Abhatoo - Centre Nationale de Documentation

This paper presents an estimated model with learning and provides evidence that learning can improve the .t of popular monetary DSGE models and endogenously generate realistic levels of persistence.

The paper starts with an agnostic view, developing a model that nests learning and some of the structural sources of persistence, such as habit formation in consumption and in.ation indexation, that are typically needed in monetary models with rational expectations to match the persistence of macroeconomic variables. I estimate the model by likelihood-based Bayesian methods, which allow the estimation of the learning gain coe¢ cient jointly with the .deep.parameters of the economy.

The empirical results show that when learning replaces rational expectations, the estimated degrees of habits and indexation drop near zero. This .nding suggests that persistence arises in the model economy mainly from expectations and learning. The posterior model probabilities show that the speci.cation with learning .ts signi.cantly better than does the speci.cation with rational expectations.

Finally, if learning rather than mechanical sources of persistence provides a more appropriate representation of the economy, the implied optimal policy will be di¤erent. The policymaker will also incur substantial costs from misspecifying private expectations formation.