False discoveries in the anomaly research:new insights from the Stock Exchange of Melbourne (1927–1987)

Are equity anomalies a product of p-hacking in the asset pricing literature? To shed new light on this question, we perform a true out-of-sample study of 30 well-known anomalies in the cross-section of returns. We replicate these anomalies in a novel hand-collected dataset of firms listed on the historical Stock Exchange of Melbourne in the years 1926 to 1987. The vast majority of return-predictive signals cannot be confirmed. Those which are observed are commonly driven by small firms with marginal economic significance. Only a handful of anomalies survive our tests, namely, the dividend yield, value uncertainty, and short-term residual reversal effects. Overall, our findings support the view that many anomalies are statistical artifacts resulting from data mining.