The Gender Influence on US District Court Decisions: Updating the Traditional Judge Attribute Model

Abstract ABSTRACT. The present study contributes to the literature on federal lower court decision‐making as well as the broader body of research on gender effects on elite behavior by showing how gender, as a determinant of behavior, is filtered through other sociological determinants of decisional behavior. Starting with the assumption that personal attributes influence decision‐making patterns, our argument is that gender effects are not direct; rather, they subtly and indirectly influence elite behavior by providing another “lens” or “layer” by which a judge will view his or her world. These effects cannot be directly observed using traditional attribute behavioral models. We test our hypotheses with data on 35,038 published decisions by federal district court judges from the 1977 to 2005 terms. The results of this analysis indicate that gender has a subtle and indirect influence on judicial behavior that is not captured by traditional judge attribute models. Keywords: Female judgesjurisprudencegender lenslower courts Notes 1. Although Richardson and Vines do not specifically use this analogy, their idea is that one could express the “pulls” on judges as a set of vectors with different magnitudes and directions, produced by the two subcultures, and the ultimate decision could be conceptualized as the mathematical sum of adding together these vectors, using matrix algebra. 2. We might also expect a “regional subculture” to exist, although it is not the major topic of investigation in this study. 3. Traditional interaction terms would not adequately capture male‐female partisan and regional distinctions, either. This reasoning is further explained in the data and methods section. 4. The analysis is limited to the extent that we look at aggregate decisions across multiple districts and circuits, which may emphasize different publication criteria. However, any limitations that may exist, we believe, are outweighed by the quality of the data set in terms of length of time period, volume of cases (the universe of published decisions), and diversity of issue areas. While not a perfect measure, the only other feasible alternative in terms of data availability would be to utilize unpublished decisions appealed to the US Courts of Appeals, which would introduce other potential biases, such as whether or not cases appealed are systematically different than cases that are not appealed. 5. Some studies have utilized Poole and Rosenthal's presidential nominate scores as an alternative measure of political ideology in order to make distinctions between same‐party presidents. We analyzed the data using the W‐Nominate scores and obtained the same general results as we did when using the appointing president's party. We chose to use the more parsimonious party variable. 6. In determining which states are “Southern,” we followed Nagel's practice of including the states that constituted the Confederacy and also those states whose predominant cultural socioeconomic character has been southern: Alabama, Arkansas, Delaware, Florida, Georgia, Kentucky, Louisiana, Maryland, Mississippi, Missouri, North Carolina, Oklahoma, South Carolina, Tennessee Texas, Virginia, and West Virginia. 7. We chose not to separately analyze district decisions by circuit. While, we knowingly acknowledge there are district and regional differences in terms of the types of cases that are filed in the district courts, and that no two district dockets will always look the same, we believe that our study design minimizes circuit biases. The issue categories are defined narrowly so that we do know that district judges in various parts of the country are responding to somewhat similar types of cases. In using logistic regression with standard errors estimated with a robust variance estimator, we minimize the bias of a particular judge's voting pattern over time. We acknowledge that there may be circuit differences that are not taken into account by the model as a limitation of our study. However, our review of the existing literature suggests that over a long time period, differences are most marked by South versus non‐South regional variation, especially in many of the issue areas that we examine. 8. The correlation between Democrat and Democrat×female is .984. 9. The correlation between South and South×female is .977. 10. Note that the female×South interaction term was dropped from this model by Stata. 11. Female‐Democrat index: 1 = Male Repub., 2 = Female Repub., 3 = Male Dem., 4 = Female Dem. 12. Female‐Democrat index: 1 = Male Repub., 2 = Female Repub., 3 = Male Dem., 4 = Female Dem. 13. Female‐Democrat index: 1 = Male Repub., 2 = Female Repub., 3 = Male Dem., 4 = Female Dem. 14. Female‐Democrat index: 1 = Male Repub., 2 = Female Repub., 3 = Male Dem., 4 = Female Dem. 15. Female‐Democrat index: 1 = Male Repub., 2 = Female Repub., 3 = Male Dem., 4 = Female Dem. 16. Female‐Democrat index: 1 = Male Repub., 2 = Female Repub., 3 = Male Dem., 4 = Female Dem. 17. Female‐Democrat index: 1 = Male Repub., 2 = Female Repub., 3 = Male Dem., 4 = Female Dem.