The Influence of Descriptive Representation on Support for Judicial Nominees and the US Supreme Court

Abstract

We argue that characteristics of unelected officials directly influence individuals’ perceptions and evaluations of them. These evaluations then have indirect, downstream consequences on evaluations of the institution. To test this, we fielded a unique survey with an oversample of Black Americans after the nomination of Justice Ketanji Brown Jackson to the Supreme Court. Using a conjoint experimental design to randomize nominee race, we find that increased racial descriptive representation elicits more favorable views of the nominee and Court among Black respondents. Causal mediation analysis confirms our theoretical expectation that descriptive representation indirectly influences views of the Court through its effects on views of nominees. The effects we uncover are not confined to co-partisan nominees, indicating that descriptive representation may matter for more than policy reasons alone. Finally, our external validity test suggests these effects generalize beyond our experimental setting, with Black (but not white) respondents equally as supportive of an anonymous profile matching Justice Jackson’s characteristics as they are of Jackson herself.

Members of the American mass public tend to evaluate political institutions based, in part, on the composition of the institution. Individuals feel better represented by leaders demographically similar to them; such representation yields greater feelings of efficacy, support, and legitimacy [(Bowen & Clark, 2014; Hayes & Hibbing, 2017; Sanbonmatsu, 2003; Stauffer, 2021; West, 2017), but see Gay (2002)]. This relationship is complicated when it comes to the federal judiciary; being uneelected, there is no constituency to whom judges must respond and the countermajoritarian function of the Court muddies any representative role. As a result, research on increased diversity on the bench has revealed both positive and negative feelings (Holst & Langvatn, 2021; Overby et al., 2005; Scherer & Curry, 2010).

Nevertheless, February 25, 2022, was seen as a historic day for representation when Ketanji Brown Jackson became the first Black woman nominated to the United States Supreme Court. Senator Elizabeth Warren noted that, “A diverse judiciary matters," and that the first Black woman justice could bring, “long overdue representation to the bench."¹ Scholarship shows that this is true. Descriptive representation is simply the idea that individuals are represented by leaders that are sufficiently like them (Mansbridge, 1999; Pitkin, 1967). While descriptive representation does not automatically confer policy congruence or substantive representation (although it can, see Lublin (1997), Minta (2011), and Whitby (1997)), it can provide avenues, particularly for underserved and underrepresented groups, for people to feel empowered, elicit feelings of procedural fairness, and create other symbolic, intangible goods (Bobo & Gilliam, 1990; Gay, 2002; Hayes & Hibbing, 2017; Pitkin, 1967). Despite the lack of a discernible representative role for federal judges, members of the public evaluate judicial nominees similarly to those who run for representative office (Sen, 2017). Thus, representation influences mass evaluations of the judiciary (Scherer & Curry, 2010), particularly among certain demographics (Evans et al., 2017; Ostfeld & Mutz, 2021).

Existing research on the role that descriptive characteristics play in judicial evaluations falls into two broad camps. First, descriptive characteristics shape support for individual nominees (Badas & Stauffer, 2018; Sen, 2017). Second, the overall diversity of state (Overby et al., 2005), federal (Scherer & Curry, 2010), and international (Holst & Langvatn, 2021) courts shapes support for the respective institution. We believe these literatures ought to intersect. The nomination of a Supreme Court justice is one of the most salient and public events the Court is (in)directly involved in outside of handing down a salient decision. We believe the public gains information and makes evaluations about both the nominee and the Court during this period. We argue that evaluations of a nominee are based on descriptive characteristics of the nominee, as well as broader racial attitudes (e.g., racial resentment). Further, we argue that these evaluations of a nominee are linked to views of the institution itself. Previous research leads us to accept two things: (1) descriptive representation influences nominee support and (2) nominee support influences institutional support. Yet, the link between descriptive representation and institutional support is far murkier; we doubt descriptive representation has a clear, direct effect on institutional support.

We contend that nominee support—which, again, is partially a function of descriptive representation—has positive and significant downstream effects for the institution on which they serve.² That is, our theory holds that the role of descriptive representation on institutional evaluations is mediated by nominee support. Indeed, our theory reflects the likeliest logical path by which one would make evaluations about the institution (outside of the context of a case, a controversy, etc.); when a new judge is nominated to the Court, individuals almost certainly evaluate the nominee first and, depending on their evaluation, then assess the institution with the new member in mind. We utilize the nomination of Ketanji Brown Jackson to the Supreme Court to examine the effects of racial descriptive representation and racial attitudes on support for the nominee and, importantly, their potential downstream impact on support for the Supreme Court.

Another important line of research on public evaluations of Supreme Court nominees examines how an individual’s political beliefs or partisan attachments influence their evaluation of nominees. Most individuals evaluate judicial nominees similarly to other political actors (Sen, 2017), and perceived ideological or policy agreement influences perceptions of both the Court (Christenson & Glick, 2015) and nominees (Badas & Simas, 2022). In the same way that shared ideology can boost support for a nominee, so can shared identity, such as race; importantly, shared racial identity can increase support above and beyond, or even in spite of, perceived political similarity (Badas & Stauffer, 2018; Kaslovsky et al., 2021). Still, partisan attachments influence both Supreme Court evaluations (Armaly, 2018a; Clark & Kastellec, 2015) and nominee evaluations (Krewson & Schroedel, 2020; Krewson, 2023), and those who “lose" (on partisan grounds) when a new justice is confirmed confer less support on the institution (Armaly & Lane, 2022). Therefore, we think it is important to understand whether shared partisanship with a potential nominee may result in heterogeneous effects for descriptive representation and its downstream effects on support for the Court. We expect that both shared racial and shared partisan identities will play a role in evaluations of nominees and courts.

