Xavier Garza - Gomez Yong Lee Jianjun Du Abstract: This paper explores a new trend in discretionary accrual models proposed in the literature by empirically testing the models’ capacity to identify earnings management on 361 cases obtained from the U.S. GAO database. Estimates of discretionary and non discretionary accruals are compared for the models to check for better fit and then their capacity to detect earnings management is assessed with samples where earnings management is expected. Initial statistical tests suggest that recent models that incorporate cash flows as explanatory variable of accruals are statistically superior to the traditional and the modified version of the Jones model. However, when tested in the sample of earnings restatement cases, the competing models show the same power to detect earnings management. |