Falsifying Marginal Treatment Effects Models

摘要：In this paper, we develop a novel specification test for IV validity assumptions within the framework of marginal treatment effects models. The IV validity assumptions are not directly tractable. However, they have the strongest testable implication involving two shape index sufficiency and monotonicity restrictions on the conditional joint density function of the outcome and treatment on the propensity score. To address these, we devise two distinct statistics, each tailored to evaluate one of these shape restrictions, which form the basis of the joint test for IV validity. We employ a multiplier bootstrap technique to approximate the null asymptotic distribution of these statistics. We propose a Monte Carlo experiment that examines the finite sample performance of the test and apply it to evaluate some commonly used instruments for college attendance. Our test provides a powerful tool to falsify marginal treatment effects models.

报告时间：2024年12月17日，下午2:00        

线下地点：yl23455永利6-304会议室

主办单位：yl23455永利数量经济研究中心、yl23455永利

协办单位：《财经论丛》编辑部

主讲人简介：毛明海，辽宁大学李安民经济研究院助理教授，博士毕业于马德里卡洛斯三世大学。主要研究领域为因果推断和应用计量经济学。论文发表于《中国农村经济》等期刊。主持和参与多项国家自然科学基金项目。担任Journal of Business & Economic Statistics等期刊匿名审稿人。