Using elements of extreme value theory, I develop a Bayesian modeling approach that is capable of capturing the extremal dependence structures characterizing energy and agricultural prices. This approach is based on asymptotic arguments that hold for many underlying distributions of prices. Positive and negative movements of prices are considered separately which allows for asymmetry. Because the model is applied only to returns designated as extreme, inference does not depend on observations in the main body of the distribution. This is appealing because there is no reason to suspect a priori that the processes generating non--extreme and extreme observations are similar.