Abstract
Care infrastructures are essential for supporting families and enabling women’s participation in the labor market, but they also have implications for family income inequality. This article examines access to childcare services in the United States as a case study. We propose that market-priced childcare systems generate inequalities in how births affect mothers’ contributions to family income, because they constrain post-birth labor supply for lower-income women more than for higher-income women, and aggravate family income inequality as a result. Using the Survey of Income and Program Participation (SIPP) merged with state-level childcare prices, we estimate individual fixed-effects regression models for the consequences of births on family income and its proximate determinants: mothers’ labor supply and earnings, and partners’ labor supply and earnings. We find that childcare prices increase post-birth earnings losses for mothers without college degrees, but not for mothers with college degrees, and these losses are not compensated for by increases in partners’ earnings or by income transfers. As a result, childcare costs exacerbate family income gaps between partnered women with and without a college degree by 34 percentage points.
Keywords
In the United States, nearly 4 million babies are born every year, each requiring 24/7 care and supervision for several years. The provision of care for newborns and young children up to school age requires a large amount of human labor that is performed by both paid and unpaid caregivers. The birth of a child often demands substantial reconfigurations of labor efforts in the household that affect family income. When someone, disproportionately the mother, drops out of the labor force to provide care for a newborn, the family loses a source of income. Access to childcare services can help families avoid income losses that come from reductions in paid work. However, when access to childcare services depends on purchasing power, as is the case in primarily market-priced childcare systems like in the United States, higher-income families can more easily rely on childcare services to avoid income losses, and this might exacerbate family income inequality.
In the context of rising family income inequality, care infrastructures and the economic consequences of births remain unexplored mechanisms. Existing research about the drivers of family income inequality has focused on labor market processes, tax and welfare policies, and demographic family-formation processes (e.g., Brady, Blome, and Kleider 2017; McCall and Percheski 2010; Parolin and Gornick 2021; Schwartz 2010; Sudo 2017; Western et al. 2016). Poverty research has long associated the number of children with poverty risks, but the focus has been more on the increase in resource needs associated with children than on the economic effects of having children (Brady, Finnigan, and Hubgen 2017; Rainwater and Smeeding 2005). However, the fact that births leave long-lasting marks on women’s earnings (Juhn and McCue 2017; Killewald and García-Manglano 2016; Kleven, Landais, Posch, et al. 2019; Musick, Bea, and Gonalons-Pons 2020), and that women’s earnings constitute a substantial and growing portion of families’ incomes (Bloome, Burk, and McCall 2019; Glass, Raley, and Pepin 2021), suggests the economic consequences of births may be important determinants of families’ economic positions. Furthermore, if the effect of births on families’ economic position is unequal, with more negative effects for lower-income families than for higher-income families, births might constitute an underappreciated mechanism contributing to increasing family income inequality.
An extensive body of scholarship documents the negative effects of births on mothers’ economic outcomes and the role of work–family policies in shaping birth penalties (for recent overviews of this literature, see Ferragina 2020; Hegewisch and Gornick 2011; Olivetti and Petrongolo 2017), but we know less about the effect of births on family income. Birth penalties on mothers’ earnings might translate into income losses for the family, but not necessarily. Indeed, mothers’ earnings losses could be offset with additional sources of income (e.g., increased partners’ earnings or income transfers). In one of the few analyses directly focused on the effect of births on family income in the United States, Stanczyk (2020) found that household income adequacy declines with births, indicating that birth penalties on mothers’ earnings translate to declines in family income. Thus, it does not appear that other sources of income compensate for mothers’ labor earnings losses. In particular, fathers’ earnings do not substantially increase with births (Killewald and García-Manglano 2016; Kleven, Landais, Posch, et al. 2019; Musick et al. 2020; Musick, Gonalons-Pons, and Schwartz 2022).
In this article, we argue that care infrastructures and birth penalties contribute not only to the production of gender inequality, but also to the production of family income inequality. To articulate this relationship, we offer a conceptual framework that focuses on how care infrastructure shapes the economic tradeoffs of replacing mothers’ unpaid care labor. We define “care infrastructure” as policies that shape the availability and costs of care services, as well as policies that deliver income transfers for unpaid caregivers. 1 We empirically evaluate this argument by examining the U.S. market-priced childcare system. We expect that market-priced childcare services exacerbate inequalities in how births affect mothers’ economic outcomes, and that these inequalities reverberate in the distribution of family income. More precisely, we expect that more expensive childcare disproportionately discourages post-birth paid work among women with lower earnings potential, and that this leads to substantial income losses for families in the lower half of the income distribution. In the absence of processes offsetting the negative effect of childcare prices on mothers’ income contributions to families, this mechanism will result in an increase in family income inequality.
We use individual fixed-effects regression models to estimate how births and childcare costs affect family income, examining major proximate determinants of family income: mothers’ labor supply and earnings, and partners’ labor supply and earnings. We use data from the Survey of Income and Program Participation (SIPP) and time-varying, state-level information on average childcare prices. Our main analyses focus on first and second births to women partnered with men; we perform supplementary analyses with the broader sample of all first and second births (which includes births to unpartnered women and to women partnered with women). Our models control for individual and state fixed effects. Individual fixed effects allow us to compare earnings before and after birth for the same woman. Given state fixed effects, we are explaining an individual mother’s earning losses as a function of the variation in childcare costs over time within a state: for example, if a mother without a college degree in Pennsylvania gives birth at a time when Pennsylvania’s childcare costs are lower, we expect her to have lower post-birth earnings losses, and thus higher family income, compared to observationally equivalent mothers in Pennsylvania giving birth at a time when childcare costs are higher.
Our results show that childcare costs fuel inequalities in women’s labor supply responses to births and exacerbate family income. In our analyses focused on partnered women, we find that a $1,000 increase in the annual price of childcare (representing a 7 percent increase over the mean childcare price of $13,4002) is associated with a 0.5-hour decline in weekly work hours and an 8 percent decline in monthly earnings for women without college degrees. In contrast, higher childcare costs have no statistically significant effect on work hours or earnings for women with college degrees. Childcare costs also have no statistically significant effect on women’s partners’ work hours or earnings. Taken together, we find that childcare costs exacerbate family income inequality, with each additional $1,000 increase in the annual price of childcare being associated with a 2.3 percent decline in family income for women without a college degree (the average decline is 16 percent for this group).
Our study makes three key contributions. First, we make a theoretical contribution by developing a framework that articulates the relationship between care infrastructure and family income inequality. In so doing, we highlight a new dimension of the institutional drivers of family income inequality. Second, our research design improves the identification strategy to estimate the relationship between childcare costs and the economic penalties of births, by relying on panel data and independently collected data on childcare prices. Finally, our study expands our understanding of the economic effects of births, broadening the current focus on mothers’ economic outcomes to study births’ effects on family income and on family income inequality.
