Within-job gender pay inequality in 15 countries

Women earn less than men in all advanced industrialized countries, and overall gender gaps are closely tied to the occupations and establishments where women and men work. Prior work argued that while overall wage differences are substantial, women and men doing the same work for the same employer receive very similar wages, implying that sorting into different jobs (and male- vs female-dominated occupations) is central to gender pay inequality. This understanding carries strong policy implications: if significant within-job pay differences exist, equal pay mandates are critical; if differences are primarily due to sorting, policies should target hiring and promotion practices and broader valuation of occupations. Yet most available data lack the detail to link individuals to employers and specific jobs. The best evidence on within-job gender pay differences is decades old and limited in country coverage. This study leverages contemporary linked employer-employee data from 15 countries to estimate within-establishment, within-occupation, and within-job gender differences in earnings, reassessing the balance between sorting processes and within-job inequality.

Foundational research (e.g., Petersen and colleagues; Blau and Kahn) emphasized that within-job gender pay differences are small and that sorting into occupations accounts for most of the gender wage gap. Classic studies using linked employer-employee data from the 1980s–1990s in a few countries found minimal within-job discrimination and a dominant role for occupational segregation. Broader literatures on job queues, comparable worth, and organizational opportunity structures highlight recruitment, segregation, and valuation as mechanisms shaping gender pay gaps. However, contemporary, cross-national evidence with detailed employer-employee linkages has been scarce, leaving uncertainty about current levels of within-job inequality versus sorting across establishments and occupations.

The study analyzes restricted-access linked employer-employee datasets from 15 countries (Canada, Czechia, Denmark, France, Germany, Hungary, Israel, Japan, the Netherlands, Norway, Slovenia, South Korea, Spain, Sweden, United States). Samples focus on workers aged 30–55 where applicable. The dependent variable is log earnings. Four sets of country-year ordinary least squares models are estimated separately by year: (1) a baseline model with basic individual-level covariates (gender, age, age-squared, education) and a fixed effect for full-time vs part-time status; (2) a within-establishment model adding fixed effects for establishment-by-full-time/part-time; (3) a within-occupation model adding fixed effects for occupation-by-full-time/part-time; and (4) a within-job model adding fixed effects for occupation-establishment-by-full-time/part-time (i.e., occupation-establishment units). Analytic samples are restricted to gender-integrated fixed-effect units in models with fixed effects. Coefficients on gender are interpreted as relative differences in geometric means of unlogged earnings (equivalently, differences in arithmetic means of logged earnings). Basic adjustment models control for age, age-squared, education, and full-time vs part-time status. Fixed-effects specifications isolate within-establishment, within-occupation, and within-job comparisons to assess the role of sorting into establishments and occupations versus pay differences among men and women doing the same work for the same employer. Trends are assessed by estimating these models across approximately a decade per country (beginning in 2005 where possible). Data processing and analyses were conducted using STATA (versions 14–17) and SAS (version 9). Table 2 summarizes data sources, years, occupational and establishment measures, education measures, unit of observation (job spells or person-years), and any sectoral omissions or irregularities for each country.

• After basic adjustments (age, education, full-time/part-time), the overall gender earnings gap among ages 30–55 ranges from about 10% in Hungary to 41% in South Korea (Table 1). All coefficients indicate women earn less (P<0.001).
• Within-job gender gaps are smaller but remain substantial, ranging from about 7% in Denmark and France to 26% in Japan.
• Within-job differences account for a sizable share of overall gaps: typically around half of the total. The proportion of the overall gap remaining within jobs ranges from roughly one-third in Israel (0.35) to more than nine-tenths in Hungary (0.96). Other examples include Japan (0.73), Spain (0.77), Slovenia (0.74), Germany (0.54), France (0.59), the United States (0.48), Denmark (0.40), the Netherlands (0.37), and Sweden (0.43).
• Sorting into both establishments and occupations makes important contributions to gender pay differences. Contrary to earlier work that emphasized occupational sorting over establishment sorting, both are influential across countries.
• Trends (circa 2005 to most recent year) show that in most countries both overall and within-job gender gaps have decreased over time. Exceptions include Central and Eastern European cases: in Czechia, within-job gaps decline but overall gaps increase (indicating rising sorting effects); in Hungary and Slovenia, overall gaps increase due to both greater sorting and rising within-job gaps.
• No country exhibits a decrease in the overall gender gap combined with an increase in within-job gaps, suggesting that sorting processes are rarely gender-egalitarian.
• In six countries with data on contractual hourly wage rates, sorting similarly explains wage differences, reinforcing the relevance of equal pay policies alongside policies targeting sorting processes.

The analyses demonstrate that both within-job pay differences and sorting into different establishments and occupations materially contribute to gender pay gaps across 15 countries. While historical consensus held that within-job inequality was negligible, current evidence shows it remains consequential—often about half of the overall gap persists when comparing men and women in the same occupation and establishment. At the same time, temporal trends indicate that sorting is becoming increasingly important while within-job differences are shrinking in many countries. Policy implications are twofold: enforcing equal pay for equal work remains necessary to address persistent within-job disparities, and reforms targeting hiring, promotion, and organizational job assignment processes are essential to counteract gendered sorting that sustains overall pay inequality.

Using recent linked employer-employee data from 15 countries, the study updates and broadens evidence on the sources of gender pay inequality. It shows that within-job pay differences remain meaningful in all countries studied, and that sorting into establishments and occupations also plays a major role—contrary to earlier conclusions that emphasized sorting almost exclusively. Trends suggest within-job gaps are shrinking while sorting effects are growing in importance in many contexts. Future research should continue to track these components over time, investigate mechanisms of sorting within and across firms, examine how institutional features (e.g., parental leave policies, part-time work prevalence, wage-setting institutions) shape within-job and sorting components, and assess the effectiveness of equal pay and organizational policies in reducing both within-job inequality and gendered sorting.

Cross-country comparability is challenged by data harmonization issues and contextual differences (e.g., parental leave policies, prevalence of part-time work, and the relevance of occupations and firms). Although the authors aligned analytic decisions across datasets, differences in gender regimes and labor market organization remain. Some countries lack certain measures (e.g., education), sectors are omitted in some datasets, and top earnings are imputed in a few cases. Robustness checks varying variable definitions, model specifications, and sample definitions suggest the main results are stable despite these limitations.