Authors: Thomas Gray and Banks Miller
Publication: American Politics Research 49: 319-327
Original source: final_version.pdf
Conversion note: this Markdown was generated by text extraction from the hosted file. Review equations, tables, notes, and citations before treating it as a final web copy.
https://doi.org/10.1177/1532673X21989019 American Politics Research 2021, Vol. 49(3) 319 –327 © The Author(s) 2021 Article reuse guidelines: sagepub.com/journals-permissions DOI: 10.1177/1532673X21989019 journals.sagepub.com/home/apr Article Questions of the quality and capacity of political leadership abound. Studies have analyzed mayors of major U.S. cities (e.g., Gerber & Hopkins, 2011) and U.S. Presidents (e.g., Canes-Wrone, 2006), as well as the effects of world leaders on economic growth (Jones & Olken, 2005). These questions are particularly acute when it comes to the leadership of collegial courts, like the U.S. Supreme Court and state supreme courts. This is because some leaders on these courts, known as chief justices or chief judges, are thought to be able to bring disparate judges together to form a consensus, while others foster dissensus (e.g., Walker et al., 1989). Consensus is important because courts presenting a united front are less susceptible to a host of potential interbranch threats, something that is particularly true for state supreme courts (Langer, 2002). State supreme courts are increasingly important venues for policy change and the role of leadership within them is understudied. Partially this is because the quantitative assessment of the qualities of any particular political leader, including chief judges, is difficult—some have said impossible (Hall & Windett 2016, p. 685). We investigate whether the personal qualities of leadership from the chief judges on U.S. state supreme courts helps to build consensus on those courts. By applying a newly created randomization-inference technique (Berry & Fowler, 2018), we overcome some of the difficulties previously thought insurmountable in the assessment of leadership effects. We find that there is no evidence that the personal leadership qualities of chief judges matter for a host of measures of consensus on state supreme courts. Our results have implications for the debate about leadership effects broadly, in that they tend to correlate with findings for U.S. Mayors. We find that institutional features of courts, such as court size and workload, are more likely consequential for consensus than who sits in the chief judge’s chair. Justice Rehnquist once likened the ability of the chief justice to control the associate justices as similar to the ability of one to control “hogs on ice” (Rehnquist, 1976, p. 637), a statement that accords with our findings on consensus. Our analysis is the first of a two-question process: (1) Is there a relationship between leadership and consensus? (2) If so, how?; and if not, why not? We answer the first question in the negative. As we explore below, there are two potential explanations for finding that chief judges do not systematically alter consensus on state supreme courts. First, it might be that chief judges are consistently motivated to create consensus, but fail in the attempt. Second, it might be that chief judges do not systematically try to foster consensus in the first place, rendering irrelevant the question of their capacity to do so. Our data are unable to distinguish between these explanations and this second question is ripe for additional analyses. In addition, we take advantage of the greater variation offered by studying state courts to analyze whether judges are more successful as chief judges compared to when they 989019 APRXXX10.1177/1532673X21989019American Politics ResearchGray and Miller research-article2021 1University of Texas at Dallas, Richardson, USA Corresponding Author: Banks Miller, University of Texas at Dallas, Political Science Program, 800 West Campbell Road, GR 31, Richardson, TX 75080, USA. Email: millerbp@utdallas.edu Swineherds and Hogs on Ice: Leadership Impacts for State Chief Judges Thomas Gray1 and Banks Miller1
Chief judges stand as visible leaders of their courts. Analyses of the Supreme Court focus on the role of the chief justice as an institution-builder seeking out public-facing consensus to protect Court legitimacy. Studying the powers of chief judges and political leadership in general is difficult. Analyzing all 50 states over 16 years we find no evidence that the identity of chief judges explains consensus behavior any better than random chance. This is true even among the subset of chief judges with additional institutional powers like opinion assignment. We show that court structures explain consensus, while leader features do not. Being chief judge correlates with an elevated likelihood of being in the majority, particularly in cases decided by one vote. These results add to our understanding of leadership on courts and imply that the office of chief judge at the state level is more symbolic than uniquely powerful. Keywords chief judge, state supreme court, consensus
