Family-friendly policies and women's wages – is there a trade-off? Skill investments, occupational segregation and the gender pay gap in Germany, Sweden and the UK Open Access

In research on the gender wage gap, new puzzles have emerged. Although women have been catching up with and, in the case of education, even surpassed men in terms of human capital investments, they still receive lower wages. In fact, the gender wage gap appears to be particularly large among highly educated employees and in highly prestigious occupations (e.g. Evertsson et al. 2009; Magnusson 2010) and is remarkably persistent even where social policies promote gender equality, as the in Scandinavian countries (Boye et al.2014).

In light of the current situation, researchers are pointing to a potential ‘catch-22’ in the quest for gender equality. Emerging from the work of both economists and social scientists is the notion of a trade-off between women's economic and occupational achievements and the ‘family-friendliness’ of jobs, occupations and welfare states. The argument ties together gender, skill investments, social policy and wages in a way that, theoretically, would provide a key to understanding the puzzling wage gap that is found also among the high-skilled and in ‘woman-friendly’ contexts.

In the trade-off argument, workplace skill investments play a central role. Several scholars argue that skill acquisition through workplace training and learning – notably, employer investments in on-the-job training – presuppose long-term employment relations and because women can be presumed to interrupt their careers due to childrearing, employers are reluctant to provide them with training (e.g. Tam 1997; Polavieja 2008). Furthermore, because the importance of such training varies between occupations, women can be excluded from occupations requiring substantial on-the-job training (Polachek 1981; Estévez-Abe 2005, 2006). Workplace skill investments can also form the rebar of the ‘glass ceiling’ that many women hit somewhere up the career ladder. Because on-the-job training is generally considered to be more important in high-skilled jobs (e.g. OECD 2003; Iversen and Rosenbluth 2012), gender differences in access to such training is likely to produce a larger wage gap among the high-skilled. In short, workplace skill development and gendered occupational tracks might devalue the human capital acquired in the educational system and contribute to horizontal and vertical stratification which, in turn, produce a gender wage gap that is larger among the highly educated (e.g. Mandel and Shalev 2009b; Mandel 2012).

Recently, prominent scholars have extended the trade-off argument to include social policy by arguing that policies designed to improve work–family reconciliation actually make matters worse. In particular, the dual-earner model of the Scandinavian countries institutionalizes female work interruptions in a way that will have negative consequences for women's relative wages, especially in higher positions (e.g. Mandel and Semyonov 2005, 2006; Mandel and Shalev 2009a; Mandel 2009, 2012). Meanwhile, labour market institutions that promote long-term employment and workplace skill investments make the consequences of interruptions more severe (Estévez-Abe 2005, 2006). By this account, the Anglo-Saxon countries – with hire-and-fire labour markets and rudimentary family policies – are seen as the most beneficial for women with career ambitions, while the Scandinavian countries – traditionally depicted as a paradise of gender equality – come out as more problematic, especially for career-minded women.

Clearly, the notion of a trade-off between work–family reconciliation on the one hand and women's careers and wages on the other deserves serious scrutiny. To date, comparative research on the gender wage gap is relatively limited and the mechanisms put forward in the trade-off hypothesis have not been thoroughly examined. In this article, we provide a tentative assessment of the trade-off hypothesis by examining how these mechanisms contribute to gender wage gaps in Germany, Sweden and the UK. Utilizing pooled wage data from the European Social Survey (ESS) 2004 and 2010, we study the relative impact of education, workplace skill investments and occupational segregation. Also, we compare the situation of high- and low-skilled employees in these countries, representing different institutional frameworks.

In human capital theory, Becker (1991) claims that, because families benefit economically from within-couple specialization, women invest less in their human capital, that is, in skill acquisition through schooling and workplace training. This claim seems outdated at a time when female students outnumber men at the universities in most OECD countries (2008). However, skill investments can be gendered in ways that are not captured by traditional human capital variables, that is, years of education and work experience. In particular, several researchers have pointed to the importance of specific human capital. Based on Becker's theory of human capital (1964/1993), they claim that investments in specific human capital – in particular, firm-specific skills developed through on-the-job training – presuppose long-term relations between employers and employees.¹ Therefore, employers will hesitate to invest in women who are expected to interrupt their work life to care for their children and families (Polachek 1981). Furthermore, it is argued that some occupations require more workplace skill development than others and that women are excluded from these occupations – or avoid them – due to anticipated discrimination (Polachek 1981; Estévez-Abe 2005, 2006). Thus, while general skills obtained through formal schooling are gender-neutral, specific skills connected to the workplace discriminate against women and provide a mechanism for occupational gender segregation.²

