“The desire for a strong leader who can identify domestic enemies and who promises to do something about them without worrying over much about legalities—those germs, mutated to fit the particular local subcultures, are latent in every democratic electorate, waiting for sufficiently widespread human suffering to provide conditions for the explosive spread.”
—Achen and Bartels (Reference Achen and Bartels2016, 316)Partisanship in American politics is inextricably linked with social identities. Increasingly, the Democratic and Republican parties have divided along racial, religious, and other identity-based lines (Mason Reference Mason2018). Most of the research into this phenomenon focuses on the role of partisans’ own social identities in shaping their political attitudes and partisan identities (Campbell et al. Reference Campbell, Converse, Miller and Stokes1960; Green, Palmquist, and Schickler Reference Green, Palmquist and Schickler2004; Mason and Wronski Reference Mason and Wronski2018). Yet, emerging research suggests that feelings toward out-groups can also affect partisan identification independently of in-group affiliations (Kane, Mason, and Wronski Reference Kane, Mason and Wronski2021). Thus, in-group closeness and out-group animosity may differentially affect political preferences depending on the political party or elite in question.
In this letter, we examine Donald Trump, whose explicit expressions of animosity are unique in modern presidential politics.Footnote 1 From claiming that President Obama was born outside of the United StatesFootnote 2 to referring to Mexicans as “rapistsFootnote 3” to accusing “low-income housingFootnote 4” of threatening suburbs, Trump’s controversial comments were often racially charged. Given this incendiary rhetoric, does Trump’s political support uniquely reflect public animosity toward social out-groups, or is it simply an extension of Republican Party affiliation?
In this heightened era of partisan polarization, public support for Trump may operate similarly to any Republican candidate or the Republican Party itself. Though Trump’s rhetoric and overt support from xenophobic right-wing groupsFootnote 5 raise the possibility that some of his political support arises from hatred of marginalized groups (Schaffner, Macwilliams, and Nteta Reference Schaffner, Macwilliams and Nteta2018), it is unclear whether this effect is unique to him. We thus seek to better understand public support for Trump and, in particular, the degree to which it represents a novel phenomenon in American politics. More broadly, by understanding how social dynamics shape presidential support, we provide insight into the role of elite rhetoric as well as the social divisions underlying American politics at large.
What about Out-Group Hate?
Brewer (Reference Brewer1999) distinguished the “in-group love” felt by group members toward their own kind from the “out-group hate” felt toward others. She makes the essential point that preferential treatment of in-group members does not always coincide with hostility toward out-group members (see also Bankert Reference Bankert2020). Instead, social group attachment can produce out-group hostility under several conditions: feelings of moral superiority, perceived conflict over resources and political power, forced collaboration, or a shared standard of relative worth.
Recently, Mason and Wronski (Reference Mason and Wronski2018) demonstrated that individuals who are closer to party-aligned groups, and cognitively understand the connection between their social identities and the parties, are more strongly attached to their party. Additionally, Kane, Mason, and Wronski (Reference Mason, Wronski and Kane2021) show that mass Republican and Democratic Party identification is shaped by sentiments towards party-aligned groups regardless of one’s membership in those groups. These works reveal that both in-group affection and out-group animosity constitute the social roots of partisan affiliation. As partisanship becomes increasingly socially defined, feelings toward political elites, parties, and their associated social groups may simultaneously affect each other. Thus, sentiments toward social groups—either positive or negative—may bolster support for particular elites or political parties.
Nonetheless, we suspect President Trump represents a distinct case. His 2016 primary election support was uniquely predicted by opposition to racial equity and immigration and negative attitudes toward African Americans and Muslims (Sides, Tesler, and Vavreck Reference Sides, Tesler and Vavreck2019). To the extent that Trump overtly portrayed his political function as protecting majority, higher-status groups (e.g., whites, Christians) from minority, lower-status groups (e.g., nonwhite immigrants, LGBT community, etc.), he potentially tapped into a reservoir of social animosity for these latter groups, even among nonpartisans. Further, research on sociopolitical sorting illustrates that these marginalized groups are disproportionately aligned with the Democratic Party, Trump’s partisan out-group (Mason Reference Mason2018). It therefore follows that, compared with other elites who avoid such incendiary rhetoric, and with the parties as a whole (which require some level of positive commitment rather than hostility alone [Brewer Reference Brewer1999]), Trump support may be significantly more rooted in group animosity.
