Abstract
Experimental research has long indicated that minority candidates are perceived as being more liberal than Whites. These expectations may hinder the electoral prospects of minority candidates campaigning for office who need the support of independents to win. Drawing upon a unique dataset of 933 informants (party delegates) and survey data from the 2006 Cooperative Congressional Election Study, we investigate the accuracy and implications of these race-based stereotypes for voting in the U.S. House of Representatives. Our analysis indicates that ethnic and racial minority candidates for Congress are typically viewed by voters as being far more liberal than objective indicators would suggest. Moreover, we find that these misperceptions of ideological extremity may harm minorities’ electoral chances at the ballot box, limiting their support among White voters, particularly independents. We conclude by discussing the implications of our findings for candidates campaigning for office.
Introduction
The number of racial and ethnic minorities in the House of Representatives has increased significantly in recent decades, reaching an all-time high of 101 non-White members in the 115th Congress. 1 Yet diversity in the population at large far outpaces this growth, with non-Whites comprising more than 38% of all Americans, compared with only 23% of the House. 2 Why are there so few minorities in Congress? Scholars attribute the low proportion of minorities in government to many factors, ranging from racial bias to candidate emergence (Juenke, 2014; Terkildsen, 1993). The fact that the vast majority of racial and ethnic minorities in Congress are elected from minority–majority districts (Boschema, 2015) suggests that White 3 voters are unsupportive of minority candidates; however, the extent to which White voters discriminate against minorities in the ballot box, once factors like partisanship and incumbency are taken into account, remains a matter of contention in experimental and observational research (Abrajano, Nagler, & Alvarez, 2005; Citrin, Green, & Sears, 1990).
While some may argue that explicit racial bias is a thing of the past, there is reason to believe that race-based attitudes play a significant role in candidate evaluations, particularly in shaping evaluations of candidate ideology. Previous research finds that minority candidates are perceived as being more liberal than Whites, and that these beliefs about ideological extremity undermine voter support for minority candidates (e.g., Jacobsmeier, 2014, 2015; McDermott, 1998; Sigelman, Sigelman, Walkosz, & Nitz, 1995; Visalvanich, 2017). We build on these findings theoretically and empirically, refining our understanding of the complicated ways race shapes voter attitudes and support for candidates.
Drawing upon theories of in-group/out-group effects (Duckitt, 2003; Tajfel & Turner, 1979), we expect that ideological evaluations are conditioned simultaneously by shared racial and partisan identities. Thus, perceptions of ideological differences will be heightened when White voters evaluate minority candidates who do not share their partisanship. Consequentially, we expect that White Republican voters will sense that minority candidates belong to not just one out-group but two (racial and partisan), amplifying perceived differences. Importantly, this intuition extends to White independents evaluating minority candidates—They should view minority candidates as discrepant on race and partisanship, and thus perceive minority candidates as more ideologically remote. In contrast to previous research which has centered on White partisans, our work is unique in theoretically and empirically engaging the perceptions and behavior of independents. Given the important role independents often play in electoral outcomes and the growth in this group in recent years (Jones, 2016), understanding their perceptions and behavior is critical.
We further build upon and extend the existing literature by incorporating a new measure of candidate ideology to add analytical rigor to our test of our theory (Adams, Merrill, Simas, & Stone, 2011; Buttice & Stone, 2012; Stone & Simas, 2010). This measure improves our understanding of whether voters’ perceptions of minority candidates’ ideology are in line or out of sync with objective indicators. Our measure of candidate ideology draws upon the perceptions of district informants. With multiple district informants rating the same candidate in the same district, we can obtain an average opinion of candidate ideology that is based on the assessments of individuals who are embedded in the congressional district and who are familiar with the candidates. When compared against these other indicators, our analysis suggests that the district informants are indeed picking up on information about candidate ideology that is overlooked by conventional measures of ideology, like DW-NOMINATE and CFScores. By integrating these heterogeneous indicators of ideology, we can better gauge the extent to which voters’ assessments of minority candidates are rooted in or disconnected from more objective indicators of ideology.
Our analysis demonstrates that perceptions of minority candidate liberalism extend beyond available indicators of their objective liberalism. Like previous research, our results show that Republican voters perceive minority candidates as more liberal, while Democrats’ perceptions of minority candidate ideology are unaffected by race. Moving beyond existing work, we further find that independent White voters display a larger gap in their perceived ideological difference with minority and White Democratic candidates than do either Republican or Democratic voters. These findings—which indicate that misperceptions of minority candidates’ ideology are both common and electorally consequential for independents—speak to the critical need to extend this line of research. If minority candidates are perceived as “too liberal” by independents, then this may be a significant barrier to minorities seeking congressional seats from districts and states with large White populations and moderate electorates. Furthermore, it may impede their electability in contests with a high degree of demographic and partisan heterogeneity, such as in Senate seats. Our results demonstrate that reshaping ideological perceptions held by independents and Republicans is crucial to minorities’ future electoral success.
