Cultural and Institutional Factors Shaping Mothers ...

Social Forces Advance Access published December 8, 2014

Mothers’ Employment in Wealthy Countries

1


Cultural and Institutional Factors Shaping Mothers’ Employment and Working Hours in Postindustrial Countries

E

xisting research shows that women’s employment patterns are not driven so much by gender as by motherhood, with childless people and fathers employed at substantially higher levels than mothers in most countries. We focus on the cross-national variation in the gap in employment participation and working hours between mothers and childless women. Controlling for individual- and household-level factors, we provide evidence that institutional and cultural contexts shape maternal employment. Well-paid leaves, publicly supported childcare services for very young children, and cultural support for maternal employment predict smaller differences in employment participation and working hours between mothers and childless women. Yet, extended leave, notably when unpaid, is associated with larger motherhood employment gaps. Mothers’ employment has sparked many debates over the past decade. In the US popular press, Lisa Belkin’s (2003) “The Opt-Out Revolution” raised questions about mothers’ ability to maintain careers, Anne-Marie Slaughter’s (2012) essay “Why Women Still Can’t Have It All” emphasized the challenges faced by working mothers, while Sheryl Sandberg’s (2013) book Lean In suggests how mothers can and should remain engaged in employment. Academic research analyzing employment participation of women and mothers similarly reflects these concerns (Boushey 2008; Damaske 2011; England 2010; Goldin 2006; Jones 2012; Percheski 2008; Stone 2008; Williams 2000). What factors support or limit maternal employment? And to what extent does work-family conflict

The authors are indebted to Karen Mason, and the staff of the Cross-National Data Center in Luxembourg for their support. They wish to thank the editor, Arne Kalleberg, the Social Forces reviewers, Janet Gornick, Catherine Bolzendahl, and Monica DasGupta, for their comments on drafts of this paper. Earlier versions of this paper were presented at the 2012 Annual Meetings of the American Sociological Association and the 2013 Meeting of the Population Association of America. The authors gratefully acknowledge the financial support by the National Science Foundation (Grants #0600926, #0751505). Direct correspondence to Irene Boeckmann, WZB Berlin Social Science Center, Reichpietschufer 50, 10785 Berlin, Germany; E-mail: [email protected]. © The Author 2014. Published by Oxford University Press on behalf of the University of North Carolina at Chapel Hill. All rights reserved. For permissions, please e-mail: [email protected].

Social Forces 00(00) 1–33, Month 2014 doi: 10.1093/sf/sou119

Downloaded from http://sf.oxfordjournals.org/ at Univ. of Massachusetts/Amherst Library on January 16, 2015

Irene Boeckmann, WZB Berlin Social Science Center Joya Misra, University of Massachusetts–Amherst Michelle J. Budig, University of Massachusetts–Amherst

2 Social Forces


reflect cultural and structural barriers for women to combine employment and childrearing? We analyze 19 countries to examine how leave policies, childcare services, and cultural norms regarding maternal employment shape differences in employment participation and working hours between mothers and childless women. Although maternal employment has risen cross-nationally, gains are uneven (England 2006; Lewis 2009; Rubery, Smith, and Fagan 1999; Tranby 2008), and sometimes reflect growth in part-time employment. As a result, substantial crossnational variation in mothers’ employment rates and working hours remains (Gornick, Meyers, and Ross 1997; Stier, Lewin-Epstein, and Braun 2001). We address several challenges in the literature. First, the focus on the gender gap (e.g., Blau and Kahn 2013; Mandel and Semyonov 2006; Pettit and Hook 2009) disguises inequalities based on motherhood. Despite women’s inroads into employment and men’s increasing participation in childcare, women remain primarily responsible for children. Thus, motherhood is an axis of inequality central to our understanding of the processes that shape women’s employment patterns. While research on the motherhood wage penalty (Anderson, Binder, and Krause 2003; Budig and England 2001; Budig and Hodges 2010, 2014; Budig, Misra, and Boeckmann 2012; Waldfogel 1997) foregrounds this important point, research on women’s employment focuses less on differences among mothers and childless women (Correll, Benard, and Paik 2007; Gornick, Meyers, and Ross 1997; Pettit and Hook 2009). By comparing differences between women with and without children in the home, we highlight the relationship between motherhood and women’s employment. Second, cross-national variation in women’s employment reflects different policy and cultural contexts. Countries have instituted measures aimed at addressing work-family conflict. These policies contain different gendered assumptions about women’s and men’s paid and unpaid work, and care for children. For example, extended childcare leaves with low levels of benefits may support parental (and more implicitly maternal) care of children in the home (Brandth and Kvande 2009; Morgan and Zippel 2003), and assume that leave-takers are supported by other income sources (e.g., the income of a male breadwinner). On the other hand, publicly provided childcare services may foster women’s employment, and support varying family forms. Thus, not all work-family policies support maternal employment equally (Korpi, Ferrarini, and Englund 2013; Lewis 2006). While policy packages in some countries lend primary support to one model of gender division of labor (e.g., West German policies tended to support male breadwinner/female part-time care-provider families),1 in other countries leave policies and childcare availability may support multiple models of organizing paid and unpaid work (e.g., in France, widely used public childcare coexists with long childcare leaves). Gendered cultural norms may also shape inequalities based on motherhood (Crompton 1999; Kremer 2007; Pfau-Effinger 2004). Using multilevel models, we investigate how specific workfamily policies and cultural attitudes are related to maternal employment, controlling for individual and household factors and recognizing that policies and cultural attitudes are embedded in a wider context.


