How to Divide Paid Work and Unpaid Care between Parents? Comparison of Attitudes in 22 Western Countries

Sharing responsibilities for paid work and unpaid care between men and women is recognised as one of the challenges that Western countries face in the 21st century. This article examines attitudes towards sharing paid work and unpaid care responsibilities in 22 Western countries by addressing the following questions. (1) How do attitudes towards different earner-carer models vary across countries? (2) Which socio-demographic and country-level factors explain differences in attitudes to an equal division of paid work and unpaid care responsibilities? International Social Survey Programme (ISSP) data 2012 is used as the data source and research methods include logistic multi-level regression analysis. Results reveal that cross-national variations in attitudes are significant: Most traditional attitudes are found in many Eastern European countries, whereas Nordic countries are the least traditional. At the individual level, those who are highly educated, in paid work, single, childless, and religiously non-active support the equal division of paid work and unpaid care responsibilities more often than other respondents. At the country level, longer father-specific parental leave, a stronger tradition of women’s paid work, and less traditional gender roles are related to stronger support for an equal division of paid work and unpaid care.


Introduction
During the last few decades, many Western countries have witnessed a rise in women's, especially mothers', labour market participation (Eurostat 2017).This has led to changes in women's and men's roles regarding paid work and unpaid care responsibilities: The traditional 'male breadwinner and female homemaker' model, where fathers are responsible for earning the income and mothers for taking care of children and home, has been losing its importance.In general unpaid care work refers to all unpaid care work that is provided within a household for its members, including care of persons, housework and voluntary community work (England 2005).In this study, unpaid care refers mainly to the care of own children.Nevertheless, the change in roles has been uneven in a sense that mothers' paid work has increased more than fathers' unpaid care work.(England 2010;Esping-Andersen 2009;Hochschild and Machung 2003;Lewis 2009) Still, sharing the responsibility for paid work and unpaid care between men and women has become one of the challenges that Western countries are facing in the 21st century.
The question of how paid work and unpaid care responsibilities are shared in families is crucial, because raising the employment rate is one of the main political objectives in many Western countries, and raising the employment rate of women and mothers is especially emphasized (European Union 2017).Increasing the female employment rate is also linked to several other policy objectives, such as accelerating economic growth, tackling poverty, combating worsening dependency ratios and enhancing gender equality (Bettio and Plantenga 2004;Lewis 2009).
The question of how paid work and unpaid care should be divided in families is important also when planning and implementing policies targeting the objectives mentioned above.For example, the availability of affordable childcare is highly relevant when considering mothers' participation in paid work (Ellingsaeter 2006;Lewis 2009).Moreover, non-transferable rights to parental leave and wage-level replacement benefits increase the parental leave take-up rates of fathers (Duvander and Lammi-Taskula 2011;Leira 2006).The importance of developing family policies that ensure childcare services are more available and fathers have (and exercise) their right to non-transferable parental leave has been emphasized, for example, by the EU (European Commission 2015) and the OECD (2017).While many countries have developed their family policies according to these principles, wide cross-national variations both in terms of childcare services and parental leave systems remain (Blum et al. 2017;Ciccia and Verloo 2012;Moss 2013).
Previous research (Chou et al. 2016;Chung 2011;Kangas and Rostgaard 2007;Misra et al. 2010;Tammelin 2009;Ylikännö et al. 2015) has focused primarily on the reality of gender differences in paid work and unpaid care responsibilities and demonstrated that despite increasing convergence in time use patterns for hours of paid work and unpaid caring duties, practices in families are still highly gendered.Research into attitudes has focused more on mothers' paid work in general than on the preferred division of unpaid and paid work in families.These studies suggest that attitudes towards mothers' paid work have become more accepting than in the past, but cross-national differences in this respect are still significant (Sundström 1999;Yu and Lee 2013).A number of researchers also argue that attitudes towards gender roles shape labour market practices, such as mothers' patterns of hours of paid work (Pfau-Effinger 2004;Salin 2014).
Regardless of earlier research tradition, there are some aspects that have not been studied profoundly before.Firstly, the vast majority of attitudinal studies on gender roles is primarily interested in attitudes concerning mothers and paid work.Therefore the perspective of fathers and unpaid work has more often been absent.Even less common are studies that combine the questions of paid work and unpaid care as well as perspectives of mothers and fathers in the same study.Secondly, in many cases, earlier studies have been interested only in one country and the cross-national comparison in this respect has been rarer (Knight and Brinton 2017;Yu and Lee 2013).The advantage of a cross-national perspective is, that it enables the analysis on how a single country appears in relation to other countries (Hantrais 2009).Thirdly, there is a shortage of studies that examine attitudes towards gender roles at the concrete level, i.e. how the division of paid work and unpaid care responsibilities should be divided between parents.To our best knowledge, the study of Edlund and Öun (2016) is the only one that examines at the concrete level attitudes towards sharing paid work and unpaid care from a cross-national perspective by comparing five European countries.
The specific contribution of this study is to examine the concrete level attitudes towards sharing paid work and unpaid care responsibilities between parents from a cross-national perspective by comparing a large amount of countries.Compared to a comparison of few countries, the advantage with a large number of countries is the possibility to put the results of single countries on a broader perspective and to identify possible country clusters, i.e. to see which countries are similar and which ones are different to each other (Ragin 1987).Hence, the aim of this article is to examine attitudes toward sharing paid work and unpaid care responsibilities for children under school age in 22 Western countries by investigating the following questions. 1) How do attitudes towards different earner-carer models vary across countries?2) Which individual and country-level factors explain the differences in attitudes to an equal division of paid work and unpaid care responsibilities?Countries included are Austria, Belgium, Bulgaria, Canada Czech Republic, Denmark, Finland, France, Germany, Iceland, Ireland, Latvia, Lithuania, Norway, Poland, Slovakia, Slovenia, Spain, Sweden, Switzerland,

