Rev Econ Household
DOI 10.1007/s11150-014-9244-y
Parenthood wage penalties in a double income society
Sara Cools • Marte Strøm
Received: 14 March 2013 / Accepted: 22 February 2014
Ó Springer Science+Business Media New York 2014
Abstract We estimate how parenthood affects hourly wages using panel data for
Norwegian employees in the years 1997–2007. Though smaller than for most other
OECD countries, we find substantial wage penalties to motherhood, ranging from a
1.2 % wage reduction for women with lower secondary education to 4.9 % for
women with more than four years of higher education. Human capital measures
such as work experience and paid parental leave do not explain the wage penalties,
indicating that in the Norwegian institutional context, mothers are protected from
adverse wage effects due to career breaks. We do however find large heterogeneity
in the effects, with the largest penalties for mothers working full time and in the
private sector. Contrary to most studies using US data and to previous research from
Norway, we find a small wage penalty also to fatherhood. Also for men, the penalty
is greater for those who work full time and in the private sector. A substantial share
of the fatherhood wage penalty is explained by paternity leave.
Keywords
Parenthood penalties Welfare state Gender wage gap
We are grateful to Jenny Clarhäll, Eva Kløve, Andreas Kotsadam, Ingrid Krüger, Jo Thori Lind, Kalle
Moene and Fredrik Willumsen for helpful comments and suggestions. The authors bear all responsibility
for errors and shortcomings. This paper is part of the research activities at the center of Equality, Social
Organization, and Performance (ESOP) at the Department of Economics at the University of Oslo, in
collaboration with the Ragnar Frisch Centre for Economic Research. We acknowledge funding from
the Norwegian Research Council (Frisch project 1110). Data made available by Statistics Norway have
been essential for the research project. ESOP is supported by the Research Council of Norway.
S. Cools
BI Norwegian Business School, Oslo, Norway
e-mail: [email protected]
M. Strøm (&)
Institute for Social Research, Oslo, Norway
e-mail: [email protected]
123
S. Cools, M. Strøm
JEL Classification
J13 J22 J24 J31
1 Introduction
Women’s labor force participation increased greatly in Norway during the last
decades of the twentieth century, closely approaching that of men. Over the same
period, as several European countries experienced falling fertility rates, the total
fertility rate in Norway increased and stayed far above the European average. In a
country where the combination of work and family is encouraged through public
policies such as universal child care, paid parental leave and extensive job
protection for parents, and where not only the two career household, but also the two
carer household, is encouraged, the double fact of high female labor force
participation and high fertility is the hallmark of these policies’ success.
Unfortunately, and puzzling to some, the gender gap in hourly wage has not declined
as successfully over the period. It seems rather that the convergence in men’s and
women’s hourly wage stagnated some time during the 1980s, only to pick up slightly
from 2000 onwards. As in most other countries, the gender wage gap is greater among
parents than among non-parents—a phenomenon known as the family gap in wages.
Quantifying the wage effects of parenthood is therefore of interest to policy
makers wishing to address the overall gender wage gap. In this paper we study how
children affect men’s and women’s hourly wages, using panel data on hourly wages
in Norway from 1997 to 2007. Using fixed effects estimation and register data
covering both the private and the public sector (in total 80 % of the working
population), this is the first broad panel data study of parenthood wage effects in
Norway. In addition to rich register data, the institutional setting provided by
Norway as a typical Nordic welfare state makes it an interesting case for the
exploration of causes behind the family wage gap.
The international empirical evidence on a negative relationship between having
children and women’s labor market outcomes, like wages and labor supply, is
substantial (e.g. Korenman and Neumark (1992), Waldfogel (1997) and Budig and
England (2001)). The evidence on the effects of fatherhood is both more scarce and less
consistent (see for instance Millimet (2000), Lundberg and Rose (2002), Simonsen and
Skipper (2008), Astone et al. (2010), Wilde et al. (2010)). Generally, the literature finds a
smaller family gap in pay for mothers in the Scandinavian countries (Datta Gupta and
Smith 2002; Harkness and Waldfogel 2003a). Differences in family policies are argued
to be key in explaining differences in family gaps across countries (Waldfogel 1998).
For example, Waldfogel (1998) show that in countries with paid parental leave, women
return to work more quickly and at the same time maintain a good job match. The
institutional context may therefore leave a smaller role for the motherhood wage penalty
being caused by an adverse effect on firm-specific human capital.
This paper aims to investigate further two main explanations for why we observe
a negative wage effect on mothers’ wages.1 Commonly, children are considered to
1
Unless tied specifically to institutional or biological gender differences, these explanations would also
apply to fatherhood wage penalties.
123
Parenthood wage penalties in a double income society
have an effect on women’s productivity—by a comparative reduction either in their
human capital or in their work effort—which again will be reflected in their wages.
As proposed by Mincer and Polachek (1974), childbearing and child rearing may
cause a comparative reduction in mothers’ labor market experience, both through
periods out of the labor market and through periods of reduced working hours,
which then adversely affects mothers’ accumulation of human capital. Several
empirical studies have shown that when controlling for experience—measured as
time in active employment, net of career breaks and/or adjusted for part time
work—a large part of the wage penalty is explained (Waldfogel 1997; Budig and
England 2001; Datta Gupta and Smith 2002; Anderson et al. 2002; Wilde et al.
2010).
We explore a human capital explanation for the family wage gap by including
actual job experience, the length of parental leave and part-time work as explanatory
variables in our estimations of parenthood wage effects. As the Norwegian
institutional setting provides, on the one hand, long periods of parental leave, while
on the other, high job-protection during leave, we expect two conditions to be met:
First, that the overall wage penalty is lower in Norway if the loss of job-specific
human capital explains a large share of the wage-penalty in other settings. Second,
that the inclusion of experience measures explains a smaller share of the wage
penalty because they only pick up the deterioration of human capital in general.
In the other part of the productivity explanation for mothers’ reduced hourly
wage, Becker (1985) theory of a conflict between effort in home production and
effort at work implies that among women working the same number of hours,
mothers will put in less effort per hour. Several studies have found a negative
relationship between housework load and wages (Coverman 1983; Shelton and
Firestone 1989; McAllister 1990; Hersch and Stratton 1997). Career break effects
may also wrongly be attributed to loss of human capital. Using Swedish data,
Albrecht et al. (1999) find that the type of career break matters for its effect on
wages, and that the effects vary by gender. As they find no effect of maternity leave
and a negative effect of paternity leave, they interpret this as parental leave
signalling worker effort.2
We explore the effort explanation in different ways. First, we expect workers to
be more free to regulate their own effort in occupations requiring higher education,
whereas workers in occupations requiring lower education will be relatively more
forced to reducing working hours rather than effort in their work. We therefore
expect parenthood to affect hourly wages more negatively in higher education
groups (whereas it affects working hours more negatively in the lower education
groups). Second, Anderson et al. (2003) suggest that the role of effort can be
investigated by studying how the parenthood wage penalty evolves as the child
grows older, since smaller children demand more effort at home than older children.
The parenthood wage penalty should consequently be smaller when the child grows
2
Mothers being expected by employers to take leave for a considerable period of time means maternity
leave is not a strong signal to the employer of their type. For men, on the other hand, the length of
parental leave is a strong signal of how much effort they will spend at work.
123
S. Cools, M. Strøm
older.3 Third, if effort is part of the explanation and there is a constant amount of
effort to be shared between parents, we would expect men’s penalties to increase as
women’s decline. Last, we should also expect to see larger wage penalties in the
private sector, where individual wage setting is more common but job protection
still applies, as long as individually bargained wages are more closely linked to
effort.
Other explanations from both the empirical and theoretical literature do not
evoke an effect of children on women’s productivity: Mothers may earn lower
wages than non-mothers due to employers’ discrimination or because they seek
employment in ‘‘mother-friendly’’ jobs, for instance offering greater flexibility or
requiring skills that deteriorate less rapidly during time spent out of the labor
force—at the expense of lower wages (Budig and England 2001; Nielsen et al.
2004). We do not test hypotheses of discrimination in this paper, and cannot rule out
that it plays a role. Public sector jobs are often viewed as more easily compatible
with family life. We explore whether a change in sector of employment can explain
some of the parenthood wage effects by including sector of employment as an
explanatory variable in our regressions.
