Measuring Extreme Price Risks by Different Statistical Methods: An In-Depth Case Study in the Crude Oil Market

Abstract

Crude oil prices are particularly volatile. Managing such price risks is vital for participants in financial markets, in particular the oil market. In the perspective of a long position, we conduct an in-depth study of popular existing statistical approaches as well as a recently developed method to estimate Value at Risk of the next day's oil price — a measurement of potential extreme price risks. We then validate the estimations via tests of accuracy, independence, and a combination of both criteria. The approaches that capture heteroscedasticity in the data, namely conditional Extreme Value Theory and Filtered Historical Simulation, perform considerably better than the pure bootstrapping method — Historical Simulation — and the (sub)asymptotic-target approach — Average Conditional Exceedance Rate.