Previous Research
Culturally determined ideas of how people should behave in various life situations differ for women and men as well as for mothers and fathers (Fagan 2001;Lewis et al. 2008).At the societal level, these ideas are concretized in expected and observed gender roles 1 .The notion of public and private spheres and the roles women play in each has a long history.Hence, there are two kinds of gender roles: Those related to the private sphere, such as unpaid care of children, and those related to the public sphere, such as paid work.It has been suggested that women's roles in the two spheres can be inconsistent: Attitudes towards gender roles in the public sphere are typically less traditional than attitudes towards the division of childcare and housework.Thus, people may have more positive attitudes to mothers' paid work than to reducing mothers' role in childcare.(Hochschild and Machung 2003;Knight and Brinton 2017;Yu and Lee 2013).
Esping-Andersen ( 2009) employs the terms 'masculinization of women's life cycles' and 'feminization of men's life cycles' to describe changes in gender roles.Masculinization of women's life cycles refers to women's and mothers' increased involvement in paid work and feminization of men's life cycles to men's and fathers' more active role in childcare and housework.This has led to a situation where the gender roles of women and men have become increasingly similar.However, the change has been asymmetrical; until now, changes in gender roles has been more profound for women and mothers than for men and fathers.Men's and fathers' increased time spent in childcare and housework is less than women's and mothers' increased time spent in paid work.Women and mothers still bear the primary responsibility for housework and childcare.Moreover, the similarity of women's and men's time use is also due to the fact that women have reduced the time they devote to housework.Hence, the experiences of women and men have become more similar because of the reduction of women's commitment to housework, not because of a rising involvement of men in care and housework.This has led researchers to talk about incomplete or even stalled gender role revolution (England 2010;Esping-Andersen 2009;Hochschild and Machung 2003; see also Golinowska 2009).
Moreover, asymmetry in gender role change has led to the phenomena of double burden which refers to a situation where women/mothers doing paid work face the second (unpaid) work shift at home in terms of childcare and housework (England 2010;Hochschild and Machung 2003).The double burden can further have various negative effects for the wellbeing of women/mothers (Hagqvist et al. 2017;Hagqvist et al. 2018), for example in terms of increased sickness absence from paid work (Bratberg and Risa 2002;Palmer 2018).
In earlier studies, different kinds of earner-carer models were employed to illustrate cross-country variations in the roles of men and of women in paid work and unpaid care responsibilities.The models are based on the ways that countries organize their family policies, how gender is perceived and treated by welfare state institutions, and more generally, how countries address women's and mothers roles in society.It could be argued from these studies that not many different models exist, as they usually vary from three to five.Moreover, most studies include the following three models: The so-called modified male breadwinner model, where the father works full-time and mother part-time; an equal dual-earning model, with public childcare; and an intermediate model between these two.Some studies include also the traditional male breadwinner model where the father works full-time and the mother stays at home (Korpi 2000;Lewis 1992;Misra et al. 2010;Pfau-Effinger 2006;Sainsbury 1994;Thévenon 2011).
Previous research has found that attitudes towards mothers' paid work and gender roles, in general, are affected by various individual-level factors (Crompton and Harris 1999;Davis and Greenstein 2009;Knudsen and Waerness 2001;Sjöberg 2004).First, attitudinal differences are gendered.