Occupational gender segregation is a prominent pattern throughout the OECD and has been related to several labour market outcomes, including wages. A vast amount of research shows that employees with equal qualifications receive lower wages in female-dominated occupations than in those dominated by men (e.g. Cohen and Huffman 2003). Thus, part of the gender wage gap can be attributed to occupational segregation and workplace skill investments have been presented as one potential mechanism behind this gap. Using measures of on-the-job training, which are considered to be proxies for specific human capital, several scholars have shown that gendered access to such training can explain part of the gender wage gap and part of the relationship between occupational gender segregation and wages (e.g. Tam 1997; Polavieja 2008).

The interest in occupational segregation and workplace skill development has led to a radically new perspective on the relationship between welfare-state institutions and gender equality. Traditionally, the social democratic welfare state of the Scandinavian countries, with extensive support for dual-earner families, has been considered to promote gender equality by stimulating the labour force participation of women, in particular mothers (e.g. Korpi 2000). In the current debate, however, these policies take on a more double-edged nature as scholars suggest there is a trade-off between ‘mother-friendly’ family policies and women's occupational and earnings achievements.

The trade-off hypothesis has been developed mainly in two strands of research. One of them comprises the works by Mandel and Semyonov (2005, 2006), Mandel and Shalev (2009a), Mandel (2012); the other was established by researchers connected to the Varieties-of-Capitalism (VoC) school. Mandel and colleagues argue that extensive family policies bring mothers into the labour market on a large scale and create family-friendly jobs in the public sector and while this may benefit women with meagre human capital endowments, it will harm more educated and career-minded women. The institutionalized rights to work interruptions – for example, through parental leave entitlements – will increase statistical discrimination of highly educated women by private employers because ‘for jobs with high training costs, they favor more stable and productive employees’ (Mandel 2012: 243). As a result, the labour market will be segregated and women's chances of attaining more lucrative positions will be hindered. Due to the policies of work–family reconciliation, then, the ‘glass ceiling’ is predicted to be lower and harder in the Scandinavian countries than in other countries, especially the Anglo-Saxon countries.

A similar argument, but with a focus on labour market institutions, has been made by researchers applying a gender perspective to the influential VoC school, most notably Estévez-Abe (e.g. Estévez-Abe et al. 2001; Estévez-Abe 2005, 2006). In the VoC perspective, policies and institutions are firmly linked to the skill requirements of production. A central idea is that the institutional framework of a country, as defined by its capacity for non-market coordination (or long-term strategic interaction between important actors of the economy), is crucial to specific skill accumulation and therefore, to companies’ production strategies. In liberal market economies (LMEs; the Anglo-Saxon countries), the lack of coordination requires firms to react promptly to supply and demand in the market. Thus, they rely on hire-and-fire strategies requiring mainly general skills that are taught in the educational system and are readily available on the market. In contrast, the coordinated market economies (CMEs; the Scandinavian and continental European countries) have developed institutions that promote long-term employment – in particular, strong employment protection legislation. These institutions facilitate investments in specific skills but because such investments are jeopardized by employee work interruptions, employers will be reluctant to invest in women. Therefore, CME institutions produce occupational and vertical gender segregation and exacerbate gender differences in skills and wages and by institutionalizing work interruptions, generous family policies further aggravate the situation (Estévez-Abe et al. 2001; Estévez-Abe 2005, 2006).

