Revisiting analyses of matched samples in corporate governance research

Abstract

We observe divergent approaches to analysis of matched samples in recent corporate governance research. Agrawal and Chadha (2005) and Srinivasan (2005) employ a conditional analysis that takes into account matching in the sample selection; Farber (2005) and Erickson, Hanlon, Mayhew (2006) apply traditional methods that do not. As has been shown in simulations, replications, and proofs in Cram, Karan, Stuart (2009), the failure to take matching into account yields biased estimates and can lead to incorrect inferences. We describe use of matched samples in these papers and how more valid and extensive results might be obtained from reanalysis. We illustrate potential impact of correct methodology upon inferences through replication of Heninger (2001), which investigated causes of audit litigation. We make testable predictions about how results of the corporate governance papers will change upon similar reanalysis. We survey other published studies in corporate governance whose findings are questionable and merit revisiting.