A Study of Quasi-Experimental Control Group Methods for Estimating Policy Impacts
This study examines the efficacy of Quasi-Experimental Control Group (QECG) methods for estimating policy impacts. It establishes that QECG estimators can outperform the conventional regression (CR) estimator when policy adoption is endogenous (nonrandom), the relationship between outcomes and policies is nonlinear, and CR equations do not correctly specify the nonlinear form of the relationship. In the case of perfect matching, QECG methods produce unbiased estimates. In the case of imperfect matching, QECG estimators will be biased. To address this and other issues, we develop a more general QECG estimator that (1) allows control places to be matched to more than one treatment place, and (2) weights observations by the “closeness” of the match. Using Monte Carlo analysis we demonstrate that our estimator substantially improves estimates of policy impacts.