320 American Politics Research 49(3) serve as a regular judge on a state supreme court. We find that judges serving as chief judge are more likely to vote with the majority of the court, particularly in cases decided by one-vote margins, than they are when they are serving as a regular judge. In addition, we show that this benefit accrues to the chief judge simply by virtue of holding the office and does not stem from special powers given to the chief judge, such as the ability to assign opinion-writing duties. However, these benefits are typically very small (less than five percentage points) and are estimated with precision due to the power provided by a large dataset with significant variation across states. Chief Judges and Court Consensus Scholarly investigations into the role of the chief justice in reducing dissent on the U.S. Supreme Court are legion (e.g., Caldeira & Zorn, 1998; Danelski, 1960; Haynie, 1992; Walker et al., 1989). Many of these investigations focus on the apparent collapse of consensual norms of decision making in the early 1940s, when the share of cases with a dissenting opinion increased 60 percentage points. Many previous works have suggested the leadership style of Chief Justice Stone was the primary culprit in the collapse of these consensual decision-making norms (e.g., Walker et al., 1989). Others have emphasized more institutionally oriented explanations, including giving the Court the ability to select its own cases and a change in the underlying composition of cases decided by the Court (e.g., Caldeira & Zorn, 1998; Hendershot et al., 2012). Yet distinguishing these explanations can be difficult because there are so few total chief justices against which to gauge leadership outcomes. 1 Considerably more variation is available when analysis moves to state supreme courts, where there are many chief judges (CJs) across all 50 states. Consensus on state supreme courts is likely to be particularly important because state supreme courts are more vulnerable to court-curbing measures than is the U.S. Supreme Court. For instance, the Brennan Center reported that in 2018 at least 18 states considered bills that would reduce the independence of state supreme court judges (Brennan Center for Justice, 2018). As one example, in North Carolina, the legislature introduced a bill targeting the Democratic CJ by reducing her access to assistance from judicial clerks and staff attorneys (Tiberii, 2019). More generally, state supreme court judges must be concerned about threats to their reelection (Langer, 2002) and reappointment (Gray, 2017, 2019) chances, to their ability to hear particular kinds of cases, and to their budgets (Langer, 2002). Indeed, less divisive state supreme courts may enjoy higher levels of public confidence (Sheldon, 1999) and better relationships with the legislature (Hunzeker, 1990). Given the potentially wide-ranging benefits of consensus, what role does the CJ play in its construction? The literature on state supreme courts has focused primarily on the ability of some CJs to assign who writes an opinion and, because writing a separate opinion is time consuming, the resources of the associate judges in explaining consensus on state supreme courts (Hall & Windett, 2016; Langer et al., 2003). An additional perk commonly granted to CJs in the states is the right to vote last at conference, as in the U.S. Supreme Court. Finally, some CJs are elected to the position by voters or their peer judges, while other CJs are placed in the position simply due to seniority or because of rotational practices. We also investigate whether variation in the power of the CJs correlates with their ability to create consensus. Yet these institutional explanations have bypassed the possibility that some state supreme court CJs might be better at creating consensus on their courts—something the literature on the U.S. Supreme Court has referred to as task and social leadership. Task leadership reflects the ability of the CJ to control discussion and to assign opinions, while social leadership is the ability to convince judges to suppress their dissents (Danelski, 1960). 2 John Marshall’s leadership abilities helped establish norms of consensus on the Court (e.g., Walker et al., 1989). Work pinning the collapse of consensual norms on the court to the leadership of Chief Justice Stone focuses on his disinterest in undertaking these types of leadership roles. A finding that being CJ does not correlate with consensus is behaviorally equivalent with two explanations: that they try and fail to achieve consensus and that they do not try in the first instance (e.g., Chief Justice Stone). 3 The first explanation, trying but failing, is built on the perceived benefits of consensus for courts, including greater legitimacy with the mass public (Zink et al., 2009), accordance with professional norms (Leonard & Ross 2014), and a reduced likelihood of interference from another branch of state government (Langer et al., 2002; Leonard & Ross, 2014). The second explanation, that CJs do not attempt to create consensus, is consistent with the notion that judges have a preference for leisure (e.g., Baum, 2010; Epstein et al., 2013) and that marshalling a court is time consuming and ultimately not likely to be successful. Further, given frequent rotation into and out of the position in state supreme courts, over-wrought efforts might harm collegiality (e.g., Brace & Hall, 1993). Therefore, by a simple cost/ benefit analysis, working for consensus may not be worth the effort. Although we cannot—and do not—distinguish between these two explanations, our analyses contribute to work on consensus on state supreme courts by narrowing down the possibilities and fixing the empirical fact that leadership itself appears to have no systematic correlation with consensus. That changes in leaders has little association with changes in consensus behavior is a significant fact, separate from whether that is due to incapacity or lack of desire on the part of leaders. Further, as explored below, we add to knowledge about how being CJ benefits the individual in the center chair.
