Content uploaded by Malgorzata Mikucka
Author content
All content in this area was uploaded by Malgorzata Mikucka on Jun 30, 2014
Content may be subject to copyright.
WORKING PAPERS
Working Paper No 2011-09
January 2011
Homemaking and women’s
well-being in Europe.
Effect of divorce risk,
selection and dominating
gender-role attitudes
Malgorzata MIKUCKA
CEPS/INSTEAD Working Papers are intended to make research ndings available and stimulate comments and discussion.
They have been approved for circulation but are to be considered preliminary. They have not been edited and have not
been subject to any peer review.
The views expressed in this paper are those of the author(s) and do not necessarily reect views of CEPS/INSTEAD.
Errors and omissions are the sole responsibility of the author(s).
L’European Values Study (EVS) est une enquête
réalisée au Luxembourg en 2008 auprès d’un
échantillon représentatif de la population résidante
composé de 1610 individus âgés de 18 ans ou plus.
Au niveau national, cette enquête fait partie du
projet de recherche VALCOS (Valeurs et Cohésion
sociale), conancé par le FNR dans le cadre du
programme VIVRE. Au niveau international, elle
est partie intégrante d’une enquête réalisée dans
45 pays européens qui a pour objectif d’identi-
er et d’expliquer en Europe les dynamiques de
changements de valeurs, et d’explorer les valeurs
morales et sociales qui sous-tendent les institu-
tions sociales et politiques européennes
(www.europeanvaluesstudy.eu).
Plus d’infos : http://valcos.ceps.lu.
Homemaking and women’s well-being in Europe.
Effect of divorce risk, selection and dominating
gender-role attitudes.∗
Małgorzata Mikucka
CEPS/Instead, Luxembourg
January 2011
Abstract
Whereas it is known that employment affects individual well-being, the lit-
erature on the effect of homemaking is so far inconclusive. The paper inves-
tigates the effect of being a housewife on well-being of women, using Euro-
pean Values Study data for 36 European countries (year 2008) and multilevel
regression methodology.
Results show that the effect of homemaking on well-being is overall positive
and it varies across countries. Three possible explanations of this variation
are examined.
First hypothesis concerns traditional gender-role attitudes in a country. Re-
sults confirm that in more traditional countries homemakers have higher well-
being, but only in western Europe. Effect of individual norms is strong:
housewives with traditional gender-role attitudes declare higher well-being.
Second hypothesis refers to the economic risk of specialization to homemak-
ing, and states that higher divorce risk decreases well-being of housewives.
Contrary to expectations, higher divorce risk in a country is accompanied by
higher well-being of housewives. I interpret this as a sign of equality concerns
incorporated into legal divorce procedures.
Third hypothesis concerns positive and negative selection to homemaking.
Results show that the relationship between prevalence of homemaking and
the well-being of housewives is curvilinear. Highest well-being gains from
homemaking occur in countries with lowest and highest prevalence of home-
making.
Keywords: well-being; homemaking; housewife; women’s employment
∗This research is part of the VALCOS project supported by the Luxembourg ’Fonds National de la Recherche’
(contract FNR/VIVRE/06/01/09) and by core funding for CEPS/INSTEAD from the Ministry of Higher Educa-
tion and Research of Luxembourg. Małgorzata Mikucka is supported by an AFR Grant (n◦PDR 2010-1) from
the Luxembourg ‘Fonds National de la Recherche’ cofunded under the Marie Curie Actions of the European
Commission (FP7-COFUND).
1 Introduction
Despite the three waves of feminism, anti-discrimination laws and gender-parity regulations in
parliamentary elections, homemakers still constitute a considerable group among working-age
women in Europe: currently, in countries of European Union there are about 45.5 millions
of housewives1. Are these women just a relic of the past? There are reasons to expect that
homemaking is bad both for them and for societies. Mainstream economics stresses the bene-
fits of earning income. Higher market productivity (obtained, among others, by entering labor
market by women) strengthens the economy, and on the individual level - enhances women’s
personal freedom.
Mainstream economics stresses the advantages of full employment. However, recent de-
velopments of happiness economics provide evidence that money makes us happy only tem-
porarily and mainly at lower income levels. For people whose basic needs are satisfied, of
major importance may be other factors: satisfying social relationships, lack of stress and time
pressure, etc. In these terms homemaking may be beneficial, both for women and for their
families.
However, other developments of happiness economics show that employment itself makes
our lives better. Contrary to the assumptions of mainstream neoclassical economics, which
treats work as a burden that must be compensated by wage (Brereton et al., 2008), employment
is overall beneficial for well-being, as long as the hours of work don’t exceed certain limit
(Glass and Fujimoto, 1994).
The debate this sums up to a question: what is the effect of homemaking on well-being
of women? Answering this question is the first goal of this paper. The issue is important not
only in the context of happiness economics. The answer may also help us better understand
the cross-country pattern of women’s employment.
Literature suggests that homemaking may affect well-being differently in different coun-
tries. In general, well-being (Bonini, 2008, Böhnke, 2008, Helliwell, 2003, Diener et al., 1995)
and also its determinants (Bonini, 2008) differ across countries. Comparative analyses (such
as Kalmijn (2010), who focuses on the effects of divorce or Boye (2009), who examines the
1LFS data: women inactive and not seeking employment due to “looking after children or incapacitated
adults” or “other family or personal responsibilities”.
1
well-being effect of women’s employment in various countries) show interesting patterns and
encouraging conclusions. For example, Kalmijn (2010) shows that the negative effect of di-
vorce on well-being is weaker in countries with stronger family networks, with less traditional
family values and with higher prevalence of divorce. As a second and complimentary goal of
this paper I want to establish if the effect of full-time homemaking on well-being of women
differs across countries. Furthermore, I explore the pattern of this variation. In particular, I re-
fer to social pressure, divorce risk and employment prevalence as possible explanatory factors.
The novelty of my work sums up in three points. First, I focus on a topic that is cur-
rently treated marginally. Relationship between women’s homemaking and their well-being
was in the center of the debate in the 70’s and the 80’s. Availability of new data, new sta-
tistical methods and theoretical developments of happiness economics create conditions for a
fruitful re-examination of the question. Secondly, cross-national analyses of determinants of
well-being are still rare, and existing literature is limited mainly to economic and psycholog-
ical factors. My analysis contributes to existing literature by focusing on social and cultural
country characteristics that may not only affect the levels of well-being but also the happiness
equation itself. Third, I use recent data covering a wide range of countries. Results of cross
country analyses may depend strongly on the countries included in the sample, which may
limit generality of the conclusions. Sample used consists of countries very diverse in terms of
history and current economic and cultural conditions. This constitutes an important advantage
in a field dominated by US data.
The paper is organized as follows. I start with a brief literature review and formulating
hypotheses; then I move to description of the data and statistical method used. Analysis starts
with inspection of descriptive results. Subsequently I estimate a set of multilevel models to
formally test the hypotheses. I conclude with summary and discussion of results.
2 Literature review and hypotheses
Contrary to unemployment, which was systematically found an undesirable and involuntary
state (Clark and Oswald, 1994, Helliwell, 2003, Lucas et al., 2004, Pittau et al., 2010, Gerlach
and Stephan, 1996), the literature on the well-being effect of full-time homemaking is incon-
clusive. Some studies find no effect of it on well-being (Hessami, 2010, Wright, 1978, Miller
et al., 1991, Lennon, 1994, Klumb and Lampert, 2004, Ferree, 1984), whereas others find neg-
2
ative (Brereton et al., 2008) (“frustrated housewife hypothesis”) or positive effect (Stokes and
Peyton, 1986). It has been also found that employment history has positive effects on psy-
chological well-being among separated and divorced (Krause and Markides, 1985) as well as
widowed women (Aber, 1992).
Homemaking seems to relate to psychological well-being in a more complex way than
unemployment. Housewives presumably lose the psychological benefits of paid work, such
as prestige and social recognition, opportunities for social interaction and building networks.
Indeed, according to Bird and Ross (1993) and Lennon (1994) homemaking is associated with
less - both intrinsic and extrinsic - gratification and rewards than paid work. Also character of
work performed by housewives is specific: more routine, more physically exhausting and with
more work interruptions. Other research shows that doing housework2correlates with higher
depression rates and lower well-being (Sironi and Mencarini, 2010, Glass and Fujimoto, 1994,
Golding, 1990). On the other hand, homemaking implies more autonomy, less time pressure
and less responsibility for things outside one’s control than paid employment (Lennon, 1994,
Bird and Ross, 1993). Routine and responsibility of things outside of one’s control are as-
sociated with greater depressive symptoms among all women, independently from their work
status (Lennon, 1994). Concluding, literature suggests that both paid work and homemaking
brings some costs and some benefits in terms of psychological well-being.
2.1 Social desirability
Individual values are important in the context of choice between homemaking and employ-
ment. As shown by e.g. Stokes and Peyton (1986) and Tolciu and Zierahn (2010) housewives
have more traditional views on gender-roles than employed women. More importantly how-
ever, individual beliefs influence also the well-being consequences of performing certain roles.
Only career-oriented women (Pietromonaco et al., 1987) and women with non-traditional gen-
der attitudes (Krause and Markides, 1985) experience lower well-being as housewives. Ap-
parently, (Townsend and Gurin, 1981, Pietromonaco et al., 1987) it is rather the conformity
2Importantly, housework is not equivalent to homemaking. “Housework” indicates all works done in and
around home, which are also performed by employed people. “Homemaking” or “being a housewife” refers to
individuals, predominantly women, who do not work for pay, and for whom housework (frequently also childrea-
ring) is the main activity.
3
between actual and desired role that raises well-being, and not the role itself.3
Additionally, considerable literature informs on the link between gender attitudes domi-
nating in a country and women’s roles (such as employment and inter-household division of
work) (Crompton and Harris, 1997, 1998, Cha and Thebaud, 2009, Tolciu and Zierahn, 2010).
Moreover, a paper by Ferree (1984) shows that not only personal values, but also social de-
sirability may be an important factor influencing satisfaction with homemaking: housewives
more strongly desiring social approval reported higher satisfaction with life and housework.
Hypothesis I expect that in countries where more traditional gender-role attitudes dominate,
housewives’ well-being is higher. When acceptance for women’s work is low (i.e. gender-role
attitudes are traditional), employed women may need to face disapproval or open criticism
of their communities; the same applies to housewives in communities where gender-attitudes
are egalitarian. Importantly, the effect of “gender climate” should act independently from the
individual beliefs and attitudes of women.
2.2 Economic risk of specialization to housework
According to the economic model of the household, developed mainly by Becker (1981), com-
plete specialization of spouses (so that the man devotes all time and energy to paid work, and
the woman - to household tasks) is more efficient than equal sharing of duties, i.e. it creates
considerable economic gains for the household. The main reason for lack of complete special-
ization is a risk of divorce - event very undesirable for the household-specializing spouses, who
- by not being employed - sacrifice not only their personal income, but also future employment
and earning chances (human capital)4.
Moreover, vast literature shows that economic consequences of divorce for women (dis-
regarding their employment status) are worse than for men (Ongaro et al., 2009, Poortman,
2000, Andress et al., 2006)5. There is also some evidence concerning housewives: Krause
3Waldron and Herold (1986) found similar positive health consequences of congruence between desired and
performed roles.
4Additionally, human capital of value within one marriage cannot be - in case of divorce and re-marriage -
easily transfered to another household, whereas the human capital related to market work is to a larger extent
transferable in case of changing workplace (Bryant and Zick, 2006).
5Mainly because women work less, often have lower education and usually get the custody over children
4
and Markides (1985) and Aber (1992) have shown that divorce and widowhood are especially
undesirable for women have never done any paid work.
Divorce rates differ a lot across countries. Country legislation may make divorce easier
or more difficult (Gonzalez and Viitanen, 2009). On top of that, also economic burden of
divorce on women varies across countries (Andress et al., 2006, Uunk, 2004). In consequence,
housewives living in various countries face different risks associated with their role.
Hypothesis I expect that in countries with higher risk of divorce, homemaking should have
worse well-being consequences.
2.3 Selection to homemaking
Conditions in a country decide not only how happy women are as housewives, but - in the first
place - how many women will choose homemaking and who will become a housewife.
In countries supporting women’s employment, homemaking may attract a selected group
of strictly family-centered women, who derive especially high satisfaction from homemaking,
which would raise the average well-being of housewives compared to employed women. Sim-
ilar effect has been suggested by Boye (2009), who discovered that well-being of employed
women is on average higher in countries where policies support traditional gender roles. Ap-
parently, such policies may encourage women to stay at home; as a result those who decide to
work are strictly career-oriented (and happy with employment)6. If better labor market chances
are equivalent to higher congruence between the roles desired and performed by women, then
higher activity rates and lower percentage of full-time homemakers in a country should be
accompanied by increased well-being of housewives.
