Predicting risk in energy markets: Low-frequency data still matter

Are high-frequency data always needed to generate precise forecasts of risk measures in energy markets? This study attempts to shed light on this question. We study whether energy market participants can rely on low-frequency volatility estimators when interested in two market risks: volatility and expected shortfall. Using ten years of data on four of the world's most liquid energy futures contracts - the crude oil benchmarks West Texas Intermediate and Brent, as well as natural gas and heating oil futures - we provide conclusive evidence that while realized volatility models lead to much more accurate forecasts in the short term, medium- and longer-term forecasts based on daily ranges are comparable and, in some cases, even more accurate than their high-frequency counterparts that are computationally more intensive and that require costly data. Next, we present an application to predict extreme price declines - expected shortfall - with low-frequency volatility estimates. For that purpose, we propose a novel complete subset quantile regression model to predict multiple-day-ahead expected shortfall. Our back-testing results show that the new model leads to well-specified price decline forecasts, particularly when used with low-frequency volatility estimates. These results show that depending on the forecast horizon and purpose, low-frequency, publicly available, free of cost and easy to process volatility estimators still matter.