Validating Monetary DSGE Models through VARs

Robust sign restrictions derived from calibrated DSGE models are used to identify structual shocks in the actual data. The dynamic behaviour of selected variables in response to these shocks is employed to measure, both qualitatively and quantitatively, the economic discrepancy between the model and the data. We design an algorithm that allows increasingly demanding diagnostics on the model, room for respecification at each stage of the process and comparison across models. We show that neither a limited participation model, nor a sticky price monopolistic-competitive model, fully accounts for the dynamics of a small set of macro variables. Furthermore simple alterations of the former fail to improve the match with the data, even in qualitative sense.