Our results show that even in the context of a double income welfare state like
Norway, there is a substantial wage penalty to motherhood. Ranging from 1.2 %
during the first years for women with lower secondary education to 4.9 % for
women with more than four years of higher education, motherhood wage penalties
in Norway are lower than estimates found for the UK (Harkness and Waldfogel
2003b) and the US (Korenman and Neumark 1992; Budig and England 2001; Wilde
et al. 2010), and comparable to estimates found for Germany (Harkness and
Waldfogel 2003b) and Denmark (Nielsen et al. 2004), among others. Contrary to
most other studies, we find a negative (though comparatively small) effect of having
children of about .3 to .4 % on the hourly wages of men with middle length
education.
The wage penalty to motherhood is not explained when we include measures of
experience, parental leave, working part time and sector of employment. The results
indicate that Norwegian family policies protect mothers from adverse wage effects
following a career break. We find evidence that changes in effort might be important
for the explanation of a parenthood wage effect. As hypothesized, the wage effects
to parenthood is larger for higher educational groups and for people working in the
private sector, where we expected a change in effort to affect wages more.
2 Estimating the wage penalty to parenthood
The obvious problem of childbearing being endogenous to wages—and to labor
market outcomes in general—is dealt with most completely in studies that use
3
However, Wilde et al. (2010) argue that if parenthood affects both wage levels and wage growth, the
effect of reduced effort during early years may have a lasting effect. If the wage trajectory changes
growth rate, the difference between parents and childless individuals may even be increasing, a pattern
found in Wilde et al. (2010) on US data. A long term persistence in motherhood wage effects is found in
several other studies, like Viitanen (2014) using UK data.
123
Parenthood wage penalties in a double income society
instrumental variables for the childbearing decision. A shortcoming of this approach
is that it cannot be applied to the study of effects of having the first child.4 The
literature on the wage penalties to parenthood is mainly concerned with effects of
the first child, although effects of subsequent children are often estimated in
addition. Given this focus, endogeneity is primarily addressed by the inclusion of
individual fixed effects in the estimation (this is done by Waldfogel (1997), Hersch
and Stratton (1997), Budig and England (2001) and Anderson et al. (2002)).
Even when controlling for individual fixed effects, there is the possibility of
selection of individuals into parenthood and of systematic differences in the career
path of individuals who have children at different ages, which would both yield
biased estimates (Korenman and Neumark 1991; Antonovics and Town 2004).
Wilde et al. (2010) address the issue of selection into parenthood by restricting their
sample to contain only individuals who are observed to become parents. This
restriction in combination with fixed effects estimation means identification of the
parenthood wage penalties comes from comparing the development in wages of
individuals with children to the development in wages of individuals who have not
yet had children—in essence, causal interpretation rests on the assumption of
exogenous variation in the timing of when to have children.5Wilde et al. (2010)
further address the endogeneity in the timing of childbearing by running separate
analyses of the wage effects for different skill levels (defined by Armed Forces
Qualification Test scores taken at ages 14–21). If the timing of children is random
within a skill group, separate analyses will yield unbiased estimates. Put less
strongly, if part of the unobserved heterogeneity that is correlated both with the
timing of children and with wages varies systematically between certain groups,
estimation of wage effects within groups will reduce the bias.
Our data do not contain any exogenous measure of ability. Given that there is a
strong tradition in Norwegian centralized wage setting of tying qualification to
education length, the length of individuals’ education will in large part determine
how their wages develop. Length of education is also a major factor influencing
when to have children. For these reasons, subsample analysis based on education
length might reduce the bias resulting from time-variant heterogeneity amongst
individuals who have children at different points in time. We divide the sample into
four educational categories.6 We include individual fixed effects in order to take out
time-invariant unobservable heterogeneity that is correlated both with the average
wage level and with the propensity to have children at different ages. The resulting
identifying assumption for a causal interpretation of our results is that, within an
education group, there is no systematic difference in the wage paths of those who
4
The most widely applied instruments for having additional children are twins and child gender
composition (first used by Rosenzweig and Wolpin (1980) and Angrist and Evans (1998), respectively).
5
As postponing childbearing is found to reduce the negative career effects of having children (Hofferth
1984; Taniguchi 1999; Buckles 2008; Miller 2011), the assumption is under siege.
6
Our division—further described in Sect. 3—is similar to one in Anderson et al. (2003), but we have a
separate group with a master degree or higher. This is a small group, but interesting because of the large
career potential, meaning motherhood may be a substantial hinder as is shown in e.g. Bertrand et al.
(2010). The timing of the first child is clearly different between education groups: Age at first birth
increases by one year on average between one educational group and the next.
123
S. Cools, M. Strøm
have children at a certain point in time and those who do not—other than what is
caused by the event of having children itself.
We estimate the following relationship between wages and having children:
ln wit ¼ p1a Child1½1 5it þ p1b Child1½6 10it þ p1c Child1½11 15it
þ p1d Child1½ [ 15it þ p2a Child2½1 5it þ p2b Child2½6 10it
þ p2c Child2½11 15it þ p2d Child2½ [ 15it þ p3a Child3½1 5it
þ p3b Child3½6 10it þ p3c Child3½11 15it þ p3d Child3½ [ 15it
ð1Þ
þ p4 Child4it þ qa Pregnantit þ qb Babyit þ df ðAgeit Þ þ et þ vi þ uit ;
where i indicates individual and t indicates year of the observation. w denotes hourly
wage, which enters log transformed. The Child1½it variables are variables indicating the age categories of the first child. Child1½1 5it takes a value between 0
and 1 according to how much of year t individual i’s first child is between the age of
1 to 5. Child1½6 10it thus takes a value between 0 and 1 according to how much
of year t individual i’s first child is aged 6 to 10, and so on. The Child2½it variables
indicate the age categories of a second child, and the Child3½it variables those of
the third child. Child4it indicates whether the individual has had a additional children beyond the third child.
Parental leave benefits in Norway entail wage replacement during approximately
the first year after a child is born, contingent on the parent’s labor market behavior
during the last 10 months prior to the child’s birth. Wages observed during pregnancy
and during the parental leave period may therefore differ significantly from wages
received prior to the pregnancy, and from the wage realized after returning to the labor
force. The variable Pregnantit , which takes a value between 0 and 0.75 according to
how much of the year was spent ‘‘in pregnancy’’, and the variable Babyit , taking a value
between 0 and 1 according to how much of that year the individual had a first-born
baby younger than 1 year, take out the variation in wages caused by these periods.
We also include a fourth order polynomial in each parent’s age (f ðAgeÞ). et
denotes year fixed effects and vi individual fixed effects. Hence, the wage path of
non-parents is given by the age function, net of year and individual fixed effects.7
Our main interest lies with the parameters p1a , p1b and p1c , the effect on wages of
having a first child aged 1–5 years, 6–10 years and 11–15 years, respectively. For
(weakly) negative values of the p1 s, p1a ¼ p1b ¼ p1c would correspond to a constant
wage effect of parenthood over time. jp1a j [ jp1b j [ jp1c j corresponds to a scenario of
parents’ wages catching up with their original wage path, whereas jp1a j\jp1b j\jp1c j
corresponds to the case in which parents’ wages keep deteriorating relative to the
wages of non-parents.
3 Data
The basis for our main sample is all Norwegian residents born between 1950 and
1980, except parents who at one point had twins, triplets or any type of multiple
7
The fourth order polynomial is chosen in order to be as flexible as possible, given that the use of age
dummies is not feasible due to multicollinearity with year and individual fixed effects.
123
Parenthood wage penalties in a double income society
birth. The sample is further restricted by the availability of data on hourly wages
(described below).
We run separate estimations in four subsamples based on the length of
individuals’ education. The education data is provided by Statistics Norway and
makes it possible to track individuals’ highest registered education over time. As an
individual’s ultimate education level may be influenced by childbearing, we face a
trade-off between a late observation of education, that more likely reflects the
individual’s actual wage path, and an early and less endogenous measure of
education. Since it is not clear how much there is to gain in terms of exogeneity by
using observations on education at younger ages, and because there is something to
be gained by moving close to the actual education of the individual during his or her
working life, we use education measured at age 25 for the main part of our analysis.
The median age at which an individual attains the highest registered education is 24
years in our sample.