1
The term gender roles has been criticized quite widely (see for example Connell 1979;Edwards 1983) for that they are socially constructed and principally there would not be differences between men and women.Women's attitudes are less traditional than men's (Alwin et al. 1992;Loslocco and Spitze 2007), which has been explained by larger changes in gender roles for women than for men.
Age is also related to gender role attitudes: Younger people are less traditional than older people.Older cohorts have lived their childhood and/or early adulthood in societies with much more traditional gender roles and a stricter division of women's and men's work.Childcare, especially, was seen as the mother's responsibility (Scott et al. 1996;Yu and Lee 2013).
Education is an important factor in determining attitudes to gender roles: Those with more years of education are less traditional in their attitudes than those with less education.Higher education typically means having a more meaningful and motivating job, which in turn leads to a more positive attitude towards paid work in general and towards mothers' labour market participation (Edlund and Öun 2016;Wall 2007).A similar relationship is found between labour market status and attitudes: Those who are in paid work are more equal in their attitudes toward gender roles than those not in paid work (Alwin et al. 1992;Yu and Lee 2013).Earlier studies (Knudsen and Waerness 2001;Sjöberg 2004) have shown that religiousness is related to more traditional family values in general and to more traditional views on gender roles.
Finally, some researchers argue that family-related factors shape people's attitudes toward gender roles.Marital status plays a role: Single and divorced people are less traditional, whereas married and cohabiting people are more traditional.Further, some researchers have hypothesised that whether a person has children, and especially children under school age, explains more traditional attitudes toward gender roles.The importance of family-related factors is due to the change in practice between paid work and care in families after the birth of a child, especially the first child.In most families, the mother stays at home taking care of the child and the father becomes (at least temporarily) the main earner in a family (Bühlmann et al. 2010;Davis and Greenstein 2009).
It is likely that, in addition to individual-level factors, country-level factors might play a role in shaping attitudes toward sharing paid work and unpaid care responsibilities.Earlier studies (Böckmann et al. 2013;Uunk 2015) have shown that country-level factors shape the actual division of paid work and unpaid care responsibilities.Moreover, different country-level factors have proven to be important in determining attitudinal differences in other respects, for example, concerning lone parent families (Mia and Milla).Previous studies (Edlund and Öun 2016;Uunk 2015) have also revealed that there are significant cross-national differences in attitudes toward gender roles at a general level.Therefore, factors related to family policies, labour market practices, and societal gender role ideology at the country-level are possible determinants of attitudes.
The role of family policy in determining parents', especially mothers', employment and care behaviour has been emphasised in many earlier studies.The organisation of family policies in a country can enhance or hinder a more equal division of paid work and unpaid care responsibilities between mothers and fathers (Hegewisch and Gornick 2011;Stejber and Haas 2009).Therefore, family policies might also shape people's attitudes concerning the division of paid work and care responsibilities between parents (Edlund and Öun 2016).In the case of children under school age (as in this study), two dimensions of family policies are considered particularly important: The parental leave system and the (public) childcare system 2 (Crompton and Lyonette 2006;Leira 2002).
From the perspective of sharing the responsibilities of paid work and unpaid care, the most important feature of the parental leave system is whether it encourages fathers to take leave.Earlier studies (Ellingsaeter 2006;Lewis 2009) have shown that in all countries mothers are using the vast majority of leave that either parent could use.It has been suggested that in order to enhance fathers' take-up of parental leave, it is of vital importance to earmark a share of paid/income related leave exclusively for them.Instead, when it is left up to the family to decide who uses the leave, it is usually 2 In this study the focus is solely on the parental leave system.Liable and comparable information on countries' (public) childcare systems was not available and therefore it was not possible to test its role empirically.
Soc. Sci.2018, 7, 188 5 of 20 the mother who uses the leave (Arnalds et al. 2013;Brandth and Kvande 2009;Rantalaiho 2010).Of the countries included in this study, the Nordic countries have the longest tradition of offering fathers paid non-transferable paternal leave.Other countries have not had this kind of leave in the past, but several have introduced it in recent years.Nevertheless, there are countries that do not offer fathers any paid non-transferable leave (Blum et al. 2017;Ciccia and Verloo 2012;Moss 2013).
In the case of labour market practices, the focus is on women's and mothers' historical roles as earners.Differences in labour market practices might be relevant determinants of attitudes about the division of paid work and care responsibilities.The last few decades have witnessed a significant increase in women's, especially mothers', paid work.Nevertheless, there are still relatively substantial differences across countries in women's and mothers' employment rates.The tradition of paid work by women and mothers is longest in the Nordic countries and their rates are among the highest in Western countries.Eastern European countries share a long tradition of paid work by women and mothers, but rates in some Eastern European countries have declined slightly since the collapse of the communist system.In English-speaking, continental, and southern European countries, the tradition of paid work by women and mothers is weaker, but rates have increased significantly in a majority of these countries over the last few decades (Eurostat 2017;Lewis 2009;Pfau-Effinger 2004).
Another vital dimension of women's and mothers' roles as earners is their hours of work.Women work part-time more often than men (Eurostat 2017; Warren 2010), an indication of their different roles as earners.There are, nonetheless, cross-national differences in women's and mothers' rates of part-time work.Especially in continental European and English-speaking countries, part-time work has been, and still is, more prevalent among women and mothers, while in Eastern European countries, and to a lesser extent in southern European countries, women and mothers tend to work full-time.Further, in Nordic countries except Finland, a large share of women, especially mothers, work part-time.However, in the Nordic countries 'part-time' work often means somewhat longer hours of work than it does in continental European and English-speaking countries (Eurostat 2017;Nergaard 2010;Salin 2014).
In addition to the actual practice of the division of unpaid care and paid work in families, gendered norms about the roles of mothers and fathers as earners and carers are important.The way gender role ideology is addressed at the country level is assumed to shape attitudes to sharing paid work and unpaid care at the individual level.Even though gender roles have been changing and becoming less rigid, it does not necessarily mean that the cultural ambience at the country-level has become more accepting of mothers' equal roles as earners or fathers' equal roles as carers (Knight and Brinton 2017;Yu and Lee 2013).Moreover, changes in actual practices and societal gender role ideology do not always go hand in hand: Regardless of the increase in the share of working mothers, the idea that mothers should take care of children under school age at home can continue to be dominant (Pfau-Effinger 2006).Earlier studies (Edlund and Öun 2016;Knight and Brinton 2017) have revealed that societal gender role ideology is least traditional in Nordic countries, and attitudes are most traditional in Eastern European countries.Continental European, English-speaking, and southern European countries sit between these two groups.