Gray and Miller 321 Being Chief Judge and Voting in the Majority An additional question is whether CJs are themselves advantaged by the office. This advantage would most likely manifest in the ability of the CJ to vote with the winning side in cases. However, to empirically assess such a concept requires a baseline against which to measure success as a CJ. In short, we need to observe judges serving as both associate and chief judges within their careers. Whether a CJ is more successful in this way is rarely studied, possibly because of too much focus on the Supreme Court. Only one modern chief justice, Rehnquist, 4 fits the pattern required to assess the impact of being chief judge on winning. A close study of the correlation of being chief justice with Rehnquist’s success suggests that any extra influence by the chief justice is likely to be “contingent and conditional, rather than direct” (Cross & Lindquist, 2006, p. 1679). Most relevant to our investigation, Cross and Lindquist (2006, p. 1691) note that Chief Justice Rehnquist was five percentage points more likely to be in the majority when he was chief justice as opposed to when he served as associate justice. This suggests that being CJ may be personally beneficial to a judge, but it is difficult to know how generalizable such a result is given that the analysis includes only one judge. In total, we analyze the votes of 738 judges, of whom 211 served as chief judge at some point in the 16 years we analyze. Of those 211, 186 also served as associate judges. Assessing the potential of CJ leadership empirically is difficult, not least because the concepts of task and social leadership remain difficult to quantify and because the pattern of service required to assess individual impact is rare. What is needed is both more variation, solved by analyzing state supreme court judges, and a robust way of inferring leadership ramifications from observable data. Our analysis below proceeds in two distinct steps. First, we investigate whether some CJs are better able to foster consensus. Second, we turn to the question of whether being CJ is personally beneficial to judges. Data and Analysis Variation in the Chief Judge To understand the correlation of judicial leadership with court unanimity, we utilize randomization inference. Specifically, we employ the RIFLE technique developed by Berry and Fowler (2018). This method leverages leader fixed effects to assess the portion of the data variation nominally explained by leadership (through the conventional R 2), and then compares that to the “explanatory power” that could be obtained through random permutations of fictitious leader fixed effects. If leaders shape outcomes, then the true set of fixed effects should be far more explanatory than randomized (counterfactual) leadership assignments. In our specific case, if state supreme court CJs impact court collegiality, then the true ordering of chief judges should have a higher R 2 than a set of randomly generated orderings of chief judges that never occurred. To conduct such a test, we construct a panel dataset organized by state-half-years. We begin with the universe of all state supreme court cases from Hall and Windett (2013). From this, we drop a small set of abnormal cases—such as those discharged without an opinion and those featuring an unusually small number of judges. 5 We then divide each year into half-years, with July 1 marking the start of the second half-year. Given the data range of 1995 to 2010, this yields 32 half-year time units for each state. States produce a sufficient volume of cases that half-years still feature, on average, more cases than the U.S. Supreme Court decides in a given term. We collapse these data within half-years, obtaining four useful measures of court consensus and collegiality. First, we measure the percentage of cases decided unanimously, which we call Percent Unanimous. In the average state-half-year, about 79% of cases were unanimous, though the values vary considerably, and are as low as 16.7%. Second, we measure the percentage of cases decided with a per curiam opinion, which we call Percent Per Curiam. About 23% of opinions were per curiam in the typical statehalf-year. Third, we measure the percentage of cases with a pivotal vote—those where the number of judges out of the majority was within one vote of the majority size. At the U.S. Supreme Court, this would be the infamous “5 to 4,” but in state supreme courts, these vary based on the size of the court and the number of judges actually hearing the case. Additionally, cases of evenly split coalitions are counted as pivotal. 6 The variable is called Percent Pivotal, and in the typical state-half-year was about 6% of cases. Finally, the fourth variable is the number of Dissents Per Case, which was, on average, 0.21. Collectively, these variables provide good evidence on the level of outward-facing agreement on courts. If CJs are able to steer their courts toward more public agreement, it should be evident in these variables. We additionally code the identity of the chief judge in each half-year. While in most cases this is straightforward, as a single person held the position the entire year, there are also regular changes-of-power that do not occur exactly on January 1 or July 1. To resolve this, we credit to each halfyear the person who was CJ for the most days. Half-years allow us to be more precise than whole years, while retaining meaningful case volume in each unit. 7 While shorter periods would allow even more precision in CJ coding, they also yield fewer cases per period, resulting in noisier patterns of key outcome variables. In total, we count 204 unique CJs who served long enough to be credited with at least one half-year. The final step before estimation is pre-processing. We demean each observation through state and year fixed effects. States vary considerably in their rates of consensus, and since each CJ is unique to their state, the data may otherwise
322 American Politics Research 49(3) credit to CJs variation owed solely to other features of states. Year fixed effects remove any national trends within given years. Once de-meaned, the data represent variation relative to the state average for 1995 to 2010, minus any national year-to-year variation. These are the values that we use to estimate leadership effects. For the estimation itself, we closely follow Berry and Fowler’s RIFLE technique: we run a CJ-fixed-effects model on the de-meaned data and estimate its R 2. This value indicates to what extent the specific timing of leadership explains the variation of the demeaned outcome (case unanimity, in the primary test). The stronger the correlation between the true timing of leaders and the outcome, the higher the R 2. We expect a higher R2 for a simple CJ fixed effects model if CJs directly impacted the outcome. We then compare it to 10,000 random permutations of alternative orderings of CJs. The counterfactual CJs are drawn from the pool of actual CJs and bound to serve the same number of years as their true counterpart, in the same number and length of “blocks” of service in the CJ position, and within their true state. 