Abstract

We propose a method to evaluate cyclical models which does not require knowledge of the DGP and the exact specification of the aggregate decision rules. We derive robust restrictions in a class of models; use some to identify structural shocks in the data and others to evaluate the class or contrast sub-models. The approach has good properties, even in small samples and when the likelihood is misspecified. We showhow to sort out the relevance of a certain friction (the presence of rule-of-thumb consumers) in a standard class of models.