Subjective and projected returns to education

•I calculate subjective wage and employment returns to education.•Subjective returns are uncorrelated with projected returns estimated using observed data.•The use of projected returns as a proxy for subjective returns is unsupported. There is significant heterogeneity over high school students in the wage and employment rate returns to education. I evaluate this heterogeneity using subjective returns derived from a data set of high school juniors and seniors in Washington State. Variation over observables in projected returns estimated using observed data is uncorrelated with variation in subjective returns elicited by directly asking students about their beliefs. These results mean that returns estimated using observed data are likely a very weak proxy for student beliefs.