To Catch a Thief: Can Forensic Accounting Help Predict Stock Returns?

Abstract

An earnings manipulation detection model based on forensic accounting principles (Beneish 1999) has substantial out-of-sample ability to predict cross-sectional returns. We show that the model correctly identified, ahead of time, 12 of the 17 highest profile fraud cases in the period 1998-2002. Moreover, the probability of manipulation estimated from this model (PROBM) consistently predicts returns over 1993-2007, even after controlling for size, book-to-market, momentum, accruals and the level of open short interest. Separating high PROBM from low PROBM firms within each of these characteristic deciles greatly improves long/short hedge returns. Further analyses show that PROBM also helps predict future earnings because of its ability to anticipate the persistence of current years’ reported accruals. Overall, our findings offer significant empirical support for the investment approach advocated by forensic accountants.

Keywords: forensic accounting, stock return prediction, earnings manipulation

JEL Classification: M41, G11, G14