Research Design
This study examines attitudes to sharing paid work and unpaid care responsibilities in 22 Western countries by addressing the following research questions.

1.
How do attitudes to different earner-carer models vary across countries?2.
Which socio-demographic and country-level determinants explain the differences in attitudes to an equal division of paid work and unpaid care responsibilities?
The first research question focuses on differences across Western countries in attitudes to six earner-carer models.The focus in the second research question turns to individual and country-level characteristics that are related to cross-national variances in the division of paid work and unpaid care responsibilities.
We drew on earlier literature and studies to construct two hypotheses for the study.Firstly, we expected to find cross-national differences in the attitudes towards different earner-carer models.Support for equal earner-carer models is assumed strongest in the Nordic countries and weakest in Eastern European countries, with Continental European, English-speaking, and southern European countries in the middle position (Esping-Andersen 2009; Lewis 2009).We assume that at the country level, stronger support for an equal division of paid work and unpaid care responsibilities is expected in countries where the parental leave system supports fathers' roles as carers, the tradition of mothers as earners is stronger, and societal gender role ideology supports equal gender roles (see e.g., Duvander and Lammi-Taskula 2011;Nergaard 2010;Sainsbury 1994;Uunk 2015;Yu and Lee 2013).
Secondly, we hypothesized that at the individual level, women, younger, more highly educated, those in paid work, single or divorced, without children, and not religious are expected to support an equal division of paid work and unpaid care responsibilities more often than persons without these characteristics (Alwin et al. 1992;Davis and Greenstein 2009;Edlund and Öun 2016;Wall 2007).
The data used in the study derives from the International Social Survey Programme (ISSP), which is a cross-sectional collaborative programme that conducts surveys in a wide range of countries around the world.We used data for the year 2012, including the module family and changing gender roles.(International Social Survey Programme, ISSP.2012 -Family and changing gender roles IV).Altogether 22 Western countries were analysed.Countries included are Austria, Belgium, Bulgaria, Canada Czech Republic, Denmark, Finland, France, Germany, Iceland, Ireland, Latvia, Lithuania, Norway, Poland, Slovakia, Slovenia, Spain, Sweden, Switzerland, the United Kingdom, and the United States.As many countries as possible were included in the analyses.However, countries with too many missing cases in some variables used and/or with no available comparable country-level data were excluded.Large enough number of cases per country and the availability of comparable country-level data were the crucial preconditions for the use of multi-level analysis.The final data includes 14,436 persons between 18 and 55 years old.The number of cases and response rates per country are presented in Appendix A. The very low response rates in some countries means the results for these countries should be considered with caution (Gendall et al. 2016).
As the number of countries included in analyses is more than 20 it is not possible to focus each country separately.Therefore, when appropriate, results are interpreted by employing country clusters.However, the purpose is not to produce any regimes or typologies as such, but country clusters are used to the extent they might help the understanding of differences and similarities across countries (see also Daly 2000).Therefore no specific country clusters based on earlier classifications (Korpi 2000;Pfau-Effinger 2006;Sainsbury 1994) are predetermined.However, our terminology of Continental European, English-speaking, Eastern European, Nordic and southern European countries have to some extent been used for example by Esping-Andersen (2009), Lewis (2009) and Thévenon (2011).

Dependent Variable
Respondents' attitudes towards different earner-carer models were measured with the question, 'Consider a family with a child under school age.What, in your opinion, is the best way for them to organize their family and work life?' Respondents were asked to select one of the following options: The remaining six options were presumed to measure not only the respondent's opinion of how paid work and unpaid care responsibilities should be divided between the parents but to measure more generally the respondent's attitude towards different earner-carer models.The models are: Male breadwinner model (option 1); modified male breadwinner model (option 2); two full-time earners model (option 3); two part-time earners model (option 4); modified female breadwinner model (option 5); and female breadwinner model (option 6) (see, for example, Edlund and Öun 2016;Lewis 1992;Pfau-Effinger 2004).
In addition to six different earner-carer models, the options separate mothers' and fathers' relative roles as earners and carers into three broad categories.Options one and two represent a model where the father's role as earner is more important than the mother's and the mother's role as a carer is more important than the father's.This is the 'male-dominated earner model'.Options three and four represent an equal earner-carer model where the mother's and father's earner and carer roles are equal to each other.Lastly, options five and six represent a model where the mother's role as earner is more important than the father's and the father's role as a carer is more important than the mother's.This is the 'female-dominated earner-model'. 3 Descriptive statistics for the dependent variable along with the independent variables are shown in Table 1.
Firstly, we analysed possible cross-national variations in attitudes to the six different earner-carer models.Secondly, we analysed the differences across countries and the factors that were associated with these differences by employing two combined earner-carer models, the male-dominated earner model, and the equal earner-carer model.Unfortunately, we could not include the female-dominated earner model in our analyses, because there were too few responses that identified this model (Table 1).The primary reason for employing the two-category dummy variable instead of the previously presented six-category variable is the problem of defining part-time work: The terms 'full-time' and 'part-time' refer to different numbers of hours in different countries.Further, respondents in different countries might have different perceptions about how many hours of work in a week constitutes 'part-time' how many hours constitutes 'full-time' (Haataja et al. 2010;Nergaard 2010).

