Gender compositions of occupations and firms jointly shape switches from gender-atypical towards more gender-typical positions

ABSTRACT

Research on sex segregation in the labor market has repeatedly found that women and men are more likely to exit from occupations and firms in which they are the numerical minority and subsequently seek positions that are more represented by their gender. However, this research looked at mobility either across occupations or across firms, leaving unclear how the simultaneous exposure to gender compositions of occupations and firms shapes attrition from gender-atypical positions. We draw on linked employer-employee register data from the German social security insurance system (SIEED, 2012-2018) to highlight that some occupations can be found in firms with various gender compositions, indicating that gender compositions of occupations and firms do not always align and thereby may independently affect mobility. Conditional relative risk ratios for mobility between gender-typed occupations and firms reveal substantial switches from gender-atypical towards more gender-typical positions. This gendered labor market mobility is most pronounced for men across occupations. For women, gender compositions of firms drive not only mobility across firms but also switches out of gender-atypical occupations. Our findings underscore that gender compositions of occupations and firms jointly shape attrition from gender-atypical positions, which ultimately perpetuates labor market segregation.

KEYWORDS:

• Sex segregation
• gender
• occupations
• firms
• labor market mobility

Acknowledgments

We thank Laura Lükemann, Maximilian Sprengholz, the participants of the workshop Longitudinal Research of Income Inequality at Haifa University, and three anonymous reviewers for feedback and advice on this article.

Disclosure statement

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

Notes

1 RRRs are calculated as non-linear combinations of the predicted probabilities for switching the gender-type occupation (firm) from seemingly unrelated linear probability models. The results do not change substantially when predicting the probabilities via logistic regressions or obtaining the RRRs from log-binomial regressions.

2 Daily wages compromise the product of hourly wages and daily working hours; randomly selecting one of the co-occurring spells leads to the same results.

3 When calculating the share of female employees, we could only apply the sample restrictions for the primary sampled firms, because the necessary information is unavailable for firms where employees work prior or posterior. However, as the correlation between the restricted sample- and unrestricted sample-based share of female employees across primary sampled firms is 0.959, this should not affect our results.

4 While the 5th percentile across occupations denotes a cell-size of 59 observations per occupation and the median entails 699 observations, we observe fewer than 10 observations for six occupations; nevertheless, due to their small sample size, they marginally affect our estimations.

5 Only for men, the RRRs of switches from gender-atypical compared to gender-typical occupations are substantially bigger for the 33.3/66.7 percent and smaller for the 20/80 percent cut-offs, which is based on mobility between gender-typical and integrated occupations across occupations with 20 to 33.3 percent female employees; men’s risk of switching from gender-atypical towards more typical occupations stays rather constant across cut-offs.

6 Although the Dissimilarity Index has been critiqued for insufficiently accounting for marginal changes (changing gender gaps in labor force participation or (dis-)appearing occupations; e.g., Charles and Grusky 2004), our illustration does not invoke any assumptions about marginal changes.

7 Notably, integrated positions are positioned in between the extremes, with the risks for switches out of integrated occupations and firms being also greater than switches out of gender-typical positions (Figure S.1 in the Supplementary Materials), indicating that women and men in positions of mixed gender compositions are still seeking positions that are even more represented by their gender (see also mobility out of integrated occupations in Table 1).

8 Promotions are operationalized as switches from not working in a supervisory position in a given year $t$, but doing so in the following year $t+1$. Supervisory positions are identified via the fourth digit of the KldB2010-codes (supervisory positions task within an occupational group) and the codes 7110 (Managing directors and executive board members), 7121 (Legislators), and 7122 (Senior officials of special interest organizations) among those with high complex tasks (cf. Eisenmenger et al. 2014). Although this measure of promotions in registry data comes with measurement error that seems to be larger for men (Collischon 2023), it is similarly to the measure of occupational gender compositions based on the KldB-4-digits occupational codes and thus most directly speaks to promotions as a mechanism for men’s change of their occupational gender-type.