"In this paper we study whether location-specific price variation likely affects statistical inference and theoretical interpretation in the empirical implementation of human capital earnings functions. We demonstrate, in a model of local labor markets, that the "return to schooling" is a constant across locations if and only if preferences are homothetic--"a special case that seems unlikely to generally certain. Examination of U.S. Census data (for 1980, 1990, and 2000) provides persuasive evidence that the return to a college education, relative to a high school education, does indeed vary widely across cities, e.g., in 1990 the return in Houston is 0.54 while in Seattle it is only 0.33. We provide theoretical reasons to suspect that the returns to education are relatively lower in expensive high-amenity locations, and present evidence consistent with this prediction. Finally, we raise concerns about standard empirical exercises in labor economics which treat the returns to education as a single parameter"--Federal Reserve Bank of St. Louis web site.