On the other hand however, selection may also work in the opposite direction. In particular,
in countries strongly supporting employment of women (by e.g. reconciliation of work and
family) housewives may be a negatively selected group of women who are unable to do any
paid work, e.g. because of psychological or health problems. We may expect that the negative
selection is stronger where more women are active in the labor market, therefore, labor market
(Poortman, 2000).
6Kalmijn (2010) shows a similar mechanism for cross-country differences in the effect of divorce.
5
activity may be also associated with on average lower levels of well-being of housewives.
Hypothesis I expect that - due to selection to homemaking and employment - the well-being
of housewives may vary according to the prevalence of homemaking and activity rates of
women. It is hard to predict the overall effect of both positive and negative selection to home-
making. It is however probable, that the first described type of selection takes place at lower
levels of employment, whereas for higher levels the second, negative selection dominates. This
would suggest curvilinear, inverted-U shaped relationship7.
3 Data and method
3.1 Data
The data come from the fourth edition of the European Values Study (EVS) (EVS Founda-
tion/Tilburg University, 2010) conducted in years 2008-2009 in 39 European countries and
regions. EVS is a cross-sectional survey program dating back to 1981 and a rich source of
information on beliefs, attitudes and opinions of European citizens on wide range of topics,
such as family, work, religion, politics, society and others8.
I use data for 36 countries9, which sums up to over 53 600 individuals, among which over
30 100 are women. The range of countries is particularly wide, including:
•post communist countries: both central-eastern (Czech, Hungary, Poland, Slovakia and
Slovenia) and southern countries (Albania, Bulgaria, Romania, Bosnia and Herzegovina,
Serbia and Montenegro),
•former Soviet Union, including the European (Russian Federation, Belarus, Ukraine,
Estonia, Latvia, Lithuania and Moldavia) and Caucasian countries (Armenia, Azerbaijan
and Georgia),
•Mediterranean (Malta, Cyprus, Greece, Spain and Portugal),
7Such shape of relationship corresponds also to the idea of freedom of choice. The more homogeneous is a
group (dominated be either housewives or employed women) the lesser freedom of individual choice. The higher
the entropy, the better are the chances to follow her/ its own preferences.
8see: www.europeanvaluesstudy.eu
9The database contains information for 39 countries and regions, however country/region level statistics for
Northern Cyprus, Northern Ireland and Kosovo are hardly available, therefore I excluded them from the analysis.
6
•western (Austria, Belgium, France, Germany, Ireland, Luxembourg, Netherlands and
Switzerland), and
•Scandinavian (Denmark and Finland).
The analysis is performed on a sample consisting of women. Sample size per country varies
between over 1 000 (Germany and Russian Federation) and over 550 (Cyprus and Finland).
For details on sample size and composition in particular countries see table 2 (page 16).
3.2 Measurement of the dependent variable
Subjective well-being The dependent variable in the analysis is self-assessed well-being.
EVS contains two indicators of it:
1. happiness (“Taking all things together, would you say you are: very happy / quite happy
/ not very happy / not at all happy”) and
2. life satisfaction (“All things considered, how satisfied are you with your life as a whole
these days? (1) dissatisfied - (10) satisfied”).
Both these variables were proved reliable indicators of psychological well-being. It has been
shown that national-level happiness correlates with hypertension and with prevalence of hearth
diseases in a country (Blanchflower and Oswald, 2008); that individuals who declare to be
happier have lower levels of salivary cortisol (stress hormone), reduced fibrinogen stress re-
sponses, and lower heart rate (Steptoe and Wardle, 2005); their areas of the brain associated
with processing of pleasure are more active (Urry et al., 2004), and their neurological re-
sponse to negative information is weaker (van Reekum et al., 2007). Self-ratings of subjective
well-being were also proved to correlate with judgments made by a third person (Schneider
and Schimmack, 2009), to be stable for individuals (Schimmack et al., 2010, Kahneman and
Krueger, 2006), and to associate with satisfaction with particular domains of life (Schimmack
et al., 2010).
It is also recognized that self-assessment of well-being fluctuates day-by-day under the
influence of random events (such as finding a coin or the weather) (Kahneman and Krueger,
2006), and (on a social level) depends on propensity toward positivity in responding (Diener
et al., 2000). This however does not undermine the reliability of these measures. Happiness
may have an effect on behavior even if it is triggered by random events. More importantly
7
however, despite above mentioned randomness, judgments of well-being systematically differ
along individual and social-level factors (such as prosperity, equality, social security, political
freedom), which supports the claim that they well reflect living conditions faced by individuals
(Ouweneel and Veenhoven, 1991).
Happiness and life satisfaction are regarded as separate but close measures of the same
phenomenon. Helliwell and Putnam (2004) consider life satisfaction a better tool for assess-
ing effects of stable characteristics of social context, but note that their central results do not
change when happiness measure is used. On the other hand, Peiró (2006) considers happiness
and life satisfaction two distinct spheres of well-being, of which the second one depends more
on economic factors.
Taking this into account and following the example of Kalmijn (2010), I use both indicators
to construct my dependent variable. Correlation between them is .5110 for sample including
all countries11.
After re-coding (so that higher values of both measures indicate higher well-being), I stan-
dardize both variables (because they have different metrics) and sum them. The resulting
variable is recoded into percentile scores, which makes it (and the later obtained regression
coefficients) easier to interpret (because 99 means the highest, and 1 - the lowest possible level
of well-being). Obtained variable ranges from 1 to 94, and has the mean of 46.7. Switching
from standardized values to percentile scores reduces also the skewness on the left side of the
distribution.
Acknowledging that the measure used in analysis is not a standard one, I further validate
obtained results with the use of happiness and life satisfaction variables.
3.3 Measurement of the individual-level independent variables
Employment status I code employment status as a set of dummy variables. I distinguish
five broad employment categories:
1. employment, including
10After reversing scale of happiness.
11But within particular countries it ranges between .08 for Azerbaijan and .71 in Finland, which suggests that
using each of three possible measures in countries such as Azerbaijan may give different results.
8
•employed full-time (30 hours or more per week),
•employed part-time (less than 30 hours per week),
•self-employed,
2. housewife not otherwise employed,
3. unemployed,
4. retired and
5. other not employed (containing students, disabled, military service and otherwise not-
classified situations).
Here, the most important categories are homemaking and employment (which serves as the
reference category12.)
Traditional gender-role attitudes EVS contains set of questions capturing opinions on roles
of men and women. These are:
1. “A working mother can establish just as warm and secure a relationship with her chil-
dren as a mother who does not work”,
2. “A pre-school child is likely to suffer if his or her mother works”,
3. “A job is alright but what most women really want is a home and children”,
4. “Being a housewife is just as fulfilling as working for pay”,
5. “Having a job is the best way for a woman to be an independent person”,
6. “Both the husband and wife should contribute to household income”,
7. “In general, fathers are as well suited to look after their children as mothers” and
8. “Men should take as much responsibility as women for the home and children”.
All questions use 4-points answering scale from “agree strongly” to “disagree strongly”.
I construct a measure of traditional gender-role attitudes in two ways. The first measure is
a general one, constructed as an average of eight above mentioned items (after reversing the
scale of some questions so that higher values indicate more traditional attitudes).
12Analysis taking as the reference category full-time employment and overall employment lead to identical
conclusions, therefore, for simplicity I present results referring to overall employment (i.e. distinguishing 5 and
not 7 employment states).
9
Second, recognizing that these items may refer to various dimensions of gender-related be-
liefs, I use exploratory factor analysis (with oblique factors) and distinguish three dimensions
of gender-role attitudes.
1. First dimension, “women do not need to work for pay” is related to statements 5 and 6
(both factor loadings over .5);
2. second dimension, “home is not men’s business” strongly corresponds to statements 7
and 8 (both factor loadings over .45);
3. third dimension, “for women and children is better if women stay at home” is captured
by statements 2, 3 and 4 (factor loadings over .5, and for statement 2 - .4).
Statement 1 forms a separate factor with statement 2; however, taking into account their low
factor loadings (.34 and -.26) I decide not to include this dimension into the analysis.
Again, original variables are recoded, so that higher values mean more traditional beliefs.
In the model I also include a list of individual-level control variables.
Age Age is measured in years and centered for easier interpretation of model’s intercept. I
also include age squared, to allow for the U-shaped relationship between age and well-being.
Family situation I control for family situation of a woman with a set of four dummy vari-
ables.
1. Being married is proved to positively contribute to well-being. In the model I control
only for the effect of being married vs. not being married, i.e. I don’t distinguish between
never married, widowed, divorced, separated or persons in registered partnership.
2. Having children is known to reduce well-being, presumably because of increased stress
and time-pressure. For this reason I control for presence of any children in the household
of the respondent13.
3. I include variable that indicates being a parent, also if the child does not live with the
respondent.
13The data does not allow checking if the child in the household is a child of the respondent, his/her grand-
child, relative or other person.
10
4. Since the combination of being married and having children may have an effect on well-
being, especially among women, I include a dummy variable taking the value of 1 for
women who are married, have children and live with some children in their households.
Education With two dummy variables I control for having secondary or tertiary education,
as opposed to having vocational, primary or lower education.
Social trust Social trust has been shown to correspond to higher levels of well-being. I
construct the measure of social trust as an average of two variables (“Do you think that most
people would try to take advantage of you if they got the chance, or would they try to be fair?”
and “Would you say that most of the time people try to be helpful or that they are mostly
looking out for themselves?”), of which both use 10-points answering scale. The correlation
between variables is .49, and higher values designate higher level of trust.
Health problems State of health is known to be a strong determinant of psychological well-
being. For this reason I control for declared level of health problems, using answers to the
question: “All in all, how would you describe your state of health these days? Would you say
it is (1) very good . . . (5) very poor?” Higher values indicate more health problems.
Because endogeneity of self-declared health may be a problem, I validate the robustness
of final results by excluding this variable from the model.
Household income Income is another strong correlate of well-being, therefore in the model
I control for total monthly household income14. In EVS income is coded as an ordinal variable
which assigns to each respondent only the information on the income range to which the
respondent belongs, so that exact value of income is not known. To construct the measure
of income, I use the country-specific variables15. First, I replace the category codes with the
categories’ middle values (separately for each country) and then make them comparable with
14The question in EVS questionnaire is as follows: “Here is a list of incomes and we would like to know in
what group your household is, counting all wages, salaries, pensions and other incomes that come in. Just give
the letter of the group your household falls into, after taxes and other deductions.”
15A common variable (unified for all countries, with all income ranges expressed in Euro) is available, however
it offers very poor representation of income differences in poorer countries. For this reason I use a set of country-
specific variables.
11
the use of PPP exchange rates, so that resulting values are expressed in international PPP
dollars16. In regression, I use income in the logarithmic form, which I subsequently center
around the country mean.
As usually, for considerable percentage of respondents (over 40% in Malta, Portugal and
Ireland, over 30% in Denmark) the information on household income is missing. In order to
include also these respondents in the analysis, I substitute missing values with country mean
and mark answers “don’t know” and “refusal” with two separate dummy variables. This way
of dealing with missingness lowers the reliability of the income variable. Alternative solutions
might be either to completely ignore the level of income, or to perform analysis on strongly
narrowed sample with non-missing income - both may lead to estimate bias. For this reason
I use the income variable with mean-imputed missing values. Considering its weaknesses, I
perform additional robustness checks of the final model.
3.4 Measurement of country-level variables
Traditional gender-role attitudes in a country Similarly to individual-level measures, I
construct four different country-level indicators of traditional gender-role attitudes in a country.
They are constructed as country-level averages (for both men and women) of individual scores,
and include:
1. overall measure of traditional gender climate, and measures referring to three specific
dimensions:
2. “women do not need to work for pay”,
3. “home is not men’s business”,
4. “for women and children is better if women stay at home”.
In the regression model I use centered scores.
Risk of divorce Statistical databases usually report numbers of divorces per year or crude
divorce rates (i.e. divorces per 1 000 of population). Since prevalence of marriages differs
across countries, I construct another measure of divorce risk, which is number of divorces in
16Because at the time of writing the paper PPP exchange rates were available only for 2005 (and data come
from 2008/9), before PPP conversion I also deflate the incomes to the 2005 values.
12
a year in relation to the population of married women (aged 18-64) in that year. This way,
the measure reflects the average probability of divorce faced by women in that country. The
measure takes highest values in Belarus, Latvia and Russian Federation (20-25 divorces yearly
per 1 000 married women) and lowest in Georgia and Bosnia and Herzegovina (with values of
2.2-2.3). In the model I include centered values.
