THE RELATIONSHIP BETWEEN PAID MATERNITY LEAVE AND THE GENDER WAGE GAP: A COUNTRY-LEVEL ANALYSIS

Caroline Johnson

Abstract: While in recent decades, countries have progressed toward greater gender equality, an average gender wage gap of over 12% still exists in Organization for Economic Cooperation and Development (OECD) countries. Although paid maternity leave has been proposed as a policy approach for advancing pay equity, existing literature suggests that longer maternity leave entitlements are generally associated with wider gender wage gaps. I use country-level panel data on all 38 OECD countries through 2018 to investigate the relationship between paid maternity leave and the gender wage gap. This paper represents a geographic and temporal expansion on previous work because the most recent study in this area only used data for 22 OECD countries through 2010. I also contribute to the literature by focusing specifically on the length of paid, rather than employment-protected, maternity leave. In contrast to the existing scholarship on this topic, I find evidence of an inverted U-shaped relationship, in which relatively shorter leave entitlements are associated with widening gender wage gaps, while relatively longer leave policies are associated with narrowing wage gaps. However, this relationship is very small in magnitude, suggesting that maternity leave policies may not be an effective means of reducing the gender wage gap.

Introduction

Despite widespread gender equality policies, a gender wage gap — defined as the difference in the median earnings of full-time male and female workers as a percentage of the median earnings of full-time male workers — persists among Organization for Economic Cooperation and Development (OECD) countries (OECD 2021b). Among these countries, the most recent data suggests that the average wage gap is about 12%. There is significant cross-country variation in this metric: as of 2018, Luxembourg’s gap of -3.1% was the smallest, while South Korea’s gap of 34% was the largest (OECD 2021b) [1].

The gender wage gap negatively affects women, their families, and the societies of which they are a part. As more women have entered the workforce, their earnings have become increasingly closely related to household economic stability (Milli et al. 2017). A persistent gender wage gap driven by depressed female earnings is negatively related to household income, especially among families headed by single women. Milli et al. (2017) concluded that an estimated 60% of women and two-thirds of working single mothers would experience a pay increase if the gender wage gap was closed. The authors calculated that these increased earnings would halve the poverty rate among working single mothers. Pay inequality extends beyond a woman’s working life. Since women have lower lifetime earnings compared to men, women’s retirement savings tend to be lower, and they tend to receive less income from pensions and social insurance programs (Groom 2018; Gough 2001). 

The gender wage gap may also stifle economic growth. Milli et al. (2017) found that pay inequality is associated with human capital misallocation, in which women tend to sort into less productive careers than they would have in the absence of a wage gap. The authors further estimated that in the United States, equal pay to women would increase the dollars flowing into the economy by over $500 billion annually. Similarly, Wodon and de la Brierre (2018) calculated that in 2014, high and upper-middle-income countries experienced a combined $140.2 trillion loss in expected future earnings related to the gender wage gap. 

The gender wage gap can be decomposed into explained and unexplained components. The explained component consists of explanatory factors such as occupational segregation, differences between men’s and women’s hours worked (Blau and Kahn 2017), and the fact that women tend to enter and exit the workforce more frequently than men (OECD 2018). Some scholars theorize that unexplained gender pay differences are linked to discrimination (Goldin 2014); others suggest that, compared to men, women may have different attitudes and preferences about work (Bensidoun & Trancart, 2018); and still others contend that employers set wages with the expectation that women will experience more career interruptions (Amano-Patino et al. 2020). 

In an analysis of the gender wage gap in the United States over the last forty years, Blau and Kahn (2017) concluded that a significant reduction in the unexplained component of the gender wage gap coincided with a narrowing of the overall gap in the 1980s. However, they found that the size of this component has remained stable in subsequent decades. The authors also found evidence of a reduction in the explained component between 1980 and 2010, which they attributed to a convergence of human capital characteristics: relative to men, women’s level of educational attainment and labor market experience increased. Thus, the authors concluded that the overall gender gap decreased in recent decades, but that during this period, the unexplained component accounted for a larger share of the total, increasing from 71% of the gap in 1980 to 85% in 2010. 

Paid maternity leave has been proposed as a way of reducing the gender wage gap. In 2013, the OECD recommended that member countries adopt paid maternity leave policies to facilitate female participation in the workforce (Recommendation of the Council on Gender Equality 2013) [2]. The authors of the recommendation reasoned that these entitlements might strengthen women’s attachment to their employers and reduce their time spent out of work, which could help boost their earnings. However, the existing literature generally concludes that longer maternity leave entitlements are associated with wider gender wage gaps (Thévenon and Solaz 2012; Christofides et al. 2013; Cukrowska-Torzewska and Lovasz 2020). Thus, while these policies are important to overall gender equality, they may not be an appropriate tool to reduce wage inequality. 

Using OECD data on the gender wage gap and the length of paid maternity leave, the present study examines the relationship between the generosity of paid maternity leave and the size of the gender wage gap in OECD countries from 2000 to 2018. Because my data encompass all 38 OECD countries through 2018, this analysis represents a geographic and temporal expansion of previous studies, the most recent of which only used data for 22 OECD countries through 2010. Additionally, while most previous studies focus on employment-protected maternity leave, this analysis primarily focuses on paid leave, which may have a different relationship to the gender wage gap than protected leave. 

