Files

Abstract

In-sample and out-of-sample Granger causality tests are applied to determine whether the real trade-weighted agricultural exchange rate helps predict monthly real prices and export sales of wheat, corn, and soybeans. An ARIMA model, alternative univariate and bivariate autoregressive models, and a restricted bivariate autoregressive model based on Hsiao's procedure are specified for each variable. Results of the causality are shown to be sensitive to specification choice. Forecasting performance of the models is compared and out-of-sample Granger causality is determined using univariate and bivariate models with the best (lowest mean-square forecast error) forecasting accuracy. These tests provide evidence supportive of Granger causality from the exchange rate to export sales, while the evidence on causality from the exchange rate to prices is mixed.

Details

Downloads Statistics

from
to
Download Full History