The Use of Exponential Smoothing (ES), Holts and Winter (HW) and ARIMA Models in Oil Price Analysis

Abstract

This paper compares the performance accuracy of three types of univariate models in oil price prediction given so many complicated models exist. It may be just as easy to apply a simpler model with economy of costs and accuracy in prediction. We investigate the Exponential smoothing (ES), Holt-Winters (HW) and Autoregressive integrated moving average (ARIMA) models. Six strategies were used to determine selection prediction accuracies using data from the West Texas Intermediate (WTI) crude. The results show that the HW model performed better than the ES model (95%), while an ARIMA (2, 1, 2) model was most accurate of the three. The most sophisticated of the three was robust thus useful as a quick and economical model to use in the oil market.