These arguments – pointing to a catch-22 in the quest for gender equality – form a new perspective that calls for empirical investigation. To date, comparative research on the gender wage gap has investigated differences between countries in terms of their overall wage structures and female labour supply (Blau and Kahn 2003). Some studies have focused on the relationship between family policies and the gender wage gap, noting that the overall gender wage gap tends to be largest in the Anglo-American countries and smallest in the Nordic countries with the continental European countries in-between (e.g. Harkness and Waldfogel 2003). However, the work of Mandel and colleagues provides a new focus by highlighting the effects of welfare-state policies and potential class–gender interactions. A focus on class–gender differences is also motivated by studies showing that the gender pay gap is largest among the highly educated (Evertsson et al. 2009), in high-prestige occupations (Magnusson 2010) and at the top end of the wage ladder (e.g. Arulampalam et al. 2007). Still, several gaps remain to be filled. Most empirical studies assessing the trade-off hypothesis have focused on horizontal and vertical segregation (Estévez-Abe 2005, 2006; Mandel and Semyonov 2006; Mandel and Shalev 2009a). The gender wage gap has been studied by Mandel and Semyonov (2005) and Mandel (2012) who find that welfare state interventions – an index comprising both family policies and public sector employment – do not affect the overall gender wage gap; however, they increase the gap in higher socio-economic groups and decrease it in lower classes. Neither of the above-mentioned studies explores how skill investments and occupational segregation – the mechanisms proposed to link policy and outcomes – affect wage gaps in different strata. Interestingly, however, comparative studies which investigate earnings differences between mothers and non-mothers – the so-called motherhood wage penalty – show rather large wage gaps between mothers and non-mothers in the USA and the UK, while these differences seem to be small or non-existent in the Scandinavian countries (e.g. Sigle-Rushton and Waldfogel 2007; Gash 2009). This is in line with the arguments made by Mandel and colleagues, who suggest that in countries which institutionalize maternal work interruptions even women who are firmly committed to their careers (e.g. non-mothers) will be penalized (through processes of statistical discrimination).

In this article, we acknowledge the insights from the two strands of research discussing the trade-off hypothesis. These are based on different theoretical perspectives and have different focuses, but share some central arguments. While Mandel and colleagues emphasize how extensive family policies institutionalize female work interruptions, Estévez-Abe and other VoC-scholars explain why such interruptions are particularly problematic for employers in some countries. However, both sides agree that occupational segregation and workplace skill development are important mechanisms linking institutional frameworks to gender inequalities and that institutionalized work interruptions are particularly harmful for high-skilled women.³ By examining the gender wage gap among high-skilled and low-skilled employees in different institutional contexts and exploring the relevance of the proposed mechanisms, we hope to bridge the gap between the two perspectives and develop the debate on policies and gender inequalities.

To this end, we take a close-up look at three countries that represent different combinations of family policy and labour market institutions. In Table 1, we illustrate the character of these institutions. As mentioned, the VoC literature classifies the UK as a LME based on hire-and-fire principles and defines Germany and Sweden as CMEs promoting long-term employment relations (Hall and Soskice 2001) and the difference is reflected in the strength of employment protection. In classifications of family policy (Korpi 2000, Korpi et al. 2013, cf. Thévenon 2011), Sweden is labelled a dual-earner model. Here, policy instruments such as earnings-related parental leaves, public day-care services at a full-time basis for small children and individual taxation for spouses combine to support female employment and the reconciliation of work and family. As seen in Table 1, public spending on children is much higher in Sweden than in the other countries and the earnings-related parental leave longer. Germany, as a contrast, is considered a traditional family model. Here, taxes subsidize families with only one provider and care of small children in the home is encouraged by the lack of public day-care services and through cash-for-care allowances as well as long parental leaves with low compensation.⁴ In the UK, family policy is classified as a market model, since the state does not strongly support either dual-earner or traditional households. Thus, there is little public support for childcare and as we can see in Table 1, parental leaves are shorter than in the other countries.

A well-documented effect of the dual-earner policies is that women remain in work through the child-rearing years. These work-and-care practices are illustrated in Table 1, where Sweden stands out regarding female and – particularly – maternal employment. Theoretically, this is a main point of the trade-off hypothesis. In the words of Mandel and Shalev (2009a: 1879), extensive family policies stimulate an ‘unselective entry of women into the labour force’ that will affect gender inequalities on the labour market because ‘if mothers with relatively meager endowments of human capital and weak career motivation are drawn into paid work, they can be expected to achieve limited wage and occupational attainments’.⁵ In the other countries, particularly Germany, women often exit from the labour force when their children are small, labelled here work-or-care practices (Table 1).

Based on the reasoning in the trade-off hypothesis, we argue that the different combinations of family policies and labour market institutions will affect the mechanisms behind the gender wage gap differently in the three countries. Table 2 summarizes the theoretical propositions and in the following, the predictions for each country will be clarified.

