Tanzania's economy remains highly dependent on imported petroleum, with petrol price fluctuations significantly impacting inflation, transportation costs, and household welfare. The purpose of the study was to conduct a comparative analysis of ARIMA and Exponential Smoothing methods for short-run forecasting of petrol prices in Tanzania to identify the most accurate model for predicting future petroleum price trends. Using monthly petrol price data from 2005 to 2024 obtained from the Bank of Tanzania (BOT), we applied the Box-Jenkins methodology to compare the Autoregressive Integrated Moving Average (ARIMA) and Exponential Smoothing models. The ARIMA (1,1,4) with drift model was identified as optimal based on Akaike's Information Criterion (AIC), Bayesian Information Criterion (BIC), and maximum log-likelihood criteria. This model achieved superior forecasting accuracy with a Mean Absolute Percentage Error (MAPE) of 2.44%, compared to 2.58% for Exponential Smoothing. The model forecasts a steady monthly increase of 0.3-0.5% in petrol prices, projecting prices to reach 3,500.65 TZS/L by September 2026. While the model demonstrates strong predictive performance (Ljung-Box p-value = 0.265), its limitations in anticipating sudden price shocks highlight the need for complementary risk management strategies. These findings provide policymakers and market participants with a reliable tool for budget planning and economic forecasting. The study underscores Tanzania's vulnerability to global oil market volatility and emphasizes the importance of developing strategic fuel reserves and alternative energy sources to enhance energy security.
Ilembo et al. (Fri,) studied this question.