Bank for International Settlements publishes research showing extreme heat lowers Argentina soybean yields and early-season climate models can beat USDA forecasts

The Bank for International Settlements published a working paper that models and forecasts soybean yields in Argentina using high-frequency climate data, with an emphasis on quantifying how weather shocks and ENSO phases affect yields and how climate-based models can be used for forward-looking exercises. Using a novel delegation-level dataset covering 1980–2023, the paper estimates a fixed effects spatial error model that links yields to measures of extreme heat, precipitation and El Niño–Southern Oscillation indicators, alongside economic and technology controls such as transgenic seed adoption, the soybean-to-fertilizer price ratio, and a land-use change indicator. The results indicate that extreme heat materially reduces yields, precipitation supports yields up to a nonlinear threshold, El Niño conditions are associated with higher yields and La Niña conditions with lower yields, while technological adoption and favourable price signals also improve productivity; a statistically significant spatial error term points to regionally correlated unobservables. The authors then translate these relationships into national and region-specific forecasting models, including mixed-frequency specifications, and benchmark them against the US Department of Agriculture’s monthly yield forecasts using rolling-window exercises and Giacomini–White tests. Several early-information models using data available by December or January outperform the corresponding USDA forecasts and, in some comparisons, produce forecasts comparable to later USDA reports.