Identification and Estimation of a Nonparametric Panel Data Model with Unobserved Heterogeneity ∗
Identification and Estimation of a Nonparametric Panel Data Model with Unobserved Heterogeneity ∗
复制标题
具有未观察到的异质性的非参数面板数据模型的识别和估计*
DOI：
--
复制
发表时间：
2009
期刊：
影响因子：
--
通讯作者：
Kirill S. Evdokimov
中科院分区：
文献类型：
--
作者：
Kirill S. Evdokimov
本文考虑具有非加性不可观察异质性的非参数面板数据模型。与标准线性面板数据模型一样，模型中存在两种类型的不可观察量：个体特定效应和特殊干扰。个体特异性效应不可分离地进入结构函数，并允许以任意方式与协变量相关。特异扰动项可与结构函数相加分离。建立模型所有结构元素的非参数识别。识别不需要参数分布或函数形式假设。识别结果具有建设性，只需要两个时间段的面板数据即可。因此，该模型允许使用短面板对异质边际效应进行非参数分布和反事实分析。本文还开发了一种非参数估计程序并推导出其收敛速度。作为副产品，获得了条件反卷积问题的收敛率。所提出的估计器易于计算并且不需要数值优化。蒙特卡罗研究表明，估计器在有限样本中表现良好。 ＊本文是我论文第一章的修订。我非常感谢 Donald Andrews、Xiaohong Chen、Yuichi Kitamura、Peter Phillips 和 Edward Vytlacil 的建议、支持和热情。我特别感谢我的导师北村雄一为我的智力培养所付出的努力和时间。我还从与 Joseph Altonji、Stephane Bonhomme、Martin Browning、Victor Chernozhukov、Flavio Cunha、Bryan Graham、Jerry Hausman、James Heckman、Stefan Hoderlein、Joel Horowitz、Ilze Kalnina、Lung-Fei Lee、Simon Lee、Yoonseok Lee、Taisuke Otsu、James Powell、Pavel Stetsenko 和 Quang Vuong 的讨论中受益匪浅。我还要感谢耶鲁大学、布朗大学、芝加哥布斯商学院、芝加哥经济学院、杜克大学、EUI、麻省理工学院/哈佛大学、西北大学、普林斯顿大学、图卢兹大学、加州大学伯克利分校、加州大学洛杉矶分校、沃里克大学、2009 年计量经济学会北美夏季会议、SITE 2009 年非参数计量经济学进展夏季研讨会以及 Chateau 暑期学校统计的研讨会参与者提供的非常有用的评论。所有错误都是我的。感谢考尔斯基金会通过卡尔·阿维德·安德森奖提供的财政支持。 †电子邮件：kevdokim@princeton.edu。
This paper considers a nonparametric panel data model with nonadditive unobserved heterogeneity. As in the standard linear panel data model, two types of unobservables are present in the model: individual-specific effects and idiosyncratic disturbances. The individual-specific effects enter the structural function nonseparably and are allowed to be correlated with the covariates in an arbitrary manner. The idiosyncratic disturbance term is additively separable from the structural function. Nonparametric identification of all the structural elements of the model is established. No parametric distributional or functional form assumptions are needed for identification. The identification result is constructive and only requires panel data with two time periods. Thus, the model permits nonparametric distributional and counterfactual analysis of heterogeneous marginal effects using short panels. The paper also develops a nonparametric estimation procedure and derives its rate of convergence. As a by-product the rates of convergence for the problem of conditional deconvolution are obtained. The proposed estimator is easy to compute and does not require numeric optimization. A Monte-Carlo study indicates that the estimator performs very well in finite samples. ∗This paper is a revision of the first chapter of my thesis. I am very grateful to Donald Andrews, Xiaohong Chen, Yuichi Kitamura, Peter Phillips, and Edward Vytlacil for their advice, support, and enthusiasm. I am especially thankful to my advisor Yuichi Kitamura for all the effort and time he spent nurturing me intellectually. I have also benefited from discussions with Joseph Altonji, Stephane Bonhomme, Martin Browning, Victor Chernozhukov, Flavio Cunha, Bryan Graham, Jerry Hausman, James Heckman, Stefan Hoderlein, Joel Horowitz, Ilze Kalnina, Lung-Fei Lee, Simon Lee, Yoonseok Lee, Taisuke Otsu, James Powell, Pavel Stetsenko, and Quang Vuong. I also thank seminar participants at Yale, Brown, Chicago Booth, Chicago Economics, Duke, EUI, MIT/Harvard, Northwestern, Princeton, Toulouse, UC Berkeley, UCLA, Warwick, North American Summer Meeting of the Econometric Society 2009, SITE 2009 Summer Workshop on Advances in Nonparametric Econometrics, and Stats in the Chateau Summer School for very helpful comments. All errors are mine. Financial support of Cowles Foundation via Carl Arvid Anderson Prize is gratefully acknowledged. †E-mail: kevdokim@princeton.edu.
固定效应面板数据模型的非参数估计和测试。
影响因子：
6.3
作者：
Henderson,DanielJ;Carroll,RaymondJ;Li,Qi
通讯作者：
Li,Qi