Demographic characteristics have been increasingly prominent in the politics of recent nominations to the federal judiciary. For instance, Sonia Sotomayor's confirmation as the first Hispanic and woman of color to serve on the U.S. Supreme Court was heralded as a “benchmark”Footnote 1 for Latino politics and a “victory”Footnote 2 for Latino groups. In producing an “excited Latino base,”Footnote 3 Sotomayor's nomination evoked similarities with Clarence Thomas's nomination in 1991, for whom African Americans provided the strongest supportFootnote 4 despite holding mostly unfavorable views of President Bush. While the nominations of Sotomayor and Thomas, and, before them, O'Connor and Marshall were historic, they also reflected presidents' political calculations. More recently, President Trump was advised to select a female to replace retiring Justice Anthony Kennedy in an effort to recruit greater support for his choice.Footnote 5 The demographic characteristics of judges and justices, therefore, are widely believed to provide a way for presidents to make history while also appealing to valuable political constituencies.
In this paper, we study how Supreme Court nominees' descriptive characteristics affect public support for their nominations. Despite the contemporary political relevance of judges' demographic attributes, a growing body of scholarship on the influence of descriptive characteristics on judicial outcomes (e.g., Scherer, Reference Scherer2004; Boyd et al., Reference Boyd, Epstein and Martin2010; Glynn and Sen, Reference Glynn and Sen2015), and a large literature on descriptive representation in legislative settings (Lublin, Reference Lublin1997; Gay, Reference Gay2002; Preuhs, Reference Preuhs2006), surprisingly little research focuses on how public opinion toward judicial nominees is shaped by the nominees' descriptive profiles (for a recent exception, see Badas and Stauffer, Reference Badas and Stauffer2018). The lack of attention to descriptive representation as a predictor of public opinion toward judicial nominees is particularly surprising given the importance scholars attribute to public opinion for nominees' confirmation votes in the Senate (Overby et al., Reference Overby, Henschen, Strauss and Walsh1994; Kastellec et al., Reference Kastellec, Lax and Phillips2010).
We present new evidence about how judicial nominees' race/ethnicity and gender affect public attitudes. Our study examines the presence of out-group discrimination and in-group favoritism in support for judicial nominees at the individual level. We examine these relationships using data from a conjoint experiment conducted on an original survey with a national sample of 2500 Americans. Our study's design improves upon approaches used in existing scholarship on descriptive representation and the judiciary and enhances our ability to evaluate causal claims about the effect of descriptive characteristics on public opinion.
Overall, we find mixed evidence that descriptive representation affects public attitudes toward judicial nominees. At the aggregate level, we find that Americans are more supportive of nominees who share their racial or ethnic identity. These effects are strongest among white Republicans and Black Democrats, for whom race may serve as an especially salient cue of the nominees' political views. However, we find no evidence that gender affects Americans' support for judicial nominees; our null results are estimated relatively precisely and are consistent across subgroups of respondents based on gender and partisanship. Overall, our results suggest that the descriptive characteristics of judicial nominees may play an important role in shaping public attitudes toward judicial nominees in specific contexts but cast doubt on the political wisdom of selecting judicial nominees on the basis of demographic characteristics rather than other criteria, such as the nominee's political views, that may be more salient for influencing public opinion.
1. Demographic representation and the judiciary
Women and minorities are severely underrepresented at the highest levels of the judiciary. Of the 114 justices who have served on the Supreme Court, only four have been women; just three have been people of color. The historical underrepresentation of women and people of color in the judiciary motivated several recent presidents to place particular priority on the appointment of persons from these groups to federal judgeships at rates much higher than their predecessors.Footnote 6 These presidents have often asserted a link between the demographic characteristics of federal judges and how Americans view them. For instance, President Barack Obama argued that:
I think there are some particular groups that historically have been underrepresented—like Latinos and Asian-Americans—that represent a larger and larger portion of the population. And so for them to be able to see folks in robes that look like them is going to be important.Footnote 7
Justices themselves have argued for the importance of nominating justices that reflect the American population. According to former Supreme Court Justice William Brennan, for instance, “the sole end of making the Court diverse and reflective of America's heterogeneity was to foster legitimacy for it in the eyes of the American people.”Footnote 8
Despite the political salience of the courts' demographic composition, little empirical scholarship directly investigates how the demographic characteristics of judges shape public opinion. The omission is especially surprising due to the importance politicians, political observers, interest groups, and the public have placed on the confirmation of nominees from underrepresented groups to the Supreme Court in recent decades, including Sandra Day O'Connor, Clarence Thomas, Ruth Bader Ginsburg, Sonia Sotomayor, and Elena Kagan. While substantial literature studies how descriptive representation affects attitudes toward legislatures (e.g., Gay, Reference Gay2002; Schwindt-Bayer and Mishler, Reference Schwindt-Bayer and Mishler2005; Preuhs, Reference Preuhs2006) and executives (e.g., Atkeson and Carrillo, Reference Atkeson and Carrillo2007; Merolla et al., Reference Merolla, Sellers and Fowler2013), strikingly little is known about how the demographic attributes of judges and judicial nominees affect public attitudes toward them (see, however, Evans et al., Reference Evans, Franco, Polinard, Wenzel and Wrinkle2017; Badas and Stauffer, Reference Badas and Stauffer2018).