Previous Literature
What challenges do racial and ethnic minorities face when running for congressional seats in majority-White areas? A number of scholars have explored White voters’ unwillingness to support minority candidates. Much of this work—both at the congressional and presidential levels—has found significant evidence of a direct effect of racial bias on Whites’ candidate evaluations and vote choices (Citrin et al., 1990; Dwyer, Stevens, Sullivan, & Allen, 2009; Huddy & Carey, 2009; Kinder & Sears, 1981; Terkildsen, 1993; Tessler & Sears, 2010; Washington, 2006). However, some studies (e.g., Highton, 2004; Matsubayashi & Ueda, 2011) have found limited evidence of a direct White racial bias toward minority candidates, particularly when campaign and voter characteristics were controlled for.
Beyond direct racial bias, scholars have increasingly looked to race-based stereotypes and cues as a source of White opposition to minority candidates. This research has demonstrated that voters—often short on time and information—rely on candidate identities to infer candidate ideology, stereotyping minority candidates as more liberal than their White male counterparts (McDermott, 1998; Sigelman et al., 1995; Williams, 1990). As these and other scholars point out, voter expectations regarding candidate ideology are likely rooted in the fact that, nationally, most non-White candidates are Democrats as are many non-White voters. 4 As a result, when lacking additional information, voters are likely to believe that minority candidates are more liberal than their White peers and use these beliefs when making vote choices.
Experimental studies such as those outlined above may result in particularly robust findings, given the controlled setting in which they take place, where subjects lack additional knowledge of the candidate, and the electoral context is held constant (Jacobsmeier, 2015). Compared with the experimental literature in this field, survey-based literature is somewhat limited. However, existing research largely confirms the generalizability of experimental findings, indicating that race acts as an ideological cue during elections. Using survey data from the 2006 midterm elections, Jacobsmeier (2014, 2015) explored the effect of race in shaping ideological perceptions of candidates, relying on survey data from the American National Election Study and the Cooperative Congressional Election Study (CCES). Jacobsmeier’s results indicate that White voters believe Black candidates to be more liberal than their White counterparts and that these misperceptions hurt Black candidates during congressional races. Most recently, work by Visalvanich (2017) explored the role of voter partisanship in shaping perceptions of minority candidate ideology and vote choice in the U.S. House of Representatives. Visalvanich provides convincing evidence to demonstrate that White voters are less likely to support Black and Latino candidates, and that their tepid support can be traced to misperceptions of ideological extremity and diminished perceptions of competence.
In sum, the existing literature indicates that candidate race may affect White voters’ attitudes and behaviors both directly (through racial bias) and indirectly (through race-based stereotypes about candidate ideology). Some of this research further indicates that partisanship may condition the effect of race-based beliefs about candidate ideology. Our study builds upon the exemplary work of Visalvanich, Jacobsmeier, and others, further articulating how candidate race and partisanship intersect to shape voter attitudes and behaviors. However, our work is distinct in focusing on independents—a group of voters who are often pivotal to the outcome of elections, but who have previously been underexamined. Our theory of in-group/out-group effects leads us to expect that White independent voters will view minority candidates as out-group members on both partisanship and race, which should magnify perceived differences in ideology, and attenuate White independents’ willingness to support minority candidates.
Theoretical Expectations: In-Group/Out-Group Evaluations
What drives voters’ perceptions of candidate ideological positions? Our theory centers on group identity. We propose that racial and partisan identities influence the extent to which White voters perceive minority candidates as ideologically extreme and that those perceptions shape vote choice. Group identification, such as partisan and racial identities, often shape the assumptions people make about other individuals (Brewer & Kramer, 1985; Huddy, 2013). In particular, people distinguish between those “in-group” members sharing their identity and those “out-group” members who do not. In-group membership often heightens perceived similarities (Duckitt, 2003; Tajfel & Turner, 1979) and beliefs about shared interests (Huddy, 2013). On the contrary, the use of stereotypes is more frequent for individuals when evaluating out-group members (Brewer & Kramer, 1985). We argue that group identity shapes voter perceptions of candidate ideology. Given that individuals perceive greater similarity with those whom they share an identity and greater difference with those whom they do not (Brewer & Kramer, 1985; Greene, 1999, 2004), we anticipate that voters will perceive a greater ideological distance between themselves and out-group candidates than those sharing their identity.
We further expect that the combination of partisan and racial identities shape these perceptions. Given research that finds that minorities are typically stereotyped as more liberal (Hutchings & Valentino, 2004; McDermott, 1998), we expect White voters evaluating out-group minority members will perceive them as being more liberal, all else constant. While minority candidates may be perceived as out-group members by all Whites, they face a unique overlap of race and party-based stereotypes when seeking support among White out-partisans. We anticipate that these overlapping identities will lead White voters to perceive greater ideological distance between themselves and out-partisan minority candidates, than they would for either in-partisan minority candidates or out-partisan White candidates, with whom they share at least one identity.
When it comes to who constitutes an out-partisan to whom, our perspective is shaped by Green, Palmquist, and Schickler (2002), who maintain that individuals hold stereotypes about the parties and compare their assessments of themselves with the stereotypes they hold about various partisan groups:
As people reflect on whether they are Democrats or Republicans (or neither), they call to mind some mental image, or stereotype, of what these sorts of people are like and square these images with their own self-conceptions. In effect, people ask themselves two questions: What kinds of social groups come to mind as I think about Democrats, Republicans and Independents? Which assemblage of groups (if any) best describes me? (p. 8, emphasis added).