To test our theory regarding race, nominee evaluations, and Court support, we use a (quota-based) nationally-representative sample with an oversample of Black respondents. The survey was conducted after President Biden’s nomination of Jackson, but before the Senate confirmation hearings began.³ We employ a pre-registered conjoint design whereby we randomize several features—such as the race, sex, age, and legal training of the nominee, the party of the nominating president, and so on—to isolate the unique influence of particular characteristics.⁴ The oversample of Black respondents allows us to make stronger, more generalizable claims about the role of descriptive representation, particularly among groups traditionally underrepresented.

Additionally, because our survey was conducted during an actual nomination, we are able to exploit the timing to increase the external validity of our claims. Specifically, we presented respondents with a static (i.e., non-randomized) profile that described Judge Jackson’s characteristics, but in no way suggested or revealed that it was her. This profile appeared as another “hypothetical” nominee. In addition, respondents evaluated Judge Jackson specifically. We use these results to conduct an external validity check by comparing evaluations of Judge Jackson herself to an anonymous profile of Jackson. This enables us to see whether respondents evaluated Justice Jackson herself and her anonymous profile similarly. If so, then it provides suggestive evidence that the evaluations in our experimental setting mimic evaluations in the real-world, and that descriptive representation matters.

The results of our study support our theory. Relying upon causal mediation analysis (Imai et al., 2011), we find that racial considerations have an indirect influence on views of the Court via nominee perceptions. That is, for Black Americans, racial descriptive representation improves views of a nominee (e.g., trust, perceived quality), which then predicts support for the Court. But, we see no direct effect of racial descriptive representation on support for the Court. This is consistent with our theoretical expectations. We also find that descriptive representation “matters" whether or not a nominee is co-partisan, suggesting that descriptive representation may not simply cue substantive representation.⁵ What is more, we demonstrate via the external validity check that Black Americans respond in equally positive terms to an anonymous profile of Jackson and to Jackson herself, whereas white Americans seem simply to support Jackson. The external validity check suggests that our experimental findings may generalize to the real world.

We believe our study offers three important contributions: (1) the elucidation of a theory regarding how descriptive representation can causally influence views of both nominees and the Court, including a consideration of both direct and indirect effects, (2) generalizability to a real-world nomination of a Black nominee, and (3) an oversample of Black Americans, allowing us to speak better to pending representation of a historically underrepresented group.

Our results indicate that, in addition to substantive representation (i.e., offering policy benefits), judges have the potential to offer symbolic benefits to members of the American mass public. These symbolic benefits, by way of influencing views of judges, can help buttress support for the Court and increase perceptions of fairness, effectiveness, and even legitimacy. This is significant, given the increasing blows the institution suffers as a result of polarized political discourse [e.g., Armaly and Enders (2022), Armaly and Lane (2022), and Nelson and Gibson (2019)]. Thus, our results have important implications for studies on future nominations (both judicial and otherwise), mass attitudes toward the judiciary, and judicial politics more broadly. They also highlight the consequences of identity politics and representation, even within an unelected institution.

In the following section we build our theory in two parts. First we connect the two lines of literature on descriptive representation that focus on support for a nominee based on their identity traits and support for the institution based on its overall composition. We also include discussion of how ex ante racial attitudes (e.g., racial resentment) are an important factor in these evaluations and are largely missing from the existing literature. Then, we discuss how partisan predispositions are likely to influence these evaluations. Empirically, we begin by demonstrating, descriptively, preferences regarding representation on the judiciary among Black and white Americans. This is an important first step, as if individuals did not care about diversity in the judiciary we would have a more difficult time connecting descriptive representation and institutional support. We then employ structural equation models and mediation analysis to test our hypotheses, both with and without considerations of partisanship. Finally, we utilize an external validity check, comparing views associated with Jackson to views associated with an unidentified profile having her very characteristics.

Descriptive Representation, Nominees, and the Supreme Court

Individuals support the Supreme Court for a variety of reasons.⁶ Based on it’s uniquely legal nature, scholars have argued that perceptions of procedural fairness and legal symbolism bolster support for the Court (Baird & Gangl, 2006; Caldeira & Gibson, 1992; Tyler, 2006; Gibson et al., 2014). A separate and growing body of evidence, however, indicates that perceived policy congruence underlies Court support (Bartels & Johnston, 2013, 2020; Hetherington & Smith, 2007). The types of factors related to support for the Supreme Court also influence public support for nominees. For instance, the perception that one has the “legal and technical qualification necessary to be a good judge” underscores support for nominees [(Gibson & Caldeira, 2009a, p. 65); also see Hoekstra and LaRowe (2013)]. But, nominees are also evaluated based on their perceived ideological proximity and co-partisanship (Badas & Simas, 2022; Krewson & Schroedel, 2020; Krewson, 2023; Sen, 2017). Just like with evaluations of the Court, “intangible goods" related to fairness and trust help explain attitudes about nominees (Gay, 2002), but so do perceptions of more substantive benefits in the form of preference alignment. Thus, it is clear that the public supports nominees (and the Court) for substantive as well as symbolic reasons. In this study, we emphasize identity politics and representation as an important lens through which the public perceives substantive and symbolic benefits [see Tate (2004)], thereby increasing their support.