Motherhood, Employment, and Work Hours Cross-Nationally Previous research documents the cross-national variation in women’s and mothers’ employment (Pettit and Hook 2009). Highlighting employment differences among women based on motherhood, figure 1 summarizes childless women and mothers’ employment rates and average weekly working hours for women aged 25–45. The clustering of the gray data points shows that childless women’s Figure 1. Percentage of women aged 25 to 45 employed and average weekly working hours by motherhood status

Average Weekly Hours Among the Employed

45

RU

HU

40

CZ IL

RU

CZ US

ES US

IL

35

ES IT

FR CA

IT

IE

AU

GEE

CA

LU

GEE

BE AU

AT

BE

GEW

IE

FR

NL

SE

LU

30

UK

HU

SE

AT

UK

GEW

25

NL

20

.40

.50

.60 .70 .80 Employment Participation Rates Childless women

.90

1.00

Mothers

Source: Authors’ own calculations based on data (wave V) from the Cross-National Data Center in Luxembourg


Finally, previous studies vary in how they measure and conceptualize women’s employment. Analyses of employment rates may not recognize that high levels of women’s employment may mask very low weekly employment hours (e.g., the Netherlands). Similarly, a focus on employment hours may miss that in some countries, relatively few women are employed (e.g., Italy and Spain). To advance the literature on employment, we examine the impact of policies and cultural measures on both employment participation and employment hours among employed women. We begin by mapping the differences in employment and work hours by motherhood cross-nationally. Subsequently, we discuss institutional and cultural explanations for these differences.

3

4 Social Forces

Work-Family Policies: Divergent Impacts on Women’s Employment One set of institutional explanations focuses on how work-family policies may affect women’s employment opportunities. In light of ever-decreasing fertility rates, labor-market and social policies in European countries have focused on alleviating work-family conflict. These policy responses reflect concerns that welfare states are sustainable only with high levels of employment and a sufficiently large workforce (Esping-Andersen et al. 2002; Kenworthy 2008).


employment rates and weekly hours vary far less cross-nationally compared to mothers’ employment rates and hours. At least 70 percent of childless women are employed in all countries, with weekly hours ranging from 38 in Sweden to 43 in Russia, partly reflecting countries’ working-time regulations. Even in Italy and Spain, where women’s overall labor-force participation is low, childless women’s employment participation rates exceed 70 percent. The scatter of the black diamonds illustrates the dramatic cross-national differences in maternal employment. Russia and Sweden lead with 81 and 86 percent; Spain and Italy trail with around 45 percent of mothers in employment. While maternal employment rates may reflect part-time work, part-time work can mean very different levels of engagement. For example, Belgian and Dutch mothers participate in the labor market at about the same rates, yet Dutch mothers average shorter hours (21 hours) compared to Belgian mothers (32 hours). We see shorter maternal working hours in countries where policies and cultural norms favored the “male breadwinner/female care provider” gender division of labor, notably West Germany and the Netherlands, where part-time employment became an accepted form of employment for mothers to combine work and family (Crompton 1999; Kolbe 1999). These patterns contrast with the Southern European countries, where few mothers are employed, but if employed, they likely work full-time. On the other hand, despite long parental leaves and the erosion of childcare provision in Eastern European countries after 1989, employed mothers in these former “dual-earner/state-carer” countries continue to work full-time (Crompton 1999). In the “Liberal” welfare states of Canada and the United States, where non-parental childcare tends to be marketized rather than state supported (Esping-Anderson 1990; Morgan 2005; O’Connor, Orloff, and Shaver 1999), we see moderate maternal employment rates with relatively long weekly hours, whereas in the “Liberal” UK and Australia, we find lower maternal employment and higher rates of part-time work. Figure 1 highlights that welfare-state typologies only partially account for the cross-national variation in maternal employment, given the variation within countries commonly grouped together (O’Connor, Orloff, and Shaver 1999). What factors account for this variation? Some scholars focus on the role of structural factors, such as work-family policies, taxation, and economic conditions (Eliason, Stryker, and Tranby 2008; Mandel and Semyonov 2006; Pettit and Hook 2005, 2009; Stier, Lewin-Epstein, and Braun 2001), while others emphasize how cultural contexts interact with structural factors (Auer 2002; Kremer 2007; Pfau-Effinger 1996, 2004).


