The effect of state expenditure on family policy and family policy instrument types on female labor market participation and attitudes: A Repeated-Cross Section study 1988-2012

The effect of state expenditure on family policy

and family policy instrument types on female

labor market participation and attitudes: A

Repeated-Cross Section study 1988-2012

Author: Beata Berke Student ID: S1909770

Email: beata.berke@gmail.com

Supervisor: Dr. Max van Lent Master Thesis Public Administration Economics and Governance

Faculty of Governance and Global Affairs Leiden University

INTRODUCTION

II. THEORETICAL FRAMEWORK ... 7

PT.1:ECONOMIC ANALYSIS AND POLICY ...7

INDIVIDUAL CONTEXT ...8

Labor Force Participation Intensity ... 10

INSTITUTIONAL CONTEXT ... 11

Cash benefit ... 12

Child care for working mothers ... 13

PART II : SUPPORT FOR TRADITIONAL GENDER ROLES ... 14

III. DATA AND METHODS ... 16

PART I:... 18 Dependent variable ... 18 Independent variables ... 18 Controls ... 19 PART II: ... 20 Dependent Variable ... 21 Controls ... 22

IV. RESULTS AND ANALYSIS ... 23

PART I:EFFECT OF BENEFIT PRESENCE ON FEMALE HOURS WORKED PER WEEK... 23

PART II:ATTITUDES TOWARD TRADITIONAL GENDER ROLES... 44

Descriptive statistics and mean differences ... 44

Expenditure and Policy ... 49

Analysis... 57

Abstract

Declining fertility and the increasing weight of population aging is a crucial issue in today's socioeconomic environment. The implications are that there will not be enough active workforce to finance pension, health care and long term care costs. A proposed solution for the existing population aging problem is to increase fertility rates via various state policies that I look at in this thesis. The argument of the thesis is that increasing state expenditure on certain types of policies, namely benefits paid out in cash, will have a negative effect on female hours worked. This thesis will be divided into two parts: in the first part, I begin to investigate the change in hours worked for the whole sample of women, married women and married women with children, based on individual characteristics, aggregate country expenditures and dummies representing family policy presence. Additionally, an interaction term between education groups and family policy dummies are included in the analysis. In the second part of the study, I research the effect of family policy structure on attitudes toward traditionalist gender roles.

In the first part I predict that hours worked will reduce with an increase in expenditure on cash benefits for families, and hours worked will be higher or remain the same with increased spending on benefits in kind. The main basis of my paper is this argument: I expect that women who receive income substitution in the form of compensation in cash will most likely reduce their hours worked. On the other hand, women who receive child care options through higher expenditure on benefits in kind are more likely to utilize these services and have little or no work hour reduction, since the cost of childcare would be higher if they ceased to work. In the second part of the study, I expect that day care, early childhood education care and parental leave, will reduce the support for traditional gender roles. The findings of this thesis finds support that income substitution, birth grant and child allowance have negative effect on female labor force participation, in terms of hours worked, and increases support for traditional gender roles. The thesis did not find support to the statement that the presence of ECEC would lead to an increase in female hours worked per week. The second part of the thesis found support that the presence of both, benefits in cash and in kind, expenditures together with parental leave, day care and child allowance had a negative effect on support for traditional gender roles, while birth grant, ECEC, total public expenditure were a significant predictor of support for traditional gender roles.

Introduction

In the last decades the female labor force participation (FLFP) rose markedly. This increase is usually attributed to female emancipation from the household, and shifts away from traditional gender roles that have allowed increasing female investment in education (Gupta, 2007; Hakim, 1995; Thévenon & Horko, 2009). This increase in investment comes from women having more opportunities to invest in obtaining a degree, which has led to a shift away from the traditional household dynamic. Even though female participation in the workforce brings great socioeconomic benefits, it comes at a cost, too, namely, declining fertility rates. It is noteworthy that FLFP has been usually studied from the perspective of declining fertility rates and aging of the populations. Increasing FLFP and declining fertility lead to an unsustainable level old-age dependency ratio, an issue faced by many OECD countries currently, with the European regions affected the most by 2050 (Bloom, Canning, & Sevilla, 2003; Muenz, 2007). Declining fertility means that there will be less working-age individuals contributing to the state economy, and more retired persons. Making it increasingly difficult to allocate money to social expenditures (Castles, 2003).

Furthermore, in the last 15 years, the EU-28 median age has increased by 4.3 years (Eurostat, 2016), this means that full employment has gained in importance. Therefore, as part of its gender equality goals the EU aims to increase FLFP to reach 75% of the population aged 20-64 by 2020 (European Commission 2016). Alongside the EU gender equality plan, the Organization of Economic Cooperation and Development (OECD) has promoted making parental leave available and increasing the duration of maternal leave, so new parents can take care of the child without exiting the labor force (Adema et al., 2007). It is expected that by making these kinds of work-leaves more commonly available, more women would be willing to have children and participate in the labor force. Parental leave, more specifically maternal leave, might not bring the desired effect of more women remaining in the workforce. The woman is still considered the primary caretaker of the child, and will choose to withdraw from the labor force during the leave, rather than dividing the leave among the parents (Gesthuizen et al., 2011). Additionally, these policies vastly differ between countries. Hence, until there is a parental leave available to fathers to the same extent as to mothers, it is unlikely that it would move us closer to parental equality between genders. A large amount of literature argues that the main reason for increasing female LFP of mothers is the increasing opportunity costs for women of childbearing due to educational investment (Walter, 1985; Bratti, 2001). To combat this, more

governments have started offering benefits that would at least partially compensate the family for the cost of having children (Becker, 1988).

The complexity of this issue is reflected in the conflicting objectives pursued by governments in this area: simultaneously increasing FLFP and fertility. The institutional context sets out what type of role division is appropriate or makes it more or less difficult to combine both work and family. On the one hand, if the government puts more emphasis on increasing the number of first births, then a lump sum or increased cash benefit payment approach would be appropriate (Gonzalez, 2007). Moreover, since the amount of the benefit typically is not high enough for the most of the women to exit the labor market, therefore, women will decrease the intensity of their participation, namely their work hours (Sánchez-Mangas & Sánchez-Marcos, 2008). On the other hand, if a government is focused on maintaining a high FLFP, then enlarging expenditure aimed at child care and tax changes for families would be the strategies pursued (Powell, 2007; Schøne, 2004). The issue with targeting benefits is that the existing family policy often supports one kind of family model (Steffen, 2011). This tends to be the socially accepted one, as elaborated by Bourdieu in his essay “On Family as a Realised Category” (1996). Therefore, in typical male breadwinner societies one can expect the primary benefactors of social policy to be heterosexual married couples with children. Such systems focused on typical models will have secondary earners tax and little or no preschool subsidized child care (Haan et al., 2010; Kaygusuz, 2009). Policy Instruments used in family policy include non-contributory benefits(universal), based on contributions and taxes. For the purpose of this study I will focus on cash-benefits and benefits in kind.

