An analysis of price responses to public information: a case study of the USDA corn crop forecasts
Crop forecast errors were defined as the differences between the five-year revised final estimates of corn production and the monthly USDA corn crop forecasts. It was found that the monthly forecast errors were normally distributed according to the Shapiro-Wilk W Statistics and Probability; An F test, based on Hotelling's T('2) statistics showed that the means of the monthly USDA corn crop forecasts were not statistically different from the five-year revised final estimates at the 5% significance level. When a nonparametric L test was employed, the accuracy of the monthly USDA corn crop forecasts was noted to improve over the reporting months, from July to December. Regression analysis revealed that the best estimate of the final crop size was the most recent month's USDA corn crop forecast;The impact of the USDA corn crop forecasts on daily cash and futures corn prices were analyzed within the framework of the supply-of-storage theory. When the inter-temporal price spreads were assumed to be a function of the number of grain-consuming animal units and the difference of the two adjacent months' USDA corn crop forecasts, the August USDA corn crop forecast was found to be the only crop forecast to influence the cash and futures prices observed on the day immediately following the day of the crop announcement. Anticipated effect of the October forecast on cash prices was present during the three days prior to the announcement while that of the November forecast on cash prices was felt during the two days prior to the announcement. Anticipated effect of these forecasts, however, was not present in futures prices;When the past price movements were introduced into the model, the results found were not significantly different from those obtained under the earlier assumption, in that only the August forecast had an impact on the cash prices and on the September and December futures prices observed in August at the 1% significance level.