Computational Reproducibility in Finance: Evidence from 1,000 Tests

Abstract

We analyze the computational reproducibility of more than 1,000 empirical answers to six research questions in finance provided by 168 international research teams. Running the original researchers’ code on the same raw data regenerates exactly the same results only 52% of the time. Reproducibility is higher for researchers with better coding skills and for those exerting more effort. It is lower for more technical research questions, more complex code, and for results lying in the tails of the results distribution. Neither researcher seniority, nor peer-review ratings appear to be related to the level of reproducibility. Moreover, researchers exhibit strong overconfidence when assessing the reproducibility of their own research. We provide guidelines for finance researchers and discuss several implementable reproducibility policies for academic journals.

Keywords: computational reproducibility, open science, credibility of research, multi-analyst study, data-availability policy, scientific publishing

JEL Classification: C80, C87