Non-contributory benefits or universal child care benefits (including family allowance): Gonzalez (2007) examined at the universal benefit which came in one-off payment of 2,500 euros per child born after July 2007. The author found that while the benefit may have increased short-term fertility it had a negative effect on maternal labor supply as women chose to stay home to care for their child, however, their household expenditure did not increase as it would if the parents chose to use formal or informal day-care. This approach is often used to give fertility a boost, however, as Gonzalez has established the effect is mainly short term on first time births.

Benefits in kind (childcare option): the expenditure on child care subsidies and other than home care for the child, will reflect the countries attitudes toward working mothers. Powell (1988) found that when females have subsidized child care, they are more likely to increase the intensity of their

increasing expenditure on childcare has a positive effect on FLFP and consequently the number of hours worked.

In response to demographic and labor market challenges, governments have increased spending on families (Kalwij, 2010). This is done with the goal to counter the fertility decline and maintain or increase the FLFP. This thesis assumes that women are rational-actors that decide themselves about the intensity of their labor market participation if they can afford to do so. Therefore, this research will investigate how the presence of family policy affects the intensity of FLFP. The first part of the thesis will attempt to answer the research question: How does state expenditure on benefits in kind and benefits in cash affect female work hours (intensity)? And how does the presence of different family policy, namely ECEC, birth grant, parental leave, income maintenance and child allowance affect female work hours? How does this effect differ by group?

Previous literature on attitudes has some interesting findings regarding how attitudes among men regarding household work and child care have changed over last years (Andre et al., 2013). In this aspect, it is essential to look at how different family policy instruments influence men's support toward traditional attitudes. André, Gesthuizen, and Scheepers (2013) found that in countries that have higher expenditure on child care, men have less traditional perceptions of gender roles. This contrasts with the more traditional views in countries where cash benefits were a more commonly used policy instrument. In this study, heterosexual married women are the group of interest, therefore I will exclude attitudes towards single mothers, LGBTQ and homosexual couples. Hence the second part of this research will try to answer the following question: how does state expenditure in benefits in kind and in cash influence support for traditional gender roles? And how do different policy instruments influence attitudes toward traditional gender roles?

The academic relevance of this topic is to fill in the gaps in the literature with regards to mother LFP from the perspective of the mother being an economic-utility self-maximizing individual, rather than a victim of her circumstance. Secondly, with fertility decline throughout the world and resultant population aging, governments need to be careful about which policy instruments they pursue (Gauthier, 2007; Fehr, & Ujhelyiova, 2013). A thing to consider from a future challenge in fertility policy is the consequences of delayed fertility. Gustaffson (2001) argues that it is likely that alternative fertility treatments will become a significant part of government medical expenditures in the future Therefore, delayed fertility choices will have a direct influence on state medical expenditure. Third, the majority of the literature suggests that men play a significant role in women’s choice to decrease

her work hours (Gupta, 2007; Stone, 2008). As such, it is of vital importance to study how family policy systems influence perceptions of males, since it might be a counter-active measure. In other words, the family policy can either make men more or less likely to support traditional gender roles. A counterproductive family policy may then affect the disposition of males towards engaging in child-care themselves. Additionally, this can influence their preference for dividing labor by gender and becoming the primary earner (Stier et al. 2001; Riphahn & Wiynck, 2017; Neyer & Andersson, 2008). The contribution of this paper to the field is the inclusion of interaction between different levels of education and the following policies, birth grant, ECEC and parental leave to estimate the effect of these policy instruments on female work hours. Furthermore, since the second part of the study focuses on changes in attitudes, the contribution to the relevant field is the following country comparison by year and gender. By utilizing this approach between country differences and inter-country differences become clear.

This thesis consists of two parts. The first part will investigate the effect of aggregate state expenditure on family policy benefits in cash and in kind and different instruments of family policy on women's work hours. The second part of the research, the focus will be on how the different types of family policy instruments influence traditional gender role attitudes of men and women. For the goal of studying these effects, four waves of International Social Survey Data 1988-2012 will be analyzed. The paper will be structured as follows; first section presents the theoretical framework with relevant literature. The second section outlines the data and methodology used for the analysis. The third section shows the analysis section, where the outcome of the analysis will be presented. The fourth section will be concentrated on a discussion of the results. The last section will present conclusions and further policy suggestions, as well as address limitations of this study.

II. Theoretical Framework

PT. 1: Economic Analysis and Policy

The first part of the theoretical framework will focus on how aggregate level social expenditure on family policy instruments influence female work hours. The theoretical framework consists of two parts. Part one presents relevant literature regarding family policy expenditure and female hours worked. In part two, the literature regarding on societal attitudes and relevant research will be discussed. Both parts will be looked from the individual specific context and institutional context. The

could shape individual preferences while individual level characteristics shape preferences (Hakim, 1995; 1996; Shreffler & Johnson, 2013; Stadelmann‐Steffen, 2011; Gesthuizen et al., 2011).

First, I discuss the particular context of female decision making, elaborating on factors influencing her choice to reduce her work hours or exit the labor market. In this section theories regarding individual preferences of women, with regards to their investment in education and fertility timing will be discussed. Second, the institutional context will be investigated. Further I argue the motivation to include aggregate expenditures, namely total amount of public expenditure, benefits in kind, benefits in cash % of GDP, and dummies for Early Childhood Education Care (ECEC), day care, birth grant, income maintenance and parental leave. For the empirical analysis, education levels for respondents are divided in three groups – lower/primary education level, secondary/upper secondary education level completed and lastly lower/upper tertiary education level. The effect of different policy instruments per education level will be investigated.

Individual Context

FLFP is strongly linked to the concept of human capital. Becker’s (1988) theory of human capital (Walker, 1985; Becker, 1981), female opportunity cost, and time allocation. As female investment in education increases the costs of having quality children rise as well. In developed countries, low rates of infant mortality have resulted in fewer births and higher monetary investment (education and extracurricular activities), effort (time spent with children) in the children born (Becker & Lewis, 1973). This is the tradeoff between quantity and quality of children; since most children live a full life, more investment is put in these children. As parents strive for quality children, it has become more expensive to have many children. This is often cited as an alternative reason for declining fertility rates throughout the developed world and accounts for decreasing family sizes.

