This paper proposes an improved concept of educational mismatch that combines a statistical measure of over- and undereducation with the worker's self-assessment of skill utilization. In that way, we account for worker heterogeneity in skills whose omission may generate biased estimates of the...