8 Thus, if a CJ served for four consecutive half-years, they will always be assigned to a consecutive set of exactly four half-years within their actual state, but in randomly determined times not matching reality. For each of these permutations, we re-estimate a CJ-fixedeffects model and determine an R 2 value. This collection of R2s results in a distribution of the amount of apparent “explanatory power” we could expect to uncover through sheer noise alone, which we know the fake orderings are. We then compare the true R 2 to the 10,000 simulated counterfactual R2s. If CJs impact unanimity, we expect that the real ordering of CJs should yield among the very best R 2s possible. The percentage of false fixed effects sets that yield superior R 2 values to the true model’s R2 effectively becomes a p-value. We present the results for a test on Case Unanimity in Figure 1 and explain the results. In Figure 2, we present the results for the remaining dependent variables, which are highly consistent. The results are emphatic. The identity of the chief judge explains court unanimity no better than chance. There is no evidence that the identity of CJs mattered for steering their state supreme courts to unanimous decisions or consensus— or toward dissensus. About 47% of randomly permuted orders of CJ service explain the data as well as or better than the true ordering. Figure 2 shows that this result is not specific to case unanimity and is repeated also for the rates of pivotal and per curiam decisions, as well as the number of dissents per case. In each test, the actual order of CJs does no better than chance at explaining the data, with between 43% and 52% of random (false) orderings better explaining the (true) data. 9 Power analysis. Berry and Fowler frequently point to the justifiable concern that the RIFLE method may not have sufficient statistical power in certain arrangements of data. If there are too few units, too short a period of time under study, or insufficient variation in leaders, then it may be impossible to detect a true result. To show that this is not the case with our data, we simulated unanimity data using our exact panel format. For every true state-half-year, we generate simulated data where we define the relationship between leadership and consensus. We begin with zero, reflecting the de-meaning of the data, and then add a random draw from a distribution matching the within-CJ variation in the observed data (normally distributed with mean zero and standard deviation 0.07). This reflects random variation in case patterns. We then add to this a static quantity for each CJ, drawn from varying mean-zero distributions. The standard deviation of this CJ-effect distribution increases in the power of the typical CJ to impact the outcome. Specifically, we consider distributions with a standard deviation of 0.01, 0.02, 0.03, 0.04, and 0.05. The maximum of these would correspond to the typical chief judge being able to move their court’s unanimity rate by 4%. For each of the five levels (0.01 through 0.05), we generate 50 different values from the relevant distributions. We then estimate the RIFLE procedure with 100 permutations on each of the 50 simulations. In Figure 3, we plot the percentage of each set of 50 in which we reject the null hypothesis of no correlation of the CJ with unanimity. The results are reassuring as they indicate that our design should uncover even modest CJ results. Even for a distribution of (0,0.02), we would have about a 60% chance of uncovering the correlation, and this approaches 100% at (0,0.03) and reaches 100% of all simulations for (0,0.04) and (0,0.05). Thus, while our results do not allow us to say that CJs have absolutely zero effect on unanimity outcomes, they do allow us to say that any correlation is likely quite small, with the average CJ able to move consensus by, at best, less than two percentage points. Secondary tests. The nature of the RIFLE test—its reliance on fixed effects—necessarily eliminates much institutional 0 .01 .02 .03 .04 Fraction .21 .22 .23 .24 .25 .26 .27 .28 .29 .3 .31 .32 .33 .34 .35 R-squared from Permuted Data Dashed line shows R-squared from the real data. p-value = .4717
better than random chance.
Gray and Miller 323 variation we might otherwise be interested in, such as the specific powers chief judges have. To further explore this question, we conduct more traditional regression results on the same outcome variables, though now identifying clear institutional factors that might empower CJs, as well as obvious logistical candidates to drive consensus. These results are displayed in Table 1. One important institutional variation in the chief judge position is how one obtains the seat. At one extreme, some judges simply assume the position based on a rotating or random assignment, while others are appointed or elected to the position by governors or voters. We identify two methods of obtaining the position that may imply CJ leadership power: obtaining the position through Seniority and through a Peer Vote. In the former case, the CJ will have the longest existing relationships with other members of the court, and in the latter case the CJ is distinguished by being selected by the associate judges. All other selection methods are treated as a catch-all base category. 10 Additionally, some CJs have extra formal powers that may give them the capacity to shape outcomes more than others. We focus on two powers: whether the CJ Votes Last in voting conference, and whether the CJ Assigns Opinions. 11 Both powers give the CJ strategic advantages that could be exploited to promote consensus. Finally, we include two key logistical facts that make consensus harder: the size of courts measured by the Number of Judges, and the size of a Court’ s Workload, measured in the number of cases it decided in a half-year. The more cases there are, the more demanding it is for a judge to write additional dissenting opinions in the typical case, promoting consensus. However, the more judges there are on a court, the more possible voices there are to dissent. Pivotal cases are those decided by one vote (e.g., 5–4) or tied votes and per curiam opinions are those that are unsigned by members of the court. These results show no apparent relationship between chief-judge institutions and court consensus. How the CJ is picked and the powers over assignment and position in the voting order do not meaningfully covary with the rates of unanimity, pivotal cases, per curiam opinions, or dissents. The only facts that help explain these outcomes are the count of the cases a court has to decide and the number of judges. 0 .01 .02 .03 .04 .05 Fraction .18 .19 .2 .21 .22 .23 .24 .25 .26 .27 .28 .29 .3 .31 R-squared from Permuted Data Dashed line shows R-squared from the real data. p-value = .4332 0 .01 .02 .03 .04 Fraction .24 .25 .26 .27 .28 .29 .3 .31 .32 .33 .34 .35 .36 .37 .38 .39 .4 .41 .42 R-squared from Permuted Data Dashed line shows R-squared from the real data. p-value = .4968 0 .01 .02 .03 .04 Fraction .21 .22 .23 .24 .25 .26 .27 .28 .29 .3 .31 .32 .33 .34 .35 R-squared from Permuted Data Dashed line shows R-squared from the real data. p-value = .5116
Figure 2. Chief judges explain variation in pivotal cases (top left), Per Curiam decisions (top right), and dissents per case (bottom) no
better than chance.
judge results.
