USDA Livestock Price Forecasts: A Comprehensive Evaluation

Abstract

One-step-ahead forecasts of quarterly live cattle, live hog, and broiler prices are evaluated under two general approaches: accuracy-based measures and classification based measures which test the ability to categorize price movements directionally or within a forecasted range. Results suggest U.S. Department of Agriculture (USDA) price forecasts are not optimal. Broiler price forecasts are biased, and all the forecast series tend to repeat errors. While the USDA forecasts are more accurate than those of a univariate AR(4) time-series model, evidence suggests the USDA live cattle forecasts could be improved with a composite forecast that includes a time-series alternative. Despite this, the USDA correctly identifies the direction of price change in at least 70% of its forecasts over the sample period. Furthermore, actual prices fall within the USDA's forecasted range 48% of the time for broilers, but only 35% for hogs. Finally, there is some evidence that the USDA's price forecasting accuracy has improved over time for broilers, but has gotten marginally worse for hogs.

Keywords

• forecast efficiency
• forecast evaluation
• livestock prices
• usda forecasts