Revisiting Stereotypes of Non-White Politicians’ Ideological and Partisan Orientations

Abstract

This research revisits when and how voters use race as a cue for politicians’ ideological and partisan orientations. Using an embedded survey experiment that manipulates the race and policy positions of a (fictitious) Member of Congress, I provide a more comprehensive view of the role of ideological and partisan stereotypes in impression formation. Voters perceive non-White politicians as more liberal and more likely to be Democrats than otherwise-identical White politicians. This stereotyping persists even when the politician takes counter-stereotypical positions (e.g., a Black or Hispanic politician with a conservative record), and shapes non-White legislators’ approval ratings in significant ways.

Keywords

cue-taking voter perceptions ideological stereotypes partisan stereotypes non-White politicians

Introduction

How voters form impressions of politicians and their records in office is of central importance to our understanding of democratic accountability. Without knowledge of what has been done in their names, constituents cannot use electoral sanctions to punish “out of step” representatives (Ansolabehere & Jones, 2010). Sixty years of research on the American voter has, however, shown that few have the resources, interest, or capacity to gather such detailed information on every politician they are asked to evaluate (A. Campbell, Converse, Miller, & Stokes, 1960; Delli-Carpini & Keeter, 1996). Rather, citizens rely on a range of cues or “information shortcuts” to make sense of a complex political world (Popkin, 1991). In particular, voters frequently categorize newly encountered politicians as members of a general group, and use stereotypes of that group to form impressions of the individual (Fiske & Neuberg, 1990; Fiske & Taylor, 1991; Terkildsen, 1993).¹ For example, voters evaluating a female politician may rely on a belief that women are more liberal than men to infer that she holds generally liberal policy positions (Koch, 2000, 2002; M. L. McDermott, 1997).

Given the significance of race in structuring modern politics and parties (Carmines & Stimson 1989; Green, Palmquist, & Schickler, 2002, Ch. 6), it is unsurprising that numerous scholars have hypothesized that a stereotype of non-Whites as liberal has a particularly strong influence on voters’ evaluations of Black politicians. Empirical studies of these ideological stereotypes have, however, reached strikingly mixed conclusions. Some conclude that voters perceive Black politicians as more liberal than White politicians (M. L. McDermott, 1998) while others fail to uncover any differences in voters’ perceptions of politicians of different races (Colleau et al., 1990; Weaver, 2012) or find only conditional relationships (Sigelman, Sigelman, Walkosz, & Nitz, 1995). These conflicting findings may partly be due to differences in research designs—some studies infer voter perceptions from responses to candidates of different races, while others ask survey respondents to assess general groups of politicians.

This study revisits these stereotypes and proposes a direct test of whether voters systematically perceive non-White politicians’ records as more liberal than their White counterparts. I develop an experiment embedded in an Internet survey that randomly manipulates the race and policy positions of a (fictitious) Member of Congress (MC). In several ways, the experiment provides a more comprehensive view of how voters use the cue of a politician’s race than previous studies. I expand on previous work that focused on stereotypes of Black politicians to also examine stereotypes of Hispanic politicians, a group that has grown in size and prominence since early studies of cue-taking. And I extend previous work on ideological stereotypes (of non-White politicians as liberals) to explore partisan stereotypes (of non-White politicians as Democrats).

Early studies of cue-taking on the basis of race stressed that they applied to “low information” environments—where voters knew little about a politician except for their race (e.g., M. L. McDermott, 1998). In contrast, the experiment used in this study provided respondents with information about the policy positions the politician took. In doing so, it also allows me to assess whether the extent of cue-taking varies with individuating information about the politician’s stances. Previous work in social psychology has questioned whether people are more or less likely to follow group cues when presented with individuals with traits that run counter to the group stereotype. I assess whether non-White MCs are perceived as consistently more liberal/Democratic than their White counterparts, or whether the extent of these perceived differences varies with the ideological slant of the legislators’ records.

The results show that voters’ perceptions of politicians’ ideological and partisan orientations are strongly shaped by the cue of race. Voters perceived Black and Hispanic politicians as more liberal and more likely to be Democrats than otherwise-identical White politicians. The “cue” of race does not overwhelm the effect of policy positions on voters’ impressions. No matter whether they took mostly liberal or mostly conservative stances, however, voters placed a non-White politician to the left of a White politician. In turn, these stereotypes lead to distorted approval ratings for legislators: liberal Democrats are more likely to approve of a non-White politician, conservative Republicans less likely, even given identical levels of policy congruence. Even in “high information” environments that provide details of a legislator’s record, partisan and ideological stereotypes of non-White politicians can distort perceptions of what they have done in office and skew their approval ratings in significant ways.

Stereotypes of Politicians as Shortcuts for Voters

Few (if any) voters have the resources, interest, and capacity to carry out detailed information searches every time they are faced with a new politician or political situation to evaluate. Rather than engaging in a data-driven, individuating, process of evaluation, voters often categorize newly encountered individuals as members of a general group, and rely on generalizations about the group to form impressions of the individual (Fiske & Neuberg, 1990; Fiske & Taylor, 1991; Terkildsen, 1993). In this characterization, stereotypes are cognitive schema that link category labels with associated attributes. A stereotyping-based process of impression formation thus consists of two steps: An individual is examined to assess whether they fit into a relevant category, and then the attributes associated with that category are ascribed to the individual. For example, voters encountering a politician for the first time may categorize them as a Democrat and then infer that they are liberal, as the trait of “liberal” is associated with the category of “Democrat” in their minds (Judd & Downing, 1995; Rahn, 1993).

Voters categorize politicians in a range of ways beyond party labels. Stereotypes about gender (Koch, 2000, 2002; M. L. McDermott, 1997, 1998), religion (D. E. Campbell, Green, & Layman, 2011; M. L. McDermott, 2009), occupation (M. L. McDermott, 2005), and sexual orientation (Golebiowska, 2001) all serve as shortcuts that voters use to form impressions of politicians and their records. In this study, I focus on one of the “most chronically salient categories” (Bodenhausen, Kang, & Peery, 2012, p. 315), race, and the ideological and partisan stereotypes voters hold of non-White politicians.² Previous research on the hypothesis that non-White politicians are stereotyped as more liberal than White politicians has been extensive but produced conflicting results, while the hypothesis that non-White politicians are stereotyped as more Democratic than White politicians has received almost no attention despite some suggestive evidence.

Ideological Stereotypes of Non-White Politicians

To assess whether non-White politicians are seen as more liberal than White politicians, previous studies have followed one of three basic research designs: “indirect” studies that infer stereotype usage from voters’ candidate choices, “direct” studies that ask voters for their perceptions of a particular politician’s ideology, and studies that probe voters’ perceptions of general categories of politicians. Within each camp, however, researchers have reached conflicting conclusions.