The first education group is the group of individuals who have completed nine
years of lower secondary education or less (this group also includes individuals for
whom education information is missing). This group constitutes about 40 % of our
sample, based on the observation of education at age 25. Typical occupations for
this education group are for instance shop clerks, transport workers (men), cleaning
personnel (women). The second group has completed additional three years of upper
secondary education, and constitutes roughly 30 % of our sample. Typical
occupations are technicians and electricians (men) and secretaries and health care
personnel (women). The third group has a ‘‘lower degree’’ in higher education (four
years or less of higher education, typically a bachelor’s degree or the equivalent),
and is typically employed as engineers (men) or nurses (women). This group
constitutes almost 30 % of the sample. Finally, the fourth education group consists
of those with four years or more of higher education (and thus a ‘‘higher degree’’).
Constituting only about 2 % of the whole sample, this group is typically employed
as doctors, architects and engineers.
Our measure of hourly wages is constructed using information on contracted
hours and monthly wages in Statistics Norway’s ‘‘Wage statistic’’ (‘‘Lønnsstatistikken’’). The Wage statistic is based on employer reports for a sample of
Norwegian enterprizes on all employees by the 1st of October. 8 Every year all
public enterprizes and all private enterprizes with more than a certain number of
employees are included (the number varies with industry and year), for the
remaining private sector a 50 % sample of medium size enterprizes and a 20 %
sample of small enterprizes is drawn every year. Employment in agriculture,
hunting and forestry is left out. So are enterprizes with 3 or less employees. On
average, the Wage statistic covers about 80 % of Norwegian employees (100 % of
the public sector employees and 70 % of the private sector employees) every year.
The information on birth year, education and the linking of parents to their children
comes from Statistics Norway’s demography, family and education registers.
8
Results are therefore representative of employees only. The share of self-employed in the labor force is
about 4 % for women and almost 10 % for men during the period under study, with smaller shares for
individuals with small children (Villund 2005).
123
S. Cools, M. Strøm
The resulting panel is unbalanced, both due to the sampling of medium size
enterprizes, and because people may move in and out of employment covered by the
Wage statistic. A missing observation for a given year may mean either that the
person in question does not work that year, that she is self-employed, or that she is
employed in an enterprize that is not included in that year’s sample.9
In Sect. 5, we include four covariates that are clearly endogenous to the
childbearing decision (so-called ‘‘bad controls’’) in order to see to what extent they
explain the parenthood wage penalties. In Sect. 6 we divide the sample into
different subsamples based on these four variables. As some of these variables are
only observable from 1993 onwards, the samples used in the analyses in Sects. 5
and 6 will be restricted to parents of children born between 1994 and 2007.
Experience is constructed counting the cumulative number of years the
individual is registered with occupational income above the basic amount of the
Norwegian social security system (G), which is a common measure of labor force
participation based on Norwegian income data. Parental leave counts the
cumulative number of days the individual has been registered with paid parental
leave. This register has reliable information from 1993 onwards. Third, the variable
Part time is constructed as a dummy variable equal to one if the individual is
registered as working 30 h or less per week, zero if the registered number of weekly
hours is more than 30. Lastly, the Public sector variable used in Sect. 5 is based on
whether the individual is registered with a public sector employment code in the
Wage statistic. When we split the sample according to which sector individuals
worked in the year they became parents, we again use information from the
Employer/Employee register, going back to 1993.
Descriptive statistics for the samples and subsamples we use in the analyses in
Sects. 4–6 are given in Table 1.
4 Baseline results
Table 2, columns (1)–(5) shows the results from fixed effects regression on Eq. 1
for our main sample of women. Estimated on the whole sample together, having
children on average reduces women’s earnings by 1 % during the first years. This
estimate for the whole sample is smaller than in any of the subsamples.10 The effect
9
Using other official registries we find that about 20 % of women and 7 % of men who have a missing
wage observation when their first child is 3 years old have earnings below the basic amount (G) of the
Norwegian social security system, meaning that they are marginally or not employed at all that year (from
January 1 2010, G is NOK 72 881, approximately USD 12 500). About 50 % of the women and 75 % of
the men with missing wage observations are registered as employed. Some of the remaining missing
observations may indicate self-employment.
10
When the timing of children differs between educational groups, the estimate for the full sample will
typically be different from the average over the within-group estimates. Estimating the effect of children
relative to the counterfactual wage path for the full sample means individuals in the highest education
group are over-represented among the (still) childless, while individuals in the lower education group are
over-represented among those who have already had children. The same pattern is found in Anderson
et al. (2002).
123
13.30
38.24
Age of first child
Age
(8.40)
(9.92)
(1.20)
(37.00)
SD
41.71
18.20
1.90
123.02
Mean
139,037
N
10.96
Age of first child
(8.42)
(9.36)
(1.26)
(70.08)
(0.49)
(0.40)
(0.43)
(0.39)
(0.49)
(0.49)
(0.77)
(5.32)
40.67
14.27
1.60
145.59
28,075
0.49
0.72
0.61
0.78
0.52
0.52
0.80
10.90
Experience (years)
12.24
(5.94)
13.07
Men: observations in 2002, sample with first child born 1994–2007
37.89
Number of children
Age
160.31
1.41
Hourly wage (1998 NOK)
Men: observations in 2002
0.59
Public sector before children
0.82
Takes leave with first child
0.68
0.61
Public sector
0.80
0.42
Part time
Full time before children
0.75
Cumulated parental leave (years)
First child leave (years)
10.18
Experience (years)
(7.02)
(8.05)
(9.25)
(1.28)
(62.53)
(0.50)
(0.45)
(0.41)
(0.41)
(0.50)
(0.50)
(0.75)
(6.45)
(7.42)
(8.79)
(1.14)
(29.49)
SD
12.50
35.55
8.71
1.27
158.98
46,876
0.48
0.77
0.68
0.83
0.51
0.44
0.78
10.59
35.31
9.79
1.42
131.06
Mean
SD
(35.76)
(5.47)
(8.01)
(8.55)
(1.22)
(67.28)
(0.50)
(0.42)
(0.42)
(0.38)
(0.50)
(0.50)
(0.76)
(5.08)
(7.80)
(8.57)
(1.15)
11.18
36.63
8.48
1.28
181.07
60,722
0.73
0.85
0.70
0.82
0.74
0.36
0.71
9.59
35.85
8.78
1.35
148.06
Mean
(5.30)
(8.28)
(8.95)
(1.26)
(75.53)
(0.45)
(0.36)
(0.44)
(0.39)
(0.44)
(0.48)
(0.79)
(4.87)
(8.43)
(9.15)
(1.23)
(39.60)
SD
Lower deg.
Lower
Upper
Higher education
Secondary education
Women: observations in 2002, sample with first child born 1994–2007
133.31
1.61
Number of children
Mean
All
Hourly wage (1998 NOK)
Women: observations in 2002
Table 1 Descriptive statistics
11.08
37.57
9.26
1.45
221.34
3,364
0.47
0.95
0.74
0.85
0.47
0.21
0.78
9.21
34.08
5.94
1.16
183.79
Mean
Higher deg.
(5.57)
(8.19)
(8.81)
(1.30)
(86.48)
(0.50)
(0.21)
(0.44)
(0.36)
(0.50)
(0.41)
(0.83)
(4.59)
(7.34)
(7.78)
(1.18)
(63.15)
SD
Parenthood wage penalties in a double income society
123
123
0.32
137,198
Public sector before children
N
(0.47)
(0.28)
(0.08)
(0.50)
(0.46)
(0.27)
(0.12)
SD
37,825
0.23
0.90
0.03
0.47
0.22
0.10
0.05
Mean
(0.42)
(0.29)
(0.06)
(0.50)
(0.41)
(0.29)
(0.10)
SD
56313
0.25
0.93
0.04
0.57
0.24
0.07
0.06
Mean
SD
(0.12)
(0.43)
(0.25)
(0.07)
(0.50)
(0.43)
(0.25)
38767
0.50
0.90
0.05
0.60
0.50
0.08
0.07
Mean
(0.50)
(0.29)
(0.09)
(0.49)
(0.50)
(0.28)
(0.14)
SD
Lower deg.
Lower
Upper
Higher education
Secondary education
4,293
0.37
0.96
0.06
0.63
0.35
0.05
0.09
Mean
Higher deg.
