摘要：This study suggests that confidence intervals for WASDE forecasts of corn, soybean, and wheat prices may be improved if they are estimated using an empirical approach. Empirical confidence intervals are calculated following Williams and Goodman's (1971) method and use historical forecast errors to estimate forecast error distributions which is then used to predict confidence limits for future forecast errors. Three procedures for empirical distribution estimation are compared: 1) histogram, 2) changing distribution, 3) fixed distribution. The results suggest that the fixed distribution approach using logistic distribution provided accurate confidence intervals for WASDE corn, soybean, and wheat price forecasts.
