Learning in an Estimated Medium-Scale DSGE Model

In this paper we evaluate the empirical relevance of learning by private agents in an estimated medium–scale DSGE model. We replace the standard rational expectation assumption in the Smets and Wouters (2007) model by a constant gain learning mechanism. If agents know the correct structure of the model and only learn about the parameters, both expectation mechanisms result in a similar fit, and only the transition dynamics that are generated by specific initial beliefs are responsible for the differences between the two approaches. If, in addition, agents use only a reduced information set in forming the perceived law of motion, the implied model dynamics change and for some initial beliefs the marginal likelihood of the model is further improved. The learning models with the highest posterior probabilities add some additional persistence to the DSGE model that reduce the gap between the IRFs of the DSGE model and the more data-driven DSGE-VAR model. However, the additional dynamics that are introduced by the learning process do not systematically alter the estimated structural parameters related to the nominal and real frictions in the DSGE model.