324 American Politics Research 49(3) For simple reasons, a smaller court will find it easier to agree and larger courts have more available judges to disagree and write dissenting opinions. Similarly, as court workloads increase, judge’s expressing individual opinions at odds with the majority becomes more costly and the threshold of importance should increase, yielding higher rates of apparent consensus. These variables have predictable and intuitive associations, but those for chief judge institutions do not. Collectively, these results point in a consistent direction: state supreme court judges manifest no apparent ability (or willingness) to impact their courts’ consensus and unanimity. We find no evidence that CJs impact consensus through either formal or informal means. Institutional powers such as opinion assignment and strategically useful voting orders do not in effect lead to higher consensus. And the identity of the chief judge does not matter to any significant degree for court consensus. Moving In and Out of the Chief Position As a Judge A second approach to assessing the importance of the CJ position is to look at how outcomes change for judges as they move in and out of the position. In the preceding section, we saw that variation in who serves as the CJ does not seem to explain court-level outcomes. In this section, we consider whether being the CJ is associated with different individuallevel outcomes for judges. State supreme courts are conducive to this analysis because the chief judgeship changes frequently and many judges serve on the court as an associate judge before—and even after—a period as the CJ. To assess this, we construct a dataset of every judge-vote in all state supreme courts—including the two court-of-lastresort criminal courts (in Texas and Oklahoma)—in the Hall and Windett dataset. We then model whether the judge was in the majority as a function of whether they were the Chief Judge in the case in question. We expect that the various formal and informal powers of the CJ position should make it so they are more likely to be in the majority than they were as associate judges. Other factors surely impact the likelihood of being in the majority on a case. We still expect that vote choices are primarily driven by other preferencessuch as policy attitudes. Thus, it is important to control for whether we should expect, ideologically, the CJ to be part of a majority coalition. Existing ideology scores, such as the SDIRT scores of Windett, Harden, and Hall (2015), are inappropriate because they rely on exactly the sets of votes we analyze here to estimate the scores. Thus, we opt for a separate measure: the Judge’ s Party’ s Share of Seats on the state supreme court. To achieve this, we classify each judge in the dataset as a Republican or Democrat. For many judges, this is straightforward, due to running in a partisan primary or general election, being appointed by a partisan governor, or receiving formal endorsements from state parties. In a few cases, these are difficult to assess due to a lack of partisan involvement. Because our goal is to get a measure of general ideological alignment as indicated by partisanship, but not to make
Unanimity % Pivotal cases % Per Curiam % Dissents per case Workload (n cases) –0.00 (0.00) 0.00 (0.00) 0.00* (0.00) –0.00** (0.00) Number of judges –0.07** (0.01) 0.01* (0.00) –0.01 (0.02) 0.07** (0.01) CJ by seniority –0.05 (0.04) 0.01 (0.01) 0.06 (0.07) 0.05 (0.04) CJ by peer vote –0.04 (0.03) 0.01 (0.01) 0.06 (0.06) 0.03 (0.04) CJ votes last 0.04 (0.04) –0.01 (0.01) 0.03 (0.09) –0.02 (0.03) CJ assigns opinions 0.04 (0.03) –0.01 (0.01) –0.04 (0.06) –0.04 (0.04) N 1,599 1,599 1,599 1,599 Clusters 50 States 50 States 50 States 50 States R2 0.34 0.05 0.14 0.33 Note. All models contain a time trend.
All cases Non-unanimous cases Pivotal cases Chief judge 0.01* (0.00) 0.04** (0.01) 0.05* (0.02) Judge’s party’s share of seats 0.04** (0.01) 0.06^ (0.04) 0.12* (0.05) N 1,031,616 259,294 64,184 R2 0.06 0.06 0.05 Standard error clustering 738 Judges 731 Judges 724 Judges Note. Legal area, year, and judge fixed effects are included in both models. ^p < .10. *p < .05. **p < 0.01.
