New Paper: Chernozhukov et al. Characterize Omitted Variable Bias in Difference-in-Differences Designs
RexDouglass · x · 2026-09-20
Wang、Sant'Anna、Chernozhukov、Cinelli 在 arXiv 发布新论文《Omitted Variable Bias in Difference-in-Differences Designs》,研究 DID 设计中未观测混杂违反平行趋势假设时的遗漏变量偏差(OVB)问题。
- 给出了处理组平均处理效应(ATT)偏差的 OVB 公式新刻画,表明偏差主要由处理分配机制中混杂的强度决定
- 提出三种量化混杂强度的方式:处理组平均处理几率的变化、处理组与对照组的混杂不平衡、未处理组中处理几率的解释变异
- 提供可常规报告的敏感性统计量:推翻 DID 结论所需的最小混杂强度,以及基于观测协变量或前期趋势的混杂强度形式化边界
- 给出高效的统计推断方法,可结合现代机器学习算法估计
- 以最低工资对青少年就业影响为例演示方法
More from Research
- Long trajectories go for ~$1k each, data mining breaks even at 2k samples — HanchungLee · 2026-09-20
- Sitzmann clarifies: only learned systems generalize to unknown in-the-wild objects — vincesitzmann · 2026-09-20
- Brain evolved from two primitive nervous systems merged together, Stanford study finds — AnnaCiaunica · 2026-09-20
- Harvard lab releases one year of LLM inference traces: 6.12B production requests across 9,174 models — airesearch12 · 2026-09-20
- NASA releases new Curiosity postcard as rover crosses 5,000 Martian days — XFreeze · 2026-09-20
- A Relational Lens on How LLMs Construct Genuinely New Mathematical Objects — Endless-monkey · 2026-09-20