Forecasting the conditional distribution of realized volatility of oil price returns: The role of skewness over 1859 to 2023

Abstract

We examine the predictive value of expected skewness of oil returns for the realized volatility using monthly data from 1859:11 to 2023:04. We utilize a quantile predictive regression model, which is able to accommodate nonlinearity and structural breaks. In-sample results show that the predictive impact of expected skewness on realized volatility can be both positive and negative, with these signs contingent on the quantiles of realized volatility. Moreover, we detected statistically significant forecasting gains that arise at the extreme ends and around the median of the conditional distribution of realized volatility. Our results have important implications for investors and policymakers.