Standard Errors in Event Studies

Abstract

Even if true abnormal returns are uncorrelated, estimated abnormal returns are not. This paper presents a simple formula for the variance of estimated cumulative abnormal returns. Both returns and dummy variable procedures for estimating the standard error correctly, taking account of both intertemporal and contemporaneous correlation of estimated residuals, are discussed. They are then applied to an event study of post-merger performance. It is shown that ignoring either the intertemporal or contemporaneous correlation of residuals can result in significant underestimates of standard errors.