Debiased Difference-in-Differences Estimation and Inference of Causal Effects Under Adaptive Controls
讲座通知 2026 年 7 月 7 日（星期二）， 上午 10:00 - 11:30 信息管理与工程学院308室 上海市杨浦区武东路100号 主题 Debiased Difference-in-Differences Estimation and Inference of Causal Effects Under Adaptive Controls 主讲人 Sentao Miao University of Colorado Boulder Sentao Miao is an Assistant Professor of Operations Management in Leeds School of Business at University of Colorado Boulder. Previously, he was an Assistant Professor in Bensadoun School of Retail Management & Desautels Faculty of Management at McGill University. His research interests are mainly in developing efficient learning and optimization algorithms with various applications in Operations Management. For methodologies, Sentao Miao focuses on statistical and machine learning algorithms such as online learning, multi-arm bandit problem, reinforcement learning; he is also interested in approximation algorithms with provable performance. For applications, he mainly works on operations management problems such as dynamic pricing, assortment selection, inventory control, etc. Sentao Miao obtained his PhD degree in Department of Industrial and Operations Engineering at University of Michigan. 讲座简介 Differences (DiD) method is a widely used empirical strategy for identifying causal effects when randomized experiments are infeasible. By comparing outcome trends between treated and untreated units before and after an intervention, DiD isolates treatment effects while controlling confounders exhibiting parallel temporal trends. However, as adaptive experiments and data-driven decision systems become increasingly prevalent in managerial and operational settings, the validity of standard DiD methods is challenged when data collection depends on past outcomes. This paper studies the problem of estimating and conducting inference on causal effects within the DiD framework when data are adaptively collected under policy control. We make three main contributions. First, we develop a general debiasing method that converts the bias of standard DiD estimators under adaptive data collection into additional variance. Second, we provide finite-sample bounds on the bias and variance of the proposed estimator and establish the asymptotic validity of the resulting confidence intervals. Third, we demonstrate the practical relevance of our approach through dynamic pricing examples, showing that our estimator achieves a minimax-optimal trade-off between estimation accuracy and policy regret. Together, these results offer a robust framework for applying DiD methods in sequential decision-making environments, extending their applicability to modern adaptive data settings where classical assumptions no longer hold.
上海财经大学（武东路校区）
