Combining the determinants of earnings management : a propensity score-matched model

Our work focuses on earnings management, a common practice according to which managers exercise their discretion, in the accounting process, in order to achieve their goals through the manipulation of some components of reported income. The main objective of this study is to propose a new approach that is able to improve the performance of the existing models in detecting this practice. Previous researchers, in fact, show that traditional models (e.g., Jones and modified-Jones models) usually exhibit a strong misspecification that can be associated to the effect of omitted variables that affect earnings management. Kothari, Leone and Wasley (2005) try to overcome this problem providing a new approach that controls for firm's performance. Their results suggest that, including an additional determinant in the analysis, namely the Return on Assets, they actually achieve better results in terms of specification, compared to the existing models. Consistent with this finding, we decide to verify whether including other determinants, that the literature about earnings management identifies as potentially relevant, may help in improving the performance of the existing models. In order to effectively combine the effect of several determinants in an individual variable, we adopt an approach based on the propensity score matching technique (Rosenbaum and Rubin, 1983). Among the potential determinants of earnings management identified by previous studies, those that combine in a parsimonious and well-performing model are Ohlson's O-Score, Earnings-to-Price ratio, Debt-to-Assets ratio and account receivables scaled by total assets. Simulations results suggest that the model based on propensity score matching tends to be the best method in detecting earnings management, since it appears to exhibit a higher statistical specification than existing models without a loss of power across a wide variety of simulated event conditions.