Forecasting Crop Prices using Leading Economic Indicators and Bayesian Model Selection

Corn, wheat and soybeans are very important to the US agricultural sector as the main sources of many farmers’ income. Thus, forecasting the prices of these three crops is important. When considering model specification of crop price forecasting models, this paper focuses on potential benefits from including leading economic indicators variables, both those clearly related to agriculture such as the crude oil price and interest rate and those not clearly related such as the purchasing managers index or the S&P500 stock price index. To do this, our paper tests whether leading economic indicators can be used to improve the forecasts of corn, wheat, and soybean future prices. We take a Bayesian approach to estimate the probability that a set of leading indicators belong in the forecasting model where specification uncertainty is explicitly modeled by assuming a prior distribution over a very large set of models. Model specifications considered vary by different lag lengths for leading indicators and crop prices as well as which variables are included at all. We apply this method to corn, soybean and wheat monthly spot price data from 1985 to 2016. The results show that several leading economic indicators appear to be useful for forecasting crop prices.