Utilizing the ARIMA Model to Forecast Rice Prices in Bangladesh

Abstract

Aims: This study aimed to forecast rice prices in Bangladesh by identifying the most suitable Autoregressive Integrated Moving Average (ARIMA) model, thereby providing insights for policymakers, researchers, and farmers. Study Design:  Quantitative time series analysis using Box–Jenkins ARIMA methodology. Place and Duration of Study: The study utilized national-level secondary data from Bangladesh, spanning the period from 1991 to 2021. This period was selected because it represents the most consistent and reliable annual rice price dataset available from FAOSTAT, while also capturing major economic reforms, trade liberalization, and structural changes in the rice sector. Methodology: Annual wholesale rice price data (USD/metric ton) were obtained from FAOSTAT (2023). Stationarity of the series was tested using the Augmented Dickey–Fuller and Phillips–Perron tests. The Box–Jenkins ARIMA approach was employed, involving model identification, estimation, diagnostic checking, and forecasting. Model selection was based on the statistical significance of coefficients, correlogram analysis, and residual diagnostics. Forecasting was conducted using STATA 14 software. Results: The rice price series was found to be non-stationary but achieved stationarity after first differencing. The ARIMA (2,1,0) model was selected as the best-fit model. Residual diagnostics confirmed model adequacy. Forecasts for 2022–2031 indicated a generally upward trend in rice prices, with minor deviations between actual and predicted values. Conclusion: The ARIMA (2,1,0) model provides a reliable framework for medium-term rice price forecasting in Bangladesh. Given rice’s central role in national food security and household consumption, these forecasts are valuable for policy formulation on procurement, import planning, and farmer decision-making regarding crop allocation and storage. However, future studies could incorporate seasonal and external economic factors to improve accuracy.