Women s Opportunities under Different Family Policy Constellations: Gender, Class, and Inequality Tradeoffs in Western Countries Re-examined

Social Politics 2013 Volume 20 Number 1 Women s Opportunities under Different Family Policy Constellations: Gender, Class, and Inequality Tradeoffs in Western Countries Re-examined Walter Korpi 1,*, Tommy Ferrarini 1 and Stefan Englund 1 This article explores tradeoffs reflecting interaction effects between socioeconomic class and different types of family policies on gender inequalities in terms of agency and economic inequality in eighteen Organization for Economic and Cultural Development countries. We identify multiple dimensions in family policies, reflecting the extent to which legislation involves claim rights supporting mothers paid work or supporting traditional homemaking. We use constellations of multidimensional policies in combination with multilevel analysis to examine effects on class selectivity of women into employment and glass ceilings with respect to women s access to top wages and managerial positions. Our results indicate that while major negative family policy effects for women with tertiary education are difficult to find in countries with well-developed policies supporting women s employment and workfamily reconciliation, family policies clearly differ in the extent to which they improve opportunities for women without university education. Introduction Can public policies decrease gender inequalities to the benefit of all women, or do they primarily result in welfare state paradoxes (Mandel and Semyonov 2006) in which gains in women s employment levels come at the expense of opportunities especially for well-educated women a reshuffling of inequalities among citizens (as, e.g., Cooke 2011 contends is characteristic of many countries)? This question has stayed alive in debates on effects of welfare states. Over recent decades social scientists views on this issue have shown drastic twists and turns. Long seen as fortifying patriarchy, a change came when comparative research indicated that Western welfare states 1 Swedish Institute for Social Research (SOFI), Stockholm University, 10691 Stockholm, Sweden * Corresponding address. Swedish Institute for Social Research (SOFI) Stockholm University, 10691 Stockholm, Sweden. walter.korpi@sofi.su.se Socpol: Social Politics, Spring 2013 pp. 1 40 doi: 10.1093/sp/jxs028 # The Author 2013. Published by Oxford University Press. All rights reserved. For permissions, please e-mail: journals.permissions@oup.com

2 Korpi involve complex structures and actors possibly generating woman friendly policies, to use Hernes (1987) evocative term, with potentials to reduce gender inequalities, especially by reducing gender employment gaps (see, e.g., Crompton 2006; Hobson 1990; Koven and Michel 1993; Leira 1992; Lewis 1993; O Connor, Orloff, and Shaver 1999; Orloff 1993, 2009; Sainsbury 1996). The rise in women s labor force participation rates since the 1960s, partly fueled by state policies and public employment expansion, could be expected to lead to declining gender inequalities. Recently, however, some analysts have questioned such interpretations. Critics argue that in reducing gender employment gaps by expanding women s employment in public sectors, woman-friendly policies such as those seen in the Nordic welfare states create major tradeoffs in terms of reducing inequality, which may have different implications for women of different classes. These analysts argue that social policies have these effects via unintended consequences: segregating women into female ghettos within public sectors; offering a way to ease work family conflicts without challenging the gender division of household labor; increasing employer statistical discrimination; and reinforcing glass ceilings (Albrecht, Björklund, and Vroman 2003; Arulampalam, Booth, and Bryan 2007; Booth 2006; Datta Gupta, Smith, and Verner 2008; Estevez-Abe 2006; Mandel and Semyonov 2005, 2006; Mandel and Shalev 2009a; Mandel 2009). The gains such policies achieve in expanding women s paid employment are therefore perverted when the very same policies thwart career opportunities for women, although at least some of these analysts (e.g., Mandel and Shalev 2009a) conceptualize the problem as different policies favoring women of different classes. (See Hegewisch and Gornick (2011) for a thorough review of this literature.) The purpose of this article is to explore the effects of different types of family policies on opportunities among women occupying different positions in socioeconomic class structures, in the Western countries. We argue that scholars have overstated the negative effects of generous policies promoting women s employment which we call dual-earner and dual-carer or earnercarer policies, and which are found most notably in the Nordic countries on women s economic opportunities. We combine a multidimensional institutional perspective on family policy with an intersectional approach to gender inequality. Coined to address the double disadvantage of black women in terms of race as well as gender (Crenshaw 1989), the term intersectionality has come to be widely used in studies of gender, signaling an assumption that we are likely to observe complex inequalities (McCall 2001) of race, class, gender, and other dimensions of inequality and difference in contemporary societies, which combine in nonadditive ways (see, e.g., Choo and Ferree (2010) on this point). Intersectionality draws attention to patterns of interactions between different aspects of power and inequalities within categories of individuals too often seen as relatively homogeneous. Emerging from a US context in which