Gray and Miller 325 inferences about partisanship itself, we force each judge to a Republican or Democratic code that best matches their career behavior. We expect that as this percentage grows, the probability of the judge being in the majority should also grow, as they would probably be in a likeminded majority coalition. Different legal areas are also likely to lead to different results. Some types of cases may feature more unanimous decisions, while others may be more divisive, leading to different base rates for a judge being in the majority. Thus, we control for the case legal area, as defined by the Harden and Windett dataset. These are included as fixed effects. Additionally, we include fixed effects for the year of the case, to account for any unmeasured heterogeneity over time in rates of consensus on state courts of last resort. Finally, we also include judge fixed effects. This allows for better identification of the impact of being chief judge, by narrowing in on the difference within judges between being the CJ and not being the CJ. As each judge serves only within one state, these fixed effects also internalize between-state variation in institutions and other structures that drive the likelihood of any judge being in a majority. Because we observe many votes from each judge, we cluster the standard errors on each judge. We present the results in Table 2. The analysis suggests that there is a statistically significant, but substantively small, impact of CJ status on being in the majority. CJs are 1% more likely to be in the majority in all cases and 5% more likely to be in the majority in pivotal cases. Notably, the result in pivotal cases closely mirrors findings of a similar comparison of Justice Rehnquist’s time as an associate justice and chief justice of the Supreme Court (Cross & Lindquist, 2006, p. 1691). The impact of CJ status on being in the majority is between one-quarter (all cases) and one-half the size of the impact of the share of seats held by the chief judge’s co-partisans on being in the majority. The fixed effects in the regression make clear whether the powers of CJs alter the likelihood that they are in the majority in cases. In Figure 4, we present the results of a slightly modified model that breaks the chief judge result into a specific coefficient for each state. We display these coefficients across four potential combinations of extra powers available to the CJ, with increasing authority from left to right. The main takeaway from Figure 4 is the lack of correlation between available powers and the likelihood of being in the majority. 12 We find some evidence that being the CJ is beneficial to the individual who holds the position, in that they are slightly more likely to prevail, especially in pivotal cases. There is no correlation between being in the majority and extra power that accompanies service as the CJ in some states. In other words, extra powers do not increase the chances that a CJ is in the majority. This suggests that opinion assignment, often seen as the key tool available to the chief justice of the U.S. Supreme Court (e.g., Bonneau et al., 2007; Wahlbeck, 2006), is not what enables the success of CJs in state supreme courts. The chief justice of the Supreme Court serves a lengthy, open-ended term—since 1986 there have been only two. In contrast, rotation in the center chair is more common in state supreme courts. In our data, 211 judges served at least once as CJ in a state supreme court, suggesting significantly shorter terms and a potential return to associate judge status once time as CJ ends. Perhaps such a power-sharing arrangement reduces incentives to use the power of opinion assignment in a consistently strategic manner and increases collegiality (e.g., Brace & Hall, 1993). In addition, high workloads in many state supreme courts may dull the ability of the CJ to act strategically. Under norms of equality in opinion writing, workload pressures might make strategic opinion assignment difficult. The power to vote last should benefit a CJ by allowing them to strategically join a majority coalition and assign an opinion so as to keep such an opinion relatively ideologically proximate (e.g., Epstein & Knight, 1998). Nevertheless, even in state supreme courts where both the power to vote last and the power of opinion assignment belong to the CJ we observe no obvious increase in the likelihood that the CJ is in the majority. 13 The California State Supreme Court is an outlier in this group, as we observe that the CJ is approximately 8 percentage points more likely to be in the majority, but of course the California Supreme Court is more like the U.S. Supreme Court in its prestige (Caldeira, 1983) and professionalization (Squire, 2008) than any other state supreme court. All of this suggests a need to develop a sui generis theory of the power of the CJs in state supreme courts, rather than imposing the theory developed at the Supreme Court level on the states. Discussion and Conclusion It has become a common theme of coverage of the Roberts Court to assume the chief justice’s preference for and ability to bring about court consensus and collegiality. However, AL CA HIIA LA ME MI MT NJNY PA SD TX UTVA WI AK AZAR CO CT DE FL GAID IL KS KY MD MA MN MS MO NE NV NH NM NC OK OR RI SC TN VT WA WV WY -.1 -.05 0 .05 .1 Increased Probability ofB eing in Majority as Chief No Extra Powers Last Vote (LV) Opinion Assigner (OA) LV + OA
Note. Statistically significant coefficients are black and in larger text, while states without statistically significant coefficients are presented in gray and smaller text.
