Abstract
In this paper, we explore the long-range memory on energy markets volatility and value-at-risk (VaR). The main question is: can we estimate better the VaR if long memory exists? To investigate this question several GARCH-type processes, including the FIGARCH process, have been implemented to some main energy products' daily prices (January 1986 to July 2007). Value-at-risk was estimated for both the short and the long trading positions and at various confidence levels. We suggest using normal, Student and skewed Student distributions for divers GARCH-type processes. The GARCH-type VaR performance is assessed by estimating the failure rate of the Kupiec test statistic. Consistent with previous studies, our results show that energy price volatility exhibits a long-range memory. The VaR computed through a skewed Student-t FIGARCH process provides the best performance for both the short and the long trading positions.