Inference with “Difference in Differences” with a Small Number of Policy Changes
针对双重差分应用中政策变化组数很少的情况，提出一种利用大量未变化组信息的替代推断方法，可构造检验统计量和置信区间。
In difference-in-differences applications, identification of the key parameter often arises from changes in policy by a small number of groups. In contrast, typical inference assumes that the number of groups changing policy is large. We present an alternative inference approach for a small (finite) number of policy changers, using information from a large sample of nonchanging groups. Treatment effect point estimators are not consistent, but we can consistently estimate their asymptotic distribution under any point null hypothesis about the treatment. Thus, treatment point estimators can be used as test statistics, and confidence intervals can be constructed using test statistic inversion. © 2011 The President and Fellows of Harvard College and the Massachusetts Institute of Technology.