(0.48)
(0.19)
(0.10)
(0.48)
(0.48)
(0.22)
(0.18)
SD
Sample is women and men born between 1950 and 1980 who are registered with employment in Statistics Norway’s Wage statistic. Working less than 30 h per week is
classified as part time employment, working 30 h or more is classified as fulltime
0.04
0.92
Full time before children
Takes leave with first child
First child leave (years)
0.31
0.55
Public sector
0.06
0.08
Part time
Mean
All
Cumulated parental leave (years)
Table 1 continued
S. Cools, M. Strøm
Age 1–5
Third child
Age [15
Age 11–15
Age 6–10
Age 1–5
Second child
Age [15
Age 11–15
Age 6–10
Age 1–5
First child
(0.0021)
(0.0012)
(0.0016)
0.0066***
0.010***
0.0024
(0.0020)
(0.0011)
(0.00087)
0.0087***
0.0089***
(0.0017)
0.013***
(0.0026)
(0.0014)
(0.00093)
0.016***
0.015***
0.0062***
(0.0024)
(0.0013)
0.012***
0.017***
0.016***
(0.0014)
(0.0021)
(0.0011)
0.0016
0.014***
0.013***
(0.00070)
(0.0017)
(0.00079)
0.0095***
0.012***
0.0098***
(0.0016)
0.011***
(0.0024)
0.026***
(0.0021)
0.029***
(0.0017)
0.025***
(0.0012)
0.019***
(0.0027)
0.053***
(0.0023)
0.051***
(0.0019)
0.044***
(0.0014)
0.033***
(0.0014)
0.018***
(0.0022)
0.034***
(0.0019)
0.031***
(0.0015)
0.027***
(0.0011)
0.021***
(0.0023)
0.046***
(0.0021)
0.042***
(0.0017)
0.035***
(0.0012)
0.025***
(0.0076)
0.041***
(0.013)
0.043***
(0.011)
0.040***
(0.0086)
0.035***
(0.0059)
0.037***
(0.014)
0.029**
(0.011)
0.042***
(0.0088)
0.045***
(0.0061)
0.049***
(0.0012)
0.0086***
(0.0017)
0.015***
(0.0015)
0.015***
(0.0012)
0.012***
(0.00094)
0.010***
(0.0018)
-0.0045**
(0.0016)
0.0015
(0.0014)
0.0027**
(0.0010)
0.0067***
(6)
Lower deg.
(4)
Lower
(2)
(1)
Higher deg.
(5)
All
Higher education
Secondary education
All
Upper
(3)
Men
Women
Table 2 The effect of children on women and men’s hourly wage
(0.0020)
0.0083***
(0.0027)
0.015***
(0.0025)
0.015***
(0.0021)
0.011***
(0.0017)
0.0090***
(0.0030)
0.00016
(0.0027)
0.0013
(0.0024)
0.0036
(0.0018)
0.0026
Lower
(7)
(0.0019)
0.0083***
(0.0030)
0.0023
(0.0025)
0.0046*
(0.0020)
0.0049**
(0.0015)
0.0040***
(0.0031)
0.0096***
(0.0026)
0.0086***
(0.0022)
0.0067***
(0.0016)
0.0028*
Upper
(8)
Secondary education
(0.0023)
0.00072
(0.0035)
0.0015
(0.0030)
0.00049
(0.0024)
0.00077
(0.0018)
0.0015
(0.0037)
0.020***
(0.0032)
0.017***
(0.0026)
0.013***
(0.0019)
0.0039**
Lower deg.
(9)
(0.0064)
0.0010
(0.011)
0.0036
(0.0089)
0.011
(0.0072)
0.0032
(0.0053)
0.0016
(0.011)
0.0043
(0.0095)
0.00059
(0.0079)
0.0043
(0.0058)
0.0017
Higher deg.
(10)
Higher education
Parenthood wage penalties in a double income society
123
123
(0.0022)
0.0070***
(0.0014)
3,11,879
170,462
1,039,057
(0.0029)
0.0067**
(0.0025)
0.016***
(0.0020)
0.018***
157,188
1,167,112
(0.0024)
0.030***
(0.0021)
0.027***
(0.0018)
0.025***
7,340
51,957
(0.017)
0.056***
(0.013)
0.055***
(0.011)
0.050***
567,321
3,325,989
(0.0020)
0.018***
(0.0018)
0.015***
(0.0015)
0.012***
256,282
1,377,709
(0.0031)
0.018***
(0.0028)
0.014***
(0.0025)
0.0098***
Lower
(7)
192,227
1,120,625
(0.0037)
0.014***
(0.0030)
0.011***
(0.0025)
0.0097***
Upper
(8)
Secondary education
106,054
736,930
(0.0041)
0.0013
(0.0036)
0.0019
(0.0030)
0.0016
Lower deg.
(9)
12,758
90,725
(0.012)
0.0010
(0.010)
0.0075
(0.0084)
0.0048
Higher deg.
(10)
Higher education
Each column provides FE estimates from a regression based on Eq. 1 on the sample of Norwegian women and men born between 1950 and 1980. The columns (2)–(5) and
(7)–(10) correspond to subsamples according to individuals’ length of education measured at age 25. Year dummies, a polynomial in age and individual fixed effects are
included in all specifications. So are controls for pregnancy, having a baby younger than 1 year and for having more than 3 children. Robust standard errors are in
parentheses. * p\0:10;** p\0:05; *** p\0:01
646,869
0.0034
(0.0013)
Individuals
(0.0021)
0.011***
1,927,527
0.0068***
(0.0011)
4,185,653
0.0070***
(0.0019)
0.013***
Observations
Age [15
Age 11–15
Age 6–10
(6)
Lower deg.
(4)
Lower
(2)
(1)
Higher deg.
(5)
All
Higher education
Secondary education
All
Upper
(3)
Men
Women
Table 2 continued
S. Cools, M. Strøm
(0.0039)
0.056***
(0.0019)
66,353
Experience
Age 11–13
Age 6–10
Age 1–5
First child
(0.0042)
(0.0020)
0.0047***
0.030***
0.059***
(0.0011)
(0.0034)
(0.0015)
0.0021***
0.027***
0.048***
(0.00059)
0.021***
(0.0026)
0.031***
(0.0011)
Panel B: including explanatory variables
257,691
0.027***
(0.0013)
Individuals
(0.0030)
0.046***
298,690
0.024***
(0.00090)
1,439,805
0.018***
(0.0021)
0.029***
Observations
Age 11–13
Age 6–10
Age 1–5
First child
Panel A: baseline
0.034***
(0.00099)
0.0028***
(0.0035)
0.050***
(0.0026)
0.048***
(0.0020)
(0.00098)
0.00045
(0.0030)
0.046***
(0.0022)
0.039***
(0.0016)
0.029***
94,454
91,566
0.038***
619,874
(0.0029)
0.045***
(0.0020)
0.037***
(0.0013)
0.027***
485,626
(0.0032)
0.046***
(0.0023)
0.044***
(0.0016)
(0.0066)
0.0070
(0.016)
0.042***
(0.011)
0.045***
(0.0079)
0.048***
5,318
35,615
(0.015)
0.047***
(0.0095)
0.049***
(0.0063)
0.052***
(0.0013)
0.0042***
(0.0024)
0.028***
(0.0017)
0.017***
(0.0012)
0.0051***
268,544
1,449,591
(0.0024)
0.029***
(0.0016)
0.018***
(0.0011)
0.0059***
(6)
Lower deg.
(4)
Lower
(2)
(1)
Higher deg.
(5)
All
Higher education
Secondary education
All
Upper
(3)
Men
Women
Table 3 The effect of children on women and men’s hourly wage
(0.0023)
0.00050
(0.0043)
0.0074*
(0.0031)
0.0035
(0.0022)
0.0019
88,101
402,689
(0.0042)
0.0070*
(0.0030)
0.0030
(0.0021)
0.0020
Lower
(7)
(0.0025)
0.00027
(0.0037)
0.019***
(0.0026)
0.012***
(0.0018)
0.0040**
110,893
595,752
(0.0037)
0.021***
(0.0026)
0.014***
(0.0017)
0.0066***
Upper
(8)
Secondary education
(0.0022)
0.0049**
(0.0047)
0.012***
(0.0032)
0.0096***
(0.0022)
0.0024
62,652
405,242
(0.0045)
0.016***
(0.0031)
0.013***
(0.0021)
0.0060***
Lower deg.
