How reliable is it?
Average error when the model forecast past years it had not seen (tests from 2010 to 2019).
What drives the forecast
Year by year
Method. Six statistical models are fitted to each series (1990–2024, log scale): naive, drift, damped trend (Holt), Theta, recent linear trend and a driver model linking growth to population and GDP. Each model forecasts 2010–2019 from the data before, and the three most accurate are combined, weighted by their accuracy. The 80% range comes from those past errors. Events are applied on top of the statistical forecast. This is a statistical projection, not an economic market model such as the OECD-FAO Aglink-Cosimo model shown for comparison.
Sources. History and Outlook: OECD/FAO Agricultural Outlook 2026–2035 (wheat, fish). Population: FAOSTAT, UN World Population Prospects 2024. GDP: IMF World Economic Outlook, real GDP growth.