Due to the unrepresentative nature of the judicial branch, the connection between identity politics and judicial evaluations is not always as explicit as it is when studying representative branches of government. As Weissberg (1978) makes clear in discussing representation, there are two primary types of representation: dyadic and collective. Collective representation relates to how well a legislative body, as a whole, represents constituents. In contrast, dyadic representation reflects how well an individual is represented by their particular representative. Both forms relate to descriptive representation, as a governing body can reflect diversity collectively or one can be dyadically represented by one who looks like them [see Harden and Clark (2016)]. When collectively represented, members of traditionally underrepresented groups feel more represented, and they perceive government outcomes as more fair and legitimate (Clayton et al., 2019; Hayes & Hibbing, 2017; Hutchings et al., 2004; Stauffer, 2021; Tate, 2004). Despite not having a true representative component, many scholars believe the same theory should apply to the judiciary. Results, however, are mixed. Collectively, positive evaluations increase among Black Americans when told the federal judiciary is diverse (Scherer & Curry, 2010), though Overby et al. (2005) do not find the same at the state level. Similarly, more women on the bench in international courts and as chief justices in state appellate courts is beneficial for women’s perceptions (Holst & Langvatn, 2021; Lee et al., 2022).

While we are unable to study dyadic representation in the judiciary as one would in an electoral context, we think individual representation, or the idea that an individual nominee or judge can provide descriptive or other forms of representation, is a more apt way to conceive of judicial representation. That is, we think adding even a single descriptively representative judge is important, even if doing so does not necessarily make the judiciary more collectively representative. We believe this accurately reflects how members of the mass public make evaluations of the courts. After all, people likely assess the Court in one of two circumstances: After a salient decision or when there is a nomination. In the decision context, we would anticipate no effect of racial descriptive representation. When there is a nomination, it is possible (albeit unlikely, see (Franklin, 2022)), that one assesses the overall diversity of the Court and therefore determines whether it is collectively representative.⁷ It is likelier, however, that individuals evaluate the individual nominee—based on shared physical traits and other considerations (Pitkin, 1967)—making descriptive representation operative.

Dyadic (or individual) representation matters in legislative and judicial contexts. In the legislative context, shared identity traits play an important role in candidate evaluations, often leading to more positive evaluations (Ansolabehere & Fraga, 2016; Bejarano et al., 2021; Jones, 2016; McDermott, 1998; Philpot & Walton Jr, 2007; Stout, 2018). This is in part due to the fact that co-ethnic representatives are more likely to provide substantive representation (Lublin, 1997; Tate, 2004; Whitby, 1997) and better constituent assistance (Butler & Broockman, 2011; Grose, 2011; Lowande et al., 2019; Minta, 2011). The same is true in the judicial context for evaluations of Supreme Court nominees. Simply, Americans are more supportive of co-racial nominees (Kaslovsky et al., 2021). This shared identity can, at times, help overcome ideological incompatibility due to its more symbolic benefits. For example, we see this in the relatively high levels of support for Clarence Thomas and Sonia Sotomayor among liberal Black Americans and conservative Latinos, respectively (Badas & Stauffer, 2018). Beyond the nomination context, identity also relates to evaluations of judges once they are on the bench in both positive and negative ways; Achury et al. (2023), Bracic et al. (2023), and Ono and Zilis (2022) find that ascriprive characteristics underscore perceptions of judicial fairness and (im)propriety.

In both judicial and legislative contexts, descriptive representation affects more than support for government actors. In addition to descriptive representation boosting candidate evaluations, it also relates to normatively positive political behaviors (Bowen & Clark, 2014; Gay, 2002; Gleason & Stout, 2014; Tate, 2004; Wolak & Juenke, 2021) and positive feelings about government in the legislative context (Banducci et al., 2004; Bobo & Gilliam, 1990; West, 2017). Some work on descriptive representation in the judicial context demonstrates similar relationships. For example, Justice Sotomayor’s confirmation increased perceptions of government responsiveness among Latinos (Evans et al., 2017; Ostfeld & Mutz, 2021). And, diverse judicial nominations work to boost presidential support as well (Badas & Stauffer, 2023). Yet, researchers have failed to connect identity and individual descriptive representation of nominees to overall support for the Court, which is consequential since the Court relies on the public’s support to ensure its decisions are implemented and enforced (Bartels & Johnston, 2020).

Given their salient nature, we believe that highly publicized nominations are the likeliest scenario in which descriptive representation will affect views of nominees and the Court. It is well documented that the public tunes into these hearings (Gibson & Caldeira, 2009; Kastellec et al., 2010) and evaluates the Court when doing so (Armaly & Lane, 2022; Glick, 2023; Krewson, 2023). That is, the salient nomination and confirmation of a descriptively representative judge can lend potential substantive or symbolic benefits to groups the judge descriptively represents, especially when individuals in those groups have a high level of group consciousness (Minta, 2011). We believe this will be particularly true for a group that has historically been underrepresented on the Court (Gibson & Nelson, 2018). Substantively, the judge may vote in a way that represents their groups’ policy goals. Symbolically, the judge’s presence on the bench may elicit perceptions of procedural justice, fairness, trust, and satisfaction with decision-making (Hayes & Hibbing, 2017; Schwindt-Bayer & Mishler, 2005; Tate, 2004), all of which are related to evaluations of the Supreme Court (Easton, 1975; Scherer & Curry, 2010).

