Chapter 22 The Determinants and Political Consequences of Prejudice

Vincent L. Hutchings and Spencer Piston
Researchers have been interested in the distribution of prejudice in the population, as well as its effects on policy preferences, vote choice, and economic and social outcomes for racial minorities, since the dawn of the social sciences. One of the earliest scholars to address these questions was W. E. B. DuBois (1899) in his classic work, The Philadelphia Negro. Referring to prejudice in chapter 16 of his book, DuBois wrote, “Everybody speaks of the matter, everyone knows that it exists, but in just what form it shows itself or how influential it is few agree” (322). Although these words were written more than 100 years ago, this observation still does a good job of summarizing our understanding of prejudice. There have been, to be sure, significant advances in this literature since the time of DuBois, but social scientists continue to disagree about the influence of racial prejudice in modern American politics. The aim of this chapter is to explore and adjudicate some of these differences, paying particular attention to the strengths, limitations, and contributions provided by the use of experimental methods.
Before examining the ways that scholars have studied prejudice, it is important that we define this term. What is prejudice? Social psychologists were among the first to answer this question. For example, Allport (1979) defined (ethnic) prejudice as “an antipathy based upon a faulty and inflexible generalization. It may be felt or expressed. It may be directed toward a group as a whole, or toward an individual because he is a member of that group” (9). As we show, subsequent scholars would expand and modify this definition in significant ways. However, the importance of faulty generalizations or stereotypes has remained a central component of virtually all definitions that would follow Allport. We therefore rely on this broad definition unless otherwise indicated.

1. The Study of Prejudice in the Political Science Literature

The political science literature on prejudice1 has focused primarily on the role that prejudice plays in structuring policy preferences and whether it influences candidate preferences. The latter has centered on contests involving two white candidates, as well as elections between a white candidate and an African American candidate. Each area of study has focused almost exclusively on white attitudes and has sought to determine whether prejudice remains a dominant, or at least significant, predictor of preferences in the post–Civil Rights era.2 As we discuss in the next section, although observational work on each area has produced a number of contributions, it has often failed to isolate the precise role that prejudice plays in public opinion, as well as the circumstances in which its influence is more or less powerful. In this chapter, we highlight some of the ways in which experiments have helped address these questions.
Racial Policy Preferences
Much of the early work on prejudice in the political science literature relied on observational studies. Sears and Kinder (1971), for example, used cross-sectional survey data to explore the impact of prejudicial attitudes on policy preferences and candidate support in Los Angeles. They would go on to develop the theory of symbolic racism, which maintains that a new and subtler form of racism emerged in the aftermath of the Civil Rights movement, spurred in part by the urban riots of the late 1960s. Sears and Kinder argue that symbolic racism, unlike previous manifestations of antiblack prejudice, did not posit the biological inferiority of African Americans. Rather, the latent antipathy that many whites still felt toward blacks was now combined with the belief that blacks did not try hard enough to get ahead and violated traditional American values such as hard work and individualism (Kinder and Sears 1981; Sears 1988).3 To test this theory, Sears and Kinder (1971) relied on an attitudinal scale composed of several different survey items. Respondents are asked to what extent they agree with the following statements, among others: 1) “Irish, Italian, Jewish and many other minorities overcame prejudice and worked their way up. Blacks should do the same”; and 2) “It's really a matter of some people not trying hard enough; if blacks would only try harder they could be just as well off as whites.” In general, the symbolic racism scale has been shown to powerfully and consistently predict candidate support and racial policy preferences.4
Although the symbolic racism scale has frequently emerged as the most powerful correlate of racial policy preferences, a number of critics have challenged this construct (Sniderman and Tetlock 1986; Sniderman and Piazza 1993; Sidanius and Pratto 1999). Their critiques take a variety of forms; however, for our purposes, the most relevant questions involve conceptual and measurement issues. Sniderman and his coauthors offer an alternative, although perhaps not wholly incompatible, view of the role of race in modern American politics. This perspective focuses on the institutional or political forces that structure these views. In other words, they argue that the ways in which politicians frame racial issues determine how most Americans express their racial policy preferences.5 Thus, although it is true that Americans disagree on racial policy questions, this disagreement owes more to partisan or ideological differences than it does to racial attitudes per se (Sniderman and Carmines 1997).
To support this alternative view, Sniderman and his colleagues rely heavily on question wording experiments embedded in national surveys. This approach has the advantage of strong internal and external validity. In one study, Sniderman and Carmines (1997) develop what they refer to as the “Regardless of Race Experiment.” In this experiment, a random half of their sample is asked about their support for job training programs for blacks after being told that “some people believe that the government in Washington should be responsible for providing job training to [blacks] because of the historic injustices blacks have suffered” (italics added). The second half of the sample is also asked about job training programs, but with the following rationale: “Some people believe that the government in Washington should be responsible for providing job training to [blacks] not because they are black, but because the government ought to help people who are out of work and want to find a job, whether they're white or black.” Sniderman and Carmines expect the second frame, owing to its more universal character, to be much more popular among whites. As expected, they find that support is higher for race-targeted job training programs when they are justified in universal terms rather than strictly racial ones (thirty-four percent vs. twenty-one percent), consistent with their broader argument.
