Open access

Research Article

17 October 2023

Child Penalties in Canada

Authors: Marie Connolly, Marie Mélanie Fontaine, and Catherine HaeckAuthor information and affiliations

Publication: Canadian Public Policy

Volume 49, Number 4

https://doi.org/10.3138/cpp.2023-015

PDF EPUB

Résumé

Le fait d’avoir des enfants a un impact considérable sur les résultats des femmes sur le marché du travail, mais pas sur ceux des hommes. Les effets différentiels du parentage selon le sexe, ce qu’on appelle la pénalité liée aux enfants ou la pénalité parentale, sont désormais documentés pour de nombreux pays. Dans cet article, nous exploitons les données de l’Étude longitudinale et internationale des adultes pour estimer les pénalités liées aux enfants au Canada, tant à l’égard des revenus que de l’emploi, sur une période allant de cinq ans avant la naissance du premier enfant à dix ans après. À partir d’une méthode fondée sur l’étude d’évènements, nous constatons des effets négatifs importants et durables de la parentalité chez les mères, mais pas chez les pères. Les revenus des mères diminuent de 49% l’année de la naissance, et la pénalité est toujours de 34,3% dix ans plus tard; en ce qui concerne l’emploi, la pénalité correspondante est de 14,2%. Nous documentons également les effets négatifs plus importants de la parentalité pour les femmes qui ont de nombreux enfants et pour celles qui ont un niveau de scolarité inférieur. Enfin, nous fournissons des données qui suggèrent que des politiques familiales telles que le congé parental et les services de garde subventionnés pourraient contribuer à réduire les pénalités liées aux enfants.

Abstract

Having children has a sizeable impact on women's labour outcomes, but not on men's. The differential effects of children by gender, referred to as child penalties, are now documented in many countries. In this paper, we exploit the Longitudinal and International Study of Adults to estimate Canadian child penalties in both earnings and employment for a period going from five years before the birth of the first child to ten years after birth. Using an event study methodology, we find large and persistent negative effects of parenthood for mothers, but not fathers. Mothers’ earnings decrease by 49 percent at the year of birth, with a penalty still at 34.3 percent ten years after birth; the corresponding penalty in employment is 14.2 percent. We also document larger negative impacts of parenthood for women who have multiple children or those with a lower education level. Finally, we provide suggestive evidence that family policies such as parental leave and subsidized childcare may help reduce child penalties.

Introduction

Despite notable improvements in the economic and social situation of women in recent decades, inequalities between men and women persist in many fields, a trend which is seen both in Canada and internationally. Indeed, women's wages have increased progressively over time, but women continue to be paid less than men on average. Between 1998 and 2016, the ratio of women's to men's hourly wage in Canada rose from 80 percent to 86 percent (Fortin 2019). Québec¹ has a slightly higher gender wage ratio than Canada overall, with a women-to-men ratio of 89 percent in 2016, up from 84 percent in 1998 (Cloutier-Villeneuve 2018). In the United States, the hourly wage ratio for full-time employees increased significantly during the 1980s before slowing down, rising from 64 percent in 1980 to 74 percent in 1989 and then 82 percent in 2010 (Blau and Kahn 2017).

Although the rise in hourly wage has been favourable for women in Canada over the last twenty years, the trend in annual earnings reveals a different story, one in which larger gaps remain. The women-to-men ratio of average annual earnings was around 69 percent in 2015, an increase of around 15 percentage points since 1978 (Bonikowska, Drolet, and Fortin 2019). For the younger generations, the earnings ratio has improved, but a dramatic decline past the age of 30 is still observable, coinciding with the arrival of children for many households (Fortin 2019). Whereas the gender wage ratio only considers differences in the average hourly wages of men and women, annual earnings also take into consideration differences in the number of hours and weeks worked by women and men. It therefore appears that there are still large differences between men and women in terms of not only wage rates but also labour intensity.

Many studies have investigated the causes of these inequalities and concluded that women have been unable to catch up to the earnings level of men in part because of parenting responsibilities (Angelov, Johansson, and Lindahl 2016; Kleven, Landais, and Søgaard 2019; Waldfogel 1998). It is generally considered that children have a negative effect on women's productivity in the labour market by substantially reducing their human capital or work effort, something which translates into a significant decrease in their earnings (Mincer and Polachek 1974). Indeed, when children arrive, mothers do more often turn towards part-time jobs, positions offering flexible working hours, or a situation offering work conditions more favourable to family life—all of which are less profitable and pay lower wages (Joshi, Paci, and Waldfogel 1999). Employers in return see part-time employees as less committed to their employers and performing at a lower level and hence offer them fewer bonuses and promotions (White 2019). Goldin (2014) argues that the gender pay gap could be substantially reduced if long work hours and inflexible schedules were not rewarded as much by firms. This situation may also explain why women have a hard time entering the top decile of the earnings distribution, a phenomenon that can account for more than 50 percent of the gender pay gap in Canada (Bonikowska et al. 2019).

It thus appears that women pay a majority of the direct and indirect childcare costs, resulting in a deterioration of their economic situation. In the literature, this phenomenon is known as the “motherhood pay gap,” the “family gap,” or “child penalties” (Kleven, Landais, and Søgaard 2019; Waldfogel 1997). Henrik Kleven, in a series of papers on the subject, sheds light on the actual extent of the phenomenon that women disproportionately bear parenting-related costs in the labour market (Kleven, Landais, and Søgaard 2019, 2021; Kleven et al. 2019, 2022; Kleven 2022). For example, Kleven, Landais, and Søgaard (2019) use administrative data from Denmark, which is considered to be one of the most egalitarian countries in the world, and reveal that the earnings of mothers diminish considerably following the birth of their children, whereas men's pay is not affected, leading to a long-run child penalty of 19.4 percent ten years after the birth of the first child.

The main goal of our study is to document child penalties in Canada. We do so for parents who had their first child between 1989 and 2009 and follow Kleven, Landais, and Søgaard's (2019) methodology, focusing on annual earnings and annual employment. In order to achieve this aim, we exploit the Longitudinal and International Study of Adults (LISA), a database that includes a longitudinal study coupled with historical administrative fiscal records. We then present evidence comparing mothers and fathers, but we also look at heterogeneity among mothers along marital status, number of children, education level, and whether they are born in Canada or not. Our contribution is to present the first descriptive evidence on child penalties in Canada for a nationally representative sample that allows us to follow individuals over the course of 15 years surrounding the birth of their first child.

In earlier contributions, Phipps, Burton, and Lethbridge (2001) find that mothers’ incomes are 13 percent lower than those of women without children (even when taking into account work interruptions), but they do so using a cross-sectional dataset (the 1995 General Social Survey). Zhang (2010) follows mothers from three years preceding the birth of their first child to nine years after and observes that earnings of mothers dip 30 to 40 percent at child birth while regaining losses after seven years. However, Zhang (2010) had to use maternity leave as an indicator for childbirth given the nature of the data (the Longitudinal Worker File), thus narrowing the focus of his study to employed women. Karademir, Laliberté, and Staubli (2023) also present evidence on child penalties based on two administrative data sources that are scarce in terms of sociodemographic variables such as education level. First, they use the Intergenerational Income Database and a sample representative of Canadian parents born between 1963 and 1985 who had their first child before age 40 and between 1981 and 2016. Second, they exploit a sample from the Longitudinal Administrative Databank (LAD).² We thus believe our paper is the first to provide a comprehensive picture of child penalties in Canada both by showing the impact of children on the labour market outcomes of women relative to men and also by being able to look into heterogeneous effects.

