Race, Partisanship, and Democratic Politics: The Role of Racial Attitudes in Motivating White Americans’ Electoral Participation

Abstract

While there is considerable research on the role racial attitudes play in shaping white political preferences, relatively little is known about how racial attitudes influence white participation in democratic politics. We present a model examining the relationship between racial attitudes and political participation in the 2016, 2018, and 2020 U.S. national elections. Using a variety of measures of political participation, our analysis presents a clear finding: the direction of the relationship between latent conservative racial attitudes and political participation is asymmetrical among partisan sub-groups, with conservative racial attitudes motivating participation among white Republicans and, to a greater degree, depressing participation among white Democrats. This finding has stark implications for how racialized appeals are likely to be deployed in an era of increasing affective partisan polarization.

Following the 2016 presidential election, there was active debate on the importance of white voters’ racial attitudes in determining the outcome of that contest. ¹ Scholarship in the following years has conclusively established that racial attitudes affected whites’ vote choice (e.g., Hooghe and Dassonneville 2018; Schaffner et al. 2018) and that the impact was exaggerated in key swing states, aiding Trump in the Electoral College (Sides et al. 2017). Indeed, the presidencies of Obama and Trump have brought a renewed focus on the centrality of race in American electoral politics. Since 2008, racial attitudes have strongly affected vote choice for president (Hooghe and Dassonneville 2018; Jardina 2019; Knuckey and Kim 2015; Lewis-Beck et al. 2010; Piston 2010; Reny et al. 2019; Schaffner et al. 2018; Sides et al. 2018; Tesler 2013), Congress (Abrajano and Hajnal 2015; Algara and Hale 2019; Hale 2019; Luttig and Motta 2017; Petrow et al. 2018), and even the 2016 Republican presidential primary (Tucker et al. 2019).

However, vote choice is only part of the story. While we now know that racial attitudes affected vote choice, much less is known about how racial attitudes shaped voters’ decision to turn out to vote in the 2016 election, let alone participate in other types of pre-election activity like volunteering, donating, or even displaying a lawn sign. This is an especially salient question, given the historic unpopularity of both major party candidates in 2016—a factor we might reasonably expect could affect participation in unpredictable ways.

Despite the comprehensive bodies of work on political participation and on the relationship between racial attitudes and vote choice, there has been little examination of how racial attitudes condition political participation. A small body of research has found that racially conservative attitudes may decrease white political participation in the mass public (Cepuran and Berry 2022; Luttig 2017; Pasek et al. 2014) while studies on the effect of racial attitudes on voter turnout (Chong and Rogers 2005; Luttig 2017; Pasek et al. 2009) have come to conflicting conclusions.

While these studies suggest that in the aggregate racial attitudes may have a depressing effect on turnout, there is reason to believe that the effects may be conditioned by partisanship. Research on the effect of racial attitudes on political participation has found that conservative racial views may increase some forms of civic activity like tea party activism (Maxwell and Wayne Parent 2013; Tope et al. 2015) or joining anti-tax protests (Sears and Citrin 1982). Recent work by Banda and Cassese (2021) and Cepuran and Berry (2022) also provides insight. Both these studies find that white Americans are less likely to participate in politics when they have higher levels of racial resentment.

In this paper, we seek to expand on this nascent body of literature by comprehensively examining the relationship between racial attitudes and multiple types of white political participation in 2016, 2018, and 2020. First, we improve on the methods used in previous studies by generating our measure of racial attitudes from a new set of national survey questions recently introduced by DeSante and Smith (2020) in the Cooperative Election Study (CES). ² Second, we determine the relationship between individual-level racial attitudes and a variety of political activities among white Americans, including turning out to vote, running for office, volunteering for a campaign, and making a political donation. Third, we break down how these relationships differ by partisanship and how racial attitudes operate within the major party coalitions. Finally, we build on existing research by analyzing midterm elections (which are traditionally understudied) and are thus able to assess whether the relationship between racial attitudes and political participation is solely a byproduct of Trump’s presidential campaigns.

We find that the effect of racial attitudes on the political participation of white voters is asymmetrical. Republicans are more likely to participate in most of our measured forms of political activity when they have highly conservative racial views. By contrast, Democrats with more conservative racial attitudes are significantly less likely to participate in politics (a relationship that also holds to a lesser degree in the aggregate among all voters). Each of these results holds across the 2016, 2018, and 2020 general elections. These findings strongly suggest that while racially conservative views diminish political participation among white Democrats (and to a lesser degree independents), those views mobilize white Republicans to participate in a variety of political activities.