Hypothesis (1a): Well-paid, job-protected parental leave of moderate duration should help mothers keep their attachment to the labor force and therefore correlate with higher maternal employment and higher maternal working hours. Hypothesis (1b): No leave entitlements and very long leaves may weaken mothers’ labor-force attachment and should be associated with lower maternal employment and lower working hours. Other work-family policies may promote both maternal employment and work hours when children are young. Childcare services with opening hours


Policies such as maternity leave, parental leave, and childcare provisioning have shaped women’s employment. As Guerrina (2002, 63) notes, reconciliation policies target women “despite the artificial gender neutrality enshrined in the language.” Lewis’s (1992) early formulation grouped countries as strong, modified, or weak male-breadwinner countries, with associated differences in women’s employment rates. Studies that explore the relationship between women’s employment and favorable welfare-state contexts, including work-family policies, mostly argue for a positive relationship between generosity of policy and employment effects (Daly 2000; Gauthier 1996; Gornick, Meyers, and Ross 1997; Gornick and Meyers 2003; Kenworthy 2008; Korpi 2000; O’Connor, Orloff, and Shaver 1999; Orloff 2002; Stier, Lewin-Epstein, and Braun 2001). We consider how specific policies shape employment inequalities based on motherhood, as we believe that it is important to tease out differing effects of these policies. Maternity and parental leave policies may impact both women’s labor-market attachment and employers’ preferences. To the extent that job-protected leaves facilitate employment reentry after a childrearing break, these leaves may maintain mothers’ labor-force attachment. Well-paid, short parental leaves (less than one year) enable maternal care of infants, without risk to their jobs. However, long or poorly compensated leaves, often geared toward maintaining maternal caregiving at home, may have a paradoxical effect, dampening women’s employment and weakening their opportunities in the labor market (Bainbridge, Meyers, and Waldfogel 2003; Kenworthy 2008; Lewis 2006; Morgan and Zippel 2003; Pettit and Hook 2005, 2009; Rønsen and Sundström 2002; Tranby 2008). This may result in less secure, part-time positions post-leave. In addition to human capital losses, such as foregone experience and job seniority, very long parental leaves might reinforce employers’ expectations that mothers will spend long periods outside the labor market. Employers may be less likely to hire, promote, or otherwise support workers who are likely to leave for long periods, and may reroute mothers to part-time jobs or dismiss mothers upon return after the job-protected period (Glass and Fodor 2011). Moreover, mothers on very long parental leaves may experience eroded network ties or fall behind on keeping skills current. Indeed, Ondrich et al. (2000, 2003) found that the extensions of parental leave entitlements in Germany in the 1980s and 1990s were associated with longer post-birth employment breaks, and more foregone experience. This leads us to the following expectations for curvilinear effects of leave on employment, and weekly working hours:

5

6 Social Forces

Hypothesis (2): Public childcare provisioning should help mothers keep their attachment to the labor force; thus, public childcare should be positively correlated with both maternal employment and longer hours. In a twist in this literature, Mandel and Semyonov (2006) examine women’s labor-force participation and part-time employment. They find that “well- developed” welfare states (defined by maternity leave policies, childcare, and public-sector employment) have higher employment rates, but also more parttime employment. In supplementary analyses, they note that women in generous welfare states reduced their hours of employment. We contend that not all generous welfare states are the same, and the types of policies, such as parental leave policies and publicly funded childcare, and the design of these policies have different implications for maternal employment and work hours. We separate policy indicators for parental leave and childcare to assess which policies encourage or deter maternal employment and work hours. However, we recognize that policies are embedded in broader socio-political contexts. In robustness analyses, we therefore examine relationships between leave policies and the motherhood employment and hours gaps, controlling for childcare availability, and in turn include leave policy measures when examining childcare availability.

The Importance of Cultural Factors in Shaping Employment Levels Institutional explanations may not fully explain the variation in women’s employment outcomes. For example, British mothers have access to somewhat better leave policies (18 weeks of paid maternity leave and 13 weeks of unpaid leave) compared to American mothers (12 weeks unpaid leave for some workers, but no federal statutory paid-leave entitlement).2 Yet, British mothers have lower levels of employment. Cultural factors may illuminate these patterns. Kremer (2007) suggests that welfare states promote certain “ideals of care,” which define both what good care is and who provides it; these ideals are also embedded in welfare-state policies. Pfau-Effinger (1996, 2004) similarly argues that women’s employment must be read in relation to the gender culture (values), the gender order (institutional arrangements), and the gender arrangements


corresponding to regular working hours have positive effects on women’s employment (Korpi 2000; Lewis 2009; Pettit and Hook 2005, 2009; Stryker and Eliason 2004). Yet, quality childcare is costly and may exceed the potential wages parents might earn. Publicly subsidized and public-sector universal childcare reduces parental costs for childcare. Publicly provided childcare, particularly for very young children (aged 0–2), is associated with higher women’s employment rates (Pettit and Hook 2005, 2009; Tranby 2008). While childcare available through markets may encourage women’s employment, childcare costs have a significant negative impact on mothers’ labor supply (Han and Waldfogel 2001; Powell 2002). We model these effects, exploring the association of childcare with both maternal employment and work hours, to provide more nuanced understandings of these processes.


7

(gender divisions of labor in the home). Indeed, Budig, Misra, and Boeckmann (2012) show that work-family policies are associated with higher maternal earnings in contexts where cultural support for maternal employment is high, but have less positive or even negative relationships where cultural ideals reflect maternal care and paternal breadwinning.

