【主 题】Decision Theory for Treatment Choice Problems with Partial Identification
【报告人】Chen Qiu（助理教授，美国康奈尔大学）
【时 间】2023年12月25日周一15:00-16:30
【地 点】经济学院701会议室
【语 言】英文
【摘 要】A policy maker must decide between implementing a new policy or preserving the status quo. Her data partially identify payoff-relevant parameters. We show that the application of statistical decision theory to such treatment choice problems with partial identification presents important theoretical and practical challenges, as well as interesting research opportunities. The challenges are: In a general class of problems with Gaussian likelihood, all decision rules are admissible; it is maximin-welfare optimal to ignore all data; and there are infinitely many minimax-regret optimal decision rules, all of which randomize the policy action for some data realizations. The opportunities are: We introduce a profiled regret criterion that can reveal important differences between rules and also render some inadmissible; and we uniquely characterize the mimimax-regret optimal rule which least frequently randomizes the policy recommendation. We illustrate these results in three examples: aggregation of experimental estimates for policy adoption, extrapolation of Local Average Treatment Effects, and policy making in the presence of omitted variable bias.
【报告人简介】Chen Qiu is an Assistant Professor of Economics at Cornell University. Before joining Cornell, he was a Postdoctoral Fellow at the Department of Economics and Institute of Fiscal Studies, University College London, during 2020-2021. He is an econometrician and his research interests are causal inference, treatment choice and statistical decision theory. He obtained the PhD in Economics from London School of Economics in 2020.