1. Racial attitudes & political participation

Racial attitudes have long been understood to affect modern American political behavior. Classic research by Carmines and Stimson (1989) identifies racial attitudes as the basis of a major partisan realignment beginning in the 1950s, and subsequent research by Kinder and Sanders (1996) and Hillygus and Shields (2014) has noted that this shift coincided with a shift from explicit racist appeals to “dog whistle” language that primed white voters’ fears and prejudices without offending social norms against explicit racism. Party sorting largely incorporated racial attitudes into voter partisanship in the following decades (Carmines and Stimson 1989; Engelhardt 2019; Layman and Carsey 2002; Lee 2002; Santucci and Dyck 2022; Sides et al. 2018), but the candidacy of Barack Obama in 2008 signaled a return of explicit racial attitudes as a strong predictor of voting behavior (e.g., Tesler 2013). This finding echoes an earlier study by Petrow (2010) which finds that both white racial liberals and white racial conservatives are more likely to participate in politics when their racial attitudes are activated by a black congressional candidate. In recent years, explicit racial appeals have returned in force to American politics, with minimal blowback (Valentino et al. 2018).

In the Obama and Trump eras, there is a large body of work demonstrating that racial attitudes are predictive of vote choice and policy preferences. Conservative racial attitudes drove down Obama’s vote share among whites in both the 2008 (Clarke et al. 2011; Lewis-Beck et al. 2010; Piston 2010) and 2012 (Jardina 2019; Knuckey and Kim 2015) presidential elections. The effect was not constrained to Obama: Trump and Republican congressional candidates since 2008 have benefited from increased support among racially conservative white voters, even without Obama on the ballot (Abrajano and Hajnal 2015; Algara and Hale 2019; Luttig and Motta 2017; Petrow et al. 2018; Sides et al. 2017; Tolbert et al. 2018). Importantly, these effects appear to be driven by individual-level racial attitudes rather than the racial composition of the voter’s geographic context (Windett et al. 2013).

Beyond vote choice, decades of political science research have established that racial attitudes affect whites’ preferences on a variety of racialized policy areas. Research prior to the 2008 election finds that racial attitudes have large effects on whites’ policy attitudes in racially coded issue areas like welfare and school integration (e.g., Federico 2006; Hurwitz and Peffley 2005; Mendelberg 1997, 2001; Sears et al. 1980; Winter 2008). Since 2008, public opinion has polarized further on the basis of racial attitudes, in large part in reaction to the Obama presidency (Abramowitz and McCoy 2019; Craig and Richeson 2014; Luttig and Motta 2017; Petrow et al. 2018; Tesler 2013).

Despite the wealth of research on the salience of racial attitudes for voters’ policy preferences and vote choice in modern American politics, there has been a relative paucity of research on the effect of these attitudes on political participation writ large. Pasek et al. (2009) find little effect of racial attitudes on turnout in 2008. Research in this arena has often focused on linked fate sentiment among minorities and the benefits of descriptive representation for minority participation (Chong and Rogers 2005; Rocha et al. 2010; Whitby 2007). Among whites, studies have found that racial resentment increased turnout among Republicans and decreased it among Democrats in the 2010 (Luttig 2017) and 2016 (Banda and Cassese 2021) elections.

Recent studies on group status threat and affective polarization also suggest that a linkage between whites’ racial attitudes and political participation exists. Research on group status threat has found that when whites view their racial identity as no longer a benefit to their place in the social hierarchy, they are less likely to participate in politics (Gest 2016). Recent studies show that many white Americans express concern about their decreasing power economically (Gest 2016; Mutz 2018), culturally (Abrajano and Hajnal 2015), politically (Parker and Barreto 2014), and as a share of the national population (Craig and Richeson 2014). These anxieties have been shown to directly affect whites’ partisanship (Jardina 2019), ideology (Craig and Richeson 2014), immigration policy preferences (Abrajano and Hajnal 2015), satisfaction with democracy (Enders and Thornton 2022), and vote choice (Mutz 2018).

While status threat alone is unlikely to increase political participation among whites (and may in some cases decrease it), there is cause to believe that conservative racial attitudes may have a motivating effect. Studies have shown that racialized appeals by political parties and politicians such as Trump (Jardina 2019; Lamont et al. 2017; Tolbert et al. 2018; Tucker et al. 2019) mobilized racially resentful white voters in both primary and general election contexts in 2016. These appeals are particularly effective mobilizing forces in a political climate where both parties have made race a salient feature of their messaging and policy platforms (Schaffner et al. 2018; Sides et al. 2017; Tesler and Sears 2010; Tolbert et al. 2018) and where racial attitudes are a salient predictor of attitudes towards government (Filindra et al. 2022).

Beyond turnout and vote choice, some studies have shown that racial attitudes can have effects on white political participation. Hassell and Visalvanich (2015) find that whites are less likely to write their member of Congress in support of a non-racial political cause if they are given racial cues that it benefits minorities. More recent research by Banda and Cassese (2021) and Cepuran and Berry (2022) finds a dampening effect of racial resentment on multiple types of political participation, such as attending political meetings or donating to a campaign, in recent presidential elections.

