一个真实随机效应共同随机前沿模型的设定和估计  

Setting and Estimation of a True Random Effects Stochastic Meta-Frontier Model

作　　者：冯冬发 张涛[2] 褚冬琳 FENG Dong-fa;ZHANG Tao;CHU Dong-lin(National School of Development,Peking University,Beijing 100871,China;University of Chinese Academy of Social Sciences,Beijing 102488,China;Guangxi Minzu University,Nanning 530006,China)

机构地区：[1]北京大学国家发展研究院,北京100871 [2]中国社会科学院大学,北京102488 [3]广西民族大学,广西南宁530006

出　　处：《数理统计与管理》2023年第5期856-868,共13页Journal of Applied Statistics and Management

基　　金：国家自然科学基金项目(71991475);研究生科研创新支持计划项目(2020-KY-048)。

摘　　要：在经典随机前沿模型的基础上引入随机效应与群组效应,构建出一个真实随机效应共同随机前沿模型,可在考虑个体异质性的前提下,将决策单元的综合技术效率分解为组间技术差距与组内技术效率。基于极大模拟似然估计法估计模型参数,使用Halton序列缩减模拟抽样量,并结合KNN算法预测决策单元的技术效率。模拟实验结果表明,模型参数估计量与技术效率预测量的精确度显著高于经典模型,且会随样本量扩大而不断提升。实例应用发现我国区域经济存在个体效应和群组效应,造成东中西部技术效率差异的主要原因在于区域间技术壁垒阻碍了先进技术的自由流动。This paper constructs a true random effects stochastic meta-frontier model by incorporating random effects and group effects into classical stochastic frontier model,which can decompose DMU's technical efficiency into technical eficiency within group and technology gap between groups under the premise of considering individual heterogeneity.This paper estimates model parameters based on maximum simulated likelihood estimation,utilizes the Halton sequence to reduce sampling size,and uses the KNN algorithm to predict technical efficiency.The simulation experiment results show that the accuracy of model parameters'estimator and technical eficiency predictor is significantly higher than the classical model and will perform better with the expansion of the sample size.The application found that China's regional economy has individual effects and group effects.The main reason for the difference in technical efficiency between the eastern,central and western regions is that technical barriers between regions hinder the free flow of advanced technologies.

关 键 词：层次面板数据 随机前沿模型 共同前沿 极大模拟似然估计 

分 类 号：F224.0[经济管理—国民经济] O212[理学—概率论与数理统计]