(9)
(0.0076)
0.011
(0.013)
0.013
(0.0089)
0.011
(0.0063)
0.0070
6,898
45,908
(0.013)
0.011
(0.0086)
0.0085
(0.0060)
0.0043
Higher deg.
(10)
Higher education
Parenthood wage penalties in a double income society
123
123
66,353
91,566
94,454
619,874
(0.0010)
0.020***
(0.00061)
0.013***
(0.0011)
5318
35615
(0.0040)
0.0031
(0.0032)
0.0056*
(0.0057)
0.0062
260,210
1,402,030
(0.00069)
0.0059***
85,578
390,462
(0.0015)
0.010***
(0.0025)
0.021***
0.0017
(0.0014)
(0.0061)
0.011*
(0.0027)
0.0098***
Lower
(7)
107,422
576,088
(0.0011)
0.00080
(0.0025)
0.016***
(0.0046)
0.032***
Upper
(8)
Secondary education
60,559
391,283
(0.0012)
0.0087***
(0.0021)
0.030***
(0.0044)
0.036***
Lower deg.
(9)
6,651
44,197
(0.0036)
0.017***
(0.0079)
0.034***
(0.012)
0.039***
Higher deg.
(10)
Higher education
Each column in each panel provides FE estimates from a regression based on Eq. 1 for the sample of individuals who had their first child in 1994 or later. Year dummies, a
polynomial in age, individual fixed effects and a set of dummies for having a second and a third child at various ages are included in all specifications. So are controls for
pregnancy, having a baby younger than one year and for having more than 3 children. In the lower panel additional controls are included. Robust standard errors are in
parentheses. * p\0:10 ;** p\0:05 ; *** p\0:01
Including controls for experience, parental leave, part time and sector of employment
257,691
Individuals
485,626
(0.00094)
(0.0012)
(0.00060)
298,690
0.00067
0.0066***
1,439,805
(0.00077)
(0.0010)
(0.00043)
0.023***
0.0055***
0.026***
0.017***
0.0029*
(0.0015)
0.0025**
0.0041**
(0.0020)
0.0030***
(0.00081)
(6)
Lower deg.
(4)
Lower
(2)
(1)
Higher deg.
(5)
All
Higher education
Secondary education
All
Upper
(3)
Men
Women
Observations
Public sector
Part time
Parental leave
Table 3 continued
S. Cools, M. Strøm
Parenthood wage penalties in a double income society
of having children is strongest for the highest education group, and weakest for the
lowest education group. During the first years after having children wages are
reduced by 4.9 % for those who have a higher university degree and by 1.2 % for
women who have lower secondary education.11 In all education groups except the
highest one, and in the sample as a whole, the effect of the first child grows over
time.
Restricting the sample to mothers only, as is done by Wilde et al. (2010), gives
more negative estimates in the sample as a whole, whereas the estimates in the
subsamples based on education length change only marginally. The results indicate
that women who do not have children are on a lower wage-path, yielding a smaller
estimated impact of children when the control group includes these women. If we
further restrict the sample to parents of at least two children, effects are again
somewhat larger in the aggregate, but the main pattern across subsamples and child
ages remains the same. The estimated effect of having a second child is stronger for
women in all education groups and at all ages of the second child. Lastly, splitting
the sample into subgroups based on education measured at age 22 instead of 25—
though yielding too few observations to get meaningful estimates in the group with
the highest education—the pattern across the other groups is still one of larger wage
penalties with higher education.12
Table 2, columns (6)–(10) gives the results on the wage penalties to fatherhood.
In the whole sample, the estimated effect is positive initially, at .67 %, and only
becomes negative as the child is older than 15, at .45. However, when we estimate
the effect on wages separately for the four education groups, there are negative
effects in the two middle education groups already at the outset, at .28 and .39 %
respectively. Also for men in these groups, the wage penalty to parenthood grows as
the child gets older, ending up at 1 % in the group with upper secondary education,
and 2 % for those with 4 years or less of higher education.
Restricting the sample to fathers only, the estimated effects of the first child tend
to become stronger, in both directions, but the general picture remains the same.
Further restricting the sample to fathers of at least two children, effects of having a
second child are estimated to be negative, indicating that fathers of two children are
positively selected with respect to wages. There are no significant differences when
we use the observed level of education at age 22 instead of 25.13
The statistically and economically significant wage penalty to fatherhood that we
find here contrasts with the findings of other studies on US data (Lundberg and Rose
2000, 2002), and also those by Simonsen and Skipper (2008) for Denmark. Wilde
et al. (2010), who restrict their sample to those who become fathers, also find small
negative effects on men’s wages of having children.
The estimated effects of further children are positive and statistically significant
for men in the two lower education groups—and in the sample as a whole. However,
11
For comparison, the average return to one more year of education for women in Norway in this period,
lies between 5 and 6 % (Hægeland and Kirkebøen 2007, p.14) and the gender wage gap in Norway lies
around 16 % (NOU 2008:6).
12
Results are available from the authors upon request.
13
Results are available from the authors upon request.
123
S. Cools, M. Strøm
when we restrict the sample of men to those who are eventually observed to have at
least two children, the estimated impact of the second child is negative.
5 Explaining the wage penalties to parenthood
We examine the determinants of the motherhood wage penalty in two different
ways. In this section, we investigate how changes in observable circumstances
explain the wage effect of children. Specifically, we study the respective roles of
work experience, parental leave, part time work and sector of employment.
As discussed in Sect. 3, the sample we use are those who have their first child
between 1993 and 2008. Panel A gives the baseline results from estimating Eq. 1 for
this sample. Results are very similar to those for the whole sample, given in Table 2.
In Panel B, the potential explanatory variables are included in the regression.
Comparing the two panels of Table 3 for women, we see that inclusion of these
variables does not significantly alter the coefficients. If anything, the estimated
impact of children is larger when they are included. We explored the results further
by including only one variable at the time. This showed that adding measures for
experience, maternity leave and sector of work does not alter the estimated impact
of having children on women’s wages. However, when a dummy for working less
than 30 h per week is added, the estimated wage penalty becomes greater. It is thus
the case that working part time, in itself positively associated with hourly wage in
all education groups (except in the highest one where there is no significant
association), is also positively correlated with having children. When the part time
dummy is omitted from the analysis, the effect of children on wages is smaller
because switching to part time work ameliorates the wage penalty to motherhood.
The economically insignificant role of experience, part time work and parental
leave in explaining the motherhood wage penalties point to the importance of labor
market policies, as these variables are found to explain a large part of the
motherhood wage penalty in other countries (Waldfogel 1997; Lundberg and Rose
2000; Budig and England 2001; Datta Gupta and Smith 2002; Anderson et al. 2002;
Wilde et al. 2010). A possible explanations for the small role of career breaks is that
policies of job protection during child related absence from work have an effect14. A
less optimistic explanation is that career breaks around birth are already accounted
for in the wage offer that the woman gets even before pregnancy. This is the
reasoning in Albrecht et al. (1999), who do not find a negative effect of child related
career breaks for Swedish women. It could also be that women have already chosen
jobs where the negative effects of a career break are small. If women select into
more child-friendly occupations even before they have children, some of the
motherhood wage effect is already realized. Adda et al. (2011) suggest such a prebirth wage penalty is substantial.
14
The positive effects of job protection on mothers’ wages may explain parts of the differences in the
size of motherhood wage penalties across countries (Sanchez-Marcos 2013).
123
Parenthood wage penalties in a double income society
Table 4 The effect of children on women’s hourly wage
All
Secondary education
Higher education
(1)
Lower
(2)
Lower deg.
(4)
Upper
(3)
Higher deg.
