Did Accrual Earnings Management Decline and Real Earnings Management Increase Post-SOX? A Re-examination and Replication

Abstract

A widely cited paper, Cohen, Dey, and Lys (2008, hereinafter CDL), examines accrual (AEM) and real earnings management (REM) pre- and post-Sarbanes-Oxley. It has been more than a dozen years since CDL's publication and almost 20 years since SOX became law. Our re-examination analyses investigate whether CDL's findings are sensitive to a battery of robustness checks regarding sample construction, model specification, and variable definitions, and whether CDL's conclusions hold after 2005, the end of their sample period. We find support for many of CDL's conclusions, but also evidence suggesting the need to adjust our understanding of the use of AEM and REM post-SOX. CDL's time trends evidence of decreasing AEM and increasing REM is sensitive to some research design choices; and the substitution between AEM and REM that CDL find is attenuated post-SOX, thus cautioning against automatically assuming AEM and REM are substitutes. We also replicate CDL with mixed success.