Forecasting U.S. Shelled Pecan Prices: A Cointegration Approach

The relationship between shelled pecan prices and inventories is examined drawing using Engle-Granger and Johansen approaches to cointegration. Forecasts are made using the univariate ARIMA and cointegration methods. Monthly data are used (January 1992 to December 2004). Results suggest that both shelled pecan inventories and shelled pecan prices are nonstationary. Both Engle-Granger and Johansen cointegration tests found evidence of long run relationship between pecan prices and inventories. The ARIMA model outperformed the coointegration techniques at shorter forecast periods. The Cointegration methods, however, outperformed the ARIMA model with the Engle-Granger cointegration having the smallest MAPE statistic.