Selection I use two measures of selection into homemaking:
1. activity rate of women, and
2. prevalence of homemaking status in the population of women aged 18-64.
Both variables are computed from EVS data as percentages of - respectively - (1) economically
active (either working or unemployed) women and (2) housewives among women aged 18-65.
Again, I use centered variables in the model.
In the analysis I also control for some country-level indicators that possibly have an effect on
well-being.
Unemployment rate Literature shows that unemployment affects well-being also for em-
ployed individuals (Hessami, 2010, Di Tella et al., 2001), which is explained by tougher job
market17, but also social problems, such as social exclusion or crime18 that rise in periods of
high unemployment (Edmark, 2005).
GDP Although Easterlin (1974) found little evidence on the link between GDP and happi-
ness (neither cross-nationally nor in time-trends) which initiated discussion on the “Easterlin
paradox”, part of the literature shows a positive relationship between the two (Stevenson and
Wolfers, 2008, Hessami, 2010). Therefore I control the level of national income by using
country’s GDP (data for year 2007, corrected by PPP).
17Luechinger et al. (2008) reports that the well-being of public sector employees responds less to unemploy-
ment rate changes than the one of private employees, which suggests the importance of economic insecurity.
18At the same time however, research finds no effect of crime on well-being, presumably because of unrelia-
bility of criminal records (Alesina et al., 2004).
13
Gender inequality Inequality in employment between men and women was found to explain
almost 1/3 of the cross-country women’s happiness, after accounting also for gender gap in
education and political power - over 40% (Sironi and Mencarini, 2010). Therefore I control
for gender inequality by including a variable being the reverse of gender pay gap (1−pay gap)
in a country. Higher values correspond to higher gender inequalities of earnings between men
and women.
Income inequality Income inequality in a country is associated with lower well-being, es-
pecially among the poor (Alesina et al., 2004). I control for income inequality by including a
measure of Gini coefficient for each country.
Region I include a dummy variable marking eastern European countries (both post commu-
nist and former Soviet union), which are known to experience systematically lower levels of
subjective well-being (Kalmijn, 2010).
All variables used in the analysis are summarized in table 1, table 2 presents summary of
the country-level factors.
3.5 Statistical method
To test the hypotheses I use linear multilevel regression, which allows modeling the depen-
dent variable as a function of predictors measured on various levels, e.g. using individual and
country characteristics. I use multilevel, rather than regular regression, because hierarchical
data (such as multi-country EVS with individuals nested within countries) do not satisfy the
basic assumption of independence of observations. This may lead to biased estimates of pa-
rameters and their standard errors, which in turn can result in wrongly rejecting or supporting
theoretically important conclusions (Luke, 2004, Bryk and Raudenbush, 1992).
I test a two-level model, with individuals (level 1) nested within countries (level 2). The
average subjective well-being is allowed to vary randomly across countries (random intercept),
and effects of both homemaking and unemployment are allowed to vary across countries (ran-
dom slopes). Formally, the model is presented in equations 1-5.
14
Table 1: Means and standard deviations of variables used in the analysis
mean sd min max count
subjective well-being, percentiles 46.11 28.32 1 94 29547
fulltime 0.35 0.48 0 1 29959
housewifea0.14 0.34 0 1 29959
housewife x NOT married 0.03 0.18 0 1 29848
unemployeda0.09 0.28 0 1 29959
retireda0.24 0.43 0 1 29959
other employment statusa0.08 0.27 0 1 29959
age (centered) 0.50 17.89 −29.0 61.0 30019
age2(centered) 48.94 1796.93 −2205.8 9134.2 30019
married 0.51 0.50 0 1 29986
married, with own children in the hh 0.34 0.47 0 1 29356
ever had children 0.75 0.43 0 1 29921
children in the hh 0.48 0.50 0 1 29623
secondary educationb0.46 0.50 0 1 29923
tertiary educationb0.24 0.43 0 1 29923
social trust 5.01 2.19 1 10 29925
health problems 2.42 0.97 1 5 30031
hh income (ln, PPP) −0.04 0.72 −3.81 3.40 30114
income not known 0.11 0.32 0 1 30121
income refused 0.08 0.26 0 1 30121
traditional (overall) 2.10 0.42 1 4 30008
traditional (1) 1.79 0.61 1 4 29801
traditional (2) 1.80 0.61 1 4 29818
traditional (3) 2.62 0.66 1 4 29891
Observations 30121
EASTc0.58 0.50 0 1 36
COUNTRY GDP (K)c0.16 15.29 −20.0 56.9 36
INCOME INEQUALITY (GINI) c0.04 4.09 −7.15 8.95 36
SEX PAY GAP c−0.00 0.09 −0.15 0.22 36
UNEMPLOYMENT RATE c−0.07 5.62 −8.56 20.8 36
DIVORCE RATEc−0.12 6.12 −9.85 13.5 36
TRAD. COUNTRY (overall)c0.00 0.14 −0.40 0.33 36
TRAD. COUNTRY (1)c0.00 0.17 −0.24 0.49 36
TRAD. COUNTRY (2)c−0.00 0.16 −0.34 0.36 36
TRAD. COUNTRY (3)c0.00 0.23 −0.76 0.45 36
WOMEN ACT. RATEc−0.00 0.09 −0.29 0.13 36
WOMEN ACT. RATE2c0.01 0.01 0.000061 0.082 36
% OF HOUSEWIVESc0.00 0.11 −0.14 0.40 36
% OF HOUSEWIVES2c0.01 0.03 0.00000013 0.16 36
Observations 36
Source: European Values Study, 2008;
Source for GDP, Gini, gender pay gap, unemployment rate: United Nations Statistics Division (unstats.un.org)
Note: household income is mean-centered within each country, country-level variables (except “East”) are centered to the overall mean
15
Table 2: Summary of country-level statistics
(1) (2) (3) (4) (5) (6) (7) (8) (9) N
Albania 36.79 −4.14 2.13 1.83 1.89 2.60 4.76 0.13 0.68 775
Azerbaijan 33.48 −7.64 2.46 2.28 2.18 3.11 6.51 0.04 0.77 750
Austria 55.23 4.60 2.15 1.81 1.86 2.65 15.73 0.12 0.65 854
Armenia 36.34 4.84 2.24 1.96 1.91 2.84 4.35 0.34 0.53 856
Belgium 59.93 5.68 2.00 1.79 1.68 2.50 16.51 0.12 0.68 781
Bosnia Herzegovina 48.62 −3.13 2.15 1.89 1.87 2.61 2.29 0.17 0.65 828
Bulgaria 33.25 9.34 2.02 1.63 1.86 2.55 10.37 0.07 0.76 868
Belarus 36.70 10.56 2.05 1.72 1.67 2.67 20.98 0.05 0.76 890
Cyprus 48.55 −2.58 2.23 1.71 1.92 2.83 8.62 0.28 0.63 556
Czech Republic 46.95 11.18 2.18 1.84 2.01 2.60 16.78 0.07 0.64 995
Denmark 68.21 −13.18 1.73 1.80 1.55 1.89 13.81 0.01 0.80 760
Estonia 39.56 4.95 2.22 1.89 1.91 2.75 19.37 0.08 0.77 982
Finland 50.32 −35.59 1.96 2.16 1.59 2.33 13.93 0.00 0.78 557
France 52.76 −4.20 1.89 1.60 1.50 2.46 14.87 0.12 0.71 818
Georgia 31.91 −2.70 2.21 1.64 1.95 2.98 2.22 0.22 0.68 943
Germany 42.44 −4.13 1.99 1.75 1.82 2.30 13.17 0.09 0.75 1085
Greece 45.42 −6.32 2.25 1.72 2.00 2.89 5.50 0.30 0.53 850
Hungary 41.40 2.62 2.09 1.67 1.70 2.72 13.93 0.03 0.64 790
Ireland 64.44 −5.78 2.19 2.12 1.81 2.56 5.47 0.32 0.62 605
Latvia 38.69 4.69 2.10 1.74 1.80 2.63 21.34 0.12 0.74 949
Lithuania 35.28 7.87 2.33 2.04 1.91 2.91 19.17 0.09 0.73 817
Luxembourg 60.81 1.62 1.95 1.73 1.49 2.55 16.89 0.17 0.58 815
Malta 59.27 0.31 2.38 2.02 1.93 2.99 7.63 0.54 0.38 936
Moldavia 37.27 7.45 2.21 1.82 1.93 2.82 15.88 0.15 0.62 843
Montenegro 51.84 −0.85 2.10 1.76 1.71 2.71 4.17 0.07 0.76 844
Netherlands 65.72 −2.55 2.22 2.33 1.99 2.44 9.62 0.18 0.72 853
Poland 49.43 2.30 2.24 1.94 1.84 2.72 8.26 0.12 0.63 842
Portugal 37.69 −3.42 2.19 1.81 1.97 2.63 11.73 0.11 0.74 925
Romania 39.63 −9.09 2.21 1.77 2.11 2.75 7.40 0.17 0.49 838
Russian Federation 37.48 10.77 2.20 1.91 1.74 2.90 25.55 0.11 0.73 1002
Serbia 45.88 −1.68 2.08 1.74 1.83 2.55 4.20 0.10 0.70 810
Slovak Republic 43.94 3.10 2.06 1.74 1.88 2.51 9.78 0.04 0.69 904
Slovenia 51.89 −6.13 2.10 1.90 1.78 2.55 7.03 0.04 0.66 738
Spain 51.60 −4.31 2.01 1.75 1.70 2.44 18.64 0.21 0.70 842
Switzerland 63.31 4.92 2.16 1.90 1.78 2.66 15.93 0.17 0.72 685
Ukraine 34.00 6.55 2.11 1.70 1.51 2.97 18.02 0.14 0.63 935
Total 46.56 −0.39 2.13 1.85 1.82 2.65 11.96 0.14 0.67 836.69
Note: for Malta, where there is no legal divorce, data on divorce refer to separations and anullments
Source: 1-6, 8: European Values Study 2008; 7, 9: United Nations Statistics Division (unstats.un.org)
Variables:
(1) average subjective well-being in a country (scale 1-94)
(2) average subjective well-being of housewives vs. full-time employed women
(3) average traditional gender role attitudes (overall) in a country (scale 1-4)
(4) average traditional gender role attitudes (“women do not need to work for pay”, 1-4)
(5) average traditional gender role attitudes (“home is not men’s business”, 1-4)
(6) average traditional gender role attitudes (“for women and children is better if women stay at home”, 1-4)
(7) number of divorces per 1000 married women (per year)
(8) % of adult women in population being housewives
(9) activity rate of women
16
Individual-level equation:
wbij =α0j+α1j·hwij +α2j·unij
+α3·x1.ij +. . . +α2+k·xk .ij +ij
(1)
Country-level equations:
α0j=β00 +β01 ·z1.i +. . . +β0m·zm.i +µi(2)
α1j=β10 +νi(basic model)(3)
α1j=β10 +β11 ·T radi+β12 ·Divi+β13 ·Activityi+νi(f ull model)(4)
α2j=β20 +ηi(5)
Equation 1 shows the individual-level model for individual iin country j. The dependent
variable, wbij , is the self-assessed well-being of the individual ij. The intercept α0jcontains
subscript j, which indicates that different intercepts may be estimated in various countries. The
cross-country variation of the intercept is described by equation 2, which models the country-
specific intercept as a function of mcountry level variables (z1.i . . . zm.i). The error term µi,
which corresponds to the country-level random intercept in the multilevel model.
Equation 1 contains also individual-level variables hwij and unij - dummies indicating
if the individual is - respectively - either a full-time housewife or unemployed. The coeffi-
cient α1jis fundamental in the analysis, because it describes the effect of homemaking on
well-being. Subscript jin coefficients α1jand α2jindicates that the effects of homemak-
ing / unemployment may differ across countries. Equations 3-5 describe, how these effects
are estimated. In a simple case of a basic model (equations 3 and 5) no additional country-
level variables are introduced, and the effects of homemaking (equation 3) and unemployment
(equation 5) are assumed to vary randomly across countries. This random variation is captured
by elements νiand ηi, which correspond to random slopes in the multilevel model. In the more
complex case of full model, the effect of homemaking on well-being is modeled as a function
of country-level variables (equation 4). The variables (T radi,Diviand Activityi) correspond
to the three hypotheses formulated above. Coefficients β11 . . . β13 inform if and how does the
well-being effect of homemaking depend on country characteristics.
