On regression based event study

This paper revisits event-study methodology based on regression estimation of abnormal returns. The paper reviews the traditional event study and gives a more detailed discussion of the regression based approach with quantitative event variables. The paper discusses also briefly the the dummy variable regression which is a special case of the quantitative case. Use of GARCH to predict event period volatility is suggested as a partial remedy for the event induced volatility. Furthermore, it is demonstrated that in order to optimally estimate the average (cumulative) abnormal returns, maximum likelihood (ML) or weighted least squares (WLS) type estimators should be used. As an empirical illustration the impact of inflation shocks on stock market index returns is examined. Particularly it is demonstrated that although the data suggest a volatility increase around the inflation release dates, modeling the residuals with GARCH adequately captures the changing volatility, and no actual level shift in volatility can be empirically detected.