Other Institutional and Economic Factors Shaping Employment Levels Finally, explanations for variation in women’s employment rates cross-nationally may include a variety of other institutional and economic conditions, such as tax policies, business cycle, economic performance, regulation of normal working hours, promotion of part-time work, and public-sector employment (Eliason, Stryker, and Tranby 2008; Gornick and Heron 2006; Gornick and Jacobs 1998; Huber and Stephens 2000; Pettit and Hook 2005, 2009; Tranby 2008). Higher taxation of second earners’ incomes in dual-earner families may be a disincentive to women’s (full-time) employment. Indeed, cross-nationally, higher taxes on second earners’ incomes (compared to single earners) have been associated with lower women’s employment and full-time work (Jaumotte 2003). However, studies of tax reforms in different countries show complex relationships between income taxation and employment, with uneven effects across different types of households (e.g., Francesconi, Rainer, and van der Klauuw 2009).3 Unemployment should depress women’s employment rates (though its effects on working hours are less clear), while economic performance should stimulate it. Public-sector employment, especially public-sector service delivery, is often filled by women, and therefore associated with higher women’s employment (Eliason, Stryker, and Tranby 2008; Huber and Stephens 2000; Tranby 2008). While we do not focus on how these economic and structural factors mediate differences in employment participation and working hours among women, we do control for economic performance (GDP per capita), the size of the public sector, men’s unemployment rates, and second earner’s income taxation to examine whether the relationships between maternal employment, policies, and cultural norms are robust under different economic and structural conditions. Finally, mothers’ work hours and decisions about whether to enter the labor market may be shaped by the number of hours (full-time) workers are expected to spend at work, and the availability of part-time work. Countries in our sample differ in the extent to which they promoted part-time work as a work-family reconciliation strategy, and for other goals (Fagnani and Letablier 2004; Gornick and Heron 2006; Jenson and Sineau 2003). In robustness analyses, we control for normal weekly hours and countries’ part-time employment rates.4


Hypothesis (3): Ideals regarding maternal employment will condition mother’s employment, and the number of hours worked by women. Where support for maternal employment is high, mothers will be more likely to be employed and work longer hours.

8 Social Forces

Individual- and Household-Level Controls

Data and Methods We use multiple data sources. Individual-level data come from the CrossNational Data Center in Luxembourg (LIS). LIS harmonizes national labormarket survey data. With one exception, we use data from around the year 2000 (wave V) for 18 countries.5 We examine Eastern and Western Germany separately, due to persistent differences in employment patterns and policy legacies (Rosenfeld, Trappe, and Gornick 2004), resulting in 19 cases included in the analysis. Table 1 presents the original data sources and sample sizes for each country. For the models estimating the gap in employment, the sample consists of women aged 25 to 45 (prime years for childrearing), who are neither in the military nor self-employed (subsample I).6 For the models estimating


Although we are primarily interested in the institutional and cultural factors outlined above, we also include controls for individual- and household-level factors associated with employment decisions. This ensures that the cross-national variation we try to explain is not due to differences in individual- and household-level factors, including human capital, such as education (Cogan 1980; Heckman 1974; Leibowitz and Klerman 1995; Morgenstern and Hamovitch 1976; Powell 2002) and labor-market experience (Heckman 1974, 1980; Henkens, Meijer, and Siegers 1993; Lehrer 1999; Powell 2002). We control for educational attainment, though we are unable to control for labor-market experience with our data, controlling instead for age, which, to some extent, reflects labor-market experience. Age also takes into account that older mothers who are more established in the labor market may find it easier to take time off from paid work to care for children (Ondrich et al. 2003), and cohort differences in employment participation. At the household level, partnered women’s decisions to be wage-earners may be based in joint economic calculations with their partners, taking into account each partner’s mix of human capital and preexisting gender differentials in pay in the relevant labor market (Becker 1981; Verbakel and de Graaf 2009). Yet, not all women are partnered, and partnership may have varied effects, depending on the partner’s resources (Abroms and Goldscheider 2002). Having a partner who earns more curbs possible financial pressures for partners to work (Abroms and Goldscheider 2002; Bernasco, de Graaf, and Ultee 1998; Cogan 1980; Heckman 1980; Henkens, Meijer, and Siegers 1993; Lehrer 1999; Leibowitz and Klerman 1995; Morgenstern and Hamovitch 1976; Powell 2002; Schultz 1980; Verbakel and de Graaf 2009), while transfer income from the state also may affect women’s employment participation and work hours (Flood, Hansen, and Wahlberg 2004; Schultz 1980). In addition to individual-level control variables, we control for partnered status, other household labor income, and non-family-related transfer income. Once we control for these factors, do we still find significant motherhood employment and working hours gaps? If so, do institutional and cultural explanations help explain at least part of the remaining variation?

Household Expenditure Survey

Survey on Household Income and Wealth

Socio Economic Panel

Socio-Economic Panel

Russia Longitudinal Monitoring Survey

European Community Household Panel

Income Distribution Survey

Israel

Italy

Luxembourg

Netherlands

Russia

Spain

Sweden

2000

2000

2000

1999

2000

2000

2001

2000

17,164

8,181

4,000

1,602

1,198

2,011

973

2,307

2,299

889

592

3,167

926

3,588

8,964

9,745

994

770

2,267

Subsample I 25–45-year-olds

12,434

5,614

3,586

836

883

1,571

638

1,188

1,408

564

334

2,164

717

2,662

6,792

7,887

773

581

1,450

Subsample II Employed 25–45-year-olds


Current Population Survey

2000

Living in Ireland Survey / ECHP

Ireland

1999

2000

United States

Household Monitor Survey

Germany West

Hungary

2000

2000

1999

German Social Economic Panel Study

Germany East

1996

2000

2000

2000

2001

Survey year

United Kingdom Family Resources Survey

Household Budget Survey

German Social Economic Panel Study

France

Survey of Labour & Income Dynamics

Czech Microcensus

Panel Study of Belgian Households

Belgium

Czech Republic

European Community Household Panel (ECHP)