In another experiment, dubbed the “Color-Blind Experiment,” Sniderman and Carmines (1997) test their argument about the appeal of universal programs more directly. With this experiment, white respondents were divided into three groups and asked whether the federal government should seek to improve conditions for “blacks who are born into poverty…because of the continuing legacy of slavery and discrimination” or to “make sure that everyone has an equal opportunity to succeed.” The last group is distinguished from the other two in that, instead of asking about blacks, respondents were asked about efforts to alleviate poverty for “people” because, as in the second condition, it is the government's role to “make sure that everyone has an equal opportunity to succeed.” They find that support for antipoverty programs increases by eighteen percentage points when the policy is described in universal terms (condition three) relative to race-specific, and racially justified, terms (condition one).
Sniderman and Carmines (1997) acknowledge that their finding, by itself, does not rule out the possibility that it could be the result of antiblack attitudes. However, they argue that if prejudice is driving the lower support for policies targeted at blacks, then whites who are less committed to racial equality should be the most likely to embrace programs that do not mention race. Although plausible, this is not what they find. White respondents who are committed to racial equality are much more likely to support government assistance to the poor regardless of how the program is framed. More important, moving from racially targeted and racially justified characterizations to a more universal frame increases support for these programs at similar rates for those scoring high or low on their racial equality scale.
Kinder and Sanders (1996) employ a similar experimental manipulation of question wording in their examination of support for government efforts to assist blacks. In 1988, half of the American National Election Studies (ANES) sample was asked about efforts to improve the social and economic position of blacks (condition 1), whereas the other half was asked about blacks and other minorities (condition 2). As with Sniderman and Carmines (1997), Kinder and Sanders (1996) find that support for this policy increases by about seven percentage points in the latter condition. They get larger effects in the same direction among blacks, although these results fall short of statistical significance. They conclude that “race neutral programs do appear to be more popular among the American public, black and white” (184).
Experiments such as these do provide some conceptual clarity that is often missing when researchers rely primarily on observational studies. Instead of asking the reader to accept a particular interpretation of an attitudinal construct, Sniderman and Carmines (1997) simply manipulate the rationale behind the policy (i.e., group specific or universal) or, in the cases of Sniderman and Carmines as well as Kinder and Sanders (1996), whether a policy refers specifically to African Americans or is applied to some broader group. They typically find diminished support for the racially targeted program, which might be the result of prejudice. However, even when Sniderman and Carmines focus only on racially targeted programs, as in the “Regardless of Race Experiment,” they find that more universal justifications lead to greater support among whites. Similarly, with the “Color-Blind Experiment,” they and Kinder and Sanders find that the role of prejudice in prompting this greater support is minor because universal programs are more popular with liberals, conservatives, and African Americans. Still, even results derived from an experiment can be open to interpretation. To draw an inference that differences across conditions are due to race and only race, the experimental conditions must differ on this dimension and nothing else. When this is not the case, it is impossible to isolate the source of the difference (or lack of difference) across conditions.
For instance, the “Regardless of Race Experiment” purports to show that universal justifications are more compelling than race-specific ones, but, in this case, the appropriate inference is uncertain. The two aforementioned conditions do not just differ in terms of whether the rationale for the policy is group specific. The justifications also differ in terms of their emphasis on the past and in their characterization of the unemployed as eager “to find a job.” They may also differ in terms of their intrinsic persuasiveness in that a reference to “historic injustices” might be too vague and, consequently, less convincing than references to unemployed workers seeking to find a job. In short, we cannot be sure whether justifications framed in universal terms are inherently more persuasive or whether this particular race-specific rationale is simply inferior to the specific universal rationale they employed. A more convincing race-specific rationale might have asked about support for job training programs “because government ought to help blacks who are out of work and want to find a job.” In this example, the only differences across conditions would be the use of the term “black,” and so any difference could only be attributed to the race-specific nature of the justification.
The Kinder and Sanders (1996) experiment is open to a somewhat different criticism. Although the authors conclude that race-neutral programs are more popular, the inclusion of other minority groups in the question (condition 2) arguably does not diminish the role of race and is therefore not race neutral. It is clear that referencing other minority groups increases the popularity of the program, but it does not necessarily follow that the absence of any reference to race would also increase popularity. The discussion of the results also implies that blacks and whites are more supportive of programs to assist blacks and other minorities for the same reasons (i.e., universal programs are more popular). It is possible, for example, as various studies discussed later indicate, that those whites who find broader programs more appealing adopt this view because of their negative attitudes about blacks. African Americans, however, may be motivated by a sense of solidarity with other nonwhite groups and may or may not view truly race-neutral programs more favorably.
Some studies have addressed these concerns about precision in experimental designs. Iyengar and Kinder (1987) provide one example. They were interested in the impact of television news reports on the perceived importance of particular issues. Instead of modifying survey questions, Iyengar and Kinder unobtrusively altered the content of network news accounts on various national issues. In one experiment, subjects viewed one of three stories about an increase in the unemployment rate. The content was constant across conditions except that in the two treatment conditions, the discussion of employment information was followed by an interview with a specific unemployed individual. In one case, this person is white, and in the other, the person is an African American. Given that the only salient difference across the two treatment conditions is the race of the unemployed worker, lower levels of concern with the unemployment problem when the black worker is shown would represent evidence of racial prejudice. This is exactly what the authors find. Moreover, consistent with the prejudice hypothesis, they find that whites who have negative views of blacks are most likely to diminish the importance of unemployment when the worker is an African American. We should note, however, that these results are not entirely unassailable because they achieve only borderline statistical significance, raising concerns about the reliability of these findings. Further discussion of this experiment can be found in Iyengar's chapter in this volume.
