Event Studies for Merger Analysis: An Evaluation of the Effects of Non-Normality on Hypothesis Testing

Abstract:     
Financial event studies using daily stock returns are frequently employed in the analysis of mergers to estimate the sign and magnitude of stock movements to particular merger announcements. A common method of conducting the event study is least squares regression with dummy variables. Daily stock returns, however, are typically non-normally distributed, potentially rendering the hypothesis tests on the least squares coefficients invalid if based on asymptotic critical values. We present evidence on the non-normality of daily stock returns and the consequences of it on critical values using a bootstrap technique. We find that non-normality can lead to substantial departures from the asymptotic critical values and large asymmetries. Both under- and over-rejection of the null hypothesis are possible depending on the particular form of the non-normality.

Number of Pages in PDF File: 27

Keywords: Event Study, Normal Distribution, Hypothesis Testing, Bootstrap, Generalized Bootstrap

JEL Classification: C12, C14, C15

working papers series