A project used a mixed quantitative-qualitative approach, drawing from the Job Corps database and site visits at five sites, to generate knowledge that can guide policymakers and program planners as they act to increase retention in the Job Corps or similar programs. Quantitative data on 343,097 students who enrolled in Job Corps between July 1993 and December 1998 were analyzed. Of this group, 86 percent remained in the program at 30 days, and 64 percent remained at 90 days. Students were more likely to leave if they were in centers in which they were greatly in the minority by race, and younger students were much more likely to leave under negative circumstances than older students. Qualitative data showed that intrinsic unmeasured characteristics of students, such as commitment, attitude, motivation, confidence, maturity level, emotional status, willingness to change, and ability to interact with others, are of great importance in determining whether students will stay. Barriers to retention include intrinsic student factors, personal or institutional barriers, and student-staff interactions. The study concluded that although the easiest way to improve retention statistics would be to discourage youth at greatest risk of dropping out from enrolling in the program, this approach would be inconsistent with Job Corps' mission of serving the nation's most needy youth. Among recommendations for staff were that they demonstrate they expect the best from all youth; build students' self-confidence and sense of connectedness to peers and staff; learn to communicate effectively with youth from different cultures; teach stress reduction and coping strategies; foster a multicultural environment; and communicate effectively with youth at different developmental stages.. (Contains 110 references.) (KC)