(5)
Panel A: no parental leave
First child
Age 1–5
Age 6–10
Age 11–13
0.026***
0.034***
0.033***
0.0099*
0.033
(0.0036)
(0.0074)
(0.0063)
(0.0052)
(0.038)
0.042***
0.040***
0.036***
0.017**
0.026
(0.0044)
(0.0088)
(0.0079)
(0.0067)
(0.050)
0.050***
0.039***
0.035***
0.026***
0.15**
(0.0057)
(0.011)
(0.010)
(0.0091)
(0.066)
Observations
245,061
63,128
73,200
103,659
5,074
Individuals
58,494
20,352
16,888
20,253
1,001
Panel B: short length parental leave
First child
Age 1–5
Age 6–10
Age 11–13
0.020***
0.0096***
0.027***
0.022***
0.047***
(0.0017)
(0.0036)
(0.0028)
(0.0025)
(0.011)
0.035***
0.013**
0.036***
0.031***
0.051***
(0.0025)
(0.0053)
(0.0041)
(0.0038)
(0.017)
0.042***
0.014*
0.038***
0.035***
0.023
(0.0035)
(0.0070)
(0.0057)
(0.0056)
(0.026)
Observations
397,366
89,395
140,172
158,203
9,596
Individuals
68,809
18,111
25,993
23,322
1,383
Panel C: middle length parental leave
First child
Age 1–5
Age 6–10
Age 11–13
0.029***
0.017***
0.037***
0.029***
0.043***
(0.0017)
(0.0038)
(0.0029)
(0.0026)
(0.013)
0.047***
0.021***
0.051***
0.040***
0.039*
(0.0026)
(0.0055)
(0.0044)
(0.0041)
(0.020)
0.059***
0.027***
0.054***
0.048***
0.019
(0.0036)
(0.0073)
(0.0060)
(0.0059)
(0.031)
Observations
351,706
75,906
125,978
142,738
7,084
Individuals
59,327
14,507
22,847
20,931
1,042
Panel D: long length parental leave
First child
Age 1–5
Age 6–10
Age 11–13
0.025***
0.015***
0.035***
0.021***
0.059***
(0.0020)
(0.0055)
(0.0037)
(0.0027)
(0.012)
0.040***
0.026***
0.046***
0.033***
0.057***
(0.0029)
(0.0075)
(0.0052)
(0.0040)
(0.017)
0.055***
0.041***
0.053***
0.046***
0.057**
(0.0040)
(0.010)
(0.0071)
(0.0056)
(0.025)
123
S. Cools, M. Strøm
Table 4 continued
All
Secondary education
Higher education
(1)
Lower
(2)
Lower deg.
(4)
Upper
(3)
Higher deg.
(5)
Observations
396,786
60,440
130,141
193,584
12,621
Individuals
62,649
11,366
22,853
26,721
1,709
Subsample analysis according to relative length of parental leave
Each column in each panel provides FE estimates from a regression based on Eq. 1 for the sample of
individuals who had their first child in 1994 or later. Year dummies, a polynomial in age, individual fixed
effects and controls for further children, being pregnant and having a baby younger than one year are
included in all specifications. Robust standard errors are in parentheses. * p\0:10; ** p\0:05;
*** p\0:01
In themselves, experience and part time work are generally positively associated
with the level of wages, whereas maternity leave hardly seems to matter. Public
sector work is negatively associated with wages in all education categories except
for the one with lower degree higher education, where the association is strong and
positive. If women choose to work in the public sector because they plan to have
children (and believe that a public sector job entails better job protection, greater
flexibility or lower attrition of skills during career breaks), the true effect of having
children on women’s wages is larger than our estimates, as some of the cost is taken
even before birth.
In Table 3, column (6)–(10), the corresponding results are given for men who had
their first child in 1993 or later. Panel A shows that in this sample the average effect
of children on wages for all men, irrespective of education level, is negative, at
.59 %. This sample thus differs significantly from both the baseline sample used in
Table 3 and a sample excluding childless men. This is reflected also in a stronger
point estimate for the wage penalty in the two middle education groups, at .66 and
.60 % respectively (columns (8) and (9)). The difference is due to the exclusion of
parents to children born before 1994. The results are consistent with the observation
that men increasingly share in the caring for children, consequently also sharing in
the wage effects.
When we include the potentially explanatory variables in the regression the
impact is significant, although most of the effect estimated for the whole sample is
left unexplained. The coefficient on having a first child aged between 1 and 5 years
is no longer significant for the group with higher education, lower degree (column
(9)), and for the group with upper secondary education it is cut in half. Also the
persistence of the effect, reflected in the coefficients on having a first child aged
6–10 years and 11–14 years, is reduced, although not by as much.
Including each of these variables separately shows that the wage penalty to
fatherhood hardly covaries with experience, part time work or sector of employment. It is the inclusion of parental leave that matters for the estimated effect on
men’s wages of having children. In the two groups where having children does have
a statistically significant impact on men’s wages, paternity leave explains a
substantial part of the effect.
123
Parenthood wage penalties in a double income society
The variation in parental leave in itself explains a great deal of men’s hourly
wages. For the three highest education groups (columns (8) to (10)), one year of
parental leave is associated with a 3–4 % reduction in wages—about the whole
wage penalty for women, who typically take somewhat less than a year of leave. We
are not able to determine to what extent this is due to the adverse effect of paternity
leave on human capital accumulation or whether the length of paternity leave
reflects men’s propensity to spend more effort at home relative to market work,
thereby reducing wages. The finding that parental leave is more strongly associated
with the wage effects of parenthood for men than for women is in accordance with
Albrecht et al. (1999)’s finding in their study on Swedish data. They find that career
breaks around birth are associated with negative wage effects only for men, not for
women. Their interpretation is that the negative wage effect stems from the
employer offering lower wages to women even before they have children because
they expect women to have a career break around birth. Employers do not expect the
same from men, and parental leave is therefore a stronger signal of the strength of
men’s commitment to market work. Another possible explanation that would give
the same empirical pattern, however, is that men to a larger extent than women
choose jobs where a career break has a larger effect.
Part time work is associated with higher wages for those with secondary
education (columns (7) and (8)) and with lower wages for those with higher
education (columns (9) and (10)). Public sector work is generally associated with
having lower wages, except in the group with lower degree higher education—like
for women.
6 Heterogeneous effects
In this section we investigate whether the wage penalty is stronger for some groups
than for others. We do this by running separate regressions on different subgroups;
on those who take relatively longer or shorter parental leave, on those who work full
time and part time, and on those who work in the public or in the private sector.
Individuals are sorted into groups based on the observed length of parental leave
taken with the first child, or according to sector or working time category the year
before their first child is born. As the variables we condition on are available from
1993 onwards, we limit the sample to individuals having their first child after 1993
(like in Sect. 5).
6.1 Parental leave
Probably the most direct labor market effect of having children is the period of
parental leave. In Tables 4 and 5, we investigate whether parenthood wage effects
are stronger for those who take relatively longer leave. We have divided the samples
in four: Those who take no, little, middle and long leave. The division into ‘‘little’’,
‘‘middle’’ and ‘‘long’’ is based on the assigned percentile in the distribution of all
parents of the same sex with children born in the same quarter of the year. The
123
S. Cools, M. Strøm
distribution is divided in three. For men, though, there is very little spread in the
number of leave days taken, and the middle group—containing less than 1 % of the
sample—has therefore been included in the group of men taking a short period of
leave.
Table 5 The effect of children on men’s hourly wage
All
Secondary education
Higher education
(1)
Lower
(2)
Upper
(3)
Lower deg.
(4)
Higher deg.
