Split-Adjusted Stock Price and the Cross-Section of US Stock Returns [Early Draft

Abstract

The price of a security potentially contains incremental information regarding its (unobservable) required rate of return, since price depends partly on required return. We construct a simple model in which we use Bayesian updating to show that the cross-sectional expectation of the required rate of return should be negatively related to the logarithm of split-adjusted price. Our model provides a theoretical explanation of the split-adjusted price anomaly first documented by Brown & Pfeiffer (2007). There is strong empirical evidence linking split-adjusted price with subsequent realised US stock returns. Among more than 100 anomaly hedge portfolios constructed using the same US stock return data and portfolio construction methods, the split-adjusted price decile hedge portfolio generates the highest value-weighted mean return (190 bp/month, t-statistic 8.25) and the highest time-series alphas (Fama & French (1993) 3-factor alpha 171 bp/month, Carhart (1997) 4-factor alpha 212 bp/month and Fama & French (2015) 5-factor alpha of 185 bp/month). We argue that as one of the largest and seemingly most robust anomalies yet documented the split-adjusted price anomaly deserves more attention than it has received so far.