Experience

Political leaders are often selected based on experience (e.g., Reference Chiru and GherghinaChiru and Gherghina 2017). There are several reasons for this practice. As an initial matter, it is common to associate experience with expertise (Reference Rusbult, Insko and LinRusbult, Insko, and Lin 1995). To the extent these concepts overlap, choosing the most experienced candidate may benefit institutional performance. Allocating rewards based on experience can also enhance external perceptions of legitimacy (Reference FischerFischer 2008, 168–69). Internally, relying on experience values longevity and incentivizes stability. It also mitigates group conflict by minimizing decision costs and reducing discretion that may otherwise facilitate biased decision making. Avoiding conflict over small-group leadership choices is important for managing collegial relations, particularly because common dispute resolution mechanisms such as bargaining and third-party interventions are less applicable in these contexts (Reference Baron, Murphy and SaalBaron 1990).

The effect of experience on leader selection may be nonlinear. Evidence on acclimation suggests new judges sometimes face a steep learning curve (Reference BoyeaBoyea 2010; Reference Hettinger, Lindquist and MartinekHettinger, Lindquist, and Martinek 2003; Reference Hurwitz and StefkoHurwitz and Stefko 2004; but see Reference Heck and HallHeck and Hall 1981). In light of this potential performance deficit, relatively inexperienced judges are unlikely to be considered good candidates for leadership positions. At least initially, the probability of selection should be expected to increase with additional years of service. This experience allows for more effective peer evaluation, increased political visibility, and enhanced skill concerning the court’s internal and external workings. However, there may be diminishing and eventually negative returns to additional experience. Longer tenure, for example, may increase the threat of reduced mental capacity and declining performance. And of course it increases the probability of institutional exit (Reference BoyeaBoyea 2011; Reference Curry and HurwitzCurry and Hurwitz 2016; Reference HallM. G. Hall 2001). As a result, the probability of being selected for a state judgeship increases with additional years of experience before steadily declining (Reference GoelzhauserGoelzhauser 2018b). Likewise, additional years of experience may initially increase the probability of being selected chief, with diminishing and eventually negative returns.

Political officials may also value seniority as an objective candidate sorting heuristic. There is evidence, for example, that the practice of assigning congressional committee chairs based on seniority persisted even after the norm ostensibly eroded (Reference Deering and WahlbeckDeering and Wahlbeck 2006; Reference Polsby, Gallaher and RundquistPolsby, Gallaher, and Rundquist 1969; but see Reference CannCann 2008). In the judicial context, seniority norms for rotating leaders are often conditioned by prior service. Federal law dictates that the most senior circuit court judge who has not been chief assumes office at the end of an incumbent’s seven-year term, and numerous states have similar rules. Absent such a rule under choice by peer vote, default to a seniority norm would be understandable as a way to manage group conflict. Seniority is likely to be one of the few objective sources of overlapping consensus (Reference RawlsRawls 1987) concerning valued leader characteristics. As a proxy for leader quality, seniority is over- and under-inclusive, but that may be an acceptable tradeoff when it allows heterogeneous group members to reach an incompletely theorized agreement (Reference SunsteinSunstein 1995) about valued characteristics rather than arguing from first principles during each selection event or depleting group capital by making contentious choices. Moreover, default to seniority rotation would be sustainable if group members believe they will benefit from the norm in the future.

• Experience Hypothesis: The probability of being selected chief increases with additional years of service before declining.

• Seniority Hypothesis: The most senior justice who has not been chief is most likely to be selected chief.

