# 5. Is it where you work, what you do, or what you get? Unpacking the gender wage gap and its evolution over the life-course

To provide a better understanding of the gender wage gap, it is important to take account of two key stylised facts. First, a significant gender wage gap persists even among similarly-skilled women and men in all OECD countries (Goldin, 2014[1]; Blau and Kahn, 2017[2]). Indeed, the gender wage gap tends be larger and more persistent when focusing on similarly-skilled women and men than when focusing on the raw gender wage gap. This reflects the fact that the gender gap in education has largely closed and that in many countries young women now have higher levels of education than men on average. Second, the gender wage gap tends to increase over the life-course in most countries. This is likely to reflect to an important extent the role of childbirth in shaping the career progression of women across occupations and firms (Kleven et al., 2019[3]; OECD, 2018[4]). Both stylised facts suggest that a better understanding of the gender wage gap requires paying more attention to the firms in which women and men work (“where they work”), their tasks and responsibilities in those firms (“what they do”) and the way they are rewarded for them (“what they get”). Since the importance of these elements depends in part on the role of childbirth, a life-cycle perspective is needed that allows following women and men across jobs and firms throughout their careers.

The objective of this chapter is to contribute to a better understanding of the gender wage gap over women’s professional career by focusing on the gap in pay between women and men with equivalent skills - defined in terms of their level of education and potential experience - within firms and between firms at each age. The between-firm component captures the role of differences in wage premia (“wage-setting practices”) between firms in the gender wage gap due to the sorting of women into low-wage firms. The within-firm component captures differences in pay between women and men within firms related to differences in tasks and responsibilities, or differences in pay for work of equal value (e.g. bargaining and discrimination). Previous studies have typically found mixed results with respect to the importance of pay differences within and between firms and their evolution over the working life – see a.o. Masso, Merikull and Vahter (2020[5]) for Estonia, Coudin et al. (2018[6]) for France, Casarico and Lattenzio (2019[7]) for Italy, Card et al (2016[8]) for Portugal, Bruns (2019[9]) for Germany, Goldin et al (2017[10]) and Barth et al. (2021[11]) for the United States. It is a priori unclear to what extent these differences are genuine or reflect differences in data treatment and empirical methodology.

The chapter provides new cross-country evidence on the gender pay gap within and between firms at each age based on a harmonised approach using linked employer-employee data for sixteen OECD countries.2 It is shown that, on average across the countries covered by the analysis, the bulk of the gender wage gap between similarly skilled women and men is related to pay differences within firms due to differences in tasks and responsibilities (“what you do”), but that differences in firm wage-setting practices between firms (“where you work”) and differences in pay for work of equal within the same firm (“what you get”) also can play a significant role. The gender wage gap within and between firms tends to increase over the working life due to the role of motherhood. This reflects to an important extent unequal opportunities for upward mobility between and within firms and the effect of career breaks at the age of childbirth on the career progression of women. Consequently, bringing down the gender wage gap crucially requires policies that promote opportunities for career advancement within and between firms, while policies that tackle pay discrimination also have a role to play, particularly in some countries.

The remainder of this chapter is structured as follows. Section 2 presents a number of key stylised facts to motivate the analysis. Section 3 lays out the conceptual and empirical framework that will be used to analyse the gender wage gap between similarly-skilled women and men within and between firms over the working life. Section 4 provides the results. Section 5 concludes with a discussion of the policy implications.

A better understanding of the gender wage gap requires taking account of two important stylised facts: i) the presence of large and persistent wage gaps between similarly qualified women and men; ii) the tendency for the wage gap to increase over the working life.

On average across European countries, women earned about 10% less per hour than men in 2018 (Figure 5.1, Panel A). Controlling for skills - in terms of education and potential experience – tends to increase the gender wage gap to about 15% as working women tend to be better educated than working men on average (18% for the European countries analysed in Section 4).3,4 The gender wage gap therefore reflects differences in the firms for which women and men work, differences in tasks and responsibilities in those firms and differences in pay for work of equal value rather than differences in their skills. Controlling for skills also renders the gender wage gap more persistent, as educational attainment has increased more quickly for women than for men (Panel B). The raw gender wage gap declined by about 35% between 2002 and 2018 on average across countries (from almost 16% to 10%), while the conditional gender wage gap that controls for differences in skills declined by less than 15% over the same period (from 17% to 15%).

In most countries, the gender wage gap increases over women’s professional careers as the wages of women tend to grow less quickly than those of men (Figure 5.2).5 This points to the presence of barriers to upward mobility related to the scope for learning on the job, promotions to better jobs in the same firm, or job switches to better paying firms. To a large extent, the more limited scope for career advancement for women is likely to reflect the role of childbirth and the uneven distribution of household responsibilities between mothers and fathers (Kleven et al., 2019[3]). This may have important implications for the incentives of mothers to look for jobs in firms offering short and/or flexible working hours and those of firms, where long and unpredictable working hours are common, to disproportionately hire men (Goldin, 2014[1]; OECD, 2017[12]; OECD, 2019[13]). This not only reduces the scope for upward mobility to better paying firms and occupations of women, but also potentially weakens their bargaining position by reducing the number of alternative job opportunities.

This section describes the conceptual framework for analysing the wage gap between women and men with equivalent skills within and between firms, the empirical approach and the data that will be used.

The gender wage gap at each age can be decomposed into a between-firm and a within-firm component. The between-firm component captures the role of differences in firm wage premia (or “wage-setting practices”) between firms in the gender wage gap between similarly-skilled women and men due to the sorting of women into low-wage firms and industries. Firm wage premia refer to the component of wages that is determined by the characteristics of the firm and not the characteristics of workers. The within-firm component captures differences in pay between similarly-skilled women and men within firms related to either differences in tasks and responsibilities or differences in pay for work of equal value. The present analysis abstracts from the role of skill composition, which can also contribute to differences in average pay between firms as well as differences in pay by gender within firms. Public policies can help to reduce gender wage gaps between similarly-skilled women and men within firms as well as those arising from the sorting of women in low-wage firms. A schematic representation of the decomposition is presented in Figure 5.4.

