Two-Step Conditional Alpha-Quantile Estimation via Additive Models of Location and Scale

We propose an estimation procedure for value-at-risk (VaR) for conditional distributions of a time series of returns on commodity prices. Our approach combines an additive B-spline/back tting estimator of conditional mean and volatility functions in a conditional heteroscedastic autoregressive nonlinear model with Extreme Value Theory for estimating quantiles of the conditional distribution. We apply our methodology to estimate value-at-risk associated with return series for various commodities.