Event Study Testing with Cross-Sectional Correlation Due to Partially Overlapping Event Windows
31 Pages Posted: 23 Apr 2018 Last revised: 14 Jul 2020
Date Written: May 2, 2018
Abstract
This article re-examines the issue of cross-sectional correlation. Kolari and Pynnonen (2010) find that, in the case of event-date clustering with the same event window for all firms, relatively low cross-sectional correlation among abnormal returns can seriously bias standard tests to over-reject the null hypothesis of zero average abnormal return. We generalize their approach to take into account windows that partially overlap in calendar time. For this purpose we modify their test statistics to adjust for the average percentage of overlaps in event windows. Also, we report empirical evidence on the application of the proposed test statistics to actual event studies.
Keywords: Abnormal Returns; Event Day Clustering; Clustering Robust Standard Errors; Cross-Sectional Correlation
JEL Classification: G14, C10, C15
Suggested Citation: Suggested Citation