Forecasting Agricultural Prices Using a Bayesian Composite Approach

Abstract

Abstract Forecast users and market analysts need quality forecast information to improve their decision-making abilities. When more than one forecast is available, the analyst can improve forecast accuracy by using a composite forecast. One of several approaches to forming composite forecasts is a Bayesian approach using matrix beta priors. This paper explains the matrix beta approach and applies it to three individual forecasts of U.S. hog prices. The Bayesian composite forecast is evaluated relative to composites made from simple averages, restricted least squares, and an adaptive weighting technique.