326 American Politics Research 49(3) there is little empirical evidence that chief judges actually have this ability. We investigate this at the state supreme court level and find no evidence for chief judge leadership on court consensus. Although this may be because CJs try and fail to achieve consensus or because they do not try in the first place, our results are clear under a variety of approaches. Using novel and more traditional methods, and approaching the task with a number of possible specifications, we consistently find no impact by CJs on consensus. Instead, more obvious factors such as the size of courts and their dockets impact the likelihood of dissenting voices. Extrapolating from these results to the U.S. Supreme Court, the Court’s relatively large size and clear preference splits, as well as its rigorously maintained small docket, would appear to be far more consequential than the identity of the chief justice for improving court consensus. Our results are consistent with a body of work that finds that democratic institutions constrain rather than enable leadership effects. CJs are given a small set of powers to lead their courts, but exist within a much more significant tapestry of court institutions and logistical hurdles that mute their ability to steer the court. Despite this, individual salient cases and the tendencies of journalistic and historiographic coverage of the court may inflate the perceived impact of individual chief justices and chief judges. We do find that being CJ is beneficial to those who serve, even if who serves has little correlation with consensus. Utilizing the greater variation available by analyzing CJs on state supreme courts, we find that CJs are slightly more likely to find themselves in the majority, as compared to when they served on the court in a role other than as chief judge. A similar result appears in work on Chief Justice Rehnquist’s time in the center chair (Cross & Lindquist, 2006). Our results further suggest that it is the title alone, and not special powers that might accompany the office, that helps chief judge’s win more often. Therefore, we can say that court leadership seems to have a small benefit for the leader and little or no correlation with court consensus. Declaration of Conflicting Interests The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article. Funding The author(s) received no financial support for the research, authorship, and/or publication of this article. ORCID iD Banks Miller https://orcid.org/0000-0002-7948-8947 Notes 1. Throughout we use the term CJ to refer to chief judges on state supreme courts. Although judges on state supreme courts are often referred to as justices, in order to help make clear distinctions between U.S. Supreme Court Justices and state supreme court justices, we refer to the former as justices (either chief or associate) and the latter as judges. 2. Interpretations of the notion of task and social leadership have put further gloss on these concepts. For instance, Danelski (1989) suggest that CJs who are task leaders tend to be more interested in convincing the other justices of the correctness of their position, while caring little about the social aspects of their arguments. Conversely, social leaders are CJs who are more likely to suppress their own views in the interest of building consensus. Either is a plausible mechanism for influence and our analytical approach does not allow us to distinguish them. Given our results, however, it is likely that neither approach is noticeably effective. 3. Technically, it is also possible that state CJs simply do not vary in their leadership effectiveness and preferences for consensus. While possible, we find this explanation extremely implausible. A group of many dozens of different individuals will almost certainly vary in goals and talents. 4. Before Chief Justice Rehnquist, who served as an associate justice for 14 terms before becoming chief justice, we must go back to Chief Justice Stone, who served as chief justice from 1941 to 1946 after a long tenure as an associate justice. 5. The cases vary widely in the specific facts leading to abnormality, but each ultimately did not feature the normal set of a full or nearly full court deciding a case and releasing an opinion announcing the decision. These abnormalities may imply a different data generating process for vote choices. In total, these cases are less than 0.4% of the entire docket. We believe their removal is a sound decision, though note that including them would not impact results. 6. Evenly tied cases are sufficiently rare (0.2% of all cases, or about one every 2 to 3 years per state) that they are not useful for analysis and instead fit better combined with cases decided by one-vote margins. Tied cases are the purest form of “pivotal” as each judge could change the result by switching their vote. In a one-vote margin, only those in the majority could change the result by changing their vote. 7. For example, imagine a year in which Chief A serves until July 15, and then Chief B serves the remainder of the year. In a whole-year method, we would credit the year to Chief A, coding 46% of the year’s data incorrectly. With the halfyear coding, January-June are exactly correct, while the July- December half-year is coded 92% correctly. 8. Thus, if in State Z, CJs A, B, and C served for 10, 10, and 12 half-years consecutively, the set of counterfactual fixed effects for State Z would feature sequences of A, C, B; B, C, A; B, A, C; C, A, B; and C, B, A, retaining the true lengths of service for A, B, and C in each case. This occurs across all 50 states to produce a considerable, but finite, number of permutations. 9. We considered a variety of possible methods of subsetting the data, including based on the methods of selection of a CJ, the specific powers of CJs, and with various forms of pre-processing, and consistently found no explanatory power for CJs. 10. This baseline category includes chief judges who obtain the position through election, appointment by the governor, or who are placed in the position by a nominating commission. 11. Data on chief judge powers comes from Hughes et al. (2015).
Gray and Miller 327 12. One objection to this finding is that it is confounded with our measure of success—being in the majority. In other words, readers may anticipate that opinion assignment on state supreme courts operates as it does on the U.S. Supreme Court—that the CJ only assigns when in the majority. However, six of the eleven state CJs with opinion-assignment power assign the majority opinion regardless of whether they are in the case majority. 13. Despite expectations, this result does not appear to vary based on whether the CJ must be in the majority to assign cases, as is the practice at the U.S. Supreme Court, but only in some states.