Indirect evidence of ideological stereotypes comes from studies that show differences in how liberal and conservative voters evaluate politicians of different races. For example, M. L. McDermott (1998) draws on a nationwide survey that asked respondents to choose between (hypothetical) candidates for President whose race was randomly manipulated to be either Black or unspecified. Liberals were more likely to vote for a Black candidate than conservatives were. From this, she infers that voters stereotyped the Black politician as more liberal, leading liberals to support (and conservatives oppose) their candidacy. Other studies that use the same inference strategy, however, find no interaction effects between a voter’s ideology and the race of a politician (see Colleau et al. (1990, p. 393) for experimental evidence and Gay (2002, p. 723) for observational data).

Studies that directly ask voters to assess the ideologies of politicians of different races produce equally mixed findings. Sigelman et al. (1995) asked White voters from a jury pool in Arizona to evaluate a (fictitious) politician whose race and campaign platform were randomized. Respondents did rate a conservative Black or Hispanic candidate as less conservative than a White candidate who took the same positions. Ratings of the ideology of politicians who took liberal or moderate stances, however, did not vary with his race. Furthermore, Weaver (2012) uses a similar experimental design but reports no differences in perceptions of a White or Black politician’s ideology (p. 174).

Studies that ask for voters’ perceptions of general categories of politicians also reach conflicting conclusions. Schneider & Bos (2011) asked a sample of college students to rate various groups in terms of their liberalism. Respondents gave a mean score of 6.06 (out of 7) to the category “Black politicians” and 3.72 to “Politicians,” indicating that they saw Black politicians as much more liberal than (presumably mostly White) politicians in general. In contrast, Williams (1990) reports the results of a survey suggesting that White voters felt the phrase “liberal” was equally applicable to Black or White candidates for office, even though they rated them differently on character traits.

In short, previous research on whether voters stereotype non-White politicians as more liberal than their White counterparts has produced decidedly mixed results. Furthermore, prior studies have usually provided voters with little information besides the race of the politician they are to evaluate, either because they ask about general categories of politicians (e.g., Schneider & Bos, 2011;Williams, 1990) or because they describe fictitious candidates without including individuating information about their stances (e.g., Colleau et al., 1990; M. L. McDermott, 1998). Whether voters perceive non-White politicians as more liberal than White politicians—even when given information about their policy positions—is still an unanswered question.

Partisan Stereotypes of Non-White Politicians

If research on the stereotype of non-White politicians as more liberal than White politicians has been plentiful but inconclusive, research on the stereotype of non-White politicians as more Democratic than White politicians has been essentially nonexistent. This is a surprising omission—because party identification is generally considered more central to voters’ evaluations than ideology and because voters’ images of the parties are strongly influenced by the social groups they associate with them (D. E. Campbell et al., 2011; Green et al., 2002; Miller, Wlezien, & Hildreth, 1991). Especially given the importance of race to the modern party system and its coalitions (Carmines & Stimson, 1989), we might expect voters to use a politician’s race as a cue for their party affiliation.

Some direct data on this question come from the 2007 Cooperative Congressional Election Study, as discussed by D. E. Campbell et al. (2011). Respondents were asked about the partisan composition of various social groups: 76% believed that Blacks were “mainly Democrats” (compared with 2% saying “mainly Republicans” and 18% saying “a pretty even mix of both”), while 56% believed Hispanics to be mainly Democrats (compared with 7% saying Republicans and 33% saying an even mix). This association of Blacks with the Democratic party is, of course, not new: Using NES survey data from the 1960s and 1970s, Bastedo and Lodge (1980) show that the attitude “favors Blacks” was seen as strongly characteristic of a typical Democrat, and helped differentiate the “Democrat” and “Republican” labels for voters. Although they are a relatively recent addition to the Democratic coalition, it is possible that partisan stereotypes of Hispanics are in turn strengthening.

There is no other work, that I know of, that explores whether voters hold stereotypes of non-White politicians’ party affiliations. While it seems plausible that voters would perceive non-White politicians as more likely to belong to the Democratic party than White politicians, there is scant empirical evidence for this—and nothing exploring the persistence of such stereotypes in the face of information about a politician’s positions.

When Group Stereotypes and Individuating Information Conflict

The main question explored in this research is whether voters perceive non-White politicians as more liberal and more Democratic than White politicians who take identical policy positions. A related question is whether the specific policy stances that the politicians take moderate the degree to which voters rely on generalizations about the group to evaluate their record. In particular, does counter-stereotypical information (e.g., when a non-White politician takes conservative positions) lead voters to rely on a more individuating approach rather than relying on group generalizations?

Previous work in political science suggests that such “conflicting signals” lead voters to rely less on category cues and more on individuating information (Arceneaux, 2008; Bartholow, & Dickter, 2008; M. L. McDermott, 1997). For example, Arceneaux (2008) finds that counter-stereotypical position-taking by politicians leads voters to rely more on individuating information than on party cues. When presented with (for example) a conservative Democrat, voters are less likely to categorize them as stereotypical Democrats and more likely to rely on the policy positions they took. As the findings relate to this study, when voters assess a conservative Black politician, they may be less likely to rely on her race and assume she is liberal than when faced with a Black politician who takes (expected) liberal positions. As such, we would expect differences in perceptions of non-White and White politicians to be reduced when they take more conservative policy positions.

Taken collectively, the preceding considerations suggest several hypotheses about how voters perceive non-White politicians:

Hypothesis 1: Voters perceive non-White politicians as more liberal than White politicians.

Hypothesis 2: Voters perceive non-White politicians as more likely to be Democrats than White politicians.

Hypothesis 3: These perceived differences are reduced when non-White politicians take counter-stereotypical (i.e., conservative) policy positions.

Testing these hypotheses requires data that compare perceptions of otherwise identical non-White and White politicians. These comparisons are difficult to make using observational data as race and policy stances are often correlated in the real world. Instead, I designed a randomized experiment, embedded in a national survey, that I describe in the next section.

Experimental Data

I take advantage of a July 2011 survey of around 1,850 U.S. adults that Knowledge Networks (KN) sampled from their online panel.³ The sample was stratified by race, resulting in roughly equal numbers of Black (N = 623), Hispanic (N = 611), and White (N = 618) respondents. The use of this somewhat unusual sample raises concerns about one element of the experiment’s external validity, whether the results can be generalized to the broader population that I acknowledge here.

The sample mirrors the general population well in terms of demographics—respondents are slightly less well-educated than the U.S. adult population, but on average there is about the same age, income, gender, and regional composition (Table A-1 in the online supplementary materials compares the sample to the Census’ Current Population Survey for each of these measures). Indeed, the sample does a better job of approximating the population than previous studies of stereotypes that rely on samples of students (e.g., Schneider & Bos, 2011) non-Hispanic Whites alone (e.g., Sigelman et al., 1995) or residents of a single county (e.g., Terkildsen, 1993). Where it differs, most obviously, is in its racial composition: Blacks and Hispanics are significantly oversampled, and other minorities excluded.