Reyna et al. (2006) address some of the limitations identified in earlier experimental work on prejudice. These scholars rely on a question wording experiment in the 1996 General Social Survey. In this within-subjects experiment, respondents were presented with a question about racial preferences with either blacks or women identified as the target group. Reyna and her colleagues find that, overall, respondents were significantly more supportive of this policy when women were the target group rather than African Americans, suggesting that not all group-specific policies are created equal. Moreover, following up on the source of these effects, Reyna and her colleagues report that political conservatism among college-educated respondents was significantly related to greater opposition to preferences for blacks. They report that these effects are mediated by “responsibility stereotypes,” which turn out to be measured by one of the standard items in the symbolic racism index.6 Providing additional confidence in their results, Reyna and her colleagues replicate their basic findings with a convenience sample of white adults from the Chicago area (n = 184).
Increasingly, researchers have begun to focus on the circumstances under which racial prejudice influences the policy positions of whites, rather than whether these attitudes play any role in structuring public opinion. In one of the earlier examples of this approach, Nelson and Kinder (1996) had their subjects view and evaluate several photographs of individuals engaged in routine, exemplary, or scandalous activities. Specifically, eighty-four University of Michigan undergraduates were randomly assigned to one of three conditions, where they either viewed photos of whites engaged in activities such as gardening (the control group) or pictures of blacks engaged in stereotypical activities such as illegal drug use, or counterstereotypical imagery of blacks interacting with their family or in school settings. If, as some have argued, negative attitudes about African Americans are often dormant among whites, then exposure to frames that serve to remind them of these stereotypes might enhance the influence of prejudice. Consistent with this view, Nelson and Kinder find that the relationship between attitudes about blacks and support for racial preferences is significantly stronger for subjects exposed to stereotypical depictions of African Americans.
Although the symbolic racism researchers have primarily relied on observational studies, they have recently called on experimental methods to defend various elements of the theory. One particularly persistent criticism was that symbolic racism theorists had never empirically demonstrated that this new form of racism derives from a blend of antiblack affect and traditional American values. Sears and Henry (2003) sought to address this criticism using a split ballot design in the 1983 ANES Pilot Study. They adapted the six-item individualism scale so that each item referred specifically to either blacks or women. Respondents received only one of these scales, depending on which ballot they were provided, but all participants received the general individualism scale that made no reference to race or gender. If individualism in the abstract is primarily responsible for white opposition to group-specific preferences, then the individualism scale adapted to apply to women should be as strongly linked to opposition to racial preferences as the individualism scale that was modified to apply to blacks. Similarly, the general individualism scale should also share the same predictive properties as the black individualism scale. If, on the other hand, opposition to racial preferences is primarily driven by individualistic principles applied only to blacks, then one of the core assumptions of the symbolic racism theory would be sustained. Consistent with their theory, they find that only the black individualism scale is significantly correlated with opposition to racial preferences.
Although Sears and Henry (2003) resolve one of the outstanding criticisms of the theory of symbolic racism, it is still unclear whether the construct is confounded with political ideology. Feldman and Huddy (2005) provide some support for this contention. These researchers relied on a question wording experiment embedded in a representative sample of white New York State residents (n = 760). Respondents were asked whether they supported college scholarships for high-achieving students. The treatment consisted of manipulating whether the students were described as white, black, poor white, poor black, middle-class white, middle-class black, middle class, or simply poor. The racial group categories alone do not produce any evidence of a double standard, but, once class was introduced, Feldman and Huddy find evidence of racial prejudice. Specifically, respondents were much more supportive of college scholarships for white middle-class students (sixty-four percent) than they were for black middle-class students (forty-five percent). Also, they find that the symbolic racism scale is associated with this racial double standard, but only for self-identified liberals. Conservatives scoring higher on the symbolic racism scale are more likely to oppose the scholarship program for all groups, not just blacks.7 This suggests that the symbolic racism scale acts more like a measure of political ideology among conservatives. These results are not consistent with the theory of symbolic racism, but it is possible that Feldman and Huddy's results might differ if run on a national sample.
Nonracial Policy Preferences
In addition to experiments designed to explore the role of prejudice in shaping attitudes on racial policies, scholars used this tool to examine the influence of prejudice on ostensibly nonracial policies. Most of this work focused on crime or welfare policy. For example, Gilliam and Iyengar (2000) explore whether the race of criminal suspects influenced levels of support for punitive crime policies among viewers of local television news. Unlike most of the experimental work in political science, they include whites and African Americans in their convenience sample (n = 2,331). Gilliam and Iyengar present their subjects with one of three different (modified) newscasts: one featuring a black criminal suspect, one featuring a white suspect, and one in which there is no photograph or verbal description of the suspect. With the exception of the race of the alleged perpetrator, the newscasts are identical in every way. As anticipated, they find that support for punitive crime policies increases by about six percentage points when subjects view the black suspect, but the effects are much weaker and statistically insignificant when the suspect is white or ambiguous. Interestingly, Gilliam and Iyengar find that this only applies to the whites in their study because the effects are either insignificant or run in the “wrong” direction for blacks. This result provides additional support for their claim that some form of racial prejudice contributes to white support for punitive crime policies. In light of these findings, experimentalists assessing the role of prejudice in shaping policy preferences should include minority subjects whenever possible (see Chong and Junn's chapter in this volume).
