What VAR Tell us about DSGE Models?

Summary

We examine the consequences of extracting structural shocks in VAR models using standard standard inertial restrictions, when the data has been generated by two stochastic dynamic general equilibrium (DSGE) models featuring different types of microfundations and different sources of sluggishness. We find that, in general, misspecification is substantial: short run coefficients often have wrong signs; impulse responses and variance decompositions give misleading representations of the dynamics; inexistent puzzles are created. We show that an omitted variables bias accounts for the results and propose an alternative identification technique which can cope with the inherent underidentification displayed by the DSGE models currently used in macroeconomics.

We would like to thank Harald Uhlig, Lucrezia Reichlin, Jerome D’Adda, Morten Ravn, Vincenzo Quadrini, John Faust, David Bowman and the participants of seminars at a number of universities, conferences and summer schools for comments and suggestions. Canova acknowledges the financial support of DGYCIT grants; Pina acknowledges financial support from Sub-Programa Ciência e Tecnologia do 2 ^o Quadro Comunitário de Apoio. An earlier version of this paper has circulated with the title “Monetary policy misspecification in VAR models”.