The Limits ofp-Hacking: Some Thought Experiments

Abstract

Suppose that the 300+ published asset pricing factors are all spurious. How muchp-hacking is required to produce these factors? If 10,000 researchers generate eight factors every day, it takes hundreds of years. This is because dozens of publishedt-statistics exceed 6.0, while the correspondingp-value is infinitesimal, implying an astronomical amount ofp-hacking in a general model. More structure implies thatp-hacking cannot address≈100 publishedt-statistics that exceed 4.0, as they require an implausibly nonlinear preference fort-statistics or even morep-hacking. These results imply that mispricing, risk, and/or frictions have a key role in stock returns.