Ethnic antagonism erodes Republicans’ commitment to democracy Ethnic antagonism erodes Republicans’ commitment to democracy Larry M. Bartelsa,1 aDepartment of Political Science, Vanderbilt University, Nashville, TN 37203-5721 Contributed by Larry M. Bartels, July 10, 2020 (sent for review April 29, 2020; reviewed by Donald R. Kinder and Paul Sniderman) Most Republicans in a January 2020 survey agreed that “the tradi- tional American way of life is disappearing so fast that we may have to use force to save it.” More than 40% agreed that “a time will come when patriotic Americans have to take the law into their own hands.” (In both cases, most of the rest said they were unsure; only one in four or five disagreed.) I use 127 survey items to measure six potential bases of these and other antidemocratic sentiments: par- tisan affect, enthusiasm for President Trump, political cynicism, eco- nomic conservatism, cultural conservatism, and ethnic antagonism. The strongest predictor by far, for the Republican rank-and-file as a whole and for a variety of subgroups defined by education, locale, sex, and political attitudes, is ethnic antagonism—especially concerns about the political power and claims on government resources of immigrants, African-Americans, and Latinos. The corrosive impact of ethnic antagonism on Republicans’ commitment to democracy under- lines the significance of ethnic conflict in contemporary US politics. ethnic antagonism | political polarization | support for democracy Political developments in the United States and around the worldhave drawn attention to the question of “how democracies die” (1). While the role of ordinary citizens in democratic backsliding is by no means settled (2, 3), concerns about “democratic deconsolidation” and “democratic erosion” have prompted renewed attention to public attitudes regarding democracy and democratic norms (4–7). The frailty of public commitment to democratic norms in the contemporary United States is illustrated by the responses of 1,151 Republican identifiers and Republican-leaning Independents* inter- viewed in January 2020 to survey items contemplating transgressions of a variety of essential democratic principles, including the rejection of violence in pursuit of political ends and respect for the rule of law and the outcomes of elections†. A majority of respondents (50.7%) agreed that “The traditional American way of life is disappearing so fast that we may have to use force to save it.” A substantial plurality (41.3%) agreed that “A time will come when patriotic Americans have to take the law into their own hands.” A near-majority (47.3%) agreed that “Strong leaders sometimes have to bend the rules in order to get things done.” Almost three-fourths (73.9%) agreed that “It is hard to trust the results of elections when so many people will vote for anyone who offers a handout.” In each case, most of those who did not agree said they were unsure; only 1 in 4 or 5 or 10 said they disagreed. These responses are detailed in Table 1.‡ Why do so many people endorse these undemocratic proposi- tions? Political scientists have mostly conceptualized democratic norms as “consensual” and interpreted lack of commitment to them as a product of insufficient social learning: “Those who are actively interested in political events should encounter little diffi- culty in understanding the principles on which the system oper- ates. Others, less discerning in their powers of observation, more circumscribed in their social roles and experiences, or perhaps more parochial in their perspectives, will be less likely to learn the norms” (ref. 9, p. 403). However, the antidemocratic sentiments reported in Table 1 are not primarily products of social isolation or insufficient education or political interest. Rather, they are grounded in real political values—specifically, and overwhelmingly, in Republicans’ ethnocentric concerns about the political and social role of immigrants, African-Americans, and Latinos in a context of significant demographic and cultural change. Measuring Antidemocratic Sentiment Social scientists analyzing survey data from the 1950s documented the shaky allegiance of ordinary Americans to supposedly consen- sual democratic values. Prothro and Grigg found overwhelming majorities endorsing “the basic principles of democracy when they are put in abstract terms” but observed that “consensus breaks down completely” when “broad principles are translated into more spe- cific propositions” regarding, for example, the rights of communists, atheists, or Negroes (ref. 10, pp. 284–286). Less-educated people were especially willing to abandon democratic principles in specific cases; on average, 58% of those with low education but only 36% of those with high education expressed antidemocratic views. Significance Growing partisan polarization and democratic “backsliding” in vari- ous parts of the world have raised concerns about the attachment of ordinary Americans to democratic institutions and procedures. I find that substantial numbers of Republicans endorse statements con- templating violations of key democratic norms, including respect for the law and for the outcomes of elections and eschewing the use of force in pursuit of political ends. The strongest predictor by far of these antidemocratic attitudes is ethnic antagonism—especially concerns about the political power and claims on government re- sources of immigrants, African-Americans, and Latinos. The strong tendency of ethnocentric Republicans to countenance violence and lawlessness, even prospectively and hypothetically, underlines the significance of ethnic conflict in contemporary US politics. Author contributions: L.M.B. designed research, performed research, analyzed data, and wrote the paper. Reviewers: D.R.K., University of Michigan; and P.S., Stanford University. The author declares no competing interest. This open access article is distributed under Creative Commons Attribution-NonCommercial- NoDerivatives License 4.0 (CC BY-NC-ND). 