1.1. Descriptive representation and public support for judicial nominees
Existing theoretical perspectives and empirical evidence generate competing expectations about how descriptive characteristics affect attitudes toward the judiciary. One perspective suggests citizens may express preferences for demographically representative judges through the mobilization of group-based loyalties and consciousness. For instance, citizens from historically marginalized groups may link their life circumstances to the political success and representation of other group members (Dawson, Reference Dawson1994). Under these conditions, citizens may place an intrinsic value on descriptive representation and express greater support for officeholders from their group. These accounts have been used to explain positive associations between descriptive similarity and public evaluations of legislative candidates and elected officials, though these findings vary somewhat across racial groups (e.g., Gay, Reference Gay2002), gender (e.g., West, Reference West2017), and political party (e.g., Fairdosi and Rogowski, Reference Fairdosi and Rogowski2015). The historical underrepresentation of minorities and women on the federal bench may lead citizens from these groups to place particular value on descriptive characteristics when evaluating prospective judicial nominees irrespective of the nominees' political views.
The public may also use judges' descriptive characteristics as a heuristic to evaluate their political leanings. According to Bartels and Johnston (Reference Bartels and Johnston2012), the American public perceives the judiciary in explicitly political terms and desires that Supreme Court justices are chosen on the basis of their political beliefs. Consistent with this perspective, Sen (Reference Sen2017) demonstrates that political views are the most important factor in how the public evaluates judicial nominees (see also Reference Rogowski and StoneRogowski and Stone, Reference Rogowski and StoneForthcoming). Given the salience of political considerations for attitudes toward judges and justices, descriptive characteristics may be especially important in the context of nominations because nominations exemplify low-information environments. The public generally has little knowledge about the legal opinions expressed by judicial nominees and nominees typically go out of their way to avoid making their political opinions known during the confirmation process. In the absence of these political cues, demographic information may play a particularly important role in influencing public opinion toward judicial nominees. For instance, women and people of color are often perceived as more liberal than similar or white officials, respectively (e.g., Huddy and Terkildsen, Reference Huddy and Terkildsen1993; McDermott, Reference McDermott1998; Sanbonmatsu, Reference Sanbonmatsu2002; Lerman and Sadin, Reference Lerman and Sadin2016), and the use of these stereotypes may lead voters to infer whether a particular judge is likely to share their own political views.
An alternative body of research suggests few opportunities for descriptive characteristics to influence Americans' support for Supreme Court nominees. Some scholars argue that the public perceives the judiciary as fundamentally distinct from other political institutions, such that political considerations are minor contributors to public opinion in the context of the courts (see, e.g., Gibson and Caldeira, Reference Gibson and Caldeira2009). If the public believes the courts are mostly insulated from political influences, demographic-based stereotypes about the political attributes of judicial nominees may not affect attitudes toward nominees. Alternatively, if citizens are well-versed about the political positions of judicial nominees, the political stakes associated with their nominations may reduce the opportunities for descriptive characteristics to counteract the nominees' political beliefs. Evidence from Redman (Reference Redman2017) is consistent with this perspective, in which an experimental design manipulating the gender of federal judges shows that demographic characteristics do not affect public acceptance of those judges' decisions; instead, the content of the judge's decision, regardless of the judge's gender, is the dominant consideration for the public's evaluation of the case outcome. This finding is generally consistent with a growing literature in the study of elections and legislators which employs research designs that provide greater leverage for identifying the causal effect of descriptive similarity. This literature reaches largely null conclusions about the effects of descriptive representation on political attitudes and behavior (e.g., Broockman, Reference Broockman2014; Fraga, Reference Fraga2016a,Reference Fragab; Henderson et al., Reference Henderson, Sekhon and Titiunik2016). To the extent public opinion toward judicial nominees is shaped by the same influences that affect attitudes toward other political officials, descriptive representation may have little effect on public attitudes toward its (prospective) members.