In this conception, partisanship (and the absence of partisanship) is a social identification that prompts Democrats to view Republicans as an out-group, and vice versa. But, independents who view themselves as being neither Democrats nor Republicans will also perceive both partisan groups as their out-group. This perspective that some Americans socially identify as independents has been substantiated in other work (Greene, 1999, 2004). Because independents do not see Democrats or Republicans as part of their in-group, we expect that they should maximize differences and minimize similarities with partisans.
Previous research supports our expectation that group identity structures candidate evaluations. For example, Branton, Cassese, and Jones (2012) found that voters’ racial and partisan identities shape perceptions of Black and Latino members of Congress, with White independent and Republican voters less likely to recognize and positively evaluate minority representatives (compared with White House members). Conversely, they found little difference in White/non-White candidate recognition and evaluations among White Democrats. Research by Hood and McKee (2015) also highlighted the interactive role voter partisanship and candidate race play in White voter attitudes. In their analysis of three senatorial and gubernatorial elections, Hood and McKee (2015) found that White Republican voters supported Republican minority candidates at a similar rate to their White Republican counterparts, concluding that “for white conservatives, ideological purity trumps race” in the contests they explored (p. 50). Other scholars have proposed that the overlap between identities may affect ideological perceptions of some candidates. For example, Koch (2002) found that female Democrats are often perceived as more liberal than their Republican female or Democratic male peers. These studies inform our expectations here, indicating that attitudes toward candidates may be a function of the interaction between racial and partisan identities.
Given that the overwhelming majority of minority candidates running at the time of data collection (2006) were Democrats, our hypotheses and analysis focus on evaluations of minority and White Democratic candidates. 5 As race signals out-group status to all White voters, we first hypothesize that White voters will perceive minority Democratic candidates as relatively more liberal than White Democratic candidates (Hypothesis 1). Following our expectation that voters will consider both partisan and racial in- or out-group status, we anticipate that when White Democratic voters form attitudes toward minority candidates sharing their own party, they may employ only race-based evaluations and as a result perceive a smaller ideological distinction between themselves and these candidates (compared with independents and Republicans). In contrast, Democratic minority candidates being evaluated by White independent and Republican voters are likely perceived as belonging to not one but two out-groups (racial and partisan). As such, we expect that perceived ideological distance between minority and White Democratic candidates will be greater among White independent and Republican voters, compared with White Democratic voters (Hypothesis 2). Finally, we expect perceptions of ideological distance to have consequences at the polls, hypothesizing that perceived ideological distance will shape vote choice among all White voters (Hypothesis 3). We now turn to a discussion of the methods and data we employ to test our expectations.
Method and Data
Because we are interested in how Democratic minority candidates are viewed by individuals from different partisan and racial groups, we concentrate on the perceptions of White Democrats, Republicans, and independents, and exclude minority respondents. We test these propositions by exploring perceptions of candidate ideology and the electoral implications of those perceptions for Democrats, independents, and Republicans. We draw our data from the 2006 CCES. The CCES is a nationally representative study that was conducted during the 2006 midterm election. 6 Respondents were asked about their political perceptions, vote choice for the House of Representatives, and demographics. Our key variable of interest is the respondent’s perceived ideological distance to the candidate, which we construct by subtracting the respondent’s own ideological orientation (on a 1 = “extremely liberal” to 7 = “extremely conservative” scale) from their perception of the Democrat’s ideological position (on a 1 = “extremely liberal” to 7 = “extremely conservative” scale). Thus, we obtain a variable that ranges from −6 (the respondent views the candidate to the left) to +6 (the respondent views the candidate to the right), with 0 indicating that the candidate is perceived to be at the respondent’s ideal point. 7
We supplement the CCES data with information about the district (e.g., prior presidential and congressional vote, percent minority) and information about the Democratic and Republican candidates for Congress such as incumbency status, racial and ethnic identity, and ideology operationalized as DW-NOMINATE (Poole & Rosenthal, 1984) and CFScores (Bonica, 2014). Because we wish to examine whether White voters perceive minority candidates as being more ideologically liberal than they actually are, the inclusion of the DW-NOMINATE score and CFScores are especially important in enabling us to compare voters’ perceptions of the candidates’ ideology with more objective indicators of ideology. By including these measures of ideology, we can assess the extent to which White voters perceive minority candidates as being more liberal than objective indicators would suggest.
However, one shortcoming of the DW-NOMINATE score is that it only incorporates data on actual roll-call votes and neglects positions that candidates might take on issues that never make it to the floor of the House. Likewise, the CFScore provides important insights into candidate ideology by mapping the relationship between candidates, interest groups, and contributors. It has the advantage of providing ideological estimates for all candidates running for office, not just incumbents as is the case with DW-NOMINATE. However, it is dependent upon Federal Election Commission (FEC) data about contributions, and thus omits candidates who fail to report to the FEC, even if those candidates received a sizable proportion of the vote.