1Email: larry.bartels@vanderbilt.edu. This article contains supporting information online at https://www.pnas.org/lookup/suppl/ doi:10.1073/pnas.2007747117/-/DCSupplemental. First published August 31, 2020. *Independents who report feeling closer to one party than the other generally think and behave much like partisans (8). Thus, Republican-leaning Independents are included along with Republicans in all analyses reported here. †The survey was conducted by the online survey firm YouGov. YouGov maintains a large opt-in panel of respondents, using a combination of matching and weighting to produce nationally representative samples. Vanderbilt University’s Institutional Review Board de- termined that the study posed minimal risk to participants and was exempt from the requirement for informed consent (IRB 192011). All analyses reported here are based on weighted data employing weights constructed by the YouGov staff. Further weighting the data to reproduce the distribution of 2016 presidential votes alters average levels of antidemocratic sentiment and the magnitudes of the key parameter estimates relating ethnic antagonism to antidemocratic attitudes by less than 1%. ‡Levels of agreement with the statements in Table 1 are inflated by the general tendency of survey respondents to agree rather than disagree with statements regardless of their content. However, the structure of the survey items arguably reflects the reality of dem- ocratic backsliding, which invites ordinary people to acquiesce in norm violations couched in terms of patriotism, tradition, and social order. My statistical analysis takes explicit account of the impact of acquiescence on response to these items. 22752–22759 | PNAS | September 15, 2020 | vol. 117 | no. 37 www.pnas.org/cgi/doi/10.1073/pnas.2007747117 D o w n lo a d e d a t C a rn e g ie M e llo n U n iv e rs ity o n A p ri l 5 , 2 0 2 1 http://crossmark.crossref.org/dialog/?doi=10.1073/pnas.2007747117&domain=pdf https://creativecommons.org/licenses/by-nc-nd/4.0/ https://creativecommons.org/licenses/by-nc-nd/4.0/ mailto:larry.bartels@vanderbilt.edu https://www.pnas.org/lookup/suppl/doi:10.1073/pnas.2007747117/-/DCSupplemental https://www.pnas.org/lookup/suppl/doi:10.1073/pnas.2007747117/-/DCSupplemental https://www.pnas.org/cgi/doi/10.1073/pnas.2007747117 McClosky found even sharper differences in support for the “rules of the game” between the general public and a sample of “political influentials”—national convention delegates and alter- nates. (An apparently unintentional echo of this finding appears in a recent study comparing support for a variety of democratic norms between the general public and a sample of “expert” political sci- entists; ref. 6.) He concluded that “it is the articulate classes rather than the public who serve as the major repositories of the public conscience and as the carriers of the Creed” (ref. 11, p. 374). These findings helped to spawn a substantial literature focusing on polit- ical tolerance of disliked groups (12); for the most part, however, the broader issue of public support for McClosky’s “rules of the game” was reduced to a fixation on support for democracy in the abstract, neglecting the crucial distinction between “broad princi- ples” and “specific propositions.” More recently, the limitations of Americans’ commitment to specific democratic norms has begun to come back into focus. For example, a series of surveys conducted between 2010 and 2017 employing items originally designed for use in Latin America found 23–36% of US respondents agreeing that a military coup would be justified “when there is a lot of crime” or “when there is a lot of corruption.” Researchers concerned that the respondents might be misunderstanding the questions tested a variety of alternative question wordings—for example, specifying that the military would “take power over the U.S. government by removing the president by force.” However, they concluded that “respondents’ opinions are not conditional on clarifying the target and nature of the coup” (13). In another series of surveys conducted in 2017–2018 as part of the Bright Line Watch project, only 73% of Americans said it was “important” or “essential” that “government does not interfere with journalists or news organizations,” and only 68% said it was “im- portant” or “essential” that “government effectively prevents private actors from engaging in politically-motivated violence or intimida- tion” (ref. 6, pp. 703–704). Unlike surveys in which respondents are asked to assess the “importance” of democratic norms in the abstract, the survey items presented in Table 1 mimic real politics in pitting demo- cratic values against other cherished values such as patriotism, strong leadership, and the “traditional American way of life.” Faced with such trade-offs, relatively few Republicans—1 in 4, or 5, or 10, depending on the item—decline the invitation to “bend the rules” or “take the law into their own hands.”§ Republicans are not alone in their uncertain attachment to democratic values. For example, Kalmoe and Mason found that majorities of Democrats and Republicans alike viewed the other party as “a serious threat to the United States and its people,” while Democrats were slightly more likely to say that “violence would be justified” if the opposing party won the 2020 presi- dential election (ref. 14, pp. 18, 19, 23). In the Bright Line Watch surveys, differences in the perceived importance of a variety of democratic norms between supporters and opponents of Presi- dent Trump were “generally small,” except on specific issues of salient partisan controversy such as foreign influence in elections and partisan gerrymandering (ref. 6, pp. 705, 706). The willingness of both Democrats and Republicans to sanction abuses of democratic procedure is especially clear when survey items refer explicitly to the president. For example, in 2017, with a Republican in the White House, Republicans were twice as likely as Democrats (24–11%) to agree that “when the country is facing very difficult times it is justifiable for the president of the country to close the Congress and govern without Congress.” However, in 2014, with a Democratic president and a Republican majority in Congress, that partisan difference was reversed—30% of Demo- crats but only 6% of Republicans were willing to countenance the president closing Congress (15). While antidemocratic sentiments clearly exist in both major political parties, the nature and bases of those sentiments are likely to differ significantly between Republicans and Democrats. The specific survey items in Table 1, with their references to “the traditional American way of life” and politicians offering hand- outs, tap frustrations that are more common among Republicans than among Democrats. Moreover, the attitudes that turn out to be highly predictive of agreement with these items among Re- publicans are quite rare among Democrats.