Identifying the sources of model misspecification

Abstract

In this paper we propose empirical methods for detecting and identifying misspecifications in DSGE models. We introduce wedges in a DSGE model and identify potential misspecification via forecast error variance decomposition (FEVD) and marginal likelihood analyses. Our simulation results based on a small-scale DSGE model demonstrate that our method can correctly identify the source of misspecification. Our empirical results show that the medium-scale New Keynesian DSGE model that incorporates features in the recent empirical macro literature is still very much misspecified; our analysis highlights that the asset and labor markets may be the source of the misspecification.