My results suggest that there is an inverted U-shaped relationship, in which relatively shorter leave entitlements are associated with widening gender wage gaps, while relatively longer leave policies are associated with narrowing wage gaps. However, this relationship is very small in magnitude. Thus, while paid leave policies are important for a myriad of reasons, including giving women the flexibility around childbirth necessary to balance work and family, these policies alone are unlikely to be effective tools for reducing the gender wage gap.

Background

Of the 38 OECD member countries, the United States is the only one that does not guarantee paid maternity leave at the national level (OECD 2019c) [3]. Maternity leave is a type of parental leave that guarantees new parents a certain amount of time off from work prior to or following the birth of a child [4]. While the terms are often used interchangeably, maternity leave is an entitlement that is allocated to mothers around the time of birth; paternity leave is an entitlement allocated to fathers around the time of birth; and parental leave is an entitlement that can be used by one or both parents after they have used their available maternity or paternity leave (OECD 2019c) [5]. 

Parental leave policies can be disaggregated along two dimensions: payment and protection. In terms of the first of these two dimensions, leave can be paid or unpaid. This paper focuses on paid maternity leave. All countries in the OECD, with the exception of the United States, provide some level of wage replacement during maternity leave [6]. The level of income support varies by country. Within countries, the wage replacement rate may also depend on the individual’s income before the birth of their child. As of 2014, the wage replacement rate for paid leave in OECD countries ranged from 30% to 100% of average national earnings (OECD 2017) and averaged 75% for maternity leave and 45% for parental leave (Adema et al. 2015). Typically, these benefits are funded through social insurance systems and social security contributions from employees and employers (Olivetti and Petrongolo 2017). 

The second dimension relates to whether mothers’ leave is “protected.” While leave may be paid or unpaid, all parental leave entitlements in the OECD are employment-protected, which means that employers are prohibited from dismissing employees during periods of leave. Such provisions guarantee that individuals can resume the same (or equivalent) job that they held prior to taking leave (Adema et al. 2015). 

Maternity leave policies first emerged in Germany in the nineteenth century and were gradually implemented in other industrialized countries throughout the twentieth century (Kamerman and Moss 2011). Early iterations of these policies tended to take the form of mandatory bans on female work soon after giving birth. These bans were rarely accompanied by job protection or income replacement, which became the norm in the twentieth century (Olivetti and Petrongolo 2017). As female labor force participation grew, the shape of parental leave changed as well. In the 1960s, these policies expanded in both length and generosity of income replacement to allow women to balance their families and careers (Olivetti and Petrongolo 2017).

Though the OECD’s 2013 Recommendation on Gender Equality in Education, Employment, and Entrepreneurship urged member countries to adopt paid parental leave policies, there is no OECD-wide policy on maternity leave. Nonetheless, in recent decades, consistent with the push towards gender equality, parental leave entitlements in OECD countries have expanded greatly. Entitlements have become increasingly ubiquitous and generous, with their average length increasing from 17 weeks in 1970 to 51 weeks in 2018 (OECD 2019c). Figure 1 displays the distribution of paid maternity leave across the OECD as of 2018. Estonia currently has the most generous policy, guaranteeing 166 weeks of paid leave for mothers (OECD 2019b). Aside from the United States, which guarantees no paid leave, Mexico is the only country that fails to meet the International Labour Organization’s minimum guideline of 14 weeks (Maternity Protection Convention 2000) [7]. More recently, many countries have introduced leave for fathers, either in the form of paternity leave or parental leave which can be shared among both parents. While in 1970 only three countries had reserved any leave for fathers, by 2018, 36 OECD members offered some form of paid father-specific leave (OECD 2019b) [8]. 

Paid parental leave is associated with economic and health benefits for mothers and families. Studies have found that paid maternity leave is related to a reduction in the infant mortality rate (Patton et al. 2017), an increase in the average length that mothers spend breastfeeding (Baker and Milligan 2008b), and a decrease in mothers’ postpartum depression symptoms (Chatterji and Markowitz 2008). Paid parental leave is also linked to higher female labor force participation (Thévenon and Solaz 2012) and an increased likelihood that women return to work after birth (Baker and Milligan 2008a), as discussed in more detail in the next section.

Literature Review

A sizable body of literature examines the relationship between parental leave and female labor force outcomes. Some papers focus specifically on the association between leave and the gender wage gap. Generally, scholars found that parental leave is associated with greater female labor force participation, though there are competing findings regarding the correlation with wages. Evidence suggests that the length of leave has a positive or U-shaped relationship with the gender wage gap. However, there is no clear consensus on the mechanism through which this relationship operates, particularly in terms of the relative importance of labor supply and demand factors.

Parental Leave, Women’s Labor Force Participation, and Women’s Wages

An extensive literature focuses on the link between parental leave and female labor force outcomes. These outcomes, particularly labor force participation and wages, are important contributors to the size of the gender wage gap. The gap may shift depending on the extent to which parental leave policies affect female and male wages differently or change the profile of women in the labor force. Studies have generally found that longer parental leave is associated with increased female labor force participation. However, the evidence on wages is mixed.

Ruhm (1998), Thévenon and Solaz (2012), Akgunduz and Plantenga (2013), and Del Rey et al. (2021) found evidence of a positive relationship between parental leave and female labor force participation. However, Thévenon and Solaz (2012), Akgunduz and Plantenga (2013), and Del Rey et al. (2021) concluded that this relationship had an inverted U-shape, in which increased parental leave entitlements were associated with higher levels of participation at relatively low levels of leave, but with reduced participation at higher levels of leave. Conversely, Schönberg and Ludsteck (2014) found that expansions in Germany’s maternity leave policy were associated with lower short-term maternal employment, but that there was no relationship over the long term [9].