Women s Opportunities under Different Family Policy Constellations 3 racial divisions are very marked, intersectionality studies have been especially important in uncovering patterns of race gender interaction (e.g., Collins 1990), but scholars also have investigated gender and class interactions (see, e.g., Acker 2006; McCall and Orloff 2005). McCall (2005, 1788) notes that comparative research can use familiar statistical methods, such as interaction and multilevel analyses, to evaluate such inequalities. In evaluation of family policy effects on class and gender inequalities, we consider the two-pronged nature of gendered outcomes, including economic achievements as well as inequalities in agency and capabilities. In Western societies, inequality has traditionally been defined in terms of individuals or households achievements with respect to income and material standards of living. Yet there are important gender-relevant differences, such as those between homemakers and gainfully employed persons, in access to resources, even within families. Thus, we must examine whether women have access to employment, whatever their household standard of living (Orloff 1993). Moreover, alongside a growing number of feminist scholars influenced by Sen s (1992) and Nussbaum s (2000) theories of capabilities, we argue that gender inequalities also include agency and capabilities. Individuals are purposive actors with different resources and capabilities; an individual s capabilities represent her freedom to choose between alternative ways of living (see also, e.g., Hobson 2011; Lewis and Guillari 2005; Robeyns 2005). In Western countries, being inside or outside the labor force is a key factor differentiating women with respect to agency and capabilities. 1 Compared with homemakers, women in paid work tend to be able to choose within wider sets of accomplishments and to have better capabilities to direct their own lives. Yet clearly capabilities differ depending on what happens in employment with respect to pay and opportunities. Those arguing that there is a welfare state paradox admit the positive effects of increasing women s employment, but contend that the character of employment conditions may suffer under certain policy configurations. Family policies affect the extent to which women join the labor force, and we contend that in these processes family policies interact with class in ways relevant for the socioeconomic composition of women in employment a selection effect (Korpi 2000; Mandel and Shalev 2009a; Montanari 2000a; Shalev 2008). To properly assess the impact of family policy on women s chances to attain high wages and influential positions, then, we must take into account that these policies are likely to influence the class composition of women selected into paid work. In all countries, women with tertiary education have high employment rates (see, e.g., Evertsson et al. 2009); but many analyses of family policy have missed the fact that among women without university education, labor force participation rates are patterned by different constellations of family policies in some countries, women with low or middling levels of education are more likely to be in employment than in others. Country differences in socioeconomic heterogeneity among

4 Korpi employed women must thus be considered in analyses of women s opportunities to reach top wages and powerful positions. Early research on family policies tended to see such policies as unidimensional more or less woman friendly and causal interpretations have often been based on classification of countries into regime typologies. (In addition, the analytic imprecision of the term woman-friendly has been exposed; see, e.g., Borchorst and Siim 2002.) Yet family policies are multidimensional in ways likely to interact with class understood in terms of resources and orientations to employment to differentiate women s labor force attachment. Thus, a long-term low flat-rate homecare allowance for care of a child is likely to be more attractive for a lower-earning mother than that for a higher-paid professional woman; a parental leave program of moderate length with earnings-related benefits, in combination with affordable and high-quality daycare, might appeal to both of them. We define family policy constellations in terms of combinations of separate sets of legislated programs likely to differ in their effects on women s choices between paid work and homemaking and to do so along graded dimensions. This enables us to examine policy effects in terms of direction as well as intensity. Our hypothesis is that, to a major extent, multidimensional structures and inequality outcomes of family policies reflect pressures by political forces with partially differing values and goals (see also Estevez-Abe 2006, 2009; Ferrarini 2006; Mandel 2009, 2011; Morgan and Zippel 2003; Orloff 2006). We next review some previous research related to the interactions between gender and class, focusing on the empirical bases for claims about major tradeoffs between family policies and gender inequalities. Then, we analyze the family policies of eighteen industrialized countries, categorizing legislated programs into dimensions likely to have different effects on mothers choice between homemaking and paid work. 2 Focusing on the relative importance of each set of programs, we examine the relation between policy constellations and gender inequalities. Our analyses of class/gender interactions are based on examinations of interaction effects between family policies and gender inequalities through extensive analyses of relevant datasets, the main one being the Luxembourg Income Study (LIS), which offers comparable individual-level data on key variables for most of our countries. The first step is a multilevel examination of interaction effects between family policy dimensions and women s class positions in terms of the extent to which they enable women with different socioeconomic backgrounds to join the labor force. In the second step, we continue to examine whether family policies differ for women in different class positions in terms of their chance of reaching top wages. In the third step, we analyze effects of different family policy constellations on potential glass ceilings for women in terms of top earnings attainment and managerial positions in large companies. We conclude by restating our case that the earner-carer policies of the Nordic

