Modelling Nigeria Crude Oil Prices using ARIMA Time Series Models

Abstract

This paper identified the best ARIMA time series model for monthly crude oil price in Nigeria spanning from 2006 to 2020. At first, the stationary condition of the data series are observed by time plot, Autocorrelation Function (ACF) and Partial Autocorrelation Function (PACF) plots, and then confirmed using Kwiatkowski-Phillips-Schmidt-Shin (KPSS) and Phillips-Perron (PP) test statistic, which has been found that the crude oil price series is non-stationary. After taking first difference of logarithmic values of data series, the crude oil prices data become stationary. Box-Jenkins four-step iterative methodology comprising of model identification, model fitting, diagnostic and forecasting is also applied to the crude oil prices data. Two optimal time series models were selected namely; ARIMA (2,1,1) and ARIMA (3,1,1) based on the three information criteria AIC, BIC and HQC. Thus, based on the criteria of mean square error; root mean square error; mean absolute error; the ARIMA (3,1,1) model best fits the data with minimum values of predictive measures.