(5)
0.011
Panel A: No parental leave
First child
Age 1–5
Age 6–10
Age 11–13
0.015***
0.011***
0.0093***
-0.012***
(0.0019)
(0.0031)
(0.0030)
(0.0039)
(0.011)
0.032***
0.017***
0.019***
-0.023***
0.023
(0.0027)
(0.0044)
(0.0043)
(0.0054)
(0.016)
0.047***
0.024***
0.031***
-0.026***
0.0070
(0.0037)
(0.0060)
(0.0059)
(0.0077)
(0.023)
Observations
618,598
207,085
242,647
153,265
15,601
Individuals
127,953
50,982
48,601
25,839
2,531
0.00090
Panel B: short or middle length parental leave
First child
Age 1–5
Age 6–10
Age 11–13
0.0013
0.0044
0.0031
0.0014
(0.0016)
(0.0032)
(0.0024)
(0.0030)
(0.0096)
0.0076***
0.0077*
0.0097***
0.0041
0.0015
(0.0024)
(0.0046)
(0.0036)
(0.0044)
(0.014)
0.016***
0.0050
0.016***
0.0079
0.0094
(0.0035)
(0.0066)
(0.0052)
(0.0066)
(0.020)
Observations
651,320
161,248
286,487
184,084
19,501
Individuals
111,740
30,629
51,018
27,233
2,860
0.014***
0.0018
0.016***
0.014***
0.025**
(0.0032)
(0.0078)
(0.0057)
(0.0047)
(0.012)
0.020***
0.012
0.021**
0.018**
0.028
(0.0048)
(0.011)
(0.0083)
(0.0072)
(0.018)
0.019**
0.023
0.023*
0.017
0.000059
(0.0075)
(0.016)
(0.013)
(0.012)
(0.028)
Observations
132,112
22,129
46,954
53,934
9,095
Individuals
20,517
3,967
7,803
7,487
1,260
Panel C: long length parental leave
First child
Age 1–5
Age 6–10
Age 11–13
Subsample analysis according to relative length of parental leave
Each column in each panel provides FE estimates from a regression based on Eq. 1 for the sample of
individuals who had their first child in 1994 or later. Year dummies, a polynomial in age, individual fixed
effects and controls for further children, being pregnant and having a baby younger than one year are
included in all specifications. Robust standard errors are in parentheses. * p\0:10; ** p\0:05;
*** p\0:01
123
Parenthood wage penalties in a double income society
Among the parents who are not registered with taking any paid parental leave are
both individuals who have have not earned the right to paid parental leave and
individuals who may have earned the right but do not use it. It is very uncommon for
women not to use the right to maternity leave, hence the group of women taking no
leave can generally be thought of as consisting of women who did not earn the right.
For men, it is much more common not to take the paid leave granted them. Also,
men’s right to wage compensated paternity leave does not only depend on their own
earned right but also on the child’s mother having earned the right. Hence the group
of men taking no leave is more mixed, consisting of men who did not earn the right,
men whose spouse did not earn the right, and men who have the right to paid
parental leave but do not use it.
The results for women are given in Table 4. On average, there is not much
variation in the immediate effect of having children between the four groups, and
though there is some variation within each education group according to the relative
length of maternity leave taken with the first child, there is no clear pattern of
statistically significant difference between the groups. In the group of women with
the highest education (column (5)), taking longer leave, relative to taking no leave at
all, is clearly associated with a greater wage penalty to having children the first four
years after the wage compensated maternity leave period is over. For the two groups
with no more than secondary education (columns (2) and (3)), the wage penalty
during the first years is stronger for the group who does not take paid maternity leave,
and then it increases in the relative length of the leave period. All in all, the wage
penalty to motherhood is not very clearly linked to the length of maternity leave (as is
also found by Albrecht et al. (1999)). This is consistent with the insignificant role of
maternity leave in explaining motherhood wage penalties discussed in Sect. 5, and is
a further indication that in the Norwegian institutional context career breaks are less
important for how wages develop after having children.
The results for men are given in Table 5. For men, there is a U-shaped
relationship between the penalty and the propensity to take leave. The wage penalty
to fatherhood is greatest for the group of men taking no leave and the group of men
taking longer leave than what is usual. The average wage penalty in the two groups
of men is considerable, at about 1.5 %. As mentioned, the group of men taking no
leave is a diverse group, making the interpretation of the results for this group
harder to interpret. A stronger parenthood wage effect for men taking longer
parental leave is on the other hand in line with the findings of Albrecht et al. (1999).
6.2 Full time versus part time employment
Having children is associated with a reduction in working hours both for men and
women—but mainly for women (Cools and Strøm 2012). As documented in Sect. 5,
changing working time status to part time is what correlates most with a motherhood
wage penalty. Here we investigate whether wage penalties are smaller for those who
already work part time before having children. However, if part time work is chosen
in advance because the individual plans to have a child, this would cause an upward
bias in our estimates of the effect of having children.
123
S. Cools, M. Strøm
Table 6 The effect of children on hourly wage
All
Secondary education
Higher education
(1)
Lower
(2)
Upper
(3)
Lower deg.
(4)
Higher deg.
(5)
0.035***
0.020***
-0.039***
0.033***
-0.056***
(0.0011)
(0.0026)
(0.0019)
(0.0015)
(0.0067)
0.054***
0.030***
-0.052***
0.045***
-0.053***
(0.0016)
(0.0039)
(0.0028)
(0.0023)
(0.010)
0.067***
0.036***
-0.057***
0.056***
-0.039**
(0.0023)
(0.0052)
(0.0040)
(0.0033)
(0.016)
Observations
941,388
155,330
311,908
444,460
29,690
Individuals
152,488
29,385
55,171
63,681
4,251
0.020
Panel A: full time, women
First child
Age 1–5
Age 6–10
Age 11–13
Panel B: part time, women
First child
Age 1–5
Age 6–10
Age 11–13
0.014***
0.0055
0.020***
0.011***
(0.0024)
(0.0046)
(0.0038)
(0.0043)
(0.036)
0.022***
0.0070
-0.027***
0.012**
0.029
(0.0035)
(0.0065)
(0.0055)
(0.0061)
(0.051)
0.026***
0.012
-0.027***
0.0071
0.044
(0.0048)
(0.0086)
(0.0076)
(0.0085)
(0.067)
Observations
230,963
61,801
89,058
78,610
1,494
Individuals
42,296
12,490
16,703
12,852
251
0.00088
Panel C: full time, men
First child
Age 1–5
Age 6–10
Age 11–13
0.0084***
0.0021
0.0075***
0.0087***
(0.0013)
(0.0024)
(0.0019)
(0.0024)
(0.0065)
0.026***
0.0071**
0.019***
0.020***
0.0069
(0.0019)
(0.0035)
(0.0029)
(0.0035)
(0.0093)
0.040***
0.015***
0.027***
0.030***
0.014
(0.0027)
(0.0050)
(0.0041)
(0.0053)
(0.014)
Observations
1,083,899
275,061
467,116
304,117
37,605
Individuals
187,504
53,613
83,455
44,952
5,484
0.0089*
0.00051
0.0096
0.013*
0.021
(0.0046)
(0.0083)
(0.0078)
(0.0079)
(0.036)
0.023***
0.016
0.022*
0.034***
0.012
Panel D: Part time, men
First child
Age 1–5
Age 6–10
Age 11–13
123
(0.0065)
(0.012)
(0.011)
(0.011)
(0.055)
0.036***
0.023
0.028*
0.070***
0.032
(0.0092)
(0.017)
(0.016)
(0.015)
(0.080)
Parenthood wage penalties in a double income society
Table 6 continued
All
Secondary education
Higher education
(1)
Lower
(2)
Upper
(3)
Lower deg.
(4)
Higher deg.
(5)
Observations
95,634
28,386
33,631
32,228
1,389
Individuals
17,845
6,080
6,356
5,190
219
Subsample analysis according to being employed full time or part time the year before having children
Each column in each panel provides FE estimates from a regression based on Eq. 1 for the sample of
individuals who had their first child in 1994 or later. Year dummies, a polynomial in age, individual fixed
effects and controls for further children, being pregnant and having a baby younger than one year are
included in all specifications. Robust standard errors are in parentheses. * p\0:10; ** p\0:05;
*** p\0:01
The results for women are given in Panel A and B in Table 6. On average, the
negative wage effects of having children are two to three times stronger for the
women who were registered with full time employment the year before they had
children, compared to the women who had part time employment before having
children. Also, on average, there is a clear pattern of the wage penalty increasing
over time for the group of full time working women, reaching a wage reduction of
6.7 % when the first child turns 11 years old.
In the very small group of women with the highest education who worked part
time before having children, we see the first instance of a wage premium to having
children, reaching 7.8 % and significant at the 5 % level by the time the child turns
11 years old. This may indicate that for women who already started part time work
before having children, the extra effort put into child rearing does not intervene as
strongly with the effort put into work.15
Panel C and D in Table 6 give the corresponding results for men. Here, the
difference between the two groups is even more striking. The men who worked full
time before becoming fathers on average experience a wage penalty to fatherhood,
at .88 %. This is a 50 % larger penalty than the .51 % wage penalty estimated for
the whole sample given in Table 6. The penalty grows significantly over time, like
the case is for women who work full time before having children. The wage penalty
estimated for the sample of full time working men is on average stronger than for
the part time working women, especially as the first child grows older.
Consistent with the conjecture that part time workers face less of a trade-off in
effort between home and market work (they may thus even spend more effort at
work after they have children), for men who worked part time before becoming
fathers we find statistically significant wage premia to fatherhood when looking at
the whole sample (column (1) in Panel D) and in the two middle education groups.
