro le. Furthermore, the order in which the attributes were listed on the country pro les were also randomized to avoid primacy or recency e ects. Attributes and levels of each attribute are reported in the table below. A screen capture image of what was displayed to respondents is also included. Table 4: : Country Pro le Attributes and Levels Attribute Level The bottom 10% of people make $10,000 $20,000 $30,000 $40,000 The richest 10% of people make $100,000 $200,000 $300,000 $400,000 The chance that someone born poor becomes rich over time very low- 5% low - 10% high - 25% very high - 40% Respondents were also asked to indicate how much they agreed or disagreed with statements about inequality, wealth, and poverty. As in the rst study, statements such 15

16 as People are rich or poor mostly because of circumstances within their control attempted to capture prior a respondent s beliefs about individual agency determining one s success in life. Also included were more direct statements that probed respondent s egalitarian or su cientarian leanings, such as Large di erences in income and wealth are acceptable so long as everybody has enough to meet their basic needs, or Our society would be a lot better o if we reduced the gap between the rich and poor. The full list of these questions is included in the Appendix. This study was elded by YouGov in 2015 as part of an omnibus survey that covered several topics including inequality, the ethics of war, immigration, law and morality, and labor unions. The sample was composed of a panel of 1000 American adults, balanced across gender, age, education, race, and political party according to US census targets. All 1000 respondents were asked to evaluate three randomly assigned conjoint pairs of country pro les, bringing the total number of observations in this study to Key Findings Assuming no carryover or pro le order e ects, the conjoint design allows for the use of simple linear regression to estimate average marginal component e ects (AMCEs) of each attribute [Hainmueller et al., 2013]. For this study, I estimated the AMCEs of di erent levels of each attribute on the probability that a pro le with those attributes was chosen as more fair by the respondent. More speci cally, I regressed the dichotomized 16

17 outcome variable choose on a set of variables that corresponded to the di erent attributes of a given country pro le. The regression coe cient for each variable indicates the AMCE of that particular level of an attribute relative to the reference category of that attribute in this study, the lowest level of income or chance of social mobility in each attribute group. Standard errors are clustered by respondent, and are reported along with regression coe cients for the full set of variables in the Appendix. Figure 3 reports the estimated AMCE of each attribute level on the probability that a pro le with that feature is chosen by the respondent as more fair. The bars indicate 95% con dence intervals, and the estimates are interpreted relative to the reference category (indicated at the vertical line at 0). In line with the ndings of the rst study, evaluations of fairness were highly sensitive to the su ciency condition. Respondents were roughly 9.5% more likely to choose a country as more fair when the incomes of the poor were raised from $10,000 to $20,000. Doubling the incomes of the poor from $20,000 to $40,000 more than doubled the likelihood that a respondant would view that country as more fair than the alternative. These changes in probability are not only statistically signi cant, but relatively large, especially when comparing them to the changes in probability associated with other attributes. Compared to changing the incomes of the bottom 10%, changing the incomes of the top 10% appears to have little e ect on respondents evaluations of fairness. Doubling the incomes of top earners from $100,000 to $200,000 made respondents only about 5% more likely to choose a pro le with that attribute as more fair than the alternative. Furthermore, there is no discernable pattern as the top incomes continue to increase. For example, when the incomes of top earners are raised to $400,000 (the highest level of that attribute), the likelihood that a respondent chose that pro le as more fair is about the same as when the incomes of top earners was only $200,000 (about 5.3% and 5.6%). None of these changes in probability associated with changes in the incomes of the top earners are statistically signi cant. Compared to the e ects of other attributes, estimates of the e ect of the income gap computed as the di erence between the displayed levels of the top 10% s earnings and the bottom 10% s earning are relatively insigni cant. As the gap between incomes increases, there is no discernible pattern in the change in probability that a respondent chooses the pro le with that particular level of inequality. For example, in the entire sample, when the gap between the top and bottom is $80,000, respondents are about 6% more likely to choose that pro le as more fair than the alternative. Widening the gap by a factor of 2 to $160,000, however, apparently decreased the likelihood of choosing that pro le to 2.5%. When the income gap at the maximum observed level of $280,000, the likelihood of choosing that pro le is about 8%. 17

18 Figure 3: 18

19 The social mobility attribute had the largest e ect on respondents evaluations of fairness. On average, increasing the chances of upward social mobility from 5% to 10%, increased the probability of choosing that pro le by about 8%; and, in the (rather unrealistic) scenario in which the probability of upward social mobility was a whopping 40%, respondents were about 30% more likely to choose that pro le. Thus, these ndings suggest that Americans can and do weigh a plurality of values di erently when evaluating the fairness of an unequal distribution of economic, though inequality itself plays a minor role in shaping their views. Su ciency and mobility not equality of condition are the biggest drivers of respondents concerns about fairness. The palpable e ects of increasing the levels of social mobility suggest that, to the extent that respondents are concerned about inequality, that concern is best understand as an aversion to unfair opportunity for upward mobility, rather than as an inherent objection to di erences in income. The trivial e ects of increasing the gap between the rich and poor and raising the incomes of the rich (which, in the most extreme cases, could have been up to 40 times as much as those of the poor) suggest that intrinsic egalitarian and limitarian views are, at best, weak inputs into American beliefs about inequality and fairness Partisanship and Evaluations of Fairness Table 5 and Figure 4 report the average marginal component e ects of attribute levels by self-reported ID. Several ndings are consistent with the previous study and reinforce the strong link between partisanship, and normative values associated with fairness. Democrats in the sample were highly sensitive to changes in levels of income at the bottom. A change in the level of the bottom 10% s earning from $10,000 to $30,000, for example, is associated with a 20% increase in likelihood that a pro le with that characteristic was rated more fair. By contrast, Republicans in the sample appear indi erent to this attribute; there is no discernible pattern nor statistically signi cant change in probability associated with changes in the earnings of the poor. At least on this dimension, respondents who reported themselves as Independents align more closely with Democrats than with Republicans; changes in the earnings of the bottom 10% were associated with a positive change in probability that a pro le was chosen as more fair than the alternative. Similarly, the e ect of social mobility on the probability of choosing a pro le held across Democrats and Republicans, though Republicans were on average more sensitive to changes in the level of upward mobility than Democrats. Increasing the chance of upward social mobility from 5% (the reference category, or baseline) to 10%, for exam- 19

