Semiparametric weak instrumental variables models for panel and cross-sectional data

In the literature on instrumental variables estimation some works have been done to weak instrumental variables of simultaneous equations cross-section data models. In this dissertation, we consider the weak instrumental variables panel data models and the functional coefficient weak instrumental variables cross-section models. We first construct the two stage least-squares dummy variables estimator (the instrumental variables within estimator) for the weak instrumental variables panel data models. Then, two-stage estimator (the first stage is to estimate a matrix of unknown parameters of the reduced form equation by using the least squares estimator, and the second stage is local linear regression using the estimated reduced forms as regressors) is proposed.; Consistency and asymptotic normality of the instrumental variables within estimator when both cross-section dimension and time-series dimension tend to infinity are given. And we investigate how the limiting distribution of the resulting estimator changes as we vary our parametrization to allow for the increasing degrees of weakness of instrumental variables. The finite sample properties of the instrumental variables within estimator for the weak instrumental variables panel data models and of the two-stage estimator for the functional coefficient weak instrumental variables cross-section models are examined by Monte Carlo simulations.