2. The cross-pressuring role of partisanship

There is strong reason to expect that any relationship between white racial attitudes and political participation will be conditioned by partisanship, as Republican politicians (like Donald Trump) more often appeal to voters’ racial resentment and Democrats (like Hillary Clinton) more often rhetorically embrace racial inclusivity and pluralism (Sides et al. 2018). Previous studies have demonstrated that racialized rhetoric from candidates increases the salience of voters’ racial attitudes on their vote choice (e.g., Mendelberg 2001; Valentino et al. 2018). While the Republican party may be effective at mobilizing racially conservative white Republican voters through appeals that align with their perceptions of status threat, racially conservative white Democrats are less likely to be mobilized. For these white Democrats, politicians from their own party are largely choosing not to appeal to their racial conservatism and appeals by Republican politicians are likely to be dulled in effectiveness by the cross-pressuring forces of partisanship—much as individuals with cross-pressuring group identities are less likely to be politically engaged (Mason 2016, 2018). We expect that the end result of this asymmetry is that while racially conservative white Republicans will be more engaged, racially conservative white Democrats will not only be less likely to vote (Krupnikov and Piston 2015; Luttig 2017), but less likely to participate in any form of political activity.

This partisan asymmetry is especially likely to appear given the rise of affective polarization (sympathy for a partisan citizen’s in-party and hostility to their out-party) in American politics (Abramowitz and McCoy 2019; Carmines and Nassar 2021; Iyengar et al. 2019). This affective polarization extends beyond partisanship to groups associated with partisan identities—including racial groups (Robison 2019). As affective polarization increases, partisan citizens are likely to update their racial attitudes in line with shifts in their feelings towards the out-party, and vice versa (Westwood and Peterson 2020). This process wherein partisans are increasingly polarizing affectively towards out-partisans and towards out-groups perceived to be in the out-partisan party coalition (e.g., blacks in the Democratic Party) creates an environment wherein conservative racial attitudes are especially likely to be galvanizing for Republican partisans and demotivating for Democratic partisans. The increasing inseparability of race and partisanship in the United States (Westwood and Peterson 2020) as well as increasing antipathy among citizens towards social groups perceived to be affiliated with out-partisans (Robison 2019) suggest that the relationship between racial attitudes and political participation will be conditional on partisanship.

Empirically, some existing work points to such a conditional relationship. Conservative racial attitudes increase whites’ likelihood of engaging in right-wing protests (Sears and Citrin 1982) and affiliating with the Tea Party (Maxwell and Wayne Parent 2013; Tope et al. 2015). In a recent study, Banda and Cassese (2021) find that racially conservative attitudes decrease some types of political participation among Democrats, but not among independents or Republicans. In his study of the 2010 election, Luttig (2017) finds that racial attitudes affect political participation distinctly among different partisan sub-groups.

Evidence from studies of vote choice also suggests that partisanship is a key conditioning factor in determining both the direction and magnitude of the effect of racial attitudes. Phoenix (2020) finds that Republican presidential candidates in the 1980s and 1990s benefited from strategically appealing to the racial anger of white voters. Studies of recent elections show that Republican candidates do better among voters with conservative racial attitudes (e.g., Algara and Hale 2019; Luttig and Motta 2017; Petrow et al. 2018; Tesler 2013). In conjunction, these findings suggest that Republican voters may be motivated to participate by conservative racial attitudes, unlike Democrats.

3. Hypothesis

We expect that our analysis of the 2016, 2018, and 2020 elections will reveal that the relationship between white Americans’ racial attitudes and political participation is conditional. White Republicans with conservative racial attitudes will be motivated to participate at higher rates. By contrast, white Democrats with conservative racial attitudes will be cross-pressured by partisanship and the Republican party’s affiliation with and appeals to racial conservatism.

* ${H_1}$ : More conservative racial attitudes are associated with increased political participation among white Republicans, but decreased participation among white Democrats in 2016, 2018, and 2020.

Our research design provides a strong test of our hypothesized relationship. While previous studies have confirmed that Trump was successfully able to activate conservative racial attitudes in 2016 (Banda and Cassese 2021; Cepuran and Berry 2022), our study further examines whether such a relationship emerges in the 2018 midterm—where Trump did not appear on the ballot and where the white electorate was substantially less sympathetic to his messaging.

