Discussion of An Analysis of the Theories and Explanations Offered for the Mispricing of Accruals and Accrual Components

摘要

KLW offer two pieces of evidence that seem to be inconsistent with the earnings fixation hypothesis as a causal explanation for the accrual anomaly. First, they show how potential sample selection biases in prior studies may have overstated evidence in favor of the earnings fixation hypothesis. Second, they use robustness tests to identify and eliminate 1% of the data, and find that the remaining 99% of the data does not support the earnings fixation hypothesis. How much this evidence changes the reader's priors depends on the reader's assessment of the robustness tests: Is it reasonable to delete outliers and conduct hypothesis tests on the remainder of the data? KLW's robustness tests may be appropriate in their specific setting, but in general, deleting data based on the robust regression techniques employed and advocated by KLW seems inappropriate. As shown in detail by Kothari, Sabino, and Zach [2005], cautioned by Knez and Ready [1997], and illustrated here, deleting extreme observations from skewed return data leads to biased estimates and can bias inferences. KLW have generated an interesting set of results, but the way the results are generated—a biased estimator based on outlier deletion from skewed return data—leaves questions and some discomfort.