Finally, equation 1 includes also a set of individual-level control variables x1.ij . . . xk.ij . I
17
assume that the effects of these variables on well-being are the same in all countries, therefore
coefficients α3. . . α2+kdo not contain the subscript j. Last element of equation 1, ij , is
the individual-level error - the part of variation unexplained by the model, which cannot be
attributed to cross-country variation.
To test the hypotheses I focus on interaction of homemaking with the country-level indi-
cators. This method has an advantage over inspecting determinants of well-being on a sample
limited to housewives, because average well-being of housewives and employed women in
a country are strongly correlated (country-level correlation coefficient r=.66, for 36 coun-
tries). Using this model would therefore fail to focus on well-being premium of homemaking,
but would instead largely inform on determinants of women’s well-being in general.
4 Results
4.1 Descriptive results
Depending on circumstances, homemaking may take different forms. It may be a comfortable
option for women who don’t need to work for pay, but also the only solution for women
poorly educated, without much labor market prospects. Therefore, before moving to testing
the hypotheses, I check how homogeneous group are housewives and how much they differ
from employed women.
Table 3 informs what percentage of homemakers and employed women have tertiary ed-
ucation, and what percentage lives in households enjoying the top 25% of per capita income
in the country; t-test of mean difference indicates the statistical significance of the difference.
Not surprisingly, in the majority of countries employed women are better educated and more
wealthy than housewives19 . Interestingly, all countries where there is no difference are post
communist countries20. In some of them the share of housewives with tertiary education is
substantial (over 25% in Azerbaijan, Belarus, Denmark, Georgia, Russia and Ukraine, almost
19Exceptions are Azerbaijan, Estonia and Lithuania, where high per capita incomes are more frequent among
housewives. Only in Estonia the t-test indicates significant difference.
20This is consistent with results showing that in eastern Europe temporary homemaking is a childbearing-
related experience of vast majority of women, which stands in contrast to the West, where this path is chosen by
smaller percentage of women who stay in this role for longer (Mikucka, 2010).
18
Table 3: Characteristics associated with being housewife or employed
Higher education Income per capita over 75th percentile
Housewives Employed T-test Housewives Employed T-test
Albania 0.02 0.30 *** 0.19 0.48 ***
Azerbaijan 0.28 0.56 ** 0.34 0.28
Austria 0.03 0.13 ** 0.26 0.50 ***
Armenia 0.22 0.42 *** 0.30 0.32
Belgium 0.14 0.50 *** 0.12 0.40 ***
Bosnia Herzegovina 0.00 0.27 *** 0.12 0.57 ***
Bulgaria 0.23 0.32 0.44 0.51
Belarus 0.31 0.40 0.31 0.51 *
Cyprus 0.02 0.23 *** 0.18 0.43 ***
Czech Republic 0.13 0.16 0.30 0.37
Denmark 0.33 0.51 0.43 0.52
Estonia 0.19 0.33 *0.51 0.41
Finland 0.00 0.57 0.00 0.70
France 0.10 0.46 *** 0.13 0.40 ***
Georgia 0.30 0.64 *** 0.32 0.49 ***
Germany 0.09 0.22 ** 0.38 0.48
Greece 0.04 0.34 *** 0.26 0.42 ***
Hungary 0.11 0.29 0.21 0.40
Ireland 0.05 0.32 *** 0.14 0.36 ***
Latvia 0.16 0.40 *** 0.31 0.32
Lithuania 0.47 0.56 0.42 0.35
Luxembourg 0.11 0.34 *** 0.41 0.58 ***
Malta 0.02 0.30 *** 0.67 0.86 ***
Moldavia 0.16 0.30 ** 0.36 0.41
Montenegro 0.08 0.33 *** 0.32 0.44
Netherlands 0.11 0.45 *** 0.14 0.32 ***
Poland 0.18 0.31 *0.44 0.60 **
Portugal 0.03 0.18 *** 0.69 0.79 *
Romania 0.01 0.20 *** 0.26 0.56 ***
Russian Federation 0.28 0.40 *0.34 0.42
Serbia 0.00 0.35 *** 0.09 0.44 ***
Slovak Republic 0.10 0.17 0.21 0.42 *
Slovenia 0.00 0.33 *** 0.06 0.39 ***
Spain 0.06 0.25 *** 0.23 0.32 *
Switzerland 0.11 0.26 ** 0.41 0.52
Ukraine 0.37 0.62 *** 0.22 0.48 ***
Source: European Values Study, 2008
∗p < 0.05,∗∗ p < 0.01,∗∗∗ p < 0.001
19
50% Lithuania), similarly for income (over 25% of housewives in Azerbaijan, Estonia and
Russia declare having per capita income belonging to the top 25%). Most of these are former
countries of the Soviet Union, where housewives constitute a small percentage of working-age
women (see table 2 page 16). Apparently, in the East housewives constitute a less homogenous
(less selected) group, and the differences between them and the working women are smaller.
Table 4 shows the correlations between macro-level variables. Some of the variables are
considerably correlated: especially gender climate variables, as well as the prevalence of being
a housewife and women’s activity rate. For this reason, I employ the strategy of including them
in the regression models separately before moving to the complete model.
Table 4: Correlations between country-level variables
(1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11) (12)
(1) 1.00
(2) 0.48∗1.00
(3) 0.78∗∗∗ 0.34∗1.00
(4) 0.84∗∗∗ 0.10 0.46∗1.00
(5) −0.39∗0.03 −0.25 −0.41∗1.00
(6) 0.45∗0.13 0.26 0.40∗−0.77∗∗∗ 1.00
(7) 0.29+−0.12 0.25 0.42∗0.07 −0.33+1.00
(8) −0.40∗0.10 −0.41∗−0.48∗−0.03 0.07 −0.76∗∗∗ 1.00
(9) 0.43∗0.12 0.30+0.44∗0.07 0.14 0.16 −0.21 1.00
(10) 0.38∗0.00 0.31+0.41∗−0.31+0.46∗−0.13 −0.00 0.16 1.00
(11) 0.05 −0.14 −0.00 0.10 0.15 −0.08 0.33+−0.32+0.26 0.15 1.00
(12) −0.26 −0.13 −0.43∗−0.11 0.33+−0.33+−0.08 0.24 −0.01 −0.38∗−0.32+1.00
Source: European Values Study, 2008
+p < 0.10,∗p < 0.05,∗∗∗ p < 0.001
Variables:
(1) traditional country (overall)
(2) traditional country, dimension 1 (“women do not need to work for pay”)
(3) traditional country, dimension 2 (“home is not men’s business”)
(4) traditional country, dimension 3 (“for women and children is better if women stay at home”)
(5) women’s activity rate
(6) prevalence of housewives
(7) eastern country (post communist or former Soviet Union)
(8) country GDP per capita (k)
(9) household income inequality in a country (Gini coefficient)
(10) gender pay gap
(11) country unemployment rate
(12) country divorce rate
How do these variables co-vary with average well-being premium for homemakers? I
inspect the difference between the average well-being of housewives and of employed women
in a country (see the second column of table 2). Overall (and without controlling for other
factors, such as education, income etc.), housewives report slightly lower well-being than
employed women (-.39 on a scale from 1 to 94). Differences between countries are large:
from Finland and Denmark, where housewives are on average 36 and 13 percentage points
less happy, to Belarus and Russian Federation, where they report about 10 percentage points
20
higher well-being.
Scatter plots presented on figures 1 and 2 show how this cross-country variation relates
to macro-level factors. All scatter plots are accompanied by two fitted lines: a solid one for
all countries, and a dashed one - excluding Denmark and Finland (which are outliers and
moreover countries with very few housewives). Traditional “gender climate” is related to
higher gains from homemaking only when it comes to belief that “for women and children
is better if women stay at home” (last graph in figure 1). Contrary to expectations, for the
three remaining dimensions of gender attitudes the relationship is negative (after excluding
Denmark and Finland).
Concerning the selection effect (figure 2), lower prevalence of housewives and higher
women’s activity rates are related to higher well-being gains from homemaking (dashed lines
on graphs 1 and 2 in figure 2), although the trend is not very clear, and largely influenced by
outlier countries (Malta, Armenia, Greece and Romania). First two graphs in figure 2 present
also the quadratic fit line (dashed), however in none of the two cases there is a sign of hypoth-
esized, inverse-U shaped relationship.
The relationship between divorce rate and well-being premium of being a housewife (graph
3 in figure 2) is clear and stable even after including Denmark and Finland, but - contrary to
expectations - positive: in countries with higher divorce rates women gain more from home-
making in terms of well-being. Interestingly, the divorce rate is correlated with attitudes sup-
porting sharing of home duties, lower gender pay gap and higher activity rate of women (see
table 4), which suggests that rising divorces accompanies higher gender equality. Countries
with higher divorce rates and highest well-being premium for homemaking are almost exclu-
sively former Soviet Republics (Russia, Belarus, Latvia, Lithuania, Estonia and Ukraine, see
last graph on figure 2). After excluding them the relationship holds (results not shown), how-
ever the validation of final results will require additional test excluding this group of countries.
4.2 Multivariate analysis
The results so far were based on country-level bivariate correlations. In order to test the hy-
potheses I turn to a set of multilevel regression models which include - apart from individual-
level predictors - also macro-level indicators and their interactions. The results are presented
in tables 5-9.
21
FR
FI
EE
GR
LU SK
PT
SI
BE UA MD
BG
BA
DK
BY
HU
AL AZ
GE
RS
LVAT
RO
ME
CH AM
CZ
ES
DE NL MT
LT
CY
PL
RU
IE
−40 −30 −20 −10 0 10
Well−being effect of being a housewife (women)
1.8 2 2.2 2.4 2.6
TRADITIONAL GENDER ROLES (overall)
FR
FI
EE
GR
LU
SK
PT SI
BE
UA MD
BG
BA
DK
BY
HU
AL AZ
GE RS
LV AT
RO
ME
CHAM
CZ
ES
DE NL
MT
LT
CY
PL
RU
IE
−40 −30 −20 −10 0 10
Well−being effect of being a housewife (women)
1.6 1.8 2 2.2 2.4
TRADITIONAL GENDER ROLES (women’s work)
FR
FI
EE
GR
LU SK
PT
SI
BE
UA MD
BG
BA
DK
BY
HU
AL AZ
GE
RS
LVAT
RO
ME
CH AM
CZ
ES DE NL
MT
LT
CY
PL
RU
IE
−40 −30 −20 −10 0 10
Well−being effect of being a housewife (women)
1.4 1.6 1.8 2 2.2
TRADITIONAL GENDER ROLES (man at home)
FR
FI
EE
GR
LU
SK
PT
SI
BE UA
MD
BG
BA
DK
BY
HU
AL AZ
GE
RS
LV
AT
RO
ME
CH AM
CZ
ES
DE NL MT
LT
CY
PL
RU
IE
−40 −30 −20 −10 0 10
Well−being effect of being a housewife (women)
2 2.5 3
TRADITIONAL GENDER ROLES (women need home)
Figure 1: The association between country-level factors and the well-being consequences of
being a housewife (vs. employed), women.
Country codes:
AL Albania, AM Armenia, AT Austria, AZ Azerbaijan, BA Bosnia and Herzegovina, BE Belgium, BG Bulgaria, BY Belarus, CH
Switzerland, CY Cyprus, CZ Czech, DE Germany, DK Denmark, EE Estonia, ES Spain, FI Finland, FR France, GE Georgia, GR Greece,
HU Hungary, IE Ireland, LT Lithuania, LU Luxembourg, LV Latvia, MD Moldavia, ME Montenegro, MT Malta, NL Netherlands, PL
Poland, PT Portugal, RO Romania, RS Serbia, RU Russia, SI Slovenia, SK Slovakia, UA Ukraine
22
FR
FI
EE
GR
LU
SK
PT
SI
BE
UA
MD
BG
BA
DK
BY
HU
AL
AZ
GE
RS
LV
AT
RO
ME
CH AM
CZ
ES
DE NL MT
LT
CY
PL
RU
IE
−40 −30 −20 −10 0 10
Well−being effect of being a housewife (women)
0 .2 .4 .6
% OF HOUSEWIVES
FR
FI
EE
GR
LU SK
PT
SI
BE
UA
MD BG
BA
DK
BY
HU
AL AZ
GE
RS
LV
AT
RO
ME
CH
AM
CZ
ES DE
NL
MT
LT
CY
PL
RU
IE
−40 −30 −20 −10 0 10
Well−being effect of being a housewife (women)
.4 .5 .6 .7 .8
WOMEN ACTIVITY RATE
FR
FI
EE
GR
LU
SK
PT
SI
BE
UA
MD
BG
BA
DK
BY
HU
ALAZ
GERS
LV
AT
RO
ME
CH
AM
CZ
ES
DE
NL
MT
LT
CY
PL
RU
IE
−40 −30 −20 −10 0 10
Well−being effect of being a housewife (women)
0 5 10 15 20 25
DIVORCE RISK
Figure 2: The association between country-level factors and the well-being consequences of
being a housewife (vs. employed), women. (continued)
Country codes:
AL Albania, AM Armenia, AT Austria, AZ Azerbaijan, BA Bosnia and Herzegovina, BE Belgium, BG Bulgaria, BY Belarus, CH
Switzerland, CY Cyprus, CZ Czech, DE Germany, DK Denmark, EE Estonia, ES Spain, FI Finland, FR France, GE Georgia, GR Greece,
HU Hungary, IE Ireland, LT Lithuania, LU Luxembourg, LV Latvia, MD Moldavia, ME Montenegro, MT Malta, NL Netherlands, PL
Poland, PT Portugal, RO Romania, RS Serbia, RU Russia, SI Slovenia, SK Slovakia, UA Ukraine
23
I start with a null model, shown in the first column of table 5. It is basically an empty
model, containing only intercept and random effects at the country level. Fit of the model is
significantly better than of linear one (χ2(3) = 4 016.9, P rob(χ2) = 0.000). Random effects at
country level are significant (justifying the use of multilevel technique) and account for 23%21
of variation unexplained by the model.