We supplement DW-NOMINATE and CFScores with a more holistic indicator of ideology that draws from a 2006 study of informants from 155 congressional districts who shared their perceptions about the candidates’ ideology (Adams et al., 2011; Buttice & Stone, 2012; Stone & Simas, 2010). Informants were sampled from the delegates to each major party’s national conventions in 2004, and thus are knowledgeable experts in their congressional districts. 8 These informant data were collapsed to the district level to produce aggregate impressions about the candidates’ ideology and then merged with the CCES data from respondents. 9 Appendix A outlines the minority candidates included in the analysis, and Appendix B reports the variables we use in our models and their coding. The informant data provide us with a critical measure of the candidates’ ideology which, combined with DW-NOMINATE and CFScores, can be used to compare against voters’ ideological perceptions of the candidates to separate objective ideology from the effects of perceived ideological liberalism associated with being a minority candidate.
For the purposes of our analysis, we define a minority candidate as either Black or Hispanic. In our initial sample, there were 12 Black Democrats, six Hispanic Democrats, three Black Republicans, and one Hispanic Republican, for a total of 22 minority candidates. The small numbers of Black and Latino candidates led us to group them into a single category of “minority candidates.” We do not mean to imply that these representatives are the same; however, the limited number of them in the 2006 election impedes our ability to examine them separately in this analysis. We also estimated the models presented for African Americans only, which resulted in similar findings. 10 In the 155 districts for which we have informant-based candidate ideology data, there are 18 Democratic and four Republican minority candidates. 11 As mentioned earlier, because of the limited number of Republican minorities in the sample, we restrict our analysis to Democratic minority candidates. All of our analyses are weighted to reflect the probability of a response and are clustered by district to account for the fact that districts are not sampled independently.
Empirical Analysis
Our first dependent variable is respondents’ perceptions of ideological distance. Table 1 reports respondents’ perceived ideological distance to minority and nonminority Democrats, split by the respondent’s partisanship. Respondents reported their partisanship on a 7-point scale ranging from strong Democrat (1) to strong Republican (7), which we collapsed into a 3-point scale for ease of interpretation. We classify Democrats as strong, weak, and leaning (1, 2, 3); independents as “pure independents” (4); and Republicans as strong, weak, and leaning (5, 6, 7). 12
Perceived Ideological Distance to the Democratic Candidate.
p < .05. **p < .01. ***p < .001 (two-tailed tests).
Recall that our first hypothesis predicted that Whites would perceive minority Democratic candidates as more liberal than their White peers. The data in Table 1 show that Republican and independent voters perceive minority candidates as more liberal, but this difference is not statistically significant for White Democratic voters. In keeping with our second hypothesis, the race of the candidate interacts with the partisanship of the respondent to influence perceived ideological distance. Consistent with our expectations, we observe independents and Republicans perceiving minority candidates as more ideologically remote than nonminority candidates (−0.672 and −0.370, respectively) compared with Democrats (0.045) which is statistically indistinguishable from zero. This finding corroborates our theory that individuals who view minority candidates as out-group members on two dimensions (race and partisanship) will exaggerate ideological differences as compared with individuals who view them as discrepant on just one dimension (race).
Next, we examine whether perceived ideological distance is rooted in reality or is a reflection of inaccurate stereotypes. In other words, is the variance observed in Table 1 a product of voter perceptions or candidates’ actual ideological positions? Minority candidates are more often Democrats, compared with White candidates. As such, voters’ perceptions of them as more liberal may be an accurate assessment of their ideology, a product of race-based stereotypes, or some combination of the two. To assess the underlying cause of the perceived differences, we must examine the independent impact of candidate race and objective indicators of their ideology on perceived ideological distance.
To test whether the differences we find in Table 1 are simply an artifact of Democratic minorities being more liberal, we use ordinary least squares (OLS) to examine voters’ perceived ideological distance from candidates after taking into account more objective indicators of ideology. To capture ideology, we include three measures—two employed by previous research in this field (DW-NOMINATE and CFScore) and one unique informant-based measure of ideology. We expect that the inclusion of the informant-based measure of ideology will contribute ideological information that is unique from DW-NOMINATE and CFScores. Because our models incorporate all three measures of ideology, our analysis constitutes a strict test of the hypothesis that independents and Republicans misperceive their ideological distance to Democratic minority candidates.