Background
The existing literature about birth penalties on mothers’ economic outcomes suggests that having children might significantly affect families’ economic positions, but family income has not been a focus in this literature. Studies show that births leave long-lasting marks on mothers’ economic outcomes (employment, hours of paid work, wages, earnings) but have no significant effects for fathers’ economic outcomes, providing little indication that men’s earnings could offset mothers’ earnings losses (Aisenbrey and Fasang 2017; Budig and England 2001; Budig and Hodges 2010; Cooke 2014; Doren 2019; England et al. 2016; Florian 2018; Gangl and Ziefle 2009; Killewald 2013; Killewald and García-Manglano 2016; Musick et al. 2020).
Studies have also found that social policies such as parental leave or public childcare can reduce the size of birth penalties for mothers’ economic outcomes (Budig, Misra, and Boeckmann 2012; Landivar, Ruppanner, and Scarborough 2021; Landivar, Scarborough, et al. 2021; Ruppanner et al. 2021; Washbrook et al. 2011), and there is some evidence that these policies have a stronger effect on some groups of women more than others (Del Boca, Pasqua, and Pronzato 2009; Hook and Paek 2020; Keck and Saraceno 2013). Hook and Paek’s (2020) cross-national comparative study, for instance, found that the association between countries’ spending on early childhood education and the difference between mothers’ and non-mothers’ employment is stronger for mothers without a college degree than for mothers with a college degree. This finding is consistent with research on childcare costs, which indicates that childcare costs worsen birth penalties for mothers with lower earnings potential more than for mothers with higher earnings potential (Anderson and Levine 1999; Baum 2002; Kaestner, Lubotsky, and Qureshi 2016; Landivar, Scarborough, et al. 2021; Ruppanner et al. 2021; Tekin 2007). 3
Despite the pattern of findings pointing in a direction consistent with the idea that birth penalties and the care infrastructure can shape family income inequality, we do not have a conceptual framework nor an empirical record to precisely quantify this relationship. While family earnings losses can be implied from studies on motherhood penalties, direct estimates of the effects of births on family income are often lacking. Furthermore, research on income sources that could potentially offset the losses of mothers’ earnings at the family level has not been as comprehensive as the research on penalties on mothers’ outcomes. The effect of births on family income has been the focus in only a handful of studies (Aassve, Mazzuco, and Mencarini 2005; Bould, Crespi, and Schmaus 2012; Sigle-Rushton and Waldfogel 2007), with just one of these in the United States (Stanczyk 2020). These studies document cross-national variation in the average effect of births on family income or well-being, but they do not focus on family income inequality nor on care infrastructure. To conceptualize these relationships, the next section introduces a framework that explicitly articulates how birth penalties shape family income inequality depending on the features of the care infrastructure.
Care Events, Care Infrastructure, and Family Income Inequality
We propose a replacement costs framework to articulate the role of care infrastructure and births in the production of family income inequality. The purpose of this framework is to center the intensity of structured constraints on mothers’ labor as key determinants of birth penalties. This framework conceptualizes births and childcare as part of society’s care needs (Folbre 2012; Folbre and Wright 2012) and is grounded in the tradition of social reproduction theory that highlights the volume of human labor needed to sustain and care for human beings (Bhattacharya 2017; Laslett and Brenner 1989).
Our framework begins with the premise that a central challenge that arises with births is a labor challenge: how to meet the need for human care labor demanded by newborns. In contrast to conceptual models of mothers’ labor supply that emphasize mothers’ attributes, work orientations, or their job opportunities as predictors of labor supply behavior, our framework focuses instead on the constraints and economic tradeoffs implied by different ways of meeting the care labor demands that arise with births. Newborns’ need for 24/7 care and supervision for several years represents a substantial increase in care labor demand that needs to be met by someone: often the mother, paid caregivers, partners, relatives, or a combination of these. Characterizing births as care labor demand events connects births to other episodes that pose similar labor challenges because they likewise increase individuals’ responsibility to meet care needs, such as children or adults developing disabilities or recovering from accidents or surgical interventions. Our replacement costs framework can be applied to model individuals’ employment behavior after births as well as after other episodes that give rise to care labor demand.
The replacement costs framework models mothers’ labor supply as a function of the economic tradeoffs involved in replacing their unpaid care labor, because gender norms and structures assign primary caregiving responsibility to mothers (Connell 1987; Federici 2004; Folbre 2018; Glenn 2010). In absence of gender norms, the default unpaid caregiver could be gender neutral, but existing evidence shows that gender plays a major role guiding childcare decisions. Extensive research indicates that gender, more than economic rationality, dictates who takes on caregiving responsibility (Bittman et al. 2003; Dunatchik 2023; Killewald and García-Manglano 2016; Kleven, Landais, and Sogaard 2019; Musick et al. 2020), and that childcare arrangements are evaluated against mothers’ unpaid labor as the default (Damaske 2011; Rao 2020; Stone 2008). 4
Our model expects mothers’ labor supply to be the key driver of birth penalties on earnings and family income, and that birth penalties will be larger when the ability to find someone other than the mother to care for newborns is very limited or not cost effective. When replacements for mothers’ unpaid labor exist, our model expects birth penalties to be importantly determined by the costs of these replacements. If procuring replacement is expensive, adopting this option will generally be cost effective only for individuals whose rewards from paid work are high enough to offset the costs of replacement. In other words, all else being equal, our model expects that average birth penalties will decline when replacement costs are low and increase when replacement costs are high.
Consider the following simple model to understand the relationship between a woman’s earning potential and her decision to work after childbirth. Childcare costs is p per hour and mother’s wage is w per hour. A working mother’s income y can be defined as
if she works h hours; if free childcare is provided by the government or relatives, then p = 0. Note that childcare costs p act as a tax on wages, implying that take-home pay is lower than it would be in the absence of childcare costs. Assume there is a care income b (i.e., a policy income transfer program for individuals with care responsibilities). The income for a mother who is not employed is given by y = b. If no such care income exists, then b = 0, and the non-employed mother receives no income.
We focus on monetary incentives and take other factors shaping mothers’ decisions about work as given. It is clear that if
Based on this condition, we see that when the cost of childcare p increases or the caregiver’s income b increases, it is more likely for any particular woman to have a low enough wage to satisfy the non-working condition above, that is, to be in a situation where it is more financially advantageous to meet the newborn’s care labor demand herself than to find a substitute for her unpaid care labor and be employed. 5
Our model focuses on mothers’ wage-earning potential, but other factors can also shape the effects of childcare costs on birth penalties, including mothers’ partnership status and the availability of informal caregivers. In general, the work disincentive introduced by childcare costs will be stronger when families’ incomes depend less on mothers’ earnings. That is, mothers with employed partners will generally face a stronger childcare cost disincentive than will unpartnered mothers whose main income source is their own earnings. However, high childcare costs can still lower the earnings of unpartnered mothers or mothers who rely on informal caregivers. High childcare costs may lead unpartnered mothers to reduce work hours or choose lower-paying jobs that are more flexible or closer to kin who can help with childcare (Tekin 2005). Research shows that informal childcare is often less stable and reliable than formal paid care (Gordon, Kaestner, and Korenman 2008; Shattuck 2022; Tekin 2007), and that informal childcare arrangements are often utilized as alternatives to unaffordable formal paid childcare, rather than being mothers’ first preference (Chaudry 2004; Henly and Lyons 2000; Huston, Chang, and Gennetian 2002; Rachidi 2016). 6
The replacement costs framework offers straightforward predictions about the effects of the care infrastructure on birth penalties. Two types of care infrastructure policy directly intervene into replacement costs: (a) policies that shape the availability and prices of paid caregiver services (determines the parameter p in the simple model above), and (b) policies that provide income transfers for unpaid caregiving (determines the parameter b in the simple model above). By structuring the feasibility and tradeoffs associated with replacing unpaid care labor, care infrastructures shape the distribution of birth penalties across families and determine whether or not births result in increases in family income inequality.
