A1038
Title: Detecting many weak instruments
Authors: Zhenhong Huang - The University of Hong Kong (Hong Kong) [presenting]
Chen Wang - University of Hong Kong (Hong Kong)
Jianfeng Yao - The Chinese University of Hong Kong-Shenzhen (China)
Abstract: A new specification test is developed for the instrument weakness when the number of instruments and sample size goes to infinity proportionally. We proposed the test based on the fact that the asymptotic difference between the two-stage least squares (2SLS) estimator and OLS estimator disappears under the many weak instruments asymptotics, but converges to a non-zero limit under the alternative asymptotics. We establish the limiting distribution of the difference within two specifications and introduce a delete-d Jackknife procedure to consistently estimate the asymptotic variance/covariance. Monte Carlo experiments demonstrate the performance of the test procedure for both single and multiple endogenous variables. Additionally, we reexamine an analysis of returns to education by using our proposed test. Both the simulation results and empirical analysis indicate the reliability of the test.