Nevertheless, Becker (1981) argues that both time and goods are inputs in the production of this “commodity” (the child) which cannot be purchased on the market place but is purchased and consumed by households during market purchases; such commodities include “children, prestige & esteem, health, altruism, envy and pleasures of the senses” (Becker, 1981; p25). As mentioned above, these commodities do not have market prices as they are deemed not purchasable. However, they do have shadow prices resembling the costs of production (Becker, 1981; Chapter 1). A shadow price is the forgone earnings of the woman during child care. Therefore, it is not only the foregone market income but also time allocation from paid labor to unpaid labor. This shadow cost is the opportunity

cost for mothers, which Becker argues has led to decreased fertility and increased FLFP. Following this line of thought, governments increase their spending on family policy to compensate for the costs of the child in order to address the declining fertility and the LFP dip by mothers.

Moreover, with much focus on different labor division by gender, this distribution has been attributed to particular investment choices of women and men, respectively (Becker, 1981). This approach states that women are biologically predetermined to have higher human capital in childcare due to their time allocation to bearing and raising the child. The human capital approach has recognized that the time spent in that activity positively relates the motivation to invest human capital in a particular activity. By assumption of complete equality, Becker argues that women biologically have comparative advantages of child care, since they have considerable biological and time investment in their children and historically have spent more time in childcare compared to men, for whom investing in market capital has been the main focus (Becker, 1985). Following this theory of particular investments of the human/market capital, it increases household efficiency if the allocation of time is done accordingly by comparative advantages.

Hakim (1995) builds on Becker's economic utility arguments. Hakim establishes that individual preferences of these women are just as important as institutional structures. Some women have invested more in education; therefore for highly educated women, the opportunity cost of foregone earnings is higher than for medium or low educated women (Bratti, 2001; Walter, 1995). Hence, Hakim argues that the dominating feminist view of women as victims of patriarchy is not always correct. She elaborates against the assumption that women are victims and they stay at home to take care of children/family because they have no real assistance in reducing their burden. In her article "5 feminist myths" (1991) Hakim goes against the popular feminist notion that women, if they could, would choose full-time employment. In reality, it depends on the education level of the woman, which translates into investment costs of education. She builds upon household utilization theory by Becker (1981) in an attempt to better understand the behavioral choices of women. Hakim finds that the response to changes to the generosity of benefits will have different effects based on education levels. Furthermore, females have more autonomy over their career-choices than the majority of literature wants to give them (Gupta, 2007; Shafer, 2011). Hence, the goal is not to "free" women from their household responsibilities, instead establish which groups react to different kind of policies and how they react.

H1: Higher educated women will react differently to benefits, than medium and low educated women. Higher educated women will report a higher effect of parental leave compared to medium and low educated women.

Labor Force Participation Intensity

The main difference between married and single FLFP is the intensity of participation (how many hours per week). The rise in participation rates might be out of economic necessity for dual earner households or due to actual preference to work, therefore, it is essential to emphasize that not all employment is equal (Powell, 1998). It is important to distinguish between full-time employment, part-time employment, and choice to do unpaid labor (household). It is expected that different employment types will show varying levels of labor market attachment. A woman might have high-intensity work (in terms of hours), but not be that attached to the job. This is most common for poorly educated women who have to work out of economic necessity, if they could reduce their hours worked or completely exit the labor force, they may do so. A woman who has a preference to work will temporarily leave the labor force instead of complete exit (Powell, 1988; Shafer, 2011). Such differentiation among work types is essential, as the women who work full time will be more committed to their jobs than part-time workers, but might not need them as an economical means of survival (Powell, 1998; Gupta, 2007). This can be attributed to different preferences of the women – which most often arise from differences in education and family history (Shafer, 2011; Hakim, 2002; Walter 1975).

While some policymakers focus on increasing FLFP, of no lesser importance is creating the attachment to the labor force for women. One of the leading causes for women exiting the labor market or reducing the full-time position to part-time is childbirth, which explains the dip in LFP for mothers. It has been argued by many that the absence of affordable child-care before primary school or kindergarten increases the time the woman has been out of the labor market and this, in turn, leads to difficulties obtaining a job which would be combinable with family care (Sheffler, 2013). However, while the relation between affordable child care and FLFP has support in literature, the assumption that all women will start working full-time if child-care is readily available is not supported by evidence (Hakim, 1996; Gupta, 2007). Nevertheless, Powell finds that affordable child care leads to increased work hours for lower and medium education women (1998). Often it is emphasized in literature that both , temporary exit from labor force and child care options for the mother, are factors

that influence hiring of female employees. Furthermore, to fulfill household and family commitments, women require more flexible work hours, which lead to employers not willing to invest in such employees as they might appear less dedicated compared to their male counterparts (Gupta, 2007). H2: Medium and low educated women will increase their hours worked in presence of ECEC, while for high educated women there will be no or nonsignificant effect. High educated women are more likely to have low effect due to outsourcing child care, while medium and low educated women more often use available public child care options or relatives.

Institutional Context

Bourdieu in his essay “On Family as a Realized Category” (1996) argued that states social policy reflects the norms and beliefs dominant in society, reinforcing roles through social policy tools. The main benefactors of the majority of family policies are heterosexual, married couples. This leaves out single mothers or fathers, as well as homosexual couples, with the former being at higher risk of poverty. Therefore, while there is much discussion regarding policy influencing attitudes or attitudes influencing policy, we cannot exclude the institutional factor that sets the societal norms through family policy structure (Steffen, 20011; Shreffer & Johnson, 2013).

Epring-Andersen in his work “ Three Worlds of Welfare Capitalism” (2013) created a typology consisting of three leading welfare state types: conservative, social-democratic and corporative. Social-democratic states are expected to have more policies that would allow family-work life reconciliation for working parents, as they are more egalitarian in their structure. For the conservative and corporative states, where the typical breadwinner model is dominant, there will be less investment in subsidized child care, than child benefits. Only evaluating the institutional context on welfare regimes is not a great predictor by itself, since there has been much discussion regarding the adequacy of the typology (Steffen, 2008). Larsen (2008) found evidence that the basic regime characteristic such as the degree of universalism in welfare policy, the differences in economic resources between the lower and upper economic groups and the degree of job opportunities influence support for welfare policy. Overall, research has established a link supporting the credence that institutional context and has influence for support for comprehensive family-work life policy and/or traditional gender roles (Stier & Braun, 2001).