{ Thus, an examina- tion of Democrats’ allegiance to democratic values would require somewhat different measures and very different expla- nations from those offered here. In any case, the dramatic shifts in public support for closing Congress underline the extent to which the expression of anti- democratic sentiments is likely to depend on specific political con- texts. Perhaps the most important contextual factor facilitating the translation of antidemocratic sentiments into consequential political behavior is political leadership. Because antidemocratic tendencies loom larger in the leadership of the contemporary Republican Party—and especially in the rhetoric of President Trump—than among Democrats, I focus on the willingness of Republicans to countenance violations of democratic norms. Table 1. Republicans’ antidemocratic attitudes (YouGov survey) January 2020 Strongly agree, % Agree, % Neither; unsure, % Disagree, % Strongly disagree, % The traditional American way of life is disappearing so fast that we may have to use force to save it. 24.0 26.7 27.7 15.2 6.5 A time will come when patriotic Americans have to take the law into their own hands. 15.0 26.3 36.3 14.0 8.4 Strong leaders sometimes have to bend the rules in order to get things done. 12.5 34.8 29.2 17.4 6.1 It is hard to trust the results of elections when so many people will vote for anyone who offers a handout. 34.1 39.8 16.2 7.0 2.9 The survey included Republicans and Republican-leaning Independents (n = 1,151). §Graham and Svolik used survey experiments involving fictitious candidates to assess “the commitment to democratic principles among the American public.” They found that “only a small fraction of Americans prioritize democratic principles in their electoral choices when doing so goes against their partisan identification or favorite policies,” suggesting that “conventional measures of support for democracy have a fundamental blind spot: they fail to capture voters’ willingness to act on their commitment to democ- racy precisely when democracy is at stake.” When they limited their analysis to a subset of more realistic candidate pairings, the estimated electoral penalties for norm violations were even smaller—just 3.5 percentage points (ref. 7, pp. 394, 406, 408, 407). {The key explanatory variable in my analysis, ethnic antagonism, has a mean value of zero and a standard deviation of 0.96 among Republicans and Republican-leaning Indepen- dents, and a mean value of −2.21 and a standard deviation of 1.02 among Democrats and Democratic-leaning Independents. Only 2% of Democrats would be in the top half of the Republican distribution, while 87% would be in the bottom decile. Bartels PNAS | September 15, 2020 | vol. 117 | no. 37 | 22753 P O LI TI C A L SC IE N C ES D o w n lo a d e d a t C a rn e g ie M e llo n U n iv e rs ity o n A p ri l 5 , 2 0 2 1 Bases of Republicans’ Antidemocratic Attitudes Fig. 1 shows how Republicans’ antidemocratic responses in the January 2020 survey were related to education, political interest, and locale. These relationships provide only modest support for the hypothesis that allegiance to democratic values is a product of “political activity, involvement and articulateness,” as McClosky had it (ref. 11, p. 374). Although people with postgraduate edu- cation were clearly less likely than those with less education to endorse violations of democratic norms, the overall relationship between education and antidemocratic sentiments is rather weak. Similarly, people in big cities were only about 5% less likely than those in rural areas to endorse norm violations, while people who said they followed politics “most of the time” were about 7% more likely to do so than those who said they followed politics “hardly at all.” Given the distributions of these social characteristics in the Republican sample, the most typical antidemocrats were not “men and women whose lives are circumscribed by apathy, ignorance, provincialism and social or physical distance from the centers of intellectual activity” (ref. 11, p. 375), but suburbanites with some college education and a healthy (or unhealthy) interest in politics. A more promising place to look for the antecedents of anti- democratic sentiments is in the substance of political attitudes (16). Here, I distinguish six latent dimensions of Republicans’ political attitudes: 1) partisan affect, 2) affect toward President Trump, 3) economic conservatism, 4) cultural conservatism, 5) ethnic antagonism, and 6) political cynicism. These six latent dimensions are tapped by 127 distinct items in the January 2020 survey, with an average of 29 indicators per latent dimension. The most important indicators of each dimensions are summa- rized in Table 2, and the complete measurement model and question wording are reported in SI Appendix, Table S1. The most influential operationalization of ethnocentrism in recent political research (17) focuses on stereotypical negative views regarding the personal qualities of people in specific racial and ethnic groups. My measure of “ethnic antagonism” is broader in scope, incorporating not only unfavorable feelings toward Muslims, immigrants, and other out-groups, but also— and especially—concerns about these groups’ political and social claims. Survey items tapping perceptions that immigrants, African-Americans, Latinos, and poor people have more than their fair share of political power and get more than their fair share of government resources are powerful indicators of ethnic antagonism (with item-scale correlations ranging from 0.671 to 0.788).# So are concerns about African-Americans “using racism as an excuse” (R = 0.750), concerns about discrimination against Whites (0.736), and perceptions that “immigrants contribute a great deal to American society and culture” (−0.727) and that “speaking English is essential for being a true American” (0.681). The resulting scale is strongly correlated with the more familiar “racial resentment” scale (18–20), but better reflects the multi- faceted demographic threat currently perceived by many White 0 10 20 30 40 50 60 70 80 90 100 A ve ra ge a gr ee m en t ( % ) Hard to trust elections Use force to save way of life Strong leaders bend rules Take law into own hands Education Political interest Locale Fig. 1. Social bases of Republicans’ antidemocratic attitudes. #Of course, immigrants, poor people, and welfare recipients are not ethnic groups. How- ever, Republicans’ attitudes toward these groups are correlated with their attitudes to- ward Latinos and African-Americans, suggesting that, for many, views of immigrants, the poor, and welfare recipients are colored by ethnic antagonism. The measurement model in SI Appendix, Table S1 treats perceptions of these groups as potential reflections of multiple latent attitudes. The results indicate that concerns about the political power and access to government resources of poor people are almost equally shaped by ethnic antagonism and economic conservatism, while cultural conservatism tends to mitigate those concerns. In some cases, relationships allowed by the measurement model turn out to be absent in the data. For example, thinking that “people on welfare often have it better than those who work for a living” is strongly affected by ethnic antagonism but unaffected by cultural conservatism. 