The few studies that have explored the relationship between descriptive representation and attitudes toward the courts have reached mixed conclusions about its nature. Focusing on how citizens evaluate the federal courts as an institution, Scherer and Curry (Reference Scherer and Curry2010) argue that group consciousness mediates how individuals respond to the demographic composition of the courts. They present evidence from a novel survey experiment to show that perceptions of the courts' legitimacy increases among Blacks when told that the federal bench contains a higher percentage of Black judges, but that this same information decreases whites' perceptions of legitimacy. Moreover, they find that these patterns are moderated somewhat by ideology, such that increased Black representation on the judiciary has a stronger negative effect among conservative whites and a stronger positive effect among liberal Blacks. Evans et al. (Reference Evans, Franco, Polinard, Wenzel and Wrinkle2017) apply this insight to Hispanics and argue that the nomination of justices from historically underrepresented groups to the Supreme Court improves evaluations of the Court from members of those communities. Consistent with their argument, they show that Hispanic approval of the Supreme Court increased with the nomination and subsequent confirmation of Sonia Sotomayor. Though this research is not focused specifically on public opinion toward Supreme Court nominees, the findings suggest the primacy of racial group membership for how Americans evaluate the judiciary and its members.
In an examination of public support for the nominations of Clarence Thomas, Sonia Sotomayor, and Elena Kagan, Badas and Stauffer (Reference Badas and Stauffer2018) argue that shared group identity moderates the negative effect of political disagreement on individuals' support for judicial nominees. This argument predicts that individuals from underrepresented groups will be more likely to support nominees with whom they disagree when the nominee shares the individual's descriptive characteristics. Badas and Stauffer (Reference Badas and Stauffer2018) use polling data contemporaneous to each nomination to show that evaluations of the nominees were much less responsive to ideological proximity among citizens who shared these nominees' race and/or gender. These findings imply that a nominee's racial and/or gender identity may help insulate nominees from public opposition even if that nominee holds beliefs that the public finds politically objectionable. As a result, “[b]y strategically nominating members of underrepresented groups,” Badas and Stauffer (Reference Badas and Stauffer2018, 139) conclude, “presidents may be able to achieve greater congruence between themselves and their nominees than they otherwise would by appointing white men.”
We make three main contributions to research in this area. First, we test these competing theoretical perspectives to study whether Americans are more supportive of judicial nominees who share their descriptive attributes, focusing specifically on race/ethnicity and gender. Second, in doing so, we address limitations of existing research on public opinion and the courts. We are particularly attentive to relevant counterfactuals and the generalizability of our findings. Research designs such as those used in Badas and Stauffer (Reference Badas and Stauffer2018) and Evans et al. (Reference Evans, Franco, Polinard, Wenzel and Wrinkle2017) are not well-suited for identifying whether shared race and/or gender affected support for those nominees differently than an otherwise-similar nominee with different descriptive characteristics.Footnote 9 Moreover, though focused on a different dependent variable, the relatively small sample sizes and unrepresentativeness of the study population found in Scherer and Curry (Reference Scherer and Curry2010) limit the generalizability of the findings and their application to politically relevant subgroups. And third, we provide a more comprehensive study of the potential mechanisms implicated by the competing theoretical arguments presented above. In particular, we explore differences in the effects of descriptive characteristics across respondents' gender, race/ethnicity, and partisanship. So doing, we distinguish between competing theoretical expectations and provide new evidence about how demographic characteristics affect public support for judicial nominees and the potential channels through which any effects may flow.
2. Data
To investigate the effect of descriptive characteristics on Americans' evaluations of nominees to the judiciary, we embedded a conjoint experiment in a survey we conducted in January 2017. Our study coincided with a existing vacancy on the Supreme Court, providing a particularly strong claim to external validity in the realm of survey experiments. President Trump inherited a vacancy on the Supreme Court upon his inauguration due to the death of Justice Antonin Scalia in February 2016. The survey was fielded beginning on January 21 through January 30, the day before Trump nominated Neil Gorsuch to the Supreme Court. Thus, respondents evaluated prospective nominees during an actual Supreme Court vacancy and at a time when a nomination was pending, providing a degree of realism that would be absent if we asked respondents to evaluate nominees in a hypothetical context. The survey was conducted by YouGov with a sample of 2,500 respondents weighted to characteristics of the national population.Footnote 10 Table B.1 provides descriptive statistics for the sample. As the unweighted figures show, our study contains relatively large samples of Black and Latino/a respondents (230 and 195, respectively).Footnote 11
Our experimental setup proceeded as follows. First, we presented respondents with the following statement: “As you may know, the U.S. Supreme Court currently has one vacancy due to the death of Justice Antonin Scalia in February 2016. President Trump will need to nominate a replacement justice.” Respondents were then presented with a hypothetical profile of a potential Supreme Court nominee that included the nominee's background characteristics and political views, including their age, the law school they attended and their position on abortion. Respondents were randomly assigned to receive these attributes, the full text of which is displayed in Table B.2. The profiles were accompanied by text which instructed respondents to “Suppose Trump is considering nominating the following individual to serve as a justice on the Supreme Court.” Crucially, we varied two of the nominee's descriptive characteristics: race/ethnicity (white, Black, or Hispanic) and gender (man/woman).Footnote 12
The random assignment of nominee characteristics to respondents allows us to identify the causal effect of shared descriptive characteristics on a series of outcome variables relevant for evaluations of prospective Court nominees.Footnote 13 To do so, we create binary indicators that denote whether the respondent shared the same racial (Shared race) and gender identification (Shared gender) as the nominee in the profile. As noted above, we also varied other characteristics of the potential nominees, including age, educational background, career experience, and political views. Because these factors may also be important predictors of attitudes toward judicial nominees, the inclusion of these variables allows us to compare the effects of descriptive congruence against other relevant benchmarks.
