Bias in Securities Litigation Event Studies When Volatility and Beta Shift

Abstract Traditional event study methodology can lead to erroneous conclusions—incorrect identification of an insignificant price change as statistically significant (and vice versa)—when there are volatility shifts, and the model is estimated from a prior low volatility period. Earlier studies have used the empirical distribution of the test statistic to adjust for the fatter tails created due to volatility shifts. We show that the empirical distribution approach (EDA) can only account for a shift in volatility and leads to biased results when there is a simultaneous shift in the beta. We propose using a regime switching (RS) model to estimate beta and volatility for different regimes. We show that using regime-specific beta and volatility removes the bias in excess returns caused by estimating the parameters from a prior period, with different beta and volatility. Our findings have practical implications for securities litigation, where event studies serve as key evidence. We highlight the need for re-evaluating excess return methodologies in high volatility environments and provide a framework for improving event study robustness in legal and regulatory contexts.