The previous literature suggests that an individual’s race should not moderate the use of ideological stereotypes. M. L. McDermott (1998), for example, finds that Black and White respondents were equally likely to form impressions of candidates on the basis of ideological stereotypes and notes that any analysis that controls for those stereotypes should show no effects of respondent’s race (p. 910). As there is no theoretical reason to assume that different racial groups would be any more or less likely to stereotype non-Whites as liberal Democrats, the sample’s racial composition is of less concern (see also Druckman and Kam (2011) and R. McDermott (2002) for broader arguments about the focus on representative samples in experimental work).

Given the experimental method that allows us to isolate the causal effect of a politician’s race, these limitations on the sample’s representativeness are of less concern. Combining the randomized experiment with a large, national, sample allows broader conclusions to be drawn than previously possible. To assuage any further concerns that the results are due to the racial make-up of the sample, I follow M. L. McDermott (1998) and control for the respondent’s race in all of the models that follow. I also ran the analyses on each racial group separately; there was no evidence that the results varied significantly across races.

Initial Survey Items

The survey began by asking for respondents’ opinions “about some of the main issues being discussed in politics today” (full question wording is in the online supplementary materials). I selected four high profile bills that Congressional Quarterly and the Washington Post identified as recent “key” votes in Congress, and asked respondents if they favored or opposed: (a) the health care reform of 2010, (b) the stimulus bill from 2009, (c) immigration reform creating a pathway to citizenship, (d) increasing taxes on those earning US$250,000 or more, and (e) the use of racial profiling by airport security officials. This final issue was not on the congressional agenda, but was included to assess whether stereotypes are applied most on racial issues. As I discuss in the conclusion, there is no evidence that this racialized issue activated stereotypes any more than the nonracial issues did.

Experimental Manipulation

After several questions about other issues, respondents were told,

As you know, many Members of Congress use websites as a way of communicating with constituents. We are interested in how well these sites communicate information to voters. We’d like you to look at a screenshot from the current website of one U.S. Representative, Congressman [first name] [last name], and then ask you some questions about it.

Respondents were randomly assigned to see a site for a (fictitious) Black, Hispanic, or White MC. The names of the MCs were chosen to be as distinctively associated with a particular race as possible, in keeping with other experiments that manipulate race (e.g., Bertrand & Mullainathan, 2004). Using 2000 Census data (Word, Coleman, Nunziata, & Kominski, 2000), I selected surnames that were overwhelmingly associated with one race: the Black MC was named Joe Washington (in 2000, 90% of all adults with the surname Washington were Black); the Hispanic MC was named Jose Gonzalez (94% of all those named Gonzales were Hispanic); and the White MC was named Joe Mueller (97% of all those named Mueller were White). The website included a prominent image of the MC in the banner heading. Stock photos were used: In each case, the photo was a close-up shot of a smiling middle-aged man wearing a suit and tie (the online supplementary materials include all three screenshots).

The screenshot shown was of an “Issues and Legislation” page. I sampled the official websites of 20 House Democrats and 20 House Republicans, and designed the page to be as similar as possible. The screenshot included no mention of the MC’s party (none of the 40 sites I sampled included the MC’s party on their issues/legislation pages either). In keeping with the real sites, “links” to sign up for a newsletter, contact the MC, and view his schedule were shown.

The text of the page read, “Congressman [last name] continues to work on the major legislation that matters most to our district, including:” followed by a list of his positions on five bills. These mirrored the five policy questions respondents had answered at the start of the survey. The positions (included in the online supplementary materials) were described in the same way they were by actual MCs during congressional debate.⁴ Respondents were randomly assigned to an MC who either took the same positions as them on one of the five policies, or took the same positions on four of the five policies. Which issues they agreed on, and their order on the page, were also randomized.

Randomization checks confirmed that the distribution of the MC’s positions was orthogonal to other key variables. The mean number of policies on which the MC took a conservative position (which could range from 0 to 5) did not vary significantly by the race of the MC (2.50 for the Black MC condition, 2.53 for the Hispanic MC, and 2.57 for the White MC), by the race of the respondent (2.50 for Black respondents, 2.52 for Hispanic respondents, and 2.54 for White respondents), or by the ideology of the respondent (2.53 for those who took liberal positions on four or five of the policies, 2.54 for those who took conservative positions on four or five policies).

Evaluations and Independent Variables

After viewing the screenshot, respondents assessed the legislator’s ideology and party affiliation. Respondents were told to “Imagine a scale that measures how liberal or conservative a politician is. 0 would mean that they are extremely liberal, 100 would mean that they are extremely conservative. Where on this scale would you put Congressman [last name]?” and shown a sliding scale on which to record their answer. MC’s ideology thus ranges from 0 to 100. Respondents were then asked whether the Congressman was a Democrat, a Republican, an Independent, or whether they didn’t know. In the models predicting perceptions of the MC’s party, “Republican” is the excluded category. Job approval was measured with: “Although Congressman [last name] is not your current Representative, do you approve or disapprove of the job he is doing as a Congressman?” This ordered categorical variable has response options of: strongly disapprove (the excluded category), somewhat disapprove, neither approve nor disapprove, somewhat approve, strongly approve.

I code a number of independent variables. The MC’s race and the respondent’s race are coded as a series of dummy variables: “Black,” “Hispanic,” and “White” (which serves as the omitted category in the models). The respondent’s party is captured by a series of dummy variables: “Democrat” (the omitted category), “Republican,” or “Independent.” The MC’s conservativism is the proportion of the five policy positions on which the MC took a conservative stance. The variable ranges from 0 (taking no conservative positions) to 1 (taking a conservative position on all of the issues). I create a measure of the respondent’s conservativism in an identical way based on their survey responses. Both of these conservatism measures are centered around their sample means, so that the estimates for the race dummies in the models that interact race and conservatism can be interpreted as the effect of race when conservatism is at its mean. Descriptive statistics for all of these variables are available in Table A-2 in the supplementary materials online.

Race and Perceptions of Ideology

I begin by modeling perceptions of the MC’s ideology. The dependent variable ranges from 0 (liberal) to 100 (conservative), and I fit an OLS regression model that includes the MC’s race and his conservatism as independent variables. To assess H3—whether any differences in perceptions of White and non-White politicians are moderated by the ideological record of the legislator—I also fit a model that includes an interaction term between the MC’s race and his policy positions. As discussed earlier, in both models I control for the respondent’s race.⁵ Table 1 presents the results of these models: In each case, a negative coefficient for a non-White MC would indicate he was seen as more liberal than the White MC, a positive coefficient, that he was seen as more conservative.

Table 1.

OLS Regression Models Predicting Constituents’ Perceptions of MC’s Ideology.

	Model 1(a)	Model 1(b)
Intercept	53.76 (1.19)***	53.75 (1.19)***
Black MC	−2.76 (1.30)*	−2.76 (1.30)*
Hispanic MC	−2.44 (1.32)^†	−2.44 (1.32)^†
MC conservatism	24.47 (2.10)***	24.59 (3.58)***
Black MC × MC conservatism		−0.86 (5.06)
Hispanic MC × MC conservatism		0.56 (5.19)
Black respondent	0.43 (1.31)	0.43 (1.31)
Hispanic respondent	−0.65 (1.31)	−0.64 (1.31)
N	1,710	1,710
R ²	.07	.07

Note. White MC and White respondent are excluded racial categories. Conservativism of MC centered around sample mean. Significance levels are based on two-tailed tests. MC = Member of Congress.

