Detecting macroeconomic impacts on agricultural prices and export sales :a time series forecasting approach
Bradshaw, Girard W
MetadataShow full item record
The effect of movements in the real exchange rate on agricultural prices and agricultural export sales is assessed based on the principle of Granger causality. An out-of-sample forecasting procedure is used to conduct tests for Granger causality from the exchange rate to agricultural prices and export sales. Technical time series issues such as stationarity, the method of lag-length selection, in sample versus out-of-sample tests for Granger causality, and long-range versus short-range forecasting are considered in relation to the outcome of Granger causality tests. Theoretical and empirical studies are reviewed which indicate the importance of working with stationary data series when testing for Granger causality. Differing methods of lag-length selection are found to affect the outcome of both in-sample and out-or-sample tests for Granger causality. The usual in-sample tests for Granger causality are compared to out-of-sample tests; the results of the comparison reveal that the in-sample tests do not in-general agree among themselves, nor do they agree with the out-of-sample tests' results. This indicates the importance of searching the model space for the best specification before conducting Granger causality tests. Long-range forecasts are compared to the I-step ahead forecasts used to test for Granger causality; these forecasts corroborate the out-of- sample tests for Granger causality in finding significant impacts from the exchange rate to agricultural export sales and agricultural prices.
- Masters Theses