Table 5: Regression of well-being of women on set of explanatory variables. Basic model
including individual and country-level predictors.
(1) (2)
Null model Basic model
subjective well-being, percentiles
housewifea2.646 (4.03)∗∗∗
housewife x NOT married −1.655 (−1.73)+
unemployeda−4.147 (−4.50)∗∗∗
retireda2.569 (4.89)∗∗∗
other employment statusa1.901 (3.11)∗
age (centered) −0.652 (−12.35)∗∗∗
age2(centered) 0.006 (10.68)∗∗∗
married 6.909 (14.75)∗∗∗
married, with own children in the hh 1.058 (1.70)+
ever had children 1.486 (2.77)∗
children in the hh −2.663 (−4.96)∗∗∗
secondary educationb1.000 (2.62)∗
tertiary educationb1.543 (3.37)∗∗∗
social trust 2.039 (29.39)∗∗∗
health problems −10.394 (−59.19)∗∗∗
hh income (ln, PPP) 2.500 (11.37)∗∗∗
income not known 0.024 (0.05)
income refused 1.239 (2.20)∗
EASTc−4.222 (−1.72)+
COUNTRY GDP (K)c0.243 (2.98)∗
INCOME INEQUALITY (GINI) c−0.461 (−2.12)∗
SEX PAY GAP c−22.065 (−2.16)∗
UNEMPLOYMENT RATE c0.287 (1.78)+
DIVORCE RATEc−0.212 (−1.45)
TRAD. COUNTRY (overall)c8.821 (1.20)
Constant 46.745 (26.59)∗∗∗ 58.474 (33.03)∗∗∗
Random effects:
country RS: housewife 5.711 †2.066 †
country RS: unemployed 7.624 †3.933 †
country RI 10.491 †4.341 †
individual-level error 26.294 †23.220 †
N27857 27857
AIC 261489.7 254492.5
Source: European Values Study, 2008
+p < 0.10,∗p < 0.05,∗∗∗ p < 0.001;tstatistics in parentheses
For random effects standard deviations are presented; RS - random slope, RI - random intercept
†standard deviation of the random effect is at least double of its standard error
areference category is employment (full-time, part-time and self-employed)
breference category is primary and vocational education
ccountry-level variables
Basic model (second column of table 5) contains only individual-level predictors and
country-level control variables; I use it to discuss the individual-level effects. Value of Akaike
21(5.72+ 7.62+ 10.52)/(5.72+ 7.62+ 10.52+ 26.32)≈23%
24
Information Criterion (AIC) of basic model will serve as a reference value for assessing the fit
of subsequent models22. In contrast to data shown in table 2, homemaking is ceteris paribus re-
lated to higher well-being than employment. Presumably, lower average well-being of house-
wives results not from employment status itself, but from other factors, such as their on average
lower incomes and education. Being retired or in other employment status is related to higher
levels of well-being; well-being of unemployed is on average lower23.
Results obtained for other coefficients are consistent with the literature. Age has a negative
effect, but - due to the significant quadratic component - the relationship is U-shaped. Marriage
(compared to being single, divorced or widowed), being married with children in the household
and being a parent increase well-being, whereas current presence of children in the household -
decreases it. Additionally, housewives get an extra well-being bonus for being married. Higher
and secondary education raise well-being; the same goes for income and social trust; health
problems lower well-being.
Among country-level factors, well-being of women is on average lower in post communist
countries, in countries with higher income inequality and with larger gender pay gap. Higher
country GDP raises happiness; surprisingly the same is true for unemployment rate24.
Individual and country-level predictors explain much of the variance unexplained by the
null model. Decrease of AIC is substantial and statistically significant, which indicates im-
provement of the model fit25. Random variation associated with country level is much lower,
still the random effects remain statistically significant.
Tables 6 and 7 show results of models including the theoretically interesting variables:
models 3 and 4 account for values, model 5 - for divorce risk, 6 and 7 - for selection and 8 is
the full model.
22Despite common usage of Bayesian Information Criterion (BIC), for mixed models AIC is the preferred
measure of (badness of) model fit, because dependence of BIC on the sample size makes its usage for multilevel
models problematic.
23Including in the model also part-time and self-employment does not change the results. Coefficients of both
variables (with full-time employment as the reference category) are insignificant.
24This result is at odds with the existing literature (which however rarely focuses exclusively on women) and
robust across models presented in current paper. This result calls for further exploration.
25χ2(25) = 6997.2, P rob(χ2)=0.000
25
Social desirability Models 3 and 4 include variables referring to gender role-attitudes (on
individual and country level) and interactions of housewife status with gender-role attitudes
(again, on both levels); model 3 accounts for overall measure of gender attitudes, model 4 -
for three separate dimensions. Comparison of AIC values and significance of random slopes
of these models shows that gender attitudes in the form of a single dimension (model 3) have
less predictive and explanatory power than three dimensions (model 4).
In model 4, individual gender-role attitudes have an effect on well-being of housewives.
The second dimension of individual-level gender attitudes (“home is not men’s business”) is
associated with lower well-being of both housewives and women overall, possibly because it
reflects the experience of sharing household tasks. The two other dimensions (“women do not
need to work for pay” and “for women and children is better if women stay at home”) increase
the well-being of housewives. In other words, housewives believing that women don’t need to
be employed and their presence at home is good for them and their families, get more well-
being premium from homemaking.
However, the estimations concerning country-level variables do not confirm the social de-
sirability hypothesis. Although individual-level attitudes are related to well-being of house-
wives, still living in a traditional country (interaction terms: “woman x TRAD. COUNTRY”)
has no effect: the well-being of housewives does not depend on how traditional or how gender-
egalitarian are the attitudes in the country.
Economic risks of specialization to homemaking The second hypothesis concerned the
effect of country divorce risk (third column in table 6). Previous inspection of country-level
correlations suggested that higher divorce risk is associated with higher well-being premium
for housewives. The results of model 5 confirm this, contradicting the formulated hypoth-
esis. Importantly, the result concerns country-specific risk of divorce26 and not individual
experience of marital breakdown: married housewives are happier than non-married. Despite
significant coefficient, however AIC value of model 5 higher than of basic model, which in-
dicates worse fit: including the country divorce rate does not improve our understanding of
position of housewives.
26Divorce rates in a country are associated with legal regulations which may make divorce more or less difficult
(such as length of waiting periods or mandatory counseling necessary before granting a divorce).
26
Table 6: Regression of well-being of women on set of explanatory variables. Models account-
ing for gender-role attitudes and divorce risk.
(3) (4) (5)
Values 1 Values 2 Divorce risk
subjective well-being, percentiles
housewifea2.970 (4.34)∗∗∗ 2.445 (3.69)∗∗∗ 2.780 (4.33)∗∗∗
housewife x NOT married −1.683 (−1.75)+−1.806 (−1.88)+−1.768 (−1.84)+
housewife x traditional (overall) 0.749 (0.71)
housewife x TRAD. COUNTRY (overall)c−0.084 (−0.02)
housewife x traditional (1) 1.270 (1.89)+
housewife x traditional (2) −1.370 (−1.99)∗
housewife x traditional (3) 1.660 (2.42)∗
housewife x TRAD. COUNTRY (1)c0.367 (0.10)
housewife x TRAD. COUNTRY (2)c−5.727 (−1.38)
housewife x TRAD. COUNTRY (3)c4.741 (1.53)
housewife x DIVORCE RATEc0.190 (2.12)∗
unemployeda−3.974 (−4.33)∗∗∗ −4.062 (−4.40)∗∗∗ −4.188 (−4.55)∗∗∗
retireda2.660 (5.06)∗∗∗ 2.631 (5.01)∗∗∗ 2.522 (4.80)∗∗∗
other employment statusa1.902 (3.11)∗1.866 (3.06)∗1.904 (3.12)∗
age (centered) −0.662 (−12.53)∗∗∗ −0.652 (−12.37)∗∗∗ −0.648 (−12.27)∗∗∗
age2(centered) 0.006 (10.91)∗∗∗ 0.006 (10.72)∗∗∗ 0.006 (10.62)∗∗∗
married 6.942 (14.83)∗∗∗ 6.883 (14.72)∗∗∗ 6.894 (14.72)∗∗∗
married, with own children in the hh 1.102 (1.77)+1.011 (1.63) 1.063 (1.71)+
ever had children 1.486 (2.77)∗1.456 (2.72)∗1.486 (2.77)∗
children in the hh −2.653 (−4.94)∗∗∗ −2.632 (−4.91)∗∗∗ −2.677 (−4.98)∗∗∗
secondary educationb0.840 (2.19)∗0.963 (2.51)∗0.976 (2.55)∗
tertiary educationb1.255 (2.72)∗1.463 (3.18)∗1.513 (3.30)∗∗∗
social trust 2.028 (29.23)∗∗∗ 2.000 (28.82)∗∗∗ 2.040 (29.41)∗∗∗
health problems −10.357 (−58.97)∗∗∗ −10.370 (−59.12)∗∗∗ −10.390 (−59.16)∗∗∗
hh income (ln, PPP) 2.491 (11.34)∗∗∗ 2.580 (11.75)∗∗∗ 2.490 (11.33)∗∗∗
income not known 0.013 (0.03) −0.003 (−0.01) 0.021 (0.05)
income refused 1.222 (2.17)∗1.234 (2.20)∗1.248 (2.22)∗
traditional (overall) −2.188 (−5.40)∗∗∗
traditional (1) 0.123 (0.44)
traditional (2) −2.222 (−8.04)∗∗∗
traditional (3) 0.336 (1.35)
EASTc−4.126 (−1.69)+−3.649 (−1.46) −4.238 (−1.73)+
COUNTRY GDP (K)c0.244 (3.00)∗0.196 (2.29)∗0.242 (2.96)∗
INCOME INEQUALITY (GINI) c−0.458 (−2.12)∗−0.389 (−1.76)+−0.464 (−2.13)∗
SEX PAY GAP c−22.223 (−2.19)∗−14.338 (−1.26) −22.020 (−2.15)∗
UNEMPLOYMENT RATE c0.282 (1.76)+0.251 (1.50) 0.286 (1.77)+
DIVORCE RATEc−0.211 (−1.45) −0.211 (−1.32) −0.227 (−1.55)
TRAD. COUNTRY (overall)c10.945 (1.49) 8.935 (1.21)
TRAD. COUNTRY (1)c9.309 (1.88)+
TRAD. COUNTRY (2)c−1.146 (−0.17)
TRAD. COUNTRY (3)c−0.775 (−0.16)
Constant 62.997 (32.30)∗∗∗ 61.199 (30.97)∗∗∗ 58.526 (33.02)∗∗∗
Random effects:
country RS: housewife 2.141 †1.912 1.890 †
country RS: unemployed 3.907 †3.944 †3.916 †
country RI 4.317 †4.310 †4.349 †
individual-level error 23.209 †23.174 †23.220 †
N27857 27857 27857
AIC 254460.5 254368.6 254493.1
Source: European Values Study, 2008
+p < 0.10,∗p < 0.05,∗∗∗ p < 0.001;tstatistics in parentheses
For random effects standard deviations are presented; RS - random slope, RI - random intercept
†standard deviation of the random effect is at least double of its standard error
areference category is employment (full-time, part-time and self-employed)
breference category is primary and vocational education
ccountry-level variables
27
Table 7: Regression of well-being of women on set of explanatory variables. Models account-
ing for selection and the full model.