Scandinavian countries, represented by Sweden, combine institutions promoting long-term employment relations and on-the-job training with family policies promoting dual roles in work and family (work-and-care). Thus, work–family issues – notably, female work interruptions – become a concern for employers, who become wary of hiring and promoting women (statistical discrimination). Here, horizontal and vertical segregation and limited access to training will negatively affect wages for high-skilled women, while low-skilled women can still benefit from secure and ‘family-friendly’ public employment. As a result, the gender wage gap among high-skilled Swedish employees can be expected to be larger than in other countries. Moreover, a large part of the gap should be explained by segregation and on-the-job training. Among the low-skilled, the gender wage gap should be smaller than in the high-skilled group and smaller than in the other countries. Also, it should be less well explained by segregation and training.

In the Anglo-Saxon countries, represented by the UK, both employment protection and family policies are weak. Here, skills obtained through general education are more important than workplace training and therefore, work interruptions are less consequential for employers. In the UK, public spending on families is low. Presumably, this is less of a problem for high-skilled employees who can purchase childcare on the private market. Because of this, and because the value of educational merits are less compromised by requirements for workplace training, formal education can be expected to explain more of the gender wage gap in the UK than in Sweden. Consequently, the gender wage gap among the high-skilled should be smaller and that among the low-skilled larger than in Sweden. Segregation and on-the-job training should be less important for explaining the wage gap than in the other countries, while work experience and tenure should be more important than in Sweden.

Germany, representing the conservative welfare regimes of continental Europe, provides something of a mixed case. Here, family policies encourage a gender-traditional division of work while labour market institutions emphasize long-term employment. In this situation, women often have to make a choice between children and a career and mothers tend to exit the labour market (work-or-care). The German fertility level is considerably lower and as seen in Table 1, the share of employed mothers with small children is considerably smaller than in the UK and, particularly, Sweden. Thus, female employees – and consequently, employers – are less preoccupied with work–family balancing issues. In fact, in this context, mothers who are indeed working may signal strong work commitment to employers and be better rewarded than mothers in Sweden where continuous employment is the norm (Evertsson and Grunow 2012). Presumably, then, the gender wage gap among the high-skilled should be smaller than in Sweden and education should be more important for explaining the overall gender wage gap. Because German mothers are likely to leave the labour market for an extended period, work experience and tenure should also be more important than in Sweden, particularly among low-skilled employees who are more inclined to exit the labour force (Konietzka and Kreyenfeld 2010). Segregation and on-the-job training should be more important in the CME of Germany than in the UK, but considering the selection processes described above, these factors should explain less of the gender wage gap than in Sweden.

Summing up, we propose the following hypotheses:

Our analysis is based on data from the ESS 2004 and 2010. The ESS is an academically driven cross-national survey that is conducted every two years across Europe. The ESS aims to achieve high methodological standards, striving for optimal comparability in the data collected across all countries. The questionnaires consist of a core section and a rotating module and in 2004 and 2010, the theme of the rotating module was work, family and well-being (see europeansocialsurvey.org). For this analysis, the sample was restricted to employees aged 20–65 working at least 10 hours a week and data from the two years were pooled to increase the sample size. The final sample comprises 1288 respondents for Germany, 1447 for Sweden and 1156 for the UK.

As explained above, the aim of our analysis was to explore the mechanisms behind the gender wage gap with a focus on the relative impact of education, workplace skill development and occupational gender segregation. Here, we should clarify that although the term ‘mechanism’ refers to a causal link between gender and wages, our cross-sectional data set does not allow for causal inferences. Therefore, we can only test if the indicators are mediating variables that explain the relationship in a statistical sense.

In the analysis, we perform ordinary least square (OLS) regressions with the logged hourly wage (converted into Euros) as the dependent variable. In a logarithmic model, a change by one unit in the independent variable produces a percentage change in the dependent variable (Allison 1999). The wage variable for respondents in 2004 was standardized according to the consumer price index in each country to correspond to the wages in the year 2010. Separate regressions are carried out for the three countries with a stepwise inclusion of the independent variables as described below (for descriptive statistics, see Table 1A in Appendix).

