Electoral institutions, support for judges, and legitimacy

A voluminous literature studies the impact judicial elections have on both judges and the Americans tasked with selecting them. Scholars debate the degree to which elections effectively generate political competition and engage voters in the electoral process. Studies illustrate high incumbent reelection rates (Streb and Frederick Reference Streb and Frederick2009; Bonneau and Cann Reference Bonneau and Cann2011) and low contestation rates (Nelson Reference Nelson2010), and uncover evidence of a significant incumbency advantage (Olson and Stone Reference Olson and Stone2023). However, certain conditions, including high levels of campaign spending or advertising and partisan elections, improve the success of challengers (Hall Reference Hall2014) and generate voter enthusiasm for and knowledge about elections (Bonneau and Hall Reference Bonneau and Hall2009). Given the importance of legitimacy to judiciaries, studies have examined the impact elections have on legitimacy evaluations, with debate over whether and under what conditions elections improve or harm legitimacy (Cann and Yates Reference Cann and Yates2008; Gibson Reference Gibson2012; Woodson Reference Woodson2017).

Existing studies place relatively little emphasis on studying how selection methods are associated with support for a judge, attitudes that are fundamental for understanding representation and accountability. Scholarship that studies evaluations of judges in judicial elections tends to focus on aggregate outcomes rather than individual survey respondent attitudes – investigating how candidate (e.g., quality or campaign resources) or district (e.g., crime rates) characteristics shape district or state-level voting behavior. This contrasts with the emphasis studies of the federal judiciary place on measuring support for individual judges (Sen Reference Sen2017; Badas and Stauffer Reference Badas and Stauffer2018; Kaslovsky, Rogowski, and Stone Reference Kaslovsky, Rogowski and Stone2021) and on the link between individual judges and the judiciary’s broader institutional standing (Krewson and Schroedel Reference Krewson and Schroedel2020; Carrington and French Reference Carrington and French2021; Glick Reference Glick2023). Similarly, studies of legislators or executives emphasize the importance of measuring support for or approval of these actors (e.g., Canes-Wrone and De Marchi Reference Canes-Wrone and De Marchi2002; Ansolabehere and Kuriwaki Reference Ansolabehere and Kuriwaki2022). Institutional design, including electoral institutions, can shape public evaluations of politicians (e.g., Fox and Jordan Reference Fox and Jordan2011; Coll Reference Coll2021), with (would-be) voters possibly valuing selection methods perceived as fairer or more democratic for either intrinsic or instrumental reasons (Grimes Reference Grimes2006; Schwenk Reference Schwenk2024).

There are a number of reasons to expect that selection methods may affect voters’ evaluations of judges, even outside of the context of an electoral campaign. Efficacious voters may relish the opportunity to select public servants (Rosenstone and Hansen Reference Rosenstone and Hansen1993). Voters may have experience with democracy and democratic processes, good or bad, that lead them to differentially evaluate public officials selected under different institutions (Streb and Frederick Reference Streb, Frederick and Streb2007; Karp and Milazzo Reference Karp and Milazzo2017). Constituents may infer that local elected officials are more ideologically proximate to them than those appointed by state-level officials (Wilson and Gronke Reference Wilson and Gronke2000). We elaborate on these mechanisms more below in our discussion of Study Two.

The effects of selection method may extend beyond the individual judge to perceptions of judicial legitimacy. Legitimacy is relevant for a broad class of institutions and authorities that exercise influence over others (Tyler Reference Tyler2006). Political institutions including legislatures (Hibbing and Theiss-Morse Reference Hibbing and Theiss-Morse2002), the police (Peyton, Sierra-Arévalo, and Rand Reference Peyton, Sierra-Arévalo and Rand2019), and the judiciary (Gibson Reference Gibson2009) rely on legitimacy as a source of power. Legitimacy is an especially pressing concept for courts as they regularly rely on other political actors for the implementation of their decisions (Gibson and Nelson Reference Gibson and Nelson2014). Given the role that elections play in conferring legitimacy to political institutions (Anderson et al. Reference Anderson, Blais, Bowler, Donovan and Listhaug2005), it is important to assess the relationship between selection methods and legitimacy. Understanding this relationship provides leverage to assess the normative consequences of institutional choice.

