Linear and Non-Linear Modelling of Nigerian Crude Oil Prices

Abstract

To model Nigeria crude oil prices, this analysis compared univariate linear models to univariate nonlinear models. The data for this analysis was gathered from the Central Bank of Nigeria (CBN) Monthly Statistical Bulletin. The upward and downward movement in the series revealed by the time plot suggests that the series exhibit a regime-switching pattern: the cycle of expansion and contraction. At lag one, the Augmented Dickey-Fuller test was used to test for stationarity. For univariate linear ARIMA (p, d, q)) and univariate non-linear MS-AR, seven models were estimated for the linear model and two for the non-linear model. The best model was chosen based on the criterion of least information criterion,  AIC (2.006612), SC (2.156581), and the maximum log-likelihood of   (-150.5480) for the crude oil prices were used to pick MS-AR (1) for the series. In analysing crude oil prices data, the MS-AR model proposed by Hamilton outperforms the linear autoregressive models proposed by Box- Jenkins. The model was used to predict the series' values over a one-year cycle (12 months).