Decomposing the impacts of overeducation and overskilling on earnings and job satisfaction: an analysis using REFLEX data

Abstract

This article assesses the extent to which the impact of overeducation and overskilling on labour market outcomes such as earnings and job satisfaction relate to mismatches in particular competency areas. The analysis uses REFLEX data, which collects information about 19 key competence areas related to job performance. We find that the penalties to both forms of mismatch are insensitive to the inclusion of controls for overskilling in a wide range of job-specific competencies. The research suggests that the problem of mismatch relates to an inability to fully utilise general or innate ability as opposed to specific areas of acquired learning. We conclude that the problem of mismatch can only be effectively addressed by raising general levels of job quality within developed labour markets.

Keywords:

• overeducation
• overskilling
• job satisfaction
• earnings

JEL Classifications:

• J24
• J31

Notes

1 The countries included in the analysis are Italy, Spain, France, Austria, Germany, the Netherlands, UK, Finland, Norway, Chez Republic, Portugal, Belgium and Estonia.

2 There are also data for Japan which we exclude in order to focus on European countries.

3 Where 1 was not at all and 5 to a very high extent.

4 We tested the sensitivity of our analysis to variations of this definition and found that our results remained largely unchanged.

5 The controls are not reported in the paper; however, detailed results are available from the authors on request.

6 We check four country groupings: Central Europe countries: Austria, Germany, France, the Netherlands and Belgium; East Europe countries: Check Republic and Estonia; Nordic countries: Finland, Norway and UK; and finally Mediterranean countries: Portugal, Spain and Italy.

7 Results available from the authors.

8 Results available on request.

9 A KMO of above 0.5 is generally considered desirable for PCA.

10 According to Kaiser's rule only factors with eigenvalues greater than 1 is retained, but in our case, this criterion is not useful because only 1 factor achieves this value. So, we try retaining 3, 4, 5, and 6 factors and conclude that 4 factors are the most appropriate for our analysis considering both the explained variance and the interpretable results.

11 Varimax rotation is the most popular rotation method. Formally Varimax searches for a rotation of the original factors such that maximises the variance of the loadings.

12 Scoring coefficients could be available on request.