Abstract

We quantify the extent to which justices in the US Supreme Court learn from each other when voting on the merits of a case. We analyze conference votes, which have been historically cast behind closed doors in order of seniority. We exploit exogeneous variation in justices’ voting position and provide causal evidence that junior justices systematically incorporate the votes of their senior colleagues. Casting a vote one position later in the sequence increases the probability that a junior justice follows her senior colleagues in 25%. To assess the effect of the voting rule on justices’ learning beahavior and Court’s outcomes, we develop an empirical model of sequential voting in the Court. In the model, justices make decisions under incomplete information and incorporate their preferences, public and private information, as well as the choices of previous justices in the voting sequence. We show that the information contained in the voting history affects the outcome of the Court in 18% of cases. We compute the effect of sequential voting by seniority on the quality of Supreme Court’s rulings and compare it to alternative voting mechanisms such as simultaneous and anti-seniority voting.