Detecting Lack of Identification in GMM
Jonathan H. Wright
Board of Governors of the Federal Reserve System - Trade and Financial Studies Section
FRB International Finance Discussion Paper No. 674
In the standard linear instrumental variables regression model, it must be assumed that the instruments are correlated with the endogenous variables in order to ensure the consistency and asymptotic normality of the usual instrumental variables estimator. Indeed, if the instruments are only slightly correlated with the endogenous variables, the conventional Gaussian asymptotic theory may still provide a very poor approximation to the finite sample distribution of the usual instrumental variables estimator. Because of the crucial role of this identification condition, it is common to test for instrument relevance by a first-stage F-test. Identification issues also arise in the generalized method of moments model, of which the linear instrumental variables model is a special case. But I know of no means, in the existing literature, of testing for identification in this model. This paper proposes a test of the null of underidentification in the generalized method of moments model.
Number of Pages in PDF File: 34
Keywords: Generalized Method of Moments, Identification, Asset Pricing, Instrumental Variables
JEL Classification: C10working papers series
Date posted: November 22, 2000
© 2013 Social Science Electronic Publishing, Inc. All Rights Reserved.
This page was processed by apollo7 in 0.344 seconds