Optimal policy under model uncertainty: A structural-Bayesian estimation approach

In this paper we assess the relevant sources of model uncertainty for the optimal conduct of monetary policy within (parameter uncertainty) and across (specification uncertainty) a set of nested models by applying Bayesian model estimation techniques to EU 13 data. While parameter uncertainty is found to play a minor role, optimal monetary policy is highly sensitive with respect to specification uncertainty. As our main result, we find that following policy prescriptions that are optimal in the model, which includes all features and frictions, does not guard against the risk of model uncertainty.