Recent research in economics suggests a positive association between self-esteem and earnings. A major problem in this literature is that from simple cross-sectional wage regressions it is not possible to conclude that self-esteem has a causal impact on earnings. While classical measurement error leads to an attenuation bias, reverse causality and omitted variable are likely to drive the OLS coefficient on self-esteem upward. Using the National Longitudinal Survey of Youth (NLSY) that administered the Rosenberg Self-Esteem Scale during the 1980 and 1987 interviews, I provide further evidence for the existence of a self-esteem premium by exploiting variation in these measures in the two years. I show that the estimated impact of self-esteem in 1987 on earnings is about two times greater than previous OLS estimates would imply. The main explanation for this result is the large extent of measurement error in the reported self-esteem measure.