The primary goal of the Duke Law Journal’s Symposium on Evaluating Judging, Judges, and Judicial Institutions was to bring together judges and academics researching judges. Conversations between these groups can be constructive on both sides. Judges may benefit from learning about studies that show the influences on judicial performance or that demonstrate which reforms can improve the quality of judging. Academics may benefit by discovering new ideas that have not yet been researched or by understanding how judging in the real world compares with their view of judging.
Some of the discussions at the Symposium highlighted an area where academics’ perceptions of judging conflict with judges’ actual experiences. Judges and academics view the significance of judicial dissents quite differently. Whereas many of the judges believe that dissents primarily reflect the level of cohesiveness and collegiality of the court, academics typically place much more significance on the meaning of judicial dissents. For example, recent academic studies have asserted that judicial dissents often reveal the influence of judges’ retention concerns,the level of judges’ independence, or certain judges’ higher propensity for risk taking.
In contrast, many judges at the Symposium considered academics’ emphasis on judicial dissents to be misguided. The judges maintained that dissents reveal less about judges’ retention concerns, impartiality, or risk preferences, and more about the culture of collegiality on the court. Many of the judges believe that dissents primarily reflect the level of cohesiveness among judges, and that various factors and institutions influence this cohesiveness. For example, they explained that the personal relationships, amount of professional and social interaction, and diversity among judges on a court might influence the level of dissent. Many of the judges hypothesized that courts that are more collegial, either because the judges are better friends or share more common values or backgrounds, should have lower dissent rates. Judges on these courts should be less likely to openly criticize the opinions of their colleagues by dissenting.
Although academics have long recognized that institutions such as opinion-assignment procedures and voting order might influence the propensity to dissent, empirical studies have failed to consider the impact of collegiality and personal relationships on dissent rates. Thus, in this short Essay, I empirically test whether some of the judges’ assertions are consistent with the data. I test whether various measures of diversity are associated with dissent rates in state supreme courts. I find that diversity in many areas—gender, race, age, religion, home state, and political affiliation—is associated with higher levels of dissent. In contrast, diversity in the jobs that judges had before taking the bench is associated with lower dissent rates.
I also test whether the length of time judges have served on the court is associated with dissent rates. Presumably, judges that have served on a court together for many years would have stronger friendships than newer judges, and thus may be more collegial and less likely to dissent. However, my empirical analysis finds the opposite: the greater the number of judges with lengthy tenures on the court, the higher the dissent rate.
To test the influence of diversity and tenure on judges’ propensity to dissent, I use data from the State Supreme Court Data Archive. This data includes an almost universal sample of state supreme court cases in all fifty states from 1995 to 1998. The data include more than twenty-eight thousand decisions involving more than 470 individual state supreme court justices.The data include variables that reflect case histories, case participants, legal issues, case outcomes, and individual justices’ behavior. I supplemented these data with institutional variables that describe aspects of each state’s judicial system, and with detailed information about each judge’s personal characteristics, background, and career.
Thus, my data consist of individual judge-level votes in each case before the state supreme courts. I use an ordinary probit model to test whether diversity and tenure on a court influence judges’ likelihood of dissenting. The dependent variable in my estimation is an indicator variable for whether an individual judge casts a dissenting vote in each case.
My estimation includes several measures of the diversity of judges’ personal characteristics and backgrounds on each court. I measure the diversity of each personal characteristic on each court with an index that is essentially one minus a Herfindahl index of each characteristic:
For example, diversity along the racial dimension would be measured with:
The diversity measure ranges from zero, when the court is composed of judges of only one “type” (that is, race), to one, when each type is represented equally on the court. Thus, increases in this measure indicate an increase in diversity on the court.
I include variables that measure diversity along several different dimensions: gender, race (White, African American, Asian, Hispanic, and other), age (under 45, 46 to 55, 56 to 65, and over 65), religion (Protestant, Catholic, Jewish, Greek Orthodox, Mormon, and other), home state (from the state where the court is located or from another state), political party affiliation (Republican, Democrat, or Independent), and job held before taking the bench (prosecutor, attorney general, elected official, lower-court judge, or nonelected official).
