CoAnomaly: Correlation Risk in Stock Market Anomalies

Abstract

I propose a simple time-series risk measure in trading stock market anomalies, CoAnomaly, the time-varying average pairwise correlation among 34 anomalies, which helps to explain both the time-series and the cross-sectional anomaly return patterns. Since correlations among underlying assets determine the portfolio variance, CoAnomaly is an important state variable for arbitrageurs who hold diversified portfolios of anomalies to boost their performance. Empirically, I show that, first, CoAnomaly is persistent and forecasts long-run aggregate volatility of the diversified anomaly portfolio. Second, CoAnomaly positively predicts future average anomaly returns in the time series. Third, in the cross-section of these 34 anomaly portfolios, CoAnomaly carries a negative price of risk. These return patterns suggest that arbitrageurs take the time-varying correlation into account and their intertemporal hedging demand plays an important role in setting asset prices.