What`s the Matter with Tie-breaking? Improving Efficiency in School Choice

Very little is known about the student-optimal stable mechanism when school priorities are weak. In current practice, the student proposing deferred acceptance algorithm is applied after indifferences in priority orders are broken with a lottery. Although such a tie-breaking procedure preserves stability, it adversely affects the welfare of the students since it introduces artificial stability constraints. We propose a simple procedure to compute a student-optimal stable matching when priorities are weak. The idea behind our construction relies on a new notion which we call a stable improvement cycle. Abdulkadiroglu, Pathak, and Roth (2006) report that had our algorithm been applied to the preference data of the 2003-2004 New York City High School Match, 6,854 students (10.5% of the 63,795 matched students) would have been matched with schools higher on their preference lists without hurting the others. We run simulations to understand the qualitative effects of correlation in preferences and of locational preference on the size of the efficiency gain. We also investigate the strategic properties of the class of student-optimal stable mechanisms.