Peffley and Hurwitz (2007) engage in a similar analysis, although they focus on the issue of support or opposition to the death penalty. Specifically, they are interested in whether exposure to various arguments against the death penalty would reduce support among blacks and whites. In the first of their two treatment groups, drawn from a national telephone sample, Peffley and Hurwitz present their respondents with the following preamble before asking their views on the death penalty: “some people say that the death penalty is unfair because too many innocent people are being executed.” Their second treatment group is presented with a different introduction: “some people say that the death penalty is unfair because most people who are executed are African Americans.” Finally, respondents in the baseline condition are simply asked their views on this policy without any accompanying frame. They report that both frames are persuasive among African Americans. In the innocence condition, support for the death penalty drops by about sixteen percentage points and in the racial condition it drops by twelve percentage points. Whites, in contrast, are entirely unaffected by the innocence treatment, but, surprisingly, support for the death penalty increases by twelve percentage points in the racial condition. Peffley and Hurwitz attribute this to a priming effect, wherein individual attributions of black criminality become much more predictive of support for the death penalty in the racial condition than in either of the other conditions.
The finding that white support for the death penalty increases when respondents are informed that blacks are more likely to be on death row is striking, but we should interpret this result with some caution. For starters, unlike many others, Peffley and Hurwitz (2007) do not find that antiblack stereotypes contribute to white support for the death penalty. In addition, their experiment was designed to examine reactions to particular antideath penalty appeals, rather than to isolate the specific role of antiblack prejudice in shaping death penalty views. If this latter aim were their goal, then perhaps they would have designed conditions noting that “most people who are executed are men” or “most of the people executed are poor.” Both statements are true, and if support for the death penalty increased among whites in the race condition but not the others, then this would represent strong evidence of a racial double standard.
The other prominent, and ostensibly nonracial, policy domain that may be influenced by antiblack prejudice is welfare policy. Gilens (1996) examined this issue using a question wording experiment embedded in a 1991 national survey. In addition, these results were supplemented with a mail-in questionnaire delivered to the respondents who completed the telephone survey. The respondents in this study were assigned to one of two conditions: in the first condition, they were asked their impressions of a hypothetical welfare recipient characterized as a thirty-something black woman with a ten-year-old child who began receiving welfare in the past year. In the second condition, these attributes are identical except that the woman in question is described as white. Gilens finds that the white and African American welfare recipients are evaluated similarly; however, in the black condition, negative attitudes about welfare mothers are much more likely to influence views of welfare policy. Gilens concludes that whites often have such negative attitudes about welfare because their prototypical recipient is a black woman rather than a white woman.
Experiments and Prejudice beyond the Black–White Divide in the United States
Although most of the experimental work in political science on the subject of prejudice has focused on white attitudes about blacks and related policies, some recent work has also begun to focus on attitudes about Latinos. The key policy domain in this literature is typically immigration. For example, Brader, Valentino, and Suhay (2008) employed a two-by-two design manipulating the ethnic focus of an immigration story (Mexican or Russian) as well as the tone (negative or positive). Their subjects participated over the Internet and were drawn from the random digit dial–selected panel maintained by Knowledge Networks. The authors find that opposition to immigration rises substantially among whites when the negative story highlights immigration from Mexico and that respondent anxiety is the principal mediator of this result.
There is also an emerging literature in political science that uses experiments to explore the influence of prejudice outside the United States. This research must often confront challenges that are not present in the American context, such as depressed economic and social conditions, as well as multiple national languages. Gibson (2009) manages to overcome these hurdles in his study about the politics of land reconciliation in South Africa. In one experiment embedded in a nationally representative sample, Gibson exposes black, white, colored, and Asian respondents to one of several vignettes about a conflict over land ownership. In each version of the vignette, one farmer claims that land currently occupied by another farmer was stolen from him and his family during the apartheid era. There were multiple versions of these vignettes, but the key manipulations involved the race of the farmer claiming current ownership of the land and the judicial judgment as to who rightfully owned the property. In some cases, the dispute involves two black South Africans, and in other cases, the contemporary occupant is white, and the farmer making the historical claim is black. In examining respondents’ views as to whether the outcome was fair, Gibson finds that the race of the respondent as well as the race of the claimants and the nature of the judgment affected perceptions of fairness. White and black South Africans differed most significantly. Whites were much more likely to judge the outcome as fair if the contemporary owner of the land was awarded ownership and also described as white. Blacks, in contrast, were considerably more likely to view the outcome as fair if the farmer with the historical claim was awarded the land and if the losing claimant was white.
The work of Sniderman et al. (2000) represents another intriguing experiment conducted outside the United States. The goal of the experiment was to determine whether expressed prejudice was more a function of the attitude holder than the attitude object. That is, rather than prejudice being “bound up with the specific characteristics of the out-group” (53), the authors hypothesize that prejudice is a function of an intolerant personality; that is, intolerant people will express prejudice against any out-group. To test this hypothesis, the authors assign Italian citizens to conditions in which two out-groups, immigrants from Africa and immigrants from Eastern Europe, are evaluated along two dimensions: the extent to which they have negative personal characteristics, and the extent to which they are responsible for social problems in Italy. Subjects are randomly assigned to evaluate Eastern European immigrants on both dimensions, African immigrants on both dimensions, or Eastern European immigrants on one dimension and African immigrants on the other. The authors find that prejudice toward Eastern Europeans is as strong a predictor of prejudice toward Africans as is prejudice toward Africans on another dimension. These results lend strong support to the authors’ argument that prejudice rests more in the eye of the beholder than in the characteristics of the out-group being evaluated.

