True or spurious long memory in volatility: Further evidence on the energy futures markets

The main goal of this paper is to investigate whether the long memory behavior observed in many volatility energy futures markets series is a spurious behavior or not. For this purpose, we employ a wide variety of advanced volatility models that allow for long memory and/or structural changes: the GARCH(1,1), the FIGARCH(1,d,1), the Adaptative-GARCH(1,1,k), and the Adaptative-FIGARCH(1,d,1,k) models. To compare forecasting ability of these models, we use out-of-sample forecasting performance. Using the crude oil, heating oil, gasoline and propane volatility futures energy time series with 1-month and 3-month maturities, we found that five out of the eight time series are characterized by both long memory and structural breaks. For these series, dates of breaks coincide with some major economics and financial events. For the three other time series, we found strong evidence of long memory in volatility.