Forecast evaluation of AveAve forecasts in the global VAR context

Abstract

The article “Forecasting Economic and Financial Variables with Global VARs” is on the frontier of current model building based on vector autoregressions. The constructed model links country/region specific models into a unified framework. This global VAR (GVAR) consists of 26 regions and estimates equations for 134 variables. The models for most regions consist of both real and financial endogenous variables. It is thus possible to explore the dynamics and interrelationships of both the real and the financial sectors. The authors explain the issues involved in building a model, evaluating the model and doing the final testing. This paper builds on research by Pesaran, Schuermann, and Weiner (2004) and Dees, di Mauro, Pesaran and Smith (2007) by exploring the forecasting ability of the GVAR. The paper ties together a large number of diverse literatures. Consider the topics that are covered: VARs, model averaging and combinations, time series benchmarks, cointegration, structural breaks, estimation periods, etc. The model itself takes into account co-trending restrictions, cross country cointegration and trade relations, etc. The paper also develops a panel version of the Diebold-Mariano predictive test statistic. The breath of the topics covered makes