There is competing evidence on the link between parental leave and women’s wages. Lalive and Zweimüller (2009) and Akgunduz and Plantenga (2013) found evidence of a negative relationship between the length of parental leave and female wages [10]. However, Waldfogel (1998) and Rossin-Slater et al. (2013) found evidence of a positive relationship between these variables. Likewise, Houser and Vartanian (2012) found that women who took paid leave were more likely than women who did not take paid leave to receive a wage increase in the year after their child’s birth.

Parental Leave and the Gender Wage Gap

The studies described above tend to find that parental leave policies are associated with increased female labor force participation, while the evidence on wages is mixed. Given these findings, it is unsurprising that parental leave is associated with the size of the gender wage gap, which is influenced by both factors.

Using data from ten OECD countries between 1970 and 2010, Thévenon and Solaz (2012) found that longer employment-protected parental leave entitlements were correlated with wider gender wage gaps. Similarly, in a pair of analyses using survey data from 26 European countries, Cukrowska-Torzewska and Lovasz (2020) concluded that the size of the gender wage gap was positively correlated with parental leave length, while Christofides et al. (2013) found that longer maternity leave entitlements were associated with an increase in the size of the unexplained component of the gender wage gap.

Some studies identified a nonlinear relationship between parental leave and the gender wage gap. Using data on nine European countries from 1969 to 1993, Ruhm (1998) found that shorter paid entitlements (three months or less) were not significantly related to the relative wages of women and men, but that longer leave policies (nine months or more) were associated with a reduction in women’s relative wages [11]. Somewhat relatedly, Olivetti and Petrongolo (2017) analyzed the relationship between several family-friendly policies and female labor market outcomes in 22 OECD countries from 1970 to 2010. The authors found evidence of a U-shaped relationship between the availability of employment-protected parental leave and the gender wage gap: while shorter leave entitlements were associated with a reduction in the gender wage gap, the direction of the relationship reversed as entitlements became more generous. However, this relationship was only statistically significant when they controlled for the share of leave that is paid, the average wage replacement rate, and the level of early childhood spending [12].

Potential Mechanisms Linking Parental Leave and the Gender Wage Gap

Various supply and demand-side mechanisms have been proposed to explain the link between parental leave and female labor force outcomes, which subsequently affect the gender wage gap. On the demand side, employers may set wages and calibrate their demand for female workers by considering the costs and benefits of parental leave to the firm. The costs include reductions in productivity (Amano-Patino et al. 2020), the requirement to find replacements for an employee who takes leave, and the need to compensate for employee skill deterioration after leave (Ruhm 1998). As a result, firms may have a reduced demand for (Olivetti and Petrongolo 2017) and therefore offer lower wages to (Amano-Patino et al. 2020) workers who are expected to take parental leave. These dynamics can lead to discrimination in hiring (Blau and Kahn 2013) or to slower promotion of women of childbearing age (Thévenon and Solaz 2012; Christofides et al. 2013). Because employer expectations regarding parental leave are focused on women, these factors could exacerbate the gender wage gap [13].

However, employers do realize some benefits from parental leave. For example, parental leave entitlements are associated with a greater likelihood that employees return to their pre-birth employer (Baker and Milligan 2008a) and a reduced likelihood that women leave the workforce after birth (Olivetti and Petrongolo 2017). Employers could benefit from a return on their human capital investment, potentially resulting in increased demand for female labor (Klerman and Leibowitz 1994). These factors could also help mitigate the adverse impact on the wage gap described above.

On the supply side, maternity leave can foster attachment to the workforce during periods of temporary career interruption (Olivetti and Petrongolo 2017). In the presence of parental leave, female labor supply may increase. Additionally, women may be induced to enter the workforce before they have a child to qualify for parental leave entitlements (Blau and Kahn 2013). However, the wage gap could be affected if the women who decide to enter the workforce, who qualify for leave (or who remain in the workforce because of leave entitlements) differ from the women who would be in the workforce regardless of occupation or educational attainment. Alternatively, longer parental leave entitlements may encourage women to remain out of the workforce for longer periods, which could impact their ability to return to their pre-birth career track and earnings, thus increasing the gender wage gap (Olivetti and Petrongolo 2017; Blau and Kahn 2013).

Some scholars have taken into consideration these competing supply and demand effects. For example, Ruhm (1998) proposed that parental leave entitlements result in only a small decrease in the demand for labor, since governments—not employers—typically bear most of the cost. The author also posited that these policies would lead to a large increase in labor supply as women enter the workforce to qualify for paid leave entitlements. The large increase in supply and the small decrease in demand theoretically increase women’s employment while decreasing their wages. However, Olivetti and Petrongolo (2017) argued that the impact of parental leave on the gender wage gap is ambiguous and instead depends on both the relative wage elasticities of labor supply and demand and the size of any shifts in labor supply and demand.

Present Study

Existing literature on the link between parental leave entitlements and the gender wage gap generally concludes that longer entitlements are associated with increases in the wage gap (Ruhm 1998; Thévenon and Solaz 2012; Christofides et al. 2013; Olivetti and Petrongolo 2017; Cukrowska-Torzewska and Lovasz 2020). The present study will extend this literature in two ways. First, while several scholars have studied the gender wage gap in the OECD, the most recent study only used data for 22 countries through 2010 (Olivetti and Petrongolo 2017). My period of analysis extends through 2018, and I expand my geographic coverage to all 38 current OECD member countries. These enhancements to my data provide broader and more timely evidence as to the relationship of interest. Second, only one OECD-focused study (Thévenon and Solaz 2012) focused on paid, rather than employment-protected, leave. Given that paid leave may have a stronger relationship with women’s labor decisions, it is my primary focus.

