LAUSR

Abstract We evaluate the predictive power afforded by crude oil price volatility relative to widely used variables in the financial literature, such as the dividend yield, earnings-to-price ratio, the default yield spread as well several crude oil price-based variables. From a statistical viewpoint, predictions employing the suggested crude oil price volatility-based measures display a similar pattern as predictions using dividend ratios and interest rates, namely, they have relatively weak out-of-sample power. However, we find that gains in utility for an investor that uses predictions produced under the model employing crude oil price log-realized semivolatilities are statistically significant higher than an investor relying on predictions produced under the competitors as well as the historical average benchmark. We discuss and explain the reasons for our results. Overall, we argue that it is hard not to justify more attention to crude oil price semivolatilities relative to widely used financial and macroeconomic variables.