Women s Opportunities under Different Family Policy Constellations 5 countries have been successful in enhancing the employment levels of women across the class spectrum but have not undermined the opportunities for women, especially if they are well-educated, to gain access to top occupations and wages. Issues and Arguments in Previous Research on Tradeoffs A lively debate has emerged in response to claims made by a number of scholars that family policies that attempt to decrease gender inequalities by transferring care work to the public sector, simultaneously undermine other aspects of women s situation, most notably their occupational and income achievements. Claiming to have discovered a Welfare State Paradox, sociologists Mandel and Semyonov (2006, 1917) state that in highly developed welfare states the glass ceiling has become lower and wider [resulting in] low access for women to positions of power, authority, and high economic rewards. They maintain that even when states facilitate women s employment, state actions do not enhance women s occupational and economic achievements, since none of them seriously challenges the traditional distribution of market-family responsibilities of men and women (1911) women remain the primary caregivers and thereby forfeit the advantages that might come from more fully challenging the gender division of labor. In a parallel study, Mandel and Semyonov (2005, 950) argue that constellations of family policies employing women in large public sectors are likely to increase rather than to decrease earnings gaps between men and women [since] the nature of these jobs and convenient work conditions available in the public sector do not appear to enhance the economic opportunities of women in terms of occupational positions and earnings. Rather they appear to reinforce women s tendency to compromise on convenient working conditions in the female-typed jobs and to deter them from attaining high-paying positions Mandel and Shalev (2009a, 1878 79) also maintain that the Scandinavian experience of egalitarian family policies teaches us that public care... contributes to the concentration of women in feminized service jobs, lowering their representation in better-paid male-dominated jobs. The Welfare State Intervention Index, the key independent variable used by Mandel and Semyonov (2005, 2006), however, fails to differentiate between family policies with contrary effects on gender inequalities. Furthermore, in examining probabilities for women in a country to reach top wages, they focus on country differences among employed women rather than assessing the situation of all women, including those outside the labor market (housewives), a category quite uneven in size across the countries we

6 Korpi examine. Thus, the policy-related differences among countries with respect to the likelihood of women with different levels of education (a proxy for class position) are disregarded; we argue that these differences are of relevance in assessing the gender-inequality-reducing impact of policy, and women s opportunities, including their chances to reach top wages. (Indeed, Mandel (2009, 2011) has subsequently contextualized her earlier findings and, in an argument similar to that advanced by Orloff (2006) with respect to the United States and Sweden, argues that different policies affect differently situated women differently, and that this reflects different political goals.) In comparing gender wage gaps in the United States and in Sweden, economists Albrecht, Björklund, and Vroman (2003) define glass ceilings in a country in terms of the increase in gender wage gaps when we move from the 50th to the 80th percentile of the total wage distribution. Finding a higher difference in Sweden than in the United States, they take this as evidence that glass ceilings are more problematic in the former than the latter. Moreover, Sweden s glass ceilings are seen as resulting primarily from a policy-induced negative selection of women into employment and concomitant increases in employers statistical discrimination against women. Thus Albrecht et al. (2003, 172) state: Daycare and parental programs give Swedish women a strong incentive to participate in the labor force.... At the same time benefits may discourage strong career commitment on part of the parents mostly involved in child rearing. In practice it means that women may have strong incentives to participate in the labor force but not to do so very intensively. Using similar indicators on differences in gender wage gaps between median and top levels, Arulampalam, Booth, and Bryan (2007) and Booth (2006) have reported glass ceilings in many European countries, and Datta Gupta, Smith, and Verner (2008, 80) claim that in Denmark family policies have generated welfare state-based glass ceilings. These approaches are problematic in a number of respects, including the failure to consider alternative explanations of differences in gender wage gaps over the total earnings distribution. Moreover, different factors may affect gender wage gaps at median and at top levels of earnings distributions; in countries with broad-based union involvement or legal interventions into wage setting, gender wage gaps at median and lower wage levels have been decreased. Differences in the size of gender wage gaps between median and top wage levels may primarily reflect country differences in the size of gender wage gaps at median levels but say little about country differences in wage gaps at top levels. Indeed, Mandel and Shalev (2009b, 177) conclude that while welfare states generally are in the interests of both men and women in the lower classes, their pernicious effects on statistical

Women s Opportunities under Different Family Policy Constellations 7 discrimination... are most likely to limit the attainments of more privileged women. We argue below that such negative effects are overstated and that socioeconomic differences in the selection of women into employment are as critical an issue to examine. The extent of public sector employment differs among countries; comparisons between public and private sectors in wages among women as well as men are therefore relevant for comparisons of wage distributions. Comparing seven Western countries, Gornick and Jacobs (1998) explored earnings differentials between public and private sector employees, with a focus on the extent to which public sectors provide high-paying jobs for women. In all countries, earnings compression was higher in the public than in the private sector, and public/private earnings ratios tended to be negatively correlated with public sector size. In Sweden, with a large public sector, private sector employees had higher wages than their counterparts in the public sector but intra-sectoral wage differences were greater among men than among women, a finding pointing to the major role of structural factors rather than family policies. Confirming the above observations, le Grand, Szulkin, and Thålin (2001) found that among comparable individuals, wage differences between public and private sectors were greater for men than for women. Effects of the public/private sector divide for earnings are thus likely to be consequential not only for women but also for men. Dimensions and Constellations of Family Policies Our starting premise is that family policies are multidimensional and that we should assess the impact of different policies on differently situated women and men. Other scholars, too, have argued for the multidimensionality of policy (e.g., O Connor, Orloff and Shaver 1999). Pettit and Hook (2009) argue that national policies and individuals resources shape gender inequalities and tradeoffs in the workplace and in labor markets with respect to employment, occupation, work hours, and wages. Each family-relevant policy is examined in terms of two dimensions: the extent to which it includes or excludes women from labor markets and the extent to which it affects inequality among working women. These authors underline that the term family policy conflates a variety of specific policies that may have countervailing effects on women s employment and other economic outcomes (Pettit and Hook 2009, 11). They also emphasize the interplay between individual-level characteristics and national-level conditions that influence gender inequalities in economic outcomes (175; see also Korpi 2000; Montanari 2000a; Sainsbury 1996 on multi-dimensionality of family policies). Similarly, Cooke (2011, 3 4) is less sanguine than are we about the potential of certain kinds of family policies to promote the interests of all groups of women.

