11
1998-1999) ranges from six to twenty
5
. Court size (The Book of the States, 1998-1999) is coded by the
size of each state supreme court, with office size varying from five to nine justices.
Lastly, court
professionalization (Brace and Hall, 2001) is a factor score of several state supreme court characteristics,
including salary, workload, and staff size.
Regarding the political environment and the costs related to office, two independent variables are
included within the model, electoral competition and state ideological extremism. First, Holbrook and
Van Dunk’s (1993) measure of electoral competition is used as proxy measure for judge level
competition.
Second, the extreme state ideology variable is measured using Erikson, Wright, and
McIver’s (1993) state liberalism measurement, which is converted into the distance of each state’s
liberalism score subtracted from the national mean using the absolute value of the difference.
Consensus Model
For the model of consensus, the unit of analysis is at the case level. The dichotomous dependent
variable, unanimity, is coded 1 for those decisions that are unanimous and 0 for all other non-unanimous
decisions.
Because an ordinary least squares design is inappropriate when using a dichotomous
dependent variable (see Long, 1997) the parameters for the models were estimated using a maximum
likelihood estimation, specifically a logit design.
Furthermore, the model is operationalized using a
pooled cross-sectional time-series design to account for multiple states within the analysis.
6
To control
for heteroskedasticity, the model utilizes Huber-White robust standard errors to provide correct standard
errors in the presence of my violations to the assumptions within the model (see Long and Freese, 2001).
The primary explanatory variable of interest is the measure of ideological diversity. To better
understand the relationship between court diversity and decisional outcomes, including consensus, the
model uses ideological diversity as a determinant of unanimous outcomes within murder appeals.
Ideological diversity is again derived using the range of ideology (Brace, Langer, and Hall 2000; Langer,
2001) within every state supreme court from 1995 to 1997.
7
To test the hypotheses related to court structure, the model includes measures of court size, the
presence of lower appellate courts, and non-discretionary forms of opinion assignment. As within the
diversity model, the continuous measure of court size ranges from six to twenty. Intermediate appeals
courts (American Judicature Society, 2001) is a dichotomous measure of the presence of a lower appellate
courts. Where lower appellate courts operate, the variable is coded 1 and where no lower appellate courts
exist, the variable is coded 0. Lastly, random opinion assignment rules(Hall, 1990) are coded 1 for those
5
Restricted term lengths actually vary from six to fourteen years with another three states granting lifetime
appointments upon selection. The value of twenty represents those states with lifetime tenure appointments.
6
To adjust for year effects, 1996 and 1997 are controlled within the analysis. Measures of gender, racial, and tenure
diversity are also included to control for varying forms of diversity and their impact on consensus.
7
The Oklahoma Supreme Court and the Texas Supreme Court where excluded from this analysis, as they only hear
civil cases. All murder decisions were heard in each state’s criminal court of last resort.
Convention