Austria

Canada

Survey of Income and Housing Costs

Original data source

Australia

Country

Table 1. Origins of Individual-Level Data and Sample Sizes

Mothers’ Employment in Wealthy Countries 9

10 Social Forces


the motherhood gap in employment hours, we further restrict the sample to employed women (subsample II). Individual-level control variables include relationship status (= 1 cohabiting/ married, = 0 if “single”), respondent’s age (in years), educational attainment, other household income (total household earnings minus respondent’s earnings), and non-family transfer income. We convert all monetary variables into 2000 US dollars. We measure educational attainment with three dummy variables indicating high educational attainment (specialized vocational education and no less than university/college education), medium educational attainment (secondary general or vocational education and postsecondary education), and lower educational attainment (compulsory education, basic/initial vocational education or less; reference category) based on the 1997 UNESCO International Standard Classification of Education. We present means and standard deviations for all individual-level variables by motherhood status in table 2. We use the appropriate tests to test for significant group differences. With the exception of Russia, Sweden, and Hungary, mothers are significantly less likely to be employed. Among the employed, mothers average fewer work hours relative to childless women, although this difference is not significant in Russia or Hungary. The size of the differences in employment rates and hours vary considerably across countries: In Luxembourg, Australia, West Germany, Ireland, and Spain, mothers’ employment rates are between 28 and 35 percentage points lower than childless women’s rates, while we find differences of 10 percentage points or less in Belgium, Hungary, Sweden, and Russia. Similarly, differences in weekly work hours range from 10 or more hours in Luxembourg, Britain, the Netherlands, and West Germany to less than two hours in the Czech Republic and Russia. Across countries, mothers are older, more likely to be partnered, but less educated than childless women. And mothers tend to live in households with greater earnings and greater transfer income. Our policy measures, presented in table 3, are taken from our Work-Family Policy Indicators database (Boeckmann, Budig, and Misra 2012). The p olicy measures are lagged two years prior to the LIS survey year.7 We focus on p ublicly supported childcare for children aged 0–2 and 3–5, and leave policies. We distinguish between birth-related maternity leave, generally reserved for women, and parental leave available to both parents. We capture five different aspects of leave schemes: Leave generosity, that is, the number of fully paid weeks of leave, is measured as the number of paid weeks of leave available to women multiplied by the level of benefits (maternity and parental leave combined). Leave length captures the maximum number of weeks of (paid and unpaid) job-protected leave available to women. Furthermore, we interact the length of parental leave available to women with the level of parental leave benefit to examine whether the availability and generosity of benefits matters for the impact of the length of leave on employment inequalities among women, transforming flat-rate benefits into a percentage of average production workers’ wages (APW). Finally, we test whether the relationship between leave policies and motherhood gaps is sensitive to the inclusion of a measure that captures whether women on parental leave may work parttime while receiving leave benefits (Fagnani 1999; Gornick and Meyers 2003).8

Ireland

Hungary

Germ. W.

Germ. E.

France

Czech R.

Canada

Belgium

Australia

Partnered relationship status Age

High educ. attainment

Medium educ. attainment

Low educ. attainment

.88

(.33)

(.49)

(.43)

(.46)

.60

.77

(.28)

(.49)

.70

.91

(.32)

(.42)

.62

.89

(.36)

(.46)

.78

.84

(.28)

(.44)

.69

.91

(.37)

(.45)

.75

.84

(.35)

(.43)

.73

.86

(.34)

(.50)

.76

.87

(.31)

(.50)

.56

.90

.72

33.0

(6.8)

39.0

(9.5)

40.1

(9.1)

36.5

(7.2)

42.1

38.7

(10.7)

28.2

(9.6)

39.7

(8.4)

37.5

(8.7)

41.9

(13.2) (11.2)

24.3

(12.1) (12.1)

36.3

(9.7)

33.0

(6.2)

40.7

(11.6)

37.0

(.29)

.91

(.27)

.92

(.30)

.90

(.34)

.86

(.33)

.87

(.33)

.88

(.36)

.85

(.31)

(10.4) (10.8)

33.4

.89

(.42)

.78

(.30)

.90

38.3

32.2

(16.9) (15.7)

15.5

(11.0)

29.1

(.49)

.62

(.47)

.68

(.47)

.67

(.48)

.64

(.49)

.59

(.48)

.64

(.49)

.62

(.47)

.68

(.49)

.62

(.49)

.59

(5.3)

37.3

(5.8)

36.3

(5.2)

36.7

(5.3)

37.3

(5.5)

36.5

(5.9)

35.7

(5.5)

36.6

(5.1)

37.1

(5.2)

35.9

(5.6)

36.7

(5.8)

32.5

(6.9)

34.9

(5.9)

33.5

(.38)

.17

(.39)

.18

(.40)

.20

(.48)

.36

(6.7)

(.43)

(6.4)

.25

(.28)

.09

(.37)

.17

(.49)

.42

(.36)

.15

(.29)

.09

33.2

32.6

(7.2)

