Forecast Butter Prices Using ARIMA and ARIMAX Models

This paper constructs a feasible ARIMA/ARIMAX model to forecast butter prices. The preferred model is an ARIMA (1,1,0) for the butter price series. We add exogenous control variables to reduce forecast bias, and multi-collinearity is not a problem for this forecasting objective. Based on the sample data, the mean absolute margin of error of the model is less than 4.5%, which is an acceptable bias for dairy processors seeking a reference for long-term production planning and for stakeholders developing investment plans.