Based on the scholarship regarding support for nominees and the Court, as well as scholarship describing the nature of descriptive representation, our specific hypotheses about descriptive representation and its effects on Court support are as follows:

• Hypothesis 1 Support for the nominee will be higher among Black (white) respondents who evaluate a potential Black (white) nominee.

• Hypothesis 2 Support for the Supreme Court will be higher among Black (white) respondents who evaluate a potential Black (white) nominee.⁸

• Hypothesis 3 The influence of same-race judges on the evaluations of the Supreme Court are mediated through nominee evaluations.

It would be remiss to ignore the powerful influence of existing racial attitudes when considering evaluations of nominees. The same work that finds Black respondents support the judiciary more when told there are many Black federal judges also finds that white respondents are less supportive of the institution when told the same (Scherer & Curry, 2010). Thus, in the same way that a shared race political figure can increase support for institutions, the converse is also true. This is likeliest among those who harbor racial animus. It is sensible that those who do not hold racially egalitarian views would be less supportive of Black nominees and of the Court should a Black judge join. This is particularly important in the context of the Jackson nomination. Thus, we also account for respondents’ levels of racial resentment (Kinder & Sanders, 1996), a measure combining racial animus and perceived violations of American hard-work values. Racial resentment relates to evaluations of political figures [see Jardina (2021) and Knuckey (2011)], so it should naturally extend to evaluations of judges.⁹

• Hypothesis 4 Racially resentful respondents will be less supportive of Black nominees relative to white nominees.

Finally, we account for the potential role of individual partisanship, which influences assessments of both nominees and the Supreme Court (Bartels & Johnston, 2013, 2020; Hetherington & Smith, 2007; Kaslovsky et al., 2021). Badas and Stauffer (2018) demonstrate that shared racial identity with a nominee increases support, but perceived alignment in terms of political predispositions strengthens the effect. As such, we expect that shared racial and partisan identity will work together to affect support for the nominee, and as a result support for the Court. Therefore we expect:

• Hypothesis 5 The positive effects of a same-race judge will be stronger when the respondent and nominee share party identification.

While we expect the effects of descriptive representation and partisanship to be the same for both Black and white respondents, we acknowledge that they may be more apparent among Black respondents. Though the effect of shared white identity has grown stronger in recent years as a result of perceived threats from other groups encroaching on their political power (Jardina, 2019), Black respondents may have stronger in-group consciousness and a perception of linked fate that makes descriptive representation particularly impactful (White & Laird, 2020). As for partisanship, White et al. (2014) and White and Laird (2020) suggest that upwards of 80% of Black Americans support Democrats regardless of self interest and personal ideology because of strong social constraints from other members of their racial group. That is, Black Americans must evaluate if expressing their personal political preferences is more or less important than maintaining good standing within their racial group. This racialized social constraint creates strong incentives for Black Americans to support Democratic candidates and, in our case, Democratic nominees, which suggests that partisanship could condition descriptive representation effects among Black respondents.

We test these hypotheses by exposing individuals to hypothetical profiles of nominees. We ask them to evaluate the hypothetical nominee as well as the Court if the nominee were to join. Our primary expectations are that (1) individuals will evaluate nominees more positively when they are of the same race and (2) variation in nominee support will strongly predict variation in Court support but descriptive representation will not directly affect Court support. We test for these indirect and direct effects using causal mediation analysis. To manufacture the hypothetical nominees, we employ a conjoint design wherein race is randomly assigned and uncorrelated with other nominee attributes. Doing so allows us to capture the causal effect of descriptive representation. We control for racial resentment and a host of other variables. We also test for heterogeneous causal effects by party identification.

Data and Measures

We utilized Lucid to field an original survey of 1223 Black and white respondents in March 2022.¹⁰ Limiting our analysis to Black and white respondents allows us to speak directly to previous research on the effects of descriptive representation on Court support [e.g., Scherer and Curry (2010)].¹¹ Theoretically, we argue that strong in-group consciousness and a perception of linked fate should make descriptive representation particularly impactful for Black respondents (White & Laird, 2020), whereas whites may not feel threatened by the nomination of a single Black individual on a Court where whites have always been, and still are, well-represented (Jardina, 2019). Finally, we are interested in explaining the substantive effects of Ketanji Brown Jackson’s nomination on support for the Court among Black individuals, which naturally leads us to focus on Black and white respondents.

Once in the survey, respondents provided information about their demographic characteristics, like race, sex, age, income, and partisanship. After providing this information, respondents were introduced to a hypothetical nominee profile and questions associated with it. Once they answered questions associated with a profile, they were provided another profile and the same questions. In total, respondents observed four profiles and their associated questions. For each profile, respondents were prompted to “consider a hypothetical nomination in which the US president nominates the following individual to the US Supreme Court.” They were asked to “[c]arefully review the nominee profile before answering any subsequent questions.” The profile’s attributes were created using a conjoint design that randomized the order and content of a hypothetical nominee’s sex, age, race, religion, ideology, judicial philosophy, legal education, and appointing president. Table 1 provides an example of what a profile might look like.