Cash benefit

High state compensation for child-related costs often are accompanied by lower FLFP (Shafer, 2011). Since both demographic rejuvenation and prosperity are both crucial concerns to country governments, there has been an attempt to strike a balance between these both needs. The conviction regarding the cost of the child as being the main reason for reducing number of births governments have steadily increased their expenditure on family policies (Becker, 1986; Kalwij, 2010). However, targeting of these benefits will influence the likely result, and as aforementioned, some groups will react differently than others (Powell, 1998).

The phenomenon of fertility and mother LFP are both intervened, since most women at some point in their life will give birth and temporarily exit the labor force. Bratti (2003) in his study on LFP and fertility behavior of married Italian women, found that while most women bear opportunity costs, there is a difference in timing of births depending on the level of education. The main findings were; higher educated women have a more considerable stock of pre-marital work experience, leading to delayed births, as for these women an early withdrawal from the labor market is costly, both in terms of current opportunity cost and future accumulation of capital (Shraffler & Johnson, 2013; Bratti, 2003). In other words, highly educated women have higher opportunity costs of having a child in their younger years, since this time is crucial in order to accumulate work experience. Therefore, it is expected that higher educated groups of women will have less negative or significant effect on their work hours by cash benefits, as the gains from this are low. It is expected, that also higher educated women will have the highest effect on their work hours per week from parental leave, it is expected that effects of parental leave will be more evident among the higher educated women.

Boca and Sauer (2009) use a developed dynamic utility maximization model of FLFP and fertility choices of married women. The authors looked at married woman labor activity in France, Spain, and Italy. In line with Haan and Wrohlock (2010), the authors find that first-order state dependence is an essential factor in FLFP and fertility timing choices in all compared countries, an additional important factor is the local labor market flexibility and characteristics. The authors found that women in France are the most active due to their flexible labor market and family policy, followed by Italy and Spain which have similar rates. To test the possible impact of French policy the authors created and applied a simulation in which they tested the female labor market response with the French system. They found that it did lead to increased FLFP both in Italy and Spain, while the Italian and Spanish systems led to reduced labor activity.

H3: Cash benefits will have a negative effect on medium and low educated women's hours worked per week

H4: The presence of a birth grant will have a negative effect on hours worked married female with children

H5: The presence of a birth grant will have a negative effect on hours worked for low and medium educated women, no effect for high educated women

Child care for working mothers

It is argued that increasing expenditures for child care would increase mother labor participation, since it would help them combine paid labor and household labor. Jamolette (2003) found that higher public investment in childcare stimulates full-time female employment, as women are enabled to combine work and family responsibility. Nevertheless, the findings Shafer (2011) reinforce the findings by Powell (1998), that the effect differs per group. Haan and Wrohlich (2011) come to similar conclusions in their study of reform in Germany that was implemented in 2013. The reform aimed to provide a working woman with subsidized childcare for their younger children. The authors found that by providing affordable early childcare options, in the model, the average participation rate rose by 1.6% and working hours by 2.4%. There were no significant effects on fertility of the reform, however for subgroups of highly educated women and women who will give birth for the first time an increase in subsidized child care has a positive effect on fertility. For many women, the cause for exiting the labor force is the difficulty to find available and affordable childcare. Consequently, Carta and Rizzica (2018) studied the effect of a reform in Italy which introduced early childcare for 2-year-olds. They found that the policy did have a positive effect on FLFP through females lowering their reservation wages thus increasing their probability of getting a job. This is supported by previous research that women respond to changes in childcare costs in places where subsidized child care is more expensive and scarce female labor participation is lower (Connely, 1992).

H6: Child care options, such as early child education care , that provide a possibility for out of home care will have a positive effect on hours worked of married mothers.

Part II : Support for Traditional Gender Roles

A considerable amount of women name the choice to reduce hours worked influenced by difficulty of combining paid work and household work. Two most popular theories, namely economic dependence and gender display, attempt to explain the causes of this household work disparity. The former is a collaborative assumption of the household dynamics; the man and the woman in the household concentrate their efforts in the labor market to create joint utility for the household members (Becker, 1981). Because men's wages were historically higher, they spend more time working for pay and women spend additional time doing domestic labor (Becker, 1981; 1991). The economic dependence approach is gender neutral, because the individual with fewer earnings will take on more unpaid household labor. The latter approach, in contrast, states that partners who have atypical economic circumstances, a woman earning much more than the husband, are expected to overcompensate their gender-normative housework performance.

Stone (2008) finds that women privilege their husband's career over their own when the husband is passively supportive of hers. However, that does not mean they exit the labor market due to a return to traditionalist values. Women emphasize the difficulty of combining careers and families. The role of the husband in much of the research has been with regards of his work hours, income and profession, nevertheless, the level of support for traditional gender roles males have is as much of interest. Nevertheless, the study of Stone(2008) was done on highly educated women, with successful careers, who could leave the labor force due to their husbands income being high enough. As stated in the paragraphs above, husbands work hours and income have a significant influence on their wives decisions to leave the labor force. Based on previous literature (Gupta, 2007; Shafer, 2011) men still do not spend as much time on household work and family care, when compared to women. However, attitudes with regards to what is expected of both parents are changing, as currently parental leaves for both parents are gaining traction in policy (Adema, & Whiteford, 2007). A study by André, Gesthuizen, and Scheepers (2013) found where governments spend more on childcare, there is a noticeable drop in traditionalist gender attitudes.

Therefore, for the second part of the research, I will investigate attitudes toward traditional gender roles. While it is unknown if the policy affects beliefs or attitudes affect policy, as the laws form beliefs by internalization of social norms, however, laws can also change as the attitudes of social change, and some deeply held beliefs become rejected in modern society (Larsen, 2007). One can expect countries that provide more options for childcare and the opportunity to maintain a work-family

balance, to be more likely to reject traditional gender roles. Furthermore, in countries where there is higher FLFP (not only mothers), it is expected that society will be less traditional. In the second part of this study, I will build upon the research of André, Gesthuizen, and Scheepers(2013), who used the same data set. The authors of this study build on two theories, the theory of socialization by Starke (2004) and rational choice theory by Downs (1975). For the former, the study used an exposure based explanation in which socialization is seen as a lifelong process which is internalized, instructing how to behave and rationalize in society.

Furthermore, since girls and boys experience different socialization based on the expectations of gender roles from their socializing agents (parents, teachers). The latter, interest-based explanation maintains that individuals choose to further their interests. However, the study focused more on differential socialization than rational choice theory. The authors looked at contextual and individual level analysis. Contextual determinants are family policies in countries, which the authors propose that family policies affect female role attitudes of men and women differently. The individual is the differential socialization persons go through as girls and boys. The authors specify multiple hypotheses with factors that could affect traditional gender role support, such as if the respondent had a working mother and comes from a younger cohort, if the subject is invested in education, how attached the individual is to the labor market, marital situation, and lastly is religious and how frequently attend religious services. The study had two research questions, first focused on the extent of difference in traditional gender roles between countries and years, secondly what is the effect of total public expenditure, benefits in kind and cash, as well as family policy instruments on individual and country level support for traditional gender roles.

