Abstract: How do parents choose among schools when they are allowed to do so? In this paper, we analyze detailed information of 70,000 fourth-graders attending about 1,200 publicly subsidized schools in the context of the Chilean voucher system. We model the school choice of a household as a discrete choice of a single school, based on the random utility model developed by McFadden (1974) and the specification of Berry, Levinsohn, and Pakes (1995), which includes choice-specific unobservable characteristics and deals with potential endogeneity. Our results imply that households value some attributes of schools, with the two most important dimensions being test scores and distance to school. Interestingly, at the same time, our results suggest there is a lot of heterogeneity in preferences because the valuation of most school attributes depend on household characteristics. In particular, we find that while proximity to school is an inferior attribute, test scores is a normal attribute. We present evidence that our results are mainly driven by self-selection and not by school-side selection. As a nal check, we compute the average enrollment elasticity with respect to all school attributes and find that higher elasticities are correlated with higher supply of the attribute, especially in the case of test scores-enrollment elasticities for private schools.