The Effect of White Social Prejudice on Support for American Democracy

Abstract

Social prejudice constitutes an unwillingness to associate with individuals whose cultural or racial background differs from one's own group. Such prejudice is a particularly thorny problem in the context of democracy, which requires citizens to minimally respect such differences. In this paper, we assess the relationships between these attitudes and support for democratic institutions. Using World Values Survey data from 1995 to 2011, we find that prejudice toward cultural, ethnic, or racial “others” reduces the value that white Americans assign to democracy. We also find white Americans who exhibit these attitudes are more likely to dismiss the value of separation of powers and are more likely to support army rule. These findings imply that exclusionary rhetoric targeted toward non-white groups is accompanied by lower baseline support for democracy. We close with a discussion of how our analyses inform the study of Americans' attitudes toward democracy

INTRODUCTION

Democracy has been relatively durable in the United States. The irony of its longevity, however, is that many of the freedoms commonly associated with democracy have only slowly been extended to every group in the mass public (Bateman 2018). The uneven application of certain constitutional protections and guarantees to non-white citizens is of particular importance when highlighting democracy's shortcomings in the United States. Simply put, the United States possesses a fraught racial history in which prejudice often manifests in behavior not befitting a democratic polity.

While there has been obvious progress with respect to the extension of various civil freedoms to non-white Americans over time, the racial attitudes of white Americans remain mixed. The democratization of authoritarian enclaves in the South by the middle of the 20th century (Mickey 2015) coincided with a civil rights movement that generated considerable racial resentment in its immediate wake (e.g. Kinder and Sanders 1996; McConahay and Hough 1976). In fact, decreases in racial resentment (Hopkins and WA 2019) and the endorsement of negative racial stereotypes notwithstanding (DeSante and Smith, n.d.), negative racial affect among whites continues to play a pervasive role in mass politics (e.g. Banks and Valentino 2012; Gilens 1999; Jardina 2019; Tesler 2016; Valentino, Neuner, and Vandenbroek 2018).

To the extent that democracy involves institutionalizing the protections of minority groups, it is also possible that prejudice towards such groups undercuts support for it. Past research, for example, reveals that social prejudice drives support for exclusion within the polity (Glenn 2004; Kivisto and Rezaev 2018; Tileaga 2015). Yet, the older work on support for democracy has not sufficiently wrestled with whether or not prejudice spills into preferences for democratic self-governance. This gap is particularly relevant in light of observations from organizations like Bright Line Watch, which finds public support for democracy's core features is weaker than presumed.

In this paper, we investigate the relationship between prejudice and support for democracy. Our analysis of white Americans across four waves of World Values Survey data from 1995 to 2011 finds support for the notion that the rejection of immigrants, those who speak a different language, or those from a different race as neighbors—attitudes that describe a “social” form of prejudice—are associated with increased support for strongman and army rule, as well as decreased support for democracy, in general. We conclude with a discussion of the important implications these findings have for what we know about support for democracy in the United States.

A THEORETICAL ARGUMENT LINKING SOCIAL PREJUDICE TO SUPPORT FOR DEMOCRACY

Political solutions regarding what is “best” or “right” for a citizenry are sure to embody different functional forms in pluralistic societies. When these competing demands fracture along existing racial or ethnic cleavages, then politics becomes particularly tense (Kivisto and Rezaev 2018). As a result, a mutual commitment to basic norms of tolerance is an important condition of peaceable exchange in democracy (cf. Gibson 2011). While classic scholarship on democracy builds in assumptions that majorities must tolerate the right of minorities within a democratic institutional design (Dahl 1989), these idealized assumptions have often failed in the United States. Its history is replete with examples where antipathies toward the immigration and integration of non-whites into society have generated considerable political and social strife.¹

If democracy involves serious commitments to pluralism (Bohman 2000), then interacting with members of different racial groups in social settings is necessary. Given the racial tensions involved in mid-century social and political life, Allport's (1954) early work in The Nature of Prejudice suggested that social contact between (racial) groups might reduce prejudice under certain conditions. Of course, these conditions are rarely met (c.f. Pettigrew and Tropp 2006),² and research findings involving the social contact thesis are often mixed regarding the relationship between contact and discrimination (e.g. Scacco and Warren 2018). Still, the ideas found in this work are relevant here. At least minimally, a commitment to tolerance and engaging in neighborly actions toward others seems to embody the commitments associated with a healthy community and, more broadly, democracy itself. What happens, however, when members of a racial group express a preference for avoiding having racial or ethnic “others” as their neighbors—when they express that they literally want to avoid such explicit contact? Do white Americans who report a desire to avoid having an immigrant, a person who speaks a foreign language, or a racial minority as a neighbor—attitudes that we believe constitute a “social” form of prejudice—also exhibit less positive orientations toward democracy?

