Misspecification testing and robust estimation of the market model and their implications for event studies

Abstract

The consequences of regression misspecification of the market model are examined in the context of event study methodology. Using a well-known event study data set, many of the security return regressions are shown to yild residuals displaying extreme evidence of excess kurtosis, thus calling into question the assumptions underlying the calculation of cumulative abnormal returns (CARS) and associated test statistics. A variety of robust estimation techniques are employed, demonstrating that estimates of the market model parameters can change dramatically. Such changes alter the estimated CARS considerably and, hence, make any inferences drawn from them rather questionable.