Data and Methods
To test the proposition that Trump support is uniquely predicted by animosity toward groups aligned with the Democratic Party, we turn to the Democracy Fund’s Voter Study Group (2018), which conducted multiple survey waves in partnership with the Cooperative Campaign Analysis Project (CCAP) and YouGov.Footnote 6 This publicly available dataset includes several thousand respondents who were reinterviewed online across four survey waves: 2011, 2016, 2017, and 2018.Footnote 7 Importantly, the first wave of respondents were likely unfamiliar with Trump as an explicitly partisan figure and can be used as a baseline for comparison with later waves. In other words, those who would become Trump supporters in later waves were (likely) not yet Trump supporters in 2011.Footnote 8 This allows us to empirically gauge the extent to which reservoirs of group animosity existed in the mass public, regardless of extant party identity prior to Trump, that he could then draw upon during his political career.
Importantly, while we analyze panel survey data, we cannot treat this as a true panel because we do not have measures of our dependent variables in 2011. Our primary interest, therefore, is to investigate whether a citizen’s affect toward politically aligned social groups during an earlier point predicts the degree to which that same citizen approves of Trump at a later point (Sides, Tesler, and Vavreck [Reference Sides, Tesler and Vavreck2019] use a similar approach). Our secondary objective is to analyze the predictive power and magnitude of the 2011 group attitudes on Trump support relative to their effect on later feelings toward the parties and various party elites. By examining the difference between precursors to Trump support versus support for the parties and mainstream political figures, we can discern the sociopolitical uniqueness of Trump support in the American electorate.
Respondents were asked to indicate their feelings toward four Democratic-aligned social groups in 2011: African Americans, Hispanics, Muslims, and Gays and Lesbians (alpha = 0.78, mean = 0.60, SD = 0.21). Respondents also indicated their feelings toward two Republican-aligned social groups in 2011: whites and Christians (Mason Reference Mason2018; r = 0.46, mean = 0.75, SD = 0.20)Footnote 9. Our analysis employs an OLS regression model with the following general specification:
Trump favorability (or Party/Politician favorability)
i (2018) = b
0 + b
1Group Animus
i(2011) + b
2PID
i(2011) + b
3Ideology
i(2011) + b
4
$ \delta $
i(2011) +
$ \epsilon $
i,
where Group Animus captures the average feeling thermometer rating of the party-aligned group(s) examined in each model (again, Democratic-linked groups are African Americans, Hispanics, Muslims, and Gays and Lesbians; Republican-linked groups are whites and Christians). These thermometers are recoded to range from 0 to 1, with higher values indicating greater animosity toward the group(s) in question. PID indicates respondent i’s party identification in 2011, measured on a seven-point scale ranging from “Strong Democrat” (0) to “Strong Republican” (1); Ideology indicates respondent i’s self-placement on the five-point ideology scale in 2011 ranging from “Very Liberal” (0) to “Very Conservative” (1); and
$ \delta $
includes sociodemographic variables measured in 2011 including political interest, race, religion, educational attainment, gender, age, and income (see Supplemental Appendix for additional details).
Our main analyses feature several dependent variables. Trump Support is measured using a four-point item assessing the degree to which respondents’ opinion of Trump was favorable or unfavorable (asked in 2018). Party Support is measured using a 100-point feeling thermometer (asked in 2017). We also examined feelings toward other political figures from both parties, including Paul Ryan (R), Mitch McConnel (R), Hillary Clinton (D), Bernie Sanders (D), Chuck Schumer (D), and Nancy Pelosi (D). These ratings were all conducted in the 2018 wave, using the same four-point “favorability” measure used for Trump. All variables are coded from 0 to 1 to account for the different scales of the variables (i.e., the four-point favorability ratings and the 0–100 scale for party feeling thermometers). Our model isolates the effect of the composite animus scores toward Democratic- or Republican-aligned groups on Trump Support, Republican Party Support, Democratic Party Support, and Other Political Figure Support by accounting for baseline (2011) demographic and political differences between respondents.