4. Measuring racial attitudes

In this paper, we seek to not only provide a novel comprehensive assessment of the relationship between racial attitudes and political participation but also to overcome some of the methodological limitations of prior research. Many previous studies (e.g., Banda and Cassese 2021; Feldman and Huddy 2005; Kinder and Sanders 1996; Winter 2008) have focused on the impact of racial resentment, which relies on the association of blacks with threats to whites’ quality of life (Kinder and Sears 1981; Knuckey and Kim 2015). This measure sought to account for the late $20{\rm{th}}$ century shift in American political rhetoric away from “old-fashioned” explicit racism towards dog whistle appeals. To measure this new “symbolic” strain of racial resentment (Kinder and Sears 1981), the American National Election Study (ANES) since 1986 has sought to gauge respondents’ belief that blacks do not conform the Protestant work ethic, and denial that ongoing discrimination against blacks persists. ³

While the Kinder and Sears (1981) measure of racial resentment has been central to research on Americans’ racial attitudes (including their effect on political participation), it has also been seriously questioned in recent years. A major critique is that the racial resentment index may be conflating symbolic racism with conservative ideological views on economic redistribution (Feldman and Huddy 2005). Furthermore, recent research argues that rather than measuring attitudes towards blacks, the scale is actually measuring beliefs about the existence of structural racism (Kam and Burge 2018). Given the increased centrality of race in American politics in the Obama and Trump presidencies (Sides et al. 2018), the limitations of the racial resentment index have become increasingly problematic.

In response to these criticisms of the Kinder and Sears (1981) “symbolic” racism measure, a new measure of racial attitudes called the Fear, Institutionalized Racism, and Empathy (FIRE) scale was introduced in the 2016 CES. The FIRE scale attempts to both incorporate the emotional component of racial attitudes and omit policy considerations, in contrast to questions previously asked in both the ANES and CES. While the initial FIRE scale is made up of four items, we rely on the two survey items tapping into the denial of racism dimension of white racial attitudes asked in the 2016, 2018, and 2020 CES. ⁴ As Schaffner (2022) articulates, these two FIRE scale items provide scholars with another dimension of racial attitudes by measuring the degree to which whites deny racism is prevalent in American society, reflecting “contemporary conflicts over increasing racial diversity and white identity.” Respondents are asked to state their level of agreement on a scale from (1) strongly agree to (5) strongly disagree, with the midpoint (3) indicating neither agree nor disagreement:

1. 1. White people in the United States have certain advantages because of the color of their skin.

2. 2. Racial problems in the United States are rare, isolated situations.

The introduction of this new racial attitudes CES battery (which assesses the degree of denial of racism prevalent in the United States), in conjunction with the survey’s large sample size, gives us the capability to create a stronger estimate of racial attitudes. As shown by DeSante and Smith (2020), the FIRE scale improves on the Kinder and Sears (1981) racial resentment index in key areas, measuring important dimensions of racial attitudes not captured by the latter and being more predictive of vote choice in the 2016 presidential election. Given that the dynamics of racial conservatism have shifted since the 1980s from symbolic racism to white racial group identity (Jardina 2019; Jardina et al. 2021; Mutz 2018), it is critical that the measures of racial attitudes used in analyses of vote choice reflect contemporary schisms among whites over increasing racial diversity in the United States.

With this new measure in hand, we can then evaluate the how these attitudes relate to a variety of forms of political participation, including election turnout. ⁵ The FIRE scale allows us to capture both whites’ emotional reactions to racism as well as their cognitive beliefs about the prevalence of racial conflict in an increasingly racially diverse America (DeSante and Smith 2020; Schaffner 2022).

5. Research design

5.1. Measuring dimensions of political participation & racial attitudes

To evaluate the role of racial attitudes on various dimensions of political participation, we rely on the 2016, 2018, and 2020 CES cross-sectional survey years. We limit our analysis to white respondents, since our investigation is focused on the effect of whites’ racial attitudes on their level of political participation. These large- $N$ survey data provide a large sample of white respondents from all partisan persuasions during both presidential and midterm election cycles. These three cycles starkly vary in terms of citizen involvement in politics, most notably with the 2018 midterm cycle demobilizing voters relative to presidential cycles (Burden 2000). Inclusion of the 2018 election provides an especially strong test of our theory, given that those motivated to turn out are likely to be much more negative towards President Trump and his tactic of racial appeals in comparison to the 2016 election (Kernell 1977). Another key consideration is these large- $N$ survey data provide for decent sample sizes nested across all fifty states — states which vary in terms of political competition. Some respondents have high-profile election contests for U.S. Senate or Governor on the ballot, which may contribute to variation in voter-level participation on the basis of the contextual level of party competition. ⁶ In contrast to other popular survey data sources, such as the ANES, the large sample size of the CES data allows us to better control for key variables tied to political geography and to critically test our theoretical framework across presidential and midterm election cycles.