A secondary goal of our study is to investigate the potential effect of public policies on child penalties. Recently, both academic researchers and political decision makers have become interested in solutions to close the motherhood pay gap. For example, one such potential solution is to eliminate access-to-employment barriers and reduce inequalities in unpaid work, especially those pertaining to parenting responsibilities. In contrast, Kleven et al. (2022) explore the role of family policies in Austria and find that they had virtually no impact on the motherhood pay gap. This finding may lead some to conclude that family policies are not helpful for reducing the motherhood pay gap. However, we argue that the Austrian policy mix and social norms likely explain the result and may not have external validity for family policies designed differently.³ In this paper, we focus our attention on two policies: parental leave and childcare. Using a difference-in-differences approach, we investigate whether the Québec childcare policy combined with a more targeted parental leave program helped reduce child penalties. The Québec context is very different from the Austrian setting: the labour force participation of mothers is remarkably higher, the gender pay gap lower, and the childcare participation of children aged one to two much higher.⁴

Our main findings are as follows. We find that the earnings of mothers diminish dramatically following the birth of their first child, with the loss estimated at 49 percent for the year of birth. On the other hand, the earnings of fathers are not affected by the arrival of the first child. Over the long term, which is to say ten years after the birth of the first child, women in Canada continue to be paid less than they earned before the birth of their first child, and less than men, with a long-run child penalty of 34.3 percent. When looking at annual employment, we find a smaller but still substantial long-run penalty of 14.2 percent. We do not see large differences between married mothers and those in common-law partnerships or between those born in Canada and those born in another country, but we do observe stronger effects of motherhood for those with more children and those with lower educational attainment.

Finally, we find suggestive evidence that the Québec family policies have had a positive impact in reducing earnings gaps between mothers and fathers, especially in the long run. Québec mothers who gave birth to their first child in 2001 or later see their earnings increase more quickly in subsequent years than mothers in the rest of Canada or Québec mothers whose first child was born before 2001. Additionally, the long-run impact of children was reduced by 17 percentage points in Québec, from –40 percent to –23 percent. In comparison, the impact for mothers in the rest of Canada went from –39 percent to –33 percent. The overall effect of Québec family policies on earnings is therefore 11 percentage points; for employment, it is 8 percentage points. Our main limitation stems from the number of observations in the LISA, which is often not large enough for us to detect statistically significant effects. Further work in the area should exploit data sources with significantly more observations, as done by Karademir et al. (2023).

The paper is structured as followed. We first briefly review the literature on the child penalties in next section. We then present our data and empirical strategy. Our main findings on the effects of parenthood follows, while we subsequently offer suggestive evidence on the potential impact of the Quebec family policies.

Selected Literature Review

Since the seminal work by Hill (1979), many studies have examined thse effect of motherhood on women's labour supply and earnings (Angelov et al. 2016; Lundberg and Rose 2000; Waldfogel 1995; Waldfogel 1998). The literature that has developed over the last four decades covers a wide body of research that uses sophisticated estimation methods to account for endogeneity and selection bias in order to establish a causal relationship. Given that having a child is a choice and that people who make that choice may differ from those who decide not to have a child, a simple comparison of the earnings of mothers and women without children is insufficient to declare a causal effect of having a child. For example, women who plan to have one or more children—knowing that this will interrupt their career—might be less inclined to invest in their education and therefore be subject to slower professional advancement, which is a case of endogeneity (Budig 2014). Similarly, women with lower earnings may be ones who choose to have children since the opportunity cost is lower, which creates a selection bias (Jacobsen, Pearce, and Rosenbloom 1999). Researchers therefore turn to fixed-effect models (Zhang 2010), instrumental variables (Miller 2011), or Heckman selection models (Misra, Budig, and Boeckmann 2011) to study the relationship between earnings and motherhood. These methods make it possible to overcome problems introduced by endogeneity and selection bias, and therefore make it possible to estimate a relationship of cause and effect (and not just a correlation).

A growing number of studies also use data from administrative sources to lend additional credibility to their results (Fernández-Kranz, Lacuesta, and Rodríguez-Planas 2013; Kleven, Landais, and Søgaard 2019). Many studies have also looked at the motherhood pay gap for various population subgroups, including those based on education level (Anderson, Binder, and Krause 2002), ethnic origin (Glauber 2007), marital status (Budig and England 2001; Budig and Hodges 2010), immigration status (Srivastava and Rodgers 2013), and income distribution (Budig and Hodges 2010; Kellewald and Bearak 2014). Nonetheless, comparing estimates from different studies is a difficult exercise given the differences that may exist in the samples studied and the methodology used. Several studies have therefore used harmonized international data, such as the Luxembourg Income Study⁵ (Budig, Misra, and Boeckmann 2016; Harkness and Waldfogel 2003; Misra et al. 2011) or the International Social Survey Programme⁶ (Blau and Kahn 2003; Dupuy and Fernández-Kranz 2011) to provide comparable estimates across countries.

Significant variations are generally found between countries in these studies: cross-sectional motherhood pay gap estimates vary from −42 percent to +4 percent, with large gaps generally found in less developed countries and gaps favouring mothers in Nordic countries (Grimshaw and Rubery 2015). Few of these studies, however, analyze possible sources of variation to explain these international differences. On this subject, Misra, Budig, and Boeckmann (2011) look at the relationship between various policies to establish a work–family balance and the motherhood pay gap in 21 countries. Their results show that some policies, such as maternity leave and daycare services for young children, are positively associated with mothers joining the workforce, while the effect is more ambivalent in the case of others, including parental leave. In other words, the longer the leave, the greater the negative effect felt on earnings. These results are confirmed by Budig, Misra, and Boeckmann (2016) and Olivetti and Petrongolo (2017). A relationship also exists between the generosity level of family policies and the motherhood pay gap. The penalties associated with maternity are greater in southern Europe (Dupuy and Fernández-Kranz 2011), while the lowest penalties are generally found in northern European countries (Harkness and Waldfogel 2003).

Recently, a study conducted by Kleven, Landais, and Søgaard (2019) played an important role in reopening the debate on inequalities between men and women, specifically the parenting burdens disproportionately borne by women to the detriment of their professional career. Since that study was published, scientific articles addressing this subject have proliferated, and both academic researchers and political decision makers have turned their attention from whether a pay gap exists to factors that could possibly serve to reduce it. Kleven, Landais, and Søgaard (2019) use a large Danish administrative database that is also rich in socioeconomic variables. Their base sample comprises about 470,000 births and 15,040,000 observations collected annually over the years. They estimate the effect of the arrival of children on various aspects of the parents’ work life (specifically earnings), participation in the workforce, hours worked, and hourly wage.

The results of the study by Kleven, Landais, and Søgaard (2019) show that the future parents have identical earnings trajectories during the years before the birth of the first child. However, when the child arrives, women experience a drastic loss in earnings (approximately 30 percent compared with what they earned the year preceding the birth), whereas the fathers’ earnings are unaffected following the birth of their children. During the following years, the mothers’ lost wages are partially regained, but a 19.4 percent gap between men and women remains after ten years—the so-called long-run child penalty. Kleven, Landais, and Søgaard (2019) advance that the reasons for the loss in earnings may lie with three factors: the number of hours worked, labour force participation, and the hourly wage. They further assert that these three factors are negatively associated with the birth of a child for mothers, while the same factors do not appear to have been affected in any way for fathers. These study results show that even in a country with a solid social security net (Denmark), women are not immune from the child penalty. Another important contribution of Kleven, Landais, and Søgaard (2019) is their careful argumentation—through a comparison with a difference-in-differences approach, where individuals without children are used as counterfactual, and with an instrumental variable approach using sibling sex mix—that their event study methodology is appropriate to identify causal effects.

The provisions of a parental leave policy with job protection, including maternity leave and parental leave that can be shared between the two parents, are generally implemented for the purpose of offering parents the possibility of spending more time with their newborn—while also guaranteeing that the same work conditions will be in place following the leave. A system of parental insurance that is advantageous in terms of time and financial compensation is considered to be a concrete measure to ensure a work–family balance for both women and men, but women are the main beneficiaries in the majority of cases. Earlier studies have shown that women's professional trajectories tend to progress more linearly in countries with measures promoting such a balance (Blau and Kahn 2003) and that a longer job-protected leave increases job continuity with the pre-birth employer (Baker and Milligan 2008).