20 Figure 4: 20

21 ple, bumped the probability of choosing that pro le by about 5% amongst Democrats, an estimate that is neither statistically or substantively signi cant. Republicans in the sample, however, were almost twice as likely to choose the pro le with the latter attribute a substantively and statistically signi cant di erence. These subtle heterogeneous e ects might best be interpreted as re ections of deeper di erences in partisan beliefs, particularly beliefs about the origins of wealth and poverty. In this sample, Democrats and Republicans appear sharply divided over whether or not individual e ort is enough to guarantee success in life: close to 90% of Democrats agreed that For most people, hard work is not enough to guarantee success, while just over half of Republicans (51%) agreed. (Independents, perhaps predictably, split the di erence, with 73% agreeing with the statement.) Conversely, just 33% of Democrats agreed that People are rich or poor mostly because of circumstances within their control, while approximatly 74% of Republicans agreed. Thus, Democrats greater sensitivity to changes in the earnings of the bottom 10%, for example, may stem purely from a self-interest bias (generally speaking, people who identify as low-income are more likely to identify with the Democratic party), but may also stem from prior beliefs about e ort and luck determining success in life. If success is determined not by hard work, but by luck, one may be more inclined to believe that higher social mimimums for the least-well o are de nitive features of a fair society. By contrast, for Republicans in the sample, the notion of more opportunity rather than a guaranteed minimum may cohere better with a conception of fairness that takes individual e ort as the primary determinant of success. Neither of these explanations, however, seem likely given that the conjoint setup did not include information about where incomes came from in each of the pro les. At the other end of the scale, there is little evidence suggestive of limitarian leanings within or across partisan lines. Increasing the incomes of the top 10%, even by a factor of 4, has almost no bearing on respondents evaluations of fairness. Contrary to popular outcry against The 1%, the relative lack of limitarian objections is not surprising. The overwhelming majority of respondents (approximatly 88% of Democrats, 97% of Republicans, and 90% of Independents) agreed that People who are wealthy have the right to enjoy their money however they want. This shared belief in a right to the fruit s of one s labor even for the very rich may explain why on average, the e ect of raising the incomes of the top 10%, relative to other attributes, is both statistically insigni cant and substantively small across the entire sample. Likewise, the distance between the top and bottom earnings had little e ect on perceptions of fairness, which casts doubt on the prevalence of egalitarian concerns. As discussed above, almost all of the estimated changes in probability associated with changes in the income gap are statistically insigni cant, and there is no discernable pattern across 21

22 the entire sample. There are subtle di erences in patterns between partisan groups, however. Amongst respondents who identi ed as Democrats, the e ect of widening the income gap tended to be negative, that is, the larger the income gap, the less likely respondents were to choose that pro le was the fairer of the two. Amongst republicans, though, the e ect worked in the opposite direction: income gaps at the smaller end of the range ($70,000 or $90,000 for example) were associated with a negative probability that the pro le was chosen, while income gaps at the higher end of hte range ($260,000 or $270,000 for example) were associated with a positive probability. While the standard errors are too large for the data to be conclusive, it is not implausible that Republicans would be more likely to view societies with more inequality is more fair a notion consistent with the ndings of Study 1. Thus, we might stipulate that Democrats in the sample exhibited slight (though not signi cant) inequality aversion, while Republicans revealed a weak (though again, not signi cant) preference for equality. The relative lack or weakness of any inequality aversion is striking, particularly in light of dramatic di erences in the way partisans responded to questions speci clly about inequality. Approximately 95% of Democrats agreed that our society would be a lot better o if we reduced the gap between the rich and poor, while just 49% of Republicans agreed. While one could hypothesize about whether or not individuals had instrumental or intrinsic reasons in answering this question, the statement explicitly asks the respondent to consider the gap between the rich and poor, rather than the welfare of the poorest or the earnings of the richest. Thus, when asked directly, partisans express a preference for equality at strikingly di erent rates; yet when confronted with conditions with actual di erences in income, that preference appears to play little role in their evaluations of fairness. 5 Discussion and Suggestions for Further Research This paper began with the simple premise that American beliefs about inequality are not as ambivalent or uninformed as often suggested. To that end, this study has striven to illuminate the diversity of objections to inequality, not just in theory but also in the public imagination. Three key conclusions can be drawn from this study. The rst is a broader conceptual point grounded in various normative theories: that there are reasons to object to inequality that do not need to be grounded in purely egalitarian concerns. One can object to an unequal distribution of economic goods not necessarily because one thinks that everybody having the same is morally desirable (i.e., one has a preference for equality); one can object because some people have too much (the limitarian view), or that some people don t have enough (the su cientarian view). This conceptual distinction has 22