Table 1 Example of a conjoint profile

To measure support for the nominee and support for the Supreme Court (if the hypothetical nominee were to join), we leverage questions which were asked each time a respondent saw a new profile. Three of these items are used to measure nominee support and three are used to measure Court support. Specifically, we asked about the levels of respondents’ (1) “support for nominee,” (2) “trust in nominee to reach impartial decisions,” and (3) strength of “qualifications to be a US Supreme Court justice.” This battery of items—specifically used in previous scholarship on support for nominees (Kaslovsky et al., 2021)—combines common approaches to measuring nominee support in the judicial literature, with impartiality and quality related explicitly to the legal underpinnings of nominee evaluations (Epstein et al., 2006; Krewson, 2023).¹² Kaslovsky et al. (2021) demonstrate that descriptive representation increases nominee support regardless of which item is used. In our data, the three items closely correlate with one another, they load strongly on to a single factor (0.92, 0.92, and 0.82, respectively), and they are statistically reliable, with a Cronbach’s $\alpha$ of 0.92.¹³ We scale the respondents’ responses by creating an additive index of the three items for our measure of nominee support.

Immediately after indicating their support for the hypothetical nominee, survey respondents indicated (1) their “level of approval for the Supreme Court if the hypothetical nominee became a member of the Supreme Court.” In addition, they indicated (2) how strongly they would perceive the Court “as more of a legal institution," and (3) “as more likely to make decisions I agree with" if the hypothetical nominee were to join the Court.¹⁴ The items represent the mix of factors which scholars generally argue explain Court support, including policy agreement (Bartels & Johnston, 2013) and perceptions of legalism (Gadarian & Strother, 2023; Gibson & Caldeira, 2009b). As with nominee support, we create a summative index of the items for our measure of Court support. The three measures correlate strongly, they load onto a single factor (0.68, 0.68, 0.84, respectively), and they are statistically reliable (Cronbach’s $\alpha$ of 0.78).

Table 2 summarizes our dependent variables.

Table 2 Summary of outcomes of interest

Because respondents saw four randomized profiles each, it leaves us with 4892 observations (1223 respondents $\times$ 4 profile evaluations). In our main analysis, however, we subset the data to the 2409 data points with a Black or white nominee. That is, we drop Asian and Hispanic profiles. We do this because we are interested in—like other research in this line of scholarship—the effects of shared race rather than shared minority status (and we simply have too few Asian and Hispanic respondents to make generalizable claims about their shared race).¹⁵ Analyzing how Black respondents reacted to Black nominees relative to white nominees, and how white respondents reacted to white nominees relative to Black nominees, provides for a clear test of descriptive representation, whereas incorporating Asian and Hispanic profiles could introduce a role for shared minority status (but see Gershon et al. (2019) for research on “minority linked fate”).¹⁶

After observing and evaluating the four randomized conjoint profiles, respondents proceeded with one additional experiment (which we discuss later). We then concluded the survey by measuring racial resentment and self-reported views on diversity on the Supreme Court. To measure racial resentment, we employed the Kinder and Sanders (1996) four-item battery on which respondents react, using five-point, strongly disagree to strongly agree scales. Although racial resentment is typically measured among white respondents, there is appreciable variation in the attitude among Black respondents, and it proves predictive of racial policies within this group (see Kam and Burge (2018; 2019)). We discuss the other variables related to diversity which we measured post-treatment later in the text.¹⁷

In the next section, we provide some useful context on self-reported differences in the importance of descriptive representation among Blacks and whites using the post-treatment observational data. We then proceed to test our hypotheses using the conjoint data described above. After presenting these results, we then describe our additional experiment, which we utilize as an external validity test.

Descriptive Results: Importance and Influence of Diversity

We begin with a descriptive exercise to have a better understanding of whether individuals think diversity is important for the Court. If individuals do not value diversity, we would be in murkier territory as to whether descriptive representation matters.

We asked respondents whether they find eight different types of diversity on the Supreme Court important, using a 4-point “not important” to “very important” scale. Figure 1 plots the proportion of Black (dark gray bars) and white (light gray bars) respondents indicating that each diversity type listed along the horizontal axis is very important on the Court; vertical lines represent 95% confidence intervals. For every type of diversity other than ideological, Black respondents value diversity more than white respondents.

Fig. 1

Plot of proportions of Black and white respondents indicating that each of eight types of diversity is “very important” on the Supreme Court

Black respondents view racial diversity as particularly important, with 49% selecting this category as very important, compared to only 28% of white respondents. Black respondents also find racial diversity to be more important than other types of diversity (except for ethnic diversity, $p=0.12$). Gender and educational diversity are of more importance to Black respondents than religious, ideological, geographical, and sexual orientation diversity. Similar patterns emerge for white respondents, but the level of importance is systematically lower than for Blacks (though ideological diversity is equally important for both).

These preliminary results suggest that Black individuals care more about descriptive representation on the Supreme Court than white individuals. Especially in the context of Jackson’s nomination, this hints at linkages between a racially descriptive nominee and support for a diversifying Court as a result. However, these results are based on self-reports of preferences regarding diversity, did not require respondents to make trade-offs, and are non-experimental findings. Thus, we wait for the results regarding descriptive representation using the randomized conjoint data before making broader claims about causal effects.

