What Motivates the Justices: Utilizing Automated Text Classification to Determine Supreme Court Justices' Preferences
35 Pages Posted: 10 Mar 2015 Last revised: 20 May 2015
Date Written: April 1, 2015
Abstract
Works examining Supreme Court judicial behavior take the justices' votes on the merits as the typical outcome variable. In this paper we develop and test a new method of automated text classification that makes use of multiple features to generate more accurate predictions. We use a support vector machine (SVM) learning algorithm to create preference models for individual justices. With a dataset covering five justices over seventeen terms, we show that with text classification, we can focus on inputs to judicial decision-making such as litigants' briefs which allow for a new means of assessing judicial attitudes. In particular, we show that text classification's ability to locate salient language allows us to better understand the justices' preferences than through traditional ideological explanations. The models in this paper achieve near 80% predictive accuracy of the justices' votes.
Keywords: judicial behavior, empirical legal studies, briefs, First Amendment, text classification, Supreme Court
Suggested Citation: Suggested Citation