操纵性应计利润模型检测盈余管理能力的实证分析

本文运用统计模拟方法,通过测试各模型产生第一类错误和第二类错误的频率,比较基本琼斯模型、修正琼斯模型等七种常用的截面操纵性应计利润模型在中国资本市场的检验效果.检测结果发现,基本琼斯模型、修正琼斯模型和无形资产琼斯模型相对较优,它们所犯第一类错误和第二类错误的频率较小,但在收入操纵检测上,修正琼斯模型检验能力更突出;现金流量琼斯模型、非线性琼斯模型检验盈余管理的能力较强,但存在较为严重的第一类错误,易夸大上市公司盈余管理的程度;前瞻性修正琼斯模型计算复杂,第一类错误明显,检验盈余管理能力相对修正琼斯模型没有明显提高;收益匹配琼斯模型在费用操纵的检验能力上明显较弱.综合考虑,本文认为在中国证券市场上,分年度分行业回归的截面修正琼斯模型在模型的设定和盈余管理的检验能力方面表现更佳,建议未来的盈余管理实证研究以该方法为主.

An Evaluation on Specification and Power of Discretionary Accruals Models in the Chinese Context

This paper evaluates alternative discretionary accruals models for detecting earnings management. This evaluation uses a commonly used statistics simulation method to compare the specification and power of different discretionary accruals models in the Chinese context. Specifically, using the Chinese stock market data, we evaluate the Jones Standard model, the Modified Jones model, the Jones model with intangible assets, the Jones model with ROA, The forward-looking modified Jones model, the Jones model with cash flow, and the Non-linear Jones model. By examining the probability at which alternative discretionary accruals models commit Type-Ⅰ and Type-Ⅱ errors in detecting earning management, this paper provides the following major insights. First, the Jones standard model, the Modified Jones model and the Jones model with intangible assets exhibit well specified and the better power in detecting earnings management, especially the Modified Jones model, which is more powerful than its standard version at detecting revenue-based manipulation. Second, the Jones model with cash flow and the Nonlinear Jones model exhibit the most power in detecting earning management, but the two models appear worse specified when applied to a random sample of firm-years. Third, the forward-looking modified Jones model is complicated in computation, but it doesn't improve the power in detecting earnings management than the original modified version. Finally, the Jones model with ROA performs worst at detecting expense-based earnings management. One of the implications of the above-mentioned conclusions for the future earnings management research in the Chinese stock market is that it is difficulty to get a reliable result to directly use the discretionary accruals models used by the research in the American stock market. We think it is advisable to put the cross-sectional modified Jones model in priority when we use the discretionary accruals models in our research of the earnings management in the Chinese context.