A Comparative Study Between the Linear Regression Model and ARIMA Time Series Models in Forecasting the Monthly Rates of Crude Oil Production: A Case Study of the Libyan State (2021–2027)

Abstract

This study aims to conduct a methodological comparison between the efficiency of the simple linear regression model  and  time series models (ARIMA)  in forecasting the monthly production rates of crude oil in Libya. The importance of this research stems from the strategic role of the oil sector as a primary pillar of the national economy. The research problem arises from the sharp fluctuations observed in the time series of oil production , largely driven by political and economic crises , which have reduced the accuracy of conventional forecasting approaches.

The analysis is based on a time series consisting of  (60) monthly observations  covering the period (2021–2025) . The statistical results indicate that the simple linear regression model was not appropriate , as the model lacked statistical significance and the series exhibited instability due to the presence of autocorrelation in the residuals. In contrast , the ARIMA(0,1,2)  model demonstrated superior forecasting performance after successfully passing several diagnostic tests, including  (t-tests ) for parameter significance, information criteria (AIC and BIC) , and tests confirming that the residuals follow  white noise , indicating the absence of autocorrelation.

Based on these findings, the study provides  forecasts for the years (2026-2027), which suggest a relatively upward trend in crude oil production . The study therefore recommends adopting advanced time-series models in the formulation of Libya’s financial and economic policies in order to reduce the risks associated with random estimation and to support more effective strategic planning in the energy sector