Mediated Effect of Descriptive Representation on Court Support

As a simple test of our hypotheses that racial considerations (same-race judge and racial resentment) explain both nominee support and Court support—and to obtain some initial support for our theory that greater support for courts is a downstream (or indirect) effect of descriptive representation improving views of nominees—we estimate a structural equation model (SEM) using the two following OLS regression models or equations:

$$\begin{aligned}{} & {} \begin{aligned} JudgeSupport_i = \alpha _1 + \beta _{1a}BlackRespondent_i + \beta _{1b}BlackNominee_i +&\\ \beta _{1c}BlackRespondent_i\times BlackNominee_i+\beta _{1d}RacialResentment_i+ \delta _1X_i+\epsilon _{1i} \end{aligned} \end{aligned}$$

(1)

$$\begin{aligned}{} & {} \begin{aligned} CourtSupport_i = \alpha _2 + \gamma JudgeSupport_i + \beta _{2a}BlackRespondent_i + \beta _{2b}BlackNominee_i +&\\ \beta _{2c}BlackRespondent_i\times BlackNominee_i+\beta _{2d}RacialResentment_i + \delta _2X_i+\epsilon _{2i} \end{aligned} \end{aligned}$$

(2)

The two outcome variables are Judge Support and Court Support. Both variables are modeled as a function an intercept ($\alpha$), a vector of covariates (represented by X and including respondent age, sex, income, and partisanship), an error term ($\epsilon$), the randomized experimental treatment (whether the nominee is Black or White, or Black Nominee), and the race of the respondent (Black Respondent). Furthermore, the effect of a nominee’s race (Black Nominee) is moderated by the race of the respondent (Black Respondent), as we expect shared race to influence nominee and Court support. SEM allows us to estimate the models simultaneously, treat some variables as both endogenous and exogenous, and correlate error terms. We estimate robust standard errors, clustered on respondent ID to account for the fact that respondents viewed multiple profiles.¹⁸

We present the results in Table 3. All variables are rescaled 0-1 so coefficient magnitudes can be compared and for ease of interpretation. We begin by considering the exogenous variables in the Judge Support column. First, we find that Black respondents evaluate Black Judges more favorably (per the interaction term between those variables), which supports Hypothesis 1. This strongly suggests that it is racial descriptive representation that matters, not the race of the respondent or the nominee (i.e., Blacks are not more supportive of all nominees, and people are not necessarily more supportive of Black nominees). The effect size is meaningfully large, with a Black respondent predicted to be 7.7 percentage points more supportive of a Black nominee than a white nominee. We also find that Racial Resentment, when interacted with Black Judge, reduces support for the nominee. This supports Hypothesis 4: those who are racially resentful are unlikely to support a diverse nominee.

Table 3 Structural equation model

Next, we move to the Court Support column. Beginning with the exogenous variables, we find that only racial resentment predicts Court support (though we only cautiously interpret constitutive terms in the presence of interactions). Here, we fail to reject Hypothesis 2, though part of our theory is that the direct link between descriptive representation and support for the institution is specious to begin with. Thus, when controlling for exogenous characteristics, neither the race of the judge nor respondent have any direct effect on evaluations of the Court. Nor does descriptive representation. We argue that this is a departure from the existing literature, which directly links descriptive representation and support for institutions (e.g., Scherer and Curry (2010)).

The endogenous Judge Support variable in the Court Support column provides for an initial test of indirect effects. This variable has a strong and statistically significant effect on Court support, which is consistent with our theory that racial descriptive representation directly influences evaluations of nominees (as is shown with the exogenous variables in the Judge Support column), but indirectly influences evaluations of the institution (as is shown with the endogenous variable in the Court Support column). That is, the effect of descriptive representation on the Court appears to be mediated by nominee support. The endogenous variable of Judge Support dominates all other variables, with a full unit increase in nominee support predicting a 58.3 percentage point increase in Court support.

Recall that our primary theory is that the descriptive representation a nominee provides causes an increase in Court support, but that the effect is mediated largely through nominee support. We now formally test this causal pathway using the causal mediation analysis methodology proposed by Imai et al. (2011), which separates the effect of a treatment (here, random assignment of nominee race) into its direct effect on an outcome variable (Supreme Court Support) and its indirect effect on an outcome variable through a mediator variable (Nominee Support). The methodology requires us to fit a model of the mediator variable as a function of the treatment and other control variables. We use the same specification we used to model Judge Support in the SEM. It also requires us to model the outcome variable as a function of the treatment variable, control variables, and the mediator variable. We use the same model specification we used to model Court Support earlier in the SEM.

Fig. 2

Causal mediation analysis

Figure 2 presents the moderated mediation results,¹⁹ again employing standard errors clustered on respondents.²⁰ For white respondents, the nomination of a Black nominee (relative to a white nominee) has no mediated nor direct effects on support for the Court. For Black respondents, in contrast, we observe that the total treatment effect is driven by mediation. That is, mediation through nominee support explains 66% of the overall total effect of our experimental manipulation on support for the Court. These results lend strong support for Hypothesis 3.²¹ Sensitivity analyses (provided in Appendix E.2) show that the findings are robust to the inclusion of potential omitted variables which may have otherwise confounded the relationship between nominee support and Court support (Imai et al., 2011).

We also wish to point out briefly that Black respondents react positively to descriptive representation only when the hypothetical judge is also Black. Using separate analyses (see Appendix E.3 for more information), we find that Black respondents are no more supportive of Hispanic or Asian judges than they are of white judges, but are consistently more positive toward Black Judges than white judges. That is, we identify a shared-race effect rather than a shared minority status effect. This bolsters our argument that descriptive representation has a downstream impact on support for the institution.