To be clear, such preferences may not encompass an explicit expression of hostility of the sort that is generally associated with patently racist rhetoric. Instead, such attitudes involve a resistance to multiculturalism, thereby functioning as a symbolic form of prejudice.³ Such views are problematic in the sense that they undercut the sort of social contact that embodies a plural society (Berry 2006); moreover, the acceptance of cultural differences is a core value in democratic societies (Fowers and Richardson 1996). If “prejudice occurs when people are placed at some disadvantage that is not warranted by their individual actions or qualifications” (Eagly and Diekman 2005, 23–24), then disfavoring a neighbor on the basis of their racial or ethnic group membership aptly describes a potent form of social prejudice.⁴

Our argument linking social prejudice to attitudes about democracy begins with some fundamentals regarding what we know about intergroup behavior. Prejudice involves high levels of expressive identity (Rapp and Ackermann 2016). In fact, the emergence of discriminatory attitudes requires some sort of social situation that makes categorical distinctions among groups salient (e.g. Brewer 1999). The history of racial strife in the United States, for example, ensures that race functions as a categorical distinction among citizens. This development has political implications—at least with respect to support for democracy—in the sense that restructuring the social environment into opponents and allies might generate prejudice against persons who “threaten” the prevailing racial in-group (e.g. Allport 1954; Gaertner and Dovidio 2005; Kinder and Kam 2009).

In turn, white Americans who feel threatened by racial or ethnic diversity—those persons who exhibit social prejudice—ought to hold systems of governance that extend access to these individuals in lower regard due to both perceived material and symbolic threats to their own group's status (Bahns 2017). Democracy involves, in no small part, the allocation of both goods and power. We know that general economic threat can destabilize citizens' commitment to democracy (Miller 2017), but we expect a particular discomfort with members of minority groups to lead to negative summary assessments of democracy, which values such groups' rights to social and cultural expression. Thus, we expect white Americans who harbor social prejudice toward a variety of minority outgroups to be less accepting and supportive of democracy because it provides the political pretext for undesirable minority outgroups to engage in self-expression. Prejudiced white Americans—those persons who do not want a racial other as a neighbor—should be more receptive to non-democratic alternatives of governance for the United States. In particular, they should desire authoritarian alternatives to democratic rule—a preference for strongman or army governance—and generally exhibit lower levels of democratic support.

H1: White Americans who are prejudiced against ethnic/racial outgroups should be more receptive to the idea of non-democratic forms of government for the United States than white Americans who are not prejudiced against these same ethnic/racial outgroups.

RESEARCH DESIGN

The World Values Survey (WVS) offers the best possible means to test our argument across time in the United States. Its depth of questions on democratic support is well-traveled and robust (Ariely and Davidov 2011), informing much of what we know about how individuals evaluate democracy vis-a-vis other alternatives. Further, WVS data in the United States show considerable detail to questions of prejudice that we address here. Our analysis uses the third, fourth, fifth, and six waves of WVS data in the United States, spanning observations from 1995 to 2011. Importantly, the third wave is the first wave of WVS data in which survey researchers ask important questions about diffuse support for democracy, which rules out the inclusion of earlier survey waves. We select on those respondents in these four waves that self-identify as white to test our argument linking social prejudice against minority groups to anti-democratic orientations.

