Estimating discretionary accruals in the cross-section of firms: a reinterpretation of the Jones model and its variants

Abstract

Although many studies use the Jones (1991) model and its variants to estimate discretionary accruals, the choice of accrual model is typically ad hoc. In this paper, we challenge a key underlying assumption of the Jones model that there is no information available about the cross-sectional distribution of the "true" profitability of firms. When we incorporate this neglected information into a Bayesian framework, the resulting model provides both a theoretical foundation and practical guidance for researchers on which variant of the Jones model is suitable in which setting. Specifically, we find that the appropriate use depends on the sampling of firms suspected of earnings management: (i) the modified Jones plus cash flow model for random sampling, (ii) the modified Jones plus profitability model for sampling on firms' reported earnings, and (iii) the modified Jones model for sampling on the Bayesian estimate of firms' true profitability.

Keywords: Earnings management, Jones model, Discretionary accruals, Normal accruals, Bayesian analysis, Total accruals. JEL: M41, G17

JEL Classification: M41, G17