Individual-Level Independent Variables
Based on earlier studies (Alwin et al. 1992;Davis and Greenstein 2009;Sjöberg 2004;Wall 2007), gender, age, education, labour market status, marital status, children and their age, and religiousness were used as individual-level independent variables.Gender was included as a dummy variable.Age was measured by a three-category variable differentiating 18-30, 31-45, and 46-55 year olds from each other.The variable for education recorded whether the respondent's education level was low, intermediate, or high.Labour market status recorded whether the respondent was in paid work, unemployed, in education, or otherwise outside the labour market. 4Marital status recorded whether the respondent was single, in a relationship, or divorced.Children and their age were measured by a three-category variable that recorded whether the respondent had no children, his/her youngest child was under school age, or his/her youngest child was school age.Religiousness recorded whether the respondent attended religious services or not: A dummy variable separates religiously non-active from religiously active respondents.

Country-Level Independent Variables
Based on earlier literature (Böckmann et al. 2013;Kangas and Rostgaard 2007;Stejber and Haas 2009), we used the parental leave system's support for fathers' roles as carers, the tradition of women's roles as earners, and the societal gender role ideology as country-level indicators of attitudes to shared paid work and caring responsibilities.
The parental leave system's support for fathers as carers was measured with a three-category variable indicating the length of paid father-specific leave in the country's parental leave system 5 .It identified countries with no paid father-specific leave, those with paid father-specific leave up to three weeks, and countries with paid father-specific leave longer than three weeks6 .The information is for the year 2012 and is derived from the OECD Social and Welfare Statistics.Paid father-specific leave refers to the number of paid weeks reserved for the exclusive use of fathers, including entitlements to paid paternity leave, 'father quotas' or periods of paid parental leave that can be used only by the father and cannot be transferred to the mother, and any weeks of paid sharable leave that must be taken by the father in order for the family to qualify for 'bonus' weeks of parental leave.(OECD 2018b) The term 'paid father-specific leave' (rather than paternity leave for example) is employed in this study because it is term OECD (2018b) itself uses for this measure and because it includes both paternity and parental leaves as well different kinds of 'quotas' and 'bonus' leaves.
Indicators for the tradition of women's role as earners refer to a culture of women's paid work, including part-time work.The mean of women's annual employment rates from 2000 to 2012 was used to measure the tradition of a culture of women's paid work.The data is drawn from the OECD Dataset for Labour Force Statistics. 7The variable for a culture of women's part-time work is the mean of part-time employment as a percentage of women's total annual employment from 2000 to 2012.The information is from the OECD Dataset for Labour Force Statistics (OECD 2018a). 8ocietal gender role ideology measures the acceptability of mothers' paid work in each country and the cultural ambience about appropriate gender roles.The variable acceptance of mothers' paid work records the proportion of people in each country who think a pre-school child is likely to suffer if his or her mother is in paid work.Cultural ambience about appropriate gender roles is measured by the proportion of people in each country who think that a man's job is to earn money and a woman's job is to look after the home and family.Both of these variables were constructed by combining individual-level ISSP data into an aggregate variable.The aggregate variable was obtained by calculating the percentage of respondents in each country who agree that a pre-school child is likely to suffer if his or her mother is in paid work and the percentage who agree that the man's job is to earn money and the woman's job is to look after the home and family.

Research Methods
Both descriptive and explanatory research methods were used.Descriptive statistics were used to analyse cross-national differences in attitudes towards different earner-carer models.Logistic multi-level regression analysis was employed to determine which individual and country-level factors are explaining variations in the attitude to the equal earner-carer model.Logistic analysis was chosen because the dependent variable had two categories.A multi-level analysis was chosen because both individual and country-level information were included in the analyses.The advantage of multi-level analysis compared with standard logistic regression analysis is that a multi-level model allows the intercept to vary randomly across countries, whereas it is presumed to be constant in standard regression.(Hox 2002).We employed a step-by-step procedure in our multi-level analysis, starting with an empty model.First, we included individual-level factors in the model.We then added country-level characteristics to the model.Average marginal effects, statistical significances, standard errors, BIC values, country variance, and VPC are reported below.
The small N problem at the country level is an issue related to multi-level modelling.The number of macro-level units should be sufficiently large to keep the results reliable.(Hox 2002) The reliability of country effects should be treated with caution, as the standard errors for country-level effects and the country variance tend to be under-estimated, and confidence intervals are too narrow with a relatively small number (less than 30) of macro-level units (Bryan and Jenkins 2016).These limitations should be taken into account when interpreting the results.Moreover, the number of macro-level variables in one model is restricted by the number of macro-level units in the data.Therefore, country-level variables were divided into three distinct models.

