Forecasting long-term grain supply of low- and middle-income countries

Abstract Accurate grain supply forecasting is important for addressing long-term food security. The Global Food Assessment (GFA) report, published annually by the USDA’s Economic Research Service, provides one-year and ten-year projections at the country level, focusing on grain demand, production, and the gap between demand and domestic production. Our study aims to evaluate and explore alternative long-term grain supply forecasting methods. Using a data set spanning 1980 to 2021 for 78 countries, we evaluate regression models with varying specifications (pooled vs. country-specific coefficients, linear vs. non-linear trends, and alternative predictors) and Autoregressive Integrated Moving Average with exogenous variables (ARIMAX) models with different ARIMA orders ( p, d, q ) and predictor variables. To assess model forecast accuracy, we employ a time-series out-of-sample validation approach. Our analysis reveals that ARIMAX models with exogenous variables, country-specific coefficients, linear trends, and weather variables significantly improve forecast accuracy. While the achieved Mean Absolute Error of approximately 11% indicates that there is room for further improvement, our approach represents a substantial advancement over existing methods. By combining these grain supply forecasts with demand projections, it is possible to provide valuable insights into future domestic grain supply and potential international trade dynamics.