34.9

(6.5)

33.7

(6.2)

33.0

(5.6)

33.0

(6.5)

33.5

(.50)

.48

(.44)

.26

(.46)

.31

(.49)

.41

(.50)

.49

(.37)

.16

(.46)

.30

(.50)

.56

(.46)

.31

(.43)

.25

(.49)

.41

(.47)

.33

(.49)

.61

(.49)

.58

(.50)

.45

(.49)

.39

(.46)

.70

(.48)

.36

(.46)

.31

(.46)

.71

(.47)

.33

(.47)

.32

(.49)

.58

(.50)

.55

(.47)

.33

(.48)

.38

(.49)

.61

(.47)

.32

(.46)

.31

(.47)

.67

.19 (.39)

(.49)

(.50)

.42

(.30)

.10

(.19)

.04

(.38)

.18

(.50)

.46

(.28)

.09

(.32)

.11

(.49)

.38

(.28)

.08

.42

(.50)

.49

(.40)

.19

(.24)

.06

(.46)

.31

(.50)

.52

(.34)

.13

(.41)

.21

(.50)

.54

(.40)

.20

Moms Childl. Moms Childl. Moms Childl. Moms Childl. Moms Childl. Moms Childl. Moms Childl.

Weekly working hours

(4,422)

(16454) 13,668

1,953

(15,475) (13,755) (5,648) 3,861

16,220

296 1,755 (3,826)

(915) 4,733

(3,727) 8,934

(12,870) (10,686) (4,843) 11,150

1,443

(13,909) (12,731) (5,110) 4,722

18,478

933 1,949 (3,185)

(986) 3,572

(1,871) 15,920

(18,003) (15,334) (3,959)

20,215

(2,633)

(Continued)

(1,202)

(3,661) 1,172

1,556

2,917

(21,313) (20,355) (4,098)

28,160

2,536 (4,014)

6,116

17,207

15,157

(4,147)

(780)

(3,506) 920

3,326

5,118

(26,894) (23,109) (5,185)

24,175

2,522 (3,894)

5,458

21,416

(19592)

(5,002)

6,799

4,650

14112

18065

1,886 (4,567)

5,251

13,100

Childless Moms Childless

Non-family rel. transfer income in 2000 US$

(14,839) (11,902) (4,389)

19,878

Moms

Other HH labor income in 2000 US$


Austria

Employment rates

Table 2. Individual-Level Variables: Means and Standard Errors (in parentheses)


US

UK

Sweden

Spain

Russia

Netherl.

Luxemb.

Italy

Partnered relationship status Age

High educ. attainment

Medium educ. attainment

Low educ. attainment

.82

(.39)

(.46)

(.35)

(.48)

.69

.86

(.36)

(.34)

.64

.85

(.45)

(.50)

.87

.72

(.42)

(.39)

.45

.77

(.27)

(.45)

.81

.92

(.27)

(.50)

.72

.92

.57

.72

(.45)

.48

(.50)

.77

(.42)

.59

(.49)

35.9

40.9

36.6

(10.6)

37.0

(13.3)

28.6

.95

43.2

(9.3)

34.3

(9.7)

38.0

(9.6)

41.3

(9.8)

40.5

(.39)

.81

(.43)

.76

(.32)

.88

(.21)

.95

(.37)

.83

(.28)

.92

(.27)

34.7

(7.4)

(.23)

.92

(12.7) (12.9)

31.8

(10.7)

34.3

.89

(.31)

40.4

(13.0) (12.6)

41.6

(10.5)

21.3

(12.3)

30.4

(10.7) (10.3)

33.8

(11.3) (13.1)

.58

(.49)

.60

(.46)

.70

(.50)

.55

(.39)

.81

(.50)

.54

(.45)

.72

(.47)

.66

(.48)

.65

(.50)

(5.7)

36.1

(5.4)

35.9

(5.4)

36.5

(5.4)

36.6

(5.8)

36.6

(5.2)

36.8

(5.5)

35.0

(5.0)

37.6

(5.8)

35.6

(6.4)

34.7

(6.2)

33.8

(6.4)

32.5

(5.0)

31.4

(7.0)

36.3

(6.0)

32.7

(5.5)

31.4

(5.8)

34.8

(6.0)

31.5

.32

(.47)

.34

(.33)

.13

(.44)

.27

(.37)

.16

(.50)

.54

(.42)

.23

(.39)

.18

(.27)

.08

(.47)

.61

(.50)

.50

(.46)

.32

(.48)

.37

(.47)

.33

(.49)

.61

(.50)

.43

(.50)

.42

(.41)

.22

(.49)

.46

(.50)

.52

(.49)

.59

(.49)

.60

(.45)

.28

(.48)

.36

(.50)

.52

(.50)

.45

(.49)

.41

(.50)

.30

(.49)

.42

(.50)

.52

(.50)

.54

(.49)

.38

(.45)

.28

(.50)

.46

(.50)

.45

(.50)

.50

(.46)

.21

(.35)

.14

(.45)

.29

(.33)

.12

(.50)

.56

(.31)

.10

(.44)

.25

(.48)

.37

(.50)

.51

(.41)

.09

(.26)

.07

(.37)

.16

(.28)

.09

(.45)

.29

(.31)

.11

(.31)

.11

(.33)

.13

(.45)

