Estimating oil price value at risk using belief functions

Abstract

We consider extreme value theory to study extreme price movements in crude oil market. Autoregressive-Moving-Average models are developed to describe daily log return of crude oil price. Peak-over-thresholdmodels are then used to model the log return forecasting errors (residuals). The maximum residuals are expressed in terms of value-at-risk or return level corresponding to accepted levels of risk so that appropriate risk measures can be taken. A likelihood-based belief function is constructed to quantify estimation uncertainty. As a result, we can assess the plausibility of various assertions about the value-at-risk of the idiosyncratic shocks in the world crude oil market.