Skill vs. education types of labour mismatch and their association with earnings

This work aims to determine whether the difference between skill and education-based measures of labour mismatch affects the estimates of the labour mismatch-earnings relationship as specified by Verdugo and Verdugo's (1989) version of over, required and undereducation (ORU) Mincer earnings function. The analysis employs crossectional data for 26 countries from the 1st Cycle of the OECD Survey of Adult Skills (PIAAC) conducted between 2011 and 2012. The preliminary results of the graphical analysis show that education and skill mismatch may exhibit opposite relationships with earnings at the country level. Specifically, over and under-education are found to be positively associated with median earnings, whereas over and under-skilling show a negative association. To investigate the source of the opposite correlations, an error components model is used. Additionally, the paper explores the heterogeneity in earnings and labour mismatch across a set of commonly used controls. The analysis produces mixed coefficient estimates for under and over-education but predominantly negative estimates for under and over-skilling at both individual and market levels. The market-level unobserved heterogeneity is found to be driving the coefficients away from zero. Although removing it often leads to a slight loss of magnitude, some exceptions exhibit a change of sign or loss of statistical significance. It is, thus, concluded that education and skill mismatch should be distinguished both conceptually and empirically and, if used as a proxy for each other, are unlikely to produce accurate results in the analysis of the Mincer earnings function.