A New Approach to Performance Control in Earnings Management Studies

We propose a cash flow from operations (CFO) portfolio approach to performance control issue in earnings management studies, which was initiated Kothari et al.(2005). We argue that the CFO portfolio is less costly, and yet controls performance more properly and ensures more stable and consistent measures of discretionary accruals than the Kothari et al. approach. We address the issues of alternative models as well as estimation approaches to make the study more comprehensive. We carry out diverse analyses and tests and document some important findings. First, the new models proposed by Yoon et al.(2013) are superior to the Jones and modified Jones models. Second, the industry approach results in unsigned or inconsistent regression coefficients across different industries. Third, the CFO portfolio approach significantly outperforms the industry approach, indicating that the estimation approach is more important than the models. Fourth, there are significant differences in mean nondiscretionary and discretionary accruals between different models when the industry approach is used. However, the between-model differences disappear when the CFO portfolio approach is used. Fifth, the Jones models do not properly isolate discretionary accruals particularly when the industry approach is used. We find that the application of a better estimation approach is more important than the choice of models. Sixth, CFO levels and estimation approach matter in terms of type-1 errors. There are larger differences in the rejection rates between the estimation approaches than between the models. However, the rejection rates are mostly lower than the hypothesized rate for the middle CFO level firm-years. Seventh, there are risks of drawing wrong inferences when the ROA control approach is used. Eighth, we document that a significant proportion of negative CFO firms employ income-increasing earnings management strategy to avoid loss reporting, and that there are considerable number of big-bath firms particularly when their CFO level is very low. Lastly, discretionary accruals can be overestimated (underestimated) when CFO level is negative (positive) under the industry approach, indicating that use of the CFO control approach is more warranted.