BVAR FORECASTS FOR THE G-7

The paper provides forecasts, derived from Bayesian vector autoregressive models, for the output growth, inflation and balance of payments of the G-7 countries. In constructing the models, particular attention is paid to the determination of the prior and to the choice of lag length and vector content. The forecasts derived from the models are compared with those published by the International Monetary Fund on alternative assumptions about the information set available to the forecaster. The results indicate that BVAR methods can provide a highly effective standard of comparison for forecasts produced by more traditional methods. © 1990.