Care infrastructures with high replacement costs (p is high, b is low) will tend to increase family income inequality. High replacement costs due to expensive paid care services will tend to hamper mothers’ labor supply in lower-income families more strongly than in higher-income families, resulting in systematically larger birth penalties in the former compared to the latter. High replacement costs due to no or low-income transfers for unpaid caregivers will lead to lower family incomes when mothers do not work. By contrast, care infrastructures with low replacement costs (i.e., p is low or b is high) will tend to decrease or have no effect on family income inequality. In a low replacement costs context, the size of birth penalties will tend to be less structured by family income, because everyone can equally access paid caregiver services to substitute for unpaid care labor or receive income transfers for unpaid caregiving that offset earnings losses from reducing one’s paid work labor supply.
The U.S. Care Infrastructure
The U.S. care infrastructure is characterized by high replacement costs. Access to childcare services is among the most expensive among OECD countries (OECD 2020), and there is no comprehensive policy guaranteeing affordable childcare access for all families (Chaudry et al. 2017; Cooke 2011; Folbre 2012; Gornick and Meyers 2003; Morgan 2006). Historically, attempts to develop comprehensive childcare policy have failed in large part due to efforts to protect the male-breadwinner family model and resistance to increased government spending on social services, especially for poor and non-White families (Danziger Halperin 2020; Michel 1999; Swinth 2018). 7 Furthermore, the United States does not have any substantial income transfer program for unpaid caregivers. There is no comprehensive paid leave policy, and the income transfer programs that exist for families with children are generally conditional on mothers’ paid work (e.g., from the perspective of single mothers, both EITC and TANF are conditional on employment and thus cannot be considered as transferring income for unpaid caregiving) (Cancian and Danziger 2009).
Since the 1970s, government intervention in childcare provision has been relatively marginal and resulted in a two-tiered market system of childcare (Folbre 2012; Michel 1999; Swinth 2018). For low-income families, the federal government provides childcare subsidies and the Head Start program. Neither of these policies, however, comes close to reaching all low-income families, with estimates suggesting that only 9 to 12 percent of eligible families access childcare subsidies (Ullrich, Schmit, and Cosse 2019). For families with labor earnings, the federal government provides childcare tax credits that only partially offset the costs of purchasing childcare services. Some states and cities have begun to roll out affordable pre-kindergarten programs, but these tend to only focus on 3- and 4-year-olds (Chaudry et al. 2017). Overall, this patchwork government involvement in childcare results in highly unequal and volatile childcare markets (Bouek 2022), and costly, privately-provided care services are the main source of formal childcare for most families. In practice, this means the cost of childcare is prohibitive to many families, thus constraining many families to rely on unpaid caregiving for young children.
Additionally, the cost barrier to paid caregiver services has become worse in recent years. Analyses indicate that childcare prices increased substantially between the mid-2000s and mid-2010s (Herbst 2023; NCES 2018), and high childcare costs are the most cited reason for parents who report having difficulty finding care (Cui and Natzke 2021). Such increases in childcare costs could be related to declines in the use of paid childcare among employed mothers since the mid-1990s (Laughlin 2013; Swenson and Simms 2021). Importantly, growing childcare costs have co-occurred with increases in family income inequality (Eika, Mogstad, and Zafar 2019; Gonalons-Pons, Schwartz, and Musick 2021).
Our Study
We focus on birth penalties and childcare prices to provide an empirical demonstration of the relationship between births and family income inequality as delineated by the replacement costs framework. Conceptually, we use childcare prices as the operationalization of replacement costs, that is, childcare prices as a synthetic measure of how the U.S. care infrastructure shapes the general accessibility and affordability of replacing mothers’ unpaid labor. While this operationalization is obviously a simplification, we find it justifiable in a context in which access to formal paid care is largely unsubsidized and no care income is available (i.e., parameter b above is 0).
We expect market-priced childcare provision to produce inequalities in how women respond to birth events and how this affects families’ incomes. More specifically, we hypothesize that childcare costs will disproportionately constrain post-birth paid work for women with lower earnings potential but less so for women with higher earnings potential. We expect that these earnings losses will not be offset by other sources of family income, either partners’ earnings or income transfers. As a result, because women with lower earnings potential are disproportionately in lower-to-middle-income families, market-based childcare will generate larger earnings losses for lower-income families compared to higher-income families and increase family income inequality.
Our analysis is structured in two parts. The first section presents descriptive evidence to examine the relationship between births and family income inequality. We track how family income inequality changes around birth events and explore how these changes relate to women’s income contributions to their families. The second section investigates the relationship between childcare costs and birth penalties on families’ incomes (including transfers), and how this relationship differs between mothers with high versus low earnings potential. In this section, we also examine the drivers of the effects of childcare costs on family income, by analyzing mothers’, and their partners’, labor supply and earnings.
Data and Methods
Data Sources and Samples
Our primary data source is the Survey of Income and Program Participation (SIPP) (U.S. Census Bureau 2001). The SIPP is a nationally representative household panel survey that began in 1984 and was designed as a continuous series of independent national panels with interviews every four months for up to five years. The SIPP has a large sample size—each panel includes 40,000 to 50,000 families since 1996—and tracks detailed monthly information about household and family composition, employment, and earnings of all adult members, as well as other income sources. The analyses use data from the 2008 and 2014 panels for which we can obtain state-level childcare prices. In supplementary analyses, we also use panel data from the Current Population Survey (CPS) and county-level childcare prices data from the National Database of Childcare Prices (Brown et al. 2023) to address the concern that state-level measures mask relevant within-state variation (see Tables S7 and S8 in the online supplement).
Our analytic sample includes women age 15 to 45 who gave birth over the course of the panel survey, including first- and second-order births (see the online supplement for details on construction of the analytic sample). The results presented in this article focus on the subsample of women partnered with men (N = 1,226), which represent the majority of observations around first and second births in our sample (70 and 95 percent of non-college-educated and college-educated women, respectively, are observed partnered at the time of birth). Supplementary analyses using the broader sample of all women with first and second births, including women partnered with women and unpartnered women, are available in Tables S4, S5, and S6 in the online supplement. Regression analyses remove person-months observations for the four months preceding the birth event to avoid contaminating our measures of pre-birth economic position with pregnancy-related changes.
