Amir Hossien Khalilzadeh - MSc in Pure Statistics, Petroleum Ministry Fifth Building, No. ۱۷۵, Taleghani Ave., Tehran, Iran. Corresponding address
Carol Dahl - Faculty Member of Golden School of Mines, Division of Economic and Business, Colorado University, Dir CSM/IFP Petroleum Economics and Management, USA.
Hojatollah Ghanimi fard - Faculty Member of Petroleum University of Technology, Head of N.I.O.C. International Affairs, Tehran, Iran.

In this paper we try to model oil price volatility using the well-known ARCH/GA models. We focus on volatility of the spot prices of WTI crude oil, over the period 2 Jan 1986 to 27 July 2005 representing 5386 observations. The preferred model is a ARMA GARCH (1,1) model with student’s t-errors distribution. Special emphasis is given o problem of volatility prediction and the issue of a proper measure for the quality of predi In order to compare the different predictors of squared returns, we will use two po performance measures: Mean Squared Error (MSE) of prediction and Mean Ab Deviation (MAD) of prediction. So an optimal predictor is formulated, and the usefuln the new predictor is demonstrated on a real dataset.

ARCH models, oil price volatility, volatility prediction, mean absolute error, squared error