Abstract:
Mining 29,000 accounting ratios for t-statistics over 2.0 leads to cross-sectional predictability similar to the peer review process. For both methods, about 50% of predictability remains after the original sample periods. Data mining generates other features of peer review including the rise in returns as original sample periods end, the speed of post-sample decay, and themes like investment, issuance, and accruals. Predictors supported by peer-reviewed risk explanations underperform data mining. Similarly, the relationship between modeling rigor and post-sample returns is negative. Our results suggest peer review systematically mislabels mispricing as risk, though only 18% of predictors are attributed to risk.