COWLES FOUNDATION FOR RESEARCH IN ECONOMICS
AT YALE UNIVERSITY

Box 208281
New Haven, CT 06520-8281

COWLES FOUNDATION DISCUSSION PAPER NO. 1103

Automated Forecasts of Asia-Pacific Economic Activity

Peter C. B. Phillips

June 1995

This paper reports quarterly ex ante forecasts of macroeconomic activity for the U.S.A., Japan and Australia for the period 1995-1997. The forecasts are based on automated time series models of vector autoregressions (VAR's), reduced rank regressions (RRR's), error correction models (ECM's) and Bayesian vector autoregressions (BVAR's). The models are automated by using an asymptotic predictive form of the model selection criterion PIC to determine autoregressive lag order, cointegrating rank and trend degree in the VAR's, RRR's, and ECM's. The same criterion is used to find optimal values of the hyperparameters in the BVAR's. The forecasts are graphed and tabulated. In the case of the U.S.A., the results are compared with forecasts from the Fair model, a structural econometric model of the U.S. economy.

Keywords: Automated time series model, Cointegration, Model selection, Nonstationarity, Posterior information criterion (PIC), PIC'ed model, Stochastic trend, Unit root

JEL Classification: C32