In all, we find support for our theory using both structural equation modeling and causal mediation analysis. This highlights one route by which support for a nominee ultimately influences support for the Court. Thus, when presidents nominate judges, pledges of diversity—in particular, racial diversity—are more than mere identity appeals. Indeed, such nominations can impact the institution in the eyes of the public.

Examining the Role of Partisanship and Co-partisanship

We next consider the potential for heterogeneous effects based on perceived policy alignment with the judge. We consider partisan identification itself as well as co-partisanship with the nominee. First, does racial descriptive representation matter more for Democrats than others?²² We re-estimated our structural equation models, but this time for Democratic and non-Democratic respondents separately.²³ The results suggest that descriptive and substantive representation work together interactively. Racial descriptive representation directly predicts judge support for Black Democratic respondents, with downstream implications for their support of the Supreme Court. On the other hand, we observe no evidence that racial descriptive representation affects the views of non-Democratic individuals.²⁴ In both groups, judge support strongly predicts Court support, though the descriptive representation element only factors into this endogenous variable for Democratic respondents. That is, for non-Democratic respondents, support for a nominee still impacts support for the Court, but nominee support is not a function of shared race. These findings are confirmed using causal mediation analysis with the same underlying regression models and data (see Fig. 3).

Fig. 3

Causal mediation analysis by respondent partisanship

Why do Black Democrats respond so clearly to racial descriptive representation, but not other Black respondents? That non-Democrats responded less strikingly to a Black nominee suggests one partial explanation for why diversity and description representation matter: people may be more supportive of co-racial nominees when they believe the nominee shares their politics and will (attempt to) deliver preferable policy, or preferable policy for their identity group (White & Laird, 2020). At the end of the survey—which we did to avoid biasing the experiments—we asked individuals if they “believe the Supreme Court is more responsive to the issues [they] care about when some of the justices share [their] identity” (we did not specify any specific type of identity). Fifty-one percent of Black respondents said yes, whereas only forty-one percent of white respondents did—a statistically significant difference.

Building on this foundation, we estimate our structural equation and causal mediation models separately for co-partisan nominees and out-partisan nominees, relative to the respondent’s partisanship.²⁵ We use the same underlying regression models as before and again cluster standard errors on respondent. Figure 4 presents the causal mediation results. When a judge is co-partisan (left portion of Figure), we observe a statistically significant mediated effect of a Black hypothetical nominee for Black respondents but no direct effect. This is consistent with our expectations. When a judge is not co-partisan (right portion of Figure), we still observe a statistically significant mediated effect (but an insignificant total effect when combining direct and mediated effects). That the mediated effect of descriptive representation is not confined to co-partisan judges suggests that descriptive representation is not solely a function of substantive considerations. In sum, Black Democrats are uniquely responsive to the race of a nominee regardless of that nominee’s partisanship, though the downstream implications for Court support are muddied when the nominee is not a co-partisan.²⁶ Racial descriptive representation is not simply a stand-in for policy preferences (otherwise, we should see no effect once looking at only co-partisan relationships); rather, the results show that racial descriptive representation creates additional support for the nominee and Court than would otherwise exist based on perceived policy alignment alone.

Fig. 4

Causal mediation analysis by co-partisanship

Broadly, we find that descriptive representation matters for nominee support and has downstream effects on Court support. However, the results regarding partisanship and co-partisanship lead us to uncertain conclusions regarding how, exactly, descriptive representation factors in. For some, it seems that they perceive an increase in substantive representation; perhaps they believe they will “win" more in the judiciary. Others are supportive of a shared race nominee regardless of potential policy payoff, but do not support the institution more. This suggests they still may derive symbolic benefits even if they may not “win" more. Overall, the results in Figures 3 and 4 prove inconclusive in adjudicating between the main benefits of descriptive representation.

External Validity Test: Comparing Jackson and Her Anonymous Profile

In this final empirical section, we consider the external validity of our results. We are curious how individuals react to Jackson, specifically, compared to her anonymous nominee profile. If we observe minimal differences in support among Black respondents for Jackson (the nominee) relative to their support based on her profile alone, then we can feel assured that our experimental findings have serious implications for actual nominees. However, if respondents are much more positive toward Jackson, specifically, than an anonymous profile with her prominent characteristics, we might conclude that our effects have more to do with the real-world nomination of a Black Judge than descriptive representation more broadly.

Recall that respondents answered questions about their demographic characteristics and then observed four randomized nominee profiles, each of which they evaluated in terms of their nominee support and support for the Court if the nominee were to be confirmed. After seeing four randomized conjoint profiles, all individuals evaluated a fifth, non-randomized judge profile. The profile described Jackson, but to respondents it appeared as another hypothetical nominee. Respondents evaluated the “hypothetical" nominee and answered the associated survey items. The experience was the same as with the first four randomized profiles, including the same prompt, question wording, and response options.

Respondents were then taken to a new survey page, where instead of being asked to consider a hypothetical nominee, they were told that “Ketanji Brown Jackson will likely be confirmed by the U.S. Senate to serve as a judge on the U.S. Supreme Court." In this instance, no profile was given, but they did evaluate Jackson as they had with every previous hypothetical nominee. Thus, this aspect of the survey differed from the others in that (1) no profile was provided and (2) respondents evaluated “Ketanji Brown Jackson" and the Court were she to join rather than a “hypothetical nominee" and the Court if the “hypothetical nominee" joined. The question wording and response options were identical to those used when evaluating hypothetical nominees.