After receiving the nominee's profile, respondents evaluated the nominee across three dimensions: support for the nominee, assessments of the qualifications of the nominee, and trust in the nominee's impartiality. Wordings and summary statistics for each of these questions are presented in Table 1. Each question was asked on a five-point scale; for simplicity, however, we report our main results with a collapsed binary measure of these variables.Footnote 14 (Our results are robust to using the entire five-point scale of the dependent variable, however, as we report in Figure A.1.) Respondents received and evaluated four profiles of potential Supreme Court nominees across these evaluative dimensions, providing 10,000 total nominee profile evaluations.
Table 1. Outcome variables
Note: Outcome variables were initially measured on a five-point scale before being collapsed to a binary measure. The column labeled Mean indicates the proportion of respondents who expressed positive evaluations of the nominees. N = 10,000.
The experiment addresses many of the limitations of research designs used in existing scholarship. The random assignment of respondents to judicial profiles produces multiple combinations of respondent-nominee racial and gender descriptive matches. In contrast, many observational studies focus on a single nomination contest to explore how descriptive attributes shape public opinion to the nominee and thus are not able to evaluate a counterfactual nominee in which all characteristics are held constant save for their descriptive characteristics. The direct exposure of respondents to the nominee profiles further improves upon observational studies in which respondents may have been unaware of or otherwise misperceived the descriptive characteristics of a given judicial nominee. Finally, because we presented respondents with attributes of potential nominees in a period where a Supreme Court nomination was imminent, we avoid the use of deception to characterize non-existent nominees.
Three features of the context of our study merit discussion. First, the experiment was administered in an extremely partisan environment. Not only had President Trump just been inaugurated, but the vacant seat he inherited was accompanied by a particularly high (by historical standards) level of partisan conflict. Upon Scalia's death in February 2016, President Obama nominated Merrick Garland as his successor. Yet Senate Republicans refused to hold hearings or otherwise consider Garland's nomination, which was a particular point of contention during the 2016 presidential election cycle and shaped the context in which Trump considered potential nominees. Second, because contemporary presidents are expected to nominate justices who share their political views, the respondents in our survey may have anticipated that Trump would nominate a conservative justice. Therefore, they may have projected these beliefs onto the prospective nominees referenced in our experimental setup. Third, it is possible that the experimental design provides respondents with more information about a nominee than they might otherwise receive.Footnote 15 These factors suggest that our experimental setting allows us to identify the effects of descriptive characteristics in particularly partisan, politicized, and high-information contexts. Existing research is somewhat ambiguous about whether these features would moderate the effects of descriptive characteristics. We nevertheless note that Supreme Court nominations, both contemporary and historical, are regularly partisan affairs. As judicial nominees receive greater scrutiny due to increased partisan contestation over their appointment (as Neil Gorsuch's and Brett Kavanaugh's nominations indicate), these contextual characteristics surrounding our study seem to resemble many of aspects of real-world contemporary judicial nominations. In light of the secular increase in partisan rhetoric and contestation over Supreme Court nominees since the mid-1900s (Epstein et al., Reference Epstein, Lindstädt, Segal and Westerland2006), the contemporary context can be seen as a product of earlier partisan and ideologically-driven nomination processes and shares clear parallels with instances such as the failed nomination of Robert Bork. Finally, we note that partisan disagreement over federal judicial nominees has been a component of American politics for centuries. The Senate rejected roughly one-third of all Supreme Court nominees in the 1800s, a considerably higher rate than in contemporary times.Footnote 16
Using the data from our experiment, we study the effect of descriptive representation by regressing the dependent variables on the indicators for shared race and shared gender along with indicators corresponding to the other characteristics of the nominees. This provides estimates of the average marginal component-specific effect (AMCE) of descriptive congruence (Hainmueller et al., Reference Hainmueller, Hopkins and Yamamoto2014), which represents the average difference in public opinion toward the nominees among respondents who shared the nominees' race and/or gender compared to those who did not. This quantity is calculated over all possible combinations of the nominee's other attributes, thereby eliminating potential confounding between race, gender, and other characteristics of nominees. Random assignment of the values of each characteristic further helps avoid potential confounding by ensuring that the values of one attribute are uncorrelated on average with any pre-treatment characteristics of respondents as well as the values of other nominee attributes. The AMCE is estimated with linear regression; standard errors are clustered on respondents to account for non-independence between the evaluations provided by individual respondents.Footnote 17
3. Results
We begin our analysis by investigating the overall effect of descriptive representation on attitudes toward prospective Supreme Court nominees. We test whether respondents provide more favorable evaluations of prospective Supreme Court nominees who share their race and/or gender. Below we present our main findings graphically; however, the full set of coefficient estimates is displayed in Table B.3 in the Supplementary Appendix.