Bias

Discretionary voting institutions may disadvantage members of marginalized groups such as women and people of color. The evidence from legislatures is mixed. A recent study finds no evidence that women and people of color are differentially excluded from legislative leadership positions (Reference Hansen and ClarkHansen and Clark 2020). With respect to committee type, however, one study finds that women are as likely to chair “important fiscal committees” (Reference DarcyDarcy 1996, 888), while another suggests they are less likely to chair politically prestigious committees (Reference Fouirnaies, Hall and PaysonFouirnaies et al. 2018). Women in the British House of Commons are more likely to be voted committee chairs by peers (Reference O’BrienO’Brien 2012). More generally, there is substantial evidence that women are less likely to be promoted to workplace leadership positions (e.g., Reference Blau and DevaroBlau and Devaro 2007; Reference Smith, Smith and VernerSmith, Smith, and Verner 2013; Reference Yap and KonradYap and Konrad 2009). As for mechanisms, role congruity theory demonstrates that women are often perceived to be less effective leaders as a result of purportedly lacking traits commonly associated with effective leadership such as aggressiveness and assertiveness (Reference Eagly and KarauEagly and Karau 2002). Moreover, adherence to gender stereotypes leads to biased evaluations of women, resulting in a competence discount that hinders workplace advancement (Reference HeilmanHeilman 2001).

Mirroring other societal contexts, women and people of color have long endured formal and informal barriers to equality in the legal profession. Although many explicitly disadvantageous rules such as prohibitions on women practicing law have been eliminated in the United States, subtler forms of discrimination persist. Examples of relevant empirical findings concerning law and courts include evidence that attorneys who are women or people of color are more likely to report being treated unfairly (Reference Collins, Dumas and MoyerCollins, Dumas, and Moyer 2017); judges who are women or people of color receive worse performance evaluations from members of the legal community (Reference GillGill 2014; Reference Gill, Lazos and WatersGill, Lazos, and Waters 2011); women are less likely to be advanced by nominating commissions under merit selection (Reference GoelzhauserGoelzhauser 2018b); federal district court nominees who are black or women receive lower competence ratings from the American Bar Association (Reference SenSen 2014); and black federal district court judges are more likely to be overturned on appeal (Reference SenSen 2015). ^{Footnote 3}

Bias may be explicit or implicit. Implicit bias arises from discriminatory attitudes and stereotypes that operate outside of conscious awareness. Examining judicial performance evaluations, Reference GillGill (2014, 276–77) demonstrates that implicit bias may be particularly prevalent “in hiring-related decisions where the characteristics that are stereotypical for the job are at odds with the gender or race stereotype,” and where “subjective, vague, or abstract evaluation criteria” allow for evaluatory gap filling based on attitudes and stereotypes. The chief justice position, for example, may be associated with increased time demands, and women who are attorneys are perceived to be less committed to the time-intensive nature of high-profile legal employment (Reference Rivera and TilcsikRivera and Tilcsik 2016). Being chief may also be associated with a need for assertiveness that women are perceived to be less likely to possess and judged more negatively for exhibiting (Reference Park, Yaden, Schwartz, Kern, Eichstaedt, Kosinski, Stillwell, Ungar and SeligmanPark et al. 2016). Explicit bias can be difficult to detect, but former Wisconsin Supreme Court justice David Prosser made headlines when he told his colleague Shirley Abrahamson, “You are a total bitch,” while Justice Ann Walsh Bradley reported that Prosser “put his hands around my neck, holding my neck as though he were going to choke me” (Reference CaplanCaplan 2015).

Racial bias allegations were raised in a Louisiana chief justice selection dispute. Some historical context is essential for understanding the dispute. At one point, the predominantly black Orleans Parish had been combined with three predominantly white parishes to elect two Louisiana Supreme Court justices while other parishes constituted single-member districts. Finding it “particularly significant that no black person has ever been elected to the Louisiana Supreme Court” (Chisholm v. Edwards 1988, 1,058), the U.S. Court of Appeals for the Fifth Circuit held that the state’s districting scheme violated the Voting Rights Act by diluting the black vote in Orleans Parish. To satisfy a consent decree, Louisiana created a new Court of Appeal seat with the elected judge to be assigned to the Supreme Court, increasing the latter’s size from seven to eight (Perschall v. Louisiana 1997). ^{Footnote 4} Nobody received a majority in the ensuing three-person Democratic primary, resulting in a planned runoff between the white first-round winner and a black second-place candidate. ^{Footnote 5} Before the runoff, however, pressure mounted to ensure that a black judge filled the position in light of its purpose, and the white first-round winner withdrew arguing that her candidacy “threatened to ‘permanently scar’ race relations in New Orleans” (Orlando Sentinel 1994).

