A Matching Estimator Based on a Bilevel Optimization Problem
提出一种新的匹配估计量，通过优化协变量平衡准则内生决定邻居和权重，在NSW数据中表现优异，蒙特卡洛模拟显示其性能良好。
This paper proposes a novel matching estimator where neighbors used and weights are endogenously determined by optimizing a covariate balancing criterion. The estimator is based on finding, for each unit that needs to be matched, sets of observations such that a convex combination of them has the same covariate values as the unit needing matching or with minimized distance between them. We implement the proposed estimator with data from the National Supported Work Demonstration, finding outstanding performance in terms of covariate balance. Monte Carlo evidence shows that our estimator performs well in designs previously used in the literature.