Predicting energy futures high-frequency volatility using technical indicators: The role of interaction

Abstract

In this paper, we investigate the predictability of technical indicators on energy futures volatility from the high-frequency and high-dimensional perspectives. We show that the technical indicators have significant impacts on crude oil and natural gas futures volatility based on in- and out-of-sample analysis. Further, we analyze the impacts of interactions among predictor variables on future volatility. Based on an improved conditional sure independence screening model, we find that the interactions contribute to the out-of-sample predictive power significantly. The improved model has robust and better forecasting performance relative to extant popular dimension reduction methods, forecast combination methods, and regularization methods. Moreover, we show that the out-of-sample predictability is robust during various periods. Finally, we show that technical indicators improve economic value in the crude oil market but the economic increment is not significant in the natural gas market.