Ultimately, Bernette Joshua Johnson, a black woman, was elected to the newly created Court of Appeals seat in 1994 and immediately assigned by statute to the Louisiana Supreme Court, becoming the first woman of color to hold such a position (Reference Goelzhauser, Solberg, Disascro and WaltenburgGoelzhauser 2020). Several months later, a white man named Jeffrey Victory was elected to the Supreme Court. Justice Johnson ran unopposed for reelection to a regular Supreme Court seat after redistricting in 2000. As part of that redistricting, the number of high court seats was reduced back to seven and a law passed stipulating that “[a]ny tenure on the supreme court gained by [the judge elected to fill this position] while so assigned to the supreme court shall be credited to such judge” (see Chisholm v. Jindal 2012, 706). Notwithstanding this provision, a dispute later arose concerning whether Johnson or Victory was entitled to be chief justice under the state’s seniority rotation rule. Arguing that she was “being challenged because she’s an African-American” (quoted in Reference Campbell-RockCampbell-Rock 2012), Johnson filed suit in federal court, where a judge subsequently ruled that Johnson’s tenure began with her initial election and that she was therefore entitled to be chief justice. Although this anecdote arises under a different chief justice selection system, it illustrates the high stakes associated with position allocation and how bias can impact collegial court decision making. ^{Footnote 6}

• Women Hypothesis: Women are less likely to be selected chief.

• People of Color Hypothesis: People of color are less likely to be selected chief.

Politics

Political dynamics are important for leader selection. In the formative paper on chief justice selection by peer vote, Reference Langer, McMullen, Ray and StrattonLanger et al. (2003) develop a theory that emphasizes internal and external political dynamics. Internal political dynamics may be paramount in discretionary chief justice selections. In addition to being administrative leader, state supreme court chief justices often exercise opinion assignment authority (Reference HallM. G. Hall 1990; Reference Hughes, Wilhelm and ViningHughes, Wilhelm, and Vining 2015; Reference Kaheny, Szmer and ChristensenKaheny, Szmer, and Christensen 2019). Moreover, they can influence the production of separate opinions (Reference Brace and HallBrace and Hall 1990; Reference Hall and BraceM. G. Hall and Brace 1992; Reference Hall and WindettM. E. K. Hall and Windett 2016), which may impact perceptions of legitimacy. As Reference Langer, McMullen, Ray and StrattonLanger et al. (2003, 660) put it, “associate justices want a chief justice who has a moderate ideological tenor,” and “the median member is presumed least harmful to the other justices’ policy goals.” ^{Footnote 7} Consistent with this expectation, they find that justices are less likely to be selected chief as their ideal point diverges from the court’s median.

In addition to revisiting this hypothesis, we build on Reference Langer, McMullen, Ray and StrattonLanger et al.’s (2003) theory by considering collegiality as an important manifestation of internal political dynamics. Specifically, we consider the propensity to dissent. As a former chief judge of the U.S. Court of Appeals for the Seventh Circuit once explained, “Maintaining collegiality [on multimember courts] requires continuous efforts at minimizing sources of irritation—such as dissents” (Reference PosnerPosner [2008] 2010, 33). Or as U.S. Supreme Court justice Louis Brandeis put it to then-Professor Felix Frankfurter, the “great difficulty of all group action of course is when and what concessions to make. Can’t always dissent” (quoted in Reference UrofskyUrofsky 1985, 309). Although dissents can generate long-term legal change, they can also fray short-term relations by undermining a court’s positions and generating more work for opinion writers (Reference Epstein, Landes and PosnerEpstein, Landes, and Posner 2013). ^{Footnote 8} In 2015, Wisconsin altered state law to allow for chief justice selection by peer vote in order to oust longtime Chief Justice Shirley Abrahamson. Although the decision was politically motivated, some colleagues lamented her frequent dissents, with former Justice Louis Butler noting, “I was looking to build majorities while I think she was trying to set forth her view in the law,” adding, “She always saw herself as a one of seven and had the right to do that” (Reference FerralFerral 2016).