8 Korpi As long-term outcomes of partisan politics reflecting the relative strength of Catholic, left, and secular center-right parties as well as of women s movements, family policy institutions in Western countries have gradually become multidimensional in terms of claim rights. While the main left right partisan cleavage runs through all Western countries, in Continental Europe it is crosscut by Catholic parties which during the entire twentieth century have pressed for family policies differing from those of secular parties on the left and on the right. This can be expected to have major consequences for patterns of inequalities. Comparative studies of the welfare state have often related their findings to some form of welfare state typology (e.g., Esping-Andersen 1990; Hicks and Kenworthy 2003; Lewis 1992; Pettit and Hook 2009). Yet for analyses of change over time and of causal processes behind cross-national differences in outcomes, the regime approach is less helpful: it is static and is based on conglomerates of driving forces, institutions, and outcomes. Moreover, it obscures central gender-based institutions and outcomes (O Connor, Orloff, and Shaver 1999; Orloff 2009). We decompose welfare states into dimensions and constellations of family policies expected to generate different patterns of gender inequalities among countries. In each country, family-relevant policies are specified in terms of separate dimensions. We replace country names and regime labels with countries values on family policy dimensions to examine effects on gender inequalities. Contrary to one-dimensional measures, our multidimensional policy indicators enable us to specify expected differences in directions as well as in strength of policy effects. They also give more precise descriptions of country differences among constellations of family policies and enable analyses of change over time, something of crucial relevance in periods when gender relations are re-structured by political and economic changes. When considering effects of family policies, it is important to differentiate between claim rights and liberties between freedom to and freedom from. Claim rights enable citizens to secure material support from public authorities in terms of cash and services facilitating gender equality; liberties remove discriminatory rules and practices. Liberties and claim rights relevant for gender equality have often developed in tandem (O Connor, Orloff, and Shaver 1999). Here we focus on claim rights, assuming that they have a more immediate impact than liberties on women s realized choices at this moment in history. Unpacking policies defined in existing legislation in areas of transfers, services, and taxes, we differentiate claim rights aimed at increasing gender equality into three specific policy dimensions likely to differ in their consequences for women s choices between paid and unpaid work: traditional-family policies, dual-earner policies, and dual-carer policies. Up to the mid-1960s our countries showed relatively muted differences in family policy (Ferrarini 2006, 59 60; Montanari 2000a). 3 Beginning in the early 1970s, however, there has been greater differentiation in claim-based

Women s Opportunities under Different Family Policy Constellations 9 family policies and different policy dimensions have emerged as well. In the Nordic countries, there was a gradual emergence of a dual-earner dimension, including policies strengthening the capabilities of women, particularly of mothers, for extensive labor force participation, by transferring major parts of care from the home to the public sector (see, e.g., Leira 1992, 2002). Indeed, Ferrarini (2006, 101) shows a consistent positive correlation between dual-earner support and female paid work in all eighteen countries, 1970 2000. The Nordic countries have also increasingly introduced policies to stimulate fathers to take a more active part in the care of their minor children, that is, a dual-carer policy dimension. Still in a nascent stage this is being taken up in limited ways in other countries (Hobson 2002). In Continental Europe, traditional-family policies were extended through claim rights supporting women s unpaid work within the home, presuming that women have the major responsibility for care at home and enter paid work primarily as secondary earners (see, e.g., Lewis and Ostner 1994; Saraceno 1994). Anglophone countries have seen only limited extensions of claim rights; instead women s liberties were extended by abolishing barriers contributing to gender segregation (O Connor et al. 1999). In the United States, such efforts have been effectively combined with legislative structures developed in civil rights struggles in attempts to realize equal opportunities (Orloff 2006). Family policy institutions are always embedded in wider social, cultural, and historical contexts, and result from different forces; policy configurations are alloys, not elements. In some countries, partisan conflicts and bargaining among parties in coalition governments have introduced contradictory elements into family policies (Ferrarini 2006; Hiilamo and Kangas 2009; Leira 2006; Lundquist 2011). Classifying a large number of gender-relevant policies identified in our countries in terms of their expected effects on the three dimensions of gender inequality, we arrive at a policy space that enables us to describe policy differences among countries in terms of their goals as well as in terms of their strength. To make policy indicators comparable across dimensions and countries, within each policy dimension we take raw observed figures on each indicator and standardize these into a distribution having an average of zero and a standard deviation of 1. Within each policy dimension we then add standardized values over the separate policy indicators to in a plot describing constellations of the different policy dimensions for our countries. The traditional-family dimension is based on a weighted average of four policy indicators: 1. Child allowances for minor children paid in cash or via the tax system (expressed as a percentage of a single workers net wage at the level of industrial workers in the country). 4 (Weight 1.0). 2. Part-time public daycare services for somewhat older children (from three years up to school age), relating numbers of places or children in care to children in the relevant age group. (Weight 1.0).

