I investigate longitudinal preferences for schools revealed by families' applications to middle and high schools within a large urban school district's universal enrollment system. I find that preferences for schools' racial/ethnic composition are more stable than preferences for quality and proximity to home, even after concurrently controlling for all characteristics. White and Hispanic families' preferences for composition are more durable than those of Black families, though stability overall attenuates when focusing on subgroups. Because preferences are not rigid over time, research and policy might continue exploring whether families can be guided toward options that improve academic outcomes and/or address segregation.