This paper investigates the out-of-sample value-at-risk (VaR) forecasts in gold markets by considering both oil volatilities and the flexible model construction. We used the combined BHK (Brenner, Harjes, and Kroner, 1996) and power GARCH (PGARCH) models to consider not only the effect of spot prices, but also the endogenized power term. The empirical results indicate that the PGARCH-HV model with its flexibility in power term for data transformation and the high volatility of crude oil is the best model for VaR forecasting. The findings have implications for investors, financial institutions, and futures exchanges.
