HAC Standard Errors and the Event Study Methodology : A Cautionary Note

In support of Fomby and Murfin (2005), we demonstrate empirically, rather than theoretically, the severe consequences of using HAC standard errors in regression-based financial event studies. Applying an event study to a recent merger, we show that the use of HAC standard errors render misleading conclusions. Critical values for t-tests on the event dummy variables are about 15 times larger than the nominal values using only a year of daily return data. Even with samples of only 100 returns, critical values exceed nominal critical values by a factor of 10