2. Prejudice and Candidate Choice

The question of whether white voters discriminate against black candidates is still an open one. Observational work has been suggestive, but it suffers from the inability to isolate candidate race, leaving open the possibility that confounding variables are driving the (lack of) results. We briefly review two of the best examples of observational work on this question. As we show, their limitations yield an important opportunity for experimental work to make a contribution. We argue, however, that experiments on racial discrimination in the voting booth have not yet taken full advantage of this opportunity. In particular, imprecision in the experimental design has made it difficult to rule out the possibility of alternative explanations. We therefore recommend that future work pay increased attention to this issue, and we also argue that, given the mixed results, scholars should turn from the question of whether white racism hurts black candidates to begin identifying conditions under which prejudice hurts the chances of black candidates.
Highton (2004) examines U.S. House elections in 1996 and 1998. Using exit polls conducted by the Voter News Service, he measures discrimination as the difference between white support for white candidates and white support for black candidates, controlling for such factors as incumbency, funding, experience, and demographic characteristics. Highton finds no difference between white support for white and black candidates and on that basis determines that white voters showed no racial bias in these elections.
However, because Highton (2004) uses exit poll data, he lacks a measure of racial attitudes. As a result, he cannot assess whether prejudice is tied to vote choice. To be sure, by itself this may not be much of a problem, as Highton directly examines whether there is a racial double standard. But the process that determines candidate race may be endogenous to the vote choice decision. For example, consider the hypothetical case of black political figure A who decides to run for office. His personality is no more appealing than is the norm for politicians, so he loses the primary due to racial discrimination – that is, he is not sufficiently exceptional to overcome the racial bias of some white voters. He therefore is not considered in Highton's analysis because Highton counts that contest as one without a black candidate. Now consider the hypothetical case of black political figure B who has an exceptionally appealing personality. He wins the primary because his outstanding personality overwhelms the effect of prejudice. He now counts as a black candidate in Highton's analysis. If this scenario is common, so that only black candidates who are exceptional among politicians make it through the primary and to the general House election, then it is possible that discrimination does hurt black candidates in House elections, but that such discrimination is not evident due to the effects of other candidate characteristics. If Highton had a measure of racial prejudice and found it to be uncorrelated with vote choice, then this might mitigate the aforementioned concern, but he does not.
Furthermore, Highton (2004) does not control for competitiveness of the contest; it could be that white voters are voting for black candidates simply because they lack alternative viable options. Finally, he does not measure turnout, leaving open the possibility that white discrimination operates through the failure to show up to the voting booth. Lacking a measure of racial attitudes, the ability to assign candidate race, and control over such candidate characteristics as age, name recognition, ideological orientation, and personality, Highton cannot rule out the possibility of confounding variables.
Citrin, Green, and Sears (1990) examine a case study of Democratic candidate Tom Bradley's loss to Republican candidate George Deukmejian in the 1982 gubernatorial contest in California. Unlike Highton (2004), they have access to measures of racial attitudes, making use of data from polls conducted by the Los Angeles Times and the Field Institute that were conducted among a statewide sample of white Californians. Importantly, however, Citrin et al. do not simply measure the effect of racial attitudes on vote choice because they recognize that racial attitudes are deeply implicated in policy attitudes. Racial attitudes might have an effect on voting, therefore, not due to the candidate's race, but because voters might bring their racial attitudes to bear on their evaluations of the candidate's policy platform. Citrin et al. therefore pursue the clever strategy of comparing the influence of racial attitudes on vote choice for Bradley to the influence of racial attitudes on vote choice for other Democrats pursuing such state offices as lieutenant governor. They find no additional effect of prejudice on vote choice for Bradley, and conclude that Bradley's race did not hurt him among white voters.
As Citrin et al. (1990) recognize, their attempt to control for all other relevant factors besides race is by necessity incomplete. Although other candidates for state office may have shared membership in the Democratic Party with Bradley, they surely did not share the exact same policy platform. Furthermore, Bradley's personality, experience, and name recognition were different. He was even running for a different office. We would have increased confidence in the work of Citrin and his colleagues if their controls for other candidate factors were less crude than simply having other white Democrats represent the counterfactual white Bradley. Such is the potential for experiments – to control for factors that observational studies cannot in order to be sure that any relationships (or lack thereof) between candidate race and vote choice are not an artifact of some other relationship. Indeed, given that very few of the African American members of the House of Representatives hail from majority-white districts, it seems plausible that greater control over candidate characteristics might yield a finding of antiblack discrimination among whites.
Moskowitz and Stroh (1994), for example, presented their undergraduate experimental subjects with a realistic editorial, campaign brochure, and photo, all describing a hypothetical candidate (although subjects were not told that the candidate was hypothetical). Subjects read these materials and then evaluated the candidate. Subjects were randomly assigned to one of two groups wherein the descriptions were equivalent in all respects except one – the race of the candidate. Because the race of the candidate was the only thing that varied, Moskowitz and Stroh can be more confident than can Highton (2004) or Citrin and his colleagues (1990) that any difference between groups was a result of candidate race and thus has the potential to demonstrate stronger evidence of a racial double standard. Experimental control also gives Moskowitz and Stroh the ability to assess the mechanism through which prejudice affects the vote choice because they measure subjects’ perceptions of the candidate's policy positions and personality characteristics. Indeed, Moskowitz and Stroh find that prejudiced white subjects discriminate against black candidates, and that they do so by attributing to black candidates unfavorable character traits and policy positions with which they disagree.
