Event Study Analysis with Time-Varying Alphas, Betas and Variances : The Case of M&As

This paper proposes a new approach to control the effects of time-varying parameters on the estimates of abnormal returns. Event studies usually assume that the parameters of the market model are stable. Using a sample of firm takeovers, however, I find that this assumption is indeed rejected. The beta remains stable while the alpha and the idiosyncratic variance change significantly. The ex-ante merging likelihood is updated, which in turn shifts the parameters in the pre-announcement period. Furthermore, this paper documents that a benchmark event study, which assumes that the parameters are constant, overestimates significantly the abnormal returns around the announcement date. The size of the average annual bias for the target firms is 23.81%, while for the acquirer firms it is 4.13%. The main source of the bias is an upward shift in the alpha in the pre-announcement period. The benchmark approach hence underestimates the alpha and so estimates lower expected returns in the event window, which in turn boosts the abnormal returns