Using Local Information to Improve Short-run Corn Cash Price Forecasts

Using daily prices from 134 corn cash markets from seven Midwestern states, this study examines the increase in short-run cash price forecasting accuracy provided by augmenting futures prices with recent observations from other cash markets. We utilize a Granger-causality-based criterion to determine the structure of the augmented models, i.e. how far to look for potentially relevant forecast information. For about 65% of the markets, the model consisting of prices of the futures market, the specific cash market, and its nearby cash markets (M2) forecasts better than the one only incorporating prices of the futures market and the specific cash market (M1) for five-, ten-, and thirty-day ahead forecasts based on root mean squared error (RMSE). For short-run forecasts, RMSEs tend not to be significantly different for most of the cash markets investigated, suggesting that the forecast accuracy improvement from including nearby cash markets is only moderate. However, the expanded model (M2) tends to significantly outperform the bivariate model (M1) more often as the forecast horizon increases.