Working Paper: CEPR ID: DP380
Authors: M. J. Artis; W. Zhang
Abstract: This paper provides forecasts derived from Bayesian vector autoregressive (BVAR) models for the output growth, inflation and balance of payments of the G-5 and G-7 countries. These forecasts are compared with those derived from alternative time series models and with those provided by the International Monetary Fund in its World Economic Outlook (WEO) over the period 1980-7, as well as with the out-turns. The importance of setting up a proper prior and the sensitivity of forecast performance to information are two issues addressed in the paper. The results suggest significant gains from the use of BVAR models; as a minimum, these models provide a highly effective standard of comparison for forecasts derived in more traditional ways.
Keywords: forecast accuracy; BVAR; world economy
JEL Codes: 132
Edges that are evidenced by causal inference methods are in orange, and the rest are in light blue.
| Cause | Effect |
|---|---|
| BVAR models (C32) | forecast accuracy (C53) |
| BVAR models (C32) | Theil's inequality index (D31) |
| BVAR models (C32) | improved out-of-sample forecasts (C53) |
| BVAR models (C32) | avoid overfitting (C52) |
| traditional time series models (C22) | forecast accuracy (C53) |
| IMF forecasts (F37) | forecast accuracy (C53) |
| univariate autoregressive (AR) models (C22) | forecast accuracy (C53) |
| unrestricted VAR models (C32) | forecast accuracy (C53) |