Measures
We examine the relationship between birth events and the following five outcome variables: family income, mothers’ monthly earnings and hours of work, and their partners’ monthly earnings and hours of work.
Birth transitions are measured using a dummy variable that equals 0 before the birth and 1 after the birth. This measure draws on year and month birth-date information from all own children living in the household over the course of the panel. Respondents are identified as having a birth event when we observe the birth of the first or second child during the panel. When respondents experience more than one birth during the panel, we choose the first birth as the focal birth event and include a control variable that identifies person-months observations after a subsequent birth (this measure equals 0 for all observations before a subsequent birth and 1 thereafter). Controlling for subsequent births allows the outcome variable to be affected by additional birth events while simultaneously capturing the enduring effect of the focal birth event.
Childcare prices is a continuous variable indicating the state-level average price for center- and family-based childcare services for 0-to-3-year-old children at the time of birth. This variable is fixed at the individual level and only varies between individuals. Data come from the advocacy group Childcare Aware, which provides the longest-running public data on state-level childcare prices in the United States. Their data are released in annual reports on their website (https://www.childcareaware.org/) and have been used in related research (Landivar, Ruppanner, and Scarborough 2021; Landivar, Scarborough, et al. 2021; Ruppanner et al. 2021; Ruppanner, Moller, and Sayer 2019). Childcare Aware collects data primarily from a survey to childcare resource and referral agencies in each state and from market rate surveys that are completed by states to obtain federal funds from the Child Care and Development Fund (CCDF). The survey’s key informants are usually experts in government-affiliated childcare resource and referral agencies who help families navigate childcare providers and subsidies, and regularly research childcare markets in their state. These data reflect providers’ official prices instead of what families end up paying after subsidies or vouchers. Supplementary analyses confirm the reliability of this database (see Figure S1 in the online supplement). Childcare prices are adjusted for inflation and expressed in 2023 dollars.
Family income (log) is a continuous variable that measures monthly pre-tax family income. This measure includes earnings from labor and income transfers (e.g., TANF, food stamps, or unemployment benefits) from all adult members of the family, including nonmarital cohabiting partners. This measure also includes income transfers linked to children in the family, such as the Supplemental Security Income payments received by children under age 15. Family income is adjusted for inflation and expressed in 2023 dollars.
Earnings (log) is a continuous variable that measures monthly pre-tax earnings from labor, including self-employment and business income. Earnings are adjusted for inflation and expressed in 2023 dollars.
Hours of paid work is measured as a continuous variable that ranges from 0 (not employed) to 90. We use information on usual weekly hours of work as a synthetic measure of labor supply on the intensive and extensive margins.
Women’s earnings potential is a time-fixed dummy variable indicating completion of a four-year college degree (0 = no college degree, 1 = college degree). We use women’s level of education as a proxy for earnings potential because it allows us to identify groups that vary in their long-run earnings potential.
Individual-level controls. Models include time-varying controls for mother’s age, number of adults in the household, and subsequent births. The variable number of adults in the household controls for potential co-residing informal unpaid caregivers. Unfortunately, the SIPP does not include information on potential informal unpaid caregivers outside the household, a limitation we return to in the Discussion section.
State-level controls. We construct measures for state-level, time-varying characteristics that could confound our estimates of the relationship between childcare prices and birth penalties. We measure state-level, year-to-year changes in GSP per capita, women’s employment level, and the average wages of women without and with a four-year college degree. The GSP per capita measure comes from the Bureau of Economic Analysis (http://www.bea.gov). Data on women’s employment and wages is calculated using the CPS-ASEC files (Flood et al. 2021). Women’s employment level is measured as the proportion of working-age women employed, and average wages are calculated based on the sample of full-time, full-year women workers.
Methods
We use individual-level fixed-effects regression models to examine how birth events affect families’ incomes, mothers’ earnings and labor supply, and their partners’ earnings and labor supply. These models only leverage within-individual variation in the outcome variable, thus controlling for time-constant unobserved differences across individuals that could confound the relationships of interest, such as unobserved differences in specific labor market skills that are not captured by education. Our analyses focus on estimating how birth penalties are moderated by childcare prices and testing whether this moderation varies between mothers without a college degree and mothers with a college degree. Our model can be formalized as follows:
where Yismy is an economic outcome (e.g., family income) for individual i in state s, month m, and year y. β1 is the birth indicator capturing within-person changes in the outcome variable before versus after the birth of a child for the baseline group, women without a college degree. β2 is a coefficient for the interaction between the birth event and childcare costs for the baseline group, indicating how the effect of birth varies with childcare costs for women without a college degree. β3 is a coefficient for the difference in birth penalties between women without and with a college degree. β4 is the triple-difference coefficient testing whether the effect of childcare costs on birth penalties varies by mothers’ education (COLLEGE). β4 is our key coefficient of interest, and we expect it to be positive for family income and mother’s earnings and labor supply models; a positive β4 value indicates that childcare costs have a smaller effect on birth penalties for women with a college degree compared to women without a college degree. Notice that the model does not include the main effect for women’s college degree or for childcare costs because these are time-fixed characteristics. βj are coefficients for time-varying, individual-level control variables (Xjismy): age, number of adults in the household, and subsequent births. βr are coefficients for time-varying, state-level control variables (Zrsy), such as state GSP or women’s average wage. ∂s, φy, and θm correspond to state, year, and month fixed effects, respectively. δi indicate individual fixed effects, and εismy is the individual error term.
For ease of interpretation, we calculate and report key quantities of interest instead of presenting the full set of coefficients as they appear in the regression output, which can be difficult to interpret. The key quantities of interest are (a) β1 = birth penalty for women without a college degree, (b) β2 = how childcare prices moderate birth penalties for women without a college degree, (c) β1 + β3 = birth penalty for women with a college degree, (d) β2 + β4 = how childcare prices moderate birth penalties for women with a college degree, and (e) β4 = the test for the difference between the childcare prices moderation of birth penalties for the two groups of women.
This model estimates key coefficients of interest solely leveraging within-individual, within-state over-time variation. That is, the coefficient for the interaction between births and childcare costs is estimated by evaluating how within-individual differences in outcomes pre- versus post-births vary between women who have births in different years in the same state, as childcare prices evolve over time. For example, it compares the change in earnings for two observationally identical women in California, one giving birth in 2008 and the other in 2016, with childcare prices being 27 percent higher in real terms in 2016 than in 2008. 8 It is important to note that despite our measure of childcare prices ignoring real price heterogeneity between families and within regions, our identification relies on the assumption that families in the state face a similar childcare price change, as the average childcare price, not a similar absolute price. This simplifying assumption improves on cross-sectional analyses that assume all families face the same price as the regional-level average, and it is necessary because we do not have reliable and consistent state-level price heterogeneity measures covering our study period. 9
The models provide unbiased estimates of how childcare costs moderate the effect of births if specific assumptions hold. The estimates net out time-invariant unobserved characteristics at the state level and control for several key time-varying state characteristics, but they can be confounded by unmeasured time-varying characteristics correlated with variation in childcare costs. Remember that we are comparing how childcare costs affect birth penalties for two observationally identical women in the same state, but who gave birth at different times. Thus, bias could arise from unaccounted for changes over time in either the characteristics of the women giving birth or in the socioeconomic environment in the state. Specifically, bias would arise if these unaccounted for changes were correlated with changes in childcare costs. For instance, if increasing childcare prices were correlated with increasing employer discrimination against mothers in nonprofessional jobs, the three-way interaction could be upwardly biased because it would pick up larger declines in post-birth employment for mothers without a college degree compared to mothers with a college degree, resulting from both increasing childcare prices and employer discrimination. Similarly, if changing attitudes about motherhood and employment correlated with changing childcare prices, this could also confound the key estimates of interest.