Results
First, the differences across countries in attitudes towards different earner-carer models were analysed.Figure 1 shows the cross-national variations in attitudes towards the six earner-carer models.It appears that countries do not differ from each other in one respect: Neither the female breadwinner nor the modified female breadwinner model is seen as the best way to share paid work and unpaid care responsibilities in any country.Support for these two models is so marginal that their share is even hard to see on the figure.This demonstrates that, despite the changes in gender roles, the situation where the mother's earning role is primary and the father takes the main responsibility for childcare is not perceived as an appropriate way to share these tasks (Esping-Andersen 2009).Nevertheless, cross-national variations are found in other respects.The most traditional group of countries includes the majority of Eastern European countries (Latvia, the Czech Republic, Lithuania, Slovakia, and Bulgaria): In these countries more than 50 percent of the population think that the male breadwinner model, where the father works full-time and the mother stays home, is the best way to divide paid work and unpaid care between parents.This supports the argument (Hobson et. al 2006)  Nevertheless, cross-national variations are found in other respects.The most traditional group of countries includes the majority of Eastern European countries (Latvia, the Czech Republic, Lithuania, Slovakia, and Bulgaria): In these countries more than 50 percent of the population think that the male breadwinner model, where the father works full-time and the mother stays home, is the best way to divide paid work and unpaid care between parents.This supports the argument (Hobson et al. 2006) that, despite the long tradition of mothers' paid work, attitudes in Eastern European countries have remained traditional.Interestingly, in Poland, Croatia, and especially Slovenia, attitudes are less traditional.
The least traditional attitudes are found in four Nordic countries: Denmark, Iceland, Norway, and Sweden.In these countries, the traditional male breadwinner model gained hardly any support.The modified male breadwinner model and the two full-time earners model (as well as the two part-time earners model in Sweden) were seen as preferable ways to share paid work and unpaid care responsibilities.These countries share a long tradition of mothers' paid work, provide extensive public childcare that enables mothers' paid work, and actively developed parental leave systems to support fathers as carers (Lewis 2009;Moss 2013).Finland differs slightly from the other Nordic countries, with stronger support for the traditional male breadwinner model.Earlier studies (Edlund and Öun 2016;Hiilamo and Kangas 2009) have also shown Finland to differ from its Nordic counterparts, with more traditional attitudes and practices concerning the division of paid work and childcare.
Continental European, English-speaking, and southern European countries lie between these two positions.Support for the traditional male breadwinner model is lower in these countries than in most of the Eastern European countries, but significantly higher than in the four Nordic countries.The modified male breadwinner model, where the father works full-time and the mother part-time is seen as best for sharing unpaid care and paid work responsibilities, except in Canada, Poland and Austria, where the male breadwinner model is seen as best.Excluding southern European countries, part-time work has been the main paid work practice for mothers to work in these countries and the availability of childcare has supported mothers' part-time work (Kanji 2011;Misra et al. 2010).
Figure 2 presents essentially the same information as Figure 1 but divides attitudes into two earner-carer models: The male-dominated earner model and the equal earner-carer model.The four Nordic countries form a distinct group once again.It is only in Sweden, Denmark, Iceland, and Norway that more than half the population prefer the equal earner-carer model-that is, equal division of paid work and unpaid care responsibilities-to the male-dominated earner model.In all other countries, the male-dominated earner model is preferred, indicating that the father's role as earner is more important than the mother's earning role and the mother's role as a carer is more important than the father's.Support for the male-dominated earner model varies from less than 60 percent in Finland to (nearly) 90 percent in five Eastern European countries.The results in Figure 2 reinforce the theory that, despite the changes in gender roles, attitudes are relatively traditional in many countries and significant cross-national variations exist in attitudes to sharing paid work and unpaid care responsibilities (Edlund and Öun 2016;Esping-Andersen 2009).
The final step of the analysis was to examine which individual and country-level determinants can explain the differences in attitudes to an equal division of paid work and unpaid care responsibilities.The results of the logistic multi-level regression analysis are presented in Table 2. Model 1 includes all individual-level independent variables.Most striking is the role of education: Compared with highly educated respondents, those with intermediate education have an 11 percentage points lower probability of thinking that equal sharing of paid work and unpaid care responsibilities between parents is the best way to organize work and care.For those with low education, the probability is as much as 17 percentage points lower for supporting equal sharing of paid work and unpaid caring.This result supports our hypothesis and earlier studies (Davis and Greenstein 2009;Edlund and Öun 2016) that higher education is related to less traditional gender roles and more positive attitudes towards mothers' paid work.Hence, it appears that education is important in relation to attitudes towards both paid work and unpaid care responsibilities.