As mentioned previously, we expect the use of identity-based stereotypes to be more prevalent among out-partisans, who are less likely to know as much about the candidates and who are more likely to perceive them as different. We draw upon an established literature on candidate evaluation for guidance on control variables to include in our models (Conover & Feldman, 1989; Lodge, McGraw, & Stroh, 1989). Consistent with our theory that in-partisans view out-partisans as more extreme, we expect that respondent ideology should shape perceptions of the candidates’ ideology, with more conservative respondents viewing Democratic candidates as more liberal. Toward this end, we include a measure of the respondent’s ideology. 13 Likewise, attitudes about high-profile partisan figures, like the president, should similarly color evaluations. As our data are from 2006, we expect that as a respondent’s level of approval of G. W. Bush increases, their perception of the Democratic candidate’s ideological distance should also increase. To capture this effect, we include a measure of the respondent’s presidential approval rating. 14
Respondents’ positions on the high-profile issues of the day should also condition their interpretation of the candidates’ ideology for similar reasons, with respondents who take more conservative positions on the issues perceiving Democratic candidates as ideologically dissimilar from themselves. Using a factor analysis of the respondents’ position on issues ranging from stem cell research, minimum wage, intervention in Iraq, abortion, environmental protections, immigration, social security, affirmative action, capital gains, and the Central America Free Trade Agreement (CAFTA), we construct an overall measure of the respondents’ issue position. 15
Prior research suggests that several factors should affect the accuracy of ideological judgments, including respondent’s knowledge, 16 education, 17 and age, 18 with more knowledgeable, educated, and older people more accurate in their placement of the candidates’ ideology (Koch, 2003). Last, because information is more plentiful about incumbents rather than challengers, we expect that voters should be more accurate in their assessments of the incumbent’s ideology compared with the challenger’s, and so we include a measure for incumbency. 19 Because Table 1 indicates that respondent partisanship interacts with candidate race to influence perceptions of ideological distance, we split our sample by respondent partisanship. 20
Table 2 shows the results of our OLS regression. Taking Models 1 to 3 as a whole, the first finding to emerge is that the coefficient for minority Democrat is negative and significant for independents and Republican voters, but not for Democratic voters. Independent and Republican voters view minority Democrats as significantly more liberal and ideologically remote than White Democratic candidates, whereas Democratic voters do not perceive a difference in their proximity to minority Democratic candidates.
Predicting Perceived Ideological Distance to the Democratic Candidate.
Note. Regression coefficients with standard errors in parentheses. Data are clustered by district and weighted for the probability of response.
p < .05. **p < .01. ***p < .001 (two-tailed tests).
These findings are quite remarkable given that we include controls for all three observable indicators of candidate ideology and we include multiple observable indicators of the respondent’s ideology—all of which should be correlated with perceived ideological distance. For instance, we might expect the Democratic minority candidate to be perceived as more liberal if they actually were more liberal, based on objective indicators, or if they were being assessed by a conservative respondent. The negative and significant coefficient for respondent ideology indicates that more conservative respondents view Democratic candidates as more liberal and detached from their own ideological views. But even after controlling for this relationship, independent and Republican voters continue to view minority Democrats as more liberal and distant compared with White Democrats. The negative and significant coefficient for presidential approval reflects the fact that supporters of G. W. Bush view Democratic candidates as more liberal and ideologically distant, but minority Democratic candidates continue to be viewed by independents and Republicans as more liberal and ideologically remote even after accounting for this relationship between presidential approval and perceptions of ideological distance.
Similarly, the negative and significant coefficient for respondent issue position reveals that individuals who hold conservative values on the issues tend to view the Democratic candidate as more liberal and ideologically extreme. Yet, even after controlling for this effect, the coefficient for minority Democratic candidate remains negative and significant in the model for independent and Republican voters, suggesting that taking a conservative stance on the issues does not explain the perception independents and Republicans hold that minority Democrats are out-of-step with their own ideology.
Figure 1 plots the effect of partisanship on perceived ideological distance for minority and nonminority candidates, while holding constant the other variables in the model at their mean. 21 Democrats are predicted to perceive minority and nonminority candidates as slightly to the right of their own ideological ideal point. However, the confidence intervals between minority and nonminority candidates overlap, indicating the Democratic voters do not perceive candidates as more distant based on their minority status.

Effect of partisanship on perceived ideological distance for minority and nonminority candidates.
Independent and Republican voters are distinct from Democrats in that they perceive both minority and nonminority candidates to be to the left of their own ideological ideal point. But they view minority and nonminority candidates differently: Minority candidates are perceived to be significantly more liberal than their nonminority counterparts as evidenced by the lack of overlapping confidence intervals.
These results provide robust evidence that independent and Republican voters perceive non-White Democratic candidates as more liberal and detached from their own ideological views, and that this perception of difference cannot be explained away by alternative explanations that might account for the ideological distance—like objective candidate ideology, respondent ideology, presidential approval, or position on issues. In contrast, Democratic voters’ views about ideological proximity are unrelated to the candidate’s racial identity. These findings are wholly consistent with our group identity theory which predicts that White independent and Republican voters will perceive minorities as belonging to two out-groups (racial and partisan), and thus will exaggerate their perceived ideological differences. In contrast, White Democrats only differ from minority Democratic candidates on the race dimension, and thus perceived ideological differences should be minimized.
In terms of the remaining variables in the models—incumbency, respondent knowledge, education, age, and gender—we do not have directional hypotheses. Rather, we expect that these factors ought to influence the “accuracy” of the respondent’s judgments about the candidate’s ideology. That is, more knowledgeable respondents should be more “accurate” than less knowledgeable respondents, and respondents should be more “accurate” in judging the ideology of incumbents over challengers (due to the relative advantages incumbents have in terms of name recognition and familiarity); some research suggests that male voters are more knowledgeable or “accurate” than female voters about political issues.