Although Moskowitz and Stroh's (1994) work overcomes some of the problems inherent in observational studies, they also encounter a unique set of limitations. For example, the experimental context differed from an actual campaign environment in one potentially devastating way. Subjects were asked questions about their racial attitudes just prior to being presented with material about the hypothetical candidates. As a result, racial considerations may have been primed, causing subjects to bring their racial attitudes to bear on candidate evaluations when they might not otherwise have done so. Because this characteristic of their experiment is artificial, it could be that most of the time the race of the candidate does not influence their vote choice. Moskowitz and Stroh may have identified that candidate race can matter under certain conditions, but it is not certain that those conditions occur in the real world.
Terkildsen (1993) avoids some of the problems of artificiality in the design of her study. First, she measures racial attitudes in the post-test and thus does not run the risk of priming racial attitudes just before asking subjects to evaluate candidates. Second, Terkildsen uses an adult convenience sample selected for jury duty, decreasing generalizability concerns somewhat. However, Terkildsen's analysis also shares a limitation with Moskowitz and Stroh's (1994). Unlike in an actual election, in which voters typically choose between candidates, subjects were asked to evaluate the candidate in isolation. This is a particularly important limitation, given that previous work suggests that stereotypes may work differently when candidates are evaluated alone instead of in comparison to each other (Riggle et al. 1998).
Sigelman et al. (1995), in contrast, do ask subjects to choose between two candidates, one of whom is black in one condition but white in the other. Using a convenience sample composed of adults selected for jury duty in the Tucson, Arizona area, the authors find no evidence of race-based discrimination. The authors use a nine-cell design, and candidate A is identical across cells: a conservative candidate whose race is unidentified. Candidate B varies by ideology (conservative, moderate, or liberal) and by race (described by the experimenters as “Anglo,” “black,” or “Hispanic”).
Sigelman and her colleagues (1995) also claim to find evidence of an interaction effect between race and ideology, in which racial minorities are perceived as more liberal than are white candidates. The authors manipulate ideology by changing the content of each candidate's speech; subjects are expected to infer the ideology of the candidates by reading their speeches. Unfortunately, the content of the speech varies not only in the ideological principles espoused, but also in the highlighting of racial issues. Whereas the conservative candidate argues that minorities “have become too dependent on government” and the liberal candidate claims that minorities “have been victims of terrible discrimination in this country,” the moderate candidate does not mention race at all. As a result, it is difficult to tell whether subjects evaluating the ideology of a given candidate were reacting to the candidate's race, the ideological principles espoused in the candidate's speech, the extent to which racial issues were highlighted in the speech, or, quite plausibly, interactions among the three.
The experimental design employed by Reeves (1997) overcomes many of these problems. For example, to avoid potential unintended priming effects, racial attitudes are measured six months prior to the implementation of the experiment. Furthermore, respondents, identified in a representative mail survey of Detroit area residents, evaluate the candidates in a comparative context. Evaluations are based on realistic newspaper articles describing a debate between two fictitious candidates. Finally, in a four-cell design, Reeves manipulates the race of one of the candidates (black or white) and the issue area being debated (the environment or affirmative action). Thus, his design allows him to directly examine the impact of highlighting racial issues.
Unfortunately, Reeves (1997) only analyzes those subjects who choose to respond to his questionnaire, neglecting to consider the possibility that the decision to respond is endogenous to the effect of the treatments. In his experiment, unlike with most survey experiments, subjects receive a questionnaire by mail that they can read before determining whether they want to mail it back to the experimenter. Exposure to the treatment, therefore, may have some impact on the decision to participate in the study, but Reeves does not analyze whether response rates vary across experimental conditions.
Reeves (1997) claims to find evidence of racial discrimination when the campaign issue is affirmative action, but what is actually evident in the data is an increase in the number of white subjects who claim to be undecided. To be sure, Reeves finds that the distribution of racial attitudes among these respondents suggests that they would probably not support a black candidate, but this argument is only suggestive.
A somewhat more recent vintage of studies on the influence of racial prejudice in candidate selection focused more on indirect effects. With these studies, the emphasis is on contests featuring two white candidates and the prospect that subtle racial cues are employed to the disadvantage of one of the candidates. According to this literature, political candidates in the post–Civil Rights era no longer make direct racial appeals to whites because such efforts would be repudiated by voters across the political spectrum. Instead, covert references are made to race, leading whites to bring their latent antiblack attitudes to bear on candidate preferences. This process has been dubbed “racial priming” (Mendelberg 2001).
Whether political campaigns devise subtle racial cues in order to surreptitiously activate the racial views of the electorate is a difficult issue to study. When voters bring their racial attitudes to bear on some voting decisions rather than others, we cannot be sure using observational studies whether this occurred because of specific campaign tactics or some other unrelated event. Experimental manipulation provides perhaps the only way to confidently evaluate this possibility, but, as we illustrate, even here many questions remain unanswered.
Mendelberg (2001) was among the first to examine this question by developing a series of experiments manipulating whether a fictitious gubernatorial candidate's antiwelfare appeal was racially implicit (i.e., visual references to race but not verbal), explicit (i.e., visual and verbal references to African Americans), or counterstereotypical (i.e., antiwhite rather than antiblack). Her subjects are drawn from a random sample of New Jersey households. In addition to manipulating racial cues, Mendelberg also manipulated whether participants were told that their views conformed with or violated societal norms on race. She finds that concern with violating norms prevents explicit messages from activating antiblack attitudes with one caveat: racially liberal subjects who are unconcerned when told that their views violate the norm are much more likely to support the racially conservative candidate. This result suggests that further research needs to be done to explore exactly how concern for norms moderates the effects of racial priming.