10 Korpi 3. Home care allowance to a parent for care of children below school age. (Weight 0.5). 5 4. Marriage subsidies via tax benefits to head of household having an economically non-active spouse. (Weight 0.5). Child allowances are early forms of family support likely to be neutral with respect to labor force participation of spouses. In some countries, benefits come as tax credits or tax allowances. Part-time public daycare for older children is based on the presumption that mothers are engaged in homemaking or part-time employment. Because of generally low flat-rate payments, the home care allowance tends to be chosen by the parent with the lowest earnings, typically the mother. The marriage subsidy described by Montanari (2000b) reflects differences in the net post-tax earnings between a single person and a two-person household where only one spouse is economically active; this difference is expressed as a percentage of the net average wage of a single worker. 6 Since the two first-named indicators are found in all countries and concern all families, while the latter two indicators are of relevance in fewer countries and for fewer families, we have weighted them differently. The dual-earner dimension is an unweighted average of three policy indicators: 1. Public daycare services for the youngest children (0 2 years of age), relating numbers of children in care to children in the relevant age group. 2. Full-time public daycare services for children over-threes. 3. Earnings-related parental insurance (a multiplicative variable reflecting the percentage of replacement of previous earnings and duration of benefit). 7 This dimension reflects the extent to which public polices enable a transfer of childcare from the family to the public sector enabling mothers to maintain a major and continuous occupational commitment. Provisions of daycare for the under-threes as well as full-time services for the older children are here of key importance (Korpi 2000; Rostgaard 2002). Earnings-related parental leave encourages young women to start and maintain occupational careers while enabling parents to have an interlude for the care of infants. We here use a multiplicative indicator to differentiate earnings-related parental leave from homecare allowances with low flat-rate benefits but often long duration a difference established as significant for mothers employment outcomes, and reflecting distinctive political origins in previous research (Morgan and Zippel 2003). The dual-carer dimension refers to policies to stimulate fathers to take a more active part in caring for their minor children. With earnings-related benefits, such programs cater also to men and are earmarked for fathers or permit sharing between parents. Our measure is an unweighted average of

Women s Opportunities under Different Family Policy Constellations 11 two policy indicators based on earnings-related parental and paternity insurance: 1. Weeks of paid leave which can be used either by the mother, the father, or by both. 2. Weeks of paid leave reserved for fathers. Figure 1 indicates locations of countries and country constellations within a two-dimensional space formed by the traditional-family and dual-earner dimensions. The horizontal axis reflects degree of dual-earner support and the vertical axis degree of traditional-family support in terms of the average sum of different policy indicators within each dimension, with a country s position on each dimension based on the sum of the above standardized policy indicators (cf. Methodological Appendix A). Country locations on these two dimensions are indicated by circles identified by country internet suffixes. 8 The third dimension, dual-carer support, is reflected in the relative size of the grey country circles. Up to 2000, changes in dimensions of claim rights have generated three relatively clear-cut constellations of family policies. The traditional-family Figure 1. Country Locations and Constellations on Three Family Policy Dimensions for Eighteen Countries, 2000. Size of Grey Dots Indicate Degree of Dual-Carer Dimension (Countries Are Identified by Their Internet Suffixes).

12 Korpi policy constellation has high values on traditional-family dimension but relatively low values on dual-earner and dual-carer support. This is seen in a group of countries found in the upper left corner of the figure: Austria, Belgium, France, Germany, Italy, and the Netherlands. Distinguished by the clearly highest values on the dual-earner dimension as well as on the dualcarer dimension, in the lower right corner Denmark, Finland, Norway, and Sweden form an earner-carer group. 9 The dual-carer dimension has emerged primarily in countries which pioneered the development of dual-earner dimension support; sprinklings of this dimension are also found in Canada as well as in Belgium and France. 10 The dual-earner and dual-carer dimensions are clearly correlated, indicating similar driving forces, yet having partly different effects with respect to mother s paid work and childcare within the family. Contrary to any assumption about the unidimensionality of family policies, Figure 1 shows a clearly negative relationship between earner-carer and traditional-family policy dimensions. 11 In the lower left corner, with relatively low values on all three dimensions of family policies, we find an otherwise very heterogeneous group of countries: Australia, Canada, Ireland, Japan, New Zealand, Switzerland, the UK, and the United States. They have market-oriented family policies in the sense that in the absence of strong state action to support dual-earner or traditional households, the market is the principal institution governing individuals and families access to resources. At the same time these countries are quite heterogeneous in terms of other historical and political factors relevant for gender equality, and that markets and employment discrimination can be more or less strongly regulated. While traditional-family and earner-carer countries are distinguished by their expansion of claim rights in divergent directions since the 1960s, countries in this group are characterized by the fact that they did not develop claims in either direction. Among them it is reasonable to take Canada, the UK, and the United States as something of prototypical market-oriented countries in the sense that this outcome is likely to largely reflect active political choices in the context of dominant secular center-right parties and majoritarian elections. In these countries women s movements have pressed for extension of women s liberties, especially the abolition of gender-discriminatory rules and practices and the establishment of legal institutions designed to prevent discrimination (Cooke 2011; O Connor et al. 1999; Orloff 2006). While sharing market-oriented family policies, in some key aspects the other five countries in this group distinguished by black dots in Figure 1 differ considerably from the above proto-typical market-oriented countries. In Australia and New Zealand, which share a liberal political heritage with the UK, United States, and Canada, traditions of judiciary wage-setting institutions involving public interventions in market processes have been significant (Castles 1985). Australian industrial relations commissions on federal and state levels also played important roles in narrowing gender wage gaps