.29

(.29)

834 (3,570) 2,011 (4,262)

7,693

(18,620) (19,148) (6,615) 4,048

20,222 20,172

(17,737) (16,585) (4,704)

1,791

3,700 (6,055)

1,186 (2,599) 9,308

8,837 (8,355) 14,169

(30,607) (17,686) (8,034) 23,948

2,025 (5,910)

2,924

27,903

(49,625) (41,004) (5,828)

38,963

1,761 (4,448)

(38,037) (28,582) (7,483)

(2,773)

6,790

26,236

26,086

(10,252)

1,081

(422)

(1,045)

(1,200)

(1,738) 12,366

304

322

505

964

29,372

28,012

(4,622)

932 (2,985)

7,390 (8,431)

(9,263)

10,678

2,418 (5,367)

15,312

5,170

20,174

Childless Moms Childless

Non-family rel. transfer income in 2000 US$

(26,352) (21,978) (7,836)

Moms

Other HH labor income in 2000 US$


Israel

Weekly working hours

Moms Childl. Moms Childl. Moms Childl. Moms Childl. Moms Childl. Moms Childl. Moms Childl.

Employment rates

Table 2. continued

12 Social Forces

52 85 28 25 162 159 161 161 159 14 64 48 42 16 165 161 64 18 12

52 69 13 10 134 143 147 147 135 0 52 26 26 0 145 139 77 0 12

Parental leave in weeks 0 35 30 55 24 37 14 14 54 0 0 30 89 0 25 0 65 0 0

Parental leave benefits 0 1 1 0 1 1 1 1 0 0 0 0 1 1 1 1 1 0 0

Part-time work allowed?c 13 8 20 5 1 22 34 5 10 4 19 6 4 6 21 5 41 1 6

0–2year-olds 41 77 99 53 76 99 87 75 88 56 79 85 68 68 64 77 86 71 53

3–5year-olds

School aged 18.1% 10.4% 31.1% 50.8% 22.6% 29.6% 28.7% 5.3% 20.7% 28.8% 40.1% 16.9% NA 31.0% 9.9% 35.8% 27.9% 20.4% 41.4%

Preschool aged 4.6% 3.1% 16.6% 20.4% 6.4% 10.7% 15.9% 1.4% 4.7% 12.9% 19.2% 4.9% NA 16.9% 4.2% 15.2% 9.4% 6.6% 12.7%

Attitudesb Pref. for maternal full-time empl. when children

bInternational

Work-Family Policy Indicators Social Survey Program, 1994 Family and Changing Gender Roles Module (2002 data for France and Belgium due to lack of 1994 data) c1 = yes, 0 = no or no benefits available

0 40 15 14 52 66 27 27 97 10 10 25 39 16 56 16 57 8 0

Max. length job prot. Lv.

Enrollment in publ. supp. childcarea


aUMass

Australia Austria Belgium Canada Czech Rep. France Germany E. Germany W. Hungary Ireland Israel Italy Luxembourg Netherlands Russia Spain Sweden UK US

Weeks of fully paid leave

Leave policiesa

Table 3. Country-Level Policy and Attitude Measures


14 Social Forces

Table 4. Country-Level Control Variables Total GDP Male Taxation part-time Normal Public sector per unemployment of 2nd employment weekly rateb employmenta capitab rateb earnerc hoursd 16.4

19,053

7.0

32.0

18.1

38

Austria

27.4

24,194

3.4

29.4

13.6

39

Belgium

31.2

22,623

5.8

52.6

20.0

39

Canada

19.0

23,559

6.9

35.9

11.4

40

Czech Rep.

22.2

6,011

3.4

29.9

1.9

43

France

29.5

22,547

7.1

25.9

13.8

39

Germany E.

23.2

23,114

16.3

52.9

10.0

39

Germany W.

22.0

23,114

7.1

52.9

17.9

37

Hungary

36.7

4,692

7.7

30.1

2.7

40

Ireland

18.0

25,313

4.6

30.5

18.1

39

Israel

17.0

18,423

8.9

NA

12.9

45

Italy

15.5

19,269

8.3

38.8

13.7

38

Luxembourg

11.1

46,277

1.5

28.2

13.2

40

Netherlands

25.3

26,033

2.7

40.2

25.9

37

Russia

37.9

1,775

10.6

NA

4.0

40

Spain

25.7

14,421

9.6

23.3

7.1

40

Sweden

33.7

27,286

6.3

34.1

10.9

40

UK

19.2

24,993

6.7

24.4

19.6

38

US

15.8

34,600

3.9

29.7

7.1

40

aLABORSTA;

International Labour Organization (2012) cJaumotte (2003) dUMass Work-Family Policy Indicators bOECD

Following current practice (Gornick and Meyers 2003; Korpi, Ferrarini, and Englund 2013; Mandel and Semyonov 2006; Pettit and Hook 2009), our childcare measures include the percentage of children aged 0–2 and 3–5 enrolled in publicly supported care, which taps the availability of childcare slots, even as we recognize that these slots may be a response to maternal employment, and not a trigger. The country-level measures of attitudes regarding maternal employment come from the 1994 International Social Survey Program.9 We calculated measures based on two questions: the percentage of respondents aged 18 to 65 who prefer that a “woman should work full-time when the youngest child is preschool aged,” and “when the youngest child is school aged.”10 Country-level control variables are presented in table 4 (ILO 2012; OECD 2012).11 For taxation, we use the proportion of the second earners’ income that


Australia


(

)

log(pemp _ ij / 1 − pempij = γ 00 + γ 10 × MOM + γ 11 Z j × MOM

+ γ 01 Z j + γ 20 X ij + u0 j + rij

(1)

Hoursij = γ 00 + γ 10 × MOM + γ 11 Z j × MOM + γ 01 Z j + γ 20 X ij + u0 j + rij .