Differences in wage premia between firms contribute to the gender wage gap when women are more likely to be employed in low-wage firms. Descriptive evidence in Chapter 2 already showed that, despite recent progress, women are still much more likely to be employed in low-wage firms than men. Differences in the gender composition across firms, in principle, may reflect demand-side factors, due to discriminatory hiring practices by employers (see below) or supply-side factors, due to the preferences of women to work in certain economic activities (industries), the skills these activities require or the way they are organised. However, this does not explain why women are more likely to work in low-wage firms, even within narrowly-defined economic activities. Gender complementarities in production provide one such explanation (Goldin, 2014[1]). To the extent that women are less available for jobs requiring long or unpredictable working hours, due to unequal sharing of household responsibilities, and such jobs are more common in high-wage firms, men are more likely to be hired in high-wage firms.6

Differences in pay within firms contribute to the gender wage gap when women and men with equivalent skills are rewarded differently within the same firm. Systematic differences in pay between similarly-skilled women and men within firms reflect differences in tasks and responsibilities, which may result from unequal opportunities for career progression, or differences in pay for work of equal value, which may result from bargaining or discrimination by employers. Employer discrimination may be rooted in conscious or unconscious biases against women (“taste-based discrimination”), may reflect the perceptions of employers that women are on average less productive than men (“statistical discrimination”) or be based on profit considerations by paying lower wages to women with limited outside options and a weak bargaining position (monopsonistic discrimination).7 Unequal opportunities for career progression within firms may result from a broad range of factors, including discrimination in hires and promotions by employers and the individual circumstances of women and men, related to the unequal sharing of family responsibilities, that may constrain career choices and shape preferences for non-wage working conditions (e.g. flexible hours, short commuting times).

To empirically implement the within and between-firm decomposition of the gender wage gap between similarly-skilled women and men, this chapter builds on Goldin et al. (2017[10]) and Card, Cardoso and Klein (2016[8]). This involves in a first step estimating wage equations with flexible earnings-experience profiles by gender both without and with firm-fixed effects and, in a second step, separately estimating for women and men a wage equation with firm-fixed effects (see Box 5.2 for details).8 The specification without firm-fixed effects allows documenting the overall gender wage gap at any age conditional on worker characteristics (education). The specification with firm-fixed effects allows documenting the gender wage gap within firms at any age conditional on worker characteristics, while the difference in the gender wage gap between the two specifications captures the between-firm component of the gender wage gap due to the sorting of women and men across firms paying different wage premia. The gender-specific wage equations with (gender-specific) firm-fixed effects allow providing an indication of the role of bargaining and discrimination by comparing the firm-fixed effects for women and men within the same firm.

All specifications control for cohort effects and selection into employment based on observable worker characteristics. Selection effects into employment may induce differences in the composition of employment of women and men, with potentially important consequences for cross-country comparisons of the gender wage gap and its over evolution with age (Olivetti and Petrongolo, 2008[14]). Cohort effects can affect the age profile of the gender wage gap when the composition of women and men in employment varies across birth cohorts due to, for example, gradual improvements in educational attainment of women relative to men or rising female labour force participation. The present analysis controls for cohort effects through the inclusion of decade-of-birth fixed effects by gender (Barth, Kerr and Olivetti, 2021[11])9 and for selection into employment by focusing on women and men with observationally equivalent skills (conditional on flexible earnings-experience profiles by education and gender). To get a sense of the possible role of selection on the unobservable characteristics of workers, the evolution of the gender gap with age is analysed after including worker fixed effects as a robustness check (Abowd, Kramarz and Margolis, 1999[15]; Dostie et al., 2020[16]).

Previous studies have typically found mixed results for the role of differences in pay within and between firms in the wage gap between similarly-skilled women and men as well as their evolution over the working life – see a.o. Masso, Merikull and Vahter (2020[5]) for Estonia, Coudin et al. (2018[6]) for France, Casarico and Lattenzio (2019[7]) for Italy, Card et al (2016[8]) for Portugal, Bruns (2019[9]) for Germany, Jewell et al for the UK (2019[17]), Goldin et al (2017[10]) and Barth et al. (2021[11]) for the United States. All these studies except that by Goldin et al (2017[10]) focus on the role of sorting on the one hand and bargaining and discrimination on the other in the gender wage gap, while abstracting from differences in tasks and responsibilities. Evidence for France and Germany suggests that the role of firms exclusively reflects differences in wage premia between firms due to differential sorting, whereas evidence for Estonia and to a lesser extent also Portugal suggests that bargaining and discrimination are also important. Moreover, the role of sorting in the gender pay gap tends to increase over the working life and particularly with childbirth in some countries (e.g. Germany), but not in others (e.g. Estonia). It is a priori unclear to what extent these differences are genuine or reflect differences in data treatment and empirical methodology.

The decomposition of the gender wage gap within and between firms is implemented based on a harmonised data treatment and methodology using linked employer-employee data for 16 countries (Austria, Costa Rica, Denmark, Estonia, Finland, France, Germany, Hungary, Italy, Japan, Netherlands, Portugal, the Slovak Republic, Spain, Sweden and the United Kingdom) (see the standalone Data Annex for details). The linked employer-employee data used in this chapter are mostly based on administrative records designed for tax or social security purposes, and consequently tend to be very comprehensive (covering the entire population of workers and firms in most countries) and of very high quality, notably with respect to information on wages, given the potentially important financial or legal implications of reporting errors and extensive administrative procedures for quality control. Importantly, these data allow measuring gender wage gaps with great precision, decomposing them within and between firms and analysing what the determinants of wage and employments gaps within individual firms.

