混合形式的变系数空间面板数据模型:一个多阶段估计

Varying-Coefficient Spatial Panel Data Model with Mixture Form:A Multi-Stage Estimation

当前对空间面板数据模型的研究主要集中在常系数模型上,但这类模型无法完全体现出空间异质性。本文建立变系数的空间面板数据模型,这类模型的特点是自变量系数是固定在时期上,同一时期的所有空间个体的自变量系数是相同的,而不同时期的自变量系数则会发生变化,不同时期的方程通过误差项的跨期相关性联系起来而形成SUR系统。在对模型进行一定设定的基础上,本文提出了一个四阶段估计,证明了参数估计量的一致性,并利用Monte Carlo方法对估计量进行了模拟。

At present,the research of spatial panel data model focuses on constant coefficient models,which however can not describe spatial heterogeneity completely.This paper sets up varying-coefficient spatial panel data model.In this paper,coefficients are fixed on time,and all coefficients are identical in the same time but varying in different time.The equations in different time connect through the intertemporal correlation of error term,and compose a SUR system.On the bases of proper model setting,this paper proposes a four-stage estimator,and proves the consistence of this estimator.We also simulate the small sample property of this estimator by use of Monte Carlo.