(6) (7) (8)
Prevalence 1 Prevalence 2 Full model
subjective well-being, percentiles
housewifea2.353 (3.18)∗2.230 (3.45)∗∗∗ 2.098 (3.19)∗
housewife x NOT married −1.670 (−1.74)+−1.682 (−1.76)+−1.968 (−2.05)∗
housewife x traditional (1) 1.224 (1.82)+
housewife x traditional (2) −1.321 (−1.92)+
housewife x traditional (3) 1.639 (2.39)∗
housewife x TRAD. COUNTRY (1)c−1.419 (−0.40)
housewife x TRAD. COUNTRY (2)c−1.684 (−0.39)
housewife x TRAD. COUNTRY (3)c2.712 (0.85)
housewife x DIVORCE RATEc0.187 (1.81)+
housewife x WOMEN ACT. RATEc2.014 (0.22)
housewife x WOMEN ACT. RATE2c35.030 (0.71)
housewife x % OF HOUSEWIVESc−12.801 (−1.55) −4.537 (−0.52)
housewife x % OF HOUSEWIVES2c59.277 (2.20)∗42.744 (1.57)
unemployeda−4.146 (−4.50)∗∗∗ −4.144 (−4.50)∗∗∗ −4.095 (−4.46)∗∗∗
retireda2.575 (4.90)∗∗∗ 2.586 (4.92)∗∗∗ 2.628 (5.00)∗∗∗
other employment statusa1.889 (3.09)∗1.891 (3.09)∗1.846 (3.02)∗
age (centered) −0.653 (−12.35)∗∗∗ −0.652 (−12.34)∗∗∗ −0.650 (−12.30)∗∗∗
age2(centered) 0.006 (10.69)∗∗∗ 0.006 (10.67)∗∗∗ 0.006 (10.67)∗∗∗
married 6.911 (14.75)∗∗∗ 6.921 (14.78)∗∗∗ 6.879 (14.71)∗∗∗
married, with own children in the hh 1.059 (1.70)+1.054 (1.69)+1.021 (1.64)
ever had children 1.484 (2.76)∗1.479 (2.75)∗1.451 (2.71)∗
children in the hh −2.664 (−4.96)∗∗∗ −2.670 (−4.97)∗∗∗ −2.647 (−4.94)∗∗∗
secondary educationb1.005 (2.63)∗1.016 (2.66)∗0.959 (2.50)∗
tertiary educationb1.553 (3.39)∗∗∗ 1.556 (3.40)∗∗∗ 1.457 (3.16)∗
social trust 2.038 (29.38)∗∗∗ 2.038 (29.37)∗∗∗ 2.001 (28.82)∗∗∗
health problems −10.394 (−59.19)∗∗∗ −10.394 (−59.19)∗∗∗ −10.365 (−59.09)∗∗∗
hh income (ln, PPP) 2.498 (11.36)∗∗∗ 2.498 (11.37)∗∗∗ 2.567 (11.68)∗∗∗
income not known 0.017 (0.04) 0.017 (0.04) −0.016 (−0.03)
income refused 1.236 (2.19)∗1.233 (2.19)∗1.230 (2.19)∗
traditional (1) 0.126 (0.45)
traditional (2) −2.220 (−8.03)∗∗∗
traditional (3) 0.335 (1.34)
EASTc−4.346 (−1.62) −2.841 (−1.03) −1.061 (−0.37)
COUNTRY GDP (K)c0.232 (2.63)∗0.259 (3.08)∗0.204 (2.38)∗
INCOME INEQUALITY (GINI) c−0.429 (−1.77)+−0.393 (−1.49) −0.382 (−1.48)
SEX PAY GAP c−23.003 (−2.18)∗−24.230 (−2.30)∗−14.780 (−1.33)
UNEMPLOYMENT RATE c0.307 (1.83)+0.293 (1.80)+0.268 (1.64)
DIVORCE RATEc−0.181 (−1.16) −0.184 (−1.22) −0.128 (−0.77)
TRAD. COUNTRY (1)c10.123 (1.94)+
TRAD. COUNTRY (2)c−1.446 (−0.22)
TRAD. COUNTRY (3)c−5.991 (−1.06)
WOMEN ACT. RATEc−8.272 (−0.68)
WOMEN ACT. RATE2c−9.507 (−0.13)
% OF HOUSEWIVESc8.294 (0.64) 19.449 (1.44)
% OF HOUSEWIVES2c14.265 (0.29) −10.271 (−0.21)
Constant 58.622 (27.49)∗∗∗ 57.489 (28.20)∗∗∗ 59.785 (26.97)∗∗∗
Random effects:
country RS: housewife 2.063 †1.554 1.479
country RS: unemployed 3.928 †3.922 †3.908 †
country RI 4.455 †4.397 †4.209 †
individual-level error 23.221 †23.222 †23.175 †
N27857 27857 27857
AIC 254467.2 254465.3 254342.1
Source: European Values Study, 2008
+p < 0.10,∗p < 0.05,∗∗∗ p < 0.001;tstatistics in parentheses
For random effects standard deviations are presented; RS - random slope, RI - random intercept
†standard deviation of the random effect is at least double of its standard error
areference category is employment (full-time, part-time and self-employed)
breference category is primary and vocational education
ccountry-level variables
28
The risk of divorce served as a measure of economic risk faced by women choosing home-
making career, however the results of model 5 question this assumption. It seems that despite
negative effect on individual level27, commonness of divorce is associated with improved situ-
ation of housewives. The possible reason for that may be protection of housewives by divorce
law. In many countries sharing of assets during divorce is subordinated to the rule of equality:
courts recognize worse economic position of the non-working spouse and compensate for it by
equally sharing the wealth accumulated during the period of marriage. This interpretation is
consistent with the fact that higher divorce rates improve well-being of housewives, but not of
women in general (coefficient “DIVORCE RISK” in basic model and model 5)28 . The same is
confirmed by analysis for subgroups of women (results not shown): country-level divorce risk
increases well-being of housewives only for women with household income above the country
median. For women with lower household incomes the effect is statistically insignificant.
However, if country-specific divorce risk informs to some extent about economic conse-
quences of divorce, it should vary between countries, depending on women’s employment
position. Indeed, additional analyses (results not shown, available upon request) inform, that
the positive effect of country divorce risk on housewives’ well-being is significant only in
countries where gender pay gap is lower and where homemaking is less prevalent (in both
cases: below median). In other words, whereas the effect of divorce rate on homemakers’
well-being is overall positive, it is not so in countries where women’s earning opportunities
are generally worse compared to men.
Selection to homemaking I expected that two selection effects may occur: selection to
homemaking (women most devoted to the family stay at home) and selection to employment
(women who have poor employment chances stay at home). I hypothesized that the first of
27There is also literature suggesting, that the effect of individual divorce is not so clearly negative. Gardner
and Oswald (2006) and Clark et al. (2008) have shown that - although divorce decreases happiness in the short
run - both men and women recuperate within a few years, and in longer run even improve their well-being over
the pre-divorce period. Andress and Broeckel (2007) show that although women suffer disproportionately from
economic losses after divorce, they seem to improve other aspects of their life, which makes the well-being drop
less sharp for them. Another paper (Bedard and Deschenes, 2005) after controlling for the negative selection into
divorce concludes that women who have ever divorced live in households with higher income than never-divorced
women.
28There is also literature suggesting that easiness of getting divorce may bring benefits to women. Stevenson
and Wolfers (2006) show, that introduction of unilateral (no-fault) divorce was associated with decrease of home
violence with 30%, suicide rates of women (8-16% decrease) and homicide rates (number of women killed by
their partners decreased 10%).
29
these effects may prevail at lower, and the second - at higher employment levels. Overall,
combination of both effects should produce a curvilinear relationship, in which middle em-
ployment levels are most beneficial for housewives (compared to employed women).
Results of regression models 6 and 7 (table 7) partly support this hypothesis. Coefficients
related to women’s activity rate (model 6) are insignificant, but those related to prevalence
of housewives (model 7, “housewife x % OF HOUSEWIVES2”) show curvilinear trend: the
well-being of housewives is overall lower in countries where homemaking is rare and grows
after passing the value of about 30% of housewives in the population. Contrary to expectations,
the negative selection to homemaking seems to dominate at higher (and not lower) levels of
homemaking prevalence.
Full model Models presented so far included theoretically important predictors one-by-one.
The third column in table 7 reports the results of the full model, which simultaneously accounts
for values, divorce risk and selection effects. AIC value shows that the model fit is better than
any of the models tested so far; in particular it is better than model 7, best so far, and the
random effect of homemaking on well-being is insignificant: the cross-country variation has
been appropriately controlled for.
Coefficients concerning country gender climate (“housewife x TRAD. COUNTRY”) re-
mained insignificant even after controlling for divorce rate and selection. Effect of country-
specific divorce risk remained significant, with the positive (instead of the expected negative
one) effect on well-being of housewives. After controlling for divorce rate and values, the
effect of homemaking prevalence became insignificant.
4.3 Examination of subgroups
The presented results have revealed two issues that require further investigation. The first
one relates to the effect of “country gender climate” on well-being of housewives. Although
the country-level variables don’t have an significant effect, fit of the model is relatively good.
Moreover, (table 3) the choice to work for pay or to stay at home may depend on different
mechanisms in eastern and western countries. In particular, gender-related values and attitudes
are expected to play a more important role in the West (Mikucka, 2009, 2010).
To explore this aspect deeper, I estimate an additional regression model for the East and the
30
West separately (table 8). Indeed, in the West, the traditional “gender climate” of a country (the
relevant dimensions are: “women do not need to work for pay” and “for women and children
is better if women stay at home”) is related to higher well-being premium from homemaking.
Table 8: Regression of well-being of women on set of explanatory variables. Additional mod-
els for western and eastern Europe separately.
(1) (2)
West East
housewifea2.674 (1.93)+3.022 (2.51)∗
housewife x NOT married −0.639 (−0.50) −2.233 (−1.45)
housewife x traditional (1) 0.964 (1.06) 1.761 (1.71)+
housewife x traditional (2) −1.172 (−1.23) −1.104 (−1.08)
housewife x traditional (3) 0.232 (0.25) 3.085 (2.97)∗
housewife x TRAD. COUNTRY (1)c8.704 (1.70)+−5.332 (−0.91)
housewife x TRAD. COUNTRY (2)c−1.843 (−0.26) −4.263 (−0.81)
housewife x TRAD. COUNTRY (3)c18.881 (2.63)∗−5.644 (−1.16)
housewife x DIVORCE RATEc0.694 (2.61)∗0.211 (1.94)+
housewife x % OF HOUSEWIVESc−8.815 (−0.49) 2.143 (0.16)
housewife x % OF HOUSEWIVES2c15.170 (0.40) 120.538 (1.38)
...
(remaining coefficients not shown)
...
Random effects:
country RS: housewife 1.168 0.000
country RS: unemployed 4.453 †3.271 †
country RI 2.962 †4.087 †
individual-level error 23.187 †23.138 †
N11165 16692
AIC 101920.8 152317.9
Source: European Values Study, 2008
+p < 0.10,∗p < 0.05,∗∗∗ p < 0.001;tstatistics in parentheses
For random effects standard deviations are presented; RS - random slope, RI - random intercept
†standard deviation of the random effect is at least double of its standard error
areference category is full-time employment
ccountry-level variables
Table 8 also shows that the positive effect of divorce risk on housewives’ well-being is
stronger in the West. This may indicate higher level of judicial protection of women during
(and after) divorce in these countries. This result also shows the robustness of the divorce-risk
coefficient. In particular, the effect is not driven by the countries of the former Soviet Union.
The second issue demanding further check is the effect of divorce risk, which proved to
be positive, contrary to expectations. To test the explanation that refers to economic gains
in case of divorce, I distinguish between housewives from richer and poorer households29.
Separate estimation for each group shows that the positive effect of divorce risk appears only
for richer households (results not shown, available upon request). This is consistent with the
29I used following three specifications: (1) above vs. below mean country income; (2) above vs. below median
country income; (3) highest and lowest 25 percentiles in a country. All specifications give results leading to the
same conclusions.
31
claim that housewives’ well-being is higher in high-divorce countries because costs of divorce
are accompanied / compensated by possible economic gains, which supposedly are higher in
more wealthy households.
Additional table, (table 9) shows results of separate analysis for older and younger women.
Results show, that the processes differ across these groups, in particular the positive effect
of homemaking on well-being is lower (and insignificant) among younger women, and the
negative effect of not being married is stronger for older women. Moreover, own attitudes (but
not the country climate) seem to affect the well-being of younger housewives more.
