Partially Linear Spatial Lag Model With Fixed Effects

Traditional econometrics ignore the spatial dependency and spatial heterogeneity caused by the geospatial proximity,which may lead to the failure or deviation of the methods and test results.However,the parametric spatial measurement models with pre-set parameters have a strong subjectivity.For some specific measurement problems,they are not enough to fully describe the potential relationship between the explained variable and the explanatory variables.Therefore,this paper attempts to construct a class of semi-parametric spatial lag panel model,and do the necessary theoretical derivation and numerical simulation of its estimator.Yu,Jong&Lee(2008)investigate the asymptotic properties of quasi-maximum likelihood estimators for spatial dynamic panel data with fixed effects.Su&Jin(2010)propose profile quasi-maximum likelihood estimation of spatial autoregressive models that are partially linear.Based on these,we propose a class of partially linear spatial lag panel model with fixed effects.The influence of fixed effect,parameter independent variable and nonparametric independent variable are considered synthetically in the model.The parametric and nonparametric parts of the model are estimated on the basis of the profile quasi-maximum likelihood estimation,and asymptotic properties for the estimators are deduced under the condition that the number of individuals and the number of periods are large.The results show that,under large sample conditions,the estimators are consistent and the parameter estimator satisfies the asymptotic normal distribution and the convergence rate is(?).Monte-Carlo simulation results show that the estimators of the model have good performance under finite sample,and the accuracy improves with the increase of the sample size.