H8: Countries that spend more on total public expenditure in family policy will have higher support for traditional gender roles

H9: Countries that spend more on benefits in cash will have men who are more supportive of traditional gender roles, on the other hand, countries that spend more on benefits in cash will be less supportive of traditional gender roles

H10: Child care availability ECEC and day care will have a negative effect on support for traditional gender roles

research, spanning over four waves of surveys. Third, I include in the analysis different family policy instruments dummies to measure their effect on support for traditional gender roles. The following section is data and methods, where I describe my approach. The first part of this study will first analyze the effect of different family policies (ECEC, daycare, birth grant, parental leave, income maintenance, child allowance) on FLFP (estimated as work hours). The second part of the study will focus on changes in attitudes over time, and estimating the effect of expenditure and policy on these attitudes.

III. Data and Methods

For this research, I will use data obtained from International Social Survey Program (ISSP), more specifically the module “Family and Gender roles” that has four waves: 1988, 1994, 2002, and 2012. Countries were sampled in a multi-stage random stratified sample in the first two waves. The selection procedure for the latter two waves (02 and 12) differed by individual countries. The sampling procedure used there was an either partly simple or partly multi-stage random stratified sample. The data collection was done either face-to-face interviews, by mail or written survey (in 2012 could be done partly in the web). For the estimation strategy this research will use a OLS estimation on multiple linear regressions (MLR). While other authors who used this dataset opted for a multilevel estimation and analysis, I chose the OLS estimation in order to obtain an unbiased sample. In order to obtain an unbiased estimation I chose this approach, since the country sample size is around 22, which is a smaller sample compared to André et al (2013) or Gupta (2007) in my sample.

For the comparative cross-series study, I exclude countries that were only in one wave and those who were lacking information on work hours and employment status. I exclude Taiwan, Philippines, and Brazil due to some questions having different response codes resulting in unequal responses. Additionally, I remove the countries from the sample that have only featured in one wave. And countries that had more than 50% of responses were missing for the primary variable of interest - hours worked per week.

Table 1: Operationalization of the variables used in analysis

Variable Recoded value Definition/label

Female hours worked

Sample reduced only to women, work hours of men are set as missing values

Hours worked per week

Employment status 1. Full time employment 2. Part time employment 3. Housewife

4. NILF

Indicates employment status of the respondent

Education Group Generating education groups by years spent in education:

1. Spent 13 years in education 2. 12 to 16 years

3. 15 to 19 4. 18 to 27 5. 28 and more

Grouping of education years

Education 1. Lower/primary education 2. Upper/post-secondary education 3. Lower/upper tertiary

Recode of DEGREE, education status by highest obtained degree

Dependent children in the household

Binary coding: 1 if there is at least 1 child in the household and 0 otherwise

Household had a child or toddler in the household

R mothers work experience

Binary coding: 1 if yes and 0 if otherwise Mother of respondent worked when respondent was <14

Attendance of religious services

Recoded to : 1. Weekly attendance, 2. Monthly attendance, 3. Several times per year, 4. Once or less than in year, 5. Don’t attend religious services

Respondent attendance of religious services

Income group Due to differences by country currencies, income group is coded by percentiles from country income so income groups are comparable across countries

Respondent is in the 10%, 20%, 50%, 75%, 90%, 95%+ income group

Marital status Recoded: 1. Married, 2. Civil partnership, 3. Single/divorced

Indicates respondent marital status

Policy: ECEC,

child allowance, parental leave, birth grant

Binary coding: 0 if a country has the policy and 1 if otherwise

Aggregate expenditure of GDP ; binary code for family policy kind

Part I:

The first model will investigate the relationship between female weekly hours worked and expenditure on benefits in cash and kind, the second model will include dummies for different family policy instruments.

Dependent variable

Weekly hours worked: This variable will be kept in its original form and it ranges from 0 hours per week to 89 hours per week. Since I am interested in female hours worked in particular, I create a variable that only captures the work hours of women.

Independent variables

The independent variables used for this analysis will be total public expenditure, expenditure on benefits in cash and expenditure in benefits in kind. Overall the aggregate expenditures are expected to show the general position held by a country government (Kalwij, 2010). Additional controls are dummies for family policy, coded 1 for yes if the policy is available in the country and 0 for otherwise. These indicators were obtained from OECD aggregate social expenditure data.

ECEC /Day care / Child allowance: by decreasing costs of kindergarten/nursery the household disposable income will increase if the female returns to the labor market. In this case, the opportunity costs of not working are too high for women to remain inactive. In countries where there is no subsidized child care, there might be child-care allowance that can be used for child-care, however often it leads to an opposite effect, women instead choose to stay at home and therefore have an income effect than pay for child-care. Instead of using expenditure amounts, binary variables representing if there is such a policy will be included in the analysis. The choice for including binary variables instead of the amount spent per policy is due to the main interest of the study – what is the effect of policy on the labor force intensity of women. This effect is best captured by a binary variable that records the presence or absence of the benefit.

Birth grant: I generate a dummy variable from the social expenditure data, indicating 1 if the country has a birth grant and 0 if otherwise. I include this as an independent variable, as states that employ this policy do so with a purpose to increase fertility rates. The data are obtained from OECD Social

Aggregate Expenditure Data. It is expected that female work hours will be affected if the mother receives a lump sum, family allowance or child allowance.

Total public expenditure %, of GDP: is the amount spent by country in its total public expenditure in as % of GDP. This expenditure is intended to be used as a reference point of the state policy preferences.

Benefits in kind/ in cash: I include the aggregate expenditure on benefits in kind and benefits in cash on family policy by state. Based on previous literature it is expected that countries that spend more on benefits in kind, will lead to more hours worked per week by married women , since benefits in kind also include the expenditure on family policy. Benefits in cash are expected to have a negative effect on weekly hours worked, since the amount received , while not enough, could lead to the woman substituting forgone market income and reduce her hours worked per week.

Controls

The second model includes individual level characteristics, therefore the regression equation changes. Below the changed regression equation is presented. Multiple factors are significant for FLFP, which will be discussed below. It has been established by, that the likelihood of a woman returning to work depends on her education level, the availability of subsidized child care from the state, her religious status, employment type, the number of dependent children in the household and the benefit received. More detail on each of the confounding factors and reasoning behind including them in the analysis is described below.

