Risk-Neutral Cumulants, Expected Risk Premia, and Future Stock Returns

This paper builds an empirical model to connect option-implied cumulants with expected risk premia through latent risk factors. Expected risk premia on individual stocks are estimated by applying a new partial least squares-based method on risk-neutral cumulants at different orders and various maturities. The filtered expected risk premia based on the second and third order risk-neutral cumulants exhibit a considerably large dispersion across stocks, which further generates a wide cross-sectional variation in future realized risk premia. I find a positive relationship between the ex-ante filtered expected risk premium and future realized risk premium during the period of 1996-2017. A strategy that goes long the decile portfolio with the largest filtered expected risk premium and short the decile portfolio with the smallest filtered expected risk premium yields a Fama-French-Carhart alpha of 1.06% per month (t-stat 3.75). The filtered expected risk premium of the above trading strategy is correlated (Corr: 51.5%) with the equal-weighted stock's expected excess return in Martin and Wagner (2018). Moreover, I show that the predictability of the filtered expected risk premium can be potentially explained by informed trading driven by short-selling constraints