We demonstrate the use of the small-sample econometrics principles and strategies to come up with reliable yield and acreage models for policy analyses. We focus on demonstrating the importance of proper representation of systematic and random components of the model for improving forecasting precision along with more reliable confidence intervals for the forecasts. A probability distribution function modeling approach, which has been shown to provide more reliable confidence intervals for the dependent variable forecasts than the standard models that assume error term normality, is used to estimate cotton supply response in the Southeastern United States.