One may expect that elected courts accrue greater legitimacy than unelected courts due to the built-in boost to legitimacy that elections provide. Alternatively, elected courts may be viewed as less legitimate than unelected courts if expectations for how judges ought to behave are violated via campaigning (Gibson Reference Gibson2012), or if unelected courts can effectively build legitimacy by avoiding decisions that erode it (Gibson and Nelson Reference Gibson and Nelson2014). In a third view, selection method is not significantly associated with different levels of legitimacy if both elected and unelected courts are able to maintain necessary levels of legitimacy to ensure compliance with their decisions, even if through different routes.

Empirical debate persists about what impact judicial elections have on how voters evaluate state court legitimacy. Cann and Yates (Reference Cann and Yates2008) find that respondents in partisan election systems report lower levels of diffuse support than those in appointed systems, although this negative effect weakens for respondents who report high knowledge of the judiciary. Woodson (Reference Woodson2017) finds that respondents in states with judicial elections report higher levels of judicial legitimacy than in appointment states if their states have low levels of judicial campaign activity, but lower legitimacy if states have high campaign activity. Contrastingly, Gibson (Reference Gibson2012) finds that living through a judicial election improves evaluations of legitimacy.

In this project, we contribute new insight into how judicial selection methods shape support for individual judges and evaluations of state court legitimacy. By focusing on how elections are associated with attitudes toward individual judges, we hope to clarify an important set of attitudes that have been understudied by existing scholarship. Methodologically, we employ a design – a survey experiment – that allows us to handle the challenge of endogeneity and clearly interrogate the causal relationship between selection methods and views toward judges and the judiciary as an institution. We therefore build on previous studies that compare states with different selection methods using a selection-on-observables approach (e.g., Cann and Yates Reference Cann and Yates2008; Woodson Reference Woodson2017) or examine within-respondent change over the course of a judicial campaign (Gibson Reference Gibson2012). By studying both attitudes toward individual judges and views on legitimacy, we contribute to a nascent line of scholarship at the federal level that investigates the relationship between attitudes toward individual judges and broader institutional legitimacy (Krewson and Schroedel Reference Krewson and Schroedel2020; Carrington and French Reference Carrington and French2021; Glick Reference Glick2023). We also provide evidence as to the mechanisms through which selection methods and evaluations of judges are linked.

Research design: Study One

In our first study, we assess how selection methods shape voter evaluations of judges and the judiciary. To do so, we conducted a conjoint experiment embedded in a nationally representative survey. The survey of 1,033 U.S. adults was conducted via the NORC AmeriSpeak Panel in July and August 2021. The descriptive characteristics of the sample benchmark closely to 2021 U.S. Census data; please see Table A.1 for sample information.

An experiment is well-suited as a research design to test our expectations about how electoral institutions shape voters’ evaluations of state trial court judges and the broader state judiciary. Experiments provide a high degree of internal validity for concluding that the effect of our concept of interest – the type of selection method – on voters’ attitudes is due specifically to the concept and not any confounding factors (e.g., judge ideology, behavior, or descriptive traits). Our setup allows us to assess the all-else-equal effect of moving from one selection method to another and to speak to the consequences of institutional design for an individual’s relationship with their judges and the broader judiciary. The context during which we conducted our experiment – outside of a particular electoral campaign – lends it a level of realism for studying how an individual’s experiences with the day-to-day behavior of a judge shapes their evaluation of that judge.

It is important to note that, given the pervasiveness of judicial elections as a method of selection in the American states, 92 percent of our nationally representative sample lives in a state that uses judicial elections to select at least some judges. While we can randomize respondents into conditions where they evaluate an elected and unelected judge, we cannot change the respondents’ real-world experience with judicial elections. One way to interpret our aggregate results, therefore, is that they are (largely) reflective of how individuals who have experience with judicial elections evaluate elected as opposed to appointed judges and judiciaries. Our study thus speaks to how reforms aimed at eliminating judicial elections may influence voters’ support for their judges and judicial system (and how judges and other political officials may need to adapt their behavior to the institutional setting in which they find themselves), a pressing question given the widespread use of elections and ongoing reform debates. In Study Two, we directly assess how real-world experience with elections shapes the value voters place on elections.

To begin, respondents were primed to think about state trial courts with this statement:

We felt it important to take this step for two reasons. First, we expected that respondents would typically gravitate toward the U.S. Supreme Court when thinking about judges. Therefore, we view this prime as important for ensuring that our study measures evaluations of state trial court judges. This proves useful for measuring support for these judges and the legitimacy of the state judiciary (as we do in this study) as well as in our tests of why these effects emerge (as we do in Study Two). Second, we sought to clarify for respondents the role trial court judges play in the judicial system; this also helps us obtain valid evaluations of support for the judge and state court legitimacy.