Results
We first examine whether animus toward Democratic-linked groups in 2011 is predictive of Trump support in 2017 and 2018, controlling for baseline political and demographic measures.Footnote 10 Figure 1 presents predicted values of Trump approval (“Do you approve or disapprove of the way Donald Trump is handling his job as President?”) and favorability (“Do you have a favorable or an unfavorable opinion of … Donald Trump?”) in both 2017 and 2018 to account for any potential temporal or wording effects. Overall, 2011 animosity toward Democratic-linked groups is strongly related to later Trump approval. People who felt strong animosity toward Blacks, Hispanics, Muslims, and LGBT people in 2011 were significantly more likely to be fond of Trump once he appeared on the political scene (p < 0.001 for all models). We find a nearly identical pattern of results regardless of which particular measure of Trump support or year we use. Given this latter finding, and for ease of exposition, we will use the 2018 Trump favorability measure as our dependent variable in subsequent analyses.
Figure 1. Democratic Group Animosity and Trump Support Measures
Note: Predicted values from OLS models controlling for party, ideology, education, race, religion, age, gender, income, and political interest. Democratic group animosity is the average feeling thermometer score for African Americans, Hispanics, Muslims, and LGBT, coded to indicate less warmth. Originating models for this and all following figures can be found in Appendix A.
Is Trump Approval Unique among Republican Elites?
We next investigate whether the relationship between Democratic-linked group animosity and Trump support also occurs with other Republican elites and the Republican Party itself. Specifically, we predict support for the Republican Party in 2017 and favorability toward other prominent Republican figures in 2018—namely, Paul Ryan (then Speaker of the House) and Mitch McConnell (then Senate Majority Leader). Figure 2 presents predicted values of feelings toward these four targets, based on Democratic-linked group animosity in 2011.
Figure 2. Democratic Group Animosity and Trump, Party, and Elite Support
Note: Predicted values from OLS models controlling for party, ideology, education, race, religion, age, gender, income, and political interest. Democratic group animosity is the average feeling thermometer score for African Americans, Hispanics, Muslims, and LGBT, coded to indicate less warmth. The distribution of animosity is indicated by the histogram (right y-axis). All measures of approval from 2018, except GOP feelings, which was only available in 2017.
As demonstrated in Figure 2, preexisting animus toward Democratic-linked groups does not correspond to support for the Republican Party or its more established leaders. Among those with the lowest level of animus toward Democratic groups in 2011, their predicted favorability toward Trump in 2018 is around 0.3 (on the 0 to 1 scale). This level of favorability increases to over 0.5 among those who have the most animus toward Democratic groups, representing a 23-percentage-point increase. However, 2011 feelings of animosity toward Democratic groups do not predict favorability toward the Republican Party, Paul Ryan, or Mitch McConnell (no effects significant at p < 0.10). Thus, the relationship between Democratic-aligned group animosity and Trump support is not simply a product of being a Republican. Nor is it a measurement artifact of the four-point Trump favorability rating (compared with the 100-point party feeling thermometers) because this relationship does not hold with any of the other Republican leaders’ favorability ratings. Rather, Trump support is uniquely predicted by animosity toward marginalized groups in the United States, who also happen to fall outside of the Republican Party’s rank-and-file membership. For comparison, when we analyze warmth for whites and Christians, we find that it predicts support for Trump, the Republican Party, and other elites at similar levels. In other words, it is in the substantial degree to which his support derives from animus toward these Democratic groups that Trump is unique.Footnote 11
Trump Support-Animus Link by Party
Because of partisan loyalty (Bartels Reference Bartels2000), Democrats and Republicans should be the least likely to be moved in their approval of Trump. Conversely, Independents, unbeholden to partisan-group favoritism, should be the most inclined to modify their support for Trump based on Trump’s group-based rhetoric. We examine these possibilities in Figure 3, where we interact respondents’ partisan affiliations (Republican, Democratic, and Independent) with their Democratic group animosity scores and present predicted values of Trump favorability. Unsurprisingly, average Trump support is highest among Republicans and lowest among Democrats, with Independents in between. Yet, the effect of animosity toward Democratic-linked groups is substantively identical for both Democrats and Republicans. Among partisans (including leaners), those who are most hostile toward these groups are about 15 percentage points more supportive of Trump than those who are least hostile.
