Gender differences in fairness evaluations of own earnings in 28 European countries

ABSTRACT

Women tend to evaluate their own pay more favorably than men. Contented women are speculated to not seek higher wages, thus the ‘paradox of the contented female worker’ may contribute to persistent gender pay differences. We extend the literature on gender differences in pay evaluations by investigating fairness evaluations of own earnings and underlying conceptions of fair earnings, providing a closer link to potential subsequent wage demands than previous literature. Using European Social Survey (2018/2019) data, we find no evidence that women evaluate their own earnings more favorably than men. In 15 out of the 28 analyzed countries, women actually report more intense levels of perceived unfairness. Studying fair markups on unfair earnings, i.e. the relative distance between the earnings received and earnings considered fair, we find that women report the same, if not lower, fair markups compared to men in most countries; thus indicating limited potential for perceived unfairness as a driving force to reduce the gender pay gap in Europe.

KEYWORDS:

• Gender
• fairness of earnings
• cross-country analysis
• European Social Survey

Acknowledgements

We are greatful for comments on earlier versions of this paper from members of the project ‘Perceptions of Inequalities and Justice in Europe’ (PIJE) as well as from Peter Valet, Alexandra Fedorets, and Mattis Beckmannshagen. Additionally, earlier versions of this work were presented at the SASE 2021 Virtual Conference of the Society for the Advancement of Socio-Economics as well as at the 2021 Virtual Annual Meetings of the American Sociological Association; we acknowledge helpful feedback from participants of both conferences as well as from anonymous reviewers and the editor of European Societies. Jule Adriaans received financial and intellectual support for this paper from the Socio-Economic Panel Study (SOEP) at DIW Berlin.

Disclosure statement

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

Data availability statement

All analyses rely on data from the ESS Round 9 (ESS Round 9 2018). The full data is available here: https://doi.org/10.21338/NSD-ESS9-2018. Full replication code of the analyses is available here https://osf.io/vynk4.

Notes

1 We exclude Cyprus from the analysis as no information on working hours is provided in the data and this information is central to our analysis.

2 Sample sizes and descriptive statistics by country are available in Online Appendix Table A.1.

3 All analyses are conducted in Stata 16. Full replication code is available here: https://osf.io/vynk4.

4 While our interest is in testing gender differences in the probability of being fairly paid, some concerns might emerge whether substantial gender differences exist between under- and over-rewarded respondents that might confound the probability of evaluating own earnings as unfair. Unfair over-reward is, however, rare, as highlighted in Figure 1, and, accordingly, we drop respondents who are unfairly over-rewarded from the analysis.

5 We code ISCED classes 1 and 2 as ‘low education;’ ‘medium education’ is coded if individuals belonged to ISCED classes 3, 4, or 5. High educated individuals are individuals with tertiary education.

6 We define as part-time any individual working less than 35 h during a usual working week.

7 We based our occupational classification on the 1-digit ISCO scheme. Due to the restricted sample size in some countries, we consider the following broad occupational classes: (a) managers and professionals; (b) technicians; (c) clerks; (d) sales and service workers; (e) skilled agriculture, craft workers, and plant and machine operators; and (f) elementary occupations.

8 Due to the limited sample size, we exclude industry controls, opting for a more parsimonious specification. Nevertheless, in ESS, NACE industry classification is available at the 2-digit level. We test the robustness of our results by including industry affiliation defined in 10 main classes. Results are invariant with inclusion of industry classes as further regressors. Results from this robustness check are available in Online Appendix Figures A.8.1 through A.8.4.

9 We exclude from the analysis pension and social benefits recipients, military and community services workers, as well as permanently sick or disabled individuals. We keep unemployed and out of the labor force individuals if not included in the excluded categories as support in order to control for sample selection biases.

10 Despite its great popularity, the Heckman correction has important limitations. First, it relies on the distributional assumption that the error terms in the first- and second-stage equations, ν_ic and ε_ic, are jointly normal and uncorrelated. Failure of this assumption affects the consistency of the estimator. Second, as explored in detail in Puhani (2000), collinearity between the IMR and the regressors in the outcome equation that might cause inefficient estimation is a concern. To solve these issues, inclusion of exclusion restrictions is highly recommended. However, validity of these exclusion restrictions is debatable. As explained above, we follow the literature, which relies on Mother14 and Child6 as established instruments. Mother14 is interacted with female.

11 In a few countries, at the very top of the fair markup distribution, we find markups higher than 1, meaning that the respondent declared Y^F double than Y^A.

12 Figure A.2 in the Online Appendix shows the country-specific distribution of the fair markups in our analysis sample.

13 We exclude those individuals who evaluate their earnings as unfairly high in order to avoid negative markups. In each country, these individuals are a small minority (below 2%) of the overall analysis sample.

14 We also include a dummy variable for presence of children in the responding household as a further control.

15 Figures A.3.2 and A.4.2 report the IMR coefficients from the second stage equation of the Heckman-selection for both specifications of fairness evaluations. Besides a few exceptions, IMRs are statistically insignificant in most countries, suggesting that the fairness evaluations are not systematically affected by selection issues. In line with this conclusion, we ran additional analyses without the Heckman correction. Results are robust and available in Online Appendix Figures A.9.1 and A.9.2.

16 Past research, however, highlights that women are often less successful in wage negotiations than men (Kolb 2009; Mazei et al. 2015; Sauer et al. 2021).

17 Additional analyses presented in Online Appendix Figure A.7 underline this conclusion. If the interaction between gender and intensity of perceived is included in the fair markup model, significant interaction effects in 15 countries indicate that the effect of intensity of unfairness on fair markup is significantly smaller for women.

18 At the time of writing, ESS round 10 data collection is ongoing; no data has been published. The questionnaire for ESS Round 10 data collection is available here: http://www.europeansocialsurvey.org/docs/round10/questionnaire/ESS10-Paper-Questionnaire-English-Template-FINAL_20210706.pdf (last accessed January 26, 2022).

19 Notable exceptions are Croatia and Lithuania, where women reported more intense unfairness and higher fair markups than men.

Additional information

Funding

This work was supported by the Leibniz Association (K248/2019) as part of the project ‘Perceptions of Inequalities and Justice in Europe’ (PIJE).

Notes on contributors

Jule Adriaans

Jule Adriaans is a researcher at the Chair of Social Inequality and Social Structure Analysis and a doctoral candidate at the Bielefeld Graduate School in History and Sociology (BGHS) at Bielefeld University. Her research focuses on the perception and evaluation of inequalities and justice with a special focus on a comparative European perspective.

Matteo Targa

Matteo Targa is a researcher at the Socio-Economic Panel Study (SOEP) and a doctoral candidate at the Graduate Center at DIW Berlin. His research focuses on labor economics with a special focus on job-related determinants of economic inequality.