Abstract
The loss function is a mathematical representation of the costs experienced by a forecaster when observed outcomes differ from what was predicted. Prior studies suggest that USDA forecasts are not optimal based on an assumed mean-zero quadratic loss function. This study proposes an alternative view of forecast evaluation, which assumes all USDA forecasts are produced to minimize the forecasters’ costs, and searches for the dimensions of the loss function under which optimality holds. We illustrate the degree to which USDA loss functions vary across a series of WASDE price forecasts. A better understanding of USDA forecasters’ costs will benefit forecasters and forecast users.
Keywords
animal product prices, forecast optimality, loss function
Date of this Version
12-23-2025
Recommended Citation
Fiechter, Chad; Bora, Siddhartha S.; and Kuethe, Todd H., "The Loss Function of USDA Forecasters: Evidence from WASDE Animal Product Price Forecasts" (2025). Department of Agricultural Economics Faculty Publications. Paper 43.
https://docs.lib.purdue.edu/agedocs/43
Comments
This is the publisher PDF of Fiechter C, Bora SS, Kuethe TH. The Loss Function of USDA Forecasters: Evidence from WASDE Animal Product Price Forecasts. Journal of Agricultural and Applied Economics. 2026;58(2):243-260. Published CC-BY by Cambridge University Press, the version of record and ADA Title II compliant version is available in HTML at DOI: 10.1017/aae.2025.10031.