Presence of dependent children in the household: Childcare arrangements of who takes care of the child (relatives, grandparents, public, private), influence the cost of the childcare and thus the fertility/labor market choices for women. However, there is a difference between the ages of the child in the household. If the child is <6, it is expected that females will reduce their weekly worked hours if there is no suitable child care option. The variable is coded as 0 if there are no children and 1 if there is a dependent child between 6 to <17 in the household. Since this study does not differentiate among toddlers and children, the presence of a toddler in the household as a presence of dependent children in the household.

Education groups: Previous studies have shown that women who have invested more in education will have higher work hours, as they have a higher opportunity cost of not working. For analysis, education will be divided into years spent in education. This variable is structured as: no education, respondents who have spent up to 13 years in education, 12 to 13 years spent in education, 15 to 19 years spent in education, 15 to 19 education years, 18 to 28 years and lastly more than 29 years spent in education.

Marital Status: It is expected that females that are married will reduce their hours spent in the labor market and increase their hours spent in unpaid work (Becker, 1991; Gupta, 2006). Therefore it is an essential indicator for the intensity of labor market participation for females. Furthermore, since marriage rates are declining, civil partnership is also included. However, it might be of interest to investigate the exact differences between women who are married and who are in a civil partnership, as it might be expected that married females are more traditional than those in a civil partnership. Since we are interested in married women response to work hours, widows are excluded, as they are outside the sample of interest (Gupta 2006; 2007; Shafer, 2011; Powell, 1998)

Part II:

Institutions that implement family policy frequently reflect the political will or agenda of those in political decision making. Hence, it can be expected for policy to reflect the norms and normative expectations of gender roles, both for men and women. On the one hand, countries that focus on having high FLFP will spend more on alternative child care options and therefore, hold fewer traditionalist views on females combining work and family. On the other hand, countries that focus more on nucleus family model, where women remain the primary caretakers of the family, it is expected that the availability and expenditure on outsourcing childcare will reflect more traditionalist attitudes for both men and women, as well as have lower female LFP. Second, institutions not only influence the norms but also how individuals perceive their situation. Therefore, the presence of family policy will influence how individuals, more specifically women, perceive their opportunity in the labor market. This effect is expected to differ by education levels, since ceteris paribus, countries that have a more egalitarian policy will have a higher educated female population, which will lead to higher LFP.

Dependent Variable

For the second part of the research, the focus will be on attitude changes, by utilizing the data reduction strategy done by André, Gesthuizen, & Scheepers (2013). In order to capture the effect of attitudes, factor analysis for data reduction will be used. Since the dataset it is a survey of attitudes, the questions asked to reflect specific dimensions of attitudes. Hence, this seemed the most appropriate approach to capture the right attitudes when compared to using a single answer item as the outcome variable. The authors in their research utilized the principal component analysis due to the high correlation among survey questions as they represent an underlying dimension. This thesis follows André et al. (2013) approach to the variable selection to capture attitudes. In their study, they excluded item “ A working mother can establish just as warm and secure a relationship with her children as a mother who does not work” due to its low score. However, since the dataset used in this study is a repeated cross-section instead on single year analysis, I choose to include the item:

1. A working mother can establish as warm and secure a relationship with her child just as much as mother who does not work

2. A pre-school child is likely to suffer if his/ her mother works 3. Family life suffers when the woman has a full-time job

4. A job is all right, but what women really want is a home and children 5. Being a housewife is just as fulfilling as a job

Items 1 through 5 were coded in a Likert scale ranging from 1 strongly disagree to 5 strongly agree. The item “A working mother can establish just as warm and secure a relationship with her children as a mother who does not work” was reverse coded to 1 strongly agree and 5 strongly disagree in order to fit in the scale, following the approach set by André et al. (2011) used. If the item remained as it were, then a higher score would indicate support for working mothers, while other items on the scale would score higher if they show less support. All of the items scored above .4 recommended factor value. All items selected reflect attitudes toward working mothers and the consequence of her choice on family life. The first two items reflect attitudes toward how mothers work affects her relationship with the child. The third item applies the change in family life if a woman chooses to work. Items 4 and 5 show attitudes toward women in more traditional roles. For analysis, the aggregate mean values per country and year are computed.

Country-level indicators consist of total public expenditure as % of GDP, real GDP forecast, expenditure on benefits in kind and cash as % of GDP. To answer the research question: how are attitudes toward traditional gender roles influenced the reaction of the presence of different benefits? Dummy variables for the presence of ECEC, daycare, parental leave, child allowance, family allowance, birth grant, and income maintenance are created. First, I run a MLR on the mean support of traditional gender roles on country-level aggregate expenditures. Second, I regress support of traditional gender roles on individual-level characteristics, controlling for the presence of different family policy instruments. As control variables on the individual level, attendance of religious services, marital status, employment status, mothers work experience and the presence of a dependent child in the household are included.

Controls

Marital Status: Marital status of the respondent is expected to influence the attitudes towards traditional gender roles. Therefore categories of marital status, namely married, civil partnership, single/divorced, will be included. Furthermore, it is expected that individuals who are married will be more traditional compared to single/divorced respondents or individuals who are in a civil partnership/

Attendance of Religious Services: In more conservative and religious regions, it is most likely that the typical male breadwinner model will dominate and it can be expected that more religious men or women the more traditional household labor division will be. In countries that are more secular and have progressive politics, women will likely postpone their pregnancy and have fewer children. Furthermore, attendance of religious services is an essential indicator of the intensity of religious attachment. It is hypothesized that individuals who attend religious services more often, will have more pronounced support for traditional gender roles.

H11: Respondents who attend the church more often will report having higher support for traditional gender roles compared to respondents who do not attend such services.

Mothers work experience: In previous research about attitudes toward working women, it was established that respondents mothers work history will influence their attitudes toward working women. If respondents mother worked before when they were still a child, the respondent is more likely to believe that it is acceptable for a mother to work. Respondents whose mothers did not work

when they were a child are expected to be less supportive of working women. The item will be coded as follows: 1 if the mother did work and 0 if she did not. Below the table with operationalizations of the variables used in the analysis is presented.

Education: previous studies have shown that women who have invested more in education will have higher work hours, as they have a higher opportunity cost of not working. For analysis, education will be divided into years. This variable is structured as: finished only primary school or finished secondary school, and tertiary (no difference between lower and higher tertiary in the study). Education is divided into lower/primary education, secondary/post-secondary, lower/upper tertiary.

Urban-rural: Centralization of the location where the respondent lives can play a role in their support for traditional gender roles. It is not uncommon that in highly centralized countries, individuals from rural areas have higher support for traditional gender roles. Therefore it is included as a control in the model evaluating the effect of individual-level characteristics on the support of traditional gender roles.

