BVAR models and forecasting: a quarterly model for the EMU-11

Abstract

This paper deals with the costruction and evaluation of a quarterly forecasting BVAR model for the EMU-11 countries treated as a single country. In the current stage of EMU completion, most variables are affected by turbulences, and many macroeconomic relationships are characterised by structural instability. For this reason , the forecasting models used in this paper are time varying BVAR models. There are still signs that the models we have estimated are affected by some limitations, in spite of their good forecasting properties. In the light of this, in the second part of this paper we present an innovative approach in wich we extend the BVAR time varying parameter methodology: the intensity of parameter variation is governed by a time varying variance covariance matrix of the state equation error terms. This is achieved by slightly increasing the dimensionality of the hyperparameter space. We show some preliminary, encouraging evidence on how this proposal works, based on simulated data and on a restricted version of the EMU-11 model.