Market price forecasting for wheat and corn

Abstract

In the course of defining a price forecasting model applicable for wheat and corn in Hungary, first the application of stepwise regression was attempted, however there was a poor fit, and the parameters were not in line with the assumptions either. In addition, an extreme multiple collinearity was found, therefore the ARMA model was tried. Considering that the results for wheat and corn did not show a constant dispersion, and taking into account that in the case of the ARMA model there is a constant conditional dispersion in time, it was necessary to introduce the GARCH process analogous to a conditionally parameterised ARCH(∞) model. Based on the results, the GARCH(1,1) model could be defined. This model has a good fit and can be used to forecast the market price of wheat in Hungary. In view of the results it was possible to set up the GARCH(0,3) model for corn. This model has a good fit and can be used to forecast the market price of corn in Hungary. Based on the aforementioned, we can declare that it is possible to define a price forecasting model predicting the price movements of wheat and corn in Hungary by applying the GARCH model.