This thesis attempts to deal with difficult methodological problems of program evaluation; the particular programs considered are two types of manpower training programs: institutional or "classroom" training and government-subsidized on the job training. Characteristics of cost-benefit analysis are investigated and the method is critized for use in analyzing manpower programs. Two data sources are used to study the production process. First, a group of trainees who underwent each type of training were interviewed. Second, employers who participated in the OJT program were interviewed, as were a group of the largest employers in the New Haven area. (author/eb)