In response to this evidence, we es

In response to this evidence, we estimate more general models of earnings where the coefficientonschoolinginalogearningsequationisnot, ingeneral, interpretableasaninternal rate of return. From estimated cross section earnings functions, we compute marginal internalratesofreturntoeducationseparatelyforblackandwhitemenacrossdifferentschooling levels for different decades using data from the U.S. decennial Censuses and the Current Population Survey (CPS). Our estimates account for nonlinearities and non-separabilities in earnings functions, income taxes, and tuition. When contrasting these estimated returns with the estimated schooling coefficient from a Mincer earnings function, we find that both levels and trends in rates of return generated from the Mincer model are misleading for many schooling levels.