Many systems used to predict cotton yields and production use huge amounts of climactic and other data, ranging from daily rainfall, to growing degree days and hours available for field work. The systems require a high degree of maintenance and sophistication. At the same time, USDA-NASS reports aggregate and clearly subjective information on crop conditions by state that can be used to produce a parsimonious and efficient estimate for the current yield and productivity in the field. While the USDA data is clearly a subjective measurement of the condition of the crop, those collecting it generally have years of experience with the crop and have experienced both good and bad years. The data is quite timely, with weekly updates prior to and in-between the objective USDA reports.

Using the condition data to predict yields is not new and has been applied extensively to predicting harvested yields and production, primarily for corn and soybeans. In a crop such as cotton, where state abandonment rates can vary from 1 to more than 40 percent, estimating harvested area yields directly is problematic. Estimating planted area yields and a separate, non-linear harvesting rate equation produces equation parameters that are more intuitive. The result is a parsimonious pair of equations for each state that use reported crop conditions, planting progress, and annual trend variables to forecast yields, abandonment, and production in a system that is easy to maintain. The paper looks at the performance of such a cotton model, the behavior of crop condition reports throughout the growing season, and a comparison of performance with USDA reports of the same period.