Long-memory and volatility spillovers across petroleum futures

Abstract

Return volatility usually presents a degree of persistence, which, although still consistent with an essential stationary process, cannot be adequately captured by standard autoregressive specifications. In this study, we examine the transmission mechanism across petroleum volatilities accounting for this stylized fact. To do so, we compute both time and frequency domain measures of connectedness based on variance decompositions from a fractionally integrated VAR (FIVAR). Our main findings summarize: (1) there is strong evidence of long-memory in petroleum volatilities, but no evidence of infinite variances; (2) an adequate quantification of spillovers must consider long-memory persistence explicitly; (3) spillovers are relatively high, but considerably lower than predicted in the traditional short-memory framework, especially at low frequencies. Thus, accounting for long-memory has asymmetric implications for different market participants.