A limitation of these data, particularly in a cross-country context, is that wage definitions may differ, notably due to differences in the availability of information on working time (see the standalone Data Annex for details). To allow for meaningful comparisons of the gender wage gap across the largest number of countries the gender wage gap refers to hourly wages (including bonuses and overtime payments) where available and monthly earnings otherwise, adjusted for gender differences in working time using external data sources (this concerns Austria, Costa Rica, Estonia, Finland and the Slovak Republic).10 Consequently, cross-country differences in the overall gender wage gap should not be driven by differences between women and men in working time. For countries for which no information on working time is available, it is further assumed that differences in the between and within components of the gender wage gap or their evolution over the life-course are not influenced by differences in working time. The analysis is restricted to individuals aged 25-60 and excludes workers in mini-jobs earning less than 20% of the full-time minimum wage or, if no minimum wage exists, 10% of the median, as well as all firms that do not employ at least one woman and one man.11

The resulting gender wage gaps are generally close to those contained in the OECD Earnings Distribution database (Annex Figure 5.A.1).12 If anything, there is a tendency for the gender wage gap to be higher in the present data. This reflects in part the focus of the OECD Earnings Distribution database on full-time workers, whereas the measure of the gender wage gap used in this chapter also comprises part-time workers, which typically receive lower wages and are more likely to be female. Another difference is that the OECD Earnings Distribution focuses on differences in base wages and does not take account of bonuses and overtime payments, which tend to be more important for men, understating the true gender wage gap.13

This section discusses, respectively, differences in the gender wage gap within and between firms across countries, its evolution over women’s professional careers and the implications of motherhood by focusing on the role of career breaks for the wage growth of women.

On average across the sixteen countries considered here, the gender wage gap between similarly skilled women and men amounted to 22% over the period from the mid-2000s and the mid-2010s. As noted previously, this is somewhat higher than that reported in the OECD Earnings Distribution database because the gender wage gap is measured here in terms of the mean rather than the median, is not limited to full-time workers and includes overtime pay and bonuses. The gender wage gap is particularly high in Japan. This may be related to the tendency of women to move into lower paid non-regular or part-time jobs after returning from maternity leave. By contrast, the gender wage gap is particularly low in Costa Rica and Sweden. While in the case of Sweden, this is likely to reflect relatively low levels of gender inequality in the labour market in general, in Costa Rica this is at least in part likely reflect the above-average gap in labour force participation and the fact that women tend to be positively selected in employment, even after controlling for differences in education and age (experience). The wage gap between women and men with similar skills is decomposed below into the components related to sorting of women into low-wage firms, differences in tasks and responsibilities within firms, and differences in pay for equal work within firms due to amongst others discrimination.

On average across the countries covered by the analysis, one-quarter (23%) of the wage gap between women and men with similar skills reflects the sorting of women in low-wage firms (Figure 5.5).14 More than half of the gender wage gap between firms is due to the sorting of women into low-wage firms within sectors (55%). Differences in firm size do not play much of a role, with the exception of the three Central and Eastern European countries (Estonia, Hungary and the Slovak Republic). This suggests that in general women are less likely to work in high-productivity high-wage firms irrespective of their size and sector.15 The remaining part of the gender wage gap between firms is due to the tendency of women to work into low-wage sectors (45%).16 The sorting of women into low-wage firms may reflect differences in non-wage working conditions, as women may be constrained to opt for firms with flexible working time arrangements due to childcare responsibilities and unpaid homework as well as discriminatory hiring practices by employers (Box 5.3). The sorting into low-wage sectors also is likely to reflect the tendency of women to sort into economic activities that are compatible with their past educational choices (e.g. privileging literacy over mathematical skills), driven by gendered socialisation processes earlier in life, stereotypes and social norms (OECD, 2017[18]).

About three quarters (77%) of the wage gap between women and men with similar skills reflects differences in pay within firms. The within-firm gender wage gap is likely to reflect to an important extent differences in tasks and responsibilities, but also differences in pay for equal work due to amongst others discrimination. Using the detailed decomposition as proposed by Card, Cardoso and Kline (2016[8]), Figure 5.7 shows that, on average across the countries considered, the bulk of the wage gap between women and men with similar skills (eight-nineths) reflects differences in the work they do (e.g. tasks, responsibilities) and one-nineth differences in pay for work of equal value (e.g. bargaining, discrimination).18 Moreover, differences in pay for work of equal value are particularly large in some countries, such as Estonia and to a lesser extent Portugal, where it explains respectively 25% and 13% of the gender wage gap between similarly skilled women and men, while it tends to be rather small in the other countries. This pattern is consistent with results from previous studies for those countries (Masso, Meriküll and Vahter, 2020[5]; Card, Cardoso and Kline, 2016[8]; Coudin, Maillard and Tô, 2018[6]; Bruns, 2019[9]; Casarico et al., 2019[7]). Additional evidence for Estonia suggests that differences in pay between firms and differences in pay for work of equal value within firms tend be more pronounced for high-wage women (Box 5.4).

One reason why significant pay differences persist between women and men doing work of equal value in some countries may be that individuals and the broader public often are not aware of such differences. A number of countries have recently introduced pay transparency reforms to raise awareness of systematic pay differences between women and men within firms and make it easier to enforce equal pay legislation.19 Evaluations of mandatory disclosure or reporting measures in Canada, Denmark and the United Kingdom highlight their potential for narrowing the gender wage gap within firms (Baker et al., 2019[20]; Duchini et al., 2020[21]; Bennedsen et al., 2019[22]). However, not all pay transparency measures are associated with positive evaluations, suggesting that positive outcomes do not come automatically and a good design is key (Böheim and Gust, 2021[23]; Gulyas, Seitz and Sinha, 2021[24]). For an in-depth discussion of the pay transparency measures and their effectiveness, see OECD (2021[25]).

