Prof Rick Sander, guest-blogging over at Volokh has interesting data and discussion about affirmative action in college admissions and the eventual outcomes. Note that he is being quite general here, including legacies as well as racial affirmative action - in fact, anyone who was less-prepared coming into college, accepted on the hopes that they would catch on and catch up. He tracked the progress of all who came in with lesser credentials on a set-aside.
The results were not pretty on several levels, especially in STEM majors. Affirmative action admits flunk out or switch to easier majors at such a high level that the author believes that there would be more black, hispanic, and native lawyers, doctors, physicists, and engineers if they had gone to less-elite schools in the first place, where they were more like their classmates. The mismatch may be worse than mere prejudice, in terms of eventual outcomes.
The authors have been vilified, but not controverted, as they keep noting. One of the things that they call for is more transparency. The elite schools want to have diversity numbers that look good, and so do not acknowledge the tradeoff to highschoolers applying. They don't want those 17-year-olds to know that if their goal is to be a doctor, they might have a better chance elsewhere. Thus, they scoop the top minority students, unintentionally treating them badly, leaving the B+ level schools to take the next group down, etc. Eventually, most minority students ends up slightly over their heads, all for the "honor" of getting into a "better" school.
So who benefits, long term? Only the school, not the student or the society. It's hard to fault teenagers for jumping at the short-term gain and prestige of certain schools. It's a temptation. But they should at least have the data going in. Those who run cross-country at least, even if no others, might understand the concept of starting slower and gradually reeling in those ahead of you (and the great satisfaction of that).