Measuring Diffusion over a Large Network using Instrumental Variables
Holder： Xiaoqi He（Central University of Finance and Economics）
Time：2025-04-24 15:10-17:00
Location：Room 217, Guanghua Building 2
Abstract:
Measuring diffusion over a network aims to measure the causal effect of an individual’s prior outcome on their neighbors’ present outcomes. However, in the actual process of diffusion in practice, the exogeneity of the prior outcomes is hard to obtain, because we rarely observe the very first exogenous outcomes of individuals in the process. In this paper, we consider an instrumental variable approach, by adopting a period-by-period analogue of the local average treatment effect (LATE) as our causal parameter and establishing its identification. The main challenge in the development stems from the individual heterogeneity of outcome distributions due to the complex nature of a large network that the distributions depend on. To address this challenge, we introduce a randomized LATE analogue as our causal parameter and demonstrate how this randomization makes feasible identification and estimation of the parameter in this setting with a complex, heterogeneous dependence structure.
About the Speaker:
何晓奇，中央财经大学管理科学与工程学院副教授，研究方向为计量经济学，因果推断的理论和应用。2016年在不列颠哥伦比亚大学获得经济博士学位。研究成果“Measuring diffusion over a large network”，发表于The Review of Economic Studies。
Your participation is warmly welcomed!
欢迎扫码关注北大统计科学中心公众号，了解更多讲座信息!