What are the substantive effects of independents and Republicans overstating their ideological distance to Democratic minority candidates? To investigate, we run a logistic regression on the probability of voting Democratic in the congressional race (Table 3). Based on the literature on congressional elections, we control for characteristics of the candidate and district that ought to affect the likelihood of voting Democratic in our models, for instance, whether the seat is open or incumbent dominated, 22 the challenger’s office-holding experience, 23 the log of the difference between the Democrat’s and Republican’s spending, 24 and whether the race was designated by Congressional Quarterly as a “Key” race 25 (Herrnson, 2012; Jacobson, 2012; Mayhew, 1974). In addition to the respondent’s perception of the candidates’ ideology, minority status of the Democratic candidate, and the Democratic candidate’s NOMINATE, CFScore, and informant-based measure of ideology, we include controls for incumbency, 26 the district’s previous Democratic presidential and congressional vote, 27 and the percent minority in the district. 28 We also include variables characterizing attributes of the respondent that are influential to vote choice, like ideology, presidential approval, political knowledge, education, age, gender, and issue position.
Likelihood of Voting for the Democratic Candidate.
Note. Logit coefficients with standard errors in parentheses. Data are clustered by district and weighted for the probability of response.
p < .05. **p < .01. ***p < .001 (two-tailed tests).
Table 3 presents the results of our logit. Our key expectation is that perceptions of liberalism and ideological distance will have a strong and significant effect on the probability of voting Democratic. Model 4 strongly corroborates this expectation. The positive and significant coefficient for the respondent’s perceptions of the candidate’s ideological proximity suggests that when all the other variables in the model are held constant, it is much better for candidates to be perceived to be on the right than on the left. But because minority candidates are perceived to be significantly more liberal than nonminorities (Table 1), this places them at a relative disadvantage to their White counterparts. Although the coefficient for Democratic minority candidate is insignificant in Table 3 (suggesting that voters do not explicitly discriminate), Democratic minority candidates are indirectly harmed due to perceptions of excessive liberalism, which places them at a disadvantage in attracting votes.
Because logit coefficients are not directly interpretable, we generate predicted probabilities below. Table 4 calculates the predicted probability of voting Democratic based on the coefficients presented in Model 4, varying mean ideological distance from Table 1 for each partisan group and minority candidate status. These probabilities are plotted in Figure 2. Minority candidate status has no direct effect on the likelihood of voting as shown in Model 4, but the indirect effect of minority candidate status becomes evident when perceived ideological distance is taken into account. Because independents and Republicans view minority Democratic candidates as significantly more liberal (Table 1), and because perceptions of liberalism diminish the likelihood of voting (Table 3), the effect of minority Democratic candidates works indirectly to significantly lower the likelihood of support among independents and Republicans as a result of their perceived ideological distance. Importantly, the confidence intervals for independents and Republicans do not overlap by minority status, suggesting that minorities are uniquely disadvantaged vis-à-vis their nonminority counterparts.
Predicted Probability of Voting for the Democratic Candidate, by Minority Candidate and Partisanship.
Note. Probabilities are derived from the coefficients in Model 4, varying the mean value of perceived ideological distance based on Table 1 (varying partisanship and minority candidate status), while holding all of the other variables in Model 4 constant.

Predicted probability of voting Democratic varying perceived ideological distance for minority and nonminority candidates.
The predicted probabilities show that independents have a 25.5% decreased chance of supporting a minority candidate, whereas Republicans have about a 6% diminished chance. As expected, Democratic respondents’ likelihood of support is unaffected by minority candidate status.
Why do perceptions of ideological distance exert a stronger effect on independents’ vote choice than on Democrats’ or Republicans’? Because partisanship is the strongest predictor of vote choice, Democrats will vote for Democratic candidates and Republicans will vote for Republican candidates with less regard to ideological position. For partisan voters, ideological perceptions may make a difference at the margins, but, by in large, their vote choice will reflect their partisanship. But because independents are uncommitted to either party, factors like ideological perception will exert a stronger impact.
As Democratic minority candidates are perceived as being more liberal and ideologically remote than nonminorities by independent and Republican voters, and because perceived ideology exerts such a strong impact on vote choice, Democratic minority candidates’ chances of winning may be negatively affected by misperceptions of their ideology, particularly in electoral units where independents and Republicans are plentiful. Importantly, the effects of misperceived ideology appear to affect Democratic minorities’ chances of capturing the support of independents more than any other partisan group. Because partisanship is such a strong predictor of voting, with independents often the only group that is truly “up for grabs,” our results suggest that Democratic minorities’ electoral prospects will be diminished in moderate and swing districts.
Discussion and Conclusion
Although previous research is mixed as to whether White voters explicitly discriminate against minority candidates (Hutchings & Valentino, 2004; Sigelman et al., 1995), the literature on group identities suggests that Whites may view minority candidates as dissimilar from themselves in important ways. Corroborating previous research, our results suggest that White voters perceive minority candidates as more liberal and ideologically distant than nonminority candidates. By showing that partisanship conditions these effects, with independent and Republican voters perceiving minority Democrats to be more remote than Democratic voters, we extend the theory of in-group/out-group effects to incorporate both race and partisanship. Our results indicate that ideological (mis)perceptions have a substantively meaningful and significant impact on vote choice, revealing an important indirect pathway for the influence of candidate race in voting decisions. Our work builds on previous scholarship that demonstrates that minorities are viewed as more liberal; however, our work is unique in that we develop a theory predicting that race and partisanship shape ideological perceptions, and provide empirical evidence to sustain our prediction.
