Uncertainty‐driven oil volatility risk premium and international stock market volatility forecasting

Abstract Forecasting international stock market volatility using the oil volatility risk premium (OVRP) is important for asset allocation and financial risk management. In previous literature, the issue of biased OVRP has not been well addressed. The OVRP is defined as the difference between the ex ante risk‐neutral expectation of oil return variation and the oil realized variance. This is a biased estimator as argued by previous studies. In this paper, we propose a new measure of OVRP, namely, the uncertainty‐driven OVRP (UOVRP), in which the expected realized variation is the conditional variance of oil futures returns estimated by a typical GARCH‐MIDAS model, and the long‐term volatility component of conditional variance is directly determined by several uncertainty indices. Using heterogeneous autoregressive models (HAR), we find that UOVRPs significantly and positively predict most of the international stock market volatility in‐sample. Furthermore, it significantly outperforms the realized volatility, the “model‐free” OVRP, and recently proposed volatility predictors for most American, European, and Pacific markets out‐of‐sample. On average, UOVRPs driven by the world uncertainty index and the U.S. economic policy uncertainty perform the best among all UOVRPs. The predictive ability of UOVRPs is robust across a series of checks.