美国波士顿学院教授ShakeebKhan做客南开金融学术前沿讲座

（通讯员 欧阳夫）2018年6月29日，波士顿学院 Shakeeb Khan 教授应邀访问南开大学金融学院，并为南开师生就其工作论文 “Identification in Dynamic Semiparametric Binary Response Panel Data Models” 作了精彩报告。

在该论文中，Khan 教授及其合作者分析了在各种一般性假设条件下的半参数面板数据动态二元选择模型的识别问题。此类模型所研究的选择问题综合考虑了状态依赖性，动态性与样本异质性，因而在计量经济学的理论与应用中有着越来越重要的意义。本文首先分析了包含条件平稳性假设的二元动态选择模型，并刻画了关于其隐含随机效用函数的参数的最小可识别集。其所采用的分析方法可以被广泛应用到包含离散解释变量、时间趋势与时间虚拟变量的面板数据动态选择模型中，且并不要求模型一定要包含至少一个连续型解释变量。而其所得到的最小可识别集可以被表示为一系列凸多面体的交集。随后，Khan 教授及其合作者将同样的分析方法应用到具有更强的条件可交换性条件的选择模型中，并证明假设条件可交换性的模型对未知效用参数有更强的识别能力。随后，本文探索了含有在时间序列上满足独立同分布条件的扰动项的选择模型，并证明此类模型中，更强的独立同分布条件（相比于条件可交换性），并不能增强模型对未知参数的识别能力。此外，本文还发现，如果进一步假设扰动项与解释变量之间的独立性，则应用本文所提出的分析方法所得到的未知参数的可识别集可能与应用 Honore and Kyriazidou (2000) 的分析方法所得到的可识别集并不一致。最后，Khan 教授及其合作者将其在本文中所发展的分析方法应用到含有更长时间序列的面板数据模型中，以探索更长的时间序列对模型识别能力的影响和作用。

Shakeeb Khan，波士顿学院（Boston College）经济系经济学教授，美国普林斯顿大学经济学博士，Fellow of Journal of Econometrics。Khan 教授的主要研究领域为理论计量经济学和应用计量经济学。他的研究成果曾发表于 Econometrica，Journal of the American Statistical Association，Journal of Econometrics，Journal of Applied Econometrics，Journal of Business Economics and Statistics, Econometric Theory，Econometrics Journal 等诸多计量经济学领域内顶级期刊。Khan 教授现任国际知名学术期刊 Econometrica，Econometrics Journal，Econometric Reviews，Journal of Econometric Methods 的副主编，曾任Journal of Business Economics and Statistics 的主编及副主编。

On 29th June 2018, Professor Shakeeb Khan from Boston College visited the School of Finance, Nankai University and presented his ongoing research joint with Professor Maria Ponomareva and Elie Tamer from, respectively, Northern Illinois University and Harvard University, “Identification in Dynamic Semiparametric Binary Response Panel Data Models”, at Room 234, School of Finance, Nankai University.

This paper analyzes identification in dynamic semiparametric models of binary choice under general conditions. This class of models is increasingly important in trying to fit choice data while allowing for state dependence, dynamics, and heterogeneity. First, this paper characterizes the sharp set for latent utility parameters in a dynamic panel data model of binary choice under conditional stationarity. The analysis is general in that it allows for discrete covariates, time trends and/or time dummies without requirement for the existence of continuous regressors. The identified set can be characterized by an intersection of convex polyhedrons. The same exercise under the stronger assumption of conditional exchangeability is conducted, which exhibits incremental identifying power. Next, this paper shows that the assumption that the unobserved utility components are stationary and independent over time conditional on observables has no identifying power on top of conditional exchangeability condition. Third, this paper derives the identified set under full independence and shows that this set may not coincide with the approach used in Honore and Kyriazidou (2000). The proposed identification approach is also extended to study models that allow for more time periods, which establishes the informational content of longer time series.

Shakeeb Khan, Professor of Economics from Boston College, Fellow of Journal of Econometrics, obtained his PhD degree in economics from Princeton University. Professor Khan’s research is in theoretical and applied econometrics. His research has been published on top econometrics and statistics journals including Econometrica, Journal of the American Statistical Association, Journal of Econometrics, Journal of Applied Econometrics, Journal of Business Economics and Statistics, Econometric Theory and Econometrics Journal, etc. He currently serves as an associate editor of Econometrica, Econometrics Journal，Econometric Reviews and Journal of Econometric Methods. He is a former associate and co-editor of Journal of Business Economics and Statistics.