BAYESIAN METHODS FOR DYNAMIC MULTIVARIATE MODELS

If dynamic multivariate models are to be used to guide decision-making , it is important that probability assessments of forecasts or policy projections be provided. When identified Bayesian vector autoregressio n (VAR) models are presented with error bands in the existing literatu re, both conceptual and numerical problems have not been dealt with in an internally consistent way. In this paper we develop methods to int roduce prior information in both reduced form and structural VAR model s without introducing substantial new computational burdens. Our appro ach makes it feasible to use a single, large dynamic framework (for ex ample, 20-variable models) for tasks of policy projections.