Title: A new approach to evaluating earnings management models
Authors: Giseok Nam; June Woo Park
College of Business, Hankuk University of Foreign Studies, 107 Imun-ro, Dongdaemun-gu, 02450 Seoul, Korea
Schulich School of Business, York University, 4700 Keele St, Toronto, ON M3J 1P3, Canada
Abstract: This paper proposes a new full-scale simulation approach for correctly measuring the detection ability of the modified Jones model (MJM). For ideal evaluation accuracy, we define perfect prior information on whether a firm under investigation manipulates its earnings with the allowance for doubtful accounts by differentiating firms that engage in no earnings manipulation from those that do. The MJM shows a detection rate of only 49.1%, which is lower than 50% for a fair coin toss, and thus it is not reliable. This low detection rate is due to the oversensitivity of the MJM, which causes it to misinterpret long-term increases in sales revenue from normal business activity as earnings management. Therefore, the MJM has a tendency to incorrectly detect non-discretionary accruals (NDACs) as discretionary accruals (DACs) and thus misjudge a clean firm as a dirty one, because it shows a low (moderately high) detection rate for clean (dirty) firms.
Keywords: detection rate; discretionary accruals; earnings management models; hypothetical firms; modified Jones model; model oversensitivity; perfect prior information; simulation; earnings manipulation; model evaluation; sales revenue.
Int. J. of Managerial and Financial Accounting, 2016 Vol.8, No.3/4, pp.247 - 269
Submission date: 10 Dec 2015
Date of acceptance: 27 Jun 2016
Available online: 25 Jan 2017