Our results are shown in Figure 1, which presents the treatment effects of shared race and gender on a respondent's evaluations of the nominee.Footnote 18 The plot on the left shows the results when studying the effect of shared racial identity and the plot on the right shows results for the effect of shared gender identity. The x-axis shows the estimated AMCEs, where positive values indicate that shared descriptive characteristics increased respondents' evaluations of the nominees. The points show the estimated effects and the horizontal lines are the 95% confidence intervals associated with them. The vertical dashed line at zero indicates the null hypothesis of no effect of descriptive representation on evaluations of the nominees.
Figure 1. Effect of shared descriptive characteristics on nominee evaluations. Note: The figure presents the AMCE of shared descriptive characteristics on respondents' evaluations of the nominee. Each of the dependent variables is a binary measure. The left plot shows the effects of shared race on respondents' assessments of the nominee and the right plot shows the effects of shared gender on respondents' assessments of the nominee. The horizontal bars are the 95% confidence intervals associated with the estimated effects and the vertical dashed line is the null hypothesis of no effect.
Overall, our results indicate that race plays a significant role in shaping Americans' attitudes toward judicial nominees. Across each dependent variable, shared racial identity significantly increased respondents' evaluations of the nominees. The coefficient estimate for the Support dependent variable is 0.061 and statistically significant (p < .001), indicating that respondents who shared the nominee's racial identity were, on average, 6.1 percentage points more likely to express support for the nominee. We find nearly identical results for the other two dependent variables, in which racial descriptive representation increased respondents' trust in the nominee's impartiality and perceptions of the nominee's qualifications by 6.2 percentage points (p < .002) and 4.5 percentage points (p < .02), respectively.
The magnitudes of these effects, moreover, are fairly strong in relation to the other nominee characteristics we varied in the experiment. For instance, the effects of shared race are around half the magnitude of the effects of copartisanship on support for judicial nominees reported in Sen (Reference Sen2017, Figure 1). As Table B.3 shows, the effects of shared race on Support were about two-thirds as large in magnitude as the effects of the nominee's abortion views, and are roughly similar in magnitude as the nominee's abortion views for the other two dependent variables.Footnote 19 The effects of shared race were also slightly larger than the effects of the nominee's legal training and professional experience. For instance, respondents were 4.9 percentage points less supportive of a nominee who currently works as a corporate attorney than one who currently serves on the federal bench. Similarly, respondents were 4.4 percentage points less supportive of a nominee who attended a ‘second tier’ law school relative to an ‘elite Ivy’ law school. Altogether, these findings indicate that Americans express consistently higher support for judicial nominees who share their own racial group identity, and these effects are substantively important when compared with the effects of other characteristics.
As the right plot of Figure 1 shows, however, we find considerably weaker evidence for the effects of shared gender on attitudes toward judicial nominees. While each of the three estimates is positive, they are all considerably smaller in magnitude than the effects of shared race. Moreover, while shared gender increases trust in the nominee's impartiality by 2.5 percentage points (p < .04), it is the only one of the three dependent variables where the estimated effect is statistically significant. The effects are smaller for overall support (1.3 percentage points) and perceptions of the nominee's qualifications (0.3 percentage points), and neither of the estimates is distinguishable from zero. On the whole, we do not find strong evidence that shared gender identity affects public opinion toward judicial nominees.Footnote 20
The findings presented in Figure 1 are robust across a series of additional analyses. Our experimental setup presents respondents with pieces of information about a prospective nominee that are commonly discussed during nomination and confirmation proceedings. Nevertheless, the amount of information presented to respondents may be greater than individuals typically receive, potentially diminishing the role shared race or gender plays in evaluations of prospective nominees. We evaluate this possibility in two ways. First, though previous studies highlight the importance of knowledge in shaping attitudes toward the judiciary, we find no evidence that judicial knowledge moderates our results.Footnote 21 Second, the design of our experiment randomly assigned half of respondents to receive more information about the prospective nominee (via statements about the nominee from President Trump and Senate Democrats) than the other half of our respondents.Footnote 22 We find no systematic differences in the degree to which respondents with or without access to these additional pieces of information about the nominee value shared race or gender across the dimensions of nominee support, trust, and qualifications. Taken together, these findings weigh against the possibility that our findings are driven by the information-rich environment contained in the experimental design.
