Forecasting the VaR of crude oil market: Do alternative distributions help?

Abstract

Accurate modeling of the empirical distribution of crude oil market returns is extremely important in estimating risk measures. In addition to several commonly used distributions, alternative distributions are explored in this study, some of which account for the asymmetry and heavy tails simultaneously found in distributions, and contain more tail parameters to separately depict the right and left tails when forecasting the Value-at-Risk (VaR) of crude oil markets during highly volatile periods. Seven backtests are also conducted to compare the VaR forecasting accuracy among different distributions. The empirical results indicate that a highly volatile environment challenges the commonly used distributions, and the four risk models based on commonly used distributions are rejected about 27% to 38% of the time. The alternative distributions, i.e., skewed general error distributions (SGED), generalized hyperbolic skewed Student-t distributions (GHST), and generalized asymmetric Student-t (GAST) distributions, generally produce more accurate VaR measurement, and GAST gives the best measurement accuracy. The empirical results imply that risk managers or policymakers should further consider more flexible distributions, such as SGED, GHST, or GAST in particular, when quantifying or managing the risk in turbulent market times.