Dependent Variables

We leverage three survey items widely used throughout the literature on mass-level support for democracy and orientations citizens have toward democratic and autocratic governance. The three questions, first introduced by WVS in its third survey wave, constitute particularly severe breaks with diffuse support for democracy where individuals are asked whether they thought the following were good or bad ways of governing the United States:

1. (1) Having a strong leader who does not have to bother with parliament and elections

2. (2) Having the army rule the government

3. (3) Having a democratic political system

The benefits of these metrics are multiple. On the one hand, these items reflect a generalized preference for authoritarian systems (e.g. Ariely and Davidov 2011; Magalhaes 2014). Yet, they are also interesting on their own merits. Consider that the first item taps into general support for whether democratic governance is good by proposing an alternative in which a strongman governs the United States without oversight from the national legislature (i.e. Congress) or regular elections. In hindsight, this question is as timely as ever given the election of President Trump and his open infatuation with similar leaders like Rodrigo Duterte and Vladimir Putin, who are not subject to the same institutional limitations. The third item, meanwhile, literally proposes waiving away democracy altogether in favor of an unspecified alternative. Outright opposition to American democracy also coincides with a meaningful preference for authoritarian leadership even if it is not specified in the prompt.

The second item captures the acceptability of army rule. On balance, Americans may have little grasp of the idea that military rule often means bloodshed. Autocratic regimes in which a military coup installs leadership rarely end well (e.g. Svolik 2013). Yet, the glorification of militancy by persons who harbor social prejudice in America is as common as it is normatively troubling. From the manifestos of individuals who commit murderous acts against non-white persons to the rhetoric and marches of neo-Nazis, white social prejudice routinely cloaks itself in military regalia.

The original scale of these variables is ordinal in which the respondent could think such a system is very bad, bad, good, or very good for the United States. We condense each of these variables to a binary indicator. Responses equal a 1 for those people who thought a strong leader or army rule was “good” or “very good” for those two dependent variables. We chose to recode the “having a democratic political system” variable to be a 1 if the respondent thought it was “very bad” or “bad.” We do this to facilitate reading the direction of the regression results across all three models.

Primary Independent Variable

We measure social prejudice through questions at the core of the prejudice literature, especially for those working on behavioral manifestations of prejudice in acts of exclusion. WVS prompts its respondents to say what types of people that the respondent would not like to have as neighbors. The respondent can name any they like from a set list of familiar groups. The prompt itself has also evolved over time and is sensitive to different countries and contexts. The available responses in the United States are included in Figure 1.

Figure 1. Percent of White Americans in WVS Who Would Not Want This Type of Person as a Neighbor (WVS, 1995–2011).

Notice the variety that the WVS has included in this battery of survey prompts. It does not coerce a response easily construed as prejudiced toward an ethnic/racial minority unless this represented the respondent's earnest preference. Indeed, most responses focused on groups like drug addicts, heavy drinkers, and criminals. Figure 1 shows that over 84% of respondents included drug addicts as a type of neighbor a white respondent would not want. The next most frequently mentioned groups are heavy drinkers, criminals, and emotionally unstable people.

Our review of right-wing hate group literature and a review of language that white nationalists use identifies the responses of members of a different race, immigrants/foreign workers, or those who speak a different language as responses of interest. White nationalists and nativists routinely single out these groups, with varying levels of subtexts, as symbolic threats to status. We code a dummy that equals a 1 if a respondent identified any one of those as an unwelcome neighbor. Whereas these responses are not mutually exclusive and a respondent could conceivably list every single person on this list as an unwelcome neighbor, the measure we have for white social prejudice ultimately codes a 1 for 18.7% of all white respondents across the four waves we use here.

Control Variables

We include multiple individual-level control variables. Several basic sociodemographic indicators appear in the model. We include the respondent's age in years as well as a square term for age to discern if there is a curvilinear effect of age on these anti-democratic preferences. We include a dummy if the respondent was a woman or unemployed. We measure income on the ten-point “scale of incomes” question in which the WVS prompts its respondents to conceptualize their income as deciles relative to the respondent's own country. Higher values indicate more perceived household income. We also note the importance of education as a correlate of both tolerance and democratic support (e.g. Henry and Napier 2017; Lipset 1959; Vogt 1997). Our appendix includes a series of analyses that unpack the effect of social prejudice by the level of education but the main analyses we report here include a binary variable that equals a 1 if the respondent completed a 4-year college degree. Those with at least some college experience or even less than that are a 0.