Conceptual Framework

While the literature generally suggests that longer parental leave is associated with wider gender wage gaps (Thévenon and Solaz 2012; Christofides et al. 2013; Cukrowska-Torzewska and Lovasz 2020), several studies have found that this relationship varies depending on the length of leave (Ruhm 1998; Olivetti and Petrongolo 2017). Based on these findings, I hypothesize that there is a U-shaped relationship between the generosity of paid maternity leave and the size of the gender wage gap. That is, I hypothesize that initial increases in the length of paid leave are correlated with a reduction in the size of the wage gap, but that beyond a certain point, lengthier entitlements are associated with widening the wage gap. This scenario could be driven by a combination of female labor supply and demand factors. At relatively low levels of leave, increased leave may lead to higher levels of workforce attachment among women (Baker and Milligan 2008a; Olivetti and Petrongolo 2017), which increases both their value to employers and their average length of work experience. Both factors could increase median female wages, reducing the gender wage gap. However, as the duration of leave increases, the costs associated with taking leave, such as skill deterioration and the cost to employers of hiring replacements for leave-takers (Ruhm 1998), may reduce female wages, widening the wage gap.

While I expect to find a U-shaped relationship, it is possible that the relationship instead assumes an inverted U-shape. This dynamic could result from selection effects, in which the gender wage gap changes as women systematically enter or exit the workforce in response to changes in maternity leave policies. If the women who enter the workforce at relatively low levels of leave are less experienced than the women presently in the workforce, they may drive down median female wages, increasing the gender wage gap. As leave gets longer, women may exit the workforce because of personal preferences or unfavorable employment opportunities. If the women who exit are less productive than those who remain, median female wages may subsequently increase, decreasing the gender wage gap.

In keeping with similar studies, my model controls for country-level demographic, economic, policy, and political characteristics that could be related to the size of the gender wage gap or to the generosity of paid maternity leave.  These factors are outlined in Figure 2 and described in more detail in the following sections.

Demographic Characteristics

Demographic characteristics, including those related to population composition, fertility, and female human capital, are included in the model because of their potential relationship with the gender wage gap or with the cost of providing paid maternity leave. Consistent with Ruhm’s (1998) analysis of the link between maternity leave and the gender wage gap in European countries, I control for the fertility rate. The size of the working-age female population (15-64 years) and the percentage of the female population that is of childbearing age are also included, since these factors may be associated with the cost of providing paid maternity leave (Svaleryd 2009).

Blau and Kahn (2017) identified a positive relationship between the difference in the education levels of men and women and the size of the explainable component of the gender wage gap. To account for this dynamic, I control for the difference in the percentage of the male and female populations with a college degree (i.e., the “college gap”). The authors also found evidence of a negative relationship between women’s years of labor force experience and the size of the explainable component of the wage gap. Because data on average years in the workforce are not available, I proxy for this factor by controlling for women’s average age at childbirth, which captures a portion of the variation in the average amount of work experience prior to a potential career interruption.

Economic Characteristics

Several country-level economic characteristics may be associated with the size of the gender wage gap or with the provision of paid maternity leave. To account for exogenous economic events that could be associated with the wage gap and/or with changes in leave entitlements (Thévenon and Solaz 2012), I control for annual gross domestic product (GDP) growth. My model also includes GDP per capita, which Lambert (2008) found to have a positive relationship with the length of maternity leave. Further, because unionization has been found to be related to reduced employee wage variance (Blau and Kahn 1996), and since such wage compression could reduce the size of the gender wage gap, I control for union density. Consistent with Ruhm (1998), I also include the female unemployment rate in my model.          

Blau and Kahn (2017) found that occupational segregation, in which men and women tend to sort into different occupations or industries, was related to the size of the gender wage gap. To capture this dynamic, I control for the difference in the proportion of men and women who are employed in the services sector (i.e., the “services employment gap”). I create a comparable measure that reflects the gender difference in industrial sector employment (i.e., the “industrial employment gap”). Finally, since Matteazzi et al. (2018) found evidence of a positive correlation between the prevalence of women engaged in part-time work and the size of the gender wage gap in European countries, I control for the percentage of women who are working part-time.

Policy and Political Characteristics

This study focuses primarily on the relationship between paid maternity leave and the gender wage gap. However, there are other family policies that may be related to the size of the gender wage gap, and certain political characteristics could be associated with the provision of paid maternity leave. In keeping with Olivetti and Petrongolo (2017), who found evidence of a negative correlation between the percentage of GDP spent on childcare and the size of the gender wage gap, I control for the same measure. Because Andersen (2018) found that paternity leave was associated with a smaller in-family gender wage gap, I also control for the length of a country’s father-specific leave entitlement. Finally, I control for the share of members of the national legislature who are female, which Kittilson (2008) and Rolfes-Haase (2017) found to be positively associated with maternity leave duration.