I also include two variables that represent judges’ tenure on the court. I include the percentage of judges on each court that have served for one year or less; presumably, these judges have not served long enough to develop strong personal relationships with other judges. I also include the percentage of judges on each court that have served for six years or more; if these judges have developed stronger friendships, they may be more collegial and less likely to dissent.
In addition to the diversity and tenure variables, the estimations include a series of judge-level, case-level, and state-level variables that might be related to judges’ propensity to cast dissenting votes. The judge-level variables include an indicator for whether a particular judge is the chief justice on the court, and a variable indicating the number of years until the judge’s next retention.These variables control for voting changes throughout a judge’s career and term.
All estimations also include various case-level variables that may be related to dissenting votes. First, I include indicator variables for whether a case is a criminal case, a juvenile case, a civil case involving the state government, or a civil case involving private individuals. Thus, the base category is nonadversarial cases, such as cases involving certification and advisory opinions. Finally, I include indicator variables for whether at least one litigant is a business, a person, or a representative of the state government. These indicator variables control for any relationship between dissent rates and the litigants or legal issues in a case.
Next, I include various state-level characteristics that have been found to be related to dissenting votes. First, I include an indicator variable for whether judges in the state face reelection by the voters. In a previous study, I found empirical results suggesting that judges’ reelection concerns are important influences on their propensity to cast dissenting votes.
I also include a variable that indicates whether the state has a lower appellate court, and thus, whether the court has discretionary review to hear cases. Numerous studies report that the presence of an intermediate appellate court increases dissent rates, suggesting that discretionary dockets facilitate the expression of dissent.
The state-level variables also include indicator variables for whether a court utilizes a random opinion-assignment procedure instead of a discretionary procedure, and whether voting takes place in the order of seniority. Random or rotating opinion-assignment procedures prevent judges from being rewarded or sanctioned for their opposition votes, and thus reduce the incentives for consensus.In contrast, when opinions are assigned by the chief justice, opportunities for rewards and sanctions emerge. Similarly, when voting takes place in order of seniority, the most senior judges may influence junior judges, reducing the likelihood of their disagreement.
Moreover, all estimations include year indicators to capture trends in the likelihood of dissent. In the probit estimations, the t-statistics are computed from standard errors clustered by case.
Table 1 reports the primary probit results. In this table, the top number in each cell is the regression coefficient, which indicates the magnitude and direction of the relationship of each variable with judges’ votes. A negative coefficient indicates that a variable reduces the probability that a judge will cast a dissenting vote; a positive coefficient indicates that a variable increases the probability that a judge will dissent. Under each coefficient is the corresponding t-statistic. Coefficients with t-statistics equal to or greater than 1.96 are considered statistically significant at the 5 percent level, meaning that there is 95 percent certainty that the coefficient is different from zero. A t-statistic equal to or greater than 1.96 is typically required to draw conclusions in hypothesis testing.
The results indicate that diversity among the judges hearing a case is significantly related to the propensity to dissent. The positive and statistically significant coefficients on the diversity variables for gender, race, age, religion, home state, and political affiliation indicate that diversity along these dimensions is associated with higher levels of dissent. These results are consistent with the Symposium judges’ belief that dissent reflects lack of cohesiveness on a court; greater diversity probably implies that the judges have fewer common values and experiences, reducing the incentives for collegiality.
In contrast, diversity in the jobs that judges held before taking the bench is negatively associated with dissent. This suggests that more diversity in professional backgrounds is associated with a reduction in the propensity to dissent. Thus, different backgrounds, at least in terms of the judges’ professional lives, do not appear to reduce collegiality on a court.
Moreover, the results indicate that the more judges with very short tenure on the court (one year or less), the less likely judges are to dissent. Similarly, the more judges with lengthy tenures on the court (six years or more), the more likely judges are to dissent. If lengthy tenure is a good proxy for strong personal relationships among judges, then these results are inconsistent with the experiences of some of the judges participating in the Symposium. Instead of lengthy tenures reducing the propensity to dissent because judges are better friends, my results indicate that lengthy tenures increase the propensity to dissent. Nevertheless, the results are consistent with studies that assert that junior judges feel pressure to not dissent against more experienced senior judges.Thus, even if judges build stronger personal relationships during their time on the court, their inclinations to agree with friends may be outweighed by the confidence and experience that comes with longer tenure on the court.
