Probabilistic Forecasting of Output Growth, Inflation and the Balance of Trade in a GVAR Framework

Abstract

We apply a global vector autoregressive (GVAR) model to the analysis of inflation, output growth and global imbalances among a group of 33 countries (26 regions). We account for structural instability by use of country-specific intercept shifts, the timings of which are identified taking into account both statistical evidence and our knowledge of historic economic conditions and events. Using this model, we compute both central forecasts and scenario-based probabilistic forecasts for a range of events of interest, including the sign and trajectory of the balance of trade, the achievement of a short-term inflation target, and the incidence of recession and slow growth. The forecasting performance of the GVAR model in relation to the ongoing financial crisis is quite remarkable. It correctly identifies a pronounced and widespread economic contraction accompanied by a marked shift in the net trade balance of the Eurozone and Japan. Moreover, this promising out-of-sample forecasting performance is substantiated by a raft of statistical tests which indicate that the predictive accuracy of the GVAR model is broadly comparable to that of standard benchmark models over short horizons and superior over longer horizons. Hence we conclude that GVAR models may be a useful forecasting tool for institutions operating at both the national and supranational levels.

Keywords: GVAR modelling, structural breaks, probability event forecasting and evaluation, global imbalances

JEL Classification: C32, C53, E17