Women s Opportunities under Different Family Policy Constellations 13 according to equity principles (O Connor et al. 1999; Cooke 2011). Ireland, Switzerland, and Japan appear as cases with unusual configurations of causal factors. Although the position of the Catholic Church traditionally has been very strong in Ireland, as a part of the UK up to 1922 it did not develop a Catholic party of the type found in Continental Europe. In Switzerland, federal political structures, independent cantons, and the frequent use of popular referenda have contributed to preserve a restricted family policy (Huber and Stephens 2001) and was an extreme laggard in extending women s political rights (Banaszak 1996). Market-oriented family policies are also found in Japan, dominated by a secular center-right party during the post-war period. As noted above, a central issue in the development of family policies has been whether social care should be supported as unpaid work within the home, paid work in the public sector, or paid work in the private sector. Policy making in this area has strongly been influenced by the long-term strength of political parties and normative roles of churches (Korpi 2000). Catholic parties have historically been highly influential when advancing policies supporting traditional family forms (Korpi 2006; Korpi and Palme 1998; Morgan 2006); the traditional-family dimension shows high values in six Continental European countries. 12 Among secular parties, views on gender inequalities have differed along the left right continuum. Parties on the left have long tended to be somewhat more receptive to gender equality than other major secular parties, as indicated by the relatively high proportion of women they have traditionally elected to legislatures (Tingsten 1937, chap. 1; Norris 1987, chap. 6). Secular center-right parties have held contrasting views of gender, sometimes upholding middle-class ideals of separate spheres for men and women and sometimes advancing economic opportunities for employed women. Family policies in the three country groups largely reflect the following: Catholic parties have been averse to polices increasing women s paid work; secular center-right parties have avoided extending claim rights to facilitate women s advancement; left parties have supported family policies extending citizen s claim rights transferring social care as paid work into the public sector. Class and Family Policy Interactions in Women s Employment Opportunities Drawing on our multidimensional family policy indicators, we begin by examining how interactions between class and family policies have affected selection processes bringing women into the labor force. 13 We examine interaction effects of family policy dimensions on women s labor force participation, controlling for individual-level characteristics relevant for female employment: age, education, presence of young children, and number of

14 Korpi adults in the household. 14 Lacking good indicators on social class in most comparative databases, we take women s level of formal education as a proxy for class position. Individual-level information is from LIS, where fifteen of our eighteen countries have relevant data from around 2000. 15 In multilevel regression, correlated error terms are treated not as problems but instead as contributing information enhancing our understanding of the phenomena in focus. We calculate the variance partitioning coefficient (VPC), estimating the proportion of total variance attributable to the second level, that is, the country level. 16 Analyses cover women aged 25 54. Estimates from independent variables are reported as odds ratios; for categorical variables, odds ratio indicates deviation from a reference category, here set to 1. An odds ratio of 1.2 thus indicates 20 percent higher odds when compared with the reference category; 0.8 indicates 20 percent lower odds than the reference category. The traditional-family dimension is measured as described above. 17 Due to the high correlation between dual-earner and dual-carer indicators, these two variables are here additively combined into an earner-carer dimension. Table 1 shows estimates from random intercept regressions of macro- and micro-level variables effects on women s employment. 18 Model 1, the empty model, includes only macro-level indicators in terms of traditionalfamily support and earner-carer support. Here the VPC is 0.076, indicating that 7.6 percent of the total variation in female employment appears at the country level, with the remaining proportion of variance consequently found among individuals. In multilevel analysis on this type of data, a VPC of around 8 percent can be considered as large. The earner-carer dimension is clearly associated with increasing probability for female employment and the traditional-family dimension is also reflected in somewhat decreasing female employment. Formally, odds ratios here indicate that on average an increase of 10 percentage-points in the earner-carer dimension would lead to an increase of female employment of 1.8 percentage-points; the same increase in the traditional-family dimension would decrease female employment with 0.6 percentage-points. Model 2 includes micro-level variables only. As could be expected, both lower education and presence of pre-school children are related to lower female employment. Age has a curvilinear relationship, indicating that the highest employment proportion is in the middle age group (35 44 years). Moreover, employment proportions increase with the number of adults in the household, pointing to lower employment among single mothers. In Models 3 6, effects of the two macro-level family-policy variables are examined together with each one of our four micro-level factors; results indicate that effects of family policy variables do not substantially change when including micro-level factors. In Model 7, all factors are included simultaneously, showing roughly similar correlations between macro-level variables and female employment as found in Model 1. In Model 8, cross-level

