There is a large literature on the accuracy and market impact of USDA crop forecasts for the United States—by one recent count, 77 studies on usda forecast accuracy were published between 1990 and 2023 (Isengildena Massa, Karali, and Irwin, 2024). What is striking is how little analysis has been done of USDA forecasts for other important crop producers around the world. This gap matters most for the new South American grain giants of Brazil and Argentina. Consider the numbers. Brazilian soybean production exploded from 32.5 million metric tons in 1997 to 180 million in 2025—a 5.5-fold increase that dwarfs the 1.6-fold growth in the U.S. over the same period. Brazilian corn production grew 4.4 times, and Argentina posted 2.5-fold growth in soybeans and 2.7-fold in corn.
Brazil and Argentina have transformed into grain-exporting powerhouses over the last three decades, which makes the accuracy of USDA’s production forecasts for these two countries more important than ever. In a new article in Food Policy, “The New Grain Giants: The Reliability of USDA’s Corn and Soybean Production Forecasts for Brazil and Argentina,” my co-authors Olga Isengildina Massa (Virginia Tech) and Berna Karali (University of Georgia) and I provide the first comprehensive evaluation of these forecasts, using WASDE data spanning 1997–2025. Our thanks to Elsevier for making the article open access.
The key findings include:
📊 Early forecasts have limited informational value. Forecasts released before or during South American planting are not directionally consistent with final outcomes and offer no statistical advantage over a naïve forecast of last year’s production. Not surprising when you remember these first forecasts are released in May—five months before planting even begins in the Southern Hemisphere.
📊 There is a critical inflection point around January–February of the crop year (months 8–10 of the forecasting cycle). From that point forward, USDA forecasts become directionally consistent, beat the naïve alternative, and errors decline dramatically after harvest completion around month 14.
📊 We find systematic underestimation bias in Brazilian soybean and corn production estimates—likely a byproduct of USDA not fully keeping pace with the breathtaking expansion of Brazilian area and yields.
📊 Forecast revisions are positively correlated—the familiar “smoothing” pattern long documented in USDA’s U.S. forecasts—indicating new information is incorporated too slowly
Why does this matter? These forecasts don’t live in isolation. Errors in South American production estimates flow directly into the world balance sheets, and we show they spill over into U.S. export and ending stocks estimates. They also feed USDA baseline projections used for federal budget outlays and farm program payment estimates. When early projections are used for these purposes, the potential for resource misallocation is real.
Laurence J. Norton Chair of Agricultural Marketing
University of Illinois at Urbana-Champaign