固定效应部分线性单指标空间自回归面板模型的二次推断函数估计

Quadratic inference functions estimation for partially linear singleindex spatial autoregressive panel model with fixed effects

从模型结构的特点着手,获得了一些不依赖于工具变量的矩条件.将这些矩条件与二次推断函数法(QIF)和最小二乘虚拟变量法(LSDV)结合,为模型构建了一种新的估计方法.该方法的优点是考虑了空间内生性的同时,还将个体内的相关结构包含其中.进一步,在一些正则条件下,研究了模型估计量的大样本性质,发现非参数估计量具有最优收敛速度,参数估计量渐近于正态分布.同时,采用Monte Carlo模拟评价了估计方法在有限样本下的表现,结果表明文中所述方法的表现符合大样本性质,且远远优于忽略相关性的估计方法.最后,将所述方法应用于实际数据分析中.

Some moment conditions independent of instrumental variables are attained from fea-ture of model structure at the first.Combining these moment conditions with quadratic inference functions method and least square dummy variable method,a new estimation method for the model is constructed.The advantage of this method is that it considers spatial endogeneity and the correlation structure within individuals.Furthermore,under some mild conditions,the large sample properties of the model estimators are studied.It is found that the nonparametric estimator achieves the opti-mal convergence rate and the parameter estimators asymptotically approach the normal distribution.Meanwhile,Monte Carlo simulation is used to evaluate the performance of the proposed method in finite samples.The results show that the performance of the proposed method conforms to the large sample property and is far superior to the estimation method ignoring the correlation within individuals.Lastly,the proposed method is applied to practical data analysis.