To measure racial attitudes, our key independent variable of interest, we follow the methodological approach of previous studies (i.e., Algara and Hale 2019, 2020; Schaffner et al. 2018) and treat those attitudes as a latent variable. As mentioned earlier, we rely on the two-item FIRE index questions assessing white societal racial advantages and the perceptions of the commonality of racism found in the 2016, 2018, and 2020 CES cross-sectional survey waves. As previously mentioned, these questions are designed to measure how white Americans feel about white racial privilege and the prevalence of racism in a diversifying society—a dimension Schaffner (2022) dubs “denial of racism.” We code responses to these two core FIRE index items in a direction that indicates greater conservatism (i.e., denial of racism) on racial attitudes. These questions are desirable given that they do not measure citizen policy preferences related to race, but rather the intensity of racial attitudes (e.g., Hooghe and Dassonneville 2018; Schaffner et al. 2018; Tolbert et al. 2018). ⁷ We concur with previous scholars (e.g., DeSante and Smith 2020; Schaffner 2022) asserting that this measure captures a distinct dimension of white racial attitudes relating to the denial of structural racial inequality in American society, rather than traditional measures rooted in racial resentment (Kinder and Sanders 1996; Kinder and Sears 1981) or racial stereotypes about minorities (Piston 2010). However, we consider these alternative measures of racial attitudes in the evaluation of our theory that racial attitudes motivates political participation among white Republican partisans while demobilizing political participation among white Democratic partisans in the form of successful robustness checks presented in the appendix using alternative measurements found in the 2016 and 2020 ANES cross-sectional survey waves.

To that end, following a similar methodological approach to recent studies measuring racial attitudes in the mass public (e.g., Algara and Hale 2019, 2020; Schaffner et al. 2018), we fit a graded scale item response theory (IRT) model for our pooled CES data from 2016, 2018, and 2020. ⁸ We compare the distributions of our scaled measure of latent racial attitudes in Figure 1 for the full sample, Democratic partisans, Independent partisans, and Republican partisans. ⁹ We also present these distributions by survey cross-sectional year in the appendix since we estimate racial attitudes using year-specific IRT models, confirming that the distributions of latent racial attitudes are identical independent of the specific survey questions asked in a given survey year.

Figure 1. Distribution of scaled latent racial attitudes by samples

There are two primary conclusions to be drawn from Fig. 1. First, white Democratic partisans are significantly to the left of both white Independent and white Republican partisans in terms of racial attitudes tapping into a denial of racism across survey years and in the pooled sample. Second, and confirming prior scaling work on racial attitudes among white Americans (Algara and Hale 2019, 2020), there is overlap in the distribution of racial attitudes among white Democrats, Independents, and Republican partisans. The distributions as shown in Fig. 1 also confirm research by Engelhardt (2020) showing that while partisans are becoming better sorted on the basis of racial attitudes, there is still significant intraparty variation in racial attitudes. In short, Fig. 1 reaffirms that the distributions of racial attitudes in our sample mirror existing measures and conform to expectations from the literature.

5.2. Specifying models of political participation

We now turn to specifying our models of political participation. ¹⁰ Political participation by the polity may manifest itself in differing ways, from the relatively low-cost activities of turning out to vote (Rosenstone and Hansen 1993) to more costly propositions such as contributing to campaigns (Magleby et al. 2018) and running for office (Stone and Maisel 2003). To that end, we leverage the standard battery of political participation questions used in the CES that measures whether an individual participated in the following activities in the order of weighted frequency among white Americans pooled across the 2016, 2018, and 2020 election cycles:

1. 1. Validated turning out to vote in a general election (56%)

2. 2. Validated turning out to vote in a primary election (33%)

3. 3. Contributing to a political campaign or cause (18%)

4. 4. Putting up a political sign (17%)

5. 5. Attending a political meeting (13%)

6. 6. Volunteering for a political campaign (6%)

7. 7. Running for political office as a candidate (4%)

As one can see, there is clear variation in the propensity for individuals to engage in various forms of political participation across activities, from the most common activity of voting in a general election to the least common form of participation of running for political office. However, outside of voting in general elections, most white Americans do not participate in politics. Figure 2 shows the results of an index of political participation constructed from the seven items above. As one can see, about a quarter of white Americans do not engage in any form of democratic participation in politics according to data from the full sample. ¹¹ Moreover, across all partisan groups, the distributions are right-skewed indicating a relative lack of political participation across groups. Additionally, in line with conventional wisdom and prior research (e.g., Abramson and Aldrich 1982), we find that both Democrats and Republicans are much more involved in democratic politics than political independents, with almost half of independents not being involved in any form of democratic politics. Taken together, across the full sample and all partisan categories (i.e., Republicans, Democrats, Independents), there is clear variation in the degree of political participation among white Americans that can be exploited to critically assess the degree to which racial attitudes manifest in white political participation.

Figure 2. Distribution of political participation index among white Americans

To that end, we specify a series of logistic regression models to assess how racial attitudes influence the propensity to participate in each of the seven individual political activities provided by the standard CES. The dependent variable of these seven individual activity models captures whether or not a respondent participated in that given political activity, coded 1 for participation and 0 if not. With the exception of the general and primary election turnout outcome variables that are validated by Catalist for the CES using state election records (Ansolabehere and Rivers 2013), all participation in these individual political activities is self-reported.

