A Factor Model for Correlation Risk Premium Returns

Abstract

Index options on baskets of stocks, in conjunction with the underlying stock options, imply a statistical correlation among those stocks. This implied correlation is generally higher than the realized correlation of the stocks except in tail market events where the realized correlation tends toward 1. We show how it's possible to trade the difference between realized and implied correlation, otherwise known as the correlation risk premium, through a method called dispersion trading. This paper then generates returns from selling the correlation risk premium as well as selling the volatility, skewness, and kurtosis risk premiums, strategies known to have similar return dynamics as selling the correlation risk premium. Using these returns, in conjunction with Fama-French equity factor returns, we construct a factor model for our correlation risk premium returns. Overall, we find that correlation risk premium is highly correlated with volatility risk premium while also providing additional alpha. In addition, we find that correlation risk premium is negatively and negligibly correlated with kurtosis and skew risk premiums, respectively, but only after controlling for the volatility risk premium. This indicates the assumption that the good ’carry’ from selling option risk premiums are correlated is more nuanced than assumed.