There are several mechanisms by which a higher dissent may lead to a lower likelihood of being selected chief justice. First, consistent with Posner’s irritation observation, those who do not “go along to get along” may not be well liked. More dissents may also signal a lack of ability or desire to foster collegial relations on a multimember court. Given that “the chief justice is a figure of authority on a state supreme court” (Reference Vining and WilhelmVining and Wilhelm 2010, 712), there may also be concern about the public perception of the chief seeming out of step with the court majority. Last, as exemplified by Chief Justice Harlan Fiske Stone’s purported influence on the U.S. Supreme Court (Reference Corley, Steigerwalt and WardCorley, Steigerwalt, and Ward 2013, 29–32; Reference HaynieHaynie 1992; Reference Walker, Epstein and DixonWalker, Epstein, and Dixon 1988), there may be concern that a leader who sets an institution’s tone by frequently dissenting generates spillover effects that have deleterious consequences for collegiality and performance.

Externally, chief justices:

In Iowa, for example, former Chief Justice Mark Cady lobbied state legislators for a funding restoration to prevent court closures and reduced services, to which one legislator replied, “We will do our damnedest to get you to where you need to be” (Reference RussellRussell 2017). And in Tennessee, Chief Justice Jeff Bivens expressed interest in criminal justice reform by noting, “I am proud to say that I am in talks with the governor, [house speaker, lieutenant governor], and chairmen of the legislative committees,” adding, “We are very close to announcing the formation of a statewide task force that will undertake this effort” (Reference BennettBennett 2017). Reference Wilhelm, Vining, Boldt and BlackWilhelm et al. (2020) find that state legislatures are less likely to approve judicial requests as the ideological distance between them and the chief justice increases. As a result of external political dynamics, Reference Langer, McMullen, Ray and StrattonLanger et al. (2003, 660) argue that “the ability of the chief justice to create and maintain good relations with [the legislature and governor] affects institutional goals.” Consistent with this expectation, they find that justices who are ideologically distant from important state officials are less likely to be selected chief by their peers. We expect the same.

• Internal Compatibility Hypothesis: Justices are less likely to be selected chief as their ideological divergence from the court’s median increases.

• Dissent Hypothesis: Justices are less likely to be selected chief as their propensity to dissent increases.

• External Compatibility Hypothesis: Justices are less likely to be selected chief as their ideological divergence from executive and legislative officials increases.

Modeling Strategy

Previous research analyzing chief justice selections by peer vote employ duration analysis, though with fundamentally different approaches. Reference Norris and TankersleyNorris and Tankersley (2018) analyze the time until peer-vote states seat a woman as chief justice from 1970 through 2008, with the risk set defined as state-years in which there is a peer vote. With state-year as the unit of analysis, this approach emphasizes state-level as opposed to individual-level variation. Indeed, with only one observation per state-year, this approach does not allow for examining the impact of candidate characteristics on selection outcomes. Reference Langer, McMullen, Ray and StrattonLanger et al. (2003) use a sample of state supreme court justices who first came to the bench between 1970 and 1996 to model the time until they were first voted chief by their peers. The unit of analysis is the justice-year, and the risk set is defined as state-years for justices eligible to win a peer-vote contest.

Our approach is different. The use of duration analysis in previous research indicates that the primary interest was in modeling the time until an event occurs—either selection of a woman (Reference Norris and TankersleyNorris and Tankersley 2018) or any eligible justice (Reference Langer, McMullen, Ray and StrattonLanger et al. 2003). In contrast, we are interested in understanding whether certain candidate characteristics are associated with chief justice selections from fixed choice sets. Although Reference Langer, McMullen, Ray and StrattonLanger et al. (2003) include two individual-level predictors in their analysis, their modeling strategy compares every justice in the sample to one another (across states and time) rather than comparing justices to colleagues on the same court at the time of selection. In contrast, similar to methodological advances in the legislative context (Reference CannCann 2008), we are interested in the outcomes of particular peer-vote contests, examining chief justice selections from the pool of sitting justices on a court who collectively comprise the choice set at a given point in time.