We included controls that capture important political values. We use the ten-point ideology continuum, in which the respondent lists her or his ideology from left to right on a unidimensional scale. We also include a square term for ideology to test for a curvilinear effect. Namely, ideological extremists might be differently disposed toward democracy as a political system than ideological centrists (Adler 2018). We include fixed effects for partisanship that controls the effect of being a Republican or Democrat relative to a baseline of third-party supporters and self-described independents. Our final control variable is Welzel's (2013) “emancipative values” index. Welzel's emancipative values consists of four components of “autonomy” (i.e. analogous to the child autonomy index), “choice” (i.e. the justifiability of abortion, divorce, and homosexuality), “equality” (i.e. attitudes toward gender equality on the job, in politics, and educational opportunities), and “voice” (i.e. how much the respondent believes having a say in government is a political and personal priority). Welzel (2013) argues these are syndromes of a sort that form when existential pressures decrease and individuals are at greater will to demand more freedoms (i.e. more democracy) from their government. Miller (2017) finds a robust effect of this index on democratic orientations across the world, for which the strength of this effect could influence the effect of the primary independent variable that concerns this analysis.

Model and Robustness Notes

The three dependent variables we use are binary, making a logistic transformation of regression coefficients appropriate. The descriptive statistics we report in the appendix show no real concerns for the extent of missing data and we thus employ listwise deletion for cases with missing data. However, we want to address additional concerns about temporal heterogeneity in the data. We are using only American responses from the WVS data, but we note that there is a concern for variation over time in the data and that the effects we are observing may be changing over time. Thus, we include a random effect for the survey year (i.e. 1995, 1999, 2006, 2011) and address the issue of the relatively small number of years by estimating all models in this analysis with weakly informative Wishart priors on the covariance matrices (cf. Chung et al. 2015).⁵ We additionally standardize all coefficients by two standard deviations. This is considered both good practice for those who run mixed-effects models and it has the added effect of putting all variables, roughly, on a common scale (Gelman 2008). This allows for a preliminary comparison of coefficient sizes across the models we run.

We also refer the reader to an online appendix that includes several dozen robustness tests that address sample selection issues, the measurement of our main independent variable and dependent variables, and issues of spatial and temporal heterogeneity. We note our several dozen robustness tests do not at all change the inferences we report here.

RESULTS

We start with Figure 2, a dot-and-whisker plot that summarizes the results of the three regressions that explain attitudes toward support for a strong leader, support for army rule of the government, and opposition to democracy. Here, the dots are coefficient estimates and the whiskers correspond with 95% intervals around the point estimate. A vertical line at zero represents the null hypothesis of zero effect. A dot-and-whisker that does not overlap with this vertical line communicates a statistically significant effect.

Figure 2. Dot-and-Whisker Plots of the Covariates of Democratic Orientations of White Americans in the World Values Survey (1995–2011).

Age has a consistent negative effect on attitudes in favor of various authoritarian alternatives to democratic governance in the United States even though the insignificance of age's square term suggests we cannot discern a curvilinear effect in the data. A college education also has a mostly robust, negative effect on anti-democratic orientations. The coefficient is negative and discernible from zero in both the opposition to democracy analysis and the analysis on support for a strong leader unencumbered by legislative or electoral oversight. This is consistent with a wide body of scholarship on the importance of education to democracy (e.g. Lipset 1959). We use the appendix to explore more thoroughly the nexus among education, white social prejudice, and support for democracy, finding illustrative evidence that the effect of social prejudice on lowered support for democracy may be stronger on the better educated.

There are few consistent effects across all three models, which suggest the effect of several predictors on anti-democratic orientations may have substantial variation even if all three dependent variables we use are capturing latent support for authoritarianism or opposition to democracy. There are a few discernible differences between men and women. White women are more likely than white men to oppose democracy in the United States and they are more open to rule of government by the army, though there is no discernible difference between men and women on support for strongman rule in the United States. There is no real income difference we were able to discern. Generally, white Americans higher in the income decile scale are less likely to oppose having democracy but the effect of increasing income on support for army rule and strongman rule is effectively zero.