The age profile of the gender wage gap and the firms varies across countries (Figure 5.9).20 In a number of Western European countries, including France, Germany, Italy, the Netherlands, Portugal, Spain and Sweden as well as Japan, the gender wage gap tends to increase with age (Panel A – countries with an increasing age profile of the gender wage gap). This tends to reflect growing differences in pay both between and within firms. It may indicate that men increasingly sort into high-wage jobs as they advance in their careers, while women stay behind or may even be constrained to move into lower-wage jobs which offer more flexible working time arrangements. In Central and Eastern European countries, including Austria, Estonia, Finland, Hungary and the Slovak Republic, as well as the United Kingdom, the gender wage gap increases between the ages of 25 and 35, but then declines (Panel B – countries with a U-shaped age profile of the gender wage gap).21 This pattern is mainly driven by differences in pay between women and men within firms, while the role of between-firm differences varies across countries. In Denmark and Costa Rica, the gender wage gap is broadly stable until the age 45 –with only a tiny increase in the mid-thirties – and a more significant decline thereafter (Panel C – stable or declining age profiles). This is largely driven by the within-firm component of the gender wage gap.22

The evolution of the gender wage gap within and between firms over the working life is unlikely to be driven by changes in the characteristics of women and men in the workforce in the form of cohort or selection effects. As discussed in Section 5.3.2, the present analysis controls for cohort effects through the inclusion of decade-of-birth fixed effects by gender and controls for selection into employment based on the observable characteristics of women and men (education, age). However, the analysis does not control for selection in employment based on unobservable worker characteristics. To the extent that female labour force participation displays an important life-cycle profile due to the role of motherhood, this could contribute to the age-profile of the gender wage gap. Indeed, and as discussed in the next sub-section, career breaks around the time of childbirth are particularly important in Central and Eastern countries (OECD, 2018[4]; Kleven et al., 2019[3]). However, controlling for possible selection effects due to changes labour force participation over the life-cycle through the inclusion of workers fixed effects as in Abowd et al (1999[15]) and Dostie et al. (2020[16]) does not change the qualitative results presented here (see Annex Figure 5.A.2).

Importantly, in all countries except Costa Rica and Denmark, the gender wage gap increases significantly during the initial phase of professional careers up to the age of 35. This corresponds to a period characterised by both high wage growth and high job mobility in which the long-term careers of women and men are shaped (OECD, 2015[26]; Guvenen et al., 2021[27]; OECD, 2015[26]). However, women may miss out on important opportunities during this period since this also tends to be the period during which many women get their first child. Indeed, motherhood is likely to explain much of the divergence in the gender wage gap due to its implications for career advancement within and between firms (Kleven et al., 2019[28]; OECD, 2017[18]; Barth, Kerr and Olivetti, 2021[11]). The remainder of this sub-section analyses the role of promotions (upward job mobility) for the evolution of the gender wage gap within firms and that of job-to-job mobility for the evolution of the gender wage gap between firms.

Differences between women and men in the probability of being promoted shape the evolution of the gender wage gap within firms (Figure 5.10).23 Promotions are analysed by focusing on the probability of experiencing a significant increase in pay (more than 10%). On average across countries, women are less likely to be promoted at any age, but particularly in their thirties (Panel A). The gender gap in promotions largely reflects the role of part-time work, which is associated with a considerably lower probability of being promoted than full-time work (Panel B).24 Since women are more likely to work part-time, this contributes to the gender gap in the probability of being promoted.25 Conditional on being promoted, women tend to experience similar or slightly higher higher wage increases than men (Panel C).26 On average across countries, gender differences in the incidence and nature of promotions account for an increase in the gender wage gap within firms of 5 percentage points at age 45, or about 75% of the overall increase of the gender wage gap (Panel D).27 These findings are similar to those by Bronson and Thoursie (2020[29]) who find that promotions account for 70% of the increase in the gender wage gap by age 45 in Sweden.

The increase in the gender wage gap between firms over the life course is to an important extent driven by gender differences in the incidence and nature of job mobility. Women are less likely to change firms than men, particularly around the age of childbirth (Figure 5.10, Panel A). On average across the countries considered, the gap in job mobility increases up to the early thirties and then gradually narrows until it closes in the late 40s. Moreover, when women change firms, this is less likely to take the form of promotions, i.e. significant wage increases of more than 10% (Panel B). Since promotions account for a smaller share of overall job moves for women, this results in smaller average increases in firm wage premia (Panel C). In other words, women appear to change jobs to a lesser extent for wage and career considerations and more often for personal reasons, such as having more flexible working-time arrangements, working closely from home or following a partner. Indeed, it is the nature of job moves that explains most of the increase in the gender gap between firms over the working life, while the number of job moves plays a secondary role (Panel D). On average across the countries considered, gender differences in the incidence and nature of job mobility account for about 56 of the increase of the between-firm gender wage gap up to age 45 (20% of the overall gender wage gap), with gender differences in the change in wage premia following a job move accounting for about 80% of the total effect (wage effect) and gender differences in the probability of moving for 20% (quantity effect).

Beyond the direct effects of job mobility for the gender wage gap between firms, job mobility also has potentially important indirect effects for the gender wage gap within firms. Indeed, the lower level of job mobility among women and the greater importance of non-wage working conditions for job-mobility decisions results in a lower sensitivity of female labour supply to wage differences between firms. This increases the scope for gender discrimination based on differences in the bargaining position between women and men within the same firms (consistent with the analysis in Chapter 3). The fact that female job mobility is particularly unresponsive to wage differences between firms around the age of childbirth (early thirties) makes young mothers particularly vulnerable to discrimination by employers, in relation to both their wages as well as their probability of being hired. This issue has not received much attention in the policy debate so far.28

Systematic gender differences in the extent and nature of job mobility between and within firms reflect important differences in opportunities for career advancement between women and men. Policies and institutions that can support the upward mobility of women within and between firms are therefore key to reduce the gender wage gap. These include family policies that contribute to a more equal sharing of household responsibilities (e.g. incentivising fathers to take more parental leave) as well as a more equal sharing of part-time work (e.g. universal childcare, reducing effective marginal tax rates on second earners) (OECD, 2017[12]; OECD, 2019[13]).

The evolution of the gender wage gap between and within firms over the life-course is to an important extent determined by the motherhood penalty, i.e. the shortfall in wage growth following childbirth of mothers relative to fathers (OECD, 2017[18]). Kleven et al (2019[3]) provide estimates for selected countries of the long-term motherhood penalty in terms of labour income ranging from 21-26% in Denmark and Sweden to 31-44% in the United Kingdom and the United States and 51-61% in Austria and Germany. The motherhood penalty mainly reflects adjustments in working time and wages in the Scandinavian and German-speaking countries, but adjustments in employment in the two English-speaking countries. Bruns (2019[9]) further shows that about a quarter of the long-term wage penalty associated with motherhood in Germany results from differences in the sorting of women and men across firms. Similarly, Coudin et al (2018[6]) find for France the motherhood penalty in wages is closely related to the tendency of young mothers to move to firms close to home and firms with flexible working-time policies. Masso et al (2020[5]) suggest that sorting across firms plays no role in Estonia despite a significant motherhood penalty.

