一类近似因子模型的GMM估计及其统计性质研究  

A Study on GMM Estimation for a Class of Approximate Factor Models and Their Statistical Properties

作　　者：白仲林[1,2] 白强 

机构地区：[1]天津财经大学 [2]中国数量经济学会

出　　处：《统计研究》2016年第3期18-23,共6页Statistical Research

摘　　要：对于一类异质性误差项存在截面相关性的近似因子模型,本文首先提出了估计共同因子向量和因子载荷矩阵的广义矩估计方法(GMM),该方法推广了Doz等(2012)的极大似然估计方法;其次,分别研究了模型参数广义矩估计的渐近性质和有限样本的统计性质,在适当的条件下,证明了参数的GMM估计是具有渐近正态分布的一致估计;最后,利用近似因子模型对我国各类上市公司增长性的共同驱动因素及其差异性进行了实证分析。The paper firstly presents the Generalized Method of Moments for estimation of the common factors vector and the factor loadings matrix for a class of approximate factor models with the idiosyncratic error term being crosssectionally correlated. This method generalizes the maximum likelihood estimation method of Dozet al. （2012）. Secondly, the paper studies the asymptotic properties of generalized moments estimation of model parameters and statistical properties of finite sample respectively. It is proved that the GMM estimators of parameters are consistent estimations with asymptotic normal distribution under appropriate conditions. Finally, we use approximate factor models to detect common driving factors of growth of listed companies in China and their differences empirically.

关 键 词：近似因子模型 广义矩估计 一致性 渐近正态分布 

分 类 号：F222.3[经济管理—国民经济]