We show how data from an evaluation in which subjects are randomly assigned to some treatment versus a control group can be combined with nonexperimental methods to estimate the differential effects of alternative treatments. ...