Despite this support for the racial priming hypothesis, confirmed and replicated by Valentino, Hutchings, and White (2002), a debate has emerged in the literature regarding the influence of implicit and explicit racial appeals. Employing an experimental design similar to Mendelberg's (2001), but with a nationally representative sample treated over the Internet (n = 6,300), Huber and Lapinski (2006) find that implicit appeals are not more effective than explicit messages in priming racial attitudes. In Mendelberg's (2008) response, she questions whether the treatment was delivered successfully and argues that, because subjects’ racial predispositions were measured just prior to exposure to the experimental treatments, any differences in priming effects may have been neutralized. Huber and Lapinski (2008) reject these criticisms. More important for our purposes, however, is that the racial priming literature is vulnerable to the charge that measuring racial attitudes immediately prior to or following the treatment may affect the results. When the measurement occurs prior to treatment, researchers run the risk of dampening any priming effect due to “contamination” of the control group. However, when measured after the treatment, the distribution of racial attitudes may be influenced by the manipulation (Linz 2009). Thus, it is not so much that liberals and conservatives are sorting themselves out more appropriately as a consequence of exposure to an implicit racial appeal. Rather, the candidate preferences may remain firm and the racial attitudes may be changing. Future work in this area should try to measure racial attitudes some considerable time prior to the treatment in order to avoid this potential problem.

3. Conclusion

In sum, although experiments have contributed to our knowledge of the role that prejudice may play in shaping policy preferences and candidate support, a number of questions remain unanswered. Experiments clearly do provide advantages, as they address some of the weaknesses of observational studies. The main weakness of observational research is its inability to control for all possible confounds, and the main strength of experiments is their ability to do just that. Experiments, however, are not a panacea and consequently bring their own set of limitations. The limitations of the work reviewed here include (at least occasionally) an overreliance on convenience samples that consequently undermine external validity, lack of realism in the treatments, and imprecision in the experimental design, thereby clouding the inferences that can be drawn. Imprecision in the experimental design is especially troubling, given that the main advantage of experiments is the ability of the experimenter to eliminate alternative potential confounds. Some of these concerns will always be difficult to address, as they represent inherent problems with the use of experiments in the social sciences. However, by devoting our attention to improvement in design, we can minimize some of the most glaring weaknesses of this valuable tool.

References

Allport, Gordon W. 1979. The Nature of Prejudice: 25th Anniversary Edition. New York: Basic Books.
Brader, Ted, Nicholas A. Valentino, and Elizabeth Suhay. 2008. “What Triggers Public Opposition to Immigration? Anxiety, Group Cues, and Immigration Threat.” American Journal of Political Science 52: 959–78.
Citrin, Jack, Donald Philip Green, and David O. Sears. 1990. “White Reactions to Black Candidates: When Does Race Matter?” Public Opinion Quarterly 54: 74–96.
DuBois, William Edward Burhardt. 1899. The Philadelphia Negro. Philadelphia: University of Pennsylvania Press.
Feldman, Stanley, and Leonie Huddy. 2005. “Racial Resentment and White Opposition to Race Conscious Programs: Principles of Prejudice?” American Journal of Political Science 49: 168–83.
Gibson, James L. 2009. Overcoming Historical Injustices: Land Reconciliation in South Africa. New York: Cambridge University Press.
Gilens, Martin. 1996. “‘Race Coding’ and Opposition to Welfare.” American Political Science Review 90: 593–604.
Gilliam, Franklin D., and Shanto Iyengar. 2000. “Prime Suspects: The Influence of Local Television News on the Viewing Public.” American Political Science Review 44: 560–73.
Highton, Benjamin. 2004. “White Voters and African American Candidates for Congress.” Political Behavior 26: 1–25.
Huber, Greg A., and John Lapinski. 2006. “The ‘Race Card’ Revisited: Assessing Racial Priming in Policy Contests.” American Journal of Political Science 48: 375–401.
Huber, Greg A., and John Lapinski. 2008. “Testing the Implicit–Explicit Model of Racialized Political Communication.” Perspectives on Politics 6: 125–34.
Iyengar, Shanto, and Donald R. Kinder. 1987. News That Matters. Chicago: The University of Chicago Press.
Kinder, Donald R., and Lynn M. Sanders. 1996. Divided by Color: Racial Politics and Democratic Ideals. Chicago: The University of Chicago Press.
Kinder, Donald R., and David O. Sears. 1981. “Prejudice and Politics: Symbolic Racism versus Racial Threats to the Good Life.” Journal of Personality and Social Psychology 40: 414–31.
Linz, Gabriel. 2009. “Learning and Opinion Change, Not Priming: Reconsidering the Evidence for the Priming Hypothesis.” American Journal of Political Science 53: 821–37.
McConahay, J.B. 1982. “Self-Interest versus Racial Attitudes as Correlates of Anti-Busing Attitudes in Louisville: Is It the Buses or the Blacks? Journal of Politics 44: 692–720.
Mendelberg, Tali. 2001. The Race Card: Campaign Strategy, Implicit Messages, and the Norm of Equality. Princeton, NJ: Princeton University Press.
