Measuring VaR of Oil Price Based on GARCH-type Models

The oil is the important strategic resources, which price changes can affect economic safety of a country. The return rate series of oil price are characterized by high peak, fat tail and asymmetry. Therefore, based on the assumption of gauss normal distribution and skewed Student distribution respectively, this paper comparatively studies Risk Metrics and GARCH-type models of 11's accuracy of calibrating VaR, and checks the one-step-ahead forecasting VaR by using failure rate test and dynamic quantile test. The results show that skewed student distribution is better for fitting the feature of oil price and the models of EGARCH, FIEGARCH, GJR-GARCH, FIGARCH(BBM) and HYGARCH are more exactly than others under different confidence level, which the degree of its overestimation or underestimate is receivable.