†

p < .1. *p < .05. **p < .01. ***p < .001.

The regression coefficients demonstrate that respondents perceived the non-White MCs as significantly more liberal than the White MC, even given information about their legislative record. Take Model 1(a) that does not include any interaction term between the MC’s race and his conservatism. The coefficient estimates suggest that a Black MC is seen as around 2.76 percentage points more liberal than a White MC, holding their conservatism constant (SE = 1.30, p = .03). Likewise, a Hispanic MC was seen as around 2.44 points more liberal than their White counterpart (SE = 1.32, p = .06).

The positions that the MC took on his website had a substantial effect on voters’ perceptions, as we would expect. A MC who took conservative positions on all five policies was perceived to be around 25 percentage points more conservative than a MC who took liberal positions on all five (β = 24.47, SE = 2.10, p = .00). Respondents’ perceptions of the MC’s ideology were clearly influenced by the information they received about his policy positions.

Did this policy record moderate the extent to which voters perceived the non-White MCs as more liberal than the White MC? The interaction terms in Model 1(b) mean that the coefficient for a Black or Hispanic MC should be read as the effect of his race when his conservatism is set to zero. As this variable is centered around its mean, the estimated effect is for a MC who took an average number of conservative positions (i.e., on 50% of the policy areas). To assess whether his policy record moderated the differences in perceived ideology based on race, I estimated the marginal effect of the MC’s race across the full range of MC conservatism, as recommended by Brambor, Clark, and Golder (2006) and Berry, Golder, and Milton (2012). Following the simulation methods in King, Tomz, and Wittenberg (2000), Figure 1 shows the first differences in the perceived ideology of Black and White MCs, in plot (a), and Hispanic and White MCs, in plot (b). The differences in perceived ideology are shown as solid lines, with confidence intervals as dotted lines.

Figure 1.

First differences in perceived ideology of non-White and White MCs, by the conservatism of their positions.

The marginal effects plotted in Figure 1 do not suggest any trend in the perceived differences between White and non-White MCs’ ideology across the range of MC conservatism. H3 would expect to see large perceived differences between liberal MCs of different races and smaller differences between conservative MCs of different races. The estimated first differences, however, suggest no such trend. Instead, the marginal effect of the MC’s race appears essentially constant across the range of his conservatism.

The large confidence intervals around these estimates—particularly at either end of the ideological scale, where the effects of the MC’s race cannot be distinguished from zero—mean that this should not be taken as disconfirming evidence for H3. It is possible that such a moderating relationship exists but is undetected in these data. What we can say, though, is that there is a distinct lack of evidence that the ideological record of the MC moderated voters’ use of race as a cue to his ideology. When the MCs took liberal positions on four of the five policy areas, respondents perceived the Black MC to be 2.48 [−0.95, 5.73] points, and the Hispanic MC to be 2.68 [−0.68, 6.20] points, more liberal than the White MC. When the MCs took conservative positions on four of the five policies, these estimates are indistinguishable: the Black MC is perceived to be 3.03 [−0.24, 6.39] points more liberal than the White MC, the Hispanic MC 2.26 [−1.05, 5.51] points more liberal. Overall, in other words, there is no evidence that the particular positions the MC took on his site moderated the extent to which respondents perceived non-White MCs to be more liberal than White MCs.

To summarize, these results show that voters perceive non-White MCs as more liberal than their White counterparts who take the same policy positions. Even in a relatively “high information” environment, where voters were given explicit information about legislators’ positions, voters categorized on the basis of race and attributed a more liberal ideology to non-White politicians. The ideological skew of the MC’s positions does not appear to significantly moderate the extent to which voters saw non-White MCs as more liberal than White MCs. Note that this does not mean that voters perceived conservative non-White MCs as liberals: No matter their race, MCs who took more conservative positions were perceived as more conservative. Rather, the results show that non-White MCs were seen as more liberal than White MCs who took the same positions. Conservative non-White MCs were seen as more conservative than liberal non-White MCs, in other words, but were still seen as more liberal than a White MC with an equally conservative record.

Race and Perceptions of Party Affiliation

I fit a series of multinomial logistic regressions to predict perceptions of the MC’s party. The models use the same independent variables as before. As the excluded category for the dependent variable perceives the MC to be a Republican, each block of coefficients should be read as the effect on perceiving the MC as a Democrat, as an Independent, or not knowing, compared with perceiving him as a Republican. In the first block of coefficients in Table 2, for example, a positive coefficient would indicate that respondents were more likely to see him as a Democrat than a Republican.

Table 2.

Multinomial Logistic Regression Models Predicting Constituents’ Perceptions of MC’s Party.

	Model 2(a)	Model 2(b)
Democrat vs. Republican
Intercept	−0.66 (0.18)***	−0.67 (0.19)***
Black MC	1.09 (0.20)***	1.05 (0.22)***
Hispanic MC	0.63 (0.20)**	0.68 (0.22)**
MC conservatism	−4.22 (0.35)***	−4.38 (0.62)***
Black MC × MC conservatism		0.28 (0.85)
Hispanic MC × MC conservatism		0.28 (0.88)
Black respondent	−0.30 (0.20)	−0.30 (0.20)
Hispanic respondent	−0.05 (0.20)	−0.05 (0.20)
Independent vs. Republican
Intercept	−2.31 (0.34)	−2.34 (0.35)***
Black MC	1.23 (0.38)	1.24 (0.39)**
Hispanic MC	1.40 (0.36)	1.45 (0.38)***
MC conservatism	−2.33 (0.54)	−0.81 (1.20)
Black MC × MC conservatism		−1.55 (1.51)
Hispanic MC × MC conservatism		−2.07 (1.47)
Black respondent	−0.18 (0.30)	−0.18 (0.30)
Hispanic respondent	−0.42 (0.33)	−0.43 (0.33)
Don’t know vs. Republican
Intercept	0.84 (0.13)***	0.85 (0.13)***
Black MC	0.49 (0.15)**	0.45 (0.16)**
Hispanic MC	0.33 (0.15)*	0.34 (0.16)*
MC conservatism	−2.13 (0.26)***	−2.26 (0.40)***
Black MC × MC conservatism		0.48 (0.63)
Hispanic MC × MC conservatism		−0.05 (0.63)
Black respondent	0.06 (0.15)	0.06 (0.15)
Hispanic respondent	0.19 (0.15)	0.19 (0.15)
N	1,829	1,038
Log-likelihood	−1840.8	−1838.9

Note. White MC and White respondent are excluded racial categories. Republican is excluded category for dependent variable, respondent’s perception of MC’s party. Conservativism of MC centered around sample mean. Significance levels are based on two-tailed tests. MC = Member of Congress.