We specify seven baseline additive logistic regression models assessing the baseline relationship between racial attitudes and the probability of participating in a given political activity. ¹² The baseline additive models control for standard individual-level predictors of political participation, such as partisanship, gender, education, income, age, political knowledge, presidential approval, congressional approval, congressional delegation approval, scaled ideological preferences, and political interest (e.g., Campbell 2013; Larcinese 2007; Matsusaka 1995; Palfrey and Poole 1987; Schlozman et al. 2012; Verba and Nie 1987; Verba et al. 1995). Given the temporal nature of our pooled analysis, we also add year fixed-effects to account for inherent differences in turnout across election years. ¹³ The models also control for contextual motivators of participation, such as the presence of a contested U.S. Senate, gubernatorial, or U.S. House election on a voter’s ballot (see also: Anzia 2014; Cox 2015; Geys 2006). Lastly, given the hierarchical nature of the large- $N$ data structure, with voters nested within congressional districts, we estimate our model quantities of interest clustered standard errors by a given respondent’s congressional district.

We also posit an asymmetrical conditional theory of how racial attitudes influence democratic participation, with conservative racial attitudes motivating Republican participation and suppressing Democratic participation. To evaluate this conditional framework, we re-specify our previously described additive logistic regression models by interacting racial attitudes with a respondent’s partisan identity. Given our coding of partisanship as two dichotomous variables, with each capturing Republican and Democratic voters with independent partisans serving as the baseline category, we are able to post-estimate the marginal effect of racial attitudes on the probability of political participation across each partisan category.

Lastly, we move beyond each individual dimension of participation and specify both our individual and interactive models to predict the relationship between racial attitudes and the participation index articulated in Fig. 2. These ordinary least squares (OLS) regression models help assess the racial attitudes correspond to the total degree of political participation across the full sample and within partisan groups. ¹⁴ This approach allows us to evaluate the degree to which racial attitudes correlate with overall degrees of political participation.

6. Model results: The asymmetrical role of racial attitudes

6.1. Individual dimensions of participation

We begin evaluating the results of our models by assessing the additive marginal effect of racial attitudes on the probability of participating in various political activities. ¹⁵ Fig. 3A plots the first difference marginal effect of going from the minimum value of latent racial attitudes to the maximum level on the probability of participating in a given political activity with corresponding 95% confidence intervals derived from district-clustered standard errors. Each row of Fig. 3A corresponds to a given political activity estimated by a specific model for a total of seven model marginal effects while each panel of Fig. 3B plots the substantive predicted probabilities of engaging in a specific political activity across all values of liberal-conservative racial attitudes, with tick marks articulating this distribution of racial attitudes in the full sample. We find strong baseline evidence that greater levels of conservative racial attitudes among whites are significantly associated with a lower probability of participation in democratic politics across most activities. For example, going from the minimum to maximum level of latent racial attitudes is associated with a decrease in the probability of turning out to vote by 5% in either the general election or primary election context. In substantive terms, Fig. 3B shows that this marginal effect represents a substantive decline in the predicted probability of turning out to vote in a general (primary) election from about 71% (48%) to 66% (43%). This negative finding is also replicated in other forms of political participation beyond turning out to vote, with this minimum-maximum first difference marginal effect being significantly associated with a 7% and 2% decline in the probability of contributing of contributing to a campaign or volunteering for a campaign, respectively. Paradoxically, we find that conservative racial attitudes are slightly associated with a positive increase in the likelihood of running for office by about 2%. However, as we noted earlier, running for office is a fairly rare form of political participation, with only about 4% of white respondents indicating a run for office in a given election cycle. By contrast, we find null relationships between racial attitudes and putting up a political sign and attending a political meeting. Taken together, our baseline model finds mostly that not only do latent conservative racial attitudes lower the propensity of engaging in electoral accountability (i.e., turning out to vote), but also other forms of political participation.

Figure 3. Baseline marginal effects of racial attitudes on political participation

We now turn to evaluating our interactive model expectations in Fig. 4, which plots the same relationship between racial attitudes and propensity to participate in democratic politics across each discrete partisan group. Figure 4A articulates the marginal effects of racial attitudes on the probability of engaging in individual political activities across models (again with 95% confidence intervals) while Fig. 4B presents the substantive predicted probability of engaging in a given political activity across values of racial attitudes. As one can see in Fig. 4A, there is evidence that conservative racial attitudes correlate with an increase in democratic participation among white Republicans. Going from the minimum to the maximum value of liberal-conservative racial attitudes correlates with an increase in the probability of turning out to vote in a general election, turning out to vote in a primary election, giving a political donation, putting up a political sign, volunteering for a campaign, and attending a political meeting by 4%, 6%, 11%, 10%, 2%, and 2%, respectively. Figure 4B shows that these marginal effects translate to relatively sizable substantive relationships among white Republican partisans. For example, going from the minimum (very liberal) to maximum (very conservative) value of racial attitudes correlates with going from a 17% predicted probability of giving to a political cause to 28% predicted probability. This correlation can also be seen in the propensity to engage in democratic accountability by turning out to vote. Going from the minimum to maximum value of racial attitudes correlates with going from a 68% (42%) predicted probability in turning out to vote for a general (primary) election to 72% (48%) predicted probability. We note, however, that there appears to be no correlation between racial attitudes and the probability of running for office. Taken together, there is consistent evidence that conservative (i.e., denial of racism) racial attitudes are associated with a mobilization of white Republicans to participate in democratic politics across many other types of political activity.