The effect of childcare policies on various outcomes have also been largely documented. Baker, Gruber, and Milligan (2008) and Lefebvre and Merrigan (2008) demonstrate quite robustly the impact of the Québec affordable childcare reform on workforce participation by mothers using two distinct microdata sources. Lefebvre, Merrigan, and Verstraete (2009) and Haeck, Lefebvre, and Merrigan (2015) later confirmed the results of the earlier studies and find that the positive effects persist in the long term. Studies on maternal employment and daycare services have also been conducted in other countries. The majority of these studies find a positive association between the availability of daycare services and mothers’ level of employment, for example in Germany (Bauernschuster and Schlotter 2015), Argentina (Berlinski and Galiani 2007), the Netherlands (Bettendorf, Jongen, and Muller 2015), France (Givord and Marbot 2015), and Norway (Kunze and Liu 2019). These documented effects suggest a shift in the labour force participation of treated mothers and one of the questions we investigate in this paper is whether this shift helped reduce the child penalties in employment and earnings.

While the literature offers a certain number of studies on child penalties, few tackle the effect of family policies on those penalties. Kleven et al. (2022) do exploit various changes in the family policies of Austria to estimate the impact on the motherhood pay gap. The authors find that generous parental leave of up to 35 months combined with affordable childcare did not reduce child penalties. While the authors argue that the Austrian context is not unique, it is different from the Canadian or Québec context. Parental leave can last for up to 35 months in Austria relative to 12 months in Canada. This prolonged parental leave may entice women to remain out of the labour force for a prolonged period after childbirth, which data on childcare seem to support: childcare participation of children aged one and two is indeed fairly low, at around 25 percent (Kleven et al. 2022). In contrast, childcare participation of children aged one to four increased from around 40 percent to 75 percent in Québec between 1996 and 2008 (Haeck et al. 2015), with a similar increase amongst children aged one and two. Parallel to this increase, the labour force participation of mothers of children aged one to four increased from less than 55 percent in 1996 to more than 70 percent in 2008 (Haeck et al. 2015).

Andresen and Nix (2022) provide evidence from Norway on heterosexual non-adopting and adopting couples, and same-sex couples. While they do not offer a causal evaluation of the impact of family policies, they conclude that gendered preferences towards childcare—along with gender norms and discrimination—explain in large part the child penalty among heterosexual couples. This finding is thus somewhat at odds with the conclusion of Kleven et al. (2022). However, Karademir et al. (2023) argue that grandparents may be the missing piece of the puzzle that explain the divergent findings. They posit that another factor to consider is the availability of other forms of childcare, in particular care provided by grandparents. The impact of childcare subsidies would thus depend on the pre-existing childcare arrangements and the prevalence of grandparent-provided care.

Methodology

To estimate the impact of parenthood, we use an event study methodology that exploits the longitudinal nature of our data set. We note that an increasing number of economic studies use this approach to explore issues pertaining to economic consequences associated with the birth of children (Angelov et al. 2016; Kleven, Landais, and Søgaard 2019; Kuziemko et al. 2018; Zhang 2010).

For a sample of parents, we begin by defining the event as the year the first child was born and build a series of time indicators over a 15-year period around the birth. All years are then indexed so we can observe personal earnings or employment trajectories relative to the birth of the first child (τ = 0 or event time zero). The years observed are from five years before the birth (τ = −5) to ten years afterwards (τ = +10)—in other words, the period −5 ≤ τ ≤ 10.

To analyze changes in the parents’ outcome occurring within the 15-year window around the birth of their first child, we follow Kleven, Landais, and Søgaard (2019) and estimate this multiple linear regression model using ordinary least squares:

Y_{i s t}^{g} = \sum_{- 5 \leq j \leq 10, j \neq - 2} α_{j}^{g} \cdot 1 [j = t] + \sum_{k} β_{k}^{g} \cdot 1 [k = a g e_{i s}] + \sum_{y} Y_{y}^{g} \cdot 1 [y = s] + μ_{i s t}^{g}

(1)

In this model, the dependent variable

Y_{i s t}^{g}

represents either annual earnings (T4 earnings from paid employment⁷) in calendar year s in real terms, i.e., corrected for inflation and expressed in constant 2016 dollars, for individual i (of gender g) and event time t, or annual employment, a dichotomous variable equal to one if individual i has positive employment income during year s, and zero otherwise. The term 1[j = t] is an indicator (dummy) variable indexing event time, or time relative to the birth of the first child. Parameters of interest, specifically parameters measuring changes in outcome following the birth of a first child, are represented by the series of

α_{j}^{g}

. These parameters are indexed with respect to the birth of the first child, i.e., from five years before to ten years after birth. In our estimations we omit the dichotomous variable associated with τ = −2. The estimated

α_{j}^{g}

coefficients measure the average effect of the arrival of a child on the parents’ outcomes and are expressed in relation to the outcome two years before the birth.⁸

The two remaining sets of terms include age and calendar year fixed effects. Adding age fixed effects to the model makes it possible to control for effects related to life cycle by comparing individuals of the same age. We also introduce year fixed effects to take into account temporal differences observed during the period. The final term,

μ_{i s t}^{g}

, is the model's error term. We estimate Equation (1) separately for mothers and fathers (gender g).

The resulting

{\hat{α}}_{t}^{g}

coefficients are expressed in levels, which means that they represent variations in reported earnings in dollars or annual employment in proportion. To convert these estimates to percentage differences, we apply the conversion method used by Kleven, Landais, and Søgaard (2019). This method involves dividing each estimated coefficient

{\hat{α}}_{t}^{g}

by the predicted outcome absent the birth of the child. The formula to obtain

P_{t}^{g}

is therefore

P_{t}^{g} = \frac{{\hat{α}}_{t}^{g}}{E [{\tilde{Y}}_{i s t}^{g} | t]}

, where

{\tilde{Y}}_{i s t}^{g}

is the predicted outcome absent the child of the birth, i.e. omitting the event time coefficients. Using this methodology allows us to keep in our estimation of Equation (1) observations with earnings of zero dollars while still presenting findings showing deviations in percentages. In our comparison of mothers to fathers, the child penalty is defined as the difference in P between women and men, and the long-run penalty is the one evaluated ten years after the birth of the first child. Note that the long-run penalty captures the total effect of children, not just the first child, since in our main analysis we do not restrict the sample to those having only one child. When we look at heterogeneity of the impact of motherhood among mothers according to marital status, number of children, education level, or whether born in Canada, we estimate a version of Equation (1) in which we add full sets of interactions between our model and the dimension that we want to study, resulting in separate sets of P_t by marital status, number of children, education level, or whether born in Canada.

In all our analyses, we use the sampling weights provided by Statistics Canada with the Longitudinal and International Study of Adults (LISA) data to ensure that the Canadian population was accurately represented. All standard errors are robust to heteroscedasticity.⁹

Data, Sample Construction, and Descriptive Statistics

Our study is based on the LISA coupled with individual longitudinal fiscal data (T1 forms) (Statistics Canada 2014, 2016). The LISA is administered by Statistics Canada. It was developed to provide longitudinal information pertaining to the labour market, education, family, and health from Canadian respondents. Every two years, permanent study members selected in Wave 1 in 2012 are interviewed. Individuals who have joined the members of the permanent household between waves are added to future waves, as are children in the household when they reach the age of 15. Wave 1 included about 16,000 households, for a total of about 32,000 respondents, while Wave 2 (in 2014) numbered 11,000 households, with 32,000 individuals. For this study, we use the first two waves of LISA to identify birth history and sociodemographic characteristics along with T1 forms from 1982 to 2019.

One of the leading characteristics of the LISA is the matching of database information with longitudinal tax records from the Canada Revenue Agency (CRA). Individual T1 forms (a form used for declaring income to the CRA) are available yearly since 1982. These provide detailed annual information on different sources of income, such as personal and family income before and after taxes, government transfers and capital gains, as well as marital status and province of residence. However, the tax files do not contain information on hours or weeks worked, so we are not able to estimate the impact of children on hours or hourly wages.

We select our sample according to the following specifications. First, we restrict our attention to parents experiencing their first births between 1987 and 2009 inclusively. Second, since the topic of the study is labour market outcomes, the sample is restricted to individuals who were of working age, specifically from ages 19 to 51, during the observation period (1982–2019).¹⁰

Table 1 presents some descriptive statistics for mothers in our sample according to whether they gave birth to their first child before 2001 or after 2000 and whether they resided in Québec or the rest of Canada at the time.¹¹ Given that one of the goals of this study is to provide suggestive evidence on the effect of the family policies introduced after 2000 on child penalties, it is useful to start with this table to check that the group composition is relatively stable over time. If the groups are not stable, we could attribute the changes in outcomes observed to the policy change—when in reality they might be due to changes in group characteristics.

