Model Equivalence Tests in a Parametric Framework

Résumé

In empirical research, one commonly aims to obtain evidence in favor of re- strictions on parameters, appearing as an economic hypothesis, a consequence of economic theory, or an econometric modeling assumption. I propose a new theoret- ical framework based on the Kullback-Leibler information to assess the approximate validity of multivariate restrictions in parametric models. I construct tests that are locally asymptotically maximin and locally asymptotically uniformly most powerful invariant. The tests are applied to three different empirical problems.