†

p < .1. *p < .05. **p < .01. ***p < .001.

I begin by focusing on Model 2(a), which does not include the interaction term between a MC’s race and his conservatism. The results indicate the MC’s race had a significant impact on respondents’ perceptions of his party. The Black MC was more likely to be perceived as a Democrat than a White MC (β = 1.09, SE = .20, p = .00). Likewise, respondents were more likely to identify the Hispanic MC as a Democrat than the White MC (β = .63, SE = .20, p = .00). The difference between these coefficients is weakly significant (p = .08), suggesting that Black MCs are somewhat more likely to be seen as Democrats than Hispanic MCs. However, Black and Hispanic MCs are significantly more likely to be perceived as Democrats than White MCs.

Assessing the substantive impact of these coefficients is not straightforward, given the multinomial logit model. I generate predicted probabilities of a respondent perceiving the MC as a Democrat, Republican, Independent, or responding that they didn’t know from the coefficients in Model 2(a). Table 3 shows the predicted probabilities for each response, by the race of the MC.

Table 3.

Predicted Probabilities of Perceived Party Affiliation of MC, by MC’s Race.

	Black MC	Hispanic MC	White MC
Democrat	.23 [.20, .27]	.17 [.14, .21]	.13 [.11, .17]
Republican	.15 [.12, .18]	.18 [.15, .21]	.25 [.21, .29]
Independent	.05 [.03, .08]	.07 [.05, .10]	.03 [.02, .04]
Don’t know	.56 [.52, .61]	.57 [.53, .62]	.59 [.55, .63]

Note. Probabilities predicted from Model 2(a) in Table 2. MC conservatism set to its mean, race of respondent set to be White. 90% confidence intervals shown in brackets. Probabilities of offering each response for a particular MC may not sum to 1 due to rounding. MC = Member of Congress.

Unsurprisingly, given the experimental stimulus did not offer any information about his party affiliation, the modal response is to answer “don’t know” when asked about his party. The predicted probability of a respondent offering a “don’t know” response does not vary with the MC’s race. Of those who did offer an answer to the question, however, the MC’s race had a significant effect on perceptions of his party.

Take the Black MC, for example. The probability of a respondent perceiving him as a Democrat was .23 [.20, .27]—compared with a probability of perceiving him as a Republican of .15 [.12, .18], or a probability of perceiving the White MC as a Democrat of .13 [.11, .17]. Respondents were more likely to perceive the Black MC as a Democrat than as a Republican, and more likely to perceive him as a Democrat than they were a White MC. As we would expect, these results are mirrored by the probabilities of seeing the MC as a Republican: respondents were significantly less likely to see the Black MC as a member of the GOP, .15 [.12, .18], than they were a White MC, .25 [.21, .29].

The stereotype of Hispanics as Democrats does not appear to have influenced respondents in the same way. The probability of a respondent perceiving the Hispanic MC as a Democrat, .17 [.14, .21], is indistinguishable from the probability of perceiving him as a Republican, .18 [.15, .21], or from the probability of perceiving the White MC as a Democrat, .13 [.11, .17]. There is some indication that respondents were less likely to think the Hispanic MC was a Republican than they were to think the White MC was. The probability of respondents seeing the Hispanic MC as a Republican is .18 [.15, .21], of seeing the White MC as a Republican, .25 [.21, .29]. While respondents do not infer that a Hispanic MC is more likely to be a Democrat than a White MC is, they do appear to perceive a Hispanic MC as less likely to be a Republican.

Did the policy positions that the MC took moderate the effect of the MC’s race on perceptions of his party? Model 2(b) in Table 2 includes the interaction term between the MC’s race and his stances. To assess these interactions, Figure 2 presents four plots of these probabilities. In each plot, lines show the predicted probability of seeing a Black, Hispanic, or White MC as affiliated with each party, across the range of MC conservatism. These plots allow us to compare how equally liberal, moderate, or conservative MCs of different races were perceived by respondents. To clearly present the results, Figure 2 does not include confidence intervals, but I discuss them in the text here.

Figure 2.

Predicted probabilities of perceived party affiliation of non-White and White MCs, by the conservatism of their positions.

First, note that the ideological slant of the positions the MC took again had a strong effect on voters’ perceptions. In plot (a), the probability of seeing the MC as a Democrat decreases dramatically for all races the more conservative his positions. For example, the probability of a White MC being perceived as a Democrat shifts from .25 [.20, .31] to .06 [.04, .09] when they shift from taking conservative positions on one to four of the five policies. Correspondingly, the more conservative positions the MC took, the more likely respondents were to perceive them as Republicans: For a White MC, the same shift from one to four conservative positions leads to a shift from .13 [.10, .17] to .41 [.35, .47] in the probability of being perceived as a Republican.

Given a particular set of positions, however, voters perceived MCs of different races to have different partisan affiliations. When the MC took liberal positions on four of five policies, the Black MC was more likely to be seen as a Democrat than an equally liberal White MC (probabilities of .39 [.33, .45] and .25 [.20, .31] respectively). When they took conservative positions on four of five policies, both MCs were less likely to be seen as Democrats—but the Black MC was still more likely to be seen as a Democrat than the White MC (.12 [.08, .16] vs. .06 [.04, .08]).⁶ The extent of these differences in perceptions of Black and White MCs is not diminished by the ideological slant of their positions: whether the MC took liberal positions or conservative ones, the Black MC was more likely to be seen as a Democrat, and less likely to be seen as a Republican, than a White MC with an identical record.

Partisan stereotypes of Hispanic legislators appear weaker than those of Black legislators. As Figure 2 shows, Hispanics are perceived to be between Black and White MCs—more likely to be a Democrat than a White MC is, but not as likely to be a Democrat as a Black MC. Taking overlapping confidence intervals into account, these differences are not statistically distinguishable. For example, when the MC took mostly liberal positions, the probability of being seen as a Democrat was .29 [.24, .35] for the Hispanic MC and .25 [.20, .31] for the White MC. Similarly, when the MC took mostly conservative positions, the probability of being seen as a Republican was .32 [.27, .38] for the Hispanic MC and .42 [.36, .47] for the White MC. Perceptions of the Hispanic MC’s party affiliation are not distinct from perceptions of the White MC’s, across the range of MC conservatism. This reinforces the conclusion from Table 3 that respondents are no more likely to associate Hispanic legislators with the Democratic Party than they are White legislators, in contrast with perceptions of Black legislators. It also suggests that this association is not conditional on the ideological slant of the MC’s record: Whether liberal or conservative, the Hispanic MC was not perceived to be from a different party to the White MC.

