Forecasting GNP Components Using the Method of Box and Jenkins

Abstract

Economists and business forecasters have employed various time series techniques in shortrun economic forecasting. These techniques encompass a wide variety of models ranging from the univariate autoregressive and/or moving average scheme to the system of stochastic difference equations commonly called an econometric model. The evidence points out that the simple time series models can often outperform the large econometric models in short-term forecasting. Stekler compared the forecasting accuracy of six econometric models built by various institutions and well known economists with that of two naive models of the forms Xt = Xt_1 and Xt Xt_, = Xt_1 Xt_. The author asserted that in general these econometric models did not prove to be relatively successful in forecasting economic activity [13]. In a recent article, Naylor, Seaks, and Wichern compare the forecasting accuracy of the Box-Jenkins models with the Wharton model for four variables: GNP, investment, unemployment and price. The authors concluded that the Wharton model did not fare as well as the Box-Jenkins models in forecasting these variables in the shortrun [10]. A variety of univariate time series models for nominal GNP, its seven components, the unemployment rate, two price indices and three interest rates has been investigated by