Title: Robustifying inference of DSGE models estimated by filtering methods
Authors: Martin M Andreasen - Aarhus University (Denmark) [presenting]
Abstract: Some challenges are discussed, which are related to estimating potentially nonlinear DSGE models by filtering methods. We also discuss simple ways to detect model misspecification. To make the existing (quasi) likelihood-based estimators more robust, we augment the (quasi) likelihood function by a set of unconditional GMM moment conditions. For a simulation study, we consider a standard New Keynesian model tailored to match the yield curve in the US. The simulation study shows that this penalized (quasi) likelihood function approach delivers more robust estimates when the model is misspecified than the standard (quasi) likelihood approach.