Baum, L. (2010). Motivation and judicial behavior: expanding the scope of inquiry. In D. Klein & G. Mitchell (Eds.), The psychology of judicial decision making (pp. 3–25). Oxford. Berry, C. R., & Fowler, A. (2018). Leadership or luck? Randomized inference for leader effects. https://pdfs.semanticscholar.org/2a 41/54eb0ad21523673ed0a2bf4200f7447b5f2d.pdf. Brace, P., & Hall, M. G. (1993). Integrated models of judicial dissent. Journal of Politics, 55(4), 914–935. Brennan Center for Justice. (2018, February 7). Legislative assaults on state courts—2018. https://www.brennancenter.org/analysis/legislative-assaults-state-courts-2018. Bonneau, C. W., Hammond, T. H., Maltzman, F., & Wahlbeck, P. J. (2007). Agenda control, the median justice, and the majority opinion on the U.S. Supreme Court. American Journal of Political Science, 51(4), 890–905. Caldeira, G. A., & Zorn, C. J. (1998). Of time and consensual norms in the supreme court. American Journal of Political Science, 42(3), 874–902. Caldeira, G. A. (1983). On the reputation of state supreme courts. Political Behavior, 5, 83–108. Canes-Wrone, B. (2006). Who leads whom? Presidents, policy, and the public. Chicago University Press. Cross, F. B., & Lindquist, S. (2006). The decisional significance of the chief justice. University of Pennsylvania Law Review, 154, 1665–1707. Danelski, D. J. (1960). The influence of the chief justice in the decisional process of the supreme court. Presented at the annual meeting of the American Political Science Association, New York. Danelski, D. J. (1989). The influence of the chief justice in the decisional process of the supreme court. In S. Goldman & A. Sarat (Eds.), American court systems (pp. 486–99). Longman. Epstein, L., & Knight, J. (1998). The choices justices make. Congressional Quarterly Press. Epstein, L., Landes, W., Posner, R. (2013). The behavior of federal judges: A theoretical and empirical study of rational choice. Harvard University Press. Gerber, E. R., & Hopkins, D. J. (2011). When mayors matter: Estimating the impact of mayoral partisanship on city policy. American Journal of Political Science, 55(2), 326–339. Gray, T. (2017). The influence of legislative reappointment on state supreme court decision making. State Politics & Policy Quaterly 17(3), 275–298. Gray, T. (2019). Executive influence on state supreme court justices: Strategic deference in executive reappointment states. Journal of Law, Economics, & Organization, 35(2), 394–421. Hall, M. E. K., & Windett, J. H. (2013). New data on state supreme court cases. State Politics & Policy Quarterly, 13(4), 427–445. Hall, M. E. K., & Windett, J. H. (2016). Discouraging dissent: The chief judge’s influence in state supreme courts. American Politics Research, 44(4), 682–709. Haynie, S. L. (1992). Leadership and consensus on the U.S. Supreme Court. Journal of Politics, 54(4), 1158–1169. Hendershot, M. E., Hurwitz, M. S., Lanier, D. N., & Pacelle Jr, R. L. (2012). Dissensual decision making: Revisiting the demise of consensual norms within the U.S. Supreme Court. Political Research Quarterly, 66(2), 467–481. Hunzeker, D. (1990). Legislative-judicial relations: Seeking a new partnership. State Legislative Report, 15(14), 1–10. Hughes, D. A., Wilhelm, T., & Vining Jr, R. L. (2015). Deliberation rules and opinion assignment procedures in state supreme courts: A replication. Justice System Journal, 36(4), 395–410. Jones, B. F., & Olken, B. A. (2005). Do leaders matter? National leadership and growth since World War II. Quarterly Journal of Economics, 120(3), 835–864. Langer, L. (2002). Judicial review in state supreme courts: A comparative study. SUNY Press. Langer, L., McMullen, J., Ray, N. P., & Stratton, D. D. (2003). Recruitment of chief justices on state supreme courts: A choice between Institutional and personal goals. Journal of Politics, 65(3), 656–675. Leonard, M. E., & Ross, J. V. (2014). Consensus and cooperation of state supreme courts. State Politics & Policy Quarterly, 14(1), 3–28. Rehnquist, W. H. (1976). Chief justices I never knew. Hastings Constitutional Law Quarterly, 3, 637–655. Sheldon, C. H. (1999). The incidence and structure of dissensus on a state supreme court. In C. Clayton & H. Gillman (Eds.), Supreme court decision-making: New institutionalist approaches (pp. 201–213). University of Chicago Press. Squire, P. (2008). Measuring the professionalization of U.S. State courts of last resort. State Politics & Policy Quarterly, 8(3), 223–238. Tiberii, J. (2019, May 29). NC chief justice could lose half her staff under proposed senate budget. WUNC, North Carolina Public Radio. https://www.wunc.org/post/nc-chief-justice-couldlose-half-her-staff-under-proposed-senate-budget. Wahlbeck, P. J. (2006). Strategy and constraints on supreme court opinion assignment. University of Pennsylvania Law Review, 154, 1729–1755. Walker, T. G., Epstein, L. J., & Dixon, W. J. (1989). On the mysterious demise of consensual norms in the United States supreme court. Journal of Politics, 50(2), 361–389. Windett, J. H., Harden, J. J., & Hall, M. E. K. (2015). Estimating dynamic ideal points for the state supreme courts. Political Analysis, 23(3), 461–469. Zink, J. R., Spriggs II, J. F., & Scott, J. T. (2009). Courting the public: The influence of decision attributes on individuals’ views of court opinions. Journal of Politics, 71(3), 909–925. Author Biographies Thomas Gray is an assistant professor of Political Science at the University of Texas at Dallas. Banks Miller is an associate professor of Political Science at UT Dallas.