Feedback and Learning: The Causal Effects of Reversals on Judicial Decision-Making

Do judges respond to reversals of their decisions? Using random assignment of cases across two stages of the criminal justice system in Norway and a novel dataset linking trial court decisions to reversals in appeals courts, we provide causal evidence on feedback effects in judicial decision-making. By exploiting differences in the tendencies of randomly assigned appeal panels to reverse trial court decisions, we show that trial court judges who receive a reversal of a sentence respond by updating the likelihood of imposing a prison sentence in the direction of the reversal in future cases. Consistent with a Bayesian learning model, we find that the responses are stronger for judges with weaker priors and for reversals corresponding to stronger signals. Our estimates, however, also indicate that judges overreact to reversals compared to Bayes' rule.