22754 | www.pnas.org/cgi/doi/10.1073/pnas.2007747117 Bartels D o w n lo a d e d a t C a rn e g ie M e llo n U n iv e rs ity o n A p ri l 5 , 2 0 2 1 https://www.pnas.org/lookup/suppl/doi:10.1073/pnas.2007747117/-/DCSupplemental https://www.pnas.org/lookup/suppl/doi:10.1073/pnas.2007747117/-/DCSupplemental https://www.pnas.org/cgi/doi/10.1073/pnas.2007747117 Americans and their specifically political antagonisms toward a variety of salient out-groups.‖ “Republican affect” is measured by attitudes toward Repub- lican and (especially) Democratic political figures (Obama, Pence, Pelosi), ordinary partisans, and the parties themselves. “Trump affect” is captured by feelings toward President Trump and ratings on a variety of specific traits (“inspiring,” “honest,” “knowledgeable,” “moral,” and the like). “Economic conserva- tism” entails belief in economic opportunity, a preference for limited government (especially in the provision of social welfare and environmental regulation), and sympathy for rich people and businesses vis-à-vis poor people and labor unions. “Cultural conservatism” includes views about patriotism, traditional mo- rality (including abortion and sexual harassment), and disdain for big cities, rich people, journalists, and college professors. “Political cynicism” reflects beliefs that “people like me have no say” and that government is too powerful, ineffective, and cannot be trusted, as well as negative views about “politicians.” Not surprisingly, some of these attitudes are strongly correlated, even within the restricted sample of Republicans and Republican- leaning Independents. The largest correlations are between Re- publican affect and cultural conservatism (0.759), cultural conservatism and ethnic antagonism (0.722), Republican affect and Trump affect (0.699), economic conservatism and cultural conservatism (0.679), and cultural conservatism and political cyni- cism (0.671). (The complete correlation matrix appears at the end of SI Appendix, Table S1.) In light of these correlations, analyses focusing on some of these attitudes while ignoring others are likely to misconstrue the specific bases of antidemocratic sentiments. Here, I guard against that sort of error by relating antidemocratic survey responses simultaneously to all six sets of attitudes. In addition to these substantive differences in political atti- tudes, I take account of acquiescence—individual differences in respondents’ tendencies to agree or disagree with survey items regardless of substantive content. Because all of the items in Table 1 invite respondents to endorse or reject antidemocratic rather than prodemocratic sentiments, controlling for acquies- cence guards against mistaking general agreeableness for will- ingness to violate democratic norms.** Table 3 reports the results of ordered probit regression anal- yses relating each antidemocratic sentiment to this battery of attitudes, each of which is normalized to have a mean of zero and a SD of one within the Republican sample. The apparent importance Table 2. Key indicators of latent dimensions Latent dimension Key indicators Republican affect (+) Feelings toward Pence, Republican Party, Republicans, McConnell; closeness to Republicans; (−) feelings toward Obama, Democrats, Democratic Party, Pelosi, Romney; closeness to Democrats; Democratic trait ratings: tolerant, knowledgeable, patriotic, moral, hard-working, generous. Trump affect (+) Feelings toward Trump; Trump trait ratings: inspiring, honest, knowledgeable, strong leader, intelligent, moral, tolerant, hard- working, patriotic; Trump job approval. Economic conservatism (+) Life is better for most; hard work can still achieve; feelings toward Wall Street bankers, rich people; (−) government should reduce income inequality, provide health care, provide college and childcare, raise taxes on the rich, protect the environment, provide a decent living to those who can’t work; climate change is our greatest threat; Medicare for all; feelings toward poor people, environmentalists, labor unions. Cultural conservatism (+) Respect for the flag; decline of manufacturing due to bad trade deals; unfair government resources to people in big cities, rich people; unfair political power for people in big cities, rich people; feelings toward poor people, NRA; closeness to poor people; (−) closeness to people in big cities, rich people; feelings toward journalists, Black Lives Matter, college professors, Wall Street bankers; sexual harassment; abortion choice. Ethnic antagonism (+) Unfair government resources to immigrants, African-Americans, Latinos, poor people; unfair political power for immigrants, African-Americans, Latinos, poor people; discrimination against whites; welfare recipients better off than people who work; (−) immigrants contribute to society; feelings toward welfare recipients, Muslims, immigrants, poor people; closeness to immigrants, Latinos. Political cynicism (+) People like me have no say; government too powerful; government efforts generally ineffective; (−) government does a good job of providing services; trust in government; feelings toward politicians, Congress. Item wording and complete results for the measurement model appear in SI Appendix, Table S1. kThe correlation between ethnic animosity as measured in 2020 and a three-item racial resentment scale based on responses from the same people in a 2016 survey is 0.754. The racial resentment scale is strongly correlated with specific items in the 2020 survey fo- cusing on African-Americans (R = 0.596 to 0.728), but less strongly correlated with items focusing on immigrants, Latinos, and poor people (R = 0.393 to 0.540). **Acquiescence is estimated from responses to 36 agree/disagree items. Conservative attitudes are tapped by “agree” responses for 16 items and by “disagree” responses for the other 20 items. The correlations between acquiescence and conservative atti- tudes range from −0.210 to 0.243 and average −0.012. The structural equation model used to infer Republicans’ latent attitudes (SI Appendix, Table S1) also includes distinct acquiescence factors for a variety of other survey item formats. Bartels PNAS | September 15, 2020 | vol. 117 | no. 37 | 22755 P O LI TI C A L SC IE N C ES D o w n lo a d e d a