Following the prompt, we presented our respondents with a profile of a hypothetical judge that included a set of judge attributes. These characteristics included the judge’s gender (as implied by the judge’s name), race, tenure, partisan identity, and sentencing behavior – in short, possible dimensions upon which individuals in the American states will evaluate their trial court judges. For our primary manipulation, we varied the selection method with which the trial court judge was chosen – by partisan election, nonpartisan election, or gubernatorial appointment. Randomizing these different selection contexts – while controlling for other judge characteristics – allows us to test the relationship between selection methods and evaluations of judges. Table 1 presents the possible values of each judge attribute; respondents were randomly assigned to receive one of the values from each of the seven attributes.Footnote ³ The order in which the respondent saw the attributes was randomized to forestall the possibility of order effects influencing our findings.Footnote ⁴ Respondents only evaluated one judge profile, alleviating concerns about learning or satisficing from multiple evaluations of judge profiles (Bansak et al. Reference Bansak, Hainmueller, Hopkins and Yamamoto2018).

Table 1. Characteristics of Hypothetical Trial Court Judges and Contexts

Note: One value from each attribute was randomly assigned to respondents for each hypothetical judge.

After viewing the judge profile, we asked respondents outcome questions that allow us to assess the effect of our key manipulation (selection method) on judge and judiciary evaluations. First, we measure generic support for the judge they evaluated with a question about their level of support for this judge. This has been used in previous surveys to capture respondent feelings toward judges (Sen Reference Sen2017), incumbent elected officials (Rogowski and Stone Reference Rogowski and Stone2020), and policies (Ansolabehere and Kuriwaki Reference Ansolabehere and Kuriwaki2022) and speaks to the representation relationship between Americans and public officials.Footnote ⁵ While “support” is a fairly general concept, it features a number of positive attributes for our purposes: 1) it has a relatively overt political component (as opposed to “liking” a judge, for example); 2) it encourages a complete evaluation of the judge, as opposed to focusing on a particular characteristic; and 3) it is plausibly affected by any number of the conjoint attributes that we use. The question should therefore have a familiar meaning to respondents while also allowing respondents to weigh the different conjoint attributes as they see fit in making their evaluations.Footnote ⁶

We also measured respondents’ attitudes toward the legitimacy of their state judicial system using four questions common in political science research on judicial legitimacy (e.g., Gibson Reference Gibson2012), tailored to fit the state court context we study (three questions refer to the “courts in my state,” and one refers to the “state’s highest court”). Each legitimacy question was measured on a five-point scale from strongly disagree to strongly agree; we construct an additive index of these four questions (rescaled to range from 1 to 5) to provide an aggregate measure of respondents’ views of the legitimacy of their state courts.Footnote ⁷ Table A.2 presents the question wordings and summary statistics for each of the questions that we use as our outcome variables.

An important challenge when conducting experimental research is that the confines of a survey necessarily abstract from the real world. We took a series of steps to make our study realistic and bolster the external validity of our findings. By presenting respondents with profile of interest (in our case, a judge), we are able to more closely mimic the real-world evaluation process where individuals have to draw upon a variety of characteristics when evaluating judges, as opposed to designs that focus on and manipulate a single characteristic (Hainmueller, Hopkins, and Yamamoto Reference Hainmueller, Hopkins and Yamamoto2014). Additionally, the treatment effects estimated via conjoint experiments perform well when benchmarked against real-world behavior (Hainmueller, Hangartner, and Yamamoto Reference Hainmueller, Hangartner and Yamamoto2015). The characteristics we vary in our design are unlikely to result in any implausible combinations that might not occur in the real world. Nevertheless, we recognize that isolating the independent effect of selection method abstracts away from the real world; we know, for example, that elections influence how judges behave (e.g., Brace and Boyea Reference Brace and Boyea2008). While the internal validity our experimental design provides is important for understanding the impact of selection method, we acknowledge that the effects of reform may differ in practice if other factors also vary with or are affected by selection method.

Results: Study One

Elections increase support for judges

We begin by examining our support outcome variable. For estimation, we regress our outcome survey questions on indicator variables for the judge attributes described above; this provides estimates of the average marginal component effects (AMCEs) for each attribute – the independent effect that each attribute has on support, averaging over all other attribute combinations (Hainmueller, Hopkins, and Yamamoto Reference Hainmueller, Hopkins and Yamamoto2014).Footnote ⁸ These results pool across our full sample and are not conditioned on any characteristics of the survey respondent.