H12: Individuals in rural areas will report higher support for traditional gender roles than those from urban areas.

Income group: The income group of the female is an important predictor of her hours worked. Women who are in an higher income group have a higher cost of reduced work hours, as their hourly wage is higher. However, this is a double edged sword, since Stone(2008) and Shafer(2011) found that highly educated women are the group that is most likely to exit the labor force, as they can afford often to do so.

IV. RESULTS AND ANALYSIS

Part I : Effect of benefit presence on female hours worked per week

Part one focuses on providing first descriptive statistics about the sample, then a OLS model presenting the coefficients for the dependent variable work hours in the primary sample of interest: married women and married women with children, to see if there is a difference in the effect. Additionally, to the linear model presenting the effect on mean hours worked by country on different

kinds of benefit types, including aggregate expenditure. After linear models, a table consisting of different regression models will be presented.

Table 2 contains descriptive statistics for the whole sample, not specifying any subsample characteristics. For the whole sample of the population that is employed, the mean hours worked are 39.56. On average, the respondents have spent 11.39 years in education, which is equivalent to finishing secondary school/high school education. Overall the sample means age is 46.5, with almost half the sample being employed full time (.491), with only a minor part employed part-time. Nearly half as much respondents report their occupation as a housewife when compared to part-time. Different employment statuses are not included in the analysis since the focus of the thesis is in full time employed, part-time employed, and homemakers. Furthermore, almost half of the respondents of the sample report having at least one child in the household, compared to .269 of the respondents having a toddler in the household. Lastly, the mean income group of the sample is 2.87, which would correspond with 20% - 50% percentile income. The variable was operationalized by income variable quartiles and had to be grouped in order to use it in the analysis due to differences in currency per country.

Table 2: Descriptive Statistics whole sample

Variable Mean Std.Dev. Min Max

Weekly Hours Worked 39.562 12.8895 1 80

Years in Education 11.392 4.595 0 82

Age 46.539 17.220 15 98

Married .583 .492 0 1

Employed Full Time .491 .499 0 1

Employed Part Time .051 .221 0 1

Housewife .100 .300 0 1 Child .432 .495 0 1 Toddler .269 .443 0 1 Income Group 2.87 1.65 1 6 N= 117,158 ISSP 1988, 1994, 2002 ,2012

Table 3 contains additional descriptive statistics by gender. In table 3, the age of the sample is pretty close for both genders, with mean age 46.4 for women and 46.6 for men. In the primary variable

of interest hours work per week, women work on average 35.6 hours per week, which is 8 hours less than men, who work reported 43.5 hours per week. This is in line with the expectations and previous literature that women work fewer hours per week due to other commitments, such as household duties (Gupta, 2007; Shafer, 2011). The work hours in this table for women might seem higher than expected; however, it is worth to note that this is the whole sample, not specifying age, marital status, or employment status. Men in this sample are overall minorly less educated than women, by 0.227 percentage points. More women than men in this sample are married, with .620 women compared to .560 men. The main differences between the genders are apparent in the type of work. Men are more likely than women to work full time and less likely to work part-time. However, women in the sample report occupation as a housewife (.167), which is at least 9 times more than men (.017). Comparing table 2 and table 3 descriptive statistics, the difference in hours worked by gender becomes evident, as well as employment status.

Table 3: Descriptive statistics by Gender

Furthermore, since weekly work hours differ per age group, reflecting that respondents are in different stages of their professional career, it is expected that younger respondents will work more than older respondents, due to the first years of entering the labor market having high significance for one’s career. When nearing late 20’s and entering the 30-35 age group, most respondents begin to settle down and have families. Overall, there will be in total fice groups based on age 25-30; 30-35;35-40;40-45; 45-50. In the analysis, females above 50 are excluded since their labor market status is less affected by fertility decisions, and below ages 20 are excluded since this age group is still often in education and it is unlikely that they have vast labor market experience that could be influenced by fertility decisions. Table 4, presented below, shows that for women, the most labor-intensive years are in their early 20s with on average 36 hours worked per week. With each additional unit increase in age group, hours worked by married women decrease, with the lowest reported amount of work hours in ages 30 to 35. This can be attributed most likely to fertility decisions by these women, as the hours

Age Hours worked

Years in Education

Married House wife Full time Part time

Female 46.452 35.692 11.518 .620 .167 .408 .074

men, the situation is visibly different than for women. Men regularly work more hours on average than women with the smallest difference in work hours for those aged 20 to 25. Nevertheless, for men, unlike women, hours worked per week continuously increase and remain around 45 hours per week.

Moreover, women increased their hours worked after 45, whereas men decreased their hours after age 45. A possible interpretation of the results is that work experience that allows them to reduce hours worked, and for women, the change can occur due to the reduction of household and family responsibilities. Especially in family care since children are most likely relatively grown up and do not need constant effort and attention from the mother. Overall, this supports the notion that men work more than women per week and women reduce their hours spent on labor market once they settle down and have children. Women and men have different trends after age 45, for women, the hours worked per week show a slight increase, to 35.4 from 34.8 hours per week, and for men, there is a slight reduction to 44.9 hours from 45.4 hours per week.

Table 4: Descriptive statistics - by married age cohorts and respondent sex

Age cohort Hours worked per week women Hours worked per week men

age 20-25 36.043 43.746 age 25-30 35.793 45.113 age 30-35 34.537 45.236 age 35-40 34.915 45.199 age 40-45 34.82 45.449 age 45-50 35.361 44.914

As mentioned in the beginning of this thesis, not all employment types are equal. Therefore, descriptive statistics of the whole sample, grouped by gender are displayed. This provides an insight into the labor force participation rates. Furthermore, since this thesis focuses on female labor force participation and the type of employment women have compared to men, the whole sample employment status is presented in table 5. In this table, it is visible that women are more likely than men to report the occupation as a housewife or part-time work. Furthermore, women are more likely to report being in education, as an apprentice or trainee than men.

Additionally, men are more likely than women to work full time, and women are more disposed than men to work part-time. Both groups report equal levels of unemployment, while there are slightly more women who report being retired or sick and disabled. Women are more likely than men to be NILF. In the appendix, employment status is presented for yearly comparison.