The present cross-country data do not allow looking at the role of motherhood directly due to the absence of information on childbirth. However, the data allow identifying career breaks around the age of parenthood (25-34) by focusing on non-employment spells. Career breaks are likely to account for an important fraction of the motherhood penalty, and as a result, play a potentially important role in determining the evolution of the gender wage gap within and between firms over the life-course. This sub-section documents the incidence of career breaks around the age of childbirth and their consequences for the wage growth of women.

There are important differences in the incidence and duration of career breaks across countries. Women aged 25-34 are much more likely than men to experience a non-employment spell of one or more years, while there is only small difference between women and men aged 35-54 (Figure 5.12, Panel A). While non-employment spells may reflect many factors, the difference between women and men for workers aged 25-34 is likely to be driven by career breaks of women around the age of childbirth. Such careers breaks are most common in Northern European countries (Denmark, Finland and Sweden), while they are least common in Western European countries (France, Italy, the Netherlands and Portugal). In Central and Eastern European countries (Austria, Estonia and Hungary), they are quite common and often last for more than one year. While differences across skills groups are generally small, there is some indication that career breaks are more common among low-skilled women in Northern European countries and among high-skilled women in other European countries (Panel B).

Career breaks tend to be associated with significant wage losses (Figure 5.13). Wage losses in principle may reflect the slower upward mobility within firms due to lost experience and the possible depreciation of relevant skills or the sorting of persons following a career break into lower wage firms. To examine this, Panel A documents the percentage difference in wages within firms (conditional on age and education) before and after career breaks of different duration for women, while Panel B shows the percentage difference in firm wage premia between firms.29 The evidence suggests that wage losses due to missed experience or human capital depreciation can be sizeable, amounting to about 4% for career breaks of one year, and even larger for longer career breaks.30 In Central and Eastern European countries, wage losses tend to be largest, and larger for more skilled women, while they do not depend much on the duration of the break. In Western European countries, wage losses follow the average profile across countries, whereas in Northern European countries, wage losses are small and limited to low-skilled women. Women do not tend to move to lower wage firms following a career break (if anything the opposite is observed). This reflects the fact most women return to their previous employer after a career break. Consequently, sorting to lower-wage firms does not significantly contribute to the gender wage gap between firms.

The country patterns documented in this sub-section may be indicative of the role played by policies and institutions in shaping the incidence and duration of career breaks and hence the labour market consequences of childbirth. However, they may also reflect deeply engrained cultural differences between countries in the form of social norms. This is consistent with evidence for Denmark that shows that the motherhood penalty is highly persistent over time and tends to be transmitted across generations (Kleven et al., 2019[28]). This suggests that family policies need to be complemented with other policies that can help change social norms (e.g. school interventions).

To analyse the role firms for the gender wage gap over the life-course, this chapter decomposes the gender wage gap between similarly-skilled women and men at different ages into a between-firm component that captures the sorting of women into low-wage firms and a within-firm component that captures systematic differences in pay between women and men in the same firm. On average across the countries covered by the analysis, about three quarters of the gender wage gap reflects pay differences within firms mainly due to differences in tasks, responsibilities, but to a lesser extent also differences in pay for work of equal value. The remaining one quarter reflects differences in wage premia between firms due to sorting. The gender gap tends to increase during to the initial phase of women’s professional career due to the role of motherhood. This reflects to an important extent gender differences in mobility between and within firms and the role of career breaks for the career progression of women within firms. Consequently, tackling the gender wage gap crucially requires promoting access of women to well-paying firms and well-paying jobs within firms. This involves a range of policies (OECD, 2017[18]), including:

• Family policies. Family policies can contribute to a more equal sharing of household and care responsibilities between men and women and hence enable women to take advantage of opportunities for career progression within their current firms and at other employers. This is particularly important for countries, which see strong and persistent increases in the gender wage gap as workers advance in their careers (e.g. Western European countries, Japan). Key family policies include more equal parental leave policies for men and women, which helps introduce egalitarian norms in parenting when children enter a family; providing universal childcare, out-of-school supports and supports for elder care; and reducing marginal effective tax rates for second earners (OECD, 2017[12]). While there is strong empirical support for the role of parenthood for the gender wage gap and the need for a more equal sharing of household responsibilities (Kleven et al., 2020[30]), concerns have been raised about the effectiveness of family policies for reducing the gender wage gap in a context where preferences and social norms are deeply anchored in society (Kleven et al., 2020[30]). It is important therefore to complement family policies with other policies that can help foster more gender-friendly social norms (e.g. school interventions).

• Mobility within and between firms. To make good jobs more accessible to women, the use of flexible work arrangements across occupations and firms, including telework and part-time work, should be supported and offered to all workers – not only parents (OECD, 2019[13]). This would reduce the contribution of compensating wage differentials related to the valuation of working time flexibility by women to the gender wage gap, and the segregation of women and men across firms and jobs with different non-wage characteristics. Voluntary target setting and good management practices that make managers accountable are among the measures that could also help to promote access for women to quality jobs, while at the same time foster social norms that support gender equality. Gender quotas could in principle also help, but need to be used judiciously to avoid the risk that they undermine firm performance, particularly if targets are set too high given the number of suitably-qualified women in the sector/occupation (Hwang, Shivdasani and Simintzi, 2018[31]). Finely targeted quotas such as those related to company boards seem to hold some promise in this regard. Recent evaluations suggest that although such quotas enhance the representation of women in company boards, they have limited spillover effects on the career progression of other women in those firms (Bertrand et al., 2019[32]; Maida and Weber, 2020[33]).