We present our results in Figure 1. We find that respondents prefer judges who are elected to appointed, on average. Support is systematically higher for judges whose conjoint profiles indicate that they were elected. We estimate separately the effect of partisan (0.12, p $ < $ 0.15) and nonpartisan elections (0.21, p $ < $ 0.02); while only nonpartisan elections are statistically distinguishable from appointments, the two election types are not distinguishable from each other and they both have a substantial positive effect on support. If we collapse our selection method variable into a binary “appointed” or “elected,” the difference is statistically distinguishable (0.16, p $ < $ 0.03). When considering the scale of our outcome variable, this equates to about a 4 percentage point boost in support.

The other elements of the conjoint profile, while not the central concern of our study, contribute to a growing literature that takes seriously voters’ preferences over a judge’s descriptive and political characteristics (Sen Reference Sen2017; Badas and Stauffer Reference Badas and Stauffer2018, Reference Badas and Stauffer2019; Kaslovsky, Rogowski, and Stone Reference Kaslovsky, Rogowski and Stone2021). We find little evidence that demographic characteristics, such as race or gender, shape respondent support on average.Footnote ⁹ Margin of selection is not associated with support, though longer tenures are, suggesting that respondents are more inclined to support experienced judges. The largest coefficient estimates in Figure 1 are for our one policy attribute, sentencing behavior. We find markedly lower support for judges who deviate below or above average sentencing behavior (-0.45 and -0.59, respectively, p $ < $ 0.01 for both); the average magnitude of these effects is roughly 3.25 times larger than the effect of selection method. This suggests that respondents generally value judges who follow sentencing norms and that support is partially a function of a judge’s behavior in office. Finally, we find no effects of partisanship, but these results average across both co-partisan and out-partisan respondents. We find similar results when assessing respondent evaluations of a judge’s fairness; see Figure B.1.Footnote ¹⁰

Figure 1. Judge Characteristics and Support.

Note: The figure presents estimates of how electoral institutions and judge characteristics shape evaluations of judge support (ranging from 1 to 5).

Of course, this aggregate analysis pools across a variety of respondent characteristics that may moderate evaluations of support – most obviously, partisanship (and its correlates).Footnote ¹¹ Respondents of different party affiliations may vary in their orientations toward candidates of color or women candidates for descriptive or substantive reasons (Sen Reference Sen2017; Badas and Stauffer Reference Badas and Stauffer2019; Kaslovsky, Rogowski, and Stone Reference Kaslovsky, Rogowski and Stone2021), and may have different policy preferences. To assess the partisan differences in evaluations of judges, we estimate the effect of each attribute separately by party. As this analysis splits our sample into three relatively small subgroups (451 Democrats, 350 Republicans, and 158 independents), we interpret these results with more caution than our aggregate results.

We present the results from this analysis in Figure 2. First, our selection method findings are consistent with our aggregate results; while we do find some differences in the magnitude of the treatment effect of elections across partisan groups (e.g., stronger results for Republicans and independents than Democrats), we do not wish to make too much of these differences as they are not, with one exception,Footnote ¹² statistically distinguishable from one another (e.g., the estimated effect of nonpartisan elections for Republicans is 0.19 larger than for Democrats, but p $ < $ 0.29 on this difference).

Figure 2. Judge Characteristics and Partisan Support.

Note: The figure presents estimates of the conjoint design on evaluations of support (ranging from 1 to 5) for the judge separately by respondent partisanship.

Next, partisans offer more support for judges that share their party affiliation. Democrats offer higher support for Democratic judges compared to Republican judges (0.60, p $ < $ 0.01), and Republicans offer higher evaluations of Republicans than Democrats (0.66, p $ < $ 0.01); this equates to about a 15 and 16.5 percentage point increase in support, respectively, and is approximately 4 times the magnitude of the effect of selection method.Footnote ¹³ As expected, Democrats punish above-average sentence length, while there is no statistically significant difference between below-average and average sentence lengths for these respondents. For Republicans, below-average sentences receive the lowest marks, though above-average deviations still garner lower support than average sentences do. While there are a number of attributes whose effects vary in predictable ways with the partisanship of the respondent, for the institutional feature of central theoretical interest – selection methods – we find limited variation across parties.Footnote ¹⁴

Selection methods do not impact judicial legitimacy

Our analyses thus far have focused on how selection methods shape respondent support toward a specific judge. We have yet to examine whether or how selection methods shape evaluations of the judicial branch more generally. To assess the link between selection methods and legitimacy, we draw upon the additive index of responses to four questions measuring state court legitimacy we discussed above.

