Is There a Motherhood Wage Penalty for Highly Skilled Women?

Introduction and review of the literature

There is considerable evidence across countries that each additional child lowers a mother’s wage by approximately 5% (Waldfogel 1995; Harkness and Waldfogel 1999; Korenman and Neumark 1992; Lundberg and Rose 2000), a phenomenon known as the motherhood wage penalty.

There are a number of explanations generally offered for the motherhood wage penalty: (1) Women withdraw, at least temporarily, from the labor force to have children and therefore over time become increasingly less experienced relative to men (Hill 1979; Waldfogel 1998a; 1998b; Budig and England 2001; Beblo and Wolf 2000; 2002; Jacobsen and Levin 1995; Albrecht et al. 1999). (2) Women shift to less demanding jobs as a result of having children (Amuedo-Dorantes and Kimmel 2005; Waldfogel 1997; Bratti, Del Bono, and Vuri 2004; Nielsen, Simonsen, and Verner 2004; Simonsen and Skipper 2006; Felfe 2009). (3) Employers prefer not to hire women or they discriminate against them (Correll, Benard, and Paik 2007; Browne and Kennelly 1999). (4) Mothers invest less effort in their work and hence are less productive (Becker 1985; Hersch and Stratton 1997; Phipps, Burton, and Lethbridge 2001 who argue that women spend more time on housework and therefore have less time to invest in their careers). (5) Women who intend to have children may invest less in the accumulation of human capital (Mincer and Polachek 1974; Polachek 1981).

We empirically test for the motherhood wage penalty and some of the proposed explanations using a unique data set of all individuals who graduated from university or college in Israel during the period 1995–2008. This group is particularly interesting since college and university graduates are in general more highly motivated, they tend to be more career oriented and they have higher earning potential. We find that the labor force attachment of skilled Israeli women is comparable to that of Israeli men. Thus, during the first 10 years following graduation from university, both women and men in the sample reported an average of 22 months of absence from work. However, there is a large wage gap between skilled men and skilled women, which grows over time. Thus, the wage of men immediately following graduation is 59% higher than that of women and 10 years later is 90% higher. The purpose of this research is to determine how much of the growth in the wage gap can be attributed to the differential effect of having children on women and men.

We are not the first to examine the role of the motherhood wage penalty in explaining the wages of highly skilled women. Anderson, Binder, and Krause (2002) find that mothers who did not complete high school do not earn less than their childless counterparts, but that high school and college graduates earn about 10% less per child. Wood, Corcoran, and Courant (1993) find a wage gap in favor of males among a sample of graduates from the University of Michigan Law School. Gopta and Smith (2002), using a panel of 0.5% of the Danish population for the period 1980–1995, find that children do not seem to have a negative effect on their mother’s wages when experience is controlled for. Gangl and Ziefle (2009), using longitudinal data from the British Household Panel Survey, the German Socio-Economic Panel and the National Longitudinal Survey of Youth, find that if one takes account of the potential selectivity of women’s employment after childbirth the wage penalty for a mother is between about 10% and 18% per child in fixed-effects regression models. Staff and Mortimer (2012), using the Young Development Study sample, find that the hourly wages of mothers are approximately 3% lower (per child) than the wages of women without children, even after controlling for human capital variables and time-stable sources of unobserved heterogeneity. Accumulated months not in the labor force and not enrolled in school explain the residual pay gap between mothers and other women.

The benefits and rights of men and women after birth in Israel are similar to those in countries such as Britain, France and Italy, with certain differences in the length of maternity leave and its division between the father and mother. The rights of workers after birth in Israel are better than in the US but not as good as in Denmark. A description of the benefits in Israel and a comparison to those in other countries can be found in Appendix 7.

The sample we use has several advantages. It is composed of actual monthly salary data, thus making it possible to derive actual experience and non-employment periods, particularly non-employment due to maternity. We separate the experience effect into two components: the difference in the number of months of non-employment and the difference in the premium for non-employment between men and women. In addition, we focus not only on the effect of children on women’s labor force behavior and wages but also those of men. The data also include information on the graduates’ workplaces, thus making it possible to calculate the average wage for each workplace. Felfe (2012) argues that at least part of the motherhood wage penalty is due to the shift of mothers to more convenient jobs. We do not have a direct measure of the treatment of mothers by each employer and instead use the average wage levels in the workplaces, as well as sector (public vs private) as proxies. The fact that we are using a panel enables us to investigate the dynamics of labor force performance. Thus, we can control for the individual effect and are able to examine the effect of changes in workplace and of delaying birth. Finally, the data set includes a number of additional and relevant variables, such as the score on the psychometric exam (which is roughly equivalent to the SAT and measures an individual’s abilities), as well as undergraduate field of study (which is usually more specialized in Israel than in the US) and post-graduate degrees.

