The Mathematical Institute, University of Oxford, Eprints Archive

Time Series Modelling of Monthly WTI Crude Oil Returns

Lam, Derek (2013) Time Series Modelling of Monthly WTI Crude Oil Returns. Masters thesis, Oxford University.

[img]
Preview
PDF (Derek Lam Thesis) - Submitted Version
3MB

Abstract

This paper examines the dynamics of the monthly WTI crude oil return for the past two decades. The data are divided into two ten-year periods, and we explore with two approaches. We rst build univariate time series models using the Box-Jenkins methodology. Techniques such as stationarity tests and autocorrelation plots are used to determine the orders of the nal ARIMA model. GARCH and APARCH are also used to model residuals. Then, we build regression models based on eight explanatory variables. They are consumption, production, ending stock, net import, renery utilisation rate, U.S. interest rate, NYMEX oil futures contract 4 and S&P 500 index. Stepwise AIC method is employed to determine the optimal variables to be included. Multicollinearity is not evident in the reduced models. Residual analysis suggests that the assumptions of linear regression are not violated. Lastly, the forecasting powers of the models are compared. GARCH and APARCH perform the best in terms of forecasting accuracy, with APARCH performing the best in a turbulent market.

Keywords: Linear regression, ARIMA, GARCH, APARCH, time series forecasting, residual analysis

Item Type:Thesis (Masters)
Subjects:H - N > Mathematics education
Research Groups:Mathematical and Computational Finance Group
ID Code:1738
Deposited By: Laura Auger
Deposited On:13 Aug 2013 19:15
Last Modified:29 May 2015 19:26

Repository Staff Only: item control page