Table 1: Sociodemographic Characteristics of Mothers Who Gave Birth to Their First Child before and after 2001 in Québec and in the Rest of Canada

	Québec		Rest of Canada
	Pre-2001	Post-2001	Pre-2001	Post-2001

	(1)	(2)	(3)	(4)
Sociodemographic characteristics
Median year of birth	1965	1977	1966	1977
Born in Canada	0.92	0.93	0.82	0.72
Family characteristics
Age at the time of the first birth	27.9	28.6	27.8	28.8
	(5.8)	(5.5)	(8.5)	(7.8)
Number of children (continuous)	1.96	2.15	2.22	2.04
	(0.98)	(0.99)	(1.31)	(1.36)
Number of children (dummies)
One child	0.30	0.23	0.17	0.27
Two children	0.49	0.49	0.54	0.51
Three children or more	0.21	0.28	0.29	0.23
Marital status at τ = 0
Married	0.43	0.22	0.71	0.66
Common-law partner	0.27	0.61	0.06	0.11
Separated/divorced/widowed/single	0.29	0.17	0.23	0.23
Education
Years of education (continuous)	14.1	15.2	14.7	15.4
	(4.0)	(3.8)	(4.3)	(4.0)
Highest diploma (dummies)
No diploma	0.08	0.04	0.05	0.03
Secondary diploma	0.18	0.07	0.23	0.17
Post-secondary diploma	0.42	0.45	0.39	0.37
University diploma	0.32	0.44	0.33	0.43
N (Mothers-years)	6,100	3,200	21,000	9,500
N (Mothers)	400	200	1,400	600

Notes: Year 2001 is included in the post-2001 columns. Except for the median year of birth, the numbers in the table are averages weighted with the sampling weights provided by Statistics Canada. Standard deviations for continuous variables are displayed in parentheses. Number of observations (N) rounded to the closest hundred to respect Statistics Canada's disclosure rules.

Source: Calculations made by the authors using LISA data (2012, 2014) and T1 forms (1982–2019).

We have two distinct groups separated into two time periods. The two groups appear to be similar in composition with respect to several variables. The median year of birth of mothers is similar in the two groups: 1965 or 1966 for the pre-2001 period and 1977 for the post-2001 period. Post-2001 mothers are of course born later since these are mothers who gave birth for the first time after 2000. The mother's age at the birth of her first child increased slightly before and after 2001, from around 27.8 to 28.8 years of age. This upward trend is observed in both Québec and the rest of Canada, and it can also be seen in the overall population, from 25.1 in 1982 to 28.5 in 2011 (Statistics Canada 2018).

Some differences can however be seen across the two groups. The share of Québec mothers born in Canada is similar across the two time periods (0.92 and 0.93) and higher than that in the rest of Canada, where it decreases in the post-2001 years (from 0.82 to 0.72). In robustness checks, we investigate whether the differences between Québec and the rest of Canada in terms of the influx of immigrants could be a driver behind the trends observed in our suggestive analysis of the effect of family policies on child penalties. We find that those differences do not appear to explain our findings. The average number of children for Québec women (1.96) before 2001 is lower than for women in the rest of Canada (2.22). This number increases in Québec post-2001 (2.15) but decreases in the rest of Canada (2.04). In Québec, 21 percent of women had three or more children pre-2001, compared with 29 percent in the rest of Canada. These numbers are higher than the fertility rate observed in the entire population because our sample excludes women without children. Marriage is a lot less common in Québec than elsewhere in Canada before 2001 (43 percent in Québec compared to 71 percent in the rest of Canada) and this difference increases over time (22 percent in Québec compared to 66 percent in the rest of Canada post-2001). We return to these differences when investigating heterogeneity by marital status.

Finally, the education gap between the two groups is relatively small. The gap is 0.6 years of schooling before and 0.2 years after 2001. In terms of the highest level of schooling completed, 8 percent and 4 percent of mothers in Québec have no diploma for the two periods respectively, compared with 5 percent and 3 percent of women in the rest of Canada. Post-2001, mothers in Québec are more likely to have a post-secondary diploma, but less likely to have a university diploma relative to mothers in the rest of the country.

Results

We start with a descriptive account of earnings and employment trajectories for fathers and mothers in order to provide estimates of child penalties in Canada. We then show evidence on the heterogeneity of the impact of motherhood by marital status, number of children, education level, and whether born in Canada.

Estimated Effects of Parenthood for Mothers and Fathers

We begin by presenting the estimated impact of children in earnings and employment in Figure 1.¹² The coefficients shown graphically represent variations in the earnings (top panel) and employment (bottom panel) of mothers and fathers relative to two years prior to childbirth. The effects of parenthood for mothers and fathers are shown separately on the same figure. The shaded areas represent 95 percent confidence intervals.

Figure 1: Impact of Children for Mothers and Fathers
Notes: The values shown on the y-axis correspond to changes as a percentage based on estimated coefficients. The results are weighted using the sampling weights provided by Statistics Canada. Shaded areas represent 95% confidence intervals. The standard errors are heteroscedasticity-robust. Top panel shows variations in earnings; bottom panel shows variables in annual employment. Long-run child penalties computed as difference between fathers and mothers at time τ = 10.
Source: Calculations made by the authors using LISA data (2012, 2014) and T1 forms (1982–2019).

First, the results show that the T4 earnings trajectories of men and women are similar during the years preceding the birth of their first child. This is not to say that their earnings levels are the same, but rather that their earnings variations are similar in comparison with the reference year (two years prior to the first childbirth). Stark differences emerge on the year of the birth of the first child: women experience a drop in earnings of 49 percent while men do not see their earnings affected by the birth of their first child. Although mothers recover a small amount of the lost earnings in subsequent years, ten years after the birth they are still making less than they did before having a child. On average, their earnings are down 34.3 percent, which leads to gender differences (variation for fathers minus variation for mothers) at τ = 10, or long-run child penalties of 34.3 percent since the effect of the first child for fathers is estimated at 0.05 percent. It should be noted here that the event study is with respect to the birth of the first child, but the majority of these mothers go on to have other children in this ten-year period. For this reason, the effect measured here takes into account not only the short-term effect of the first child but also the long-term effect of other children born during the observation period.

The top panel of Figure 1 raises another point: mothers’ earnings decrease by about 20 percent starting in the year before the birth, or τ = −1. Since we do not have information about the children's month of birth, we are unable to validate the hypothesis put forward earlier that this decrease in earnings is the result of early maternity leaves and preventive withdrawals from the labour force in the case of births occurring early in the year. However, Kleven, Landais, and Søgaard (2019) show that by concentrating solely on January births, all labour market variables are negatively affected during calendar year τ = −1, which supports our initial hypothesis.

The bottom panel of Figure 1 shows variations in employment for fathers and mothers. We see a similar pattern as that regarding earnings (top panel): fathers are by and large unaffected by the birth of their first child while women experienced large and persistent negative effects. A difference with the top panel is that the largest negative effect for women is not at the year of birth of the first child but four years later, with estimated effects of −15 percent in τ = 0 and −24 percent in τ = 4. The smaller effect at the year of birth could be due to women having children in the later months of the year and thus having worked that year, albeit less. After ten years, employment variations are −17.3 percent for mothers and −31 percent for fathers, leading to a long-run child penalty of 14.2 percent.

Compared to findings from other countries reported by Kleven et al. (2019), child penalties in earnings in Canada are similar to those in the United States (31 percent) and to some extent the United Kingdom (44 percent), but they are larger than those in Denmark and Sweden (21 percent and 26 percent, respectively) and smaller than those in Austria and Germany (51 percent and 61 percent, respectively). We note however that the long-run penalties in employment are in line with those of Denmark (13 percent, as reported in Kleven, Landais, and Søgaard 2019). This finding suggests that the difference between Canada and Denmark in terms of child penalties in earnings does not stem from divergences at the extensive margin of the labour market (whether or not to work) but either at the intensive margin (how much to work) or the wage rate. However, as noted earlier, we are not able to investigate directly the response of the intensive margin since the longitudinal information coming from the tax files does not contain hours of work.