Differences in earnings between two groups may be explained by differences in the skills they bring to the labor market. Neal and Johnson (1996) find that after controlling for age and performance on the Armed Forces Qualifying Test (AFQT), the black-white earnings differential among young men is only about 7% and statistically insignificant. Thus, the results indicate that the black-white wage gap primarily reflects a skill gap. Lang and Manove (2011) argue that when controlling for AFQT and education blacks earn noticeably less than whites with the same education and cognitive score, and as a result, at least some of the black-white wage differential reflects differential treatment in the labor market. The contribution of this paper is that we control not only for education, like Gopta and Smith (2002), but also for the skill that women bring to the labor market (as represented by the psychometric exam score), as well as field of study and profession.

Data

The data include all individuals who graduated with a first degree from universities and colleges in Israel during the period 1995–2008. The data for each individual start from his year of graduation and end in 2008. Thus, for example, an individual who graduated in 1995 will have 14 annual observations while an individual who graduated in 2007 will have only one. The data were gathered from various sources by the Central Bureau of Statistics: the data on the individual’s academic education were obtained from the universities and colleges; the data on wages were obtained from the Income Tax Authority; and demographic information was obtained from the Population Registry.

The data include the following for each graduate and for each year following graduation:

1. Wage and employment: average monthly wage, number of months worked, number of workplaces, the sector in which the highest monthly wage was earned and the average wage for all workers at each workplace.

2. Demographic information: gender, country of origin, father’s country of origin, age, marital status, occurrence of a birth that year, number of children and place of residence.

3. Education: psychometric exam scores, type and name of the academic institution and field of study, both for a first degree and for advanced degrees.

Results

Figure 1 presents the average wage by number of years since graduation and gender. Men earn a higher monthly wage than women during the entire period. The average wage gap grows from 59% in the year of graduation to 90% after 13 years. This translates into an average annual increase of 2.3 percentage points in the wage gap. The increase is not linear: the wage gap increases at an average annual rate of 4 percentage points during the first 5 years after graduation, 1.9 percentage points for the next 5 years and remains unchanged during the last 3 years.

Figure 1:

Average monthly wage, by number of years since graduation and gender.

Tables 1–3 present our main estimation results, which were obtained using OLS (Table 1), Fixed Effect (Table 2) and Random Effect (Table 3) to control for unobserved heterogeneity. The dependent variable is log monthly wage ^[2] of individual i in year t. The explanatory variables include: (1) a dummy for gender and its interaction with accumulated experience (represented by number of years since graduation). Explaining this variable effect, that represents the widening wage gap between men and women, is the focus of this paper. (2) Number of children effect on men and women, which is our explanation to the widening wage gap. (3) Cumulative months of non-employment ^[3] (due to maternity leave or any other reason), employment in a low-paying firm or in the public sector, which we claim are three channels of the effect of children on the widening gap. (4) Control variables such as socioeconomic background, education and ability (see comment 4 in Tables 1–3).

The estimated raw wage gap between men and women, as can be seen from column 1 in Table 1, is 42.1% on graduation and grows by 2.8% per year. If background characteristics are controlled for (column 2), the estimated initial gap narrows to only 10.4%, while it has almost no effect on the estimated gap rate of increase. Although adding the number of children only mildly decreases the estimated initial gap, it significantly reduces the annual growth rate of the gap to 1.1%. Adding the effect of cumulative months of non-employment and dummies for employment in a low-paying firm or in the public sector reduces the annual gap growth rate to only 0.3%. Similar results, though not as pronounced, are obtained using the Fixed Effect and Random Effect methods. Thus, for both methods, controlling for individual fixed characteristics (column 1 and 2 in Tables 2 and 3) does not reduce the estimated growth rate of the wage gap over time and even increases it, while the variables related to childbearing eliminate a large portion of it. These results suggest that the increase in the wage gap during the period following graduation is not primarily due to individual characteristics but rather to maternity.

The work experience effect

Previous studies have shown that the accumulation of work experience is one of the main factors explaining the increase in wages over time. The most common claim is that women are absent from the labor market for longer periods as a result of giving birth and childrearing and therefore accumulate less work experience than men. Figure 2(a) and 2(b) present the cumulative months of absence from work over time by gender. The raw data presented in Figure 2(a) show that on average men and women are absent from work for similar periods. However, limiting the data to parents who gave birth during the sample period (Figure 2(b)) shows that cumulative absence from work grows at a slower rate for men than for women only after 6 years since graduation.

Figure 2:

Cumulative months of non-employment, by years since graduation and gender. (a) All (b) “new families”.

Appendix 4 presents the estimated children effect on the cumulative number of non-employment months. The results show that though women have, on average, lower cumulative months of non-employment, having children decreases cumulative non-employment for men and increases it for women, such that every child increases the gap in non-employment by 1.6 months (OLS) in favor of men (and by 3.5 months and 2.75 months for FE and RE, respectively). Tables 1–3 present the penalty for non-employment: according to the OLS estimates in column 4, each month of maternity leave decreases the wage of a woman by 1.0%, while absence for any other reason decreases a woman’s wage by 0.9% and that of a man by only 0.6%. The fixed effect and random effect models present an even higher relative penalty for maternity leave.