• Equal-pay-for-equal-work measures. A key obstacle to reducing gender wage gaps is that employers and employees are often unaware of them. Pay transparency rules raise awareness of discrimination and make it easier to enforce equal pay legislation. Pay transparency rules come in a variety of forms in OECD countries, and can, for example, provide the right to request information on pay levels by gender within firms, require firms to report information on employment and pay by gender, or incentivise firms to undertake gender pay audits. About half of OECD countries have recently put in place pay transparency measures (e.g. Austria, France, Germany, Sweden). Recent studies have shown that mandatory reporting requirements can help reducing the gender wage gap within firms (Bennedsen et al., 2019[22]; Blundell, 2020[34]; Baker et al., 2019[20]). Equal pay for work of equal value measures are particularly important for certain countries with large initial gender wage gaps early in worker careers (e.g. Estonia).

• Investing in STEM. While in most countries women outperform men in terms of the level of education – women are more likely to hold a tertiary degree – fewer women than men complete Science, Technology, Engineering and Mathematics (STEM) degrees (Mostafa, 2019[35]). To some extent educational choices may reflect the possibility that teenage boys still perform better in STEM subjects than girls, but gender stereotypes also play an important role in driving the educational choices of girls and boys. The lower likelihood of women to choose STEM subjects is also likely to contribute to sectoral segregation. Investing in STEM) and addressing stereotypes that drive the educational choices of girls and boys is particularly important in countries with high levels of gender segregation such as Italy and Portugal.

In addition to informing policies to tackle the gender wage gap, the linked employer-employee data used in this chapter can also be employed to contribute to gender pay transparency. Since the data cover the universe of firms and workers in most countries and provide detailed information on the characteristics of workers within firms, they are ideally suited for documenting gender wage gaps within single firms for similarly-skilled women and men. Indeed, a number of countries already have taken steps to mobilise linked employer-employee data to promote gender pay transparency by providing detailed information on reference wages in a specific industry, occupation or region. For example, Statistics Estonia is developing a web application that provides detailed information on reference wages based on administrative data. Moreover, computing firm-specific gender wage gaps, adjusted for differences in skills, from linked employer-employee data, could relief firms from reporting requirements related to pay transparency laws where these exist and ensure that reporting is done in a consistent manner across firms (Breda et al., 2021[36]).

## References

[15] Abowd, J., F. Kramarz and D. Margolis (1999), “High Wage Workers and High Wage Firms”, Econometrica, Vol. 67/2, pp. 251-333, https://www.jstor.org/stable/2999586.

[20] Baker, M. et al. (2019), Pay Transparency and the Gender Gap, https://www.bloomberg.com/news/articles/2017-04-13/tech-companies-tout-gender-pay-equity-but-balk-at- (accessed on 3 June 2021).

[11] Barth, E., S. Kerr and C. Olivetti (2021), “The dynamics of gender earnings differentials: Evidence from establishment data”, European Economic Review, Vol. 134, p. 103713, http://dx.doi.org/10.1016/j.euroecorev.2021.103713.

[22] Bennedsen, M. et al. (2019), Do Firms Respond to Gender Pay Gap Transparency?, National Bureau of Economic Research, Cambridge, MA, http://dx.doi.org/10.3386/w25435.

[32] Bertrand, M. et al. (2019), “Breaking the Glass Ceiling? The Effect of Board Quotas on Female Labour Market Outcomes in Norway”, Review of Economic Studies, Vol. 86/1, pp. 191-239, http://dx.doi.org/10.1093/restud/rdy032.

[2] Blau, F. and L. Kahn (2017), “The Gender Wage Gap: Extent, Trends, and Explanations”, Journal of Economic Literature, Vol. 55/3, pp. 789-865, http://dx.doi.org/10.1257/jel.20160995.

[34] Blundell, J. (2020), “Wage Responses to Gender Pay Gap Reporting Requirements”, SSRN Electronic Journal, http://dx.doi.org/10.2139/ssrn.3584259.

[23] Böheim, R. and S. Gust (2021), The Austrian Pay Transparency Law and the Gender Wage Gap, http://www.iza.org (accessed on 3 June 2021).

[41] Booth, A., M. Francesconi and J. Frank (2003), “A sticky floors model of promotion, pay, and gender”, European Economic Review, Vol. 47/2, pp. 295-322, http://dx.doi.org/10.1016/s0014-2921(01)00197-0.

[36] Breda, T. et al. (2021), “Les inégalités salariales femmes-hommes dans les entreprises”, Notes IPP 68.

[29] Bronson, M. and P. Skogman Thoursie (2020), “The Wage Growth and Within-Firm Mobility of Men and Women: New Evidence and Theory”.

[9] Bruns, B. (2019), “Changes in workplace heterogeneity and how they widen the gender wage gap”, American Economic Journal: Applied Economics, Vol. 11/2, pp. 74-113, http://dx.doi.org/10.1257/app.20160664.

[8] Card, D., A. Cardoso and P. Kline (2016), “Bargaining, Sorting, and the Gender Wage Gap: Quantifying the Impact of Firms on the Relative Pay of Women”, The Quarterly Journal of Economics, Vol. 131/2, pp. 633-686, http://dx.doi.org/10.1093/qje/qjv038.

[7] Casarico, S. et al. (2019), “What Firms Do: Gender Inequality in Linked Employer-Employee Data”, Cambridge Working Papers in Economics, Faculty of Economics, Cambridge, https://doi.org/10.17863/CAM.41764.

[40] Ciminelli, G., C. Schwellnus and B. Stadler (2021), “Sticky floors or glass ceilings? The role of human capital, working time flexibility and discrimination in the gender wage gap”, OECD Economics Department Working Papers, No. 1668, OECD Publishing, Paris, https://dx.doi.org/10.1787/02ef3235-en.

[6] Coudin, E., S. Maillard and M. Tô (2018), “Family, firms and the gender wage gap in France”, IFS Working Paper W18/01.

[16] Dostie, B. et al. (2020), “Employer Policies and the Immigrant-Native Earnings Gap”, http://www.nber.org/papers/w27096 (accessed on 17 October 2021).

[21] Duchini, E. et al. (2020), “Pay Transparency and Cracks in the Glass Ceiling”, CAGE Online Working Paper Series 482, Competitive Advantage in the Global Economy (CAGE).

