We compare nonexperimental impact estimates based on matching methods with those from a randomized evaluation to determine whether the nonexperimental approach can "match" the so-called gold standard. The social experiment we use was carried out to evaluate a geographically targeted conditional...