Substantively, our analysis suggests that minority Democrats are perceived as “too liberal,” particularly among independents who are 25.5% less likely to support a minority candidate. In the context of increasingly polarized congressional elections, independents are a pivotal group because they can swing elections. If we are correct, then minority Democrats’ difficulty in appealing to independent voters may account for their paucity outside of heavily Democratic areas, like those found in majority–minority districts.
Our work is consistent with others who suggest that minority candidates may garner the support of Whites if they moderate their positions on issues (Abrajano et al., 2005); however, significant questions remain about which issues are the most important to emphasize, as well as the most effective way to communicate with independent voters. Given the distance White independents perceive between themselves and minority Democrats, it is unclear whether such appeals would be successful. Our work also raises the question of whether minority Democrats risk losing their core supporters when they moderate themselves, given that White Democrats do not hold the misperceptions of these candidates’ ideologies found among independents and Republicans. This research provides insight into the treacherous terrain minority Democrats must navigate to build support among independents without alienating their base.
Footnotes
Appendix A
Districts With Minority Candidates Included in the Sample.
| District | Democratic minority candidate | Number of informants | Number of respondents |
|---|---|---|---|
| CA-31 | Becerra | 4 | 46 |
| CO-04 | Paccione | 7 | 100 |
| GA-05 | Lewis | 9 | 93 |
| IL-02 | Jackson, Jr. | 1 | 80 |
| IL-04 | Gutierrez | 2 | 43 |
| MD-04 | Wynn | 8 | 75 |
| MN-05 | Ellison | 11 | 103 |
| MO-05 | Cleaver | 4 | 104 |
| NJ-13 | Sires | 3 | 41 |
| NM-01 | Madrid | 12 | 168 |
| NY-06 | Meeks | 1 | 27 |
| NY-11 | Clarke | 4 | 54 |
| PA-02 | Fattah | 7 | 36 |
| TX-09 | Green | 2 | 60 |
| TX-15 | Hinojosa | 1 | 54 |
| TX-18 | Jackson-Lee | 3 | 73 |
| TX-20 | Gonzalez | 2 | 65 |
| TX-28 | Cuellar | 1 | 70 |
| VA-03 | Scott | 5 | 69 |
| WI-04 | Moore | 7 | 85 |
Appendix B
Description of Variables.
| Variable | Coding |
|---|---|
| Respondent Partisanship | −1 = Democrat (strong, weak, lean) 0 = Independent (pure) 1 = Republican (strong, weak, lean) |
| Respondent’s Perceived Ideological Distance to Democratic Candidate | −6 = Candidate is to the left of respondent 0 = Candidate places on respondent’s ideal point 6 = Candidate is to the right of respondent |
| Respondent Vote Choice | 0 = Republican 1 = Democratic |
| Minority Democratic Candidate | 0 = Nonminority 1 = Minority |
| DW-NOMINATE Score | −1 = Liberal 1 = Conservative |
| Informant-Based Measure of Democratic Candidate’s Ideology | 1 = Extremely liberal 2 = Liberal 3 = Somewhat liberal 4 = Moderate 5 = Somewhat conservative 6 = Conservative 7 = Extremely conservative |
| CFScore | −3 = Extremely liberal –2 = Liberal –1 = Somewhat liberal 0 = Moderate 1 = Somewhat conservative 2 = Conservative 3 = Extremely conservative |
| Incumbent Party Is Democratic | 0 = Republican incumbent party 1 = Democratic incumbent party |
| Open Seat | 0 = Incumbent dominated 1 = Open seat |
| Challenger Office-Holding Experience | 0 = No office-holding experience 1 = Office-holding experience |
| Log of Relative Spending | ln(Democratic spending – Republican spending), 0 if missing |
| Congressional Quarterly Key Race | 0 = Uncompetitive 1 = Congressional Quarterly Key Race |
| District’s Previous Democratic Presidential Vote | 0-100 |
| District’s Previous Democratic Congressional Vote | 0-100 |
| Percent Minority in District | 0-100 |
| Respondent Ideology | 1 = Extremely liberal 2 = Liberal 3 = Somewhat liberal 4 = Moderate 5 = Somewhat conservative 6 = Conservative 7 = Extremely conservative |
| Respondent Presidential Approval | 1 = Strongly disapprove 2 = Somewhat disapprove 3 = Neither approve nor disapprove 4 = Somewhat approve 5 = Strongly approve |
| Respondent Knowledge | Average number correct: Member of Congress partisanship, governor’s partisanship, Senator 1’s partisanship, Senator 2’s partisanship |
| Respondent Education | 1 = No high school diploma2 = High school graduate 3 = Some college 4 = 2-year degree 5 = 4-year degree 6 = Postgraduate |
| Respondent Age | 0-100 |
| Respondent Is a Female | 0 = No 1 = Yes |
| Respondent Issue Position | Factor analysis: Positions on stem cell research, minimum wage, Iraq, abortion, environment, immigration, social security, capital gains taxes, and the Central America Free Trade Agreement (CAFTA) |
| Respondent Is White | 0 = No 1 = Yes |
Appendix C
Perceived Ideological Distance to the Democratic Candidate (Replication Analysis With Black Candidates Only).