Figure 3 presents results using the legitimacy index as the outcome measure, pooling across the respondent party.Footnote ¹⁵ The clearest conclusion to draw from this is that our results provide no evidence that selection methods and evaluations of specific judges color individuals’ attitudes toward the institution more generally. While a few judge attributes produce statistically significant, albeit substantively small, results, the findings do not generally adhere to the patterns seen for evaluations of judge support above. When viewed through the lens of institutional reform, our results suggest that pivoting from elections to appointment alters the direct representation relationship between voters and judges but not views of the broader institution.

Figure 3. Judge Characteristics and Evaluations of State Court Legitimacy.

Note: The figure presents estimates of the conjoint design on evaluations of state court legitimacy (ranging from 1 to 5).

Why do our findings here deviate from existing scholarship that demonstrates that perceptions of state court legitimacy are malleable (Cann and Yates Reference Cann and Yates2008; Gibson Reference Gibson2009, Reference Gibson2012; Woodson Reference Woodson2017)? One important point of deviation is our research design, which mimics a ceteris paribus change in selection method. It is possible that observational studies are confounded by other state- or jurisdiction-level features in a way that our experimental design is not. Additionally, perhaps voters view courts as equally legitimate when all other factors are held equal, but they are not in fact equal “in the wild” – appointed (elected) judges may, for example, behave systematically differently specifically because they are appointed (elected) (Gordon and Huber Reference Gordon and Huber2007). This connects to our theoretical discussion above that suggests both unelected and elected courts can act to accrue legitimacy. The strength and target of our treatment may also be a reason for this difference. Our design is hypothetical; this is substantively different from the impact of living through an election in the real world (Gibson Reference Gibson2012). Respondents received the profile of a trial court judge but evaluated the broader state judiciary’s legitimacy (questions about “courts in my state” or, for one question, “my state’s highest court”); exposure to a single judge profile may not be a strong enough treatment to move respondents’ attitudes about their state’s judiciary more generally. Indeed, Figure A.1 in the Supplementary Materials suggests that respondents’ answers to the judge-specific and judicial legitimacy questions were only slightly correlated. This does not imply, however, that other real-world factors such as salient reforms, exposure to campaigns, or controversial decisions do not shape perceptions of court legitimacy.

Exploring possible mechanisms: Study Two

Our first study illustrates that voters provide more positive evaluations of judges who are elected as opposed to appointed. However, our design does not allow us to test why these results hold. We conducted a second study in order to assess possible mechanisms through which elections lead voters to evaluate judges more positively than if they were appointed.Footnote ¹⁶ The core of our second study mimics the first. We conduct a conjoint experiment, embedded in a nationally representative survey, in which we present respondents with a judge profile that varies the same characteristics as in our first study, including our core manipulation of the selection method used to choose the judge. The new approach we take in this study is to also ask respondents a number of questions that allow us to tap into possible pathways through which elections improve evaluations of judges. Our second study was administered to 1,224 Americans via the NORC AmeriSpeak Panel in July 2023. The sample’s characteristics closely reflect U.S. Census data (see Table C.1); we again find reasonably good balance on respondent characteristics (see Table D.1).

We draw upon scholarship from the study of elections, political participation, and public opinion toward the judiciary to develop three possible mechanisms through which elections may improve evaluations of judges. First, we look to research on political efficacy, an individual’s belief that they can participate in and influence politics. This broader concept is broken into two types: internal efficacy, an individual’s belief in their own skill and capacity to participate in politics, and external efficacy, an individual’s belief that their participation can shape the behavior of political actors (Rosenstone and Hansen Reference Rosenstone and Hansen1993). Research has established a link between electoral institutions and efficacy (Knobloch, Barthel, and Gastil Reference Knobloch, Barthel and Gastil2020). We expect that individuals with lower levels of external efficacy will be more supportive of elections relative to appointment than those with higher levels of external efficacy, as these individuals should be the most dissatisfied with the level of control they have over politics and the most eager to exert greater control via elections. Furthermore, we expect that individuals with higher levels of internal efficacy will be more supportive of elections relative to appointment than those with lower levels of internal efficacy, as these individuals should be more confident about their ability to learn about judges when voting for them.

