Research And Application Of Parameter Estimation Based On Dynamic Panel Data Model

This thesis focuses on researching the parameter estimation of the dynamic panel data mod-el.For the balanced dynamic panel data model,the improvement and optimization in some degree is made on the research and analysis of the traditional GMM(Generalized Method of Moments),and two improved GMM estimation methods(the Differenced Forward GMM Es-timation and the Horizontal Forward GMM Estimation)are proposed;for the parameter esti-mation problem of the block-dividing unbalanced dynamic panel data model,this thesis studies and analyzes it by two improved GMM estimation methods,and obtains accurate and effective simulation results through simulation experimentsThis thesis is divided into seven chaptersThe first chapter is the introduction part,which introduces the background,purpose,litera-ture review,and main research contentsThe second chapter mainly introduces basic forms of the dynamic panel data model,the basic idea of the generalized moment estimation,and the method of instrumental variables.Fur-thermore,it introduces the process of the parameter estimation on the panel autoregressive model as well as the dynamic panel data model through the traditional GMM estimation methodThe third chapter firstly introduces four traditional GMM estimation methods as follows:the Differenced GMM Estimation,the Horizontal GMM Estimation,the System GMM Estima-tion,and the Forward Orthogonal Deviation GMM estimation.Subsequently,it conducts certain improvements and optimizations on the traditional GMM estimation method,and integrates the instrumental variables effectively to generate more moment conditions and make use of more sample information.Accordingly,two new GMM estimation methods(the Differenced Forward GMM Estimation and the Horizontal Forward GMM Estimation)are proposed;finally,two im-proved GMM estimation methods are applied to concretely study and analyze the parameter estimation problems of the block-divided unbalanced dynamic panel data model respectivelyThe fourth chapter and the fifth chapter focus on two aspects.On the one hand,aiming at the parameter estimation of the panel autoregressive model as well as the balanced dynamic panel data model,this part applies the simulation experiment research to compare the simulation results of four GMM estimation methods(the Differenced GMM Estimation,the Forward Orthogonal Deviation GMM Estimation,the Differenced Forward GMM Estimation,and the Level Forward GMM Estimation),and conducts the comparative analysis on the obtained the estimator's mean,the standard deviation,and the root mean square error;on the other hand,aiming at the parameter estimation problem of two,three and n non-equilibrium dynamic panel data models,this part applies the improved Differenced Forward GMM Estimation and the Horizontal Forward GMM Estimation to study the parameter estimation.In addition,the simulation results are analyzed and discussed.The sixth chapter seeks for the actual data of the GDP as well as the related variables of 31 provinces,municipalities,and autonomous regions in China,applies two methods(the improved Differenced Forward GMM Estimation and the Horizontal Forward GMM Estimation)to conduct empirical analysis on it,and analyzes the influence degree and the regularity of these variables on GDP.The seventh chapter summarizes the whole thesis,and discusses some issues that can be further improved.