Estimated Impact of Children for Different Groups of Mothers

Figure 1 showed that the impact of children is substantial for women, while practically nil for men. In this subsection, we focus on mothers and investigate heterogeneity in the impact of children by showing how earnings and employment trajectories around the birth of the first child vary for different subgroups of mothers. It is interesting to note here that the lack of overall impact for fathers does not seem to mask heterogeneous responses but is really a reflection of the low impact of children on men's earnings and employment.¹³ Figures 2, 3, 4, and 5 all follow the same pattern, presenting the effect of children for subsamples of mothers, with effects on earnings shown on the top panel and those on employment on the bottom panel. In practice, we estimate Equation (1) with a full set of interactions; this is equivalent to estimating on separate subsets of the data. We do not show confidence intervals on those figures because even 90 percent confidence intervals often overlap, making the figures harder to read.¹⁴ This reflects a limit of our analysis: due to relatively small sample sizes, we often cannot rule out that the observed trajectories are statistically different. Our findings are thus reflective of general patterns, but statistical significance would need to be determined using a database with a larger number of observations. Unfortunately, large databases like the LAD do not contain sociodemographic information allowing for this type of heterogeneity analyses.

Figure 2: Earnings and Employment Trajectories for Mothers by Marital Status
Notes: The values shown on the y-axis correspond to changes as a percentage based on estimated coefficients in the base model. The results are weighted using the sampling weights provided by Statistics Canada. Top panel shows variations in earnings; bottom panel shows variables in annual employment.
Source: Calculations made by the authors using LISA data (2012, 2014) and T1 forms (1982–2019).

Figure 3: Earnings and Employment Trajectories for Mothers by Number of Children
Notes: The values shown on the y-axis correspond to changes as a percentage based on estimated coefficients in the base model. The results are weighted using the sampling weights provided by Statistics Canada. Top panel shows variations in earnings; bottom panel shows variables in annual employment.
Source: Calculations made by the authors using LISA data (2012, 2014) and T1 forms (1982–2019).

Figure 4: Earnings and Employment Trajectories for Mothers by Education Level
Notes: Low education is defined as high school diploma or less; high education is more than high school. The values shown on the y-axis correspond to changes as a percentage based on estimated coefficients in the base model. The results are weighted using the sampling weights provided by Statistics Canada. Top panel shows variations in earnings; bottom panel shows variables in annual employment.
Source: Calculations made by the authors using LISA data (2012, 2014) and T1 forms (1982–2019).

Figure 5: Earnings and Employment Trajectories for Mothers by Whether Born in Canada
Notes: The values shown on the y-axis correspond to changes as a percentage based on estimated coefficients in the base model. The results are weighted using the sampling weights provided by Statistics Canada. Top panel shows variations in earnings; bottom panel shows variables in annual employment.
Source: Calculations made by the authors using LISA data (2012, 2014) and T1 forms (1982–2019).

We start in Figure 2 by splitting the sample according to whether the mothers were married at the time of the birth of the first child or living in a common-law partnership. As we have seen in Table 1, an emerging phenomenon is the large and growing proportion of couples choosing to live in such a partnership rather than marry, especially in Québec. The top and bottom panels of Figure 2 show that the two groups have similar trajectories in both earnings and employment prior to the birth of the first child. At τ = 0, the year of the birth, the impact of children is also similar for both groups. In the subsequent years, however, trajectories diverge somewhat, with mothers in common-law partnerships having smaller negative effects both in employment and in earnings. This is coherent with a situation in which mothers in common-law partnerships are more active in the labour market than married mothers, possibly because marital status itself impacts women's behaviour, but also because characteristics which are unobservable yet correlated with marital status influence this association. Moreover, married women have more children, which may influence their participation in the labour market.¹⁵ We note however that the differences by marital status are not large enough to be statistically different due to the large standard errors.

Possibly confounded with the marital status at time of birth is the province of residence, since common-law unions are much more prevalent in Québec. To investigate whether the patterns in Figure 2 can be explained by the province in residence, we split the married and common-law mothers by whether they lived in Québec or in the rest of Canada at the time of birth. The results can be seen in Figures A.4 and A.5 of the Appendix. The trajectories for those in common-law unions appear very similar. For the married mothers, the trends are not quite as overlapping, especially for earnings: while similar up to four years after birth, the trajectories diverge starting in year five, with married mothers in Québec seeing a continuous decline until ten years after birth to reach a variation of −55 percent, while those in the rest of Canada see an improving trend, reaching −33 percent after ten years. The source of this divergence is not clear and may need to be investigated in further studies.

Another dimension that can influence personal trajectories is the number of children. As mentioned earlier, the long-term effect measures the total effect of children, including the effect of other children arriving at different moments in time following the birth of the first child. The number of children may be an important factor in the decision to return to the labour market. As Figure 3 shows, the negative impact of motherhood on both earnings and employment increases with the number of children.

More specifically, while pre-childbirth trajectories and variations on the year of the first child birth are similar for the different subgroups, mothers with only one child see the effect of children go to almost zero by the time their child is ten years old. By that point, the variations relative to τ = −2 are −6.9 percent (in earnings) and −2.1 percent (in employment) for mothers with one child, compared to variations of −36.9 percent and −13.1 percent for those with two children and −46.6 percent and −32.0 percent for those with three or more children, with statistically significant differences between those with one child and those with three or more children. The effect of motherhood thus appears to be cumulative with respect to the number of children, with mothers of multiple children being affected negatively for much longer periods than those limited to one child. This pattern is in line with the findings of Karademir et al. (2023) in Canada and Kleven (2022) in the United States, and they could possibly be a reflection of gender norms, as also argued in Kleven (2022): the same views that make people have more children could also be keeping women out of the labour force.

We then investigate heterogeneity by education level. In Figure 4, we present earnings and employment variations relative to two years before the birth of the first child by whether the mother has a low level of educational attainment (high school diploma or less) or a high level (more than high school). We can see that negative impacts are larger for mothers with a low level of education and that the differences between the two groups are especially pronounced when considering effects on employment. Indeed, while highly educated mothers have earnings variations that are between 2 to 13 percentage points smaller than mothers with at most a high school diploma, the gaps in employment variations range from 10 to 22 percentage points. Ten years after the birth of the first child, mothers with low education have employment levels 31.8 percent lower than two years before the birth, and the corresponding figure for highly educated mothers is 10.1 percent. Note that these differences are not statistically significant due to low sample sizes. Kleven (2022) finds similar patterns in the United States until year three (larger effects for low-educated women), but after that the differences by education level virtually disappear, contrasting with the Canadian case.

The final dimension along which we investigate heterogeneity is whether the mother was born in Canada or not. Figure 5 presents the earnings and employment trajectories according to the country of birth. We find that mothers born both inside and outside of Canada experience similar earnings and employment variations relative to two years before birth on the year of the birth. After that, however, the trajectories diverge slightly, with mothers born in Canada remaining at a lower level of earnings or employment relative to before birth. This gap remains until the end of our observation window, with mothers born in Canada showing a variation of −36.5 percent in earnings compared to −26.3 percent for those born outside of the country, and a similar gap for employment at −18.8 percent and −15.5 percent, respectively. We note that the group of immigrant mothers can be very diverse, hailing from countries with differing gender norms, which could lead to very different behaviours regarding children and labour market attachment (Kleven 2022). That being said, we need to be careful in our conclusions since the differences between the two groups are not statistically significant.

It is worth repeating that our estimates are imprecise due to the relatively low number of observations, and that consequently the majority of the differences between subgroups that we just presented are not statistically significant. However, we believe our findings are indicative of patterns that merit further investigation with a dataset that would yield more precise estimates. To summarize our findings in this section, we find that mothers that are married, have multiple children, attained a low level of education, and are born in Canada display the largest negative impact of motherhood—on both earnings and employment.

