Testing, Estimation in GMM and CUE with Nearly-Weak Identification
分析了在强、近弱和弱识别混合系统中GMM和CUE估计量的极限分布，发现近弱工具下Wald、LR和LM检验具有标准卡方极限，对实证研究有重要价值。
In this article, we analyze Generalized Method of Moments (GMM) and Continuous Updating Estimator (CUE) with strong, nearly-weak, and weak identification. We show that with this mixed system, the limits of the estimators are nonstandard. In the subcase of GMM estimator with only nearly-weak instruments, the correlation between the instruments and the first order conditions decline at a slower rate than root T. We find an important difference between the nearly-weak case and the weak case. Inference with point estimates is possible with the Wald, likelihood ratio (LR), and Lagrange multiplier (LM) tests in GMM estimator with only nearly-weak instruments present in the system. The limit is the standard χ2 limit. This is important from an applied perspective, since tests on the weak case do depend on the true value and can only test simple null. We also show this in the more realistic case of mixed type of strong, weak, and nearly-weak instruments, Anderson and Rubin (1949) and Kleibergen (2005) type of tests are asymptotically pivotal and have χ2 limit.