Discriminating between alternative measures of over-education

Five over-education measures are evaluated empirically on the basis of encompassing tests. The measures are based on job analysis (JA), worker-assessment of the required level to do the job (WAd), worker-assessment of the required level to get the job (WAg), the mean educational level of realized matches (RMmn), and the modal level of realized matches (RMml). Over- and under-education are linked to wages, job satisfaction, mobility and training participation. For none of the outcome variables, the JA model is encompassed by another model. Given the risk on systematic errors, this is a sufficient condition to prefer a carefully conducted JA to any other measure. The most reliable solution is to use the JA measure as an instrument for the WAd measure.