Most claimed statistical findings in cross-sectional return predictability are likely true

Code

• github.com/chenandrewy/mostly-true
  OfficialIn papernone★ 5

Abstract

I develop simple and intuitive bounds for the false discovery rate (FDR) in cross-sectional return predictability publications. The simplest bounds require only summary statistics from previous papers, and show the FDR is at most 25% in eight out of nine previous studies. A more refined bound shows the FDR is at most 9%. These studies include Harvey, Liu, and Zhu (2016), who ``argue that most claimed findings in financial economics are likely false.'' I demonstrate how Harvey et al.'s own estimates imply an FDR of 9%, and that their conclusion stems from misinterpreting ``insignificant factor'' as ``false discovery.''

Reproductions