Data and Methods

My analyses draw on country-level data sourced from the OECD and the World Bank. My panel dataset spans the period from 2000-2018 (inclusive), with a total of 483 country-year observations from all 38 OECD countries [14]. Data for my independent variable, the length of paid maternity leave, were obtained from the OECD Family Database (OECD, 2021c). In keeping with the literature (Ruhm 1998; Thévenon and Solaz 2012; Olivetti and Petrongolo 2017), I measure this variable in terms of weeks [15]. While some studies account for all employment-protected leave, I follow the practice of Ruhm (1998) and Thévenon and Solaz (2012) by limiting my focus to paid leave, calculated as the total number of weeks for which a woman can access paid time off from work around or after birth using either maternity or parental leave [16].

Data for my dependent variable, the gender wage gap, were obtained from the OECD Employment Database (OECD 2021b). The gender wage gap is calculated as the difference between the median earnings of full-time male and female workers as a percentage of the median earnings of full-time male workers (OECD 2021b). In addition, as described above, I include control variables falling into three broad categories: demographic, economic, and policy and political. Data for these controls were sourced from the World Bank’s World Development Indicators Database (World Bank 2021) or from one of four OECD databases: the OECD Education Database (OECD 2021a), the OECD Employment Database (OECD 2021b), the OECD Family Database (OECD 2021c), and the OECD Social Expenditures Database (OECD 2021d).

To analyze the relationship between paid maternity leave and the gender wage gap, I estimate a two-way fixed effects regression model. Both country and year fixed effects are included in the regression to reduce bias in my estimates. Country fixed effects control for all time-invariant differences between countries, such as a country’s history of social spending and gender norms. They also partially account for characteristics that may gradually change over time, but have a fixed component, such as cultural attitudes towards women. Time fixed effects control for characteristics that change over time, but are constant across the countries of study, such as global recessions or broad-based changes in gender norms.

The inclusion of fixed effects in my model is consistent with the approaches of other authors, including Ruhm (1998), Thévenon and Solaz (2012), and Olivetti and Petrongolo (2017). In keeping with these studies, I also include a squared maternity leave term in some regressions to account for the possibility of a nonlinear relationship between the length of paid maternity leave and the size of the gender wage gap. Finally, in some specifications, consistent with Ruhm (1998) and Thévenon and Solaz (2012), I include linear country-specific time trends to control for unobserved country-level characteristics that change linearly, but at different rates in different countries.

My base model, with the country-year as the unit of analysis, is specified as follows:

  GenderWageGapit= β0 + β1MatLeaveit + δit + θit + πit + αi + λt+ εit

where i is a country index, t is a year index, δit is a list of demographic characteristics, θit is a list of economic characteristics, πit is a list of policy and political characteristics, αi represents country fixed effects, λt represents time fixed effects, and εit is the error term. My coefficient of interest is β1, which captures the relationship between the length of paid maternity leave and the size of the gender wage gap. Table 1 contains a description of all variables to be included in my model.

Descriptive Statistics

Table 2 presents descriptive statistics for my dependent, key independent, and control variables. All estimates are weighted by the average size of the country-specific female population of working age (15-64 years) over my period of analysis. The analytic sample comprises 483 country-year observations across 38 OECD countries and 19 years [17]. In this sample, the average gender wage gap was 19.1%, with a maximum of 41.7% (South Korea in 2000) and a minimum of -3.1% (Luxembourg in 2018). The gender gap has changed considerably over time: while the sample average was 19.7% in 2000, it decreased to 12.2% in 2018 (not shown in the table).

Within my sample, the average length of paid maternity leave is 31.4 weeks, or roughly eight months. Both over time and across countries, there is substantial variation in the generosity of leave entitlements. Several countries had no paid maternity leave entitlement for at least part of the 19-year period of study, including Australia, New Zealand, and the United States (not shown in the table). The most generous entitlement was 214 weeks (over four years) in the Czech Republic between 2000 and 2007. Of the 38 countries in my analytic sample, 23 (61%) experienced a change in the length of paid maternity leave between 2000 and 2018.

My demographic, economic, and policy and political controls also vary across country-years. Particularly relevant to the size of the gender wage gap are the college gap, union density, and the share of women employed part-time. The college gap—the difference between the share of men and women with a college education—ranges from a minimum of -20.3% in Estonia in 2018 to a maximum of 19.5% in Switzerland in 2000, with an average of -0.8%. Union density—the share of the employed population that are union members—averages 17.8%, with a low of 5.3% (Estonia in 2014) and a high of 93.3% (Iceland in 2002). The share of women who are employed part-time displays a similar level of variation, averaging 25.1% with a minimum of 2.3% in the Slovak Republic in 2002 and a maximum of 60.6% in the Netherlands in 2014.

Table 3 reports differences in the characteristics of country-years that are below and above the within-sample median length of paid maternity leave (47.7 weeks). Country-years with at least 47.7 weeks are characterized as having “long” entitlements, while country-years with fewer than 47.7 weeks are characterized as having “short” entitlements. As shown in the table, most differences between these two groups are significant at the 1% level. Notably, countries with longer paid maternity leave entitlements have significantly higher gender wage gaps (an average of 23.1% compared to 16.8%). Additionally, countries that offer longer leaves, on average, have a greater proportion of men than women with a college education, a greater share of the employed population that are union members, a greater proportion of women employed part-time, and longer paid father-specific leave entitlements. The significant differences shown in Table 3 reinforce the importance of controlling for these factors in my regression analyses, the results of which are reported in the next section.

