2
these statistical models typically do not concern themselves with the substantive content of written
opinions. For good or ill, these analyses discard most of the policy information provided by judges.
In effect, the empirical research ignores what is regarded by more traditional legal scholars as
crucial to an adequate understanding of judges and the law.
Can the content of judges= written opinions be regarded as data for empirical analysis? We
believe that it can. In this paper, we offer an illustration by applying a method of content analysis
to the opinions of the justices of the U.S. Supreme Court. By relying upon the Wordscore
procedure (Laver, Benoit, and Garry 2003), we are able to extract valid policy positions from the
text of written opinions for a series of decisions in the areas of religion and search and seizure.
Although we are circumspect about our results, we believe that our findings suggest that the written
opinions of the justices are a potentially significant source of data, which can be mined not only to
buttress existing quantitative models but also to offer new perspectives on judicial policy.
Supreme Court Opinions and the Problems of Content Analysis
One of the fundamental aims of any political account of the U.S. Supreme Court is to
characterize the ideological tenor of its decisions. It has long been recognized that the policy
preferences of the justices play a substantial role in their decisions, and the most comprehensive
account provides abundant evidence that the ideological dispositions of the justices are the
principal determinant of the Court=s policies (Segal and Spaeth 2002). The theory that guides this
work is straightforward enough; liberal justices produce liberal decisions, and conservative justices
make conservative choices.
Demonstrating that preferences explain votes, however, does not necessarily provide a
sufficient understanding of the relative differences between policies. More specifically, the link
Convention