[nearby] colleagues on the trial court bench for help with the legal and administrative problems
of the judgeship.” (Carp and Wheeler 1972a, p. 28). Given all that we know about judicial
interaction, we felt it more theoretically justifiable to utilize city (or “court point”), rather than
overall court, as the basis for computation.
Using textual analysis, the published opinions were evaluated for the direction of the
decision outcome and coded dichotomously, liberal or conservative. The coding standards and
use of published decisions are well established in the literature, and similar ideological coding
methodology has also been used in comprehensive decision data sets on the U.S. Supreme Court
(Spaeth 1999) and U.S. Courts of Appeals (Songer 1998).
Because prior research has revealed divergent decision making patterns in different
policy areas, the decisions were aggregated into three frequently used case categories: criminal
justice, civil liberties and rights, and economic and labor regulation. Rulings in favor of criminal
defendants, including habeas corpus motions, were coded “liberal.” On the other hand, decisions
that favored a party who alleged discrimination by an employer, or that their First Amendment
speech rights had been violated were coded "liberal," while a decision favoring closer connection
between government and religion was recorded as a “conservative” ruling. Cases that were
decided on behalf of business over the interest of labor (or government regulators) were coded
“conservative”. Cases in which there was no clear outcome, or the liberal-conservative
dimension was unclear, were excluded from analysis.
We created two sets of logit models in order to test our two hypotheses. All models
included the gender variables, coded “1” for women, “0” for men. One set of models tested for
gender influences in critical mass environments. The second set of models was created as a
means of clarifying the results of the first. In this second analysis, we estimated the extent to
which gender influenced jurists’ rulings in non-critical mass environments. Though cross model
comparisons may at times pose some interpretative risks, we believe such risks are minimal in
this instance.
We faced a challenge in determining what constitutes a critical mass of women on the
bench. Ultimately, a critical mass is about group size and the literature provides no set
guidelines for determining when a critical mass may be numerically achieved. We decided to
use a simple and straightforward definition: a critical mass of women occurs when there are two
or more women on a city’s federal district court.
9
Convention