Does Age Exacerbate the Gender-Wage Gap? New Method and Evidence From Germany, 1984–2014

ABSTRACT

Given theoretical premises, the gender-wage gap adjusted for individual characteristics is likely to vary according to age. This study adapts John DiNardo, Nicole M. Fortin, and Thomas Lemieux's (1996) semi-parametric technique to disentangle year, cohort, and age effects in adjusted gender-wage gaps. The study relies on a long panel of data from the German Socio-Economic Panel (SOEP) covering 1984–2015. The results indicate that, in Germany, the gender-wage gap increases over a birth cohort's lifetime, including in the post-reproductive age for some birth cohorts. The results suggest that age and gender are overlapping handicaps in the labor market and call for a policy intervention.

KEYWORDS:

• Gender-wage gap
• age
• cohort
• decomposition
• non-parametric estimates
• Germany

JEL Codes:

• J31
• J71

ACKNOWLEDGMENTS

We are grateful to the Editors and anonymous referees for inspiring comments. Earlier versions of this paper have benefited greatly from remarks by Magdalena Smyk, Gianna Giannelli, Karolina Goraus, Saul Estrin, Tomasz Mickiewicz, and participants of the ASSA, EEA, IAFFE, EACES, and WCCE conferences, as well as the GRAPE seminars. This research was supported by a grant from the National Science Centre (NSC), UMO-2012/05/E/HS4/01510. All opinions expressed are those of the authors and have not been endorsed by NSC. The remaining errors are ours.

SUPPLEMENTAL DATA

Supplemental data for this article can be accessed at https://doi.org/10.1080/13545701.2018.1503418.

Notes

1 In this paper, unless stated otherwise, the terms “gender-wage gap” and “gender differences in earnings” refer to the differential adjusted for individual characteristics. The unadjusted gap is identified by the term “raw.”

2 A similar argument applies to women's decisions on which career to follow. Anticipating interruptions, women might select the careers that impose a lower penalty for interruptions. For example, Claudia Goldin and Lawrence F. Katz (2008) show that women with more children tend to work in careers with lower wage penalties for interruptions.

3 Importantly, alternative individualist explanations for the persistent gender-wage gap even after accounting for individual characteristics – such as lower wage expectations of women (Blau and Ferber 1991; Reuben, Wiswall, and Zafar 2013) or taste-based discrimination (Becker 1971; Lang 1986; Duncan and Loretto 2004) – have no clear time/age-related patterns. These explanations are also weakly founded in feminist theories.

4 This approach finds some empirical support: Peter Kuhn and Kailing Shen (2013) analyze job ads from China and find that some job openings are subjected to strong bias toward young and attractive women. Moreover, firms tend to have a higher preference for hiring men when looking for older employees than when looking for employees in other age groups (Duncan and Loretto 2004; Lincoln and Allen 2004; Lauzen and Dozier 2005; Neumark, Burn, and Button 2015).

5 Charlotte Lauer (2000) and Elke Holst and Anne Busch (2009) explore the glass ceiling in wages by studying individuals in managerial positions, suggesting that the unexplained component within the top occupation is approximately 40 percent of men's wages, once selection bias is taken into account. David Reimer and Jette Schröder (2006) also explore the relation between the adjusted wage gap and the field of education, but they do so in a much more homogeneous population (university students). Their results indicate that the adjusted gap among former students was between 4.3 percent and 7.6 percent of women's wages. Though the value is much lower than in the previous case, it must be taken into account that the sample covers only individuals at the beginning of their careers. Doreen Triebe (2013) finds that while an increase in the salaries of women leads to reduced labor supply of men, as men's labor is subject to a higher shadow tax, which leads to lower marginal income. Women's response to wage increases by men differs. Namely, married women do reduce paid working hours because of tax splitting, but cohabiting women do not.

6 This legislation was changed in 2007. Benefits are not means tested, and they are proportional to previous earnings. Yet, the effects of this reform are to be observed only in the cohorts eligible between 2008 and 2015.

7 Though originally intended to measure the consequences of the changes in the unionization rate, it was adopted to measure the impacts of other variables as well, among them gender. Casey Warman, Frances Woolley, and Christopher Worswick (2010) use the DiNardo, Fortin, and Lemieux (1996) decomposition to measure the differences in earnings between university professors in different periods, from the early 1970s to the early 2000s. Their analysis bears some similarities with ours, as we also consider a time dimension. However, the most important difference is that, in our paper, we focus on the gender-wage gap at different ages. Eva M. Sierminska, Joachim R. Frick, and Markus M. Grabka (2010) also employ the DiNardo, Fortin, and Lemieux (1996) decomposition for Germany, using data from the SOEP to study the wealth gap, of which salary is just one of the components.

8 While data from East Germany are also available, the longitudinal dimension is substantially shorter. Moreover, the communist legacy and the process of economic transition suggest that trends in the gender-wage gap in East Germany might be driven by different factors.

9 Data correspond to the German national sample. Immigrants are not included in the analysis.

10 Although in Germany the minimum legal employment age is 15, in the most recent year, only 30 percent of young people entered the labor market before their 25th birthday. In 1984, however, this percentage was twice as high. Thus, analyzing the individuals under the age of 25 would have involved additional selection issues and educational choices. The employment ratio among individuals above the age of 60 remained below 10 percent during the entire sample period.

11 The literature typically relies on cross-sectional data and hence frequently utilizes parenthood or age of children as an exclusion restriction in estimating the selection bias. However, most of the men and women in our sample eventually have at least one child. Clearly, never-parents are not directly comparable to ever-parents.

12 Wages are taken in logarithms; hence we take 0.1€/CPI – that is, the real value in 2005 euros of 10 euro cents.

13 Another source of bias, particularly toward the upper tail of the age distribution, stems from the fact that if motivation to work for pay is age dependent and heterogeneous across individuals and genders – the preference argument – our estimates would be biased upward because only individuals motivated enough to work for pay will be observed in the employed sample past certain age thresholds. Thus, our estimate of the age pattern in the gender-wage gap may partially confuse pure age effects and – should they indeed be heterogeneous across age and genders – preferences.

14 The 2010 wave of the SOEP introduced a new wave of respondents to the panel. As the wave focused on families, this raised the average of couples living with children in the following years.

15 A full results output from each of the numerous probit models is available upon request.

16 The sample of respondents used to compare 1994 to 1989 is the same for both years; it might differ from the sample used to compare 1989 to 1984. This choice is motivated by panel attrition. If we require individuals to stay longer in the panel, up to thirty-two years, the sample size often drops below 150 observations. Estimates for those small groups are reported as a robustness check in Table D1 in the Supplemental Online Appendix.

Additional information

Notes on contributors

Joanna Tyrowicz

Joanna Tyrowicz is a co-founder of FAME|GRAPE, Professor at University of Warsaw, and Research Associate at IAAEU in Trier and at Institute of Labor Economics (IZA) in Bonn. Her research interests relate primarily to the economics of gender inequality, as well as longevity.

Lucas van der Velde

Lucas van der Velde is Assistant Professor in the statistics and demography department at the Warsaw School of Economics and Researcher at FAME|GRAPE. His research interests concern labor market inequality, with a particular emphasis on technological progress and economic transition.

Irene van Staveren

Irene van Staveren is Professor of Pluralist Development Economics at the International Institute of Social Studies (ISS) of Erasmus University Rotterdam. Her field of research includes feminist, heterodox, pluralist, and social economics. Professor van Staveren is on the editorial boards of the Journal of Economic Issues, Review of Social Economy, and Feminist Economics.