Experience appears to have a double negative effect on a woman’s wage: women accumulate more absence from work than men and they pay a larger premium for it. Nevertheless, this channel cannot entirely explain the widening of the wage gap and the motherhood penalty, since both are smaller yet significant in column 4 of Tables 1–3.

A shift from the private to the public sector

There is a large wage gap between the public and private sectors, particularly in the case of women. As can be seen in Table 1 (column 5), the OLS estimates of the public sector penalty are 15.4% for men and 18.8% for women. The fixed and random effect estimates are smaller but still statistically and economically significant.

Figure 3 presents the percentage of employees in the public sector over time and by gender. Thus, 28% of women work in the public sector following graduation, which increases to 46% after 13 years, while only 14% of men work in the public sector following graduation, a level which remains more or less constant over time. Therefore, the gender wage gap can be explained by the greater tendency of women to work in the public sector.

Figure 3:

Percentage of workers in the public sector, by years since graduation and gender.

Appendix 5 presents logit estimation results for the probability of working in the public sector. According to column 1, having children increases a woman’s probability of being employed in the public sector, but does not affect that of a man. Column 2 presents similar estimations for a sub-sample consisting of those who worked in the private sector 3 years after graduation. These results can be interpreted as the probability of moving from the private to the public sector, and indeed they show that having children increases the tendency of women to move to the public sector, while decreasing that of men.

A shift to lower-paying firms

We define a lower-paying firm as one that pays an average salary which is below the average for all firms in the sector (public or private) in a particular year. This definition allows us to measure the shift from high-paying to low-paying firms, which may explain the widening of the wage gap over time. Table 1 shows that men earn 50% less in a low-paying firm while women earn 42% less. The fixed effect estimate is 38% for men and 29% for women, which tends to confirm that this effect is meaningful and not only by construction.

Figure 4 presents the proportion of workers in low-paying firms by years since graduation and gender. It shows that some of the workers who start their career in a low-paying firm gradually move to a high-paying firm. This shift toward high-paying firms is stronger for men than women (9% of women and 15% of men).

Figure 4:

Percentage of workers in a low-paying firm, by years since graduation and gender.

Appendix 6 presents the estimated probability of working in a low-paying firm. According to column 1, having children increases the probability of working in a low-paying firm for both men and women, although the coefficient for women is significantly higher. Similar results are presented in column 2, where the sample is limited to those who worked in a high-paying firm 3 years after graduation.

In sum, having children leads women to shift from the private sector to the public sector and from high-paying to low-paying firms. Both result in a decrease in wages, which can be attributed to the motherhood penalty. Nevertheless, adding these variables to the regression (see column 5 in Tables 1–3) changes the estimated children effect only marginally, if at all.

Summary and conclusion

The key findings of this paper can be summarized as follows:

1. Men earned a higher monthly wage than women during the entire sample period. The average gender wage gap grew over time from 59% in the year following graduation to 90% after 13 years.

2. After controlling for individual characteristics, men earn 10.6% more than women on graduation, and the wage gap increases by 2.8% for each additional year in the labor force (OLS estimates). Accounting for individual unobserved time-invariant heterogeneity (FE and RE) has little effect on the growth of the gap over time. However, when the number of children is introduced the estimated wage gap grows by only 1.1% annually. Adding the effect of cumulative months of non-employment and dummies for employment in a low-paying firm and in the public sector (proxies for jobs that are better suited to parents) reduces the growth in the gap to only 0.3% annually. These results suggest that the increase in the wage gap over time is primarily due to maternity. Every additional child reduces a woman’s wage by 6.6% and increases that of a man by 3.4%, amounting to 10% per child (5.0% and 9.4% in the fixed effect and the random effect models, respectively).

3. Having children increases women’s detachment from the labor force (as measured by months of non-employment) while increasing men’s attachment to the labor force. Women pay a much higher premium for leaving the labor force than men. Thus, each month of non-employment due to maternity leave reduces a woman’s wage by 1.0% and non-employment due to any other reason reduces it by 0.9%, while a man’s wage is reduced by only 0.6%. According to the fixed effect estimate, the difference in the non-employment premium is even higher, such that the maternity premium is 0.9% while the premium for men’s non-employment is only 0.2%.

4. Once they have children, women tend to switch to a more flexible workplace, which often pays a lower wage. Thus, 25% of women work in the public sector on graduation, which rises to 45% after 13 years. Children increase a woman’s probability of moving to the public sector, and increase a man’s probability of leaving it. In addition, women tend much more than men to move to lower-paying firms after having children.

5. Even after controlling for time-invariant individual characteristics, career choice, non-employment and a shift to the public sector or a lower-paying job, the children effect is not fully explained. There are apparently additional factors that determine the children effect, such as a shift to part-time work or to a less demanding job in the workplace. However, we were unable to control for these factors due to the lack of data on hours worked and jobs.