There is some evidence that the positions the MC took affected how he was perceived by respondents. Plot (d) shows the probability of respondents answering that they “didn’t know” the MC’s positions. For MCs of all races, the relationship between MC conservatism and a DK response is curvilinear. Overlapping confidence intervals do not allow for definitive conclusions to be made, but the results suggest that when MCs took positions that ran counter to their group stereotype, respondents were more likely to say they did not know his party. When the Black MC took mostly liberal positions (i.e., fit the stereotype of Black politicians as liberals), the probability of a DK response is .48 [.42, .54]. When he took mostly counter-stereotypical, conservative, positions, that probability increases to .57 [.51, .63]. The same is true for the White MC: when he took stereotypical (conservative) positions, the probability of a DK response was .50 [.44, .55]; when he took counter-stereotypical (liberal) positions, the probability rose to .60 [.54, .66]. Again, I note that the confidence intervals around these predicted probabilities overlap and thus preclude any categorical conclusion. However, the point estimates suggest that respondents were more likely to be unsure of the MC’s party when he took positions at odds with the stereotype of his racial group. In this modest way, the ideological content of the MC’s record may have moderated the extent to which respondents used his race as a guide to answering the question about his party affiliation.

In summary, the results here show that voters perceived Black politicians as more Democratic than White politicians who take the same positions. In contrast, respondents did not infer that the Hispanic MC was more likely to be a Democrat than the White MC, suggesting that the partisan stereotype associated with Hispanic MCs is less fully formed than that associated with Black MCs. While the policy positions the MC took had a strong effect on respondents’ perceptions (in that conservative politicians of all races were more likely to be seen as Republicans than liberal politicians of all races), they do not appear to moderate the use of stereotypes to infer the party affiliation of Black MCs. Whether the politician took the expected positions for his race or not, voters perceived the Black MC as a Democrat more frequently than they did a White MC. Before concluding, I examine the impact these stereotypes have on voters’ overall evaluations of a politician’s performance.

The Consequences of Stereotypes

To assess whether legislators of different races are rewarded or punished based on these stereotypes, I fit ordered logistic regression models that predict the MC’s job approval rating. I follow the same empirical strategy used in “indirect” studies of stereotyping described earlier (Colleau et al., 1990; M. L. McDermott, 1998), and interact the race of the MC with the respondent’s ideology, in Model 3(a), and party identification, in Model 3(b). If voters use partisan and ideological stereotypes to evaluate MCs, then we would expect conservatives and Republicans to approve less of non-White MCs than of White MCs (i.e., a negative interaction between a voter’s conservatism or Republican identity and a non-White MC). To isolate the effects of race, I control for policy congruence between the voter and the MC. This is modeled on the measure used in Jones (2011), and is the proportion of policies on which the MC took the same position as the voter. The results are shown in Table 4.

Table 4.

Ordered Logistic Regression Models Predicting Approval of MC.

	Model 3(a) Ideology	Model 3(b) Party ID
Black MC	−0.02 (0.11)	0.15 (0.13)
Hispanic MC	−0.03 (0.11)	0.08 (0.13)
Conservative respondent	0.40 (0.28)
× Black MC	−0.99 (0.37)**
× Hispanic MC	−0.73 (0.37)*
Independent respondent		0.28 (0.45)
× Black MC		0.24 (0.68)
× Hispanic MC		−0.56 (0.70)
Republican respondent		0.50 (0.17)**
× Black MC		−0.54 (0.23)*
× Hispanic MC		−0.31 (0.23)
Policy congruence	2.46 (0.16)***	2.45 (0.16)***
Black respondent	0.12 (0.12)	0.27 (0.12)*
Hispanic respondent	0.08 (0.11)	0.16 (0.11)
Threshold 1	−2.32 (0.12)***	−2.07 (0.14)***
Threshold 2	−1.31 (0.11)***	−1.07 (0.13)***
Threshold 3	1.07 (0.10)***	1.31 (0.13)***
Threshold 4	3.31 (0.14)***	3.56 (0.16)***
N	1,843	1,843
Log-likelihood	−2320.3	−2319.14

Note. White MC and White respondents are excluded racial categories. Conservativism of respondent and policy congruence are centered around sample mean. Democrat is excluded category for voter’s party ID. Significance levels are based on two-tailed tests. MC = Member of Congress.

†

p < .1. *p < .05. **p < .01. ***p < .001.

I begin by focusing on Model 3(a) that interacts the respondent’s ideology with the MC’s race. The negative coefficients for the interaction between a voter’s conservatism and a non-White MC indicate that more conservative voters approved less of non-White MCs than White MCs, all else equal. The more conservative a voter, the less likely they were to approve of a Black or Hispanic MC than they were of a White MC (β = −.99, SE = .37, p = .01 and β = −.73, SE = .37, p = .04, respectively).

The substantive impact of these effects is shown in the first plot of Figure 3. I simulate the coefficients from Model 3(a) to estimate differences in the probability that respondents with different ideological outlooks approve of MCs of different races. These first differences show the impact of race on approval ratings for different sets of voters. Liberal voters (those who took a liberal position on every policy) respond more favorably to a Black or Hispanic MC than they do a to White MC (an increase in the probability of approving of .07 [.02, .13] and .04 [−.01, .09], respectively). Conservative voters, in turn, are less likely to approve of Black, −.11 [−.20, −.03], or Hispanic, −.07 [−.16, .02], MCs than White MCs, although these estimates for Hispanic MCs have confidence intervals that include zero. The stereotype of Black and Hispanic politicians as liberals leads conservatives to approve less, and liberals to approve more, than they would of otherwise equivalent White MCs.

Figure 3.

Difference in approval ratings of White and non-White MCs, by respondent’s ideology (left) and party ID (right).

The models that interact the race of the MC with the respondent’s party identity show a similar pattern of effects. The excluded category for party ID is a “Democrat.” As the focus here is on party cues, I compare Democratic and Republican respondents, and set aside the results for Independents. The negative coefficient in Model 3(d) for the interaction of a Republican ID with a Black MC (β = −.54, SE = .23, p = .02) indicates that Republican respondents approved of Black MCs at lower rates than they did of White MCs, compared with Democratic respondents. The coefficient for the Hispanic MC interaction does not reach standard levels of significance (β = −.31, SE = .23, p = .18), but suggests a similar negative relationship between a Republican identity and approval of a Hispanic MC.

The first differences for these estimates are shown in the right plot of Figure 3. The probability that a Democrat approves of a MC increases by .04 [.00, .08] when the MC is Black rather than White. Republicans are less likely to approve of Black MCs than White MCs: the estimated change in probability is −.07 [−.13, −.01]. As we would expect from the regression coefficients, the first differences between White and Hispanic MCs are not significant statistically but suggest a similar relationship: Democrats are more likely to approve of a Hispanic MC than a White MC, .01 [−.03, .05], Republicans less likely, −.01 [−.08, .05].

Generalizations about non-White MCs’ ideological and partisan orientations have significant consequences beyond simple perceptions of a particular non-White individual. They go on to shape approval ratings of the legislators in predictable ways. Note that this is not necessarily bad news for minority legislators: They are likely to receive higher approval ratings from liberal Democrats than a White legislator with the same record would, even as they receive lower ratings from conservative Republicans than their White counterpart would. Regardless, the stereotypes that voters hold of legislators’ ideology and party ultimately shape their evaluations of the politician’s job.