In the first part of the analysis, we study the overall gender gap in the three countries. To explain this gap, we first include a continuous variable measuring years of education, then introduce the variables of work experience and work experience squared. These are the variables traditionally used to measure human capital acquisition in school and at work, respectively. However, work experience does not fully capture the arguments regarding on-the-job training investments. Employer tenure is often regarded in public as well as scholarly debate as a proxy for employer investments in such training, but this indicator is also affected by other factors such as employment protection regulations. Thus, it might not reflect gender and/or country differences in terms of skill investments. As a complement, we use initial on-the-job training, measured by the question: ‘Apart from the competence necessary to get a job such as yours, how long does it take to learn to do the job reasonably well?’ The response alternatives have been recoded into number of months, with a top code of 72 months. This variable captures the basic training requirements that are inherent in the job and because this initial training should be comparatively easy to appreciate, the variable seems relevant for testing arguments about employer deliberations on training investments. The indicator has a strong standing in international wage research and findings from different countries confirm that women tend to have jobs involving less on-the-job training (e.g. Gronau 1988).

In the next step of the analysis, we include indicators capturing gender segregation in the labour market. Here, the prime variable used in prior wage research is occupational gender segregation, commonly measured as the percentage of females in the respondent's occupation (e.g. England 1992; Tam 1997). For Germany and Sweden, information on the gender composition of occupations is based on register data from the year 2008 (Germany) and 2009 (Sweden). Such register data were not available for the UK, so we use the Labour Force Survey of UK households from the year 2000 (see http://www.esds.ac.uk/government/lfs/). Occupations have been classified according to the International Standard Classification of Occupations (ISCO-88) and we use the detailed three-digit level. It can be argued that public sector is another relevant variable when studying segregation and the gender wage gap (e.g. Mandel and Shalev 2009a; Mandel 2012) but unfortunately, sector was not available in ESS 2004. However, as a rough proxy of public sector employment, we include a dummy variable based on the NACE classification of industries indicating if the respondent is working in the welfare sector.⁶

In the second part of the analysis, we explore the argument that the institutional framework will affect high-skilled and low-skilled groups differently. Here, we use dummy variables to compare the size of the gender gap and the relative importance of the mechanisms for high-skilled women, high-skilled men, low-skilled women and low-skilled men. The dummies are based on the education required in the job⁷ and the high-skilled group comprises employees with jobs requiring at least five years more than compulsory education. This cut-off point was chosen to get an adequate number of respondents in the high-skilled group, although in most cases five years is not enough for a full bachelor's degree. Because the low cut-off point may underestimate wage gaps among the high-skilled, the results in this second part should be regarded as more tentative. However, our sensitivity analyses using a six-year cut-off show results largely supporting our conclusions.

There are several reasons why we chose to focus on the level of education required in the job rather than on the individual's education in this part of the analysis. First, recent research has acknowledged that the requirements of the job and the education of the individual do not necessarily match (Tåhlin 2014) and since wages are determined by the actual job, the requirements of the job should provide a more stable measure. Second, because educational requirements for many jobs have changed over time, measuring these requirements provides a more up-to-date picture and facilitates comparisons between younger and older employees. The fact that educational systems vary across time and countries was an additional reason for focusing on the job since the internationally standardized variable for level of education was not available in our data.

Below, we present results from the wage regressions carried out separately for Germany, Sweden and the UK. As discussed above, we are not only interested in the size of the gender wage gap but primarily in the mechanisms producing it. First, we will report our findings for the total sample, then we consider the similarities and differences between high- and the low-skilled employees.

In Table 3, model 1, we find that the unadjusted gender wage gap is considerably lower in Sweden (13.5%) than in Germany and the UK (21.1% and 27.4%, respectively). The following models demonstrate how the overall wage gap between men and women is affected by education, workplace skill investments and occupational segregation.

As seen in Table 3, the gender wage gap in Germany clearly decreases when we control for education and work experience. Controlling for employer tenure and initial on-the-job training, the gap is further narrowed. All in all, the gender coefficient decreases substantially from model 1 to model 5. When controlling for % females in the occupation, in model 6, the coefficient is almost halved. Accounting for welfare sector employment does not further narrow the wage gap; instead, the gender coefficient tends to increase.

In Sweden, as a contrast, education and work experience do not explain any of the gender wage gap. Instead, the gender coefficient increases by more than one-fourth when we control for these factors. However, almost half of this increase is erased when initial on-the-job training is included in the regression. This suggests that Swedish men are more rewarded for their educational investments than are Swedish women. At least partly, this is explained by their greater access to on-the-job training, but all in all, we find that the gender wage gap is wider after controlling for the workplace training variables (in model 5) than in the baseline model. When controlling for % females, the gender coefficient decreases by almost one-third. However, welfare sector employment does not explain the gap further.