Mendelberg, Tali. 2008. “Racial Priming Revised.” Perspectives on Politics 6: 109–23.
Moskowitz, David, and Patrick Stroh. 1994. “Psychological Sources of Electoral Racism.” Political Psychology 15: 307–29.
Nelson, Thomas E., and Donald R. Kinder. 1996. “Issue Frames and Group-Centrism in American Public Opinion.” Journal of Politics 58: 1055–78.
Peffley, Mark, and Jon Hurwitz. 2007. “Persuasion and Resistance: Race and the Death Penalty in America.” American Journal of Political Science 51: 996–1012.
Reeves, Keith. 1997. Voting Hopes or Fears? White Voters, Black Candidates, and Racial Politics in America. New York: Oxford University Press.
Reyna, Christine, P. J. Henry, William Korfmacher, and Amanda Tucker. 2006. “Examining the Principles in Principled Conservatism: The Role of Responsibility Stereotypes as Cues for Deservingness in Racial Policy Decisions.” Journal of Personality and Social Psychology 90: 109–28.
Riggle, Ellen D., Victor C. Ottati, Robert S. Wyer, James Kuklinski, and Norbert Schwarz. 1998. “Bases of Political Judgments: The Role of Stereotypic and Nonstereotypic Information.” Political Behavior 14: 67–87.
Sears, David O. 1988. “Symbolic Racism.” In Eliminating Racism: Profiles in Controversy, eds. Phyllis A. Katz and Dalmas A. Taylor. New York: Plenum Press, 53–84.
Sears, David O., and P. J. Henry. 2003. “The Origins of Symbolic Racism.” Journal of Personality and Social Psychology 85: 259–75.
Sears, David O., and P. J. Henry. 2005. “Over Thirty Years Later: A Contemporary Look at Symbolic Racism and Its Critics.” In Advances in Experimental Social Psychology, ed. Mark Zanna. New York: Academic Press, 95–150.
Sears, David O., and Donald R. Kinder. 1971. “Racial Tensions and Voting in Los Angeles.” In Los Angeles: Viability and Prospects for Metropolitan Leadership, ed. Werner Z. Hirsch. New York: Praeger, 51–88.
Sidanius, Jim, and Felicia Pratto. 1999. Social Dominance: An Intergroup Theory of Social Hierarchy and Oppression. New York: Cambridge University Press.
Sigelman, Carol K., Lee Sigelman, Barbara J. Walkosz, and Michael Nitz. 1995. “Black Candidates, White Voters.” American Journal of Political Science 39: 243–65.
Sniderman, Paul M., and Edward G. Carmines. 1997. Reaching beyond Race. Cambridge, MA: Harvard University Press.
Sniderman, Paul M., Pierangelo Peri, Rui J. P. de Figueiredo, Jr., and Thomas Piazza. 2000. The Outsider: Prejudice and Politics in Italy. Princeton, NJ: Princeton University Press.
Sniderman, Paul M., and Thomas Piazza. 1993. The Scar of Race. Cambridge, MA: Harvard University Press.
Sniderman, Paul M., and Philip E. Tetlock. 1986. “Symbolic Racism: Problems of Motive Attribution in Political Analysis.” Journal of Social Issues 42: 129–50.
Terkildsen, Nayda. 1993. “When White Voters Evaluate Black Candidates: The Processing Implications of Candidate Skin Color, Prejudice, and Self-Monitoring.” American Journal of Political Science 37: 1032–53.
Valentino, Nicholas A., Vincent L. Hutchings, and Ismail K. White. 2002. “Cues That Matter: How Political Ads Prime Racial Attitudes during Campaigns.” American Political Science Review 96: 75–90.
Walton, Hanes, Jr., Cheryl Miller, and Joseph P. McCormick. 1995. “Race and Political Science: The Dual Traditions of Race Relations and African American Politics.” In Political Science in History, eds. John Dryzek, James Farr, and Stephen Leonard. Cambridge: Cambridge University Press, 145–74.
1 We do not discuss implicit prejudice here (see Taber and Lodge's chapter in this volume).
2 This focus on attitudes in the post–Civil Rights era is in part a practical one because political scientists showed little interest in questions of race or racial bias prior to the late 1960s (Walton, Miller, and McCormick 1995).
3 More recently, Sears and Henry (2005) identified four specific themes associated with the theory of symbolic racism: the belief that racial discrimination against blacks has mostly disappeared, that racial disparities in social and economic outcomes are due to blacks not trying hard enough, that blacks are demanding too much too fast, and that blacks have gotten more than they deserve.
4 Some later indexes, such as the modern racism scale and the racial resentment scale, are designed to capture similar or overlapping concepts (McConahay 1982; Kinder and Sanders 1996). To limit confusion, we focus on the symbolic racism scale, but the strengths and weaknesses associated with this theory can also be applied to its intellectual progeny.
5 This emphasis on the importance of framing effects can also be found in the work of Kinder and Sanders (1996) and Nelson and Kinder (1996), as discussed later in this chapter.
6 This question reads as follows: “Irish, Italians, Jewish, and many other minorities overcame prejudice and worked their way up. Blacks should do the same without any special favors.”
7 Feldman and Huddy (2005) are quick to point out that, although the symbolic racism scale does not behave like a measure of prejudice for this group, conservatives are less supportive of the scholarship program for blacks. Thus, they find evidence of a racial double standard for this group, with opposition to college scholarships particularly low when the target group is middle-class blacks.