A Bayesian Approach to Event Studies for Securities Litigation

We propose a Bayesian method for econometric event studies commonly usedin U.S. securities litigation. We show that our approach may be based on the Bayes factor, which has a simple form when inference is based on the empirical distribution function of abnormal returns; it also avoids problems related to nonnormality of abnormal returns. We use data from litigation related to alleged fraud by the Apollo Education Group (University of Phoenix's parent) to illustrate the method. Results are similar to frequentist hypothesis testing with a large event-date effect, but they can be importantly different with a small or moderate effect.