带未知异方差与半线性结构Tobit模型的半参数估计  

Semi-parametric Estimation of Partially Linear Tobit Models with Unknown Heteroskedasticity

作　　者：张征宇[1] 

机构地区：[1]上海社会科学院数量经济研究中心

出　　处：《数量经济技术经济研究》2011年第6期147-160,F0003,共15页Journal of Quantitative & Technological Economics

基　　金：上海市教育发展基金会2010年晨光计划项目"广义空间计量模型前沿理论及其在微观金融市场个体策略性博弈中的实证研究"(编号:2010CG66)的资助

摘　　要：本文将Tobit模型扩展至同时带未知条件异方差与半线性结构回归函数的场合,并提出一种计算简便的半参数二步估计法。该方法的关键之处在于连续两次施以成对相减变换,并先后消去第一步所得被解释变量非参数条件分位函数中的两类非线性冗余成分(非线性回归函数部分与未知异方差结构)。文章证明了估计量的n-一致性与渐近正态性,并通过Monte Carlo模拟研究了分位点对的选择、扰动项分布类型与样本删尾程度等因素对估计量小样本性质的影响。最后通过国内居民医疗服务利用不平等的实例验证了本文所提的方法。A two- stage semi - parametric estimation procedure is proposed for a class of Tobit models in the presence of both partially linear structure and unknown conditional heteroskedastiticy. This procedure is able to eliminate nuisance nonlinear components appearing in the regression function by pair - wise differencing twice the nonparametrically estimated conditional quantile at the first stage. The resulting estimator is shown to be root - n - consistent and normally distributed. A Monte Carlo simulation is conducted to explore the impact upon its small sample performance through choice of quantile pairs under various error term distributions and degrees of sample censoring. As an empirical example, the suggested method is used to test the inequality of healthcare service in China.

关 键 词：TOBIT模型 条件异方差 半线性模型 非参数分位点估计 

分 类 号：F224.0[经济管理—国民经济]