What Matters: Agreement between U.S. Courts of Appeals Judges
• NIPS (Machine Learning and the Law) Forthcoming
• Journal of Machine Learning Research, Forthcoming
10 Pages Posted: 15 May 2017 Last revised: 25 Dec 2018
Date Written: 2016
Federal courts are a mainstay of the justice system in the United States. In this study, we analyze 387,898 cases from U.S. Courts of Appeals, where judges are randomly assigned to panels of three. We predict which judge dissents against co-panelists and analyze the dominant features that predict such dissent with a particular attention to the biographical features that judges share. Random forest, a method developed in Breiman (2001), achieves the best classification. Dissent is roughly half-driven by case features and half-driven by personal features.
JEL Classification: D03, G02, D8
Suggested Citation: Suggested Citation