I take advantage of a sharp discontinuity in the probability of admission to an elite university at the admission score threshold, to estimate causal returns to college education quality. I use a newly constructed dataset, which combines individual administrative records about high school,...