Additionally, if selection into having births changes over time and is correlated with childcare prices, this could also affect our estimates. This correlation could be accidental, or it could be causal if childcare prices contribute to changing selection patterns. Rising childcare prices, for instance, might discourage fertility among women with disadvantageous labor market traits and make births more positively selected. Conversely, rising childcare prices might discourage fertility among women with advantageous labor market traits and make births less positively selected. 10 If childcare costs shape selection patterns, our estimates of the effects of childcare costs are still valid for the population of women who give birth, but the estimates may not represent the cost to individual women who may select into having a child when childcare costs change.
Results
Descriptive Evidence
Table 1 presents descriptive statistics for the analytic sample. The sample includes 1,226 partnered women who gave birth during the panel and 44,059 person-months observations. Of these births, 48 percent are first-order births, and the remainder are second-order births. Women’s average time in the sample is 38 months. On average, women are 30 years of age when they first enter the panel, and 50 percent hold a college degree. Descriptive statistics for economic outcomes before versus after births show well-known patterns. Women’s labor supply drops substantially after birth events, whereas men’s slightly increases. Women’s average monthly earnings are lower after birth events, whereas men’s are substantially higher. The average family income is slightly higher after births. The average childcare price was $12,008 in 2008 and $13,524 in 2016 in 2023 U.S. dollars. The within-state over time standard deviation in childcare prices ranges from $168 in Arkansas to $5,713 in Washington, DC.
Descriptive Statistics for the Analytic Sample
Source: SIPP 2008 to 2016.
Note: Earnings, income, and childcare prices are expressed in 2023 dollars.
Figure 1 provides descriptive evidence for the general claim that care events can leave a mark on family income inequality. This figure plots income inequality (measured with the coefficient of variation) for families who experience a birth during the panel over months preceding and following the birth event. The coefficient of variation is a measure of dispersion, and it is calculated by dividing the standard deviation by the mean; higher values indicate larger deviations from the mean, and thus more inequality. Because inequality measures can be sensitive to outliers, we follow standard practice and top-code the top 3 percent of the family income distribution. We obtain similar results if we use the Gini coefficient instead (see Figure S2 in the online supplement).
Family Income Inequality around Birth Events
Figure 1 shows that family income inequality increases substantially with births, beginning its increase during pregnancy months (−9 to −1), peaking at the month of birth, and declining somewhat over the first few months following the birth but remaining stable thereafter at a level higher than before birth. Family income inequality increases by 4 percent between months −12 and +12, which is a substantial increase. Eika and colleagues (2019, see Figure 11, p. 2821), for instance, estimate that family income inequality increased by 30 percent between 1980 and 2013, meaning the size of the increase in family income inequality associated with births is about 13 percent the size of the increase in family income inequality over three decades. The increase in family income inequality during pregnancy likely reflects pregnancy adjustments in labor supply and family composition that might be related or unrelated to childcare prices.
Figure 1 also displays simulated counterfactual trends showing how family income inequality would evolve if women’s earnings remained constant. To generate this trend, we calculate each woman’s maximum monthly earnings for the period before birth, and we assign this value to all observations after the birth. This exercise indicates that changes in family income inequality following births strongly reflect changes in women’s earnings, as opposed to reflecting changes in men’s earnings or other income transfers families might receive. In particular, changes in family income inequality are strongly related to changes in earnings among women without a college degree. If the earnings of women without a college degree remained at their pre-birth levels, family income inequality would not increase at birth (see Counterfactual 1). The same does not hold for women with a college degree. If the earnings of women with a college degree remained at their pre-birth levels, family income inequality would still substantially increase (see Counterfactual 2).
This descriptive evidence shows that births have a marked effect on the distribution of family income among families experiencing a birth. While it is generally recognized that birth events come with major shifts in women’s labor supply and earnings and play a key role in structuring gender inequality, the relationship between births and family income inequality has not been explored. The next section examines whether childcare prices shape birth penalties for mothers with and without a college degree in ways that might contribute to this increase in family income inequality.
Models for Childcare Prices and Family Income
We begin by presenting models on the core outcome variable of interest: family income. Table 2 reports coefficients for birth penalties and how childcare prices moderate birth penalties for women without and with a college degree, as well as the three-way interaction term that tests for differences in the moderation coefficient between the two groups of women.
Panel Regression Analyses of Birth Penalties on Family Income (Logged)
Source: SIPP 2008 to 2016.
Note: Table 2 reports key coefficients of interest from a regression model estimating the effect of births and childcare prices on family income (see Equation 1). Individual time-varying controls used in Models 1 and 2 are the following: women’s age and age squared, subsequent birth transitions during the survey, and number of adults in the household. State time-varying controls used in Model 2 are state-level year-to-year changes in GSP per capita, women’s employment, and education-specific average wages for women without and with a college degree. Birth penalties indicate the effect of births on within-individual changes in the outcome variable. Birth penalties × childcare costs indicate how within-state changes in childcare prices affect birth penalties. Both of these quantities are presented separately for the two groups: mothers without a college degree and mothers with a college degree. The three-way interaction coefficient tests for the difference in how childcare prices moderate birth penalties between the two groups of mothers, that is, 0.033 = (0.010) – (–0.023). Robust standard errors are clustered at the state level.
p < 0.05; **p < 0.01; ***p < 0.001 (two-tailed tests).
The results indicate that births have a negative and statistically significant effect on family income for women without a college degree, whereas the effect of births on women with a college degree is also negative, but smaller and not statistically significant. On average, we find that births are associated with a 16 percent decline in family income for women without a college degree, and a 7 percent decline for women with a college degree (see Table 2, Model 2, exp(–0.174) – 1 = –0.16 and exp(–0.073) – 1 = –0.07).
Consistent with our expectations, the interaction between births and childcare costs indicates that childcare prices worsen the effect of births on family income for women without a college degree but not for women with a college degree. For women without a college degree, births are associated with declines in family income that grow larger as childcare prices increase. We estimate that a one-unit increase in childcare costs (which amounts to an increase of $1,000, or a 7 percent increase on the average annual price for childcare services) is associated with an additional 2.3 percent decline in family income, and this coefficient is statistically significant (see Table 2, Model 2, exp(–0.023) – 1 = –0.023). For women without a college degree, the childcare costs coefficient is smaller and positive, indicating the decline in family income associated with births is smaller as childcare prices rise. The three-way interaction testing for the difference between the two interaction coefficients is statistically significant and indicates that childcare costs moderate birth penalties on family income differently for women with and without a college degree. The pattern of results is remarkably similar in Models 1 and 2, indicating that our key estimates are not substantially confounded by state-level, time-varying economic characteristics we control for in Model 2.
