Status characteristics and their intersectionality: majority opinion assignment in state supreme courts

ABSTRACT

We test whether justices’ traits – race, gender, age, previous judicial experience, education, and tenure – are associated with opinion assignment patterns as suggested by status characteristics theory. Female justices were more likely to be asked to write the majority opinion, particularly if the case raised a “women’s issue.” In complex cases, however, both female and African American justices were less likely to be selected to write the majority opinion. Extending intersectionality perspectives, advances in age were associated with a decreased likelihood in women and African American justices receiving the assignment, though at distinct age ranges.

KEYWORDS:

• Intersectionality
• gender studies
• judicial behavior
• race studies

Disclosure statement

No potential conflict of interest was reported by the authors.

Notes

1 The Brace and Hall State Supreme Court Dataset, an existing National Science Foundation (NSF) dataset, covers the time period, 1995-1998. As it includes the universe of state supreme court decisions in this time period, the data has been widely employed in the published literature (see, e.g., Bonneau and Rice 2009; Brace, Yates, and Boyea 2012; Christensen, Szmer, and Stritch 2012; Hall 2014; Randazzo, Waterman, and Fix 2011; Shepherd 2009; Szmer, Christensen, and Kaheny 2015).

2 These states are: Arizona, California, Colorado, Connecticut, Delaware, Hawaii, Kansas, Kentucky, Maryland, Massachusetts, New Jersey, Oregon, Pennsylvania, and Wyoming. They either follow the US Supreme Court procedure (the chief justice assigns the opinion if she is in the majority; if not, the senior judge in the majority makes the assignment), or the chief justice always makes the assignment (Hall 1990). We relied upon Hall’s (1990) categorization of assignment regimes, based on the results of a phone survey of all 50 state courts of last resort, to identify state supreme courts that utilize discretionary (as opposed to random or rotating) assignment processes. As part of a recent study, Hughes, Wilhelm, and Vining (2015) surveyed state courts to see whether they had altered assignment procedures since 1990. Unfortunately, Hughes and colleagues did not identify when changes in opinion assignment methodology were made. We chose to rely on the Hall (1990) study, because the data was collected from 1988-1989, which is much closer to the time period we analyzed - 1995 to 1998. However, when examining Hughes, Wilhelm, and Vining’s (2015) results, the states we included in the analyses have retained some form of discretionary assignment. In addition, they report that the Tennessee Supreme Court now utilizes discretionary opinion assignment (Hughes, Wilhelm, and Vining 2015). To be conservative, we do not include Tennessee in our study as it is unclear when its assignment procedure changed.

3 We thus exclude observations when a justice casts a dissenting vote to account for this tendency. In addition, as a robustness check, we ran the models for only those situations in which there was a woman or an African American justice in the majority coalition. The substantive results of these auxiliary analyses were generally the same as those we report in the tables below. The one exception is the replication of the model that includes Case Complexity (Table 3). While the stand-alone coefficient for Female Justice is significant at the 0.05 level in the presented model, in our robust check, it is only significant at the 0.056 level. Of course, since that model includes the Female Justice and Case Complexity multiplicative term, and a zero value of Case Complexity is not conceptually meaningful, the stand-alone coefficient for Female Justice is not theoretically important.

4 Majority opinion assignment studies have utilized a variety of techniques, including conditional logit (Farhang, Kastellec, and Wawro 2015), random effects probit (Maltzman and Wahlbeck 2004), and multinomial logit (Maltzman and Wahlbeck 1996). Each method has its strengths and weaknesses. Conditional logit, for example, accounts for all court-level and case-type influences (Farhang, Kastellec, and Wawro 2015). This approach would minimize the statistical problems caused by these factors. However, since we are interested in how case-type variables like issue area and complexity interact with justice demographic characteristics, this approach is problematic in this context. Moreover, the alternative-specific conditional logit model essentially requires including a series of multiplicative terms where each separate choice option is multiplied by each separate covariate. This is unworkable in our study where we have 108 different options (justices) across courts and cases. The conditional logit model also assumes the independence of irrelevant alternatives (IIA)—essentially the choice between assigning the opinion to any two judges must be independent of any of the other options (Cameron and Trivedi 2010). In the context of the US Courts of Appeals, random assignment to three-judge panels mitigates the likelihood of an IIA violation (Farhang, Kastellec, and Wawro 2015). However, state courts of last resort generally meet en banc—and the majority coalition is systematically, not randomly, constructed. Other alternatives are equally problematic. For example, while multinomial logit models do not require the IIA assumption, like the alternative-specific conditional logit, they practically require a relatively consistent, small set of choices–not 108 options varying across thousands of cases and 14 courts.

5 It is also possible that assigners will have a better assessment of a peer’s policy inclinations with increases in tenure, rendering assignment to more senior colleagues a more predictable option.

6 While our hypotheses could extend to other minority justices, due to a lack of racial diversity on the bench during this time period, we explore whether majority assignment processes differentially affect African American justices in the present study.

7 As with our determination of those states using discretionary assignment, we relied on the survey results of Hall (1990) to figure out who the likely majority opinion assigner would be. In her study, she reports whether a given state has the tradition of allowing either the chief justice (if in the majority coalition) or the senior associate justice (if the chief justice is among the dissenters) to choose the majority opinion author (see Hall 1990, 210–211). She also reported those states in which this power is unconditionally given to the chief justice (Hall 1990, 210–211).

8 While our sample size permits statistical examination of both gender and race effects, it is important to note that our dataset includes the voting records of 16 women and seven African American judges. This means that about 15% of the judges voting in the dataset are women, and about 6.5% of the judges are African American. While small, the percentage of women in the dataset is sufficient to evaluate structural patterns in the data as they pertain to the influence of justice gender. However, the small number of African American justices in the dataset may partially explain why we do not find stronger race effects.

9 Results of auxiliary regressions suggest that possible collinearity between justice age and tenure is not a concern. The variance inflation factor (VIF) for Justice Age ranges from 1.83 to 2.53 across the models. The associated VIF for Justice Tenure does not exceed 1.54 in these same analyses. We also note that multicollinearity was generally not an issue—with one exception. The number of panelists graduating from an elite law school did have a VIF close to 10. However, it was still statistically significant in the posited sign, so it is not problematic.