Table 9: Regression of well-being of women on set of explanatory variables. Additional mod-
els for younger and older women separately.
(1) (2)
Older Younger
47+ yrs. 18-46 yrs.
housewifea1.738 (1.72)+1.505 (1.57)
housewife x NOT married −2.469 (−1.94)+−0.779 (−0.52)
housewife x traditional (1) −0.119 (−0.12) 2.429 (2.61)∗
housewife x traditional (2) −0.903 (−0.91) −1.405 (−1.45)
housewife x traditional (3) 0.341 (0.35) 2.861 (2.93)∗
housewife x TRAD. COUNTRY (1)c2.152 (0.42) −6.764 (−1.26)
housewife x TRAD. COUNTRY (2)c−4.074 (−0.61) 1.880 (0.31)
housewife x TRAD. COUNTRY (3)c5.990 (1.14) 1.653 (0.36)
housewife x DIVORCE RATEc0.154 (0.95) 0.236 (1.62)
housewife x % OF HOUSEWIVESc−11.277 (−0.88) −0.689 (−0.05)
housewife x % OF HOUSEWIVES2c42.147 (1.10) 52.803 (1.26)
...
(remaining coefficients not shown)
...
Random effects:
country RS: housewife 1.853 2.445 †
country RS: unemployed 4.321 †3.714 †
country RI 4.243 †4.368 †
individual-level error 23.129 †23.152 †
N14128 13729
AIC 128957.0 125351.5
Source: European Values Study, 2008
+p < 0.10,∗p < 0.05,∗∗∗ p < 0.001;tstatistics in parentheses
For random effects standard deviations are presented; RS - random slope, RI - random intercept
†standard deviation of the random effect is at least double of its standard error
areference category is full-time employment
ccountry-level variables
4.4 Robustness checks
Excluding former soviet Union Because countries of former Soviet Union form a specific
cluster (high well-being premium from homemaking combined with high divorce risk), I esti-
mate the full regression model excluding these group of countries. Results (not shown, avail-
able upon request) remain basically unchanged.
32
Influential countries I check how much estimations results depend on particular countries
included in the sample. I do this by calculating dfbetas, which measure how much each coef-
ficient changes after excluding given country30. In other words, dfbetas help detecting if any
country is a source of instability of coefficients of interest.
Table 10: DFbetas for each country for the full model.
Country housewife housewife x housewife x housewife x housewife x housewife x housewife x
TRAD. TRAD. TRAD. DIVORCE % OF % OF
COUNTRY (1) COUNTRY (2) COUNTRY (3) RISK HOUSEWIVES HOUSEWIVES2
Albania −0.16 0.02 −0.00 0.05 0.07 0.04 −0.03
Azerbaijan −0.14 −0.58 −0.01 −0.53 0.25 0.56 −0.20
Austria 0.23 −0.06 0.14 −0.03 0.16 −0.02 −0.01
Armenia 0.14 0.14 −0.14 0.11 −0.23 0.27 −0.21
Belgium 0.10 −0.08 −0.10 −0.20 0.08 0.03 0.01
Bosnia Herzegovina −0.01 0.07 −0.07 0.08 −0.20 −0.04 −0.04
Bulgaria 0.07 −0.28 0.14 −0.17 −0.10 −0.21 0.22
Belarus 0.04 −0.00 −0.01 0.07 0.07 −0.13 0.09
Cyprus 0.24 −0.31 0.17 −0.04 0.05 0.36 −0.25
Czech Republic 0.20 −0.14 0.33 −0.15 0.21 −0.15 0.16
Denmark −0.07 0.01 0.04 0.21 0.08 0.13 −0.17
Estonia 0.10 0.02 −0.01 0.00 0.01 −0.03 0.02
Finland −0.02 −0.08 0.08 0.01 0.06 0.12 −0.09
France −0.30 0.17 0.26 0.12 0.09 0.05 −0.06
Georgia 0.05 0.10 −0.03 −0.07 0.07 −0.05 0.06
Germany −0.09 0.29 −0.32 0.75 −0.05 0.22 −0.39
Greece −0.36 0.48 −0.46 −0.24 0.03 −0.59 0.53
Hungary −0.03 0.02 0.02 −0.01 0.02 0.04 −0.04
Ireland −0.11 −0.17 0.11 0.06 0.09 −0.17 0.15
Latvia 0.10 0.07 −0.05 0.01 −0.03 −0.03 0.02
Lithuania 0.01 −0.21 −0.02 −0.26 −0.15 0.14 −0.01
Luxembourg −0.02 0.04 0.09 0.06 −0.00 −0.06 −0.00
Malta −0.06 0.04 −0.00 0.00 0.07 −0.18 0.28
Moldavia 0.01 −0.08 0.21 0.15 0.24 0.09 −0.13
Montenegro −0.05 0.04 −0.01 0.01 −0.00 −0.01 −0.02
Netherlands −0.18 0.26 0.00 −0.07 0.01 0.02 −0.06
Poland −0.02 −0.02 0.03 −0.05 0.05 0.04 0.01
Portugal −0.06 0.02 0.02 −0.03 0.00 −0.01 0.01
Romania −0.14 0.28 −0.44 0.06 −0.05 −0.08 0.12
Russian Federation 0.00 −0.10 0.04 −0.17 −0.23 0.00 0.04
Serbia 0.11 −0.13 −0.06 −0.12 −0.36 −0.26 0.19
Slovak Republic 0.11 0.05 −0.02 0.04 0.04 0.09 −0.09
Slovenia −0.03 0.10 −0.13 −0.02 −0.21 −0.28 0.22
Spain −0.14 0.19 −0.07 0.30 −0.26 −0.33 0.18
Switzerland 0.12 0.09 −0.03 0.04 0.12 0.10 −0.13
Ukraine −0.00 −0.00 0.12 −0.16 0.00 0.01 0.04
Source: European Values Study, 2008
Overall, values of dfbetas (table 10) are low: all of them are below 1, and exceede 0.5
only for Azerbaijan, Greece and Germany. I re-estimate the full model after excluding these
3 countries, and size and sign of obtained coefficients (results not shown) remain basically
unchanged: the only change concerns significance of the interaction of homemaking and the
30Dfbetas are calculated using the formula: df beta(j)=(b(j)−b(j)i)/se(j)i, where bis the baseline coeffi-
cient for variable j,bi- coefficient for the same variable jafter excluding country i, and sei- standard error of
coefficient jafter excluding country i. Since dfbetas are not formal statistics (no formal statistical test exists),
therefore there is no strict cutoff value. As a rule, values above above 2/√nor 3/√nare considered influential,
and above 1 - strongly so. In case of this analysis, cutoff values are 0.5 (3/√36) and 0.33 (2/√36)
33
prevalence of homemakers (squared), which becomes significant at the 90% level.
Dependent variable I perform a robustness check by estimating models for the two original
variables used to construct the dependent variable: self-assessed happiness and life satisfaction
recoded into dummy variables in a way that approximates the 50%/50% percent. The models
are estimated using the multilevel logistic regression (table 11).
Table 11: Robustness check. Regression of happiness and life satisfaction on set of variables
used in the final model. Logistic model, table presents odds-ratios.
Happiness Happiness Life satisfaction
(1-2 vs. 3-4) (1-3 vs. 4) (1-7 vs. 8-10)
housewifea1.163 (1.73)+1.177 (2.60)∗1.126 (2.13)∗
housewife x NOT married 0.886 (−1.04) 1.043 (0.39) 0.858 (−1.66)+
housewife x traditional (1) 0.869 (−1.56) 1.004 (0.06) 1.214 (2.98)∗
housewife x traditional (2) 1.091 (0.98) 0.871 (−1.80)+0.926 (−1.15)
housewife x traditional (3) 0.968 (−0.35) 1.096 (1.25) 1.193 (2.68)∗
housewife x TRAD. COUNTRY (1)c1.033 (0.07) 0.839 (−0.56) 1.165 (0.53)
housewife x TRAD. COUNTRY (2)c0.497 (−1.41) 1.617 (1.19) 0.555 (−1.68)+
housewife x TRAD. COUNTRY (3)c0.369 (−2.74)∗1.901 (2.06)∗1.588 (1.76)+
housewife x DIVORCE RATEc1.000 (−0.00) 1.030 (3.01)∗1.001 (0.09)
housewife x % OF HOUSEWIVESc4.732 (1.57) 0.471 (−0.90) 0.317 (−1.59)
housewife x % OF HOUSEWIVES2c0.141 (−0.61) 28.216 (1.44) 65.252 (1.95)+
...
(remaining coefficients not shown)
...
Random effects:
country RS: housewife 0.000 (−0.00) 0.000 (−0.00) 0.000 (−0.00)
country RS: unemployed 0.388 (−4.49)∗∗∗ 0.165 (−2.86)∗0.278 (−4.90)∗∗∗
country RI 0.363 (−7.65)∗∗∗ 0.350 (−8.14)∗∗∗ 0.275 (−9.91)∗∗∗
N27980 27980 28155
AIC 21367.1 23942.4 33111.0
Source: European Values Study, 2008
+p < 0.10,∗p < 0.05,∗∗∗ p < 0.001;tstatistics in parentheses
For random effects standard deviations are presented; RS - random slope, RI - random intercept
†standard deviation of the random effect is at least double of its standard error
areference category is employment (full-time, part-time and self-employed)
ccountry-level variables
Results confirm that homemaking is accociated with ceteris paribus higher happiness and
life satisfaction. The data also show that the happiness and life satisfaction of housewives have
partially different determinants. Life satisfaction depends more on marital status, own gender
attitudes concerning role of women (‘women don’t need to work fo pay’ and ‘for women and
children it is better to stay at home’), but also traditionality of the country and prevalence of
homemakers (squared). Happiness depend more on attitudes supporting the egalitarian sharing
of duties and country divorce rate. Importantly, two out od the models indicate that traditional
gender-climate (the third dimension) of a country raises well-being of homemakers.
34
Excluding problematic variables I re-estimate the full model excluding income and self-
assessed health problems (results not shown, available upon request).
•Excluding subjective health problems from the model does not significantly alter results,
however fit of the model is significantly worse.
•Analysis conducted after excluding six countries with highest percentage of missing val-
ues of household income31 and limiting sample to respondents who provided information
on household income does not change results significantly.
•Subsequent test consisted of re-estimating full model after dropping the income variable
(for all 36 countries). Sign and size of coefficients do not change considerably, but the
fit of the model is again worse than of the full model.
•In the last test I excluded from the model both health and income variables: again results
remain unchanged and fit of model worsens.
5 Summary and discussion of results
The goal of the paper was to determine how does homemaking affect well-being of women. I
wanted to check if housewives enjoy higher or lower well-being than employed women, if this
effect varies across countries and - if yes - what factors may explain the observed variation.
With the use of multilevel models, I tested three hypotheses. First, that traditional “gender
climate” in a country increases well-being gains from homemaking. Second, that economic
risk associated with specialization to housework decreases well-being of housewives. Third
hypothesis concerned selection effect. I expected a curvilinear relationship between activity
rates or prevalence of homemakers and well-being premium from homemaking.
Results show that full-time homemaking, controlling for a set of demographic, social and
country-level variables has an overall positive effect on the well-being of women. This result
(robust to various model specifications, set of countries included and choice of dependent
variable) is a novelty, taking into account the so-far inconclusive literature on the topic. Size
and significance of random slopes show that the effect differs across European countries.
Analysis of the effect of traditional “gender climate” on the well-being of housewives leads
31The excluded countries are: Denmark, Ireland, Malta, Portugal, Slovenia and Spain: all over 25% of missing
values of household income.
35
to two conclusions. First, only in western European countries the traditional gender-role atti-
tudes in a country are related to higher well-being of housewives. In the East, as well as in the
overall sample, the relationship is not statistically significant. Concluding, either a more re-
fined measure of social pressure is necessary in case of eastern Europe, or the social pressure on
employment / homemaking is there relatively weak. Importance of individual-level attitudes
in this context suggests that the link between traditional values in a country and employment
patterns may be created more by individual beliefs than by social pressure. Investigating the
effect of social pressure in smaller units, such as households or small geographical regions
seems a promising field of future studies.
Second, the analysis of “gender climate” suggests that treating gender-role attitudes or
“gender climate” as unidimensional phenomenon may not lead to valid conclusions. My re-
sults show the importance of distinguishing several dimensions of attitudes. In particular,
one dimension (“home is not men’s business”) has a negative effect on well-being of house-
wives, whereas the effect of the remaining two is positive. Traditional gender attitudes is not a
unidimensional phenomenon, therefore defining and measurement of relevant factors requires
particular care and consideration.