Table 1. Family Policy Dimensions and Women s Employment. Odds Ratios from Random Intercept Logistic Regressions of Macro- and Micro-Level Determinants in Fifteen Countries around 2000 (Women 25 54 years) Independent variables Models and Odds Ratios (p-values in Parentheses) 1 2 3 4 5 6 7 8 Traditional-family dimension 0.94 (0.077) 0.95 (0.188) 0.93 (0.016) 0.93 (0.001) 0.94 (0.000) 0.95 (0.030) 0.96 (0.306) Earner-carer dimension 1.18 (0.000) 1.16 (0.000) 1.18 (0.000) 1.20 (0.000) 1.20 (0.000) 1.19 (0.000) 1.15 (0.000) Education High (ref) # # # # Medium 0.64 (0.000) 0.65 (0.000) 0.64 (0.000) 0.62 (0.000) Low 0.25 (0.000) 0.27 (0.000) 0.25 (0.000) 0.25 (0.000) Age 25 34 (ref) # # # # 35 44 1.10 (0.000) 1.23 (0.000) 1.10 (0.000) 1.10 (0.000) 45 54 0.92 (0.000) 1.16 (0.000) 0.92 (0.000) 0.92 (0.000) Children No child (ref) # # # # Child 1 2 0.37 (0.000) 0.44 (0.000) 0.37 (0.000) 0.37 (0.000) Child 3 5 0.55 (0.000) 0.61 (0.000) 0.55 (0.000) 0.55 (0.000) Child 6 þ 0.89 (0.000) 0.93 (0.000) 0.89 (0.000) 0.89 (0.000) No of Adults 1.02 (0.000) 0.99 (0.394) 1.02 (0.002) 1.13 (0.002) Continued Women s Opportunities under Different Family Policy Constellations 15 Downloaded from http://sp.oxfordjournals.org/ at Pennsylvania State University on February 20, 2016

Table 1. Continued Independent variables Models and Odds Ratios (p-values in Parentheses) 1 2 3 4 5 6 7 8 Interactions Trad educ. medium 0.96 (0.000) Trad educ. low 1.01 (0.328) Dual educ. medium 1.02 (0.009) Dual educ. low 1.01 (0.097) N (level 1) 133 014 133 013 133 014 133 014 133 014 133 013 133 013 133 013 Level 2 variance 0.09 (0.032) 0.25 (0.093) 0.09 (0.033) 0.09 (0.033) 0.12 (0.021) 0.11 (0.037) 0.13 (0.024) 0.10 (0.033) (Standard error) Log likelihood 2 78 424.12 2 73 924.49 2 75 491.89 2 78 320.87 2 77 199.28 2 78 404.07 2 73 917.83 2 73 893.93 Empty model N (level 1) 133 014 Level 2 variance 0.27 (Standard error) (0.100) Log likelihood 2 78 432.71 VPC (V/(V þ p 2 /3)) 0.076 16 Korpi Downloaded from http://sp.oxfordjournals.org/ at Pennsylvania State University on February 20, 2016

Women s Opportunities under Different Family Policy Constellations 17 Figure 2. Interactions Between Women s Education and Family Policy Dimensions on Probability for Women s Employment, by Three Levels of Education (Values on Family Policy Variables Range from Lowest to Highest Country Observations). Based on Table 1, Model 8. interactions are introduced between family policy dimensions and educational levels with high education as a reference point. Since interaction effects are often difficult to interpret from parameter estimates, Figure 2 graphically illustrates interaction effects on probabilities of women s employment for each policy variable at different educational level as predicted from the above regression. We see that according to such estimates, an increase in the earner-carer dimension is associated with higher female employment in all three educational groups; the somewhat weaker gradient for women with tertiary education implies that the earner-carer dimension has less relevance at the tertiary level than at medium levels of education (the coefficient for dual low education is insignificant although in the right direction). The traditional-family policy dimension has the opposite relationship to female employment; increases in this dimension are associated with decreasing female employment probabilities at all educational levels. As indicated by the somewhat sharper decline for women with medium education, the traditional-family dimension could be most important for women at this level of education, which may reflect class-specific work-family conflicts. Interactions between family policy dimensions capture main tendencies but also obscure important cross-country variation within family policy constellations. One way to illustrate interactions between family policies and individuals education is to examine separate country residuals. A focus on country residuals enables us to compare differences between country