Table 5: Employment frequencies by gender, whole sample

Male Female Total

Full time 31,287 26,663 57,950

Part time 1,269 4,834 6,103

Less than part time 1,215 1,830 3,045

Apprentice/trainee 91 260 351 Unemployed 2,673 2,789 5,462 Student, in school 1,859 2,217 4,076 Retired 10,201 12,242 22,443 Housewife 928 10,912 11,840 Sick, disabled 1,158 1,448 2,606 Other, NILF 740 1,341 2,081 Total 51,421 64,536 115,957

Since I am mostly interested in the effect of the presence of policy on working women, especially mothers, I exclude apprentice/trainee, students, retired individuals, sick or disabled, and NILF. The rationale behind such exclusion is mostly due to the targeting of the policy. Women who are retired and do not work full time, and hence, will have no effect of policies such as parental leave or early childhood education centers. Additionally, persons who are sick of disabled are eligible for other types of benefits compensating for missed income. Apprentice/trainee and students are excluded since they are possibly just entering the labor force and do not qualify for certain benefits and as part of career planning, are more likely work longer hours to increase their labor market experience if able to. This thesis focuses on women who are active in the labor force full time, part-time or report their occupation as housewife.

When comparing by years, it is visible that full-time employment of women has been increasing since 1988. This can be attributed to two things: first, the increase in sample size, and second, women are more active in the labor force.

Further, in this thesis, a regression model for the whole sample of surveyed women is presented. In order to do this, the variable is limited of interest to female work hours. Table 6 presents a linear regression with the weekly mean hours worked by females. Since the regression consists of

independent variables on the aggregate level, I use the mean of female hours worked, grouped by country. The regression model used in the analysis is presented below:

𝐹𝑊𝐻_𝑖𝑡 = 𝛽₀+ 𝛽₁𝑇𝑂𝑇_𝑃𝑈𝐵 + 𝛽₂𝐵𝐸𝑁_𝐾𝐼𝑁𝐷 + 𝛽₃𝐵𝐸𝑁_𝐶𝐴𝑆𝐻 + 𝛽₄𝐼𝑁𝐶_𝑀𝐴𝐼𝑁𝑇 + 𝛽₅𝐵𝐺𝑅𝐴𝑁𝑇 + 𝛽₆𝐷𝐴𝑌𝐶𝐴𝑅𝐸 + 𝛽₇𝐸𝐶𝐸𝐶 + 𝛽₈𝑃𝐿𝐸𝐴𝑉𝐸 + 𝛽₉𝐶𝐻𝐿𝐷_𝐴𝐿𝐿 + 𝑒_𝑖

Where the dependent variable 𝑭𝑾𝑯_𝒊𝒕represents mean hours worked by whole sample, total public expenditure TOT_PUB, benefits in kind BEN_KIND, benefits in cash BEN_CASH, as well as dummies for the policies representing income maintenance INC_MAINT, birth grant BGRANT, day care availability DAYCARE, presence of early childhood education care ECEC, parental leave

PLEAVE and child allowance CHLD_ALL, 𝒆_𝒊 is the error term.

Overall, from aggregate expenditure, only benefits in kind have a significant and positive effect on female mean work hours. With increasing expenditure on benefits in kind, females in response increase their mean hours worked by .76 hours per week, significant at p<0.01 level. From the dummy variables that indicate the presence of family policy, only income maintenance, and parental leave were significant, at p<0.1. Income maintenance has a reported adverse effect on mean hours worked, with the availability of such policy the mean hours per week worked by females decrease by 1.5 mean hours. Parental leave, on the other hand, has a positive effect on mean hours worked, with 1.6 mean hours per week increase. The dummies birth grant, ECEC, daycare, and child allowance are not significant. The year variables all are significant indicators for mean hours worked per week, with the highest effect 21.4 in 1994, and the lowest 18.3 in 2012, significant at p<0.01 level. This difference can be due to changes in the sample size, or women are reducing their weekly hours worked due to switching to part-time jobs or working less due to more flexible jobs.

The reported country indicators in the regression are overall significant. The highest effect in this group is mostly focused in Central European countries. The Netherlands has the highest decrease of 17.7 mean hours significant at p<0.01 level. The second highest decrease in mean hours worked is in the United Kingdom with 16.3 mean hours and decrease of 14.5 mean hours in Belgium and 14.6 in Denmark, both significant at p<0.01 level. Lastly, a reduction of around 14 means hours of the week present in Sweden, significant at p<0.01 level. The smallest effect of this model is in the United States with a reported decrease of 3.2 mean hours worked per week, followed by the Czech Republic with a decrease of 4.6 mean hours per week, significant at p<0.05 level. Poland has minorly higher mean

work hour decrease, with reported 5.4 hour decrease, p<0.1 level, while Hungary reports a 5.7 mean hour decrease in the presence of these benefits, significant at <0.05 level.

Moreover, Slovakia, indicates 7 hour decrease, Slovenia and Switzerland around 9.1-9.4 mean hours decrease per week and lastly France with 10.6 hours decrease. All significant at p<0.01 level. Overall, on average, the presence of expenditure and family policy leads to a reduction in hours worked on country level. Majority of the countries had the coefficients ranging from 10 to 13, indicating that women in general decrease their weekly hours worked in the presence of benefits and total public expenditure. Since in the Netherlands the majority of women are employed part-time, this could be the reason for the substantial decrease in work hours. A similar case in Belgium. Table 6 with country coefficients and year coefficients is available in the appendix.

Table 6: MLR Mean hours worked per week by females

Mean weekly hours worked Coef. St.Err. p-value [95% Conf Interval] Total public expenditure, % of GDP -0.661 0.431 0.125 -1.506 0.183 Benefits in cash, % of GDP 0.167 0.122 0.171 -0.072 0.405 Benefits in kind, % of GDP 0.763*** 0.171 0.000 0.428 1.098 Income maintenance -1.511* 0.887 0.089 -3.250 0.228 Birth grant 0.687 1.671 0.681 -2.588 3.962 ECEC -0.749 0.928 0.420 -2.569 1.070 Day care -0.669 0.689 0.332 -2.021 0.682 Parental leave 1.616* 0.845 0.056 -0.040 3.271 Child allowance -0.422 0.880 0.632 -2.148 1.304 Year FE Yes County FE Yes

Mean dependent var 19.542 SD dependent var 17.821

R-squared 0.564 Number of obs 46440.000

F-test 1579.037 Prob > F 0.000

Akaike crit. (AIC) 397522.159 Bayesian crit. (BIC) 397854.504

*** p<0.01, ** p<0.05, * p<0.1

Since I am interested in different effects of these expenditures and policies by group, sample if further narrowed to mean hours worked by married women, grouped by country. The results of this regression are displayed in table 7. The regression model used is the same as above. However the dependent variable is narrowed down to include only weekly mean hours worked by married females, 𝑴𝑭𝑾_𝒊𝒕.