t C a rn e g ie M e llo n U n iv e rs ity o n A p ri l 5 , 2 0 2 1 https://www.pnas.org/lookup/suppl/doi:10.1073/pnas.2007747117/-/DCSupplemental https://www.pnas.org/lookup/suppl/doi:10.1073/pnas.2007747117/-/DCSupplemental https://www.pnas.org/lookup/suppl/doi:10.1073/pnas.2007747117/-/DCSupplemental of specific attitudes varies from item to item (across the columns of the table). For example, political cynicism is especially strongly related to distrust of elections, while enthusiasm for President Trump seems to bolster support for strong leaders bending rules and patriots taking the law into their own hands. However, in every case the factor most strongly associated with support for antidemocratic sentiments is ethnic antagonism. These statistical results suggest that ethnic antagonism has a substantial negative effect on Republicans’ commitment to de- mocracy. For example, holding other attitudes constant at av- erage Republican values, the predicted probability of agreeing that “we may have to use force” to save “the traditional Amer- ican way of life” increases from 0.226 at the 5th percentile of Republican ethnic antagonism (−1.43) to 0.813 at the 95th percentile (1.58). The corresponding increase in the predicted probability of agreeing that “patriotic Americans” will have to “take the law into their own hands,” holding other attitudes constant at average Republican values, is from 0.153 at the 5th percentile of Republican ethnic antagonism to 0.718 at the 95th percentile. The estimated impact of ethnic antagonism on the probability of agreeing with each item, based on the statistical analyses reported in Table 3, is shown in Fig. 2, Left. These relationships are not driven solely by extreme high or low values of ethnic antagonism. Fig. 2, Right shows the simple bivariate relationship between ethnic antagonism and the prob- ability of agreeing with each antidemocratic sentiment. (The agreement rates shown in the figure are derived from locally weighted [lowess] regressions employing the nearest 500 re- sponses at each point in the Republican distribution of ethnic antagonism.) The relationships are roughly linear—and similar in magnitude to the relationships implied by the multivariate statistical results—except for a noticeable leveling-off of support among the most resentful Republicans for the notion that “strong leaders sometimes have to bend the rules.”†† Nor is the strong association between ethnic antagonism and antidemocratic attitudes limited to specific segments of the Re- publican rank-and-file. Table 4 presents ordered probit regres- sion parameter estimates for ethnic antagonism paralleling those in Table 3 but for a variety of distinct Republican subgroups— men and women, people with and without college education, those in cities or suburbs and in small towns or rural areas, and those more and less favorable toward President Trump, Fox News, and the National Rifle Association (NRA).‡‡ Some of the parameter estimates are significantly larger for men, people with college education, and (especially) those most favorable toward the NRA. However, in every subgroup ethnic antagonism is strongly related to antidemocratic attitudes, even after statistically controlling for other factors. (The t statistics [based on robust standard errors] for the 48 distinct sub-groups’ ethnic antagonism parameter estimates range from 2.45 to 7.59, averaging 4.94. The complete results appear in SI Appendix, Tables S2–S7.) Discussion The support expressed by many Republicans for violations of a variety of crucial democratic norms is primarily attributable not to partisan affect, enthusiasm for President Trump, political cynicism, economic conservatism, or general cultural conser- vatism, but to what I have termed ethnic antagonism. The single survey item with the highest average correlation with antidemocratic sentiments is not a measure of attitudes to- ward Trump, but an item inviting respondents to agree that “discrimination against whites is as big a problem today as discrimination against blacks and other minorities.” Not far behind are items positing that “things have changed so much that I often feel like a stranger in my own country,” that im- migrants get more than their fair share of government re- sources, that people on welfare often have it better than those who work for a living, that speaking English is “essential for being a true American,” and that African-Americans “need to stop using racism as an excuse.” (The average correlations between these single items and antidemocratic sentiments range from 0.330 to 0.262. The average correlation between Table 3. Political bases of Republicans’ antidemocratic attitudes (ordered probit regression parameter estimates and response thresholds) Use force to save traditional way of life Patriots take law into their own hands Strong leaders sometimes bend rules Hard to trust the results of elections Republican affect 0.129 (0.083) −0.122 (0.087) −0.023 (0.089) 0.116 (0.090) Trump affect 0.042 (0.068) 0.212 (0.065) 0.235 (0.070) −0.097 (0.064) Economic conservatism −0.242 (0.078) −0.036 (0.066) 0.116 (0.073) 0.079 (0.069) Cultural conservatism −0.049 (0.121) 0.008 (0.129) −0.432 (0.121) −0.087 (0.130) Ethnic antagonism 0.547 (0.070) 0.532 (0.069) 0.495 (0.073) 0.440 (0.081) Political cynicism 0.141 (0.070) 0.126 (0.072) 0.042 (0.067) 0.318 (0.071) Acquiescence 0.352 (0.056) 0.237 (0.055) 0.414 (0.058) 0.347 (0.057) Strongly disagree/disagree −1.816 (0.076) −1.623 (0.077) −1.749 (0.072) −2.220 (0.102) Disagree/neither, unsure −0.935 (0.055) −0.888 (0.054) −0.798 (0.052) −1.528 (0.075) Neither, unsure/agree −0.028 (0.050) 0.262 (0.047) 0.080 (0.049) −0.768 (0.055) Agree/strongly agree 0.854 (0.051) 1.223 (0.057) 1.284 (0.053) 0.491 (0.048) Log likelihood −1,549.6 −1,538.3 −1,558.9 −1,348.1 Pseudo R-squared 0.11 0.10 0.07 0.11 n 1,151 1,151 1,151 1,151 Robust SEs are presented in parentheses. ††This leveling-off may reflect the fact that ethnic antagonism is strongly correlated with cultural conservatism (R = 0.722), which significantly depresses support for bending the rules. Adding a squared ethnic antagonism term to the ordered probit regression in the third column of Table 3 produces a parameter estimate of −0.030 (with a standard error of 0.030), barely altering the parameter estimates for ethnic antagonism (0.500) and cultural conservatism (0.419). More generally, the statistical results are surprisingly ro- bust to variations in model specification. For example, analyses paralleling those in Table 3 but with ethnic antagonism as the only predictor of antidemocratic sentiments produce probit parameter estimates that are just 4% smaller on average. Adding the acquiescence factor but no political covariates produces probit parameter estimates for ethnic antagonism that are just 4% larger than those reported in Table 3, on average. ‡‡The sample of Republicans and Republican-leaning Independents is too ethnically ho- mogeneous (78.4% non-Hispanic White) to permit separate analysis of Hispanics (12.6%), African-Americans (1.4%), or others (7.5%). Not surprisingly, the average levels of ethnic antagonism among African-Americans (−0.475) and Hispanics (−0.442) are much lower than among non-Hispanic Whites (0.083), with people in other groups in between (−0.039). 