MODELLING OF OIL PRICE VOLATILITY USING ARIMA-GARCH MODELS F. Merabet, H. Zeghdoudi, R. H Yahia, and I. Saba

Abstract

In this paper, the behavior of the oil price series named OIL is examined. The non-stationarity on average and variance, with the non-normality of the OIL series distribution, indicate the volatility of the series. The study is based on a combination of the Box-Jenkins methodology with the GARCH processes (Engle and Bollerslev). The first part models the lnOIL series in which, by applying the first difference the series becomes DlnOIL. Then the Box-Jenkins methodology is applied. The choice of the model was made on basis of minimization of criterion -Akaike (AIC), Shwarz (SIC)- and maximization of log likelihood (LL). Of the four models identified, ARMA (3.1) is retained. According to the statistical indicators of the ARMA model (3,1), the nature of the residuals and other tests, it is shown that the series of squares of the residuals follows a conditionally heteroscedastic ARCH model. The second part is devoted to a symmetrical and asymmetrical GARCH modelling. The model used for predicting volatility is the EGARCH model (1,2). The data available relates to 3652 daily values of the change in OIL, from 01/01/2019 to 12/31/2019. The forecast is made for the first three months of 2020; the result concludes that the predicted values and the current values are very close, and that the model ARIMA (3,1,1) + EGARCH (1,2) is the best forecast model.