Family Policies and the Impact of Parenthood on Labour Market Outcomes

In this section, we provide suggestive evidence on the potential that family policies may have to reduce child penalties. We begin by explaining the implementation of the low-fee daycare services program in Québec and then look at changes to parental leave programs. We then present our methodology and findings.

Québec's Affordable Childcare Reform

In September 1997, Québec initiated a low-fee childcare reform. At the time, the cost per day was set at C$5 per child and services were open 10 hours a day, 261 days a year. The first spaces were made available for children four years of age only as of September 1997. The program was then rolled out year over year to make the network accessible to all preschoolers. Three-year-olds became eligible in September 1998, children aged two in September 1999, and children aged zero to one in September 2000. By 2000, the total number of subsidized spaces had reached 150,000 and over 220,000 spaces were available by 2012 (Haeck et al. 2015, Figure 1), covering over 65 percent of eligible children. The single rate of C$5 per day was increased on a few occasions but still remains low at C$8.70 in 2022. Childcare expenditures provide a tax credit that varies according to income. The subsidized fee is deductible at the federal level, meaning that the actual cost to parents is lower than C$8.70. Private childcare expenditures are deductible at both the provincial and federal level such that parents who opt for unsubsidized private childcare (by choice or due to the lack of spaces at preferred daycare services) pay a low price as well. As a result, the cost of childcare in Québec appears to be comparable to that of Austria, but operating hours are longer in Québec where childcare facilities must be open at least ten hours per day, relative to nine hours in Austria (Kleven et al. 2022).

This reform has had a major impact on the lives of children and their families. As mentioned earlier, nearly 60 percent of children aged one to four were in the care of their parents before the reform (in 1996), while 75 percent were in family-based care or center-based-care by 2008 (Haeck et al. 2015). Within ten years, the world of preschoolers in Québec was completely transformed. As described in our literature review, many studies showed that mothers responded to the reform by joining the workforce. Baker et al. (2008), Haeck et al. (2015), Lefebvre and Merrigan (2008), Milligan (2014) found that the introduction of this program had a significant positive effect on mothers’ employment. The year-round availability of low-fee childcare open ten hours a day enabled mothers to return to the labour market following the birth of a child. Haeck et al. (2015) estimate that the proportion of employed mothers increased 12 percentage points following the reform and that the increase has been maintained through the years. As a result, the reform directly contributed to supporting the attachment of mothers to the labour market and to ensuring that they were able to go back to the job they held before the birth of their child. It is therefore possible that this reform helped reduce child penalties in earnings and in employment.

Parental Leave Reforms

Following implementation of the reduced-contribution daycare services program, two parental leave reforms were introduced. First, in January 2001, to encourage work–family balance and ensure a better start to life for young children, the Canadian government modified the employment insurance program (Human Resources and Skills Development Canada 2005). In practice, the federal government improved parental insurance through the employment insurance program by increasing the length of paid parental leave from 10 to 35 weeks. As part of the reform, the eligibility criterion for parental leave was also lowered from 700 hours of insurable employment in the year preceding the birth to 600 hours. The rate of coverage for the additional weeks was set at 55 percent of the insurable annual income up to a threshold of C$39,000 in 2001 (Baker and Milligan 2008; Marshall 2003).

This reform, which extended leave time considerably, makes it much more likely that employment income during the year following the birth decreased compared with parents who did not enjoy the reform's benefits. Furthermore, parents—especially mothers—who found the leave too short prior to the reform were no longer forced to leave their job to continue caring for their children of less than one year. In this context, it is possible that the reform had an overall positive effect on the earnings trajectories of new parents, at least over the long term, by encouraging continuity in labour force participation. According to Marshall (2003), the proportion of women who return to the labour market after a long leave (9 to 12 months) rose from 8 percent in 2000 to 47 percent in 2001—just one year after the reform came into effect. A large majority of mothers who return to work quickly after giving birth are self-employed workers or ineligible employees.

Baker and Milligan (2008, 2010) use the 2001 Canadian parental leave reform to document the effect of extending maternity leave on the development and wellbeing of children. Their results show that—following the reform—mothers who took a leave spent from 48 percent to 58 percent more time away from the workforce during the first year of their child's life. This extension in the length of parental benefits may have had the effect of modifying the behaviour of mothers and fathers in the labour market. Benefits based on previous earnings encourage women to work before motherhood, and women entitled to paid leave are more likely to re-enter the labour market after childbirth (Rønsen and Sundström 2002). However, this study also shows that long leave entitlements are associated with adverse labour force re-entry effects (Rønsen and Sundström 2002). Baker and Milligan (2015) find little impact of the Canadian parental leave reform on maternal employment up to five years after childbirth, as does Milligan (2014).

Second, since 1 January 2006, parents in Québec have benefited from improved leave compared with the federal program under the Québec Parental Insurance Plan (QPIP). Several aspects distinguish QPIP from the federal program, starting with the addition of dedicated leave for the father and followed by improved benefits for both parents. A shorter leave option is also offered at a higher replacement rate. Haeck et al. (2019) review in detail the QPIP. All these reforms are part of a progressive societal movement aimed at getting away from the traditional gender roles by compensating mothers for income losses around the birth of a child and facilitating the return to work by providing affordable childcare.

Methodology

To provide suggestive evidence on the impact of family policies on child penalties, we use a difference-in-differences quasi-experimental approach in which we estimate Equation (1) while adding indicators for residing in Québec at the time of the birth of the first child (compared to the rest of Canada) and for a first child birth in 2001 and later (compared to before 2001), and a full set of interactions between the two indicators and the event time dummies.¹⁶ The Canadian parental leave reform was implemented in January 2001 while the Québec childcare program first started in September 1997, but it was not until 2001 that all children under the age of five were eligible for year-round daycare services at a reduced rate. As a result, our post-reform period starts in 2001. This method provides evidence on the intention-to-treat (ITT) effect since we are not able to use administrative files to identify parents whose children attended a subsidized daycare or those who benefited from parental benefits.¹⁷ The main limit of our study, as mentioned earlier, is the small sample size and wide confidence intervals as a result. As such, we present the difference-in-differences results as suggestive evidence rather than strong causal effects.¹⁸

Effects of Family Policies on Mothers’ Employment and Earnings Trajectories

Figure 6 presents the trajectories of variations in mothers’ earnings (top panel) and employment (bottom panel), with our sample split into four groups: mothers who lived in Québec at the time of the birth of the first child (series marked by blue triangles) and those who lived in the rest of Canada (red circles), each divided by whether the first child was born in 2000 or before (full lines) or in 2001 and later (dashed lines).

Figure 6: Earnings and Employment Trajectories for Mothers by Province of Residence at Child Birth and Pre-/Post-2001
Notes: The values shown on the y-axis correspond to changes as a percentage based on estimated coefficients in the difference-in-differences model. The results are weighted using the sampling weights provided by Statistics Canada. Top panel shows variations in earnings; bottom panel shows variables in annual employment.
Source: Calculations made by the authors using LISA data (2012, 2014) and T1 forms (1982–2019).

We start by focusing on mothers from the rest of Canada in order to assess the effect of the parental leave reform on the impact of motherhood. The pre- and post-reform earnings trajectories of mothers in the rest of Canada follow a parallel trend, and both sharply fall on the year of the birth, but slightly less so for births in the pre-reform period (−47 percent compared to −51 percent in the post-reform period). Over the long term, i.e., ten years after the birth, mothers in other provinces who had their first child before 2001 display a negative impact of 39 percent, whereas those whose first child was born in 2001 or later are penalized by about 33 percent, a decrease of 6 percentage points. When considering employment (bottom panel), a similar pattern can be observed, although the post-2001 curve is always above the pre-2001 curve, meaning that the losses in employment are always smaller after the parental leave reform, consistent with findings in Baker and Milligan (2008). Ten years out, the negative effects of children on employment are 23 percent and 16 percent for the pre-reform and post-reform periods, respectively, a difference of 7 percentage points. Although they never recover their pre-birth earnings or employment levels, Canadian mothers outside Québec regain more ground and remain more attached to the labour force in the post-reform period. The more generous federal parental leave program may have led to better job continuity after birth. Note that we find very similar trajectories when we split the rest of Canada sample by whether the mother was born in Canada or in another country, leading us to believe that the influx of immigrant mothers seen in Table 1 does not explain the trends seen post-2001 in Figure 6.¹⁹

Comparing the situation in Québec with that of the rest of Canada allows us to provide suggestive evidence on the effect of the subsidized childcare program (on top of that of the federal parental leave reform). The top panel of Figure 6 shows that the situation of Québec mothers before 2001 was comparable with those from the rest of Canada but that their earnings variations started changing in 2000 relative to women in the rest of Canada. By the fifth year after the birth, the difference between mothers in post-reform Québec and other groups becomes larger, with mothers having recovered part of the lost earnings. Ten years after the birth, mothers in Québec who had their first child before 2001 display a negative effect of children in the order of 40 percent, whereas those whose first child was born in 2001 or later are penalized by about 23 percent, a decrease of 17 percentage points. A simple difference-in-differences calculation yields an effect of children on earnings of 11 percentage points (17 points in Québec compared to 6 in the rest of Canada).

