![Sample our Politics & International Relations journals, sign in here to start your FREE access for 14 days]({"type":"image" , "src" : "/sda/5947/TOC-Subject-Sample-2021-Politics-International-Relations.jpg"})
ABSTRACT
Evidence of gendered decision making by judges has been mixed at best. We argue that this is a result of a narrow focus on how female judges differ from male judges. This treats women as the “other,” and the primary object of study is often to determine why female judicial behavior differs from the “norm” of male behavior. We depart from this tradition by using male-centered theories to derive and test hypotheses about maleness and the interactive effect of judge gender and litigant gender in appellate decision making. Drawing on findings from an original dataset of immigration appeals, we find evidence that gender biases manifest themselves in patterns of appellate decision making among all-male panels. Despite our predictions, female judges may also demonstrate evidence of these biases.
Disclosure statement
No potential conflict of interest was reported by the authors.
Notes
1 We recognize that the likelihood of a remand in an individual case will be affected by many legal and factual characteristics of the case. For example, the Courts of Appeals apply a deferential standard of review to certain kinds of questions in appeals, but not to others. This is true even among different types of immigration appeals. However, these standards are inherently indeterminate and can lead to different results in the hands of different judges. We measure only the ultimate result, namely whether petition for review is denied (i.e., the order or removal is affirmed) or the case is remanded.
2 INA § 101(a)(42)(A), 8 U.S.C. § 1101(a)(42)(A).
3 The race or ethnicity of the female litigants could complicate the analysis, as chivalry might only benefit white, middle-class women (Klien Citation1973). We hope to pursue this intersectionality dimension in future research.
4 Again, intersectionality may play a key role here. The majority of immigrants in removal proceedings are members of historically emasculated groups (Lynch and Haney Citation2011; Saucedo Citation2013; Park Citation2013). We hope to explore this dynamic in future research.
5 We removed cases where the PACER record had locked documents. This happens on occasion when the documentation is “locked” in PACER, meaning that it is not available for purchase from the PACER website. Sometimes this can happen when there is sensitive information contained in the documents. In these 91 locked cases, we were unable to record the sex category of the litigant, among other things. Most of the locked cases in our sample are from the Second Circuit. Only 28 of these 91 locked petitions were eventually decided on the merits by a panel of judges.
6 A control for concurrences and dissents is insignificant and does not change the results reported below.
7 We use binary sex category as a proxy for gender. Gender is a social construct and a social performance. But, since we do not have data on the gender identification of federal judges (or litigants), we recorded their sex as a proxy.
8 In a small number of ambiguous cases, we used Gender Checker (www.genderchecker.com) to confirm our categorization. Like all existing naïve Bayesian gender classifiers, this one is of limited utility for non-Western names, although we hope to pursue more broadly inclusive name dictionaries in future research.
9 We also investigated a number of different model specifications, including Heckman selection models to investigate the possibility of selection bias (with and without circuit controls), and models with fixed effects for the circuits or standard errors clustered around the circuits. Our results were robust among these specifications. Results and replication data for these alternative specifications are available upon request from the authors. Overall, our models showed that, while the circuits vary significantly in their overall grant rates, controlling for the variable in our models left very little variance explained by unmeasured circuit characteristics.
10 Some common indictors that the case may involve a criminal conviction include INA § 212(h) waivers of a criminal ground of inadmissibility, cancellation of removal for lawful permanent residents, and challenges to removability. We control for this collection of case facts using a dummy variable.
11 Although a Hausman (Citation1978) test shows evidence of non-zero correlation, we follow the lead of Clark and Linzer (Citation2013), whose simulation research shows no significant difference in the root-mean-squared error of the slope estimates between random and fixed effects in datasets like the one we use here. In any event, the coefficients in the model are of similar magnitude and the significant p-values of the coefficients stay significant at the p<.05 level across the models.
12 This difference is significant in a one-tailed hypothesis test at the p > .05 level.
13 We replicated this analysis using both the Giles, Hettinger, and Peppers (Citation2001) ideology scores and the Epstein et al. (Citation2007) judicial common space scores. All three of these measures were insignificant in the model, and the magnitude and significance of the remaining variables were not impacted.