Figure 4. Relationship between racial attitudes & political participation among partisans

Turning to Fig. 4C, we find that conservative racial attitudes largely do not correlate with political demobilization among white independent partisans. In the pooled models, greater conservative racial attitudes are associated with a significant decrease in the probability of white independents in only volunteering for a campaign (−3%) and attending a political meeting (−4%). This suggests that political independents are largely not swayed to participate in democratic politics on the basis of racial attitudes, in line with prior findings by Lamont et al. (2017) and Jardina (2019). Taken together, there is little evidence to suggest that conservative racial attitudes correlate with lower political participation among white independents.

Turning to Democratic partisans in Fig. 4E, we find robust evidence that conservative racial attitudes lower the propensity of white Democratic partisans to participate in politics. With the exception of running for office and attending political meetings, more conservative (i.e., denial of racism) racial attitudes correlate with a lower probability of white Democrats participating in every activity across all model specifications. This negative finding for Democrats, in contrast to the positive finding for Republicans, reflects a clear asymmetry with respect to how whites’ racial attitudes manifest themselves across partisan groups. Fig. 4E shows going from the minimum to the maximum level of conservatism in racial attitudes correlates with a decline in the probability of (1) putting up a political sign by 15%, (2) casting a primary election ballot by 20%, (3) being a political donor by 27%, (4) casting a general election ballot by 19%, and (5) volunteering for a campaign by 4%. These pronounced declines are also reflected in Fig. 4F, which plots the predicted probability of engaging in each political activity across levels of racial attitudes. To highlight this substantive decline in participation among Democrats over values of racial attitudes, we focus on the propensity of voting and political donation estimates. As Fig. 4F shows, going from the lowest (very liberal) to highest level (very conservative) of racial attitudes lowers the probability of turning out to vote in a primary election from 54% to 34%, a staggering decline of 20% in predicted probability over values of racial attitudes. We observe this similar pattern in the context of general election turnout, with going from the lowest to highest level of racial attitudes resulting in a predicted lower probability from 75% to 56%, a decline of 19% in predicted probability over values of racial attitudes. However, the highest magnitude predicted decline in democratic participation on the basis of racial attitudes among Democrats is in the domain of political contribution. Moving from the lowest to highest level of conservative racial attitudes correlates with a decline in the predicted probability of giving to a political cause among Democrats from 36% to 9%, representing a 27% decline in participation in this dimension. These findings underscore an asymmetry between the political participation of both Democratic and Republican partisans on the basis of racial attitudes. Not only do racial attitudes manifest in divergent ways among Democrats and Republicans, but the suppressing marginal effect of conservative racial attitudes on political participation among Democrats is consistently greater than the mobilizing relationship for Republicans.

6.2. Levels of indexed participation

In addition to the propensity to participate in individual political activities, we also evaluate whether our results hold when our dependent variable is an index of participation in democratic politics. Figure 5 presents results of our aforementioned OLS regression models assessing the relationship between racial attitudes and our political participation index of seven items as shown in the distributions presented in Fig. 2. Marginal effects for the full sample are estimated from an additive model while the partisan marginal effects are estimated from an interactive model. Congruent with the previous analysis, we find strong evidence for the baseline effect that more conservative racial attitudes are associated with a decrease in indexed political participation among white voters overall. In line with this aggregate finding, we also find that conservative racial attitudes demobilized white independent partisans (albeit to a much lesser extent than Democrats). Among white independents, moving from the least to most conservative values of racial attitudes correlates with a predicted modest decline of .16 participation items, going from 1.61 predicted items to 1.45 predicted items. This result comports with previous studies by Cepuran and Berry (2022) and Berry et al. (2019) and aligns with the finding by Filindra et al. (2022) that whites with more conservative racial attitudes are less trusting in government and have more negative feelings towards government—factors which are likely associated with decreased political participation.

Figure 5. Relationship between racial attitudes & political participation index across samples

Most importantly, the results of these regression models confirm the asymmetrical nature of the relationship between racial attitudes and participation for Democrats and Republicans. Going from the minimum to maximum level of conservatism for racial attitudes significantly raises the predicted number of political participation activities for white Republican partisans from 1.64 to 1.95 items on the participation scale, an increase of three-quarters of an item. For white Democrats, this same relationship moves political participation from 2.24 to 1.24 items, a decline of a single participation item on the scale. The results of these regression analyses also provide additional evidence that conservative racial attitudes demobilize Democrats to a greater extent than they mobilize Republicans, contributing to the asymmetrical nature in which racial attitudes influence political participation of white Americans on the basis of partisanship.