Results

Tables 4 and 5 report my regression results. All regressions are weighted by the average size of the country’s working-aged female population (15-64 years) over my period of analysis, and robust standard errors clustered at the country level are reported in parentheses beneath each coefficient. In Table 4, Model (1) is a bivariate regression of the gender wage gap on the length of paid maternity leave. Model (2) adds the demographic, economic, and policy and political control variables discussed in the previous section. Model (3) adds country fixed effects, which account for time-invariant country characteristics. Model (4) adds year fixed effects, which control for time-variant characteristics that are constant across countries at a given point in time. Model (5) accounts for linear country-specific time trends through the addition of interactions between country dummies and a continuous year variable. This specification thus controls for unobserved country-level characteristics that change in a linear fashion, but at different rates in different countries. For example, while all countries in the sample may have become more gender-progressive during the nineteen-year period of study, some may have changed at a faster rate than others in this regard. Finally, Model (6) adds a squared maternity leave term, which allows for the possibility of a nonlinear relationship between the length of paid maternity leave and the gender wage gap [18].

In Table 5, I use Model (6) as a base model to conduct subgroup analyses examining whether my relationship of interest varies depending on the college gap, the industrial employment gap, and the services employment gap. Model (7) includes all country-year observations with a college gap equal to or less than zero, indicating that a greater share of women than men have a college education. Model (8) includes all observations with a college gap greater than zero. To create an even distribution, the industrial employment gap and the services employment gap were dichotomized using the within-sample medians. Models (9) and (11) include observations from country-years whose industrial employment gap and services employment gap, respectively, were greater than or equal to the median. Models (10) and (12) include observations for which these gaps were less than the median.

Main Regression Results

In Table 4, Model (1) suggests that there is a positive and statistically significant relationship between paid maternity leave and the gender wage gap. A one-week increase in the length of paid maternity leave is associated with a 0.03 percentage point increase in the gender wage gap. As indicated in Table 2, the average gender wage gap within my sample is 19.1% and the average length of maternity leave is about 32 weeks. A one-week increase in leave thus corresponds to a 3% change relative to the mean, whereas a 0.03 percentage point increase in the gender wage gap corresponds to a 0.2% change relative to the mean. My results therefore suggest that, in practical terms, a modest increase in paid maternity leave corresponds to an extremely small increase in the gender wage gap.

The addition of control variables in Model (2) changes the sign of this relationship, which also loses significance at conventional levels. In Models (3) and (4), the addition of country and year fixed effects, respectively, does not substantially change the sign or significance of the relationship. However, when linear country-specific time trends are added in Model (5), the relationship again becomes positive and attains significance at the 10% level. In this specification, a one-week increase in the length of paid maternity leave is associated with a 0.01 percentage point increase in the gender wage gap (a 0.05% change relative to the mean).

Finally, when a quadratic term is added in Model (6), I find that there is an inverted U-shaped relationship between weeks of leave and the wage gap. As shown by the results of the F-test at the bottom of the table, the linear and quadratic terms are jointly significant at the 5% level. This specification suggests that, on average, an eight-week increase (roughly a 0.25 standard deviation change) in the length of paid maternity leave from the sample mean of 32 weeks is associated with a 0.16 percentage point increase in the gender wage gap, or a 0.8% change relative to the sample mean [19]. Again, the magnitude of this relationship is very small. The vertex of this model falls at approximately 131 weeks, or roughly two and a half years of paid leave [20]. In contrast with the findings of previous studies, the results of this specification imply that shorter leave entitlements are associated with wider gender wage gaps, while policies granting leave in excess of two and a half years are associated with narrower gender wage gaps.

While I estimated specifications that layered on quadratic country-specific time trends through the addition of interactions between country dummies and a continuous year squared term, the key coefficients did not meaningfully change (see Appendix Tables A1 and A2). To account for the potentially delayed response of female labor supply and employer demand for female labor to changes in leave entitlements, I also tested alternative specifications in which the key independent variable was replaced with one-, two-, five-, and ten-year lagged measures of maternity leave. However, my key coefficients were insignificant and provided no evidence to support a delayed effect (results not reported here) [21].

Subgroup Results

As shown in Table 5, I also test for whether the relationship between the length of paid maternity leave and the gender wage gap differs depending on three country-level characteristics: the college gap, the industrial employment gap, and the services employment gap [22]. Akgunduz and Plantenga (2013) identified a negative relationship between the length of parental leave and women’s wages, but only for women in high-skill occupations. It is possible that the relationship between maternity leave and the gender wage gap may be different in societies in which women are more educated, and thus more likely to occupy high-skilled positions. Similarly, higher levels of occupational segregation—represented here by the industrial and services employment gaps and identified by Blau and Kahn (2017) as a contributor to the size of the explainable gender wage gap—may change the relationship of interest. While all subgroup analyses support the direction of the relationship implied by Model (6), the magnitude and significance of this relationship varies.

As described above, in Models (7) and (8), I divide my sample into high and low female education sub-samples based on the college gap. The inverted U-shaped relationship between paid maternity leave and the gender wage gap is significant at the 10% level in Model (7), as shown by the F-test for joint significance at the bottom of the table. Among country-years in which the share of women with college degrees is equal to or greater than the share of men, an eight-week increase in the length of paid maternity leave from the sample mean of 32 weeks is associated with a 0.28 percentage point increase in the gender wage gap (a 1.5% change relative to the sample mean) [23]. On the other hand, when the share of women with college degrees is lower than the share of men, there is no evidence of a significant relationship.

