Title: Forecasting oil price volatilities with multivariate fractionally integrated asymmetric DCC models
Authors: Malvina Marchese - Cass Business School (United Kingdom) [presenting]
Lorenzo Trapani - Cass Business School (United Kingdom)
Abstract: The evidences of asymmetries and long-range dependence in the volatilities of oil prices is reassessed using a multivariate fractionally integrated exponential DCC model for three markets- Brent, Dubai, and West Texas Intermediate. We estimate several MGARCH models, compare their in-sample performance and their predictive ability with three approaches: the Superior Predictive Abilty test, the Model Confidence Set (MCS) method and the Value-at-Risk approach. In doing so, we extend the MCS method to include cases where the forecast error loss differential is strongly autocorrelated as arising in time series with long memory. In the overall, our results indicate significant gains from using models that include long-range dependence and asymmetries against short-memory models.