The Generalized Moment Estimation And LM Test For Spatial Panel Data Model With Two-way Error Component

Spatial panel data model is a spatial economic model based on panel data ,and it is an important research content of modern economics,especially in the areas of regional science and economic geography,the number of the articles analysing spatial externality,the regional economic growth,knowledge spillover and diffusion of innovation ,spatial aggregation or dispersion pattern by the spatial panel data model is very large. As the model includes the spatial weight matrix ,the methods of the parameter estimation and the model test are different with that of the former’s. The aim of the paper is to estimated and test the spatial panel data model with two-way error component.In the paper,it firstly introduces and analyses the common settings of spatial weight matrix and the structures of some common spatial economic models ,with that it introduces the model the paper will studies—the spatial panel data model with two-way error component . The model assumes its spatial correlation not only exists in the dependent variable and error disturbing item ,but also exits in the individual effects. As the model contains the endogenous variables ,and the maximum likelihood (ML)estimation is complex timely when the N is large,so it chooses the fit instrument variable (IV) first,then estimated the parameters with the method of two stage least squares (2SLS)and the generalized moment (GM).The innovation is that the error component is two-way error component which contains random individual effects and random time effect,and during the general-ized moment estimation,it introduce the weighted generalized moment estiamtion .Following,it structures four lagrange multiplier (LM) test statistics to test whether the model can be simplified .The results of stochastic simulation show that the weighted generalized moment estimators is better than the generalized estimators,and the LM test statistics is feasible.