A public college in Mexico City randomly assigns applicants into a group that can immediately enroll and a group that can only do so after one year. The author shows that the standard model of educational decisions predicts no (or minimal) effect of deferral on educational attainment. He surveyed the applicants to this college for the 2007/2008 academic year. Using data from that survey, he finds that, one and a half years after the first group enrolled, individuals in that group were 19 percentage points more likely to be enrolled than those that had to wait. He finds that one additional slot increases the attainment of about 0.3 individuals of the applicant pool and that offering them to individuals of poorer backgrounds has an even larger effect. He proposes a decision-making model where wages (and opportunity costs) vary due to a random component. He derives testable implications of the model and show that they are verified empirically. He estimates the parameters of the model and show that the model can explain the observed patterns under reasonable assumptions. He uses the estimated model to project the long-term of deferred admission on long-term attainment for different groups of applicants.