Controls for social and political values also have some inconsistent effects. Increasing ideology has a significant (negative) effect in the two models explaining attitudes in favor of army rule or attitudes opposing democracy for the United States. The square term for ideology has a significant and negative effect on the models explaining attitudes in favor of strongman rule and army rule. Generally, these communicate that increasing ideology to the right decreases the likelihood of an anti-democratic orientation and that the effect might be a little stronger on those white respondents furthest to the right. However, there is considerable variation in these parameters and our analyses do not suggest this effect is robust. The emancipative values variable, which otherwise has some of the strongest effects that Welzel (2013) and Miller (2017) report in their cross-national analyses, has much more muted effects in the United States. The coefficient is insignificant in the second model that explains variation in attitudes toward army rule. Partisanship explains relatively little in our analyses as well. White Republicans are less likely than white independents and third-party supporters to support a strong leader and white Democrats are less likely than white independents and third-party supporters to oppose having a democracy for the United States.

Our primary concern in this analysis is the effect of white social prejudice, which we operationalize as whether a white American does not want members of a different race, immigrants/foreign workers, or those who speak a different language as neighbors. Here we find a robust, positive effect. A respondent that lists one or more of these groups as unwelcome potential neighbors is much more likely to support a strong leader who is unencumbered by legislative checks and balances, is much more open to rule of government by the army, and is much more likely to outright oppose having democracy in the United States. This is consistent with our argument that white Americans prejudiced against ethnolinguistic difference are much more likely to see democracy as empowering these minority groups beyond their numerical endowment, extending rights and liberties to groups that these white Americans see as unwelcome.

We illustrate the effects of white social prejudice on anti-democratic orientations as faceted ridgeline plots in Figure 3. Here, we set all the controls at zero, which creates rows for employed male independents/third-party supporters without a college education, of average age, with the average income and social and political values. Thereafter, we allow the binary indicator for white social prejudice to vary from 0 to 1 for the white social prejudice measure we devise. We run 1,000 simulations of the model to get an expected likelihood (i.e. a probability) of observing a 1 on the dependent variable (i.e. supporting strongman rule, army rule, or opposing democracy for the United States). We do the same process iterating out the white social prejudice measure and adding instead an individual component of our white social prejudice measure (e.g. respondent would not want an immigrant/foreign worker as a neighbor, a person of a different race as neighbor, etc.). We calculate average first differences between those expected values and communicate the results as probability distributions for which 95% intervals around the mean are enclosed in brackets and shaded darker than the entire distribution. We also annotate each ridge in the plot with the mean of the first differences, the 95% intervals around the mean (in parentheses), and the proportion of simulations with negative first differences [in brackets]. Negative first differences are results of simulations in which a socially prejudiced average white male had a higher likelihood of an anti-democratic orientation than an average white male without these prejudices. These results would be inconsistent with our hypothesis.

Figure 3. Ridgeline Plots of Various Indicators of White Social Prejudice on Democratic Orientations.

The results show that the 95% intervals around these distributions exclude 0 in all but two estimations. Those two exceptions are the effect of a white American not wanting a neighbor who spoke a different language on opposition to democracy and the effect of not wanting a neighbor of a different race on support for the rule of the U.S. government by the army. We qualify the latter case because the 95% intervals we report surround the mean of the distribution. These simulations are ultimately one-tail and only 2.8% of the simulated first differences, in that case, were negative.

Elsewhere, the distributions of simulated first differences we report in the ridgeline plots in Figure 3 lend support for our argument. We ran 3,000 combined simulations of the effect of our white social prejudice measure on support for strongman rule, rule of the U.S. government by the army, and opposition to democracy. None of those 3,000 simulations yielded a negative first difference in which a socially prejudiced white male was more likely to express an anti-democratic orientation than a white male who did not express prejudice against immigrants/foreign workers, members of a different race, or those who speak a different language. The effect of not wanting an immigrant or foreign worker as a neighbor was comparably as robust. Only three of the 3,000 total simulations we ran yielded first differences inconsistent with our expectation that white Americans who view immigrants/foreign workers as unwelcome are more likely to express anti-democratic orientations because democracy means empowering these groups with the opportunity of access.

The first differences are worth belaboring further. Recall that the first difference in one of our simulations is the distance between the predicted probability of an anti-democratic orientation for a socially prejudiced average white male compared to an average white male who does not exhibit social prejudice. Positive first differences are consistent with the hypothesis while a negative first difference is a simulation result in which a socially prejudiced white male was less likely to exhibit an anti-democratic orientation than a white male who was not socially prejudiced. Thus, the mean of first differences can be interpreted as an average effect size across all simulations. For example, the effect of white social prejudice is an average increase of .091 in the probability of supporting the rule of government by the army. The average effect of social prejudice, measured through our composite measure or through any constituent item, increases the probability of favoring strongman rule for the United States anywhere from .102 (when prejudice is measured through opposing a neighbor who spoke a different language) to .126 (when prejudice is measured through opposing a member from a different race). Generally, the smallest effects are observed in the opposition to democracy analysis while the strongest effects are observed on support for strongman rule. All told, however, the analyses we report in Figure 2 and Figure 3 lend strong support to our argument that the effect of white social prejudice on anti-democratic orientations is positive and precise.