Figure 2 illustrates the results more intuitively by plotting birth average marginal effects (AME) across levels of childcare costs and by women’s education using results from Model 2. 11 Childcare costs are measured in 1,000s of dollars (in 2023 dollars) and are mean centered at $13,400. The plotted range of childcare costs covers 90 percent of the observed variation in childcare prices. The first bars on the left-side represent the size of the birth penalty in the most affordable childcare cost scenario: in this context, births are associated with an 8 and 11 percent decline in family income for women without and with a college degree, respectively. The difference in birth penalties between the two groups in the most affordable childcare scenario is not statistically significant. As childcare costs increase, the birth penalties become larger for mothers without a college degree, and the birth penalties become slightly smaller for mothers with a college degree. In the most expensive childcare costs context, births are associated with a 31 percent decline in family income for mothers without a college degree and with a 1 percent decrease for mothers with a college degree, and the difference between the two groups is statistically significant.
Birth Penalties on Family Monthly Income (Logged) by Mothers’ Education and Childcare Costs
These results indicate a widening gap in family income as childcare prices rise. The change in family income gap between the two groups increases from –3 to 31 percentage points when moving from the most to the least affordable childcare costs scenario. The increase in family income inequality between the two groups of mothers remains substantial even if we re-estimate this model using simulated measures of disposable family income that subtract childcare costs from family income (see Table S1 in the online supplement). This result is consistent with the expectation that births and replacement costs shaped by care infrastructures play a role in the production of family income inequality.
The adequacy of these estimates depends on the quality of the measure of childcare prices and the absence of substantial confounders. One possibility is that the state-level childcare price change reflects shifts in childcare prices for college-educated mothers more accurately than for non-college-educated mothers. For instance, changes in childcare subsidies are more likely to affect out-of-pocket costs for non-college-educated mothers than for college-educated mothers and could confound our estimates. To address this possibility, supplementary models include a time-varying control for states’ spending on childcare subsidies, and our main estimates remain robust (see Table S2 in the online supplement). Our childcare price measure could still be a more accurate estimate of price changes for college-educated mothers for other reasons (e.g., if average prices reflect urban areas and mothers with a college degree are more likely to reside in urban than rural areas). If this were true, our model would overestimate the real price change for mothers without a college degree, and our estimates for this group would be downwardly biased (because the coefficients refer to the measured change in childcare price, but the actual childcare price change for this group would be smaller).
Our estimates could also be affected by endogeneity between changes in childcare prices and changes in earnings for mothers with or without a college degree. For instance, if childcare prices react to improvements in the wages of mothers with college degrees, we might fail to detect a relationship between childcare prices and birth penalties for this group. Alternatively, childcare prices could reflect improvements in the low-wage labor market, which would again downwardly bias the estimates for this group because increases in wages should reduce birth penalties. Because Model 2 partly addresses these forms of endogeneity by controlling for state-level, time-varying average women’s wages for women with and without a college degree, the fact that the key estimates do not change much between Models 1 and 2 is reassuring. Additionally, sensitivity analyses using county-level childcare prices are consistent with the patterns found here (see Tables S7 and S8 in the online supplement), which suggest that within-state price-change heterogeneity does not substantially bias our estimates. In all, although our childcare price measure simplifies real price heterogeneity, we think that unobserved price change heterogeneity is unlikely to be large enough to invalidate our results.
Assuming that Model 2 provides reasonable estimates, how much of the increase in inequality observed in Figure 1 could be related to childcare costs? To obtain an approximation, we use the model’s coefficients to simulate post-birth family incomes for women without a college degree as if everyone resided in the cheapest childcare cost scenario plotted in Figure 2 ($9,400). To do this, we create a simulated family income variable that adds back the income the model indicates is reduced due to childcare prices. For instance, according to our model, a mother without a college degree who lives in a state with average childcare prices ($13,400) experiences an additional 6.9 percentage-point decline in family income compared to a mother who lives in a state with the lowest childcare prices. We re-calculate family income inequality trends using this simulated family income variable and compare it to the observed family income inequality. We find that the simulated increase in family income inequality is 1 percent instead of the 4 percent decline observed in Figure 1, a reduction of 3 percentage points. Thus, under this calculation, childcare costs could be responsible for up to 75 percent of the observed increase in family income inequality associated with births. This sizeable effect of childcare prices is consistent with existing analyses about the effect of universal childcare policy on poverty levels, which we further consider in the Discussion section (Borowsky et al. 2022; Giannarelli et al. 2019). While this result is striking, we want to caution that this simulation relies on the strong assumption that everything affecting family income remains unchanged except for the estimated effect of childcare costs.
Replacement Costs, Earnings, and Labor Supply
Next, we turn to the hypothesized drivers of the increase in family income inequality associated with childcare prices. According to our replacement costs model, this increased inequality is driven by childcare costs increasing birth penalties on earnings and hours worked for mothers without a college degree more than for women with a college degree. We expect these differential penalties to increase family income inequality if the larger penalties for mothers without a college degree are not offset by other income sources. In other words, the replacement costs model expects changes in family income related to childcare costs to be mostly driven by changes in mothers’ labor supply, not changes in partners’ labor supply.
Table 3 presents results for regression models for women’s and their partners’ earnings and labor supply. Overall, the results follow patterns that are remarkably consistent with our replacement costs model. We find that births are associated with substantial declines in earnings and work hours for both groups of women, but childcare costs moderate the size of these birth penalties only for mothers without a college degree. Additionally, we find that men’s earnings or labor supply are unaffected by births and childcare costs. Here, too, Models 1 and 2 are remarkably similar, indicating that our initial estimates are not confounded by state-level, time-varying economic changes.
Panel Regression Analyses of Birth Penalties on Monthly Earnings (log) and Weekly Paid Work Hours
Source: SIPP 2008 to 2016.
Note: Table 3 reports key coefficients of interest from a regression model estimating the effect of births and childcare prices on mothers’ and their partners’ monthly earnings (log) and weekly paid work hours (see Equation 1). Individual time-varying controls used in Models 1 and 2 are the following: women’s age and age squared, subsequent birth transitions during the survey, and number of adults in the household. State time-varying controls used in Model 2 are state-level, year-to-year changes in GSP per capita, women’s employment, and education-specific average wages for women without and with a college degree. Birth penalties indicate the effect of births on within-individual changes in the outcome variable. Birth penalties × childcare costs indicate how within-state changes in childcare prices affect birth penalties. Both of these quantities are presented separately for the two groups: mothers without a college degree and mothers with a college degree. The three-way interaction coefficients test for the difference in how childcare prices moderate birth penalties between women without a college degree and women with a college degree, that is, 0.120 = (0.044) – (–0.076). Robust standard errors are clustered at the state level.
p < 0.05; **p < 0.01; ***p < 0.001 (two-tailed tests).
