Large sample size bias in empirical finance

Abstract

The vast majority of empirical studies in finance employ large enough sample sizes and use the conventional thresholds for statistical significance. This routine practice can potentially lead to spurious statistically significant results. The primary aim of this paper is to present a rule of thumb that can be used to determine the appropriate thresholds for statistical significance for a given sample size. The paper argues that the list of statistically significant findings in the broader finance literature is likely to be much shorter after accounting for large sample size bias.