22756 | www.pnas.org/cgi/doi/10.1073/pnas.2007747117 Bartels D o w n lo a d e d a t C a rn e g ie M e llo n U n iv e rs ity o n A p ri l 5 , 2 0 2 1 https://www.pnas.org/lookup/suppl/doi:10.1073/pnas.2007747117/-/DCSupplemental https://www.pnas.org/lookup/suppl/doi:10.1073/pnas.2007747117/-/DCSupplemental https://www.pnas.org/cgi/doi/10.1073/pnas.2007747117 the ethnic antagonism scale and antidemocratic sentiments is 0.382.) The powerful effects of ethnic antagonism on Republicans’ antidemocratic attitudes underscore the extent to which this par- ticular threat to democratic values is concentrated in the con- temporary Republican Party. Seventy-eight percent of Democrats and Democratic-leaning Independents in the 2020 survey had ethnic antagonism scores below the fifth percentile of the Re- publican distribution (−1.43), while 98% had scores below the Republican average. The average score among Democrats (−2.21) barely appears in Fig. 2, which excludes the bottom (and top) 1% of the Republican distribution of ethnic antagonism. In this re- spect, among others, the attitudes of Republicans and Democrats are sharply polarized.§§ The strong association reported here between Republicans’ antidemocratic attitudes and ethnic antagonism reflects a spe- cific social and political context. It certainly does not imply that ethnic antagonism is a necessary basis for antidemocratic senti- ment, or that ethnic antagonism always and everywhere erodes public commitment to democracy. One of the most politically salient features of the contemporary United States is the loom- ing demographic transition from a majority-White to a “majority- minority” country. Several years ago, reminding White Americans of that prospect significantly altered their political attitudes (21). Now, President Trump and Fox News remind them, implicitly or explicitly, on an almost-daily basis.{{ For those who view demo- graphic change as a significant threat to “the traditional American way of life,” the political stakes could hardly be higher. This perspective is forcefully illustrated by an attention-getting essay published pseudonymously during the 2016 campaign on “the Flight 93 election: charge the cockpit or you die” (22). The author, who went on to serve on President Trump’s National Security Council staff, wrote that “a Hillary Clinton presidency is Russian Roulette with a semi-auto. With Trump, at least you can spin the cylinder and take your chance.” He denounced the “tsunami of leftism that still engulfs our every—literal and figurative—shore,” and warned that “the ceaseless importation of Third World for- eigners with no tradition of, taste for, or experience in liberty means that the electorate grows more left, more Democratic, less Republican, less republican, and less traditionally American with every cycle.” “Trump, he concluded, “alone among candidates for high office in this or in the last seven (at least) cycles, has stood up to say: I want to live. I want my party to live. I want my country to live. I want my people to live. I want to end the insanity.” The political impact of this specific instance of ethnocentric alarmism may have been modest, but it is representative of a 0 10 20 30 40 50 60 70 80 90 100 -2.5 -2 -1.5 -1 -0.5 0 0.5 1 1.5 2 2.5 A ve ra ge a gr ee m en t ( % ) Ethnic Antagonism (normalized Republican distribution) Hard to trust elections Use force to save way of life Strong leaders bend rules Take law into own hands 0 10 20 30 40 50 60 70 80 90 100 -2.5 -2 -1.5 -1 -0.5 0 0.5 1 1.5 2 2.5 P re di ct ed a gr ee m en t ( % ) Ethnic Antagonism (normalized Republican distribution) Hard to trust elections Use force to save way of life Strong leaders bend rules Take law into own hands Fig. 2. Ethnic antagonism and antidemocratic attitudes. Predicted agreement based on statistical analyses in Table 3 (Left) and average agreement based on bivariate lowess regressions (Right). §§While the latent attitudes summarized in Table 2 and detailed in SI Appendix, Table S1 are derived from an analysis limited to Republicans and Republican-leaning Indepen- dents, factor scores can be computed from the same indicators for Democrats and In- dependents as well. The only substantial overlap in attitudes between Republicans and Democrats is on political cynicism. The average factor scores for Democrats and Dem- ocratic-leaning Independents on the other five dimensions are two to three standard deviations below the corresponding Republican average scores. {{Attitudes toward President Trump and Fox News are more strongly correlated with ethnic antagonism (R = 0.450, 0.217) than with antidemocratic attitudes (correlations ranged from 0.279 to 0.338 and from 0.093 to 0.175, respectively), and the statistical relationships between ethnic antagonism and antidemocratic attitudes in Table 4 are, on average, similar in magnitude regardless of favorability toward President Trump and toward Fox News. Thus, there is no evidence here suggesting that President Trump or Fox News contributes to Republicans’ translation of ethnic antagonism into antidemo- cratic attitudes. However, detailed data on exposure to specific sources and messages would be necessary to shed light on the impact of political rhetoric on ethnic antago- nism, antidemocratic attitudes, and their association. Bartels PNAS | September 15, 2020 | vol. 117 | no. 37 | 22757 P O LI TI C A L SC IE N C ES D o w n lo a d e d a t C a rn e g ie M e llo n U n iv e rs ity o n A p ri l 5 , 2 0 2 1 https://www.pnas.org/lookup/suppl/doi:10.1073/pnas.2007747117/-/DCSupplemental broadly held and consequential worldview. Analysts of the 2016 presidential election have emphasized the activation of long- standing racial resentment and concerns about immigration as important factors contributing to President Trump’s support (23). The same factors have helped to fuel political polarization more broadly (24). The relationship reported here between ethnic antagonism and expressions of support