As in Sweden, the gender wage gap in the UK increases when controlling for education and work experience and also when tenure is accounted for. As in both the other countries, initial on-the-job training clearly reduces the gender wage gap. However, since the gender coefficient is only slightly lower in model 5 than in model 2, the workplace training variables explain little of the gender wage gap. The gender wage gap decreases by almost one-third when controlling for % females in the occupation, but clearly widens when welfare sector employment is entered. This suggests that in the UK, the size of the gender wage gap may vary between the welfare sector and other industries.⁸

All in all, the gender wage gap is best explained in Germany. Here, the gender coefficient is more than halved from the first to the final model. In Sweden and the UK, the gender wage gap narrows by about one-fourth from the first to the final model. One reason for the difference between the countries lies in the differential impact of human capital variables. In Germany, these factors explain a sizeable portion of the gender wage gap. In the UK and, particularly, in Sweden, they have positive effects, suggesting that men and women are differently rewarded for their human capital investments. In summary, H1 is supported in the sense that education, work experience and tenure explain less in Sweden than in Germany, but – quite contrary to expectations – the same is true of the UK. On-the-job training has the predicted negative effect in all countries, but it does not explain more in Sweden than in the other countries and neither does occupational segregation.⁹

Looking at the first bars, we note that in both Germany and Sweden, the unadjusted wage gap is larger among the high-skilled. For the UK, the wage gap appears to be larger among the low-skilled. In fact, the gender wage gap among the high-skilled in the UK is not statistically significant.¹⁰ In particular, we note that in terms of relative wages, high-skilled women do not appear to fare worse in Sweden than in the other two countries while low-skilled women tend to fare better. Thus, H2 is not supported for the high-skilled group so central to the trade-off argument.

The difference between the first and the second bars reflects the impact of training. The largest effect is found among high-skilled employees in Germany, where the gender wage gap decreases from 23% to 13%. According to the regressions, work experience and tenure explain most of this decrease, while initial on-the-job training has a modest impact. Among the low-skilled, the reduction of the gender coefficient is much more modest. In Sweden, the gender wage gaps are no smaller after controlling for training. However, the step-by-step regressions reveal that the three training variables have different effects. When we control for work experience, the gap among the low-skilled increases, suggesting that men and women benefit differently from acquiring work experience. Controlling for tenure does not change the picture, but when initial on-the-job training is accounted for, the gender coefficient is reduced to its initial size. Among the high-skilled, a similar but less pronounced pattern can be discerned. In the UK, the pattern for the low-skilled is similar to that in Sweden. Controlling for work experience and tenure, the wage gap increases and this effect is explained by initial on-the-job training.

Comparing the third bars with the second bars, we can assess the importance of occupational segregation. Here, we note – contrary to our expectations – that the change from bar two to three is largest among low-skilled employees in the UK. In this group, the gender wage gap decreases from about 24% to 18%. In Sweden, the wage gap is reduced from about 11% to 7% among low-skilled employees and from 18% to 16% in the high-skilled group. In Germany, the gap decreases from about 13% to 9% in the low-skilled group and from about 13% to 11% among the high-skilled. Thus, occupational segregation appears to explain more of the gender wage gap among low-skilled employees.

These findings provide no support for H3, stating that in Sweden, segregation and on-the-job training should explain more of the gender wage gap among high-skilled employees, both compared to low-skilled group in Sweden and to high-skilled employees in the other countries. In fact, the findings reveal that most of the action related to training and segregation in Sweden according to Table 3 can be ascribed to the low-skilled employees. Training has the largest impact among high-skilled employees in Germany and segregation is no less important in this group than among high-skilled employees in Sweden. As the gender wage gap among high-skilled in the UK is not significant, mechanisms are not meaningful to discuss. However, the regression results do not suggest that segregation would be less important for the high-skilled in the UK than in the other countries.

