Forecasting Milk Prices With VAR Models – Application to Farm Gate Price in Finland

Abstract

The goal of our study is to analyze the associations between Finnish and global milk markets and use the results to predict the Finnish milk price. We develop a theoretical framework of the factors affecting the milk price in Finland and test the predictive power within Vector Autoregressive (VAR) model. We employ the monthly farm gate price of milk, the quantity of milk produced, and other available monthly variables that depict the global milk markets from January 2007 to December 2016. We forecast their values for January 2017–August 2019 using only the observations from the estimation period. Thus, we make 1-step to 32-steps forecasts and measure the accuracy of forecasts with the MSPE. The greatest forecasting power for the milk price in Finland are the VAR models with combination of three or four variables: the lagged price of milk in Finland, the price of oil, the world feed price, and the quantity of milk produced. We also build forecasts where the observed series of oil future from January 2017 until August 2019 replace the forecasted prices of the oil. For the first 18 months, the oil future prices improve the forecasts, but then the both schemes forecast equally well.