[19] Gallen, Y., R. Lesner and R. Vejlin (2019), “The labor market gender gap in Denmark: Sorting out the past 30 years”, Labour Economics, Vol. 56, pp. 58-67, http://dx.doi.org/10.1016/J.LABECO.2018.11.003.

[1] Goldin, C. (2014), “A Grand Gender Convergence: Its Last Chapter”, American Economic Review, Vol. 104/4, pp. 1091-1119, http://dx.doi.org/10.1257/aer.104.4.1091.

[10] Goldin, C. et al. (2017), “The Expanding Gender Earnings Gap: Evidence from the LEHD-2000 Census”, American Economic Review, http://dx.doi.org/10.1257/aer.p20171065.

[24] Gulyas, A., S. Seitz and S. Sinha (2021), Does Pay Transparency Affect the Gender Wage Gap? Evidence From Austria Does Pay Transparency Affect the Gender Wage Gap? Evidence from Austria *, https://ec.europa.eu/info/sites/info/files/factsheet-pay_transparency-2019.pdf (accessed on 3 June 2021).

[27] Guvenen, F. et al. (2021), “What Do Data on Millions of U.S. Workers Reveal About Lifecycle Earnings Dynamics?”, Econometrica, Vol. 89/5, pp. 2303-2339, http://dx.doi.org/10.3982/ecta14603.

[37] Hirsch, B. (2016), “Gender wage discrimination”, IZA World of Labor, http://dx.doi.org/10.15185/izawol.310.

[31] Hwang, S., A. Shivdasani and E. Simintzi (2018), “Mandating Women on Boards: Evidence from the United States”, SSRN Electronic Journal, http://dx.doi.org/10.2139/ssrn.3265783.

[17] Jewell, S., G. Razzu and C. Singleton (2019), “Who Works for Whom and the UK Gender Pay Gap”, British Journal of Industrial Relations, Vol. 58/1, pp. 50-81, http://dx.doi.org/10.1111/bjir.12497.

[30] Kleven, H. et al. (2020), Do Family Policies Reduce Gender Inequality? Evidence from 60 Years of Policy Experimentation, National Bureau of Economic Research, Cambridge, MA, http://dx.doi.org/10.3386/w28082.

[3] Kleven, H. et al. (2019), “Child Penalties across Countries: Evidence and Explanations”, AEA Papers and Proceedings, Vol. 109, pp. 122-126, http://dx.doi.org/10.1257/pandp.20191078.

[28] Kleven, H. et al. (2019), “Children and Gender Inequality: Evidence from Denmark †”, American Economic Journal: Applied Economics 4, pp. 181-209, http://dx.doi.org/10.1257/app.20180010.

[33] Maida, A. and A. Weber (2020), “Female Leadership and Gender Gap within Firms: Evidence from an Italian Board Reform”, ILR Review, http://dx.doi.org/10.1177/0019793920961995.

[38] Manning, A. and B. Petrongolo (2008), “The Part-Time Pay Penalty for Women in Britain”, Source: The Economic Journal, Vol. 118/526, https://about.jstor.org/terms (accessed on 18 October 2021).

[5] Masso, J., J. Meriküll and P. Vahter (2020), The role of firms in the gender wage gap, https://majandus.ut.ee/en/research/workingpapers (accessed on 2 June 2021).

[35] Mostafa, T. (2019), “Why don’t more girls choose to pursue a science career?”, PISA in Focus, No. 93, OECD Publishing, Paris, https://dx.doi.org/10.1787/02bd2b68-en.

[25] OECD (2021), Show me the money.

[13] OECD (2019), Part-time and Partly Equal: Gender and Work in the Netherlands, OECD Publishing, Paris, https://dx.doi.org/10.1787/204235cf-en.

[4] OECD (2018), “Starting close, growing apart: Why the gender gap in labour income widens over the working life”, in OECD Employment Outlook 2018, OECD Publishing, Paris, https://dx.doi.org/10.1787/empl_outlook-2018-10-en.

[12] OECD (2017), Dare to Share: Germany’s Experience Promoting Equal Partnership in Families, OECD Publishing, Paris, https://dx.doi.org/10.1787/9789264259157-en.

[18] OECD (2017), The Pursuit of Gender Equality: An Uphill Battle, OECD Publishing, Paris, https://dx.doi.org/10.1787/9789264281318-en.

[26] OECD (2015), “The quality of working lives: Earnings mobility, labour market risk and long-term inequality”, in OECD Employment Outlook 2015, OECD Publishing, Paris, https://dx.doi.org/10.1787/empl_outlook-2015-8-en.

[14] Olivetti, C. and B. Petrongolo (2008), “Unequal Pay or Unequal Employment? A Cross‐Country Analysis of Gender Gaps”, Journal of Labor Economics, Vol. 26/4, pp. 621-654, http://dx.doi.org/10.1086/589458.

[39] Russo, G. and W. Hassink (2008), “The Part-Time Wage Gap: a Career Perspective”, De Economist 2008 156:2, Vol. 156/2, pp. 145-174, http://dx.doi.org/10.1007/S10645-008-9087-0.

## Notes

← 1. This chapter has been written by an OECD team consisting of Antton Haramboure and Alexander Hijzen with contributions of: Antoine Bertheau (University of Copenhagen, DENMARK), Gabriele Ciminelli (OECD), Chiara Criscuolo (OECD), Katarzyna Grabska-Romagosa (Maastricht University, THE NETHERLANDS), Ryo Kambayashi (Hitotsubashi University, JAPAN), Michael Koelle (OECD), Balazs Murakőzy (University of Liverpool, HUNGARY), Vladimir Peciar (Ministry of Finance, SLOVAK REPUBLIC), Andrei Gorshkov and Oskar Nordström Skans (Uppsala University, SWEDEN), Satu Nurmi (Statistics Finland/VATT, FINLAND), Catalina Sandoval and Jonathan Garita (Central Bank of Costa Rica, COSTA RICA), Nathalie Scholl (OECD), Cyrille Schwellnus (OECD) and Richard Upward (University of Nottingham, UNITED KINGDOM). For details on the data used in this chapter please see the standalone Data Annex and Disclaimer Annex.

