This paper examines the roles that information and preferences play in determining whether households choose schools with high value added. We study Romanian school markets using administrative data, a survey, and an experiment. The administrative data show that, on average, households could select schools with 1 s.d. worth of additional value added. This may reflect that households have incorrect beliefs about schools' value added, or that their preferences lead them to prioritize other school traits. We elicit households' beliefs and find that they explain less than a fifth of the variation in value added. We then inform randomly selected households about the value added of the schools in their towns. This improves the accuracy of households' beliefs and leads low-achieving students to attend higher-value added schools. We next estimate households' preferences and predict their choices under the counterfactual of fully accurate beliefs. We find that beliefs account for 18 (11) percent of the value added that households with low- (high-) achieving children leave unexploited. Interestingly, for households with low-achieving children, the experiment seems to have affected both beliefs and preferences. This generates larger effects on choices than would be predicted via impacts on beliefs alone.