The Modification Of System GMM Estimation In Dynamic Panel Data Models And Its Applications

This paper emphasis on finite-sample performances of the conventional system GMM estimation in dynamic panel data model where weak instruments is present. To modify the conventional system GMM estimator, we form new instruments set consisting of instruments in level and efficient instruments proposed by Hayakawa(2009) to reduce the probability of the presence of weak instruments, and then choose continuously-updating objective function to achieve better finite-sample properties.Firstly, we review a series of studies on weak instruments problem, conventional GMM estimation and corresponding issues in the context of panel data models. And we also propose our solutions to the issue, namely, we form the new instruments set by the combination of instruments in level and efficient instruments, and then use continuously updating objective function under the framework of system GMM estimation.In the subsequent two chapters, we introduce both modified system GMM estimator and conventional GMM estimators. We show in Monte Carlo simulations that our modified system GMM estimator work very well even in the possible presence of weak instruments than conventional system GMM estimators and has good large and finite sample properties in terms of both bias reduction and size correction of overidentification test.In chapter 4, we also use both modified GMM estimator and conventional system GMM estimator to conduct the empirical research on the basis of provincial dataset.Finally, we conclude our motivation and contributions and also discuss the drawbacks so as to derive further researches.