In Model 2 for earnings, the interaction coefficient between births and childcare costs for women without a college degree indicates that a one-unit change in childcare costs is associated with an additional 8 percent decline in monthly earnings (see Table 3, Model 2, for earnings and women, exp(–0.08) – 1 = –0.08). This same coefficient for mothers with a college degree is positive but small and not statistically significant. The three-way interaction coefficient is statistically significant, indicating the interactions between births and childcare costs are systematically different for the two groups of women: the earnings of women without a college degree are significantly more negatively affected by childcare costs than are the earnings of women with a college degree. Partners’ earnings, on the other hand, are unaffected by births and childcare costs; all the coefficients are small and not statistically significant. This indicates that women’s earnings losses (due to births and due to childcare costs) are not offset by increases in men’s earnings. This is consistent with research finding that men’s earnings are unaffected by parenthood and births (Kleven, Landais, and Sogaard 2019; Musick et al. 2020).
Our models for work hours suggest that changes in labor supply are a key driver of women’s earnings losses. Similar to the model for earnings, we find that the effect of births on women’s work hours is similar for the two groups of women. Using Table 3, Model 2, for work hours and women, we estimate that births result in a 6.7-hour decline in weekly paid work hours for mothers without a college degree and a 5-hour decline for mothers with a college degree. The interaction between births and childcare prices, however, is only statistically significant for mothers without a college degree. In Model 2, this interaction indicates that an increase in one unit in childcare costs is associated with an additional half hour decline in weekly paid work for women without a college degree. The same coefficient for the college group is much smaller and not statistically significant. Despite the difference in size between the two interaction coefficients, the three-way interaction coefficient here is not statistically significant. These results are consistent with previous studies finding that the labor supply of women with lower earnings potential is more sensitive to the affordability of childcare services than is the labor supply of women with higher earnings potential (Anderson and Levine 1999; Kaestner et al. 2016; Ruppanner et al. 2021). Partners’ work hours, like earnings, show no sign of being affected by births or childcare costs. Supplementary analyses show similar differential effects of childcare costs on mothers’ employment, hours of work among the employed, and wages (see Table S3 in the online supplement).
Figure 3 summarizes these patterns. Both in earnings and in work hours, we see childcare costs creating a widening gap among women. We find that in the context of affordable childcare services, the decline in earnings and work hours associated with births is similar for the two groups of women. However, the difference in birth penalties between the two groups becomes larger as childcare prices grow. In the most expensive childcare price scenario, the decline in earnings associated with births for women without a college degree is twice the size of the decline for women with a college degree, and the decline in work hours associated with births for women without a college degree is 50 percent larger than the decline for women with a college degree. For both outcome variables, the childcare costs gradient is statistically significant for women without a college degree but not for women with a college degree. We find no evidence of partners offsetting this widening gap between women as childcare costs increase. Partners’ earnings and work hours are unaffected by births and childcare costs for both groups.
Birth Penalties on Monthly Earnings (log) and Weekly Paid Work Hours for Mothers and Their Partners, by Mothers’ Education and Childcare Costs
Supplementary and Sensitivity Analyses
We performed supplementary analyses including all women who have a first or second birth during SIPP, irrespective of their partnership status. Table S5 in the online supplement presents results and shows that all key coefficients are remarkably close to the ones obtained for the partnered sample. We also performed analyses examining heterogeneity in the relationship between birth penalties and childcare costs by partnership status, race, fathers’ education, and birth parity. We want to note, however, that due to our sample size we have limited statistical power to detect additional patterns of heterogeneity, and these results should be considered exploratory. Table S6 in the online supplement presents the analyses including interaction terms to assess heterogeneity by partnership status. None of these interaction coefficients are statistically significant, but the signs of the coefficients suggest the effect of births and childcare costs is somewhat larger for partnered than for unpartnered women. These results are consistent with previous research showing sensitivity to childcare prices among unpartnered women (Connelly and Kimmel 2003; Han and Waldfogel 2001; Kimmel 1998). Tables S11 and S12 in the online supplement present the analyses of heterogeneity by race, fathers’ education, and birth parity. We find no clear evidence that the effect of childcare costs on birth penalties is moderated by any of these variables, although analyses suggest the effect of childcare costs on birth penalties might be stronger for White women.
We performed extensive sensitivity analyses to assess the robustness of our findings to different measures of key variables and to different model specifications. Our results are robust to using a more geographically detailed measure of childcare costs at the county level (for analyses using panel data from the Current Population Survey and county-level childcare prices from the National Database of Childcare Prices, see Tables S7 and S8 in the online supplement). Our results are also robust to using an alternative measure of wage potential to stratify the sample instead of relying on college degree as a proxy (see Tables S9 and S10), and to using alternative measures of family income (see Table S1 for analyses on simulated disposable family income, and Table S13 for an analysis on equivalized family income). We also evaluated the sensitivity of our results to sample weights, to controls for additional state–time-varying characteristics (spending on childcare subsidies, unemployment rate, Medicare expansions, EITC rates), to controls for the number and ages of children in the household, and to models that allow for individual-level slopes and state-specific time trends. Our core results are robust across all these tests (see Tables S2, S12, S13, and S14).
Discussion
This article advances a replacement costs approach to describe how two core dimensions of the care infrastructure—care services and income transfers for unpaid caregiving—can shape the economic consequences of care events on family income inequality. The framework integrates the study of birth penalties within a broader framework that centers care labor needs (Folbre 2012; Folbre and Wright 2012). We provide an empirical demonstration of the relationship between care services, care event penalties, and family income inequality, focusing on birth penalties and childcare prices in the United States. Our results show that childcare costs increase birth penalties on labor supply and earnings for mothers without a college degree, but not for women with a college degree or for the male partners of either group of mothers. We find that this uneven distribution of birth penalties induced by childcare costs translates into a widening gap in family income after birth, showing that partners’ earnings and transfers are unable to compensate for the widening inequality in women’s earnings.
A primary goal of our study was to propose the care infrastructure as an important set of institutions to understand family income inequality. Prevailing research on drivers of family income inequality has largely focused on mechanisms related to the labor market, general tax and welfare policies, and family formation (e.g., Brady, Blome, and Kleider 2017; McCall and Percheski 2010; Parolin and Gornick 2021; Schwartz 2010; Western et al. 2016). Our framework highlights care infrastructures as an additional institutional dimension that contributes to the production of family income inequalities. Our study provides a conceptual framework to connect with the extensive literature on the institutional drivers of women’s employment and gender inequality (e.g., Budig, Misra, and Boeckmann 2016; Hook and Paek 2020) and with the literature examining the relationship between women’s employment and family income inequality (e.g., Cancian and Reed 1998, 1999; Esping-Andersen 2007; Gonalons-Pons and Schwartz 2017; Gonalons-Pons et al. 2021). Our perspective also adds to the emerging literature about how care infrastructure shapes parental investments and time spent with young children (Jackson and Schneider 2022).