18 Korpi Figure 3. Residuals in Fifteen Countries from Multilevel Regressions of Individual Characteristics on Women s Employment in Traditional-Family, Market-Oriented, and Earner-Carer Family Policy Constellations of Family Policies, by Three Levels of Women s Education (Based on Separate Analyses for Each Level of Education). employment rates as predicted only on the basis of individual-level variables and actually observed employment rates. The size and sign of country residuals can thus be interpreted as interaction effects of macro-level family policies. In this way, we can locate countries where, net of individual characteristics, family policies have tended to increase or decrease female employment. Figure 3 shows country residuals from separate multilevel regressions for each level of education after controls for other individual-level characteristics, that is, log odds for female employment in different educational categories. 19 Within each family policy group, countries are ordered according to residuals for women with low education. Of interest here is to examine where, in the context of the three family policy constellations, differing values on residuals are found. The location of country residuals largely follows what can be expected from family policy structures, with some differences within groups. Among countries classified as sharing earner-carer policies, ten of two educational categories have employment residuals above average; Sweden and Denmark have the highest positive residuals, which may reflect their intensive earnercarer policies. Finland here shows below average residuals for women with secondary and tertiary education, possibly reflecting that it has partly contradictory policies with somewhat less earner-carer support and more traditional-family support than other countries in the category. While not attempting to interpret patterns of residuals in all countries, we note that in the traditional-family category of countries, fifteen of eighteen educational

Women s Opportunities under Different Family Policy Constellations 19 groups have below average employment rates among women, with especially low levels appearing in Italy and the Netherlands. As discussed above, in the heterogeneous grouping of countries with market-oriented policies, we distinguished Canada, UK, and the United States as having prototypical marketoriented policies. The United States and Canada have below average residuals for all educational categories while the opposite is found in UK. In Figure 3, two countries sharing market-oriented family policies without being prototypical Australia and Ireland are distinguished by shaded columns. While Australia s above average residuals may be related to specific wage settings structures, Ireland deviates by having lower than average participation rates, especially for women with low education. Ireland has similarities with Italy and the Netherlands, two other countries where religious influences have been strong. Of course, family policy institutions may have affected some micro-level factors in the multi-level regressions, especially the number of children in different countries. Cross-national differences in the timing and structure of female employment increases may also pattern employment over age cohorts. Family Policies and Women s Top Wages In testing the hypothesis that family policies affect women s opportunities to achieve top wages, we must consider cross-national differences in the socio-economic selection of women into employment. We use multilevel regressions on LIS data to evaluate relationships between family policy characteristics and women s individual-level probabilities for attaining high-wage positions, beginning with macro-level family policy dimensions in an empty model to examine their relevance for country differences in proportions of women with top earnings. Thereafter we examine country residuals in terms of positive or negative deviations between observed values and expected values when estimates are made only on the basis of women s individual characteristics. Samples and estimation models as well as policy and individual variables are the same as in the analyses on women s employment reported above. Since information on hours worked are missing for several countries, we follow Mandel and Semyonov (2005) and use annual earnings to calculate our dependent variable female representation in the top annual earnings quintile. Table 2 shows that VPC for the empty Model is low; only 3.1 percent of the variation in female top wage representation is explained at the country level, a considerably lower proportion than that found for female employment. In Model 1, the earner-carer indicator is not statistically significant, while the traditional-family dimension is clearly linked to lower odds of women reaching the highest earnings quintile. When introduced together with institutional variables, micro-level factors have only limited influence

Table 2. Family Policy Dimensions and Women s Top Quintile Wages. Odds Ratios from Random Intercept Logistic Regressions of Macro- and Micro- Level Determinants in Fifteen Countries around 2000 (Women 25 54 years). Independent variables Model odds Ratios (p-values in parenthesis) 1 2 3 4 5 6 7 8 Traditional-family 0.94 (0.017) 0.94 (0.144) 0.94 (0.029) 0.94 (0.012) 0.94 (0.007) 0.95 (0.179) 0.91 (0.042) dimension Earner-carer dimension 1.01 (0.613) 0.96 (0.239) 1.01 (0.705) 1.01 (0.597) 1.02 (0.183) 0.95 (0.213) 0.90 (0.013) Education High (ref) # # # # Medium 0.23 (0.000) 0.23 (0.000) 0.23 (0.000) 0.22 (0.000) Low 0.07 (0.000) 0.07 (0.000) 0.07 (0.000) 0.05 (0.000) Age 25 34 (ref) # # # # 35 44 2.00 (0.000) 1.68 (0.000) 2.00 (0.000) 2.01 (0.000) 45 54 2.20 (0.000) 1.95 (0.000) 2.20 (0.000) 2.23 (0.000) Children No child (ref) # # # # Child 1 2 0.53 (0.000) 0.46 (0.000) 0.53 (0.000) 0.54 (0.000) Child 3 5 0.65 (0.000) 0.59 (0.000) 0.65 (0.000) 0.66 (0.000) Child 6 þ 0.71 (0.000) 0.76 (0.000) 0.71 (0.000) 0.72 (0.000) Adults 0.83 (0.000) 0.83 (0.000) 0.83 (0.000) 0.83 (0.000) 20 Korpi Downloaded from http://sp.oxfordjournals.org/ at Pennsylvania State University on February 20, 2016