ECONOMETRICS: Coordinated Testing for Identification Failure and Correct Model Specification; Professor Bertille Antoine (Simon Fraser University)

Abstract

In the context of GMM or Minimum Distance (MD) inference, we develop a specification test that is: (i) robust to weak identification; (ii) easy to implement and broadly applicable; and (iii) powerful. We first show that it is still possible to rely on the J-test of overidentification, but only after adjusting its critical values and only when considering its continuously-updated form. Second, to address the lack of power of such a specification test, we resort to a conditional inference approach, where we use information about the existence of a number of strongly identified directions in the parameter space to de ne data-dependent critical values. The implementation of our suggested coordinated testing approach is based on developing a powerful test of weak identification that is compatible with mixed identification strengths - that is, subvector and/or directions in the parameters space that display different strength of identification (e.g. weak and non-weak). We illustrate the performance of our different testing strategies through three applications: Discrete Choice models with simultaneity and the New Keynesian Philips curve (in a GMM framework), as well as Asset Pricing models with stochastic volatility and leverage (in a MD framework).