Discussion and Conclusion

Whether voters are able to effectively hold their representatives accountable for their record in office depends critically on whether citizens have accurate perceptions of what has been done in their names. Previous work has focused on the limited interest and knowledge that many citizens have as an impediment to democratic accountability. Here, the results show that even when given clear and specific information about a politician’s record, voters categorize them by race and infer ideology and party affiliation based on generalizations about racial groups. While “standard” theories of accountability expect voters to hold unbiased views of their representatives’ record (e.g., Ansolabehere & Jones, 2010), this study shows that these perceptions are significantly skewed by partisan and ideological stereotypes of non-White politicians.

The experiment used here shows that voters followed the “cues” of race to infer that Black and (to a somewhat lesser extent) Hispanic legislators were more liberal and more Democratic than their White counterparts. These perceptual biases were not moderated by the particular set of positions the politicians took. Whether a non-White MC took stereotypical (i.e., liberal) or counter-stereotypical (i.e., conservative) positions, respondents perceived them as more liberal and Democratic than a White MC who took the same stances. The extent to which respondents perceived non-White MCs as more liberal and Democratic than White MCs, in other words, was not diminished by information indicating that the individual politician did not fit the category stereotype.

This stereotyping has significant consequences for broader evaluations of legislators’ jobs. Liberals and Democrats [conservatives and Republicans] were more [less] likely to approve of non-White politician than a White politician. This is controlling for the actual degree of policy congruence between them. Consistent with previous work, this suggests that voters used the stereotype of Black and Hispanic politicians as liberal Democrats to guide their evaluations of their performance (M. L. McDermott, 1998). Beyond affecting “just” perceptions of these legislators’ records, ideological and partisan stereotypes spill over and shape overall job approval ratings.

Although early work on this type of cue-taking emphasized that it was most likely to be prevalent in “low-information” contexts (M. L. McDermott, 1997, 1998), these results show that they continue to be used in relatively “high” information contexts, where information about the individual’s policy positions is available. The less consistent effects for Hispanic politicians suggest that these stereotypes take time to develop. If Hispanic politicians become as closely associated with the Democratic Party as Black politicians have been since the civil rights movement, then it is certainly plausible that this stereotype would strengthen in voters’ minds.

Throughout this article, the analyses have relied on a simple scale of the politician’s five policy positions. Recall, however, that these policy areas differed significantly, with some concerning more racialized issues (immigration reform and racial profiling) in particular. One line of inquiry that I examined but did not have space to include here was whether perceptions of non-White MCs were moderated most strongly by their positions on racialized issues. I assessed this by re-fitting the models from Tables 1 and 2, substituting each policy area individually in place of the scale of positions. The results—shown in the online supplementary materials for interested readers—do not suggest that the effects of an MC’s race on perceptions varied substantially across different policy areas. On each policy, Black and Hispanic MCs were consistently seen as more liberal and Democratic than White MCs—and the size of differences between MCs of different races did not vary significantly by the policy area under study. In other words, the perceptions of Black and Hispanic MCs as more liberal or more Democratic than White MCs do not appear to be driven primarily by racialized policy areas. Future work could expand on this line of inquiry by examining more than the limited number of issues available here.

By manipulating the race and policy positions of the politician in this experiment, I am able to more accurately estimate the causal effects of race on voters’ perceptions. At the same time, and as with any experimental study, there are important limitations to the research design. Although the MC’s website, policy agenda, and explanation of his positions were carefully based on those of actual MCs, in the real world voters rarely (if ever) are asked to evaluate legislators immediately after learning about their race and policy positions for the first time. Indeed, research suggests that voters may update their perceptions of non-White incumbents and eschew stereotypes as they learn more about the politician over time (Hajnal, 2001). Exploring over-time effects would be difficult using an experiment with a fictitious MC, although future research could perhaps replicate this study and incorporate multiple survey waves that reveal new information about the MC each time.

The experiment is also limited in its exploration of the types of stereotypes that voters use, in two ways. First, it manipulates only the race of the MC, leaving his gender, age, and affective appearance consistent across all conditions. Future work could explore the effects of all of these stereotypes—and those associated with intersecting social identities such as gender and race (see, e.g., Philpot & Walton, 2007)—using a similar research design. Second, it explores only “belief” stereotypes, the beliefs associated with particular categories of politicians. It does not examine “trait” stereotypes, the competencies, and personality traits associated with groups of politicians (Huddy & Terkildsen, 1993). Previous work suggests that voters stereotype non-White politicians as more concerned with minority rights, and compassionate, but less competent than White politicians (M. L. McDermott, 1998; Sigelman et al., 1995;Williams, 1990). Re-examining these types of stereotypes may be a fruitful next step for researchers in this field.

Early work on information shortcuts raised the possibility that voters could use generalizations about groups of people to make efficient inferences about specific individuals they were evaluating (M. L. McDermott, 1998; Popkin, 1991). In the absence of any other information, it may be rational for voters to assume that non-White politicians are liberals and Democrats—since the majority of elected minorities are. In the presence of individuating information about a politician’s policy positions, the rationality of relying on stereotypes to make inferences is less defensible. Even when shown specific information about the stances a politician took—and even when those positions were overwhelmingly conservative ones—voters continued to identify Black and Hispanic legislators as more liberal and more likely to be Democrats than otherwise identical White legislators.

To the extent that voters choose candidates for office on the basis of their ideology or partisan affiliation, these findings have mixed implications for the electoral fortunes of minorities. Black and Hispanic legislators were consistently seen as more liberal and more Democratic than White legislators who took identical positions. For legislators with a liberal record, this has the effect of making them seem more ideologically extreme and further from moderate voters. For legislators with a conservative record, however, this has the effect of making them seem more centrist and closer to the median voter (see Koch (2002) for similar conclusions about gender stereotypes). In other words, a Black conservative may win more votes from a moderate electorate than an equally conservative White politician. In contrast, a Black liberal is likely to win fewer votes from a moderate electorate than an equally liberal White politician.

Ultimately, revisiting these stereotypes demonstrates the potential for such demographic cues to shape citizens’ perceptions of their leaders’ actions. Far from a world in which voters ignore race when evaluating politicians, or one in which all citizens engage in a data-gathering process based on policy positions, citizens are apt to categorize incumbents by race and infer the rest. Even in “high information” conditions, where voters are presented with specific details about a politician’s stances, the stereotypes of non-White politicians as liberals and Democrats can distort perceptions of what they have done in office, and skew their approval ratings in significant ways.

Footnotes

Acknowledgements

I am grateful to Jason Barabas, Brian Gaines, and the anonymous reviewers for extremely helpful comments on earlier versions of the article, and to Kevin Derrick for technical assistance with the politician images. All errors remain my own.

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) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: This research was made possible thanks to financial support from the Conrad N. Hilton Foundation, the UNIDEL Foundation, and the University of Delaware’s Center for Political Communication. All errors remain my own.

