Three Approaches to Time Series Forecasting of Petroleum Demand in OECD Countries

Abstract

Petroleum (crude oil) is one of the most important resources of energy and its demand and consumption is growing while it is a non-renewable energy resource. Hence forecasting of its demand is necessary to plan appropriate strategies for managing future requirements. In this paper, three types of time series methods including univariate Seasonal ARIMA, Winters forecasting and Transfer Function-noise (TF) models are used to forecast the petroleum demand in OECD countries. To do this, we use the demand data from January 2001 to September 2010 and hold out data from October 2009 to September 2010 to test the sufficiency of the forecasts. For the TF model, OECD petroleum demand is modeled as a function of their GDP. We compare the root mean square error (RMSE) of the fitted models and check what percentage of the testing data is covered by the confidence intervals (C.I.). Accordingly we conclude that Transfer Function model demonstrates a better forecasting performance.