We use evaluations of the anonymized profile containing Jackson’s characteristics and evaluations of Jackson herself (without any profile information) for our external validity test.²⁷ We argue that if respondents evaluate Judge Jackson and her anonymous profile similarly, then that is evidence that our results have some real-world generalizability. Our analytical strategy is to calculate the mean within-person difference in support for the anonymous profile containing Judge Jackson’s characteristics and support for Judge Jackson herself, for both Black and white respondents. We also calculate differences in Court support using the same approach, but our emphasis is on nominee support because our theory is that racial representation directly affects judge support, with only downstream effects on Court support.

Figure 5 contains a plot of mean differences in judge support (top half of plot) and Court support (bottom half of plot) when comparing evaluations of Jackson herself to evaluations of her anonymized profile. We disaggregate the results by respondent race as shown by the different shapes of the point estimates. Larger values indicate that respondents expressed greater support for Jackson or the Court when told Jackson was nominated, relative to her anonymized profile. Horizontal lines indicate 95% confidence intervals.²⁸

Fig. 5

Mean differences in support is the level of support in response to the Jackson appointment minus the level of support in response to the appointment of “hypothetical nominee" with her characteristics (but no name attached to profile)

We are most interested in the top half of Fig. 5 showing the results for nominee support. We find that white respondents (gray circles) are more favorable toward the judge when evaluating Jackson, relative to her anonymized profile. That is, these respondents seem to like Jackson (and the Court, should she join) more than an unnamed Black Judge with Jackson’s exact characteristics. For Black respondents (black triangles), however, there is no significant difference between evaluations of Jackson and her anonymized profile on judge support. For Black respondents, the downstream effects appear to be slightly more positive with Jackson than they were with her profile, perhaps because they were evaluating a Black nominee with a tangible and observable opportunity actually to join the Court. For white respondents, we also see more Court support with the Jackson nomination, but it appears that those downstream effects are due to something other than her race.

In all, this analysis suggests that Black respondents’ positive evaluations of Jackson (and the Court she would join) are reflective of those we observed in our experiment, whereas white respondents are responding in a limited fashion and for reasons other than descriptive representation (perhaps specifically about Jackson). We admit it is possible that some respondents brought external information to bear on their evaluation of Jackson that they did not bring to evaluating the anonymous profile. Perhaps they learned something about her upon nomination. However, we think the differences across race in our context do highlight differential reactions that indicate descriptive representation is likely at play.²⁹

Conclusion

In this article, we set out to understand better how descriptive representation affects individuals’ evaluations of the United States Supreme Court via their support for nominees. We demonstrate that Black respondents more positively evaluate same race nominees, though white respondents do not exhibit much in the way of differences across nominees based on nominee race. Importantly, for Black respondents, we also demonstrate that descriptive representation indirectly influences views of the Court. We also find that these effects are heterogeneous across respondent partisanship. Lastly, we leverage a comparison of Jackson to an anonymized version of Jackson to determine if our experimental results port to real-world scenarios; we find that they do.

In all, the data suggest that descriptive representation is a potential mechanism for bolstering Court support. This is useful in polarized times where support appears to have dropped (Armaly & Enders, 2022) and in the face of the rapidly changing demographic composition of the country. Indeed, providing representation—especially for those who have historically lacked it—positively impacts the institution, without meaningfully decreasing support among white respondents, who have always been, and still are, well-represented on the bench. This allows us better to contextualize existing research on descriptive representation in the federal judiciary. For example, Scherer and Curry (2010) indicate that white respondents are less supportive of courts with many Black Judges; we find no such effect. We suppose that this is because of individual versus collective descriptive representation, and the ways that people assess the judiciary (i.e., during nominations, when individual-level representation is likely to be operative). Or, perhaps one additional Black nominee is insufficient to turn such individuals against the Court, though clear over-representation is sufficient. This is an important avenue for future research to examine.

We recognize that these findings have some limitations. We cannot make claims about the American mass public or Black and white racial groups as strongly as we may wish, given our sample and the nature of our research design. Nevertheless, Lucid samples are consistently used in judicial politics research [e.g., Armaly (2021) and Krewson (2023)], and are known to capture population attitudes well (Coppock & McClellan, 2019). Moreover, our oversample of Black Americans allows us to understand with greater power (relative to other samples) how this nomination influences Black Americans’ views of the Court.

Evidence suggests that support for the Court via confirmation of Justice Jackson is not simply because of her novelty (Schwindt-Bayer & Reyes-Housholder, 2017); however, we cannot speak to how long these positive feelings last, particularly among the group gaining representation. That being said, even if the effect is only temporary, it can have lasting effects on future generations and their political ambition by seeing someone who looks like them in a position of power (Fraga et al., 2020; Gilardi, 2015; Wolbrecht & Campbell, 2007). This is consequential given that women and people of color possess high levels of judicial ambition (Jensen & Martinek, 2009; Williams, 2008) but also view high barriers to entry (Melaku, 2019; Rhode, 1991). Future research might examine if the effects only exist when the nominee is salient (e.g., during the nomination) or persist into the future independent of Court outputs, how descriptive representation may influence those gaining representation, and whether the mediated effects we uncover travel to institutions like Congress,³⁰ as well as minority status and the intersectional effects of representation in the judiciary (Gershon et al., 2019). Others should consider if these effects depend on exposure to or knowledge of the nomination. Additionally, are increases in positivity a function of symbolic representation, perceived substantive representation, or what? We encourage scholars to consider these and related questions.