CONCLUSION AND IMPLICATIONS

Our paper starts by noting the irony of the democratic durability of the United States. The United States is the longest-running continuous democracy in the world, but many freedoms associated with democracy have only gradually been extended to the full American public. The shortcomings of American democracy are particularly pronounced for non-white citizens and American democracy has always had an uneven balance that clusters on race. There has been progress; racial resentment and negative racial stereotypes are decreasing and the country elected the first black president in 2008. However, there has been an observable backlash as well. Attitudes about system outputs in American democracy are increasingly filtered through racial and ethnocentric dispositions for white Americans and right-wing extremists are committing more hate crimes against minorities than they were in recent years.

The prejudice we have seen percolate since 2008 does not manifest in a vacuum. We generalize what we are observing about American politics and society into an analysis of the anti-democratic orientations of white Americans across four waves of WVS data from 1995 to 2011. We construct an argument linking social prejudice to anti-democratic orientations among white Americans, articulating how social prejudice becomes a problem for diffuse support for democracy since democracy requires extending the opportunity of access to politics to these same outgroups that aggrieved white Americans perceive as threatening them. Our analysis of four waves of WVS data finds support for our argument. White Americans who would not want an immigrant/foreign worker, someone who speaks a different language, or someone from a different race as a neighbor are more likely to support strongman rule in the United States, rule of the U.S. government by the army, and are more likely to outright reject having a democracy for the United States. These findings are robust across multiple model specifications we analyze and report in the appendix as well.

Bridging scholarship on social prejudice and democratic attitudes was our primary academic goal in this paper. We know there is a substantial overlap between both topics. Indeed, democracy institutionalizes mutual tolerance for competing viewpoints. Scholarship that explores how citizens privilege democracy for democracy's “intrinsic” value emphasizes how ingrained these views of equality and tolerance are in democracies and among citizens who value living under democratic institutions. However, we know the American context is one in which its citizens have not exhibited, past or present, the sort of lofty commitment to these values. Our paper contributes to these academic discussions by building on the past and current events and proposes an argument of how social prejudice among white Americans feeds anti-democratic orientations because of democracy's institutionalization of tolerance and the extension of the opportunity of access to these unwelcome ethnic/racial outgroups. Our findings on this end are robust. White Americans who do not welcome the presence of immigrants/foreign workers, people who speak a different language, or people from a different race, are much more likely to value an institutional alternative to democracy in the United States.

However, our immediate audience is those concerned citizens and academics who want to understand and better contextualize our current discussions of democracy's trajectory in the age of Trump. These are important concerns the academic community have been raising since Trump started his presidential campaign in the summer of 2015. We want to emphasize that our analyses concern data collected before Trump's campaign started and thus what we propose here does not speak to how Trump himself affects these trends. That said, we think the most troubling and most important implication of our analysis is the fact these are trends we do observe before Trump. We observe these trends as early as 1995 in the WVS data, a full 20 years before Trump first descended his gilded escalator and began his campaign for president by declaring that Mexicans were “rapists.” Thus, our analysis is as unique as it is troublesome. We show that the new scholarly concern with American democracy's trajectory in the age of Trump belies that these trends have been hiding in plain sight in the WVS data for over 20 years. The normative implications of our analyses are quite clear. Our scholarly interest in democracy's development and growth elsewhere in the world may have glossed over democracy's gangrene in the United States. Our analyses suggest a non-trivial proportion of Americans harbor prejudiced views that spill into support for democracy. Demographic changes in the United States may stress the very foundation of American democracy. Multiculturalism and respect for minority participation in politics are bedrock features of how democracy should operate in theory and in practice. The trends we show emerging as early as the mid-1990s, compounded with current trends, paint a pessimistic picture for the development of American democracy.