For Models (9) and (10), I created high and low industrial employment gap sub-groups by dichotomizing the industrial employment gap variable—which measures the difference between the share of men and the share of women employed in the industrial sector—using the in-sample median value (21.7%). The relationship between paid maternity leave and the gender wage gap is only significant at conventional levels when the industrial employment gap is high (Model 9). My results suggest that, for this group, an eight-week increase in the length of paid maternity leave from the sample mean of 32 weeks is associated with a 0.13 percentage point increase in the gender wage gap (a 0.7% change relative to the sample mean) [24]. However, when the industrial employment gap is low, there is no evidence of a relationship.

Finally, in Models (11) and (12), I created high and low services employment gap sub-groups by dichotomizing the services employment gap variable—which measures the difference between the share of men and the share of women employed in the services sector—using the median value (-24.7%). Again, the inverted U-shaped relationship between paid maternity leave and the gender wage gap is only significant (at the 10% level) in Model (11). Specifically, in country-years with a high services employment gap, an eight-week increase in the length of paid maternity leave from the sample mean of 32 weeks is associated with a 0.24 percentage point increase in the gender wage gap (a 1.3% change relative to the sample mean) [25].

In summary, my fully specified model, which has been purged of as much bias as possible through the addition of control variables, two-way fixed effects, and linear country-specific time trends, provides evidence of an inverted U-shaped relationship, which the opposite of what I hypothesized based on the findings of related studies. This finding suggests that shorter maternity leave entitlements are associated with wider gender wage gaps, but that as policies extend beyond about two and a half years, they are associated with narrower gaps. The results of my subgroup analyses provide additional support for the direction of the relationship but suggest that it is only significant when there are high levels of female education, a high industrial employment gap, or a high services employment gap. Nonetheless, in all specifications, the size of the relationship is very small in magnitude and provides little evidence that maternity leave policies have a meaningful impact on the gender wage gap. In the next section, I discuss the limitations of my analysis and the policy implications of these findings.

Conclusion

My results suggest that there is an inverted U-shaped relationship between the length of paid maternity leave and the size of the gender wage gap in OECD countries. Specifically, my fully specified model provides evidence that the lengthening of maternity leave is associated with widening gender wage gaps until the length of leave reaches about two and a half years; after that, longer leave is associated with narrowing wage gaps. However, the magnitude of this relationship is very small.

Potential Role of Selection Effects

Interestingly, despite the fact that I adopted a similar methodological approach, my results contrast with other findings in the existing literature, which generally fall into two categories. Some authors, including Ruhm (1998), Thévenon and Solaz (2012), Christofides et al. (2013), and Cukrowska-Torzewska and Lovasz (2020) found evidence of a positive relationship between leave and the gender wage gap. Others, including Olivetti and Petrongolo (2017), found evidence of a U-shaped relationship, in which shorter leave entitlements are associated with narrower gaps, while longer entitlements are associated with wider gaps. As discussed in more detail below, this discrepancy may be due to analytical limitations, including omitted variable bias. However, the inverted U-shaped relationship may also be a function of the selection effects at play in the labor market, as was proposed in the Conceptual Framework.

The gender wage gap is calculated using the median salaries of full-time employees. If women systematically join or drop out of the labor force in response to changes in maternity leave policies, the median female salary and, consequently, the gender wage gap will change. Recall that several studies have in fact found evidence of an inverted U-shaped relationship between maternity leave and female labor force participation (Thévenon and Solaz 2012; Akgunduz and Plantenga 2013; Del Rey et al. 2021). Thus, changes in female labor force participation could be the pathway through which maternity leave affects the gender wage gap [26]. My findings suggest that this dynamic may play out in two stages.

When given access to paid maternity leave, women may be induced to enter the workforce to qualify for the entitlements (Blau and Kahn 2013). If these new entrants to the labor market have less job experience or lower levels of educational attainment than the women currently in the workforce, they may receive lower wages, which would reduce the median female salary and widen the wage gap. Thus, when leave is relatively short, female labor supply factors may dominate.

However, beyond a certain point (in my model, two and a half years), the gender wage gap begins to narrow. This phenomenon could be due to a combination of female labor supply and demand factors. First, on the supply side, women may remove themselves from the labor force after long periods of maternity leave. Lalive and Zweimuller (2009) found that longer maternity leave policies were associated with delays in mothers’ return to work beyond the leave provided, which reduced short-term maternal employment. Similarly, Ejrnæs and Kunze (2013) observed that longer parental leave was associated with a reduced likelihood that mothers return to their jobs within the first three years of their child’s life. These studies suggest that some women choose not to return to work after long periods of maternity leave, which may be attributable to individual preferences or to economic factors such as the high cost of childcare.

Second, firms may have reduced demand for female labor when leave entitlements are lengthy, since such entitlements can make women relatively more expensive to the firm to employ. Increased employment costs may take the form of reduced productivity (Amano-Patino et al. 2020), the requirement to replace employees who take leave, and the need to compensate for the skill deterioration that occurs during leave (Ruhm 1998). These demand-side factors may result in unfavorable employment opportunities for women, including lower salaries, which may induce some to leave the workforce. In this hypothetical scenario, as leave gets longer, both supply-side and demand-side factors push women out of the labor market. If the women who exit the workforce as leave gets longer are less experienced or less productive than the women who remain, median female wages could increase, reducing the wage gap.

