Implications of Multiple Structural Changes in Event Studies

The detection of unbiased abnormal returns in the classic event study depends on the validity of the assumption that the parameters of the return generating process remain constant throughout the sample period. However, given the substantial amount of evidence to support the fact that the market model parameters are not stable over time, previous research has suggested that the classic methodology may result in mismeasurement of the magnitude and timing of the market response and lead to incorrect conclusions regarding the impact of the event. This study addresses the issue of parameter nonstationarity in event studies by examining its implication for measuring the market response. Using an unconstrained estimation model that allows multiple structural changes in the security pricing model, our results indicate that the usual finding of statistically significant positive abnormal returns on and around the split announcement is robust to the more general specification of the underlying return generating process