Issues in Testing Earnings Management and an Instrumental Variable Approach
指出在检验盈余管理时，由于管理者操纵会计数字，常用预测模型存在联立性和测量误差问题，并提出一种基于工具变量的简单方法来解决这些特定情境下的检验问题。
In empirical studies of earnings management (EMI), a central issue is the estimation of the managed component (discretionary accruals) when outsiders observe only the sum of the managed and unmanaged (nondiscretionary) accounting numbers. Important methodological issues arise in part because the variables most useful in predicting the unmanaged components are themselves accounting numbers which are likely to be affected by EM. In this paper, we propose a simple method of addressing some of these issues in testing for EM in context-specific cases.1 Our method builds on the approach used by McNichols and Wilson [1988]. By modeling a specific type of accruals (bad debt expense), they enhance precision but do not capture accrual manipulations in many other accounts. Other studies attempt to capture a larger portion of managed accruals by modeling total accruals. However, we suggest that models following this approach are subject to simultaneity, errors-in-variables,