Forecasting Energy Futures Volatility Based On The Unbiased Extreme Value Volatility Estimator

This paper uses the opening, high, low, and closing prices of five energy futures to estimate and model volatility based on the unbiased extreme value volatility estimator (the Add RS estimator). We examine the statistical and distributional properties of the logarithm of the Add RS (Log(Add RS)) estimator for the energy futures and find that the distribution of Log(Add RS) is approximately Gaussian, and hence we propose the use of a linear Gaussian model which incorporates the impact of long memory for Log(Add RS) of the energy futures. We use the appropriate orders autoregressive fractional integration moving average (ARFIMA(p,d,q)) model (ARFIMA-Add RS model) based on the Schwarz information criterion (SIC) to model volatility based on the Log(Add RS) estimator. We evaluate the forecasting performance of the ARFIMA-Add RS model using the loss functions, the regression approach, and the superior predictive ability (SPA) approach and compare the corresponding results with the alternative models from the GARCH family. We make use of realised volatility based on the sum of squares of high-frequency returns as a measure of ex-post volatility for forecast evaluation. Our findings indicate that the ARFIMA-Add RS model performs much better than the alternative models in forecasting realised volatility of energy futures. The study has important implications for governments, policy makers, oils importers, and oil traders.