The Theory of career mobility (Sicherman and Galor 1990) claims that wage penalties for overeducated workers are compensated by better promotion prospects. A corresponding empirical test by Sicherman (1991), using mobility to an occupation with higher human capital requirements as an indicator for upward career mobility, was successful confirmed by Robst (1995). However, both tests include controls for the opposite phenomenon of undereeducation and report ambiguous results without offering sound theoretical explanations. Estimating random effects probits with data from the German Socio-Economic Panel we show that part of the problem is neglecting the base effect of the occupational starting position. This severely affects the stability of the results, because career mobility is mainly observed from jobs with low qualification requirements. Moreover, we show that Sicherman’s indicator for upward career mobilty and similar indicators as moving to a higher status position are not valid indicators. When using relative wage growth (and this is the strategic variable underlying the theory of Sicherman and Galor) overeducated workers are found to experience clearly lower relative wage growth rates than correctly allocated workers; the contrary is the case for undereducated workers. This pattern of results severely jeopardizes the career mobility model and can be better explained using signaling and segmentation theories. Based on the well-known positive correlation of access to training and upward career mobility the plausibility of our results is supported by the finding that access to informal and formal on-the-job training is relatively worse for overeducated and better for undereducated workers.