Our findings above indicate that Americans are more supportive of potential Supreme Court nominees who share their racial identity and are generally consistent with some previous research on descriptive representation and attitudes toward the courts (e.g., Scherer and Curry, Reference Scherer and Curry2010; Badas and Stauffer, Reference Badas and Stauffer2018). However, the results also provide little evidence of a similar relationship on the basis of gender, which contrasts with the findings reported in Badas and Stauffer (Reference Badas and Stauffer2018). While the aggregate analyses provide new findings about how demographic characteristics of nominees affect public support, however, they reveal less about the potential mechanisms through which the public processes the information conveyed by a Supreme Court nominee's race and gender.
3.1. Evaluating potential mechanisms
Our theoretical discussion posited several channels through which demographic characteristics could affect public opinion. To evaluate these potential explanations, we distinguish whether and how the results in Figure 1 vary depending on the respondent's demographic characteristics and partisan identification. If individuals use the race and gender of a judicial nominee as a heuristic to infer the nominee's political views, we would expect that individuals of a given race or gender may evaluate a nominee with given demographic characteristics differently depending on the individual's own political leanings. For instance, because women and people of color are generally assumed to be more liberal than men and whites, respectively, democratic men might be more supportive of women nominees than male Republicans and white Democrats might be more supportive of nominees of color than white Republicans. This argument would also suggest that Republican women would be less supportive of women nominees than Democratic women, and that Republican Hispanic and Black respondents would be less supportive than Democratic Hispanic and Black respondents of nominees of color.
Alternatively, groups may place intrinsic value on descriptive representation. According to this explanation, respondents would report a preference for same-gender and coracial candidates and these relationships would be expected to have similar magnitudes across political views, proxied by party identification. Given the underrepresentation of certain groups in the federal judiciary, however, some groups may express strong preferences for same-gender or coracial nominees. Accordingly, women and people of color may respond more strongly to women nominees and nominees of color, respectively, compared with men and white respondents. The key for distinguishing this account from the heuristic-based account described above is the absence of moderating effects across party lines.
To test these competing explanations, Figure 2 displays the results that distinguish the effects of descriptive representation on support for nominees by race and party.Footnote 23 The top panel of Figure 2 presents the treatment effects for shared racial descriptive characteristics for Democrats, and the bottom panel does so for Republicans.Footnote 24 Each entry along the y-axis displays the respondents' racial group (first entry) followed by the nominee's racial group (second entry). Thus, the coefficients displayed in the plot are the differences between respondents' support for coracial nominees and nominees from each of the other racial groups. Positive coefficients indicate greater support for nominees from racial groups other than the respondent's and negative coefficients indicate greater support for coracial nominees.
Figure 2. Partisanship, race, and nominee support. Note: The figure presents the estimated treatment effects of race on support for prospective Court nominees for racial subgroups for both Democrats (top panel) and Republicans (bottom panel). Each treatment effect presents the average difference in support for a respondent of a given race (first label) for a nominee of a given race (second label) compared to a nominee of the respondent's own race.
Overall, we find mixed evidence that Americans respond to demographic characteristics in the ways predicted by the theoretical accounts outlined above. Among whites, Republican identifiers provided the strongest evidence of a preference for coracial nominee. White Republicans were significantly less support of a prospective Black judicial nominee relative to an otherwise-similar white nominee. White Republicans were also less supportive of Hispanic nominees, though this estimate falls short of statistical significance. In contrast, white Democrats were somewhat more supportive of nominees of color although none of the differences are statistically distinguishable from zero. Overall, these results suggest that the preferences for coracial nominees shown in Figure 1 are driven more by white Republicans than by white Democrats.
Among Hispanic respondents, we generally do not find strong evidence of a preference for coethnic nominees. Compared to Hispanic nominees, Hispanic Democrats were somewhat less supportive of white nominees and somewhat more supportive of Black nominees; however, the coefficient estimates are small in magnitude and indistinguishable from zero. Hispanic Republican respondents provided less favorable evaluations of white nominees than Hispanic nominees; however, the difference was not statistically significant. Interestingly, however, Hispanic Republicans were 23 percentage points more likely to express support for Black nominees compared with Hispanic nominees. We do not wish to overstate the evidence given the relatively small sample of Hispanic Republicans; however, this result provides some evidence that political identity (here, partisanship) generates common evaluations of Black nominees among both whites and Hispanics.
