We construct a "universe" of over 18,000 fundamental signals from financial statements and use a bootstrap approach to evaluate the impact of data mining on fundamental-based anomalies. We find that many fundamental signals are significant predictors of cross-sectional stock returns even after accounting for data mining. This predictive ability is more pronounced following high-sentiment periods and among stocks with greater limits to arbitrage. Our evidence suggests that fundamental-based anomalies, including those newly discovered in this study, cannot be attributed to random chance, and they are better explained by mispricing. Our approach is general and we also apply it to past return–based anomalies.

We construct a “universe” of over 18,000 fundamental signals from financial statements and use a bootstrap approach to evaluate the impact of data mining on fundamental-based anomalies. We find that many fundamental signals are significant predictors of cross-sectional stock returns even after accounting for data mining. This predictive ability is more pronounced following high-sentiment periods and among stocks with greater limits to arbitrage. Our evidence suggests that fundamental-based anomalies, including those newly discovered in this study, cannot be attributed to random chance, and they are better explained by mispricing. Our approach is general and we also apply it to past return–based anomalies. Received October 22, 2015; editorial decision October 27, 2016 by Editor Andrew Karolyi.

Previous studies find mixed evidence on whether institutional investors exploit capital market anomalies. Examining a large sample of accounting-based anomalies, we find that institutions trade in the wrong direction of overreaction anomalies, but in the right direction of underreaction anomalies. These heterogenous trading patterns, rather than reflecting institutions' differential anomaly trading skills, can be simply explained by institutions’ tendency to trade in the same direction as the sentiment of news. Examining earnings news and a comprehensive sample of newswire releases, we find strong support for this explanation. Finally, institutional trading appears to exacerbate (mitigate) mispricing associated with overreaction (underreaction) anomalies.

Shorting flows remain a significant predictor of negative future stock returns during 2010–2015, when daily short-sale volume data are published in real time. This predictability decays slowly and lasts for a year. Long-term shorting flows are more informative than short-term shorting flows. Indeed, abnormal short-term shorting flows do not predict future returns or anticipate bad news. We find that short sellers exploit prominent anomalies. A comparison with the Regulation SHO data indicates that the predictability is much shorter-term during 2005–2007. Short sellers appear to have shifted from trading on short-term private information to trading on long-term public information that is gradually incorporated into prices.

We construct real-time machine learning strategies based on a “universe” of fundamental signals. The out-of-sample performance of these strategies is economically meaningful and statistically significant, but considerably weaker than those documented by prior studies that use curated sets of signals as predictors. Strategies based on a simple recursive ranking of each signal’s past performance also yield substantially better out-of-sample performance. We find qualitatively similar results when examining past-return-based signals. Our results underscore the key role of feature engineering and, more broadly, inductive biases in enhancing the economic benefits of machine learning investment strategies.