We model centralized school matching as a second stage of a simple Tiebout-model and show that the two most discussed mechanisms, the deferred acceptance and the Boston algorithm, both produce inefficient outcomes and that the Boston mechanism is more efficient than deferred acceptance. This advantage vanishes if the participants get to know their priorities before they submit their preferences. Moreover, the mechanism creates artificial social segregation at the cost of the disadvantaged if the school priorities are based on ex ante known (social) differences of the applicants.