Shared racial identity appears to have the clearest effect on support for the nominee among Black respondents. Black Democrats and Republicans alike expressed preferences for Black nominees relative to nominees from other racial groups. Black Democratic respondents were 17 percentage points less likely to express support for a Hispanic judicial nominee than a Black nominee, and 23 percentage points less likely to express support for a white nominee than a Black nominee. We find similar patterns for Black Republican respondents, who provided less favorable evaluations of both white and Hispanic nominees, though these effects are much less precisely estimated due to the relatively small number of Black Republicans in our sample.Footnote 25 Overall, these patterns suggest that Black respondents are more supportive of coracial judicial nominees due to the priority they place on descriptive representation in the judiciary rather than due to ideological or partisan reasons.
Figure 3 displays analogous results for gender across party lines. Among both Republican and Democratic respondents, we find no evidence that the nominee's gender affects support. The top panel shows that, among Democrats, both men and women evaluate opposite-gender nominees slightly less favorably than nominees that share their gender identity, yet these estimates are extremely small in magnitude and neither is statistically significant. The bottom panel shows similar results among Republicans; Republican men are slightly less supportive of women nominees compared to men nominees, while Republican women are somewhat more supportive of men nominees than women nominees. Again, however, these estimates are small in magnitude and statistically indistinguishable from zero. Overall, we continue to find no evidence that gender meaningfully affects how Americans evaluate judicial nominees.
Figure 3. Partisanship, gender, and nominee support. Note: The figure presents the estimated treatment effects of gender identity on support for prospective Court nominees among men and women for both Democrats (top panel) and Republicans (bottom panel). Each treatment effect presents the average difference in support for a respondent of a given gender (first label) for a nominee of a given gender (second label) compared to a nominee of the respondent's own gender.
The results in Figures 2 and 3 clarify the results from Figure 1 and provide some insight about the mechanisms by which demographic characteristics affect public opinion toward judicial nominees. Aggregate preferences for coracial or coethnic nominees are driven largely by Black respondents and white Republicans, whose group members evaluate coracial candidates significantly more favorably that candidates from other racial or ethnic groups. The results among Blacks are consistent with the hypothesis that historically marginalized and underrepresented groups have intrinsic preferences for descriptive representation as a means to gain more influence over government outcomes. The results for white Republicans, in contrast, could suggest that members of this group use demographic characteristics as a heuristic to infer the nominees' political views.Footnote 26 White Democrats and Hispanics, in contrast, do not express support for judicial nominees in ways that are responsive to nominees' racial or ethnic groups. Further, and perhaps surprisingly, gender appears to play no role in shaping public opinion toward judicial nominees. Neither men nor women of either political party express varying levels of support for judicial nominees based on their gender.
To further examine how descriptive characteristics shape the public's evaluations of judicial nominees, we consider whether the combination of a nominee's racial and gender identity affect levels of support. In particular, given the historical levels of underrepresentation that women and minorities have faced on the courts, we may expect that nominees who are both women and people of color are evaluated more negatively relative to men from the same racial/ethnic group. In Table A.1 of the Appendix, we present estimates of respondents' expressed support for a nominee for each unique racial and gender combination.Footnote 27 Across a series of subgroups of white respondents, our results reveal little overall evidence that women of color are doubly penalized due to both their gender and racial group membership. The sole exception concerns evaluations from white Republican men, who penalize Hispanic women nominees at greater rates than white women and Hispanic men. However, the negative evaluations for Hispanic women nominees are not statistically distinguishable from the negative evaluations for Hispanic men, and are roughly equivalent to those provided for Black men nominees. Overall, therefore, our results suggest that while a nominee's racial group membership reduces support among white Republicans, this effect is not compounded by the nominee's gender.
4. Conclusion
The current justices on the United States Supreme Court are perhaps more descriptively representative of the American population than they ever have been. In part, this reflects an explicit effort by some recent presidential administrations to prioritize justices with certain descriptive characteristics. Presidents, legal scholars, journalists, and justices themselves have expressed the view that descriptive representation helps to generate public support for justices and the Court among populations that have typically been underrepresented in the federal judiciary. For instance, as President Trump contemplated whom to nominate for the Supreme Court seat that became open due to Justice Kennedy's retirement, some observers believed a woman nominee would generate greater political support, as her gender would help inoculate against criticism on the basis of the nominee's abortion views.Footnote 28 For the most part, however, empirical evidence in support of these arguments has been in short supply.
