Accounting for dropout risk and upgrading in educational choices: new evidence for lifetime returns in Germany

ABSTRACT

We analyse the economic returns in lifetime labour income of various educational paths in Germany. Using recent data, we calculate cumulative labour earnings at different ages and for different educational paths while controlling the parental background of individuals. We find that after the age of 55, lifetime labour income is higher for individuals with a university degree compared to individuals with a vocational degree. Considering the risk of dropout and the possibility of educational upgrading, individuals who start with a vocational training after their school degree do not earn less than individuals who start with university studies.

KEYWORDS:

• Returns to education
• lifetime income
• educational upgrading
• college dropout
• apprenticeship training
• NEPS-SC6-ADIAB

JEL CODES:

• I21
• I26
• J17
• J24

Acknowledgements

We thank two anonymous referees, Colin Green, Bernhard Boockmann, Lukas Fervers and Martin Kroczek for helpful comments and suggestions. Marit Holler, Manuel Schick and Susanne Vögele provided excellent research assistance. The authors thank the Baden-Württemberg Chamber of Commerce and Industry for financial support. An early version of this study has been published as Zühlke et al. (2020). We have to thank the employees at the FDZ at the IAB for data access and further support.

Disclosure statement

No potential conflict of interest was reported by the author(s).

Notes

1 Similar to high school, graduating one acquires the Abitur, the prerequisite to be able to attend university.

2 This is necessary to be able to calculate the logarithm of earnings.

3 The majority of the increase in observation time is due to the inclusion of time spent in education. This is an important advantage over analyses using administrative data only, such as Seckler (2021). For a more detailed analysis of the NEPS-SC6-ADIAB data, see Zühlke et al. (2020).

4 Gauly et al. (2020) show that survey data is not trustable when measuring earnings.

5 We also observe a number of individuals who never start any educational episode (according to our definition and the one by the NEPS-SC6).

6 We also have to reconcile the data from some inconsistent educational biographies and remove data on degrees obtained abroad.

7 We come to similar effects in terms of significance and the same conclusions when using the nominal labour income and Poisson regression instead of a log-linear model (see Santos Silva and Tenreyro 2006, for reasons not to use the logarithm in data with excess zeros).

8 We exclude, for example, individuals who enter the labour market very late or leave it very early in life and individuals with low absolute and relative numbers of non-zero income episodes. In contrast to, e.g., Seckler (2021), who works with SIAB data only, this affects only some people, to be precise 1,105 out of 11,467 observations.

9 This means that birth cohort fixed effects can be used for early cumulative lifetime earnings only. Note also that we cannot observe some early cohorts in the IEB before 1975 for West Germany and most cohorts before 1991 for East Germany. We can, however, observe their educational biographies in the NEPS-SC6.

10 We find the same results, when using birth year fixed effects as a robustness check.

11 International Socio-Economic Index of Occupational Status (ISEI) Score based on occupational codes by the method of Ganzeboom, de Graaf, and Treiman (1992).

12 For technical reasons, we also control for the number of years we can observe each individual in the data and for old birth cohorts, for whom we do not have non-zero income information at the beginning of their working lives. An overview of the control variables used can be found in Table A.2 in the Appendix. We further provide an overview on how the observable characteristics are distributed between individuals with different educational decisions together with the respective mean lifetime income in Table A.1.

13 A detailed overview of cumulated labour incomes can be found in Zühlke et al. (2020)

14 The results in Figures 3 and 4 are presented in more detail in Table A.3 in the Appendix. Although we use different sample sizes at each age segment, an attrition analysis shows similar results.

15 Estimated returns describe the difference between the average wages of individuals with a university degree compared to individuals with a vocational education.

16 An interesting note is that craftsmen and technicians have a lower educational background compared to university graduates. However, there is positive selection on other observable characteristics, such as gender and nationality.

17 Vocational training entails both positive and negative risks, i.e., upward mobility and dropout. The dropout risk is, however, much smaller than for university studies. Only one in twenty (5%) individuals who start vocational training end up with no degree at all, whereas this figure is almost threefold higher (13%) for university students, with an additional one in ten (8%) university students not gaining a university degree but only a vocational degree.