for violations of key democratic norms suggests that the effects of millions of White Americans’ concerns regarding the prospect of demo- graphic, social, and political change may not be limited to the electoral sphere. Many people who endorse resorting to force or taking the law into their own hands in the context of an opinion survey are unlikely to engage in actual violence or lawlessness. However, the United States has experienced a cataclysmic civil war and a long history of racial and ethnic violence (25), and currently experiences thousands of hate crimes per year##; thus, it is not fanciful to suppose that expressive support for bending the rules or resorting to force to protect one’s “way of life” is consequential for actual behavior—or that it could become even more conse- quential under inflammatory circumstances. It is also possible that antidemocratic attitudes among citi- zens encourage political elites to engage in antidemocratic be- haviors. Systematic attempts to measure democratic performance, while fraught with difficulty, do provide some grounds for concern on this score. For example, Freedom House has reported a sig- nificant decline over the past decade in its rating of the quality of American democracy, from 94 on a 100-point scale in 2009 to 86 in 2018 (27). While acknowledging that “political polarization, declining economic mobility, the outsized influence of special in- terests, and the diminished influence of fact-based reporting in favor of bellicose partisan media were all problems afflicting the health of American democracy well before 2017,” the authors of the report argued that "President Trump exerts an influence on American politics that is straining our core values and testing the stability of our constitutional system. No president in living mem- ory has shown less respect for its tenets, norms, and principles. Trump has assailed essential institutions and traditions including the separation of powers, a free press, an independent judiciary, the impartial delivery of justice, safeguards against corruption, and most disturbingly, the legitimacy of elections." How concerned should we be that a president who assails “essential institutions and traditions” of democracy has found millions of followers willing to endorse significant violations of democratic norms, including resort to force in pursuit of political ends, lawlessness by “patriotic Americans,” and casting doubt on the legitimacy of elections? The simple answer is that no one knows. Antidemocratic forces in well-functioning multiparty systems tend to be isolated in minor parties, shunned as coalition partners and cordoned off from power. In contrast, as the Republican Party establishment learned in 2016, US parties are vulnerable to hostile takeover by political entrepreneurs capable of mobilizing passionate factions. Moreover, the evolution of the Republican Party over the past few years suggests that a hostile takeover may not stay hostile for long, as rank-and-file supporters respond to new leadership and elected officials adapt themselves to new political realities (28). Frances Lee has enumerated a variety of formidable institutional barriers to the consolidation of “authoritarian power over American national government.” Nonetheless, she concluded that “A racial- ized party system in an electorate with a questionable commitment to liberal values is a troubling development. It is difficult to manage racial tensions in a democracy in any case, much less when race becomes a principal line of political cleavage” (29). The findings reported here underline the extent to which race—and, more broadly, ethnic conflict—has indeed become a principal line of political cleavage, not only in American electoral politics but also in America’s broader, ongoing struggle to embrace and instantiate democracy. Data Availability. Survey data, log, and questionnaire files are in the Inter-university Consortium for Political and Social Research openICPSR repository (https://www.openicpsr.org/openicpsr/project/ 120104/version/V1/view). Table 4. Translation of ethnic antagonism into antidemocratic attitudes in Republican subgroups (ordered probit regression parameter estimates for ethnic antagonism by subgroups) Use force to save traditional way of life Patriots take law into their own hands Strong leaders sometimes bend rules Hard to trust the results of elections Some college 0.691 (0.091) 0.568 (0.089) 0.478 (0.100) 0.597 (0.108) No college 0.386 (0.113) 0.480 (0.108) 0.529 (0.110) 0.295 (0.109) (Difference) 0.306 (0.144) 0.087 (0.139) −0.052 (0.147) 0.302 (0.153) Male 0.489 (0.089) 0.632 (0.095) 0.675 (0.112) 0.506 (0.114) Female 0.643 (0.113) 0.420 (0.101) 0.350 (0.100) 0.388 (0.106) (Difference) −0.154 (0.143) 0.212 (0.138) 0.325 (0.150) 0.118 (0.155) City/suburb 0.600 (0.088) 0.561 (0.095) 0.517 (0.098) 0.438 (0.110) Small town/rural 0.466 (0.120) 0.471 (0.098) 0.462 (0.110) 0.414 (0.113) (Difference) 0.134 (0.148) 0.091 (0.136) 0.056 (0.146) 0.024 (0.157) High Trump favorability 0.521 (0.093) 0.639 (0.101) 0.519 (0.100) 0.413 (0.109) Lower Trump favorability 0.592 (0.106) 0.425 (0.096) 0.475 (0.103) 0.485 (0.119) (Difference) −0.071 (0.141) 0.214 (0.138) 0.045 (0.143) −0.073 (0.161) High Fox News favorability 0.467 (0.081) 0.469 (0.085) 0.612 (0.092) 0.449 (0.101) Lower Fox News favorability 0.668 (0.121) 0.589 (0.108) 0.346 (0.115) 0.436 (0.129) (Difference) −0.201 (0.145) −0.120 (0.137) 0.266 (0.147) 0.013 (0.164) High NRA favorability 0.586 (0.096) 0.684 (0.096) 0.635 (0.099) 0.590 (0.107) Lower NRA favorability 0.529 (0.105) 0.397 (0.100) 0.334 (0.099) 0.285 (0.116) (Difference) 0.057 (0.141) 0.288 (0.138) 0.301 (0.139) 0.306 (0.157) Robust SEs are presented in parentheses. Complete results are presented in SI Appendix, Tables S2–S7. ##The FBI Uniform Crime Report for 2018 recorded 5,566 hate crimes against persons (murder, assault, or intimidation), of which 3,445 (62%) were motivated by racial or ethnic bias; 671 (3.4 per million population) were against Whites and 2,774 (21.4 per million) against non-Whites. These figures are substantial underestimates of the actual incidence of hate crimes, since “many cities and some entire states failed to collect or report the data,” and “experts say that more than half of all victims of hate crimes never file a complaint with the authorities in the first place” (26). 