7. Discussion: White racial attitudes & partisan coalitions

In this paper, we find strong evidence that racial attitudes affect political participation among whites and that this relationship varies drastically by party. Like Cepuran and Berry (2022), we find that conservative racial attitudes decrease various types of political participation in the aggregate. Our paper uses novel data, measurement, and analysis to contribute to the emerging body of research (Banda and Cassese 2021; Cepuran and Berry 2022) that finds that partisanship cross-pressure affects the relationship between racial attitudes and political participation. In line with recent studies of racial attitudes and vote choice (e.g., Algara and Hale 2019, 2020; Schaffner et al. 2018), we generate a novel measure of racial attitudes that have not been used in prior studies of political participation. This measure overcomes many of the potential shortcomings of the long-used Kinder and Sears (1981) “racial resentment” measure that has been criticized for incorporating attitudes on economic redistribution in addition to racial attitudes.

While the effect of racial attitudes on vote choice may or may not vary based on party (e.g., Algara and Hale 2020; Luttig 2017), we find compelling evidence that the relationship between racial attitudes and political participation is strongly conditioned by partisanship. While conservative racial attitudes make Democratic voters less likely to participate in politics, Republican voters with more conservative racial attitudes participate in politics at higher rates. This finding provides a critical caveat to prior research showing that racial conservatism depresses participation. While, among whites, it does so in the aggregate and among Democrats, racial conservatism actually increases political participation among Republicans. By examining the 2018 midterm election, where Trump was not on the ballot and faced a considerably less sympathetic electorate, we are able to make the novel finding that this asymmetry is not merely a byproduct of Trump’s presidential campaigns.

This partisan asymmetry provides us with additional insight regarding the process of racial realignment in the United States. As a result of political changes in the 20th century, the Republican Party is now firmly associated with racial conservatism and the Democratic Party with racial liberalism (Carmines and Stimson 1989; Layman and Carsey 2002). As a result, white Democratic voters who are racially conservative are psychologically cross-pressured, whereas white Republican voters face no such dissonance. While many racially conservative whites have sorted into the Republican Party since the 1950s (e.g., Engelhardt 2019; Sides et al. 2018), those who do remain in the Democratic Party may be particularly disaffected with the political process—unrepresented by their own party on matters of race and ideologically unrepresented by the Republican Party.

While this paper provides a valuable confirmation of existing research, it also makes novel findings that would benefit from further study. Our finding that racial attitudes asymmetrically affect Republican and Democratic partisans’ likelihood of political participation invites a number of questions heretofore unaddressed. Future research could assess how this asymmetry affects party/candidate campaign strategy, party messaging, legislative behavior, and the shifting racial composition of the parties. In particular, we believe our findings indicate a need for additional studies of the effect of racial attitudes on voter engagement in primary elections, given that such elections tend to be dominated by high-propensity voters (e.g., Bawn et al. 2012; Schlozman et al. 2012).

Future research should seek to uncover the underlying mechanism driving our findings. We suspect that key demographic traits such as political knowledge and income differently predict conservative racial views depending on partisanship. If higher income and higher political knowledge are associated with more conservative racial attitudes among white Republicans, but more liberal racial attitudes among white Democrats, that would provide a potential explanation for the partisan asymmetry we demonstrate. Given the well-established positive predictive effect of income (Campbell 2013; Rosenstone 1982; Verba et al. 1995) and political knowledge (Lassen 2005; Larcinese 2007; Matsusaka 1995; Palfrey and Poole 1987) on participation, such a finding would have significant normative implications. In such a scenario, if white voters’ conservative racial attitudes are activated, Republicans would gain both voter turnout and activism among their white constituencies that possess attributes already associated with greater levels of political participation. While such an analysis is beyond the scope of this manuscript, we encourage future investigation that treats racial attitudes as the dependent variable.

In summary, our study improves our understanding of the relationship between racial attitudes and political behavior. In line with prior studies, we show that racially conservative attitudes are associated with lower political participation amongst white voters in the aggregate. Our findings demonstrate that while racially conservative Republican voters are more likely to participate in politics, the inverse is true of Democrats. In an era of mounting party polarization, there is increasing incentive for candidates and parties to activate their base voters rather than persuade swing voters. Our research shows that across the parties, these core partisans with the highest likelihood of political participation diverge dramatically in terms of their racial attitudes. Going forward, we expect that savvy Republican campaigns will attempt to activate their core partisan supporters by appealing to their racial conservatism. While Democratic campaigns in the past may have successfully used similar messaging to rally their supporters prior to racial realignment, our findings demonstrate that such tactics are unlikely to be effective in the current party landscape.