Social scientists are often interested in modeling selection from fixed choice sets. In the canonical paper, Reference McFadden and ZarembkaMcFadden (1973) analyzes shopping behavior, including transportation mode and destination. In political science, scholars have analyzed phenomena such as selection of a coalition government (Reference Martin and StevensonMartin and Stevenson 2001) and legislative party switching (Reference DesposatoDesposato 2006). With respect to judicial politics, this modeling strategy has been employed to understand opinion assignment (Reference Farhang, Kastellec and WawroFarhang, Kastellec, and Wawro 2015), panel assignment (Reference Hausegger and HaynieHausegger and Haynie 2003), and merit selection (Reference GoelzhauserGoelzhauser 2018b). As with those institutional settings, we are interested in modeling the selection of a particular individual from a fixed choice set rather than pooling justices across states and time. Thus, our estimation strategy leverages within-contest variation for identification.

Data and Measurement

Our original data span 1965 through 2016 and includes more than 200 selection events across 18 states and 19 apex courts that select chief justices by peer vote. ^{Footnote 9} For each contest, the outcome variable is scored one for the justice selected chief and zero for others in the choice set. Although conditional logit is well suited for modeling selection from fixed choice sets, marginal effects can only be estimated with the fixed effect set equal to zero, which potentially generates misleading and meaningless quantities of interest (Reference BeckBeck 2015; Reference Karaca-Mandic, Norton and DowdKaraca-Mandic, Norton, and Dowd 2012; Reference KatzKatz 2001, 380). Alternatively, standard models with grouping-unit fixed effects perform well in simulations as conditional logit proxies while allowing for the recovery of meaningful substantive effects (Reference BeckBeck 2015; Reference CoupeCoupe 2005; Reference KatzKatz 2001). As a result, we present results from a logit model with contest fixed effects. Table A1 in the appendix displays results from conditional logit and linear probability model specifications. The results are similar.

Consistent with our hypotheses, we include the following explanatory variables. Our seniority variable is scored one for the longest serving justice who has not been chief and zero otherwise. To account for aggregate experience, we include variables counting number of years served on the apex court and number of years squared. The quadratic term permits testing for concavity. The discrimination hypotheses are evaluated with an indicator scored one for women and zero otherwise, and a second scored one for justices of color and zero otherwise. We account for partisan dynamics with three measures available for the length of the sample period. First, we include an indicator scored one for justices who are copartisans with the court median and zero otherwise. Second, we include an indicator scored one for justices who are copartisans with a unified state legislature and zero otherwise. Third, we include an indicator scored one for justices who are copartisans with the governor and zero otherwise. The robustness section considers alternative but time-limited measures. Last, we use opinion authorship data from Nexis Uni to measure a justice’s propensity to dissent. Specifically, we divide the total number of majority and dissenting opinions written or joined the previous year by the total number of dissenting opinions written or joined that year, multiplying quotients by 100 to generate percentages.

One of the benefits of our modeling strategy is that contest fixed effects sweep up observable and unobservable intra-contest heterogeneity. Thus, as (Reference Farhang, Kastellec and Wawro2015, S74) note in an analogous setting, “the model implicitly accounts for any factors that do not vary within a [contest] (essentially holding them constant) even though we do not explicitly estimate parameters for such variables.” In our context, the contest fixed effects account for state-level variables such as opinion assignment rules, court professionalism, and retention mechanisms that may be correlated with candidate characteristics and selection outcomes. As a result, the threat of omitted variable bias is diminished. Moreover, this approach accounts for time-varying effects with respect to predictors such as gender and race since effects are calculated based on intra-contest win probabilities (cf. Reference Farhang, Kastellec and WawroFarhang, Kastellec, and Wawro 2015, S74). In short, this estimation strategy accounts for any factor that is constant within a given contest. Leveraging within-contest variation to identify the effect of individual-level characteristics on the probability of selection lends credibility to the resulting inferences.