无条件分位数回归：文献综述与应用实例

统计研究 ›› 2012, Vol. 29 ›› Issue (3): 88-96.

无条件分位数回归：文献综述与应用实例

朱平芳　张征宇

出版日期:2012-03-15 发布日期:2012-03-22

Unconditional Quantile Regression: Literature Review and Empirical Example

Zhu Pingfang & Zhang Zhengyu

Online:2012-03-15 Published:2012-03-22

摘要/Abstract

摘要： 条件分位数回归(conditional quantile regression，CQR)方法已成为经济学实证研究的常用方法之一。由于CQR结果的经济学阐释基于过多甚至是不必要的控制变量，这与人们所关心的问题有可能并不一致。例如，在劳动经济学对教育回报的研究中，无论个体的年龄，性别与家庭特征如何，教育程度对于个人收入的异质性影响是人们关注的重点，即人们想了解收入关于教育程度的无条件分位数估计。本文旨在介绍近年来发展起来的无条件分位数回归(unconditional quantile regression，UQR)技术并梳理相关文献。特别地，本文介绍三种重要的无条件分位数回归模型：Firpo, Fortin和Lemieux(2009)提出的的再中心化影响函数(recentered influence function, RIF) 回归，Frolich和Melly(2010)提出的无条件分位数处理效应模型与Powell(2010)提出的一般无条件分位数回归。另外,论文还运用一个研究居民收入分配格局变化对其医疗支出影响的实例详细说明了新方法的应用。

关键词: 条件分位数回归, 无条件分位数回归, RIF回归, 处理效应模型

Abstract: Conditional quantile regression (CQR), introduced by Koenker and Bassett(1978), has become a popular method in empirical economics. However, researchers often find it difficult to interpret their empirical results through CQR. For example, researchers are interested in learning about the heterogeneous impact of education level upon personal income, unconditional on other individual characteristics, e.g., age, gender and family background, that is, the unconditional quantile marginal effect of education on income. This paper summarizes recent advances of the literature on unconditional quantile regression. Specifically, we discuss three important models: Firpo, Fortin and Lemieux(2009)’s RIF regression, Frolich and Melly(2010)’s unconditional quantile treatment effect model and Powell(2010)’s unconditional quantile regression model. Then we provide an empirical example to illustrate the usefulness of this new method.

Key words: Conditional Quantile Regression, Unconditional Quantile Regression, RIF Regression, Treatment Effects Model

朱平芳　张征宇. 无条件分位数回归：文献综述与应用实例[J]. 统计研究, 2012, 29(3): 88-96.

Zhu Pingfang & Zhang Zhengyu. Unconditional Quantile Regression: Literature Review and Empirical Example[J]. , 2012, 29(3): 88-96.