The effect of country-specific divorce risk is positive: in countries with higher divorce
risk housewives enjoy higher well-being than in other countries. This result is opposed to the
literature stressing the economic cost of divorce for women, and I interpret it by referring to
the legal protection of non-working spouses by divorce law. The result is robust to model
specification, countries included in the analysis, and choice of dependent variable. The con-
tradiction between hypothesized and observed effects makes the relationship between divorce
regulations and well-being of women (and especially housewives) a promising area of deeper
investigation.
The results concerning selection effect show a curvilinear relationship between prevalence
of housewives and their well-being: their well-being is highest in countries where there are
particularly many homemakers. This may suggest that the selection to homemaking (stronger,
where the prevalence of homemakers is lower) is predominantly negative, i.e. women more
likely to be happy enter labour market disproportionally more often. However, alternative
explanations are also possible: the larger size of the group of housewives itself may be asso-
ciated with well-being benefits, perhaps because constituting a large group, homemakers are
36
less isolated or more easily build social networks. It is also possible, that the observed regu-
larity results from some other, uncontrolled factor (e.g. institutional aggangement) that creates
especially advantageous conditions for being a housewife.
References
Aber, C. S. (1992). Spousal death, a threat to women’s health: Paid work as a resistance resourceâ?
Journal of Nursing Scholarship, 24(2):95–100.
Alesina, A., Di Tella, R., and MacCulloch, R. (2004). Inequality and happiness: are Europeans and
Americans different? Journal of Public Economics, 88(9-10):2009–2042.
Andress, H.-J., Borgloh, B., Broeckel, M., Giesselmann, M., and Hummelsheim, D. (2006). The eco-
nomic consequences of partnership dissolution - a comparative analysis of panel studies from Bel-
gium, Germany, Great Britain, Italy, and Sweden. European Sociological Review, 22(5):533–560.
Andress, H.-J. and Broeckel, M. (2007). Income and life satisfaction after marital disruption in Ger-
many. Journal of Marriage and the Family, 69(2):500–512.
Becker, G. S. (1981). A Treatise on the Family. Number beck81-1 in NBER Books. National Bureau
of Economic Research, Inc.
Bedard, K. and Deschenes, O. (2005). Sex preferences, marital dissolution, and the economic status of
women. Journal of Human Resources, 40(2):411–434.
Bird, C. E. and Ross, C. E. (1993). Houseworkers and paid workers: Qualities of the work and effects
on personal control. Journal of Marriage and Family, 5(4):913–925.
Blanchflower, D. G. and Oswald, A. J. (2008). Hypertension and happiness across nations. Journal of
Health Economics, 27(2):218–233.
Böhnke, P. (2008). Does society matter? Life satisfaction in the enlarged Europe. Social Indicators
Research, 87:189–210.
Bonini, A. (2008). Cross-national variation in individual life satisfaction: Effects of national wealth,
human development, and environmental conditions. Social Indicators Research, 87:223–236.
Boye, K. (2009). Work and Well-being in a Comparative Perspective - The Role of Family Policy.
European Sociological Review.
Brereton, F., Clinch, J. P., and Ferreira, S. (2008,). Employment and life-satisfaction: Insights from
Ireland. The Economic and Social Review, 39(3):207234.
Bryant, W. K. and Zick, C. D. (2006). The Economic Organization of the Household. Second Edition.
Cambridge University Press.
Bryk, A. S. and Raudenbush, S. W. (1992). Hierarchical Linear Models: Applications and Data Anal-
ysis Methods. Sage Publications, Newbury Park, London, New Dehli.
Cha, Y. and Thebaud, S. (2009). Labor markets, breadwinning, and beliefs. How economic context
shapes men’s gender ideology. Gender & Society, 23(2):215–243.
37
Clark, A. E., Diener, E., Georgellis, Y., and Lucas, R. E. (2008). Lags and leads in life satisfaction: a
test of the baseline hypothesis. Economic Journal, 118(529):F222–F243.
Clark, A. E. and Oswald, A. J. (1994). Unhappiness and unemployment. Economic Journal,
104(424):648–59.
Crompton, R. and Harris, F. (1997). Women’s employment and gender attitudes: A comparative analy-
sis of Britain, Norway and the Czech Republic. Acta Sociologica, 40(2):183–202.
Crompton, R. and Harris, F. (1998). Explaining women’s employment patterns: ‘Orientations to work’
revisited. British Journal of Sociology, 49(1):118–136.
Di Tella, R., MacCulloch, R. J., and Oswald, A. J. (2001). Preferences over inflation and unemploy-
ment: Evidence from surveys of happiness. American Economic Review, 91(1):335–341.
Diener, E., Diener, M., and Diener, C. (1995). Factors predicting the subjective well-being of nations.
Journal of Personality and Social Psychology, 69(5):851 – 864.
Diener, E., Napa-Scollon, C. K., Oishi, S., Dzokoto, V., and Suh, E. M. (2000). Positivity and the
construction of life satisfaction judgments: Global happiness is not the sum of its parts. Journal of
Happiness Studies, 1:159–176.
Easterlin, R. A. (1974). Does economic growth improve the human lot? Some empirical evidence. In
David, P. A. and Reder, M., editors, Nations and Households in Economic Growth, pages 89–125.
New York: Academic Press.
Edmark, K. (2005). Unemployment and crime: Is there a connection? Scandinavian Journal of
Economics, 107(2):353–373.
EVS Foundation/Tilburg University (2010). European values study 2008, 4th wave, integrated dataset.
Technical report, CEPS/ Instead, Luxembourg.
Ferree, M. M. (1984). Class, housework, and happiness: Women’s work and life satisfaction. Sex Roles,
11:1057–1074.
Gardner, J. and Oswald, A. J. (2006). Do divorcing couples become happier by breaking up? Journal
of The Royal Statistical Society Series A, 169(2):319–336.
Gerlach, K. and Stephan, G. (1996). A paper on unhappiness and unemployment in Germany. Eco-
nomics Letters, 52(3):325–330.
Glass, J. and Fujimoto, T. (1994). Housework, paid work, and depression among husbands and wives.
Journal of Health and Social Behavior, 35(2):179–191.
Golding, J. M. (1990). Division of household labor, strain, and depressive symptoms among Mexican
Americans and non-hispanic whites. Psychology of Women Quarterly, 14(1):103–117.
Gonzalez, L. and Viitanen, T. K. (2009). The effect of divorce laws on divorce rates in europe. European
Economic Review, 53(2):127–138.
Helliwell, J. F. (2003). How’s life? Combining individual and national variables to explain subjective
well-being. Economic Modelling, 20(2):331 – 360. Henry Special Issue.
Helliwell, J. F. and Putnam, R. D. (2004). The social context of well-being. Philosophical Transactions
of the Royal Society of London. Series B: Biological Sciences, 359(1449):1435–1446.
38
Hessami, Z. (2010). The size and composition of government spending in Europe and its impact on
well-being. Kyklos, 63(3):346–382.
Kahneman, D. and Krueger, A. B. (2006). Developments in the measurement of subjective well-being.
Journal of Economic Perspectives, 20(1):3–24.
Kalmijn, M. (2010). Country differences in the effects of divorce on well-being: The role of norms,
support, and selectivity. European Sociological Review, 26(4):475–490.
Klumb, P. L. and Lampert, T. (2004). Women, work, and well-being 1950-2000: A review and method-
ological critique. Social Science & Medicine, 58(6):1007 – 1024.
Krause, N. and Markides, K. S. (1985). Employment and psychological well-being in Mexican Ameri-
can women. Journal of Health and Social Behavior, 26(1):15–26.
Lennon, M. C. (1994). Women, work, and well-being: The importance of work conditions. Journal of
Health and Social Behavior, 35(3):235–247.
Lucas, R. E., Clark, A. E., Georgellis, Y., and Diener, E. (2004). Unemployment alters the set point for
life satisfaction. Psychological Science, 15(1):8–13.
Luechinger, S., Meier, S., and Stutzer, A. (2008). Why does unemployment hurt the employed? Evi-
dence from the life satisfaction gap between the public and private sectors. Public Policy Discussion
Papers (Federal Reserve Bank of Boston), No. 08-1.
Luke, D. A. (2004). Multilevel Modelling. Sage University Paper Series on Quantitative Applications
in the Social Sciences, 07-143. Sage, Thousand Oaks, CA.
Mikucka, M. (2009). Division of household labour between spouses: How does Central and Eastern
Europe differ from the West? International Journal of Sociology, 39(1):77–95.
Mikucka, M. (2010). Labour supply of women as a rational allocation of resources. Phd thesis,
Graduate School for Social Research, Polish Academy of Science.
Miller, M. L., Moen, P., and Dempster-McClain, D. (1991). Motherhood, multiple roles, and maternal
well-being. Gender & Society, 5(4):565–582.
Ongaro, F., Mazzuco, S., and Meggiolaro, S. (2009). Economic consequences of union dissolution in
Italy: Findings from the European Community Household Panel. European Journal of Population-
Revue Eurpeenne de Demographie, 25(1):45–65.
Ouweneel, P. and Veenhoven, R. (1991). Cross-national differences in happiness. Cultural bias or
societal quality? In Bleichrodt, N. and Drenth, P., editors, Contemporary issues in cross-cultural
psychology, pages 168–184. Swets & Zeitlinger, Amsterdam, The Netherlands.
Peiró, A. (2006). Happiness, satisfaction and socio-economic conditions: Some international evidence.
Journal of Socio-Economics, 35(2):348 – 365. The Socio-Economics of Happiness.
Pietromonaco, P. R., Manis, J., and Markus, H. (1987). The relationship of employment to self-
perception and well-being in women: A cognitive analysis. Sex Roles, 17:467–477.
Pittau, M., Zelli, R., and Gelman, A. (2010). Economic disparities and life satisfaction in European
regions. Social Indicators Research, 96:339–361.
Poortman, A. (2000). Sex differences in the economic consequences of separation - A panel study of
the Netherlands. European Sociological Review, 16(4):367–383.
39
Schimmack, U., Krause, P., Wagner, G., and Schupp, J. (2010). Stability and change of well being: An
experimentally enhanced latent state-trait-error analysis. Social Indicators Research, 95:19–31.
Schneider, L. and Schimmack, U. (2009). Self-informant agreement in well-being ratings: A meta-
analysis. Social Indicators Research, 94:363–376.
Sironi, M. and Mencarini, L. (2010). Happiness, housework and gender inequality in modern europe.
Accessed online September 2010 on www.happinesseconomics.net.
Steptoe, A. and Wardle, J. (2005). Positive affect and biological function in everyday life. Neurobiology
of Aging, 26(1):108–112.
Stevenson, B. and Wolfers, J. (2006). Bargaining in the shadow of the law: Divorce laws and family
distress. The Quarterly Journal of Economics, 121(1):267–288.
Stevenson, B. and Wolfers, J. (2008). Economic growth and subjective well-being: Reassessing the
Easterlin paradox. NBER Working Paper, No. W14282.
Stokes, J. P. and Peyton, J. S. (1986). Attitudinal differences between full-time homemakers and women
who work outside the home. Sex Roles, 15:299–310.
Tolciu, A. and Zierahn, U. (2010). Women and work: What role do social norms play? MAGKS Papers
on Economics, 201009.
Townsend, A. and Gurin, P. (1981). Re-Examining the Frustrated Homemaker Hypothesis. Work and
Occupations, 8(4):464–488.
Urry, H. L., Nitschke, J. B., Dolski, I., Jackson, D. C., Dalton, K. M., Mueller, C. J., Rosenkranz, M. A.,
Ryff, C. D., Singer, B. H., and Davidson, R. J. (2004). Making a life worth living. Psychological
Science, 15(6):367–372.
Uunk, W. (2004). The economic consequences of divorce for women in the European Union: The
impact of welfare state arrangements. European Journal of Population-Revue Europeenne de De-
mographie, 20(3):251–285.
van Reekum, C. M., Urry, H. L., Johnstone, T., Thurow, M. E., Frye, C. J., Jackson, C. A., Schaefer,
H. S., Alexander, A. L., and Davidson, R. J. (2007). Individual differences in amygdala and ventro-
medial prefrontal cortex activity are associated with evaluation speed and psychological well-being.
Journal of Cognitive Neuroscience, 19(2):237–248.
Waldron, I. and Herold, J. (1986). Employment, attitudes toward employment, and women’s health.
Women & Health, 11(1):79–98.
Wright, J. D. (1978). Are working women really more satisfied? Evidence from several national
surveys. Journal of Marriage and Family, 40(2):301–313.
40
B.P. 48
L-4501 Differdange
Tél.: +352 58.58.55-801
www.ceps.lu