Second, we draw upon studies of the public’s experience with and evaluations of democracy. Popular support for democracy and democratic values plays an important role in ensuring the strength and persistence of democratic institutions and curbing antidemocratic elite behavior (e.g., Graham and Svolik Reference Graham and Svolik2020). Scholars find differing results with respect to how satisfaction with democracy is associated with participation. In some contexts, scholars find that populist attitudes – dissatisfaction with political elites and those in government – and parties can spur political participation (Huber and Ruth Reference Huber and Ruth2017). Other studies find that greater satisfaction with democracy is associated with greater levels of participation (e.g., Karp and Milazzo Reference Karp and Milazzo2017). In the context of judicial elections, positive experiences with elections are associated with increases in legitimacy (Gibson Reference Gibson2012), and participation in judicial elections can spur engagement with other races on a ballot (Reilly and Walker Reference Reilly and Walker2010). Support for and satisfaction with democracy may factor into the relationship between selection methods and support for judges, although these studies offer different predictions about how this relationship might manifest. Those dissatisfied with democracy and political elites may express greater support for elected judges as opposed to those appointed, as elections offer these individuals greater control over institutions with which they are unhappy. Alternatively, those with higher support for democracy and its related institutions may express greater support for elected judges, as they value democratic institutions like elections.

Individuals may also have actual experience with a particular set of judicial selection methods and react to their experiences under that institutional regime. This relates to the point we raised about our nationally representative sample: as over 90 percent of Americans (and respondents in both of our surveys) live in states with judicial elections, our respondents are generally familiar with these institutions. If residents of a particular place are used to electing (or appointing) judges and are generally satisfied with their state judiciary, they may be predisposed to prefer elected (or appointed) judges. There is reason to expect this is the case. For example, Gibson (Reference Gibson2012) finds that most voters in Kentucky rated the conduct of the 2006 judicial elections as appropriate and improved in their evaluations of the judiciary over the course of the election. Similarly, evidence from recent judicial reforms highlights that voters (and their elected representatives) do not generally aim to get rid of elections entirely, but instead make them work more effectively (e.g., Bonneau and Hall Reference Bonneau and Hall2009; Bonneau and Kane Reference Bonneau, Kane, Bonneau and Hall2016; Streb and Frederick Reference Streb, Frederick and Streb2007). This is reflective of a broader degree of voter resistance to institutional change, reflected, for example, in the low rate of success of ballot initiatives (Boehmke and Patty Reference Boehmke and Patty2007).

Finally, we develop expectations related to an individual’s preferences for judges who share their politics. The public values learning about the politics of their judges and wants judges to be chosen, at least in part, based upon their political records (Gibson Reference Gibson2012; Badas and Simas Reference Badas and Simas2022). Studies have shown that Americans use cues to infer a judge’s political views and behave accordingly in response to those cues (Sen Reference Sen2017; Badas and Stauffer Reference Badas and Stauffer2019). Relatedly, scholarship from outside of the judiciary shows that the public often projects their own views and beliefs upon political actors running for office with whom they favorably relate (Wilson and Gronke Reference Wilson and Gronke2000). Putting these lines of scholarship together, we expect that Americans may infer that judges who are elected, rather than appointed, more closely share their political beliefs and value this alignment, leading them to extend greater support to judges who are elected than appointed.

To assess these mechanisms, we ask a number of pre-treatment questions based on extant scholarship to measure individual efficacy, experience with democracy, and political views. We also determine a respondent’s real-world experience with judicial elections and ask post-treatment questions that allow us to measure respondent self-reported perceived political distance from the hypothetical judge they evaluated. We present information on the measures we use to investigate our proposed mechanisms in Table 2. We find considerable variation in respondents’ scores on each of our measures. This variation allows us to test how political efficacy, experience with democracy, and perceptions of the political closeness of the judge are related to respondent support for judicial elections.

Table 2. Assessing Mechanisms for Support for Elected Judges: Survey Questions

Note: The table presents information on the measures used to assess why judicial elections generate support for judges.

^a We also use an alternative measure based on high courts being elected or appointed (see Table F.1). Data for this variable come from both the 2021 and 2023 surveys.