← 2. The countries covered in this chapter are Austria, Costa Rica, Denmark, Estonia, Finland, France, Germany, Italy, Hungary, Japan, the Netherlands, Portugal, the Slovak Republic, Spain, Sweden and the United Kingdom.

← 3. For the European countries covered in the empirical analysis in Section 4, the conditional gender wage gap is on average between 2002 and 2018 was 18%. The corresponding gap measured using linked employer-employee data amounts to 23% (excluding Costa Rica and Japan). The difference may reflect the inclusion of small firms with less than 10 workers and overtime payments in the calculation of the gender wage gap using linked employer-employee data.

← 4. The two exceptions are Germany and Belgium.

← 5. In principle, this could also reflect the role of cohort effects, i.e. the possibility that the gender wage gap tends to higher among older birth cohorts. However, controlling for cohort effects through the inclusion of decade-of-birth dummies does not significantly change the pattern shown.

← 6. The argument could alternatively be phrased in terms of compensating differentials, when firms offering more attractive non-wage working conditions offer lower wage premia.

← 7. Previous studies suggest that the job mobility behaviour of women is much less sensitive to wages than that of men, suggesting that there is considerable scope for gender discrimination (Hirsch, 2016[37]). While this evidence confirms that there is scope for monopsonic gender discrimination, it does not actually show the extent to which employers exploit differences in wage-setting power across men and women to increase profits. However, because of legal constraints or concerns over fairness, employers might not fully exploit their wage-setting power in practice.

← 8. Experience is measured in potential terms using age and therefore does not take account of for example career breaks.

← 9. Those born in the 1960s are used as the reference group for the analysis.

← 10. For Germany, in the absence of information on hourly wages, the analysis is restricted to full-time workers as in Bruns (2019[9]).

← 11. Part-time status is defined either on the basis of working time or, if this information is not available, as those earning less than 75% of the full-time minimum wage or, in the absence of a minimum wage, 37.5% of the median.

← 12. This allows defining the gender wage gap in terms of average wages as well as median wages. The gender gap in median wages is used for the official OECD measure of the gender wage gap.

← 13. Taking account of bonuses and overtime payments significantly increases the measured gender pay gap in Japan.

← 14. Applying the approach proposed by Card, Cardoso and Kline (2016[8]) yields broadly similar insights with respect to the importance of sorting in the gender wage gap with the exceptions of Germany (Figure 5.7). Once worker fixed effects are included, the role of sorting in the gender wage gap in Germany increases from being negligeable to being above the average across the sixteen OECD countries considered. This suggests that controlling for unobserved differences in worker composition across firms is important for understanding the role of sorting in the gender wage gap in Germany.

← 15. In Estonia, Hungary and Slovak Republic, women are more likely to work in large firms, which tend to pay higher wages, reducing the gender wage gap.

← 16. The Netherlands is an exception since women disproportionately work in high-wage sectors.

← 17. Administrative data such as those used for this paper are not well suited to analyse the incidence of working very long hours. The reason for this is that they record contractual hours or paid overtime, whereas in practice hours beyond the contractual level are often not paid.

← 18. The detailed decomposition can only be implemented with for countries with sufficiently long panels. This means it cannot be implemented for Japan and the Slovak Republic. Moreover, the component associated with bargaining and discrimination is likely to be overstated in countries without information on working time. This is particularly an issue in countries such as Austria where the gender gap in working time is relatively large.

← 19. Pay transparency measures can cover different obligations (OECD, 2021[25]). Amongst others, these measures can provide the right to request information on pay levels by gender within firms, require firms to report information on employment and pay by gender, mandate or incentivise firms to undertake gender pay audits (which require analysis beyond the gender wage gap), mandate public disclosure of wages and/or the use of gender-neutral job classification systems. Eighteen OECD countries impose regular reporting requirements on private sector firms in relation to the gender wage gap.

← 20. These patterns are broadly comparable with those documented in Ciminelli, Schwellnus and Stadler (2021[40]).

← 21. In countries without information on working time, including Austria, Estonia and Finland, these patterns may to some extent reflect temporary increases in part-time work among women around the age of women become mothers for the first time. This is likely to be particularly an issue in Austria where female part-time employment displays a pronounced life-cycle profile with a strong increase around the age women become mothers. In Estonia and Finland female part-time employment exhibits a similar life-cycle profile, but part-time is much less common.

← 22. The stable profile of the gender wage gap until the age of 45 in Denmark may to some extent be related to the very high degree of female labour force participation throughout the life-course.

← 23. Promotions can also affect the gender wage gap between firms. The role of promotions related to moves between firms is analysed separately below.

← 24. See also Russo and Hassink (2008[39]) for similar findings based on linked employer-employee data for the Netherlands. Lastly, women’s higher rate of part-time work after having children accounts for 21% of the cumulative gap by age 45.

← 25. Apart from shaping gender wage gap, the gender gap in promotions also shapes occupational segregation. Evidence by Manning and Petrongolo (2008[38]) suggests that most of the wage penalty associated with part-time work reflects occupational segregation.

← 26. This is likely to reflect the possibility that women are positively selected in (full-time) employment. Similar observations have been documented in the literature. For example, Booth et al. (2003[41]) document that full-time women are slightly more likely to be promoted than men.

← 27. Differences in the probability of being promoted – keeping constant differences in wage increases - contribute to an increase in the gender wage gap of 6 percentage points at age 45 (quantity effect), whereas differences in the nature of promotions – keeping constant differences in probability - reduce the gender wage gap by 1.4 percentage points (wage effect).

← 28. Evidence based on the detailed decomposition of the gender wage gap over the working life does not suggest discrimination varies substantially with age (Annex Figure 5.A.3). However, this only captures changes in discrimination due to sorting of women across firms that differ in their discriminatory wage-setting practices. The analysis does not capture the possible change in discriminatory wage-setting practices within firms based on changes in the bargaining position of women due to, for example, motherhood.

← 29. The results for men are very similar to those of women (not reported).

← 30. Differences between women and men in the wage losses associated with career breaks tend to be small.