A reappraisal of the forecasting performance of corn and soybean new crop futures

Forecasting performance of December corn and November soybean futures contracts during the previous spring was evaluated using the commonly specified price‐level and percent‐change models. These models invoke different assumptions regarding stationarity. Using Stein's analytical framework, results for the price‐level model suggest avoidable social loss existed in the soybean market since 1973, because November futures provided biased forecasts. Regression R-super-2s for both corn and soybeans declined substantially between 1952–1972 and 1973–1997, suggesting total social loss increased. By contrast, results from the percent‐change model suggest only unavoidable social loss existed in the corn and soybean markets, because the futures provided unbiased forecasts. R-super-2 increased for corn but declined for soybeans, suggesting unavoidable social loss declined for corn, but increased for soybeans. The important, conflicting nature of the results from the two models underscores the importance of examining alternative model specifications when evaluating price forecasting performance. © 1999 John Wiley & Sons, Inc. Jrl Fut Mark 19: 604–618, 1999