In event studies abnormal returns are assumed to be cross sectionally independent. If the event day is common, and if the firms are from the same industry, abnormal returns may not be independent across firms. Tests for event effects are invalid under cross sectional dependence. We propose to model the dependence in abnormal returns by a spatial error model. In the model abnormal returns of firms belonging to the same sector or industry are correlated, but abnormal returns of firms belonging to different sectors or industries are uncorrelated. The spatial error model formalises weak dependence. Corrected tests for event effects under spatial dependence are derived and shown to be asymptotically normal. The correction depends on the spatial covariance matrix. A Monte Carlo study shows that moderate spatial autocorrelation causes tests for event effects to overreject the null hypothesis of no event effect, whereas the corrected tests have the nominal level and nontrivial power. An empirical application to US stock returns around Lehman Brothers' bankruptcy illustrates the importance of correcting for cross sectional dependence in abnormal returns.
|Effective start/end date||05.02.2009 → 31.12.2012|
Explore the research topics touched on by this project. These labels are generated based on the underlying awards/grants. Together they form a unique fingerprint.