Before we test our proposed mechanisms, we note that the primary conclusion we drew in our first analysis – that voters prefer judges who are elected to appointed – replicates in our second study (Figure D.1).Footnote ¹⁷ In this survey, the estimated treatment effect on the binary elections treatment was 0.12 (p $ < $ 0.07), indicating roughly a 3 percentage point increase in support for judges who were elected as opposed to appointed.Footnote ^18, Footnote ¹⁹

Internal and external efficacy

We begin with our assessment of the efficacy mechanism. Our methodological approach is to treat respondents’ answers to the pre-treatment efficacy questions as a continuous measure, and to interact this measure with our binary election indicator variable.Footnote ²⁰ If efficacy shapes support for elected versus appointed judges, we would expect our point estimates to systematically vary with respondents’ levels of efficacy or support for democracy. In other words, we would expect their responses on these pre-treatment questions to moderate their level of support for the judge. We are interested both in whether we find evidence for treatment effects for some values of the moderators but not others (whether the marginal effect of the election treatment is significant), and also whether the treatment effect of receiving an election condition significantly varies across levels of the moderator (whether the interaction between the moderator and the election treatment is significant). For ease of presentation, we present our results in the form of marginal effects plots (Hainmueller, Mummolo, and Xu Reference Hainmueller, Mummolo and Xu2019). These plots show the expected effect of receiving an election treatment condition in our conjoint design for respondents at different levels of the moderator values. Full regression results are presented in Table D.2 in the Supplementary Materials.

We present our efficacy results in Figure 4. In the top figure, we show that there is virtually no difference across the support of the responses to our external efficacy measure. The effect of receiving an election treatment is almost exactly our aggregate estimate – approximately 0.12 – across the support. We therefore find no evidence that respondents’ level of external efficacy is associated with variation in their valuation of elections. In the bottom panel, we find some evidence that respondents’ internal efficacy moderates their feelings about electing judges. Specifically, we find evidence that only respondents expressing confidence in their ability to understand politics express significantly higher levels of support for the elected judge in our conjoint profile. We note, however, that we cannot reject the null hypothesis that the slope in Figure 4b is zero (p < 0.20); in other words, we cannot statistically distinguish between the marginal effects at higher and lower values of efficacy.

Figure 4. Moderating Effects of Efficacy.

Note: Figure plots marginal effects of receiving an election treatment across values of pre-treatment moderators. Histogram on x-axis plots distribution of the moderator, with top bar indicating treated and lower bar indicating control units; 95% confidence intervals based on robust standard errors.

Experience with democracy

We employ two classes of measures to assess how experience with democracy may explain positive aggregate effects of elections on support for judges: self-evaluations of democracy and actual experience with judicial elections. We begin with self-evaluations using our pre-treatment questions from Table 2; we use the same methodological approach as for our efficacy questions. These results are presented in Figure 5. We find little evidence that beliefs that elections improve representation or that valuing local elections conditions the effects of the election treatment in any meaningful way. We do find some evidence that satisfaction with American democracy does so, with those respondents who view American democracy as less successful being most inclined to value electing a judge. As above, however, we cannot reject the null hypothesis that the effect is the same at all levels of the moderator ( $ p< $ 0.38).

Figure 5. Moderating Effects of Attitudes Toward Democracy.

Note: Figure plots marginal effects of receiving an election treatment across values of pre-treatment moderators. Histogram on x-axis plots distribution of the moderator, with top bar indicating treated and lower bar indicating control units; 95% confidence intervals based on robust standard errors.

Now we turn to the impact of real-world experience with judicial elections on evaluations of judges selected under different methods. We consider two different ways of thinking about familiarity with elections in the state judiciary: whether any judges in a state are elected and whether the state’s high court judges are elected (see Table F.1 in the Supplementary Materials for which states meet these criteria). We consider the treatment effect of receiving an elected judge profile among respondents living in states where these conditions either are or are not met. For this analysis, we set aside the other conjoint attributes and focus on a simple difference-in-means based on the selection method treatment;Footnote ²¹ because some of the subsets of states and respondents are quite small, we pool the 2021 and 2023 survey waves together for this analysis.

Our results are presented in Table 3. The top row presents our difference-in-means estimate of the treatment effect of elections for the full sample, to establish a baseline for this estimation strategy and pooled sample. The next two rows present split-sample estimates for our real-world experience measures. Across the two measures, most states and respondents fall into the “Yes” column, with a relatively small set of respondents in the “No” column. In our results, we find a consistent pattern. Among respondents in states that do elect judges, the election treatment is associated with substantially more support for the hypothetical judge (p-values below 0.001); among states and respondents who are less familiar with elected judges, the elected judge treatment is associated with substantially lower support for the judge (though these results are not as statistically reliable; p-values of 0.094 and 0.283).