Notes

Author Biography

Philip Edward Jones is an assistant professor of Political Science and International Relations at the University of Delaware. His research interests include public opinion and political behavior, and in particular how constituents perceive and respond to their elected representatives.

References

Ansolabehere

Jones

P. E.

(2010). Constituents’ responses to congressional roll call voting. American Journal of Political Science, 54, 583-597.

Arceneaux

(2008). Can partisan cues diminish democratic accountability? Political Behavior, 30, 139-160.

Bartholow

B. D.

Dickter

C. L.

(2008). A response conflict account of the effects of stereotypes on racial categorization. Social Cognition, 26, 314-332.

Bastedo

R. W.

Lodge

(1980). The meaning of party labels. Political Behavior, 2, 287-308.

Berry

W. D.

Golder

Milton

(2012). Improving tests of theories positing interaction. The Journal of Politics,74, 653-671.

Bertrand

Mullainathan

(2004). Are Emily and Greg more employable than Lakisha and Jamal? A field experiment on labor market discrimination. American Economic Review, 94, 991-1013.

Bodenhausen

G. V.

Kang

S. K.

Peery

(2012). Social categorization and the perception of social groups. In Fiske

S. T.

Macrae

C. N.

(Eds.), The Sage handbook of social cognition, 311-329. Thousand Oaks, CA: SAGE.

Brambor

Clark

W. R.

Golder

(2006). Understanding interaction models: Improving empirical analyses. Political Analysis, 14, 63-82.

Campbell

Converse

P. E.

Miller

W. E.

Stokes

D. E.

(1960). The American voter. Chicago, IL: University of Chicago Press.

10.

Campbell

D. E.

Green

J. C.

Layman

G. C.

(2011). The party faithful: Partisan images, candidate religion, and the electoral impact of party identification. American Journal of Political Science, 55, 42-58.

11.

Carmines

E. G.

Stimson

J. A.

(1989). Issue evolution: Race and the transformation of American politics. Princeton, NJ: Princeton University Press.

12.

Colleau

S. M.

Glynn

Lybrand

Merelman

R. M.

Mohan

Wall

J. E.

(1990). Symbolic racism in candidate evaluation: An experiment. Political Behavior, 12, 385-402.

13.

Delli-Carpini

M. X.

Keeter

(1996). What Americans know about politics and why it matters. New Haven, CT: Yale University Press.

14.

Druckman

J. N.

Kam

C. D

. (2011). Students as experimental participants: A defense of the “narrow data base.” In Druckman

J. N.

Green

D. P.

Kuklinski

J. H.

Lupia

(Eds.), Cambridge handbook of experimental political science (pp. 41-57). Cambridge; New York Cambridge University Press.

15.

Fiske

S. T.

Neuberg

S. L.

(1990). A continuum of impression formation, from category-based to individuating processes. In Zanna

M. P.

(Ed.), Advances in experimental social psychology, 1-74. San Diego, CA: Academic Press.

16.

Fiske

S. T.

Taylor

S. E.

(1991). Social cognition (2nd ed.). New York, NY: McGraw-Hill.

17.

Gay

(2002). Spirals of trust? The effect of descriptive representation on the relationship between citizens and their government. American Journal of Political Science, 46, 717-732.

18.

Golebiowska

E. A.

(2001). Group stereotypes and political evaluation. American Politics Research, 29, 535-565.

19.

Green

Palmquist

Schickler

(2002). Partisan hearts and minds: Political parties and the social identities of voters. New Haven, CT: Yale University Press.

20.

Hajnal

Z. L.

(2001). White residents, black incumbents, and a declining racial divide. American Political Science Review, 95, 603-617.

21.

Huddy

Terkildsen

(1993). Gender stereotypes and the perception of male and female candidates. American Journal of Political Science, 37, 119-147.

22.

Jones

P. E.

(2011). Which buck stops here? accountability for policy positions and policy outcomes in congress. Journal of Politics, 73, 764-782.

23.

Judd

C. M.

Downing

J. W.

(1995). Stereotypic accuracy in judgments of the political positions of groups and individuals. In Lodge

McGraw

K. M.

(Eds.), Political judgment: Structure and process, 65-89. Ann Arbor: University of Michigan Press.

24.

King

Tomz

Wittenberg

(2000). Making the most of statistical analyses: Improving interpretation and presentation. American Journal of Political Science, 44, 341-355.

25.

Koch

J. W.

(2000). Do citizens apply gender stereotypes to infer candidates’ ideological orientations? Journal of Politics, 62, 414-429.

26.

Koch

J. W.

(2002). Gender stereotypes and citizens’ impressions of house candidates’ ideological orientations. American Journal of Political Science, 46, 453-462.

27.

McDermott

M. L.

(1997). Voting cues in low-information elections: Candidate gender as a social information variable in contemporary united states elections. American Journal of Political Science, 41, 270-283.

28.

McDermott

M. L.

(1998). Race and gender cues in low-information elections. Political Research Quarterly, 51, 895-918.

29.

McDermott

M. L.

(2005). Candidate occupations and voter information shortcuts. Journal of Politics, 67, 201-219.

30.

McDermott

M. L.

(2009). Religious stereotyping and voter support for evangelical candidates. Political Research Quarterly, 62, 340-354.

31.

McDermott

(2002). Experimental methodology in political science. Political Analysis, 10, 325-342.

32.

Miller

A. H.

Wlezien

Hildreth

(1991). A reference group theory of partisan coalitions. Journal of Politics, 53, 1134-1149.

33.

Philpot

T. S.

Walton

(2007). One of our own: Black female candidates and the voters who support them. American Journal of Political Science, 51, 49-62.

34.

Popkin

S. L.

(1991). The reasoning voter: Communication and persuasion in presidential campaigns. Chicago, IL: University of Chicago Press.

35.

Rahn

W. M.

(1993). The role of partisan stereotypes in information processing about political candidates. American Journal of Political Science, 37, 472-496.

36.

Schneider

M. C.

Bos

A. L.

(2011). An exploration of the content of stereotypes of black politicians. Political Psychology, 32, 205-233.

37.

Sigelman

C. K.

Sigelman

Walkosz

B. J.

Nitz

(1995). Black candidates, white voters: Understanding racial bias in political perceptions. American Journal of Political Science, 39, 243-265.

38.

Terkildsen

(1993). When white voters evaluate black candidates: The processing implications of candidate skin color, prejudice, and self-monitoring. American Journal of Political Science, 37, 1032-1053.

39.

Weaver

(2012). The electoral consequences of skin color: The “hidden” side of race in politics. Political Behavior, 34, 159-192.

40.

Williams

L. F.

(1990). White/black perceptions of the electability of black political candidates. National Political Science Review, 2, 145-164.

41.

Word

D. L.

Coleman

C. D.

Nunziata

Kominski

(2000). Demographic aspects of surnames from Census 2000 (Tech. Rep.). U.S. Bureau of the Census. Retrieved from http://www.census.gov/genealogy/www/data/2000surnames/index.html