(2)

The dependent variable in model 1 is the log-odds of women’s employment for individual i in country j, γ00 is the average log-odds of employment across countries, and the coefficient γ10 associated with the motherhood dummy variable estimates the gap in employment (in log-odds) between mothers and childless women. In model 2, the dependent variable is the usual weekly working hours among employed women, with γ00 estimating average women’s weekly hours across countries, and γ 10 × MOM the motherhood gap in weekly employment hours. In both models, Xij and the associated coefficients are the vector of individual-level variables. Zj and its coefficient are the main effect of the country-level policy or cultural indicator. u0j and rij represent error terms at the country and individual level, respectively.


pays for the increased income taxes in a dual-earner household where the first earner’s wages equal 100 percent of APW, and the second earner would go from earning no wages to earning 100 percent of APW as well (Jaumotte 2003). We also control for countries’ total part-time employment rates (OECD 2012) and normal weekly working hours.12 To examine the associations between leave policies, childcare provision, cultural factors, and the motherhood gap in employment participation and weekly hours, we use multilevel models that allow us to model individual- and countrylevel characteristics simultaneously, and account for the nested nature of our data (Diprete and Forristal 1994; Raudenbusch and Bryk 2002). Our outcome variables are a dichotomous variable indicating employment status (1 = employed; 0 = not employed), and the number of usual weekly working hours among employed women. The independent variable of interest is a dichotomous variable indicating whether the respondent has children living at home (mother = 1, childless = 0). We tested other specifications of motherhood, including number of children, and dummies for mothers of preschoolers and older children. All of these specifications led to similar results.13 For ease of interpretation, we present the motherhood dummy. To examine differences in employment participation between mothers and childless women, we estimate random-intercept logistic models. The limited country-level sample size precludes the estimation of random-slopes models. To check whether significant motherhood gaps in employment and work hours remain after controlling for individual-level covariates, we estimate separate logistic regressions, and OLS models for each country. Subsequently, we estimate multilevel models based on the pooled sample of all countries. These models can be written as follows:

15

16 Social Forces

Findings Figure 2 shows the motherhood gap in predicted employment probabilities and average weekly hours net of individual-level and household-level controls. The shaded bars in the graph on the left-hand side show that in the majority of countries, significant differences in employment probabilities between mothers and childless women remain net of individual- and household-level controls. Similarly, net motherhood working hours gaps remain in all but two countries, as indicated by the solid bars in the right-hand-side graph.


To estimate how country-level factors mediate differences in employment between mothers and childless women, we include an interaction between the motherhood dummy variable and the country-level measure Zj. Since the interpretation of interactions in logistic models is problematic (Allison 1999; Mood 2010),14 we estimate average marginal effects. For ease of interpretation, we create a series of plots showing the marginal effects and the confidence intervals around them. We use a two-step Heckman selection modeling strategy (Heckman 1979) for the models estimating motherhood gaps in weekly hours among employed women to account for differential selection of mothers into employment across countries. We run a series of Probit models predicting the likelihood of employment among 25–45-year-old women within each country, using presence of a preschooler in the household, high educational attainment (i.e., postsecondary education or occupational training leading to certification), age, non-family transfer income, the presence of adults aged 18 to 65 other than the mother or a spouse/cohabiting partner, and total household earnings minus the respondent’s earnings. Based on these models, we calculate a selection term (inverse Mills ratio). We include this selection term in the main models for work hours to adjust our estimates for differential selection into employment. We first estimate models with cross-level interactions between the motherhood dummy variable (estimating the motherhood gap in employment and weekly hours) and the country-level policy and attitude measures one at a time. In subsequent robustness analyses, we test (a) whether findings are sensitive to the inclusion of measures for other work-family policies or other policy characteristics; and (b) whether our findings hold if we account for cross-country differences in economic performance using gross domestic product per capita, men’s unemployment rates, the size of public-sector employment, taxation of second earner’s income, normal weekly work hours, and total part-time employment rates. In models estimating the relationship between motherhood employment/ hours gaps and leave policies, we control for childcare enrollment (0–2-year-olds), an indicator of whether leave-takers may work part-time while on leave, and the attitudinal indicator regarding the employment of preschool mothers. Conversely, we control for the length of parental leave, leave benefits, and the attitude measure in the models estimating the impact of childcare. In models examining the impact of attitudes towards maternal employment, we include control variables for childcare enrollment (0–2-year-olds), parental leave length and benefit.


17

Figure 2. Difference in predicted employment probabilities between mothers and childless women, controlling for individual and household characteristics

–.40 –.30 –.20 –.10 .00 .10 Difference in Predicted Employment Probabilities

0

–4 –8 –12 –16 Difference in Predicted Weekly Hours

Note: Significant differences (p