In terms of employment trajectories (bottom panel of Figure 6), the pre- and post-2001 patterns for Québec mothers are similar to those for the rest of Canada: the post-2001 curve is above that of the pre-2001 curve, meaning smaller variations in employment in the post-2001 period. After ten years, employment for mothers in Québec is down 23 percent during the pre-reform period, but only 8 percent in the post-reform period, a difference of 15 percentage points. When we compare this gain with that from the rest of Canada (7 percentage points), we infer a difference-in-differences estimate of 8 percentage points.

There are important limits to our interpretation. As previously mentioned, the confidence intervals are not shown because they overlap, making the figures hard to read. We are therefore not able to claim that the differences highlighted in the previous paragraphs are statistically significant. The small number of observations in the LISA greatly limits our ability to make strong statements, but our results suggest that Québec's family policies could have helped to reduce the negative effects of children on earnings by around 11 percentage points and those on employment by 8 percentage points. For a more definitive answer, a dataset with more observations is needed, one which is along the lines offered by Karademir et al. (2023).²⁰ Our results are coherent with theirs and with previous studies on the impact of the childcare policy on labour force participation (Baker et al. 2008; Lefebvre and Merrigan 2008). However, it is important to remind ourselves that the changes between the trajectories of Québec mothers before and after 2001 are potentially influenced by both low-fee childcare and the improved parental leave program.

Conclusion

To summarize, we began with a database that includes a longitudinal study coupled with historical administrative records covering the period from 1982 to 2018. We then made use of this database to analyze the effect of the arrival of children on the long-term earnings and employment trajectories of mothers and fathers separately. By using the same estimation strategy as Kleven, Landais, and Søgaard (2019), we found that mothers’ earnings fall drastically following the birth of the first child, with the average decrease for the year of birth estimated at about 49 percent. Over the long term, i.e., ten years after the birth, Canadian women continue to earn less than they did before the birth of their first child and less than fathers, with the average penalty estimated at 34.3 percent. A similar pattern is found when considering employment instead of earnings, but with smaller penalties, and a long-run penalty of 14.2 percent, a number that is close to the employment penalty estimated using data from Denmark (Kleven, Landais, and Søgaard 2019).

The arrival of children results in large losses that are not equally distributed within couples. Fathers do not appear to be adversely affected by children in terms of earnings and employment, whereas women experience penalties that persist over the long term, especially mothers of multiple children or those with a low level of education. This impoverishment triggered by the birth of the first child could have significant economic impacts should the couple separate (Belleau et al. 2023).

In this context, it appears crucial to concentrate on measures that could eliminate or at least reduce the economic impact associated with family responsibilities on mothers’ earnings and employment. To investigate such potential measures, we presented suggestive evidence on the effect of family policies designed to encourage maternal employment and promote more equal sharing of parenting responsibilities between partners starting in 2001, specifically the extension of parental leaves in Canada and the introduction of reduced-contribution daycare services for families in Québec. Using a difference-in-differences model, we found that Québec mothers were able to recover pre-birth levels of earnings and employment more quickly than mothers in the rest of Canada, with the caveat that our sample size does not allow us to come to firm conclusions regarding statistical significance. The effect of Québec's family policies is potentially considerable: the long-term earnings gap, i.e., ten years after the birth of the first child, was reduced by 17 percentage points in Québec, from −40 percent to −23 percent. In comparison, the gap for women in the rest of Canada dropped from −39 percent to −33 percent, which is clearly an improvement but not in the same order of magnitude as the progress observed in Québec. The net effect of Québec's family policies on earnings may therefore be around 11 percentage points. For employment, we find a similar pattern, resulting in an estimated effect of 8 percentage points.

This study highlights the importance of continuing to work on developing more inclusive family policies that could reduce the child penalties. Of particular concern are mothers with a low level of education and multiple children. However, this would need to be done without neglecting the quality of services offered for children, services such as a quality education program in daycare. Baker, Gruber, and Milligan (2019) and Haeck, Lebihan, and Merrigan (2018) agree that the introduction of reduced-contribution daycare services has on average not had positive long-term effects on children—and has had some short-term negative effects for preschool children. The quality of the daycare services network is generally low. In addition, the most vulnerable children are less represented in high-quality daycares (Haeck et al. 2015). The network needs to improve its quality, especially in the most underprivileged sectors. These policies therefore have effects in several areas of people's lives. In the short term, policies have an impact on women joining and remaining in the labour force. Over the long term, women's earnings trajectories appear to be positively influenced by these policies, which serves to reduce the gap between men and women. It should be noted that this study did not undertake a cost–benefit analysis that would allow us to compare the costs of implementing such measures with their benefits.

As Goldin (2014) highlights, the solution may not necessarily take the form of government intervention but may involve changes to the labour market. In particular, the way in which jobs are structured and valued in the labour market needs to be re-examined. Nowadays, both men and women want to spend quality time with their children, and a compensation model that allows for some flexibility therefore needs to be developed.

Finally, while it takes two to conceive a child, many couples eventually face a relationship breakdown. In Canada, nearly one third of marriages end in divorce (Statistics Canada 2022b). In this context, it would be interesting to analyze the effect of changes in marital status on post-birth earnings and employment trajectories in a future study. Indeed, literature on this topic again indicates that it is women who are financially disadvantaged following a separation (Belleau et al. 2023; Le Bourdais et al. 2016). The situation may be attributable to pre-separation factors such as the unequal division of labour during the marriage and lower earnings for women, but also to women's prolonged absences from the labour force due to family responsibilities.

In addition to having a positive effect on the economic situation of women, encouraging employment for mothers could also contribute to eliminating the stigma around the division of labour within couples by specifically exposing children to a more symmetrical model of remunerated and unpaid work. A recent study showed that active mothers are more likely to transmit egalitarian values to their children both at work and at home. McGinn, Ruiz Castro, and Long Lingo (2019) analyze the relationship between mothers’ employment and their children's behaviour when they become adults using data from 29 countries. They find that girls who had mothers who are employed ended up working more themselves than girls whose mothers did not work outside the home. Specifically, they worked more hours, were better paid, and held supervisory positions more often than girls with stay-at-home mothers. This result was not observed in boys. However, boys who grew up with employed mothers were more involved in family and domestic responsibilities than were men whose mothers were not in the labour market. The girls also spent less time doing household chores. We can therefore conclude that mothers in the workforce appear to have an intergenerational impact favouring gender equality both within the family unit and in the labour market.

Acknowledgements

This work is an extension of Connolly, Fontaine, and Haeck (2018, 2020) and the authors thank CIRANO for the opportunity to work on this project. The authors would like to thank the Fonds de recherche du Québec — Société et culture for their funding (grant number 2020-SE3-270241). Cristian Stratica provided superb research assistance. The authors also thank anonymous referees and participants at seminars at the University of Waterloo and the University of Ottawa for comments. All errors remain their own. The analysis presented in this paper was conducted at the Quebec Interuniversity Centre for Social Statistics, which is part of the Canadian Research Data Centre Network (CRDCN). The services and activities provided by the QICSS are made possible by the financial or in-kind support of the Social Sciences and Humanities Research Council (SSHRC), the Canadian Institutes of Health Research (CIHR), the Canada Foundation for Innovation (CFI), Statistics Canada, the Fonds de recherche du Québec - Société et culture (FRQSC), the Fonds de recherche du Québec - Santé (FRQS), and the Quebec universities. The views expressed in this paper are those of the authors and not necessarily those of the CRDCN or its partners.

