Statistics Seminar（2015-05）

Topic:Testing for Instrumental Variable Models under Weak Identification and Local Nonexogeneity

Speaker：Sung Yong Park, Chung-Ang University

Time：Thursday,16 April, 14:00-15:30

Location：Room 217, Guanghua Building 2

Abstract：This paper proposes an adjusted LM test that has better asymptotic size performance than the classical LM test, when the exogeneity conditions are invalid. In particular, the invalid instruments are specified to be locally nonexogenous to the structure errors. All the tests studied are based on generalized methods of moment (GMM). We show that when some of the instruments do not satisfy the usual moment conditions, LM statistics converge in distribution to a noncentral chi-sqaure with a noncentrality parameter depending on the degrees of violation of exogeneity conditions. In contrast, the asymptotic performance of the proposed LM statistics are well-behaved, regardless of the validity of exogeneity conditions. The results have been extended to two applications: a simple linear model and a dynamic panel data model. Also, Monte Carlo experiments are included.