Analytical Limitations

The conflict between my findings and those found in the existing literature may be attributable to the above-discussed role of selection effects. Alternatively, this discrepancy could be the result of differences in sample composition or omitted variable bias. My methods are largely consistent with those of other studies in this field. All three related country-level analyses (Thévenon and Solaz 2012; Ruhm 1998; Olivetti and Petrongolo 2017) used two-way fixed effects, and Ruhm (1998) and Thévenon and Solaz (2012) further controlled for linear country-specific time trends (as I did). On the other hand, I controlled for a wider range of demographic, economic, and policy and political variables than past studies. However, my empirical approach differs in terms of the country-year observations included in my analytic sample. It also is subject to omitted variable bias stemming from several variables, some of which were included in past studies.

My analytic sample contains data on all 38 OECD countries from 2000-2018. Among the existing studies on this topic, Olivetti and Petrongolo’s (2017) sample is the largest and most recent, including 22 countries from 1970-2010. While I add 16 countries and an additional eight years of recent data, I did not have access to historical gender wage gap data from 1970-1999. In my sample, 23 countries experienced a change in their maternity leave policies between 2000 and 2018; however, there were a substantial number of changes in leave entitlements pre-2000. Because fixed effects regression exploits within-country changes to estimate the relationship of interest, my results may be less generalizable than they would have been if I had used a longer period of analysis.

Additionally, while my analytical approach is similar to that of Olivetti and Petrongolo (2017), those authors had access to data on the average wage replacement rate, which measures the proportion of average national earnings paid out during leave (OECD 2017). This variable has been omitted from my analysis due to data limitations and is likely a source of upward bias. I hypothesize that the wage replacement rate is positively correlated with the length of paid maternity leave, given that countries who provide more generous leave entitlements may also be more generous with their payments. I also suspect that the replacement rate is positively correlated with the gender wage gap, as shown by Olivetti and Petrongolo (2017). More generous leave policies could encourage stronger maternal workforce attachment: when less experienced or less productive women—i.e., those who would have otherwise dropped out—remain in the workforce, the gender wage gap may widen. If my reasoning is correct, the omission of the wage replacement rate from my regressions is exerting upward bias in my key estimates.

Another potentially important omitted variable is the average length of female labor force experience, which has been found to be a component of the explainable gender wage gap (Blau and Kahn 2017). Given that there is no country-level data on this variable, I use the average age at birth as a relatively weak proxy. The omission of this variable may be exerting downward bias in my estimates. Women’s work experience is probably negatively correlated with the gender wage gap, since a more experienced female workforce would earn a higher median wage. Further, it is possible that this variable is positively correlated with the length of paid maternity leave. Voters in societies in which women are more likely to work—and therefore have greater average labor force experience—may demand more generous leave policies.

The average cost of childcare is another important omitted factor. While I do control for the percentage of GDP spent on childcare, this variable does not fully capture the costs faced by individual families. I expect that the cost of childcare is negatively associated with the gender wage gap. In dual-income households, as childcare becomes more expensive, women with lower earnings may exit the workforce to care for children. This would increase median female earnings, reducing the gender wage gap. I also posit that this variable is negatively associated with the length of maternity leave entitlements, since cheaper childcare and longer leave both signal a more family-oriented society. My reasoning thus suggests that the omission of this variable is exerting upward bias in my estimates. Overall, these various sources of omitted variable biases likely pull my coefficient of interest in different directions, making it difficult to determine their net impact on the size and direction of my estimate of the relationship between paid leave and the gender wage gap.

Lastly, my analysis is limited because of the challenge of quantifying heterogenous leave policies. Two countries may have identical lengths of leave (which means that they would have identical values in my sample), but they may differ in terms of who qualifies or the level of wage replacement. For example, both Ireland and Iceland provide 26 weeks of paid leave (OECD 2019a). In Ireland, this benefit is paid at a flat rate to women who qualify through having contributed to the state social insurance program for at least 39 weeks in the past year (OECD 2019a; Citizens Information Board 2022). On the other hand, in Iceland, the benefit is 80% of earnings, up to an annually adjusted ceiling; individuals qualify if they were employed in the country consecutively for six months prior to their child’s birth (OECD 2019a; European Commission n.d.). While a quantitative analysis that takes into consideration all the granularities of paid leave policies may not be possible, the level of wage replacement is perhaps the most important missing component from this analysis, as discussed above.

Future Research and Policy Implications

Future research should attempt to resolve the differences between my results and those of other studies in this area. This endeavor may take the form of analyses that include previously omitted variables, or that expand the geographic and/or temporal scope of analysis. Additionally, the relationship between the gender wage gap and other policies that change parents’ work-life balance could be investigated. For example, past studies have found that greater access to publicly funded childcare is associated with an increased likelihood that mothers are employed (Pettit and Hook 2005) and a smaller wage penalty related to having children (Misra et al. 2011). These factors could help alleviate the impact of motherhood on the gender wage gap. While maternity leave only provides support in the early months or years of a child’s life, a family’s childcare needs persist for many more years. As such, future research could examine whether family programs that provide affordable, accessible childcare are more effective at promoting wage equality than paid maternity leave.

Ultimately, my results suggest that, if a relationship exists between the length of paid maternity leave and the size of the gender wage gap, it is very small in magnitude. However, this finding should not be taken to suggest that maternity leave policies are unimportant. Paid leave policies are associated with improved maternal and child health outcomes (Patton et al. 2017; Baker and Milligan 2018b; Chatterji and Markowitz 2008). Crucially, they also give women the flexibility around childbirth necessary to allow them to balance work and family, if they so desire. Nonetheless, while paid leave policies likely produce many other beneficial outcomes, my results suggest they may not be an effective means of reducing the gender wage gap.

Appendix