22758 | www.pnas.org/cgi/doi/10.1073/pnas.2007747117 Bartels D o w n lo a d e d a t C a rn e g ie M e llo n U n iv e rs ity o n A p ri l 5 , 2 0 2 1 https://www.openicpsr.org/openicpsr/project/120104/version/V1/view https://www.openicpsr.org/openicpsr/project/120104/version/V1/view https://www.pnas.org/lookup/suppl/doi:10.1073/pnas.2007747117/-/DCSupplemental https://www.pnas.org/lookup/suppl/doi:10.1073/pnas.2007747117/-/DCSupplemental https://www.pnas.org/cgi/doi/10.1073/pnas.2007747117 ACKNOWLEDGMENTS. The research reported here was supported by Vanderbilt University’s May Werthan Shayne Chair. Christopher Achen, Katherine Cramer, Marc Hetherington, Cindy Kam, Donald Kinder, Frances Lee, David E. Lewis, Tali Mendelberg, Jeff Mondak, Diana Mutz, Wendy Rahn, Steven Rogers, Michael Shepherd, John Sides, Paul Sniderman, John Zaller, and participants in the Vanderbilt Research on Individuals, Politics & Society (RIPS) laboratory provided invaluable advice and criticism. 1. S. Levitsky, D. Ziblatt, How Democracies Die, (Crown, 2018). 2. N. Bermeo, On democratic backsliding. J. Democracy 27, 5–19 (2016).1045-5736 3. A. Przeworski, Crises of Democracy, (Cambridge University Press, 2019). 4. R. S. Foa, Y. Mounk, The democratic disconnect. J. Democracy 27, 5–17 (2016). 5. L. Drutman, L. Diamond, J. Goldman, “Follow the leader: Exploring American support for democracy and authoritarianism” (Democracy Fund Voter Study Group Rep., 2018). https://www.voterstudygroup.org/publication/follow-the-leader. 6. J. M. Carey, G. Helmke, B. Nyhan, M. Sanders, S. Stokes, Searching for bright lines in the Trump presidency. Perspect. Polit. 17, 699–718 (2019). 7. M. H. Graham, M. W. Svolik, Democracy in America? Partisanship, polarization, and the robustness of support for democracy in the United States. Am. Polit. Sci. Rev. 114, 392–409 (2020). 8. B. E. Keith et al., The Myth of the Independent Voter, (University of California Press, 1992). 9. D. Chong, H. McClosky, J. Zaller, Patterns of support for democratic and capitalist values in the United States. Br. J. Polit. Sci. 13, 401–440 (1983). 10. J. W. Prothro, C. M. Grigg, Fundamental principles of democracy: Bases of agreement and disagreement. J. Polit. 22, 276–294 (1960). 11. H. McClosky, Consensus and ideology in American politics. Am. Polit. Sci. Rev. 58, 361–382 (1964). 12. J. L. Sullivan, J. Piereson, G. E. Marcus, Political Tolerance and American Democracy, (University of Chicago Press, 1993). 13. E. J. Zechmeister, J. Elizabeth, Should we be alarmed that one-in-four U.S. citizens believes a military take-over can be justifiable? Latin American Public Opinion Project (2018). https://www.vanderbilt.edu/lapop/insights/IMN001en.pdf. Accessed 15 August 2020. 14. N. P. Kalmoe, L. Mason, "Lethal mass partisanship: Prevalence, correlates, & electoral contingencies" in National Capital Area Political Science Association American Politics Meeting (2019). https://www.dannyhayes.org/uploads/6/9/8/5/69858539/kalmoe___mason_ ncapsa_2019_-_lethal_partisanship_-_final_lmedit.pdf. Accessed 18 August 2020. 15. G. Feierherd, N. Lupu, S. Stokes, A significant minority of Americans say they could support a military takeover of the U.S. government. The Washington Post, Monkey Cage, 16 February 2018. https://www.washingtonpost.com/news/monkey-cage/wp/ 2018/02/16/a-significant-minority-of-americans-say-they-would-support-a-military- takeover-of-the-u-s-in-the-right-circumstances/. 16. P. M. Sniderman, J. F. Fletcher, P. H. Russell, P. E. Tetlock, B. J. Gaines, The fallacy of democratic elitism: Elite competition and commitment to civil liberties. Br. J. Polit. Sci. 21, 349–370 (1991). 17. D. R. Kinder, C. D. Kam, Us Against Them: Ethnocentric Foundations of American Opinion, (University of Chicago Press, 2010). 18. D. R. Kinder, L. M. Sanders, Divided by Color: Racial Politics and Democratic Ideals, (University of Chicago Press, 1996). 19. S. Feldman, L. Huddy, Racial resentment and white opposition to race-conscious programs: Principles or prejudice? Am. J. Pol. Sci. 49, 168–183 (2005). 20. D. C. Wilson, D. W. Davis, Reexamining racial resentment: Conceptualization and content. Ann. Am. Acad. Pol. Soc. Sci. 63, 117–133 (2011). 21. M. A. Craig, J. A. Richeson, On the precipice of a “majority-minority” America: Per- ceived status threat from the racial demographic shift affects White Americans’ po- litical ideology. Psychol. Sci. 25, 1189–1197 (2014). 22. M. Anton, The Flight 93 election. Claremont Review of Books, 5 September 2016. https://claremontreviewofbooks.com/digital/the-flight-93-election/. Accessed 18 Au- gust 2020. 23. J. Sides, M. Tesler, L. Vavreck, Identity Crisis: The 2016 Presidential Campaign and the Battle for the Meaning of America, (Princeton University Press, 2018). 24. E. Klein, Why We’re Polarized, (Simon & Schuster, 2020). 25. R. Ginzburg, 100 Years of Lynchings, (Black Classic Press, 1988). 26. A. Hassan, Hate-crime violence hits 16-year high, F.B.I. reports, New York Times, 12 November 2019. https://www.nytimes.com/2019/11/12/us/hate-crimes-fbi-report.html. 27. M. Abramowitz, “The struggle comes home: Attacks on democracy in the United States” in Freedom House (Freedom in the World, 2019), pp. 17–24, https://freedomhouse.org/ sites/default/files/Feb2019_FH_FITW_2019_Report_ForWeb-compressed.pdf. 28. L. M. Bartels, Partisanship in the Trump era. J. Polit. 80, 1483–1494 (2018). 29. F. E. Lee, Populism and the American party system: Opportunities and constraints. Perspect. Polit. 18, 370–388 (2020). Bartels PNAS | September 15, 2020 | vol. 117 | no. 37 | 22759 P O LI TI C A L SC IE N C ES D o w n lo a d e d a t C a rn e g ie M e llo n U n iv e rs ity o n A p ri l 5 , 2 0 2 1 https://www.voterstudygroup.org/publication/follow-the-leader https://www.vanderbilt.edu/lapop/insights/IMN001en.pdf https://www.dannyhayes.org/uploads/6/9/8/5/69858539/kalmoe___mason_ncapsa_2019_-_lethal_partisanship_-_final_lmedit.pdf https://www.dannyhayes.org/uploads/6/9/8/5/69858539/kalmoe___mason_ncapsa_2019_-_lethal_partisanship_-_final_lmedit.pdf https://www.washingtonpost.com/news/monkey-cage/wp/2018/02/16/a-significant-minority-of-americans-say-they-would-support-a-military-takeover-of-the-u-s-in-the-right-circumstances/ https://www.washingtonpost.com/news/monkey-cage/wp/2018/02/16/a-significant-minority-of-americans-say-they-would-support-a-military-takeover-of-the-u-s-in-the-right-circumstances/ https://www.washingtonpost.com/news/monkey-cage/wp/2018/02/16/a-significant-minority-of-americans-say-they-would-support-a-military-takeover-of-the-u-s-in-the-right-circumstances/ https://claremontreviewofbooks.com/digital/the-flight-93-election/ https://www.nytimes.com/2019/11/12/us/hate-crimes-fbi-report.html https://freedomhouse.org/sites/default/files/Feb2019_FH_FITW_2019_Report_ForWeb-compressed.pdf https://freedomhouse.org/sites/default/files/Feb2019_FH_FITW_2019_Report_ForWeb-compressed.pdf