On the Power of Asset Pricing Tests

with J Li, 2022, working paper.

[Paper] Researchers often disagree on how to interpret the evidence from asset pricing tests. For instance, Fama and French (1996) show that their three-factor model helps to explain size and book-to-market related return predictability. However, Daniel and Titman (1997) argue that such evidence has low power against the characteristic-based (anomaly) view. To help resolve these debates, we propose a simulation-based approach to benchmark the evidence from asset pricing tests. Specifically, we apply asset pricing tests to simulated characteristics that predict returns as anomalies. The simulations reveal that tests that employ the same characteristics to form factors and test assets often show nontrivial “explanatory power” for mechanical reasons, and the results obtained in the original papers are based on methods that do not enable us to reject the null hypothesis that all characteristics are anomalies (e.g., Fama and French (1996), Fama andFrench (2015)). Tests where the factors are formed using different characteristics are often not mechanical (e.g., Hou, Xue, and Zhang (2015)). The recently proposed instrumented principal component analysis test (IPCA) has more power to differentiate between factors and anomalies. This simulation-based method is easy to interpret, easy to implement, and can flexibly accommodate different null hypotheses. Overall, our findings imply that proper benchmarking can help us better interpret the evidence from asset pricing tests.