Anomaly Investing: Out-of-Sample Performance and Intertemporal Considerations

Abstract

I first show that the naïve equal-weighted 1/N investing in the set of 34 stock market anomalies is a robust implementation for out-of-sample diversification. Two types of popular portfolio optimization methods, including Sharpe-Ratio-optimizing with weight constraints and Dimension-Reduction with machine learning techniques, do not achieve robustly higher out-of-sample performance. Further to explore the gains and risks in investing stock market anomalies, I take this equal-weighted anomaly portfolio to an intertemporal CAPM framework with stochastic volatility to understand the investment considerations of a specific anomaly investor. Based on my estimation, only the correlation-induced volatility news carries a significant risk premium, which highlights the economic importance of the comovement in anomaly asset prices.

Keywords: Portfolio Strategy, Diversification, Machine Learning, Hedging Demand of Sophisticated Investors

JEL Classification: G11, G23.