Improving Empirical and Theoretic Models of Judicial Choice Involving Case Facts

• In Event: Poster Session - Law and Courts & Federalism and Intergovernmental Relations
  In Poster Session: Poster Session - Law and Courts

Abstract

How do judges decide cases? This question that animates the large body of research on judicial decision making has several, sometimes competing, sometimes complementary answers. One theoretical strain of the literature proposes that judges preferences over case outcomes can be presented as a partition of a "case space," where every dimension of the space is a factual dimension that is potentially relevant for the judge(s). Another empirical strain of the literature includes case facts in empirical models of judicial case disposition choice either as predictors of interest, or to assess legal constraint on judges' behavior or changes in legal doctrine, or simply as controls when other variables are of interest. I adapt Gaussian process classification, a technology from machine learning, to show the link between these previously separate literatures, creating a common framework for formal theoretic and empirical models of judicial choice when case facts are at least one part of their decision process. This model offers improved prediction of judicial choice, corrects for potential misspecification in existing empirical models that incorporate case facts, and provides a structural estimation model for judges' preferred rules in case space games. I illustrate how this framework can be used to study legal constraint and legal change.

Author

• JBrandon Duck-Mayr, Washington University in St. Louis