Footnotes

Québec is the second-largest province in Canada by population, with around 8.7 million inhabitants in 2022 (Statistics Canada 2023).

The LAD contains longitudinal fiscal data and is representative of the Canadian population. The timing of birth can be identified so long as parents claimed child benefits at some point in the child's life. As of 1993, parents had a strong incentive to declare their child and claim benefits.

Likewise, Karademir et al. (2023) suggest that the role of grandparents in providing childcare can explain Kleven et al.'s (2022) findings.

To our knowledge, only Karademir et al. (2023) also study the short- and long-term effects of family policies on the family gap, but they use different Canadian data sets and focus on care provided by grandparents and grandmothers’ earnings and labour force participation.

The Luxembourg Income Study contains detailed microdata from around 50 countries covering five decades (Cross-National Data Center in Luxembourg 2023).

The International Social Survey Programme is a transnational collaboration running annual surveys on various topics (International Social Survey Programme 2023).

Appendix Figures A.2 and A.3 present trajectories for net self-employment income and for total employment income (T4 earnings plus net self-employment income and other employment income), respectively. Fathers appear to have higher self-employment incomes post-birth, while mothers experience a decrease, though not as sharp as when considering earnings (see Figure 1). Looking at self-employment, fatherhood does not appear to have much of an impact whereas motherhood does appear to boost the probability to engage in self-employment. This result is consistent with what Jeon and Ostrovsky (2019) find: new mothers are more likely to transition to self-employment than other women, a situation which comes with more flexibility in work hours. Note however that large confidence intervals make it difficult to make conclusions about the sign of the effects mentioned above: only the years one to five effects on self-employment for mothers are statistically different from zero at a five percent significance level (see Table A.9 in the Appendix for the bounds of the confidence intervals). Finally, in Figure A.3 we see that using total employment income yields findings very similar to those using T4 earnings only.

Given that pregnant women are eligible for maternity leave several weeks before the expected delivery date, earnings and employment are projected to decrease as early as the calendar year preceding the birth, τ = −1. Indeed, in the case of births occurring at the beginning of a year, the withdrawal from the labour market takes place during the calendar year preceding the birth. Historically, maternity leave can begin earlier in Québec than in other Canadian provinces relative to the expected delivery date. In Québec, a leave can currently begin 16 weeks before the expected delivery date, compared with 12 weeks in the rest of Canada (Canada 2023; Québec 2023). Moreover, pregnant women in Québec in certain occupations are eligible for preventative leave, a programme managed by the work safety commission that covers about 50 percent of births (Commission des normes, de l’équité, de la santé et de la sécurité du travail 2018; Takser 2013). Kleven, Landais, and Søgaard (2019), who had complete information concerning the children's date of birth and a considerably larger sample, demonstrate that in limiting their analyses to births occurring in January, a slight decrease in employment income was observed for the calendar year preceding the birth, a finding which supports our choice of year of reference.

Additional analyses (not presented here) showed that the use of standard errors clustered by individual or by province, as well as by the number of children per person, did not significantly change the value of the standard errors. For this reason, we opted to retain heteroscedasticity-robust standard errors, replicating the methodology used by Kleven, Landais, and Søgaard (2019).

Restricting to individuals between the ages of 25 to 51 does not alter our findings significantly; see Figure A.1 in the Appendix for a version of Figure 1 where individuals under age 25 are excluded.

Comparable statistics for fathers are presented in Table A.1 of the Appendix. The trends for fathers are generally similar to those for mothers, except perhaps for the proportion born in Canada for Québec fathers. For mothers, it goes from 0.92 to 0.93 between the pre- and post-reform periods, but for fathers it goes from 0.91 to 0.87. The source of this discrepancy is unclear. Figure 5 and Figure A.9 in the Appendix show earnings and employment trajectories for mothers and fathers, respectively, by whether they were born in Canada or not.

See Table A.2 in the Appendix for the complete sets of regression coefficients underlying Figure 1, i.e., before the computation of the variations in percentages as described in the Methodology section. See also Table A.3 of the Appendix for the values used to produce Figure 1.

See Figures A.6 to A.9 in the Appendix for the equivalent of Figures 2 through 5, but for fathers instead of mothers.

However, we do provide the lower and upper bounds of the 90 percent confidence intervals in Tables A.4 to A.7 of the Appendix.

In our sample, the average number of children of married women is 1.88, while it is 1.67 for women living in common-law partnerships.

Note that we do not include full interactions with the age and calendar year dummies due to the relatively small number of observations in the treated region (i.e., Québec, especially for the post-reform period), but we do keep the age and calendar year dummies separate. The inclusion of year dummies should alleviate concerns that the post-reform period indicator could simply be picking up a trend.

This information is available in the LISA data starting with Wave 2, but it concerns only children born during the reference period, i.e., between January and May 2014, thus excluding all births we study. According to Findlay and Kohen (2012), who use data from the 2010 Survey of Young Canadians, most mothers living in Québec and working the year before birth (97 percent) reported having taken paid leave of about 48 weeks. McKay, Mathieu, and Doucet (2016) show that the proportion of Québec mothers eligible for parental leave rose from 80.5 percent to 89.3 percent between 2007 and 2013, representing an increase of 8.8 percentage points over that period, while the proportion remained relatively stable (64 percent) in the rest of Canada over the same period. The relatively high use of parental leave in Québec ensures that we include in our sample a large proportion of parents likely to have benefited from the changes in Québec's family policies. Likewise, according to data from the Québec Ministry of Family (Ministère de la Famille 2023), 56.8 percent of children under the age of five attended a regulated daycare in 2013. In addition, estimating the effect on the total population rather than on those who actually benefited from the reform provides the advantage of potentially measuring the total indirect impact of the reforms

Karademir et al. (2023) present evidence of the causal effect of the reform using administrative data. In contrast to theirs, our data identify parents even if they did not file a tax report or did not claim benefits—though these are rare occurrences in the LAD according to the authors. Our results support their findings on the effect of subsidized childcare on the effects of parenthood.

See Figure A.10 in the Appendix for a version of Figure 6 where the line for rest of Canada is split in two groups: those born in Canada (in red) and those born outside of Canada (in green). While there can be small differences, notably (but not statistically significantly) on the year of the birth, the long-term variations are remarkably similar.

Karademir et al. (2023) study the multigenerational impact of subsidized childcare using the Longitudinal Administrative Databank (LAD), which contains millions of observations. They find that Québec's family policies seem to have helped reduce penalties in employment ten years after the birth of the first child by at least 6.1 percentage points on average (see Karademir et al. 2023, Figures A17 and C3).

Supplemental Material

cpp.2023-015_supplement1.pdf

Download
2.70 MB

References

Anderson D.J., M. Binder, and K. Krause. 2002. “The Motherhood Wage Penalty: Which Mothers Pay It and Why?” American Economic Review 92(2):354–58.

Format	RIS (ProCite, Reference Manager) EndNote BibTex Medlars RefWorks
Direct import

Résumé

Résumé

Abstract

Introduction

Selected Literature Review

Methodology

Data, Sample Construction, and Descriptive Statistics

Results

Estimated Effects of Parenthood for Mothers and Fathers

Estimated Impact of Children for Different Groups of Mothers

Family Policies and the Impact of Parenthood on Labour Market Outcomes

Québec's Affordable Childcare Reform

Parental Leave Reforms

Methodology

Effects of Family Policies on Mothers’ Employment and Earnings Trajectories

Conclusion

Acknowledgements

Footnotes

Supplemental Material

References

Information

Published In

Copyright

History

Keywords:

Mots clés :

Authors

Affiliations

Metrics

Related Content

Citations

View options

PDF

EPUB

Restore your content access

Share

Copy the content Link

Share on social media

Request Username

Create a new account

Change Password

Your password must have 8 characters or more and contain 3 of the following:

Password Changed Successfully

Verify Phone

Congrats!