[tomp]

> Whatever you call it, it is in fact an outcome, and one with a demonstrated influence on other outcomes of interest

Maybe, but if you optimize only this metric (enrollment) while disregarding all others (graduation), then you'll soon get ineffective results and the "demonstrated" influence will no longer hold true - simply enroll minorities, regardless of their knowledge or intelligence, and make sure most fail the next year. Will you be satisfied? Will it still correlate with the desired outcome?