In conclusion, we find that H3 is not supported while H1 and H2 are partly supported. All in all, however, the results do not provide any strong support for the trade-off argument. H1 is supported in the sense that education, work experience and tenure explain less in Sweden than in Germany, but – quite contrary to expectations – the same is true of the UK. Moreover, segregation and workplace training do not explain more in Sweden than in the other countries. H2 is supported regarding the gender gap among the low-skilled employees but the unadjusted gender wage gap among high-skilled employees is no larger in Sweden than in the other countries, which is a main point in the trade-off argument. More importantly, the factors included in the regressions – reflecting the theoretical arguments of the trade-off hypothesis – explain less among high-skilled employees in Sweden than in Germany. Thus, in the final model, the wage gap among high-skilled is larger in Sweden. Also, in Sweden, segregation explains more among the low-skilled than among the high-skilled.

The trade-off hypothesis identifies a ‘catch-22’ in the struggle for gender equality, because it suggests that policies promoting work–family reconciliation reinforce gender segregation and discrimination in the labour market. Due to costly investments in firm-specific skills, this trade-off will have a particularly negative impact for highly educated women. Ironically, then, both national and individual attempts to overcome inequalities might have the opposite of the desired effect.

The thrust of this article was to assess these arguments by examining the mechanisms behind the gender wage gap in Germany, Sweden and the UK and comparing the situation of high- and low-skilled employees. Our findings provide little support to the trade-off hypothesis.

In Sweden, with extensive family policies, the gender wage gap among high-skilled employees was hypothesized to be wider than in other countries and largely explained by labour market segregation and requirements for workplace skill development. However, the analysis shows that the gender wage gap is no larger among high-skilled employees in Sweden than in Germany and that segregation and work-related training are no more important in Sweden than in the other countries. In fact, the explanatory power of the analysis is lowest for the group assumed to be most affected by the trade-off between skill investments and family-friendliness, namely high-skilled women in Sweden.

A main point in the trade-off argument is that generous family policies promote an ‘unselective’ inclusion of mothers in the labour force. However, despite the much larger share of mothers in the Swedish labour force, neither low-skilled nor high-skilled women appear to fare worse than in Germany regarding the unadjusted gender wage gaps. However, Sweden did differ from the UK, where the gender gap among the high-skilled was not significant. Also, we note that while traditional human capital variables – education and work experience – explain a fair share of the gender wage gap among high-skilled employees in Germany, this is not the case in Sweden. In short, the mechanisms behind the large wage gap among high-skilled men and women in Sweden remain obscure.

Another important finding is that the mechanisms behind the gender wage gap differ between high- and low-skilled employees in ways not predicted by the trade-off argument. In particular, the importance of occupational segregation may be overrated at least for high-skilled employees in Sweden and Germany. This seems reasonable considering that women enter male-dominated occupations mainly through higher education (e.g. England 2010). Thus, occupational segregation is weakening in prestigious occupations; yet, it is presented in the trade-off hypothesis as a main obstacle for highly educated, career-minded women.

Obviously, there are several limitations to our study. First, it comprises only three countries and the samples are relatively small. Due to limited sample sizes, the cut-off point between low- and high-skilled employees was set relatively low. Thus, the gender wage gap among the high-skilled may be underestimated and our results do not capture the ‘glass ceiling' effect barring women from the most lucrative positions. Also, the importance of on-the-job training may be underestimated. Although the variable used has a strong standing in wage research, it does not reflect the full variation in on-the-job training as it focuses on the basic training that comes with the job. Finally, although occupational segregation is measured in line with convention in wage research, we could not directly measure public sector employment. Considering these limitations, the trade-off argument should be further examined in studies including more countries, larger samples and better developed indicators.

Most importantly, our results show that, at least in the two CME countries, there is a large wage gap among the high-skilled and that this gap is not well explained by the factors discussed in contemporary theories, including occupational segregation, on-the-job training and family policies. Also, the mechanisms behind the gender wage gap appear to vary between countries and classes in ways not foreseen by either human capital theory or the trade-off argument. Clearly, these findings provide both theoretical and empirical challenges for future research.

No potential conflict of interest was reported by the authors.

Anne Grönlund is professor of sociology, currently employed at the Department of Social Work, Umeå University, Sweden. Her research concerns a range of labour market issues as well as work-family reconciliation, with a focus on gender and gender segregation. Important themes include flexibility and autonomy at work, work-family conflict, skill development and wages.

Charlotta Magnusson is a researcher (PhD) at the Swedish Institute for Social Research, Stockholm University. Her research focuses on gender stratification in the labour market; in particular wage, working conditions and occupational prestige. Currently, she is studying how family status is related to the gender wage gap.