A Bayesian Vector Autoregresive Model of the U.S. Dairy Industry

This work develops a structural Bayesian Vector Autoregressive price forecasting model of the U.S. dairy industry based on monthly price, production, and inventory data. It also provides a relatively simple and clear understanding of the quantitative relationships between the prices of milk, cheese, butter, non-fat dry milk, whey, and dry buttermilk. The Bayesian feature allows for more efficient use of prior information, improves handling of seasonality, and solves the degree-of-freedom problem inherent in vector autoregressions. As current production and inventory data affect future prices with a lag, the autoregressive model is especially suitable for short-term price forecasting by dairy producers, processors, and wholesale distributors. Impulse response functions isolate the effects of various shocks on dairy product prices, while error bands indicate forecasting precision. Forecasting errors are found acceptable for practical business purposes.