earner is more important than the mother's earning role and the mother's role as a carer is more important than the father's.Support for the male-dominated earner model varies from less than 60 percent in Finland to (nearly) 90 percent in five Eastern European countries.The results in Figure 2 reinforce the theory that, despite the changes in gender roles, attitudes are relatively traditional in many countries and significant cross-national variations exist in attitudes to sharing paid work and unpaid care responsibilities (Edlund and Ö un 2016;Esping-Andersen 2009).The final step of the analysis was to examine which individual and country-level determinants can explain the differences in attitudes to an equal division of paid work and unpaid care responsibilities.The results of the logistic multi-level regression analysis are presented in Table 2. Model 1 includes all individual-level independent variables.Most striking is the role of education: Compared with highly educated respondents, those with intermediate education have an 11 percentage points lower probability of thinking that equal sharing of paid work and unpaid care responsibilities between parents is the best way to organize work and care.For those with low education, the probability is as much as 17 percentage points lower for supporting equal sharing of paid work and unpaid caring.This result supports our hypothesis and earlier studies (Davis and Greenstein 2009;Edlund and Ö un 2016) that higher education is related to less traditional gender roles and more positive attitudes towards mothers' paid work.Hence, it appears that education is important in relation to attitudes towards both paid work and unpaid care responsibilities.Furthermore, religious activity is critical.In line with our hypothesis, people who are not religiously active are less traditional in their attitudes: They have seven percentage points greater probability of supporting the equal earner-carer model than those who are religiously active.This result reinforces the argument (Sjöberg 2004) that religiousness is related to more traditional views about families in general and about the division of paid work and unpaid care responsibilities in particular.
Moreover, as hypothesized, labour market attachment is related to attitudes to sharing paid work and unpaid care.In earlier studies (Yu and Lee 2013) being in paid work leads to less traditional attitudes.Results here suggest the reason for being outside the labour market is critical: those who are in education do not differ from those in paid work, and those who are otherwise outside the labour market (people looking after family and home, long-term sick and disabled, temporarily sick and disabled, retired people and discouraged workers) as well as unemployed are more traditional in their attitudes than those who are in paid work.
In addition, our hypothesis about the relationship between marital status and attitudes (Wall 2007) is supported, but only partly.As expected, those who are single have less traditional attitudes: They have five percentage points higher probability of thinking that the equal earner-carer model is the better way to organize paid work and unpaid care responsibilities than those in a relationship.Divorced people do not differ in their attitudes from those who are in a relationship.
Quite surprisingly, (Davis and Greenstein 2009), children and their age have only a modest effect on attitudes toward earner-carer models.Parents whose youngest child is of school age have two percentage points lower probability of thinking that the equal earner-carer model is the best way to share paid work and unpaid care between parents than those without children, and having children under school age is not related to attitudes.Earlier studies (Chou et al. 2016;Chung 2011) of the actual division of paid work and unpaid care between parents gave us a reason to expect that having children might have a greater impact on parents' attitudes towards earner and carer roles, especially when children are young.This might imply that even though having (small) children affects actual earner and carer roles in families, children's impact on attitudes is less evident (Bühlmann et al. 2010).Lastly, in contrast to our hypothesis, neither gender nor age is related to attitudes toward earner-carer models.This result was not predicted, as women and younger people have been shown in many earlier studies (Edlund and Öun 2016;Loslocco and Spitze 2007) to be more positive towards mothers' paid work, for example.This raises the question of whether this result bolsters the argument that attitudes towards the public sphere of paid work might be less traditional than attitudes towards the private sphere of unpaid care.It might be the case that women and young people are less traditional in their attitudes to mothers' paid work than to unpaid care responsibilities (Knight and Brinton 2017;Yu and Lee 2013).
Country-level factors are included in the analyses of Models 2, 3, and 4. Model 2 tests the hypothesis (Lewis 2009) about the role of the paid parental leave system, specifically, how it supports the father's role as carer.The result is exceedingly interesting: It is not merely the existence of paid father-specific leave as such, but rather the length of leave, that is related to attitudes to earner-carer models.In countries where leave is longer than three weeks, the probability of support for the equal sharing of paid work and unpaid care responsibilities is as much as 15 percentage points higher than in countries with no father-specific leave available.A short period of paid father-specific leave of up to three weeks is not related to attitudes.Countries with short paid father-specific leave do not differ in their attitudes towards sharing paid work and unpaid care responsibilities from countries that do not have paid father-specific leave at all.Model 3 shows that our hypothesis about the relationship between attitudes and a tradition of mothers as earners is only modestly supported.When a history of mothers' paid work increases by one percentage point, the probability of support for an equal division of paid work and unpaid care responsibilities increases by one percentage point.A history of women's part-time work is not related to attitudes.One explanation for this result might lie in the history of Eastern European countries.Regardless of the long history of women's full-time paid work, earlier studies (Hobson et al. 2006) have indicated that attitudes towards gender roles have remained traditional.This might diminish the impact of a part-time work culture. 9odel 4 shows that our hypothesis about societal gender role ideology is only partly supported.The results show that a one percentage point increase in support for more traditional gender roles lowers the probability of support for the equal earner-carer model by one percentage point.Hence, it appears that the perception of appropriate gender roles at the country level is modestly associated with attitudes towards sharing paid work and unpaid care at the individual level (Pfau-Effinger 2006).Nevertheless, acceptance of mothers' paid work is not related to attitudes toward earner-carer models.This raises the issue again of different attitudes in the public sphere of paid work from those in the private sphere of unpaid care (Knight and Brinton 2017), as the country-level indicator about acceptance of mothers' paid work is not explaining the attitudes to the division of paid work and unpaid care at the individual level.

