Using Global VAR Models for Scenario-Based Forecasting and Policy Analysis
This chapter demonstrates the usefulness of the GVAR modelling framework as a tool for scenario-based forecasting and counterfactual analysis. Working with the GVAR model developed by Greenwood-Nimmo, Nguyen and Shin (2010, J. Appl. Econometrics), we first show how probabilistic forecasting can be applied to the analysis of global imbalances. Probabilistic forecasting involves evaluating the conditional probability that a given event or combination of events will occur over a defined horizon by means of model-based simulations. To illustrate the usefulness of this approach, we develop a simple four-way probabilistic classi ficatory system built around the notions of balancing improvement in the trade balance, unbalancing improvement, balancing deterioration and unbalancing deterioration. We then extend a similar approach in a counterfactual context by constructing a range of scenarios as linear combinations of generalised impulse responses. We conclude that GVAR models are particularly well-suited to scenario-based analyses such as ours as they have the potential to analyse singularly rich datasets that allow the modeller to construct a wide range of policy-relevant scenarios