USDA generates monthly season‐average price forecasts for key agricultural commodities. Uncertainty about each forecast is indicated by its publication as a price interval. USDA’s forecasting methodology is non‐public, but its uncertainty levels are anecdotally based on historical patterns of price uncertainty and informed by expert opinion. No confidence level is attached to USDA’s intervals, so it is difficult to gauge their accuracy. But in practice, realized season‐average prices regularly fall outside of USDA‐forecasted intervals, particularly those made prior to harvest and late in the marketing year. We demonstrate that forward‐looking density forecasts for the season‐average corn price can be constructed based on the market’s expectation of volatility implied by commodity options premia, combined with historical forecast errors between futures market prices and cash prices paid to farmers. Because implied volatility is forward‐looking, confidence intervals based on these densities reflect anticipatory market sentiment not present in historical data. In out‐of‐sample trials, our 95% confidence intervals contained the final season‐average price for over 92% of the 358 forecasts made between 1995/96 and 2014/15. Compared to a model based on historical data alone, the forward‐looking model is less susceptible to forecast errors. Our approach can enhance the informational value of USDA season‐average price forecasts.


Downloads Statistics

Download Full History