Table 3. Effect of Election Treatment Based on Actual State Judicial Institutions

Note: Estimates are differences-in-means comparing those that received an elected judge profile relative to an appointed judge profile, pooling both our 2021 and 2023 surveys together; higher values indicate greater support for elected judges. Estimates in the “Yes” column are from states where the conditions indicated along the left margin are met; those in the “No” column are from states where that condition is not met. The unit of observation is the respondent.

We also replicate this exercise using our legitimacy outcome measure from Study One; we find null results for both groups (see Table F.2). We further assess whether differences in the treatment effect emerge for respondents who have experience with partisan as compared to nonpartisan elections; we describe these analyses and present results in Section F.1. We find limited substantive differences across these groups, though some evidence that the magnitude of the treatment effect is larger when respondents receive the type of election treatment that matches their state election institution. This suggests that the primary driver of our real-world experience results is general experience with electing judges, though specific types of electoral systems may also matter.

We emphasize that whether respondents live in states with or without elected judges is not randomly assigned, and because of the ubiquity of judicial elections, the set of states falling in the “No” column is both small and potentially distinctive on other dimensions (e.g., other institutions, political culture, the types of individuals who serve as judges). Nevertheless, these results provide evidence that individuals respond positively to judges chosen through methods with which they are familiar.

This provides three important takeaways for interpreting our results. First, it provides a suggestion that many Americans seem satisfied with their judicial selection methods. Second, it makes clear that our aggregate results from Study One should be understood as reflecting the scope condition that most respondents have real-world experience with judicial elections. While the randomization of our treatment ensures that we recover an unbiased treatment effect, that effect emerges from a sample that largely has experience with judicial elections. Third, it clarifies the stakes of real-world proposals to reform judicial elections. Changes to systems voters are familiar with lead these voters to evaluate judges more negatively, holding other relevant characteristics of the judge constant. In the real world, this suggests that constituents will alter their perceptions of judges if judges do not otherwise compensate through different behaviors.

Perceived ideological and policy proximity

Now, we turn to our analysis of our final mechanism – perceived political proximity. With these analyses, we assess whether respondents inferred that elected judges would be more ideologically similar to them, thus increasing their support for them. Our analyses are a tough test of this mechanism as we control for the information respondents received about the judge’s politics (e.g., partisanship and sentencing behavior) – thus, we isolate the relationship between the election treatment and assessments of proximity.

We conduct two different analyses to assess this question. First, we use perceived ideological proximity as an outcome variable to establish whether respondents perceive their distance to the judge to be lower in the election treatment condition. If proximity is a mechanism as we have proposed, we should see the election treatment predicting greater perceived proximity. Second, we control for ideological and punitiveness proximity in models otherwise similar to our base conjoint model.Footnote ²² If proximity is a mechanism through which the effects of election assignment on support are realized, this should attenuate the direct effect of election treatment assignment. We create two measures of proximity, one for ideology and one for punitiveness. For ideology, we take the absolute value of the difference between respondent self-reported ideology and respondent evaluations of the judge ideology. For punitiveness, we take the absolute value of the difference between respondent self-reported punitiveness (as measured by their answer to whether they feel those convicted of crimes spend too much, too little, or the right amount of time in prison) and that of the judge in the profile.Footnote ²³

We present our results in Table 4. We find little evidence that our elected judges findings are due to inferred ideological proximity. The leftmost column shows no evidence that perceived ideological proximity varies as a function of treatment assignment. In the right two columns, we incorporate our measures of perceived ideological proximity and punitiveness proximity as control variables. While their inclusion slightly attenuates our point estimates of the election treatment, we continue to find positive, substantively meaningful direct effects of election even while controlling for proximity. While proximity does have an effect on support (as expected), accounting for it does not wipe away the treatment effect of elections. We therefore find no evidence for the argument that perceptions of political proximity serve as a mechanism through which elections boost support for judges.

Table 4. Perceived Political Proximity and The Value of Judicial Elections

Note: Estimates are from OLS regressions with robust standard errors in parentheses. The unit of observation is the respondent. Punitive distance is a three-point measure (0-2); ideological distance is a five-point measure (0-4). Higher values indicate greater perceived distance.

^∗ p < 0.1, ^∗∗p < 0.05.