| Summary: | 碩士 === 中原大學 === 會計研究所 === 93 === This study investigates the relative ability of various accrual models in detecting earnings management. Prior research shows that extant discretionary accrual models are misspecified when applied to firms with extreme financial performance. Therefore, we focus on the Jones, modified Jones model and their variations to control firm performance (e.g. inclusion of ROA or CFO in the accrual models). We evaluate specification and power of each accrual model under our investigation. The accrual models are estimated cross-sectionally using two sets of data, parent-only financial statements and consolidated financial statements. We follow the steps outlined in Dechow, Sloan, and Sweeney (1995) to assess the specification and use a simulation procedure developed by Brown and Warner (1980) to assess the power of each model.
The empirical results suggest that the Jones and modified Jones model including the change of cash flow from operations is best specified and most powerful in our study. The inclusion of return on assets in the accrual models cannot effectively improve model specification and power. On the perspective of utilizing parent-only or consolidated financial statements, some models perform better under parent-only financial statements (e.g. inclusion of ROA in the model), and some perform better under the consolidated financial statements (e.g. the original Jones and modified Jones model).
Our paper contributes to the literature by demonstrating the superiority of the model proposed by Kasznik (1999). This model has been ignored by earnings management research. We suggest that future study should test the performance of this model using other contexts. Further evidence in other countries will shed more light on the usefulness of this model.
|
|---|