Our analyses provide robust evidence that childcare costs contribute to increased family income inequality around the time of births, but this inequality could have long-term consequences through various mechanisms. The negative effects of childcare costs on mothers’ paid work can have long-lasting consequences on mothers’ lifetime earnings and retirement savings, thus negatively affecting family income in the long run. By reducing family income at a critical time for lower-income families, childcare costs might also exacerbate inequality in parents’ investments in children and thus play a role in perpetuating inequality in the next generation (Schneider, Hastings, and LaBriola 2018). The disequalizing effects of childcare costs could also compound with other demographic processes, in particular fertility differentials. If women with lower earnings potential have more children than women with higher earnings potential, they will experience the larger negative effects of childcare costs and birth penalties more times over their lifecourse, exacerbating the disequalizing effect estimated here. However, if the effect of childcare costs on birth penalties is smaller at higher parities (as our preliminary analyses suggest, see Table S11 in the online supplement), the compounding effect of fertility differentials might not be large. Additionally, childcare costs might affect fertility outcomes. Existing research is mixed and inconclusive, but one U.S. study concluded that childcare costs reduced fertility more among lower-income women than higher-income women (Blau and Robins 1988). More research is needed to determine the lifetime effects of childcare costs and birth penalties on family income inequality.
Our study broadens the scope of research on birth penalties in several directions. First, we show that birth penalties shape not only gender inequality but also family income inequality. We thus extend the research on motherhood penalties to consider more comprehensively the effect on family income inequality. This contributes to the emerging research examining how births affect economic outcomes beyond mothers’ or fathers’ labor market positions and that incorporate public transfers (Aassve et al. 2005; Bould et al. 2012; Sigle-Rushton and Waldfogel 2007; Stanczyk 2020). Furthermore, by providing a framework that connects births to other care events, we aim to encourage more systematic communication between research on births and parenthood and research on disability and old age care needs (Coe, Skira, and Larson 2018; Coe and Van Houtven 2009; Vaalavuo, Salokangas, and Tahvonen 2023). These two research strands are typically separated from each other, with only a few recent studies starting to integrate them (Gonalons-Pons and Ansari-Thomas forthcoming; Lightman and Link 2021). Our study also provides new evidence to understand the relationship between childcare costs and birth penalties in the United States, by implementing an improved and more robust identification model compared to most previous research. By leveraging panel models, we remove important sources of potential confounder bias stemming from unobserved time-fixed individual- or region-level characteristics, increasing the confidence in our estimates. The stability of key coefficients across models and sensitivity tests also speaks to the robustness of our results.
However, our study is not without limitations. Our estimates rely on a measure of childcare prices that ignores within-state heterogeneity in childcare price changes. As discussed earlier, our identification approach assumes that everyone in the state experiences a similar childcare price change as that measured by the average, but actual childcare price changes likely vary across individuals and regions within the state. Reassuringly, supplementary analyses with county-level childcare price data are consistent with the results presented here (see Tables S7 and S8 in the online supplement), and descriptive analyses indicate that price changes are similar at different levels of the childcare price distribution (see Figure S1). Nonetheless, measurement error can still affect our estimates.
The lack of data on informal caregivers is another important limitation. Our estimates of childcare price moderation average the effect of childcare prices for those who can and cannot rely on informal caregivers (akin to an intend-to-treat estimate). The effect of childcare prices likely differs between those who can and cannot rely on informal caregivers, but this heterogeneity should not bias the average estimate we focus on. However, if the unobserved availability of informal caregivers systematically varied with childcare prices, this could bias our key estimate. While such correlation is plausible to an extent, it is unlikely to be large enough to fully confound our estimates, particularly given the relatively small differences in informal caregiving by class and region (Guzman et al. 2016; U.S. Census Bureau 2011).
Our estimates could also be biased by other omitted time-varying variables at the individual or state level. The stability of our estimates across models that control for additional time-varying characteristics is reassuring, but we cannot be certain these controls remove all possible confounders that affect both childcare costs and families’ outcomes. Finally, our analyses are also limited by the relatively narrow observation window spanning two years before and after birth, as well as by the size of our analytic sample. A larger sample would allow for more robust analyses of differences by partnership status, race, and parity, and a longer panel would enable studying long-lasting effects of births on family income and family income inequality.
Our study has clear policy implications. We show that reducing childcare prices in a context with no meaningful income transfers to unpaid caregivers and in a predominantly market-provided childcare system would help reduce family income inequality at a key life period for young children. More generally, our results support the replacement costs framework’s claim that robust care infrastructures that support people through care events can help reduce family income inequality, either by facilitating access to paid caregivers or providing income transfers to offset earnings losses resulting from providing unpaid care. These results are consistent with recent analyses that simulate the effect of large-scale policy changes to the U.S. childcare infrastructure and find that it would substantially increase mothers’ employment and decrease child poverty (Borowsky et al. 2022; Giannarelli et al. 2019). Our results also speak to the limitations of the current U.S. childcare subsidy model, which fails to reach the vast majority of families who qualify for subsidies (Giannarelli et al. 2019; Ullrich et al. 2019). Increasing the budget to reduce waitlists would likely be insufficient, because the design of the policy itself imposes strict administrative and work-requirement criteria that would continue to limit access for low-income families. A broader policy to subsidize childcare provision on the supply side and/or provide income transfers for unpaid caregiving would help reduce inequality more effectively.
Human societies require and mobilize a lot of labor to fulfill the care needs of children, working-age adults, and older adults. This care labor is often taken for granted, naturalized, and made invisible, because patriarchal norms have conventionally made women take on this work as part of their gender obligation (Glenn 2010). Since the COVID-19 pandemic, the relevance of care labor has become more visible, but research on care labor continues to be largely confined to gender inequality. Our framework and empirical study highlight the broader relevance of care labor by articulating how the social organization of care provision structures family income inequality.
Supplemental Material
sj-pdf-1-asr-10.1177_00031224241297247 – Supplemental material for Care Labor and Family Income Inequality: How Childcare Costs Exacerbate Inequality among U.S. Families
Supplemental material, sj-pdf-1-asr-10.1177_00031224241297247 for Care Labor and Family Income Inequality: How Childcare Costs Exacerbate Inequality among U.S. Families by Pilar Gonalons-Pons and Ioana Marinescu in American Sociological Review
Footnotes
Acknowledgements
Thank you to Christine Schwartz, Jennifer Glass, and Kirsten Swinth for comments on earlier versions of this manuscript. Thank you to Emily Campbell for research assistance.
Funding
This research has been carried out in part using the facilities of the University of Pennsylvania Population Studies Center (R24 HD044964). The content is solely the responsibility of the authors and does not necessarily represent the official views of the funding agencies.
Notes
References
Supplementary Material
Please find the following supplemental material available below.
For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.
For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.
