Measuring Real Activity Management

Abstract

Much recent research focuses on earnings management via managers manipulation of real activities (real earnings management, REM). While tests of abnormal REM hinge critically on the measurement of normal real activities, there is no systematic evidence on the properties of commonly used REM measures. We provide such evidence by documenting which REM measures lead to well-specified hypotheses tests. We find that the REM measures used to date in the literature, as well as those we develop to capture variation in expected real activities due to characteristics of firms’ investment opportunity sets, or to control for non-linearity in the relation between performance and expected real activities, are severely mis-specified in that their Type I error rates differ significantly from nominal significance levels. While performance-matched REM measures are not well specified in each and every setting (no REM measure is), the weight of the evidence suggests that, across a wide variety of research settings such as those encountered in accounting, performance-matched REM measures will provide better-specified tests than other REM measures. Such evidence highlights the importance of using performance-matched REM measures, when testing hypotheses related to managers’ incentives to manipulate real activities, to ensure that reliable inferences are drawn.