This paper analyzes the effect of educational mismatch on wages in Germany, using data from the German Socio-Economic Panel. Educational mismatch has been discussed extensively, mostly by applying OLS wage regressions which are prone to an unobserved heterogeneity bias. This problem is approached by using FE and IV models. As a stability check, the regressions are rerun using data from the International Adult Literacy Survey, allowing for an explicit control of skills as proxy of abilities. Results show that unobserved heterogeneity does not explain the wage differences between actual years of education and years of required education. This rejects the hypothesis that mismatched workers compensate for heterogeneity in innate abilities. The results suggest a structural problem in the German educational system as skill demand and supply are not in long-term equilibrium.