Figure 3. Trump Support and Animus toward Democratic Groups by Party
Note: Predicted values from OLS models controlling for ideology, education, race, religion, age, gender, income, and political interest, and interacting party with group animus. Democratic group animosity is the average feeling thermometer score for African Americans, Hispanics, Muslims, and LGBT, coded to indicate less warmth.
For Independents, this relationship doubles in size, where those most hostile toward Democratic-linked groups are about 30 percentage points more favorable toward Trump than the least hostile.Footnote 12 Thus, animosity toward Democratic-linked groups predicts Trump support, rather remarkably, across the political spectrum. Further, given the decisive role that Independents can play in elections (Klar and Krupnikov Reference Klar and Krupnikov2016), these results suggest that reservoirs of animosity are not necessarily specific to a particular party, and may therefore be tapped by any political elite.
A Democratic Equivalent to Trump?
While Trump support may operate differently from that of other Republican targets, a similar phenomenon may exist within the Democratic Party. Specifically, animosity toward groups aligned with the Republican Party may fuel support for particular Democratic elites, including Trump’s 2016 rival Hillary Clinton, and/or for the Democratic Party itself. If true, such a finding would indicate that, while Trump’s support may differ from that of other Republican elites, this relationship between party-aligned group animosity and political support may extend beyond Trump to other high-profile elites in American politics, especially in the context of the contentious, identity-driven 2016 presidential election.
To investigate this possibility, we estimate a similar OLS model but use a composite index of 2011 Republican group animus (whites and Christians) to predict later support for the Democratic Party and a number of prominent Democratic Party leaders: Hillary Clinton (Democratic Party candidate for President in 2016), Bernie Sanders (Progressive Senator and primary challenger to Clinton), Chuck Schumer (then Senate Minority Leader), and Nancy Pelosi (then House Minority Leader).Footnote 13
In notable contrast to Figure 2, Figure 4 shows no analogue to Trump within the Democratic Party. When modeling feelings toward the Democratic Party as a whole, increasing animosity to Republican-linked groups predicts decreasing support for the Democratic Party, contrary to expectations. Of all the Democratic political figures, only Bernie Sanders and Chuck Schumer have even mildly positive relationships with animosity toward Republican-linked groups, but these effects are small and nonsignificant. For example, moving from least to most animosity toward Republican-linked groups is related to an approximately 2-percentage-point increase in support for Sanders, but this increase is statistically indistinguishable from zero (p = 0.13).
Figure 4. Republican Group Animosity and Democratic Party and Elite Support
Note: Predicted values from OLS models controlling for party, ideology, education, race, religion, age, gender, income, and political interest. Republican group animosity is the average feeling thermometer score for whites and Christians, coded to indicate less warmth. The distribution of animosity is indicated by the histogram (right y-axis). All measures of approval are from 2018, except feelings toward the Democratic Party, which was only available in 2017. While we are constrained by the groups measured in the 2011 wave, we suspect that more extreme groups affiliated with the Republican Party (i.e., White Nationalists, NRA activists) would generate more animosity and potentially reveal a significant effect on future support for Democratic elites.
Of additional note, the (nonsignificant) relationship between Republican-linked group animus and Hillary Clinton support is negative. The results in Figure 4 thus provide further evidence that the observed relationship between group animosity and Trump support is neither an artifact of his serving as a de facto party leader nor a phenomenon that manifests symmetrically across candidates in the 2016 presidential campaign. Rather, it appears that Trump support is uniquely drawn from animosity toward social groups linked with the out-party, whereas most other elites’ political support is unrelated to these kinds of sentiments.