The premium grows as the child gets older.
We also investigated the heterogeneity of effects according to sector of
employment and found that wage penalties are larger for both men and women
15
If part time work is chosen by women because they anticipate childbearing, the wage penalty may
have been realized even before the child is born, and will not be picked up by our estimates.
123
S. Cools, M. Strøm
working in the private sector prior to having the first child.16 The results are
consistent with the hypothesis that parenthood wage effects are due to reduced effort
at work. A slowing down of wages in a sector with large wage potential, has large
consequences for the individual.
7 Conclusion
In spite of the encouragement built into the labor market institutions of the Nordic
countries to combine work and family, the two remain in conflict, as is pointed out
by the wage penalties to parenthood documented in this paper. Like in most other
countries, women bear the greater share of the cost of having children also in
Norway. Wage penalties to motherhood are however smaller than what is observed
in many other countries, where the previous literature has found that career breaks
and loss of work experience explain a large share of the wage penalty. As such
measures of human capital do not seem to explain the family gap in Norway, these
findings indicate that family policies such as paid parental and job protection
contribute to reducing the adverse wage effects of parenthood.
We find indicative evidence that a reduced effort is more important than loss of
human capital in explaining the family wage gap. With a larger workload at home,
effort at work can be affected negatively. In groups where wages may be expected
to depend more on effort—higher educated women, women working full time and
in the private sector—we find larger wage penalties to motherhood.
Contrary to earlier studies using US data and to previous research from Norway,
we find negative (though comparatively small) wage penalties also to fatherhood,
consistent with men’s increasing involvement in child rearing. Contrary to the
findings for women, the length of parental leave can explain a large share of the
fatherhood wage penalty. Parental leave periods for fathers are generally short, and
the asymmetry in the importance of parental leave between men and women may
reflect that the length of parental leave serves as a signal of future effort at work,
rather than being an adverse human capital effect (Albrecht et al. 1999).
We conclude that in the Norwegian context, human capital is of little importance
in explaining the family gap in wages. It may as such serve as an explanation for
why the observed motherhood wage penalty is smaller in Norway than in many
other countries. Though suggesting that effort plays a comparatively more important
role, the evidence presented in this article is not conclusive. For a deeper
understanding of the mechanisms, further research is needed.
References
Adda, J., Dustmann, C., & Stevens, K. (2011). The career costs of children. Working paper.
Albrecht, J. W., Edin, P.-A., Sundstrøm, M., & Vroman, S. B. (1999). Career interruptions and
subsequent earnings: A reexamination using Swedish data. Journal of Human Resources, 34(2),
294–311.
16
Results are available from the authors upon request.
123
Parenthood wage penalties in a double income society
Anderson, D. J., Binder, M., & Krause, K. (2002). The motherhood wage penalty: Which mothers pay it
and why? American Economic Review, 92(2), 354–358.
Anderson, D. J., Binder, M., & Krause, K. (2003). The motherhood wage penalty revisited: Experience,
heterogeneity, work effort, and work-schedule flexibility. Industrial and Labor Relations Review,
56(2), 273–294.
Angrist, J. D., & Evans, W. N. (1998). Children and their parents’ labor supply: Evidence from exogenous
variation in family size. American Economic Review, 88(3), 450–77.
Antonovics, K., & Town, R. (2004). Are all the good men married? Uncovering the sources of the marital
wage premium. American Economic Review, 94(2), 317–321.
Astone, N., Dariotis, J., Sonenstein, F., Pleck, J., & Hynes, K. (2010). Men’s work efforts and the
transition to fatherhood. Journal of Family and Economic Issues, 31(1), 3–13.
Becker, G. S. (1985). Human capital, effort, and the sexual division of labor. Journal of Labor
Economics, 3(1), S33–58.
Bertrand, M., Goldin, C., & Katz, L. F. (2010). Dynamics of the gender gap for young professionals in the
financial and corporate sectors. American Economic Journal: Applied Economics, 2, 228–255.
Buckles, K. (2008). Understanding the returns to delayed childbearing for working women. The American
Economic Review, 98(2), 403–407.
Budig, M. J., & England, P. (2001). The wage penalty for motherhood. American Sociological Review,
66(2), 204–225.
Cools, S., & Strøm, M. (2012). The effect of children on labor supply and wage elasticities in the
household. Working paper.
Coverman, S. (1983). Gender, domestic labor time, and wage inequality. American Sociological Review,
48(5), 623–637.
Datta Gupta, N., & Smith, N. (2002). Children and career interruptions: The family gap in Denmark.
Economica, 69(276), 609–29.
Harkness, S., & Waldfogel, J. (2003a). The family gap in pay: Evidence from seven industrialized
countries. In Research in Labor Economics, 22, 369–414.
Harkness, S., & Waldfogel, J. (2003b). The family gap in pay: Evidence from seven industrialized
countries, volume 22 of worker well-being and public policy (Research in Labour Economics),
pp. 369–413. Emerald Group Publishing Limited.
Hersch, J., & Stratton, L. S. (1997). Housework, fixed effects, and wages of married workers. Journal of
Human Resources, 32(2), 285–307.
Hægeland, T., & Kirkebøen, L. J. (2007). Lønnsforskjeller mellom utdanningsgrupper. SSB Notater 36.
Hofferth, S. L. (1984). Long-term economic consequences for women of delayed childbearing and
reduced family size. Demography, 21(2), 141–155.
Korenman, S., & Neumark, D. (1991). Does marriage really make men more productive? The Journal of
Human Resources, 26(2), 282–307.
Korenman, S., & Neumark, D. (1992). Marriage, motherhood, and wages. Journal of Human Resources,
27(2), 233–255.
Lundberg, S., & Rose, E. (2000). Parenthood and the earnings of married men and women. Labour
Economics, 7(6), 689–710.
Lundberg, S., & Rose, E. (2002). The effects of sons and daughters on men’s labor supply and wages. The
Review of Economics and Statistics, 84(2), 251–268.
McAllister, I. (1990). Gender and the household division of labor. Work and Occupations, 17(1), 79–99.
Miller, A. (2011). The effects of motherhood timing on career path. Journal of Population Economics,
24(3), 1071–1100.
Millimet, D. L. (2000). The impact of children on wages, job tenure, and the division of household labour.
The Economic Journal, 110(462), C139–C157.
Mincer, J., & Polachek, S. (1974). Family investment in human capital: Earnings of women. Journal of
Political Economy, 82(2), S76–S108.
Nielsen, H. S., Simonsen, M., & Verner, M. (2004). Does the gap in family-friendly policies drive the
family gap? Scandinavian Journal of Economics, 106(4), 721–744.
Rosenzweig, M. R., & Wolpin, K. I. (1980). Life-cycle labor supply and fertility: Causal inferences from
household models. Journal of Political Economy, 88(2), 328–348.
Sanchez-Marcos, V. (2013). Female labour market outcomes and parental leave policies. Working paper.
Shelton, B. A., & Firestone, J. (1989). Household labor time and the gender gap in earnings. Gender and
Society, 3(1), 105–112.
123
S. Cools, M. Strøm
Simonsen, M., & Skipper, L. (2008). An empirical assesment of effects of parenthood on wages.
Advances in Econometrics, 21, 359–380.
Taniguchi, H. (1999). The timing of childbearing and women’s wages. Journal of Marriage and Family,
61(4), 1008–1019.
Viitanen, T. (2014). The motherhood wage gap in the uk over the life cycle. Review of Economics of the
Household (forthcoming).
Villund, O. (2005). Yrkesdata for selvstendig næringsdrivende. Dokumentasjonsnotat: Statsitics Norway
(SSB). 43.
Waldfogel, J. (1997). The effect of children on women’s wages. American Sociological Review, 62(2),
209–217.
Waldfogel, J. (1998a). The family gap for young women in the united states and britain: Can maternity
leave make a difference? Journal of Labor Economics, 16(3), 505–545.
Waldfogel, J. (1998b). Understanding the ‘‘family gap’’ in pay for women with children. The Journal of
Economic Perspectives, 12(1), 137–156.
Wilde, E. T., Batchelder, L., & Ellwood, D. T. (2010). The mommy track divides: The impact of
childbearing on wages of women of differing skill levels. Working paper 16582, National Bureau of
Economic Research.
123