Looking Under the Hood of Data-Mining

Abstract

This paper re-evaluates academic research on 92 cross-sectional stock return predictors. Researchers studying return predictability must make decisions about portfolio construction; for example, whether to rebalance annually or monthly. In sample, the returns of portfolios constructed with the precise decisions made in the predictors’ papers are 0.23% per month larger than those of portfolios constructed with a random combination of decisions made in the literature. Out of sample, more than half of this difference disappears. Predictors published in top-ranked journals show a pronounced effect. The results are consistent with decision mining that produces biased return estimates.

Keywords: Stock Return Predictability, Anomalies, Research Decisions, Statistical Biases, Portfolio Construction, Decision-Mining

JEL Classification: G11, G12, G14, G00, C12, C18, C18, C1, C2, B4