Who is overeducated and why? Probit and dynamic mixed multinomial logit analyses of vertical mismatch in East and West Germany

Overeducation is an often overlooked facet of untapped human resources. But who is overeducated and why? Relying on SOEP data 1984-2011, we use probit models for estimating the likelihood of entering overeducation and dynamic mixed multinomial logit models with random effects addressing state dependence and unobserved heterogeneity. As further robustness checks we use three specifications of the target variable, i.e. realized matches, self-assessment and twofold overeducation. We run separate analyses for men and women, East and West Germans and medium and highly educated persons. We find that overeducation is mainly state dependent. Nonetheless, even in the dynamic context staying employed proves to be risk-decreasing. By contrast, scars of past unemployment show up in a higher mismatch risk. Moreover, an employer change does not serve as a suitable exit strategy, and a dual qualification does not show up as a valid insurance against graduates' job mismatch. Overall, effects largely depend on the operationalization of overeducation. We conclude that to combat overeducation, focusing on continuous employment careers and circumventing unintentional withdrawals from the current job is crucial. Moreover, institutional impediments that restrain job match quality for certain groups (migrants, mothers) have to be tackled.