Economists generally assume, implicitly, that “the return to schooling" is invariant across local labor markets. We demonstrate that this outcome pertains if and only if preferences are homothetic--a special case that seems unlikely. Our theory predicts that returns to education will instead be relatively low in expensive high-amenity locations. Our analysis of U.S. data provides support for this contention; returns to college are especially low in such cities as San Francisco and Seattle. Our findings call into question standard empirical exercises in labor economics which treat the returns to education as a single parameter.