Discussion and Conclusions
This study has examined attitudes towards sharing paid work and unpaid care responsibilities in 22 Western countries by addressing two research questions: How attitudes towards different earner-carer models vary across countries, and which individual and country-level determinants explain the differences in attitudes towards an equal division of paid work and unpaid care responsibilities?
The cross-national examination of attitudes towards different earner-care models has revealed that-regardless of the changes in gender roles during the last few decades (Esping-Andersen 2009)-attitudes in the 2010s are still relatively traditional when it comes to sharing paid work and unpaid care responsibilities between parents.This is evident in two respects.First, no country that we examined considered that earner-carer models, where the mother's earning role was more important than the father's and the father's caring role was more important than the mother's, was the best option.In fact, support for this kind of arrangement was more or less non-existent.Second, the equal earner-carer model was seen as better than the male-dominated earner model in only four countries out of 22. Hence, in the vast majority of countries, the father's earning role was seen as more important than the mother's and the mother's caring role was more important than the father's.
This attitudinal 'traditionalism' raises further questions about what lies behind it, as many earlier studies (Sundström 1999) have shown that attitudes towards mothers' paid work, for example, have become more supportive over the last few decades.Our results suggest that the arguments put forward by Knight and Brinton (2017) as well as Yu and Lee (2013) about different attitudes in the public and private spheres might play a part here.They have argued that attitudes to paid work and unpaid care can be contradictory: While attitudes towards mothers' paid work have become more accepting, attitudes towards mothers' responsibility for (child) care might remain traditional.Earlier studies (Misra et al. 2010) on the actual division of paid work and unpaid care support this conclusion, as they revealed that, regardless of mothers' increased involvement in paid work, practices of unpaid care are still highly gendered, especially in families with small children.It has also been argued that parental leave systems address equality in sharing paid work and in unpaid care responsibilities differently.In many countries, the male breadwinner ideal continues to provide the normative reference point in parental leave policies; however, in other countries, equality and sharing of paid work are emphasised more than equality and sharing of unpaid care.Hence, equality of unpaid care responsibility is not perceived as a primary objective in many parental leave systems.(Ciccia and Verloo 2012).As Brandth and Kvande (2009) state, gender equality policies have primarily attempted to make women equal to men whereas the attempt to make men equal to women has been less important.
Nevertheless, part of the explanation of what sits behind traditional attitudes might lie in the wording of the dependent variable.In the data, the dependent variable is formulated in a static way.It does not allow a change in earner and carer roles between parents, for example, at first the mother would stay at home and the father would work.After a period, the roles would change and the father would stay at home and mother work.This kind of 'shift-change' has been supported, at least in the Nordic countries, by parental leave systems (Moss 2013).Thus, if the dependent variable had enabled a 'shift-change' in work and care roles between parents, attitudes at least for some countries might have been less traditional.
The results of this study also revealed that attitudes towards sharing paid work and unpaid care are shaped by many factors at the individual and the country level.The study has also revealed avenues for future studies.One avenue concerns the role of paid father-specific parental leave.Our results show that, in countries with paid father-specific leave longer than three weeks, attitudes to an equal sharing of paid work and unpaid care are more common than in countries without any paid father-specific leave.However, attitudes do not differ between countries with shorter paid father-specific leave (up to three weeks) and no paid father-specific leave.It might be that shorter paid father-specific leave is (in most cases) taken at the same time as when the mother is also on leave.Hence, it might not support the father's independent care role as strongly as longer leave (see also Arnalds et al. 2013;Duvander and Lammi-Taskula 2011;Bünning 2015).In any case, the impact that the parental leave system and its support for the father's care role has on attitudes-not only on the sharing of paid work and unpaid care responsibilities but on gender roles more generally-is an interesting question for prospective study.In this study, the role of parental leave system was approached at a general level; therefore, there is a need to examine them more profoundly and in detail to understand the mechanisms that lie behind parental leave systems and attitudes.
Another aspect to consider more deeply in future studies is related to gender and parenting.Earlier studies (Tammelin 2009;Chung 2011) have revealed that both gender and having children (or not) are important determinants of how paid work and unpaid care responsibilities are actually divided.Hence, it was surprising that gender was not at all related (and having children only modestly related) to attitudes to sharing paid work and unpaid care responsibilities.Does this indicate that actual practices and attitudes are inconsistent?Bühlmann et al. (2010) assert that dissonance between attitudes and practices in sharing paid work and unpaid care responsibilities is indeed prevalent, especially in families with small children, but the magnitude of this dissonance varies across countries.This study did not provide answers to the relationship between attitudes and practices, but it is very important that this is investigated in future studies to understand the dynamics and dissonances between attitudes and practices.
(1) The mother stays at home and the father works full-time; (2) The mother works part-time and the father works full-time; (3) Both the mother and the father work full-time; (4) Both the mother and the father work part-time; (5) The father works part-time and the mother works full-time; (6) The father stays at home and the mother works full-time; and (7) I can't choose.Respondents who answered 'I can't choose' were excluded from the analyses.

Figure 2 .
Figure 2. Attitudes to the male-dominated earner model and the equal earner-carer model, %.
CZ SK UK CA AT US PL IE DE HR FR ES CH SI FI NO IS DK SE Male-dominated earner model Equal earner-carer model

Figure 2 .
Figure 2. Attitudes to the male-dominated earner model and the equal earner-carer model, %.
the United Kingdom, and the United States.Data is extracted from the International Social Survey Programme (ISSP 2012), including a module on family and changing gender roles.

Table 1 .
Descriptive statistics for dependent and independent variables.
1The number of cases is greater in this variable as it includes all six categories.

Table 2 .
Logistic multilevel regression analyses of attitudes to the equal earner-carer model, average marginal effects, and statistical significances (standard errors).