Several other variables also have statistically significant relationships with the propensity to dissent. Judicial elections, discretionary dockets, random opinion assignment, and seniority voting are associated with increases in dissent rates. In contrast, an approaching retention and chief justice status appear to reduce the propensity to dissent.
In this short Essay, I tested whether the empirical evidence is consistent with the experiences of some of the judges participating in the Duke Law Journal’s Symposium on Evaluating Judging, Judges, and Judicial Institutions. Although my results confirmed that institutional and political factors are important influences on dissent rates, I also found that the levels of cohesiveness and collegiality among judges are important. These results indicate that academic studies of judging can greatly benefit if academics consider the real-world experiences of judges.
Copyright© 2018 Joanna Shepherd.
† Professor of Law, Emory University School of Law. This Article originally appeared on the Legal Workshop as part of the Duke Law Journal’s 2010 “Symposium on Evaluating Judging, Judges, and Judicial Institutions.”
 Joanna M. Shepherd, The Politics of Judicial Opposition, 166 J. Institutional & Theoretical Econ. 88, 105 (2010).
 Stephen J. Choi, G. Mitu Gulati & Eric A. Posner, Professionals or Politicians: The Uncertain Empirical Case for an Elected Rather than Appointed Judiciary, 26 J.L. Econ. & Org. (forthcoming 2010), available at http://jleo.oxfordjournals.org/cgi/reprint/ewn016.
 Paul Brace & Melinda Gann Hall, Integrated Models of Judicial Dissent, 55 J. Pol. 914, 920 (1993); Paul Brace & Melinda Gann Hall, Neo-Institutionalism and Dissent in State Supreme Courts, 52 J. Pol. 54, 59 (1990); Melinda Gann Hall & Paul Brace, Order in the Courts: A Neo-Institutional Approach to Judicial Consensus, 42 W. Pol. Q. 391, 398 (1989).
 Melinda Gann Hall, Docket Control as an Influence on Judicial Voting, 10 Just. Sys. J. 243, 243 (1985).
 A recent study has explored the relationship between political, gender, and racial diversity on opinion publication practices. Although the study finds that gender and racial diversity have little impact on publication rates, it finds that political diversity among circuit court judges decreases the number of district court judges’ opinion publications, while increasing the length of those publications. Stephen J. Choi, G. Mitu Gulati & Eric A. Posner, What Do Federal District Judges Want? An Analysis of Publications, Citations, and Reversals 23–25 (Univ. of Chi. Law & Econ., Olin Working Paper No. 508, 2010), available at http://ssrn.com/abstract=1536723.
 State dockets exceeding two hundred cases in a single year are selected from a random sample of two hundred cases. Typically, case quantities are unaffected due to the limited size of many state supreme court dockets.
 Albert O. Hirschman, The Paternity of an Index, 54 Am. Econ. Rev. 761, 761 (1964).
 This diversity index is the standard measure of diversity used by both the U.S. Census Bureau and other studies on diversity. Alberto Alesina, Reza Baqir & William Easterly, Public Goods and Ethnic Divisions, 114 Q.J. Econ. 1243, 1254 (1999).
 This variable is actually the reverse of the years to retention. Because the longest number of years to retention during my sample is twelve, the inverse years to retention is thirteen minus the years to retention.
 Shepherd, supra note 1, at 105.
 H. Glick & G. Pruet, Jr., Dissent in State Supreme Courts: Patterns and Correlates of Conflict, in Judicial Conflict and Consensus: Behavioral Studies of American Appellate Courts 199, 200 (Sheldon Goldman & Charles M. Lamb eds., 1986); Hall & Brace, supra note 3, at 398.
 Hall, supra note 4, at 250.
 See Hall & Brace, supra note 3, at 397.
 This “freshman effect”—the lower likelihood of junior judges to dissent—has been found in Virginia A. Hettinger, Stefanie A. Lindquist & Wendy L. Martinek, Acclimation Effects and Separate Opinion Writing in the U.S. Courts of Appeals, 84 Soc. Sci. Q. 792, 802 (2003).