This paper uses administrative data from Missouri to examine the sensitivity of job training program impact estimates based on alternative nonexperimental methods. In addition to simple regression adjustment, we consider Mahalanobis distance matching and a variety of methods using propensity...