Using an experimental design administered with a large sample of American public, our study reveals mixed support for these claims. Overall, shared race does strongly affect Americans' evaluations of Supreme Court nominees, as the public reports more favorable impressions of nominees from their own racial group. These effects are driven primarily by white Republicans and Black Democrats—groups which often (though not always) find themselves on opposite sides of the political divide and who may rely more heavily on racial/ethnic cues to make inferences about nominees' political views. These findings generally do not apply to other racial groups or on the basis of gender.
Our results have important implications for and raise a series of questions about descriptive representation and the judiciary. First, the limited effect of race among Hispanics and gender among women is particularly surprising given those groups' underrepresentation in the federal judiciary. It is possible that members of these groups place less priority on descriptive representation given recent increases in Hispanic and women representation on the Court. In additional analyses, we sought to study the effect of gender representation on attitudes toward the Supreme Court as the women composition of the Supreme Court changed over time. We used a question from the General Social Survey on the public's confidence in the Supreme Court which has been asked since 1973. During this time period, the number of women on the Supreme Court changed five times, beginning with the nomination of the first woman justice Sandra Day O'Connor in 1981 and most recently in 2010, when the confirmation of Elena Kagan produced three women justices on the Court for the first time. As Figure A.6 in the Appendix shows, however, women's attitudes toward the Court do not appear to be responsive to changes in the gender composition of the Court. Years with the addition of women justices tend to be marked by minimal changes in American women's views toward the Court, and any changes in women's confidence in the Court were consistent with the changes exhibited by American men. These patterns generally reflect the null findings from our experimental design and suggest that attitudes toward judicial institutions may be somewhat insulated from courts' demographic composition, at least in the context of gender. Future research could further investigate how changes in levels of descriptive representation of traditionally underrepresented groups affects how those groups make political evaluations on the basis of group identity.
Second, in an era where party polarization is at an all-time high Americans may not place as much emphasis on descriptive representation as in other periods of history. Instead, many Americans appear to understand the judiciary as political and recognize the political stakes of its members. Therefore, political context may moderate the potential for descriptive representation to affect attitudes toward judges and courts. Third, that we find the strongest effects among Black Americans could reflect that group's continued underrepresentation on the Supreme Court and in the federal judiciary more generally. The lone Black member of the Supreme Court over the last quarter-century, Clarence Thomas, is generally not representative of Black political views. Thus, political representation for Black Americans on the Supreme Court likely remains as urgent a priority as it ever has been. However, we do not wish to overstate this point as the effects of coracial nominees on attitudes among Black respondents were relatively modest in magnitude.
Our study has important limitations of its own. First, while our experiment provides high internal validity and was fielded in a way to increase its external validity, it evaluates the effects of descriptive representation in a single political context and point in time. It is unclear whether the findings would persist in a setting with a Democratic presidential administration, where the outgoing justice was a member of a historically underrepresented group, or in the context of nominations to trial or appellate courts. Second, given this context, it is possible that respondents' attitudes toward the Trump administration dominated their assessment of its potential nominees, thereby limiting the potential effect of demographic characteristics. In an era where partisanship shapes how Americans think about virtually every aspect of political life, however, it is unlikely that our estimates of the effects of demographic characteristics are artificially high. Third, our experiment may have provided more information about the nominees' backgrounds than respondents typically receive about a nominee in the real world. It is possible, therefore, that this information may have limited the opportunity for descriptive characteristics to affect evaluations of the nominees. Future research could evaluate how the information environment moderates the relationship between descriptive representation and public opinion in the context of the judiciary. Fourth, while other research explored the relationship between descriptive representation and ideology (e.g., Badas and Stauffer, Reference Badas and Stauffer2018), our study focuses on partisanship. While partisanship and ideology are highly correlated, they are distinct concepts and the correlation between them may vary across racial and ethnic groups. Thus, future research in this area could directly compare how these separate political identities interact with the descriptive attributes of judicial nominees. And finally, while our study focused on race and gender, other descriptive characteristics—such as social/income class and sexuality—could also affect attitudes toward Supreme Court justices, particularly given the Court's role in adjudicating issues regarding economic regulation and LGBT-related discrimination. These remain important questions for future research on the judiciary in particular and American political institutions more generally.
Supplementary material
To view supplementary material for this article, please visit https://doi.org/10.1017/psrm.2019.59.
Acknowledgments
We thank Stephen Ansolabehere, Chase Harrison, and Shiro Kuriwaki for helpful comments. Our research was generously supported by the Faculty of Arts and Science and the Dean's Competitive Fund for Promising Scholarship at Harvard University.
