Impact evaluations of manpower programs have had many shortcomings, especially in finding control groups for comparison. Methodological bias and inconsistency between evaluators, along with disinterest by program administrators, have prevented evaluation from reaching its potential in program planning. The use of longitudinal studies to solve control group problems, with standardized criteria for benefit-cost analysis, could eliminate much of the inconsistency in evaluation. With improved information systems at the local project level, the evaluator and policymaker together could plan projects on the basis of accurate comparisons. (BH)