In the constant search for better models to help guide policy decisions, central banks have begun to use and develop dynamic stochastic general equilibrium (DSGE) models. Although such models were until recently considered theoretically sound but overly restrictive, newly developed methods have proved successful in specifying DSGE models that fit the macroeconomic data well.
Policy institutions that use DSGE models in policymaking need a reliable method for evaluating the models’ effectiveness. This article reviews a procedure recently proposed by the authors and their colleagues. The article first describes how the linear DSGE model can be nested in a vector autoregression (VAR) and then outlines a procedure that can systematically relax the restrictions the DSGE model imposes on the VAR.
Using the resulting DSGE-VAR specification as a framework, the authors compare the fit and forecasting performance of a DSGE model with several nominal and real rigidities. They use the DSGE-VAR framework to assess the relative importance of different frictions, with particular emphasis on wage and price indexation and habit formation. The DSGE-VAR framework also provides a benchmark that can reveal in what dimensions a DSGE model needs to be improved.
This DSGE-VAR procedure, the authors believe, shows some promise in delivering robust evaluations of DSGE models.