Improving Return Predictability Using Variance-of-Variance Premiums

This paper reports that the variance-of-variance premium (VVP), the difference between the risk-neutral and physical measures of variance-of-variance, has strong predictability for stock returns, especially at very short horizons. Furthermore, pooling both information on the VVP and the variance premium (VP) can deliver a large amount of statistical and economic gain compared to using either of them alone. These results corroborate the finding of Bollerslev, Tauchen, and Zhou (2009) that volatility-of-volatility risk is a critical driver of time-varying risk premiums. Finally, the results hold in the international stock markets and are robust to traditional predictors, investor sentiment proxies, and funding constraints