Forecasting model for crude oil price with structural break

Abstract

Nowadays, in unstable economic environment, oil refining company is facing fluctuating crude oil price that causes unstable profit margin. Fluctuating crude oil price leads to difficulty in forecasting raw material procurement. Inaccurate forecast leads to inefficient decision making in optimizing refining company profit margin. In order to overcome an inaccurate in forecasting raw material procurement, an appropriate study of forecasting model is needed. Thus the objective of this study is to model fluctuating crude oil price based on geometric Brownian motion and mean reverting Ornstein-Uhlenbeck process and also to forecast fluctuating crude oil price with structural break. In modeling crude oil price, the information on whether the structural break exists is very crucial due to the long memory property might be camouflaged by the existence of the structural break. In this study, we employed long memory test to West Texas Intermediate (WTI) daily data from 2nd January 1986 to 31st August 2016 using log periodogram regression of Geweke and Porter-Hudak (1983). Bai and Perron test was applied to find break date. The result indicates that crude oil price is characterized by structural breaks. With the assumption that future price is affected by today’s price, we modeled and forecasted crude oil price using geometric Brownian motion and mean reverting Ornstein-Uhlenbeck process for 14 days, 30 days and 6 months. Results showed that forecasting crude oil price is highly accurate for short term with geometric Brownian motion compared to mean reverting Ornstein-Uhlenbeck process.