| Black Democratic candidates | Non-Black Democratic candidates | Difference | |
|---|---|---|---|
| Democratic voters | 0.156 | 0.307 | −0.151 |
| Independent voters | −1.796 | −1.171 | −0.625* |
| Republican voters | −4.003 | −3.545 | −0.458*** |
p < .05. **p < .01. ***p < .001.
Appendix D
Predicting Perceived Ideological Distance to the Democratic Candidate (Replication Analysis With Black Candidates Only).
| Model 1 |
Model 2 |
Model 3 |
|
|---|---|---|---|
| Black Democratic candidate | −0.139 (0.131) |
−0.503**
(0.194) |
−0.201 (0.136) |
| DW-NOMINATE score | 0.695***
(0.164) |
0.840 (0.504) |
0.415 (0.241) |
| CFScore | 0.221**
(0.092) |
0.301 (0.212) |
0.415***
(0.108) |
| Informant-based measure of ideology | 0.210***
(0.038) |
0.130 (0.105) |
0.239***
(0.045) |
| Incumbent party is Democratic | 0.532***
(0.150) |
0.664 (0.480) |
0.489*
(0.227) |
| Respondent ideology | −0.756***
(0.025) |
−1.180***
(0.085) |
−1.143***
(0.031) |
| Respondent presidential approval | −0.349***
(0.062) |
−0.118 (0.079) |
−0.159***
(0.033) |
| Respondent knowledge | −0.532***
(0.164) |
−0.602*
(0.277) |
−0.941***
(0.171) |
| Respondent education | −0.028 (0.017) |
0.015 (0.042) |
0.023 (0.018) |
| Respondent age | 0.001 (0.002) |
0.003 (0.004) |
−0.003 (0.002) |
| Respondent is a female | 0.092*
(0.047) |
0.022 (0.146) |
−0.016 (0.050) |
| Respondent issue position | −0.084 (0.048) |
−0.455***
(0.138) |
−0.151***
(0.047) |
| Constant | 2.767***
(0.313) |
3.818***
(0.787) |
3.969***
(0.310) |
| F test | 137.57*** | 87.19*** | 236.61*** |
| R 2 | .504 | .702 | .615 |
| Number of observations | 2,588 | 478 | 2,897 |
Note. Regression coefficients with standard errors in parentheses. Data are clustered by district and weighted for the probability of response.
p < .05. **p < .01. ***p < .001 (two-tailed tests).
Appendix E
Likelihood of Voting for the Democratic Candidate (Replication Analysis With Black Candidates Only).
| Model 4 | |
|---|---|
| Perceived ideological distance to Democratic candidate | 0.847***
(0.074) |
| Black Democratic candidate | −0.213 (0.528) |
| Respondent party identification | −0.752***
(0.099) |
| DW-NOMINATE score | 0.108 (0.560) |
| CFScore | 0.328 (0.292) |
| Informant-based measure of ideology | −0.019 (0.132) |
| Incumbent party is democratic | 0.849 (0.479) |
| Open seat | −0.119 (0.236) |
| Challenger office-holding experience | −0.017 (0.205) |
| Log of relative spending | −0.003 (0.189) |
| Congressional Quarterly Key Race | 0.317 (0.210) |
| District’s previous Democratic presidential vote | −0.044**
(0.016) |
| District’s previous Democratic congressional vote | 0.025***
(0.006) |
| Percent minority in district | −0.005 (0.008) |
| Respondent ideology | 0.435***
(0.106) |
| Respondent presidential approval | −0.969***
(0.104) |
| Respondent knowledge | 1.209**
(0.422) |
| Respondent education | 0.029 (0.056) |
| Respondent age | 0.003 (0.006) |
| Respondent is a female | −0.023 (0.147) |
| Respondent issue position | −1.124***
(0.170) |
| Constant | 1.584 (1.383) |
| F test | 33.73*** |
| Number of observations | 4,584 |
Note. Logit coefficients with standard errors in parentheses. Data are clustered by district and weighted for the probability of response.
p < .05. **p < .01. ***p < .001 (two-tailed tests).
Appendix F
Acknowledgements
We would like to give a special thanks to Walter J. Stone for generously sharing his data. Judd Thornton provided theoretical and empirical guidance that contributed positively to the paper’s development. We are grateful for the research assistance of Liam Hayes and Alex Hegner, who helped with data collection. We appreciate the time and energy of the editors and reviewers, whose efforts have significantly strengthened the paper.
Declaration of Conflicting Interests
The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.
Funding
The author(s) received no financial support for the research, authorship, and/or publication of this article.