The Intricacies of Group Animosity
The previous analyses examined how animosity toward groups in 2011 linearly predicted later Trump support. Yet, it is not necessarily the case that the effect of Democratic group animus on support for Trump follows a linear pattern. For example, it may be that positivity toward Trump stems from animus toward Democratic groups more than negativity toward Trump stems from warmth toward Democratic groups, or vice versa.
We examine this possibility by specifying a quadratic term for the Democratic-aligned group sentiments measure within the aforementioned general model. In Figure 5, we see a noticeable nonlinear effect (quadratic term significant at p < 0.001), which indicates that lower values of animosity (i.e., warmth) toward Democratic groups matter little for Trump support. In contrast, above the midpoint of the scale (i.e., animosity), Trump support increases rapidly as animus toward Democratic groups increases. The effect of animosity on Trump support, therefore, does not appear to be primarily driven by warmth toward Democratic-linked groups generating lower support. Instead, most of the significant linear effect observed in previous figures arises from disdain toward Democratic-linked groups generating greater support.
Figure 5. Nonlinear Relationship between Trump Support and Animus toward Democratic Groups
Note: Predicted values from OLS models controlling for party, ideology, education, race, religion, age, gender, income, and political interest. Democratic group animosity is the average feeling thermometer score for African Americans, Hispanics, Muslims, and LGBT, coded to indicate less warmth. This figure allows for a nonlinear relationship between animosity and Trump support.
Which Groups Are Most Important?
To what the extent do feelings toward each of the four Democratic-linked groups—African Americans, Hispanics, Muslims, and Gays and Lesbians—contribute to Trump support? In Figure 6, we present the effects of group-based animosity on Trump support for each Democratic-aligned group individually, including both the linear and nonlinear specifications. We also feature histograms (right y-axes) indicating the distribution of feelings toward each group, where scores above 0.5 indicate animosity toward the group. Reports of negative feelings ranged from 13% of the sample (for Blacks) to 49% of the sample (for Muslims), indicating that a substantial share of Americans harbor animus toward these minority, Democratic-aligned social groups.
Figure 6. Trump Support and Animus toward Each Democratic Group
Note: Predicted values from OLS models controlling for party, ideology, education, race, religion, age, gender, income, and political interest. Democratic group animosity is the average feeling thermometer score for African Americans, Hispanics, Muslims, and LGBT, coded to indicate less warmth. This figure includes a nonlinear relationship between animosity and Trump support. Histograms represent the distribution of feeling thermometers for each group, with their percentage reported on the right y-axis.
The top-left panel of Figure 6 displays the relationship between animus toward African Americans in 2011 and Trump favorability in 2018. While both the linear and nonlinear terms are positive and significant (p < 0.006 and p < 0.001, respectively), the nonlinear specification provides more nuanced information: At low levels of animus towards Blacks, there is little effect on Trump support. But, as animus toward African Americans increases from 0.7 to 1.0, support for Trump grows from 0.45 to 0.56, an increase of 11 percentage points.
The top-right panel examines animus toward Hispanics. Both the linear and nonlinear terms attain statistical significance (p < 0.001 and p < 0.007, respectively), though we see a somewhat stronger linear relationship than we saw with African Americans. Trump approval steadily increases as animus toward Hispanics increases. As with the previous models, the nonlinear specification indicates that there is some acceleration of this relationship at higher levels of animus, with support for Trump increasing by roughly 11 percentage points moving from the midpoint of the 2011 animosity measure to its maximum value.
For feelings toward Muslims and Gays and Lesbians (bottom left and right panels, respectively), the linear and nonlinear specifications reveal similar results (p < 0.02 in all cases). Remarkably, respondents who are the most hostile toward Muslims are approximately 20 percentage points more supportive of Trump than those who view Muslims most favorably. Similarly, those who are most hostile toward Gays and Lesbians are about 15 percentage points more favorable toward Trump than those who hold the most favorable attitudes. Thus, while sentiments toward some groups relate to Trump support more prominently than others, these analyses reveal that the pattern of results discussed above is not confined to any one particular social out-group.Footnote 14
Conclusions
To examine the nature of Trump support, we identified a unique and powerful predictor of his popularity—animus toward Democratic-linked and traditionally marginalized groups. As the Republican Party grows increasingly white, Christian, and male, it may be tempting to explain Trump’s appeal with partisanship alone. However, that is not the case in these panel data. Trump appears to have been uniquely able to attract support based on preexisting animosity toward these groups. The same cannot be said for other Republican Party officials or the Republican Party itself. Similarly, we find no such pattern among Democratic officials or the Democratic Party.
It is important to highlight that these Democratic-aligned group feelings were measured before Trump became a serious nominee for president of the United States, making them exogenous to Trump’s vitriolic rhetoric. Trump, therefore, seems to have attracted those who were already feeling hostile toward marginalized groups, regardless of their extant partisan affiliations. Thus, rather than generating such feelings in the electorate, Trump acted more as a lightning rod, attracting those who were already harboring animus toward Democratic-aligned groups.
An important implication is that, given Trump’s success, future candidates may attempt to create a winning coalition based on activating group-based animosities through the explicit use of anti-out-group rhetoric. Without these kinds of explicit connections, citizens are left to merely infer whether their own out-group sentiments actually map onto elites’ own sentiments. In that case, out-group sentiments may only be weakly related to support for elites who avoid explicit out-group rhetoric. Thus, future research could measure the degree to which politicians couch their stances in explicitly negative sentiments toward out-groups.
Future research should also examine the relationship between caustic political elite statements and media attention. On one hand, Trump’s preexisting celebrity afforded him free media attention. However, as both mainstream and social media amplify outrage, relatively unknown politicians might make a national brand for themselves by publicly issuing combative group-based statements. In doing so, they can provoke outrage within the mainstream media and the political left, further satisfying the psychological needs of constituents harboring group-based prejudice (Cikara and Fiske Reference Cikara and Fiske2013). Therefore, as with Trump, future Republican leaders may similarly be able to translate incendiary rhetoric levied at marginalized groups into a stronger base of political support.
Finally, this research reveals a wellspring of animus against marginalized groups in the United States that can be harnessed and activated for political gain. Trump’s unique ability to do so is not the only cause for normative concern. Instead, we should take note that these attitudes exist across both parties and among nonpartisans. Though they may remain relatively latent when leaders and parties draw attention elsewhere, the right leader can activate these attitudes and fold them into voters’ political judgments. Should America wish to become a fully multiracial democracy, it will need to reconcile with these hostile attitudes themselves.
SUPPLEMENTARY MATERIALS
To view supplementary material for this article, please visit http://dx.doi.org/10.1017/S0003055421000563.
DATA AVAILABILITY STATEMENT
Research documentation and data that support the findings of this study are openly available at the American Political Science Review Dataverse: https://doi.org/10.7910/DVN/Q4VN7A.
ACKNOWLEDGMENTS
The authors would like to thank Didi Kuo and Lee Drutman for the invitation to write a piece for Vox Polyarchy that would eventually develop into this publication. We also thank Jeremy Pope and Chris Karpowitz for organizing the Partisanship Reconsidered conference at BYU that led to a book chapter that helped us generate the final idea for this publication. We also thank the other participants in that conference who helped us understand what questions we still needed to answer, including John Aldrich, Lisa Argyle, Michael Barber, Jim Curry, Jay Goodliffe, Marc Hetherington, Gary Jacobson, Samara Klar, Yanna Krupnikov, Paul Light, Frances Lee, Matt Levendusky, David Magelby, Kelly Patterson, Jessica Preece, and Caroline Tolbert.
CONFLICT OF INTEREST
The authors declare no ethical issues or conflicts of interest in this research.
ETHICAL STANDARDS
The authors affirm this research did not involve human subjects.
Comments
No Comments have been published for this article.