Abstract
Early childbearing is associated with poor outcomes for parents and children, potentially exacerbating inequality within and across generations. Building on the structural sexism and health perspective, we argue that systemic gender inequality is a conceptually important—and understudied—factor in early childbearing. Using survey data from the National Longitudinal Study of Adolescent to Adult Health (N = 5,052 female respondents and 32,595 person-years) merged to state, county, and school characteristics, we investigate how exposure to structural sexism during adolescence shapes early childbearing. We measure structural sexism in institutional domains with state and county-level measures of economic, cultural, and reproductive health. We also generate a novel measure of structural sexism in interactional domains, capturing gendered expectations, power dynamics, and conservative religious beliefs among school peers. We find that institutional sexism is associated with a higher likelihood of early childbearing among young women racialized as Black, but not among those racialized as White. Interactional sexism is also a significant predictor of early childbearing, and is especially salient for adolescents as they turn to peers for social cues about gender roles. Our results underscore the importance of institutional and interactional forces for shaping families and raise questions about increasingly polarized gender climates post-Dobbs.
The timing of childbirth shapes population characteristics; reflects cultural attitudes about gender, sex, and the transition to adulthood; and has important implications for the intergenerational transmission of social and economic resources (Cha and Weitzman 2023; Sweeney and Raley 2014). Early childbearing is associated with poor outcomes for parents and their children, and has long been depicted as a critical social problem (Furstenberg 2003; Wolfe, Thomeer, and Bauldry 2024). Despite steep declines in teen childbearing in recent decades (Osterman et al. 2022), age at birth in the United States remains substantially younger than in other wealthy countries, with more pronounced differences across social groups (Brady, Baker, and Finnigan 2024; Lesthaeghe and Neidert 2006; Musick and Michelmore 2018). Young age at first birth is disproportionately concentrated among historically marginalized groups, including individuals racialized as Black or Hispanic, and those with less education (Guzzo and Hayford 2020; Sweeney and Raley 2014). Disparities in the timing of parenthood by education have grown over time, along with a constellation of other family characteristics associated with the “diverging destinies” of U.S. children (Guzzo and Hayford 2020; McLanahan 2004).
Although difficult to disentangle from the socially disadvantaged backgrounds of many young parents, evidence suggests that early childbearing may constrain opportunities and serve as an avenue for reproducing inequality within and across generations (Aizer, Devereux, and Salvanes 2022; Diaz and Fiel 2016; Furstenberg 2007; Geronimus and Korenman 1993; Hoffman 1998; McLanahan 2009; Miller 2011). Parents in their teens and early 20s are more likely to report that their pregnancies came earlier than desired (Guttmacher Institute 2019; Musick 2002). Early births may accelerate the progression of romantic relationships, which tend to be less stable in young adulthood (Glass and Levchak 2014; Martin 2004; Musick and Michelmore 2018; Teachman 2002), often with negative implications for the financial and parenting resources available to children (Aizer et al. 2022; McLanahan 2004). Early childbearing may also truncate educational attainment and limit career opportunities, particularly among mothers (Furstenberg 2007; Geronimus and Korenman 1993; Hoffman 1998; Miller 2011). As women’s education and employment opportunities have expanded over time, the educational penalties associated with early motherhood have also increased (Wolfe et al. 2024). Understanding the interplay between early childbearing and broader processes of social stratification remains a critical issue for sociology and policy.
Prior research provides a rich accounting of the perceived opportunity costs, barriers to contraception, and relationship dynamics that shape early parenthood at the individual and couple level (Bell et al. 2018; Cha and Weitzman 2023; Edin and Kefalas 2005; Kusunoki and Upchurch 2011; Manlove, Ryan, and Franzetta 2007; Manning et al. 2009; Musick et al. 2009; Rosenbaum et al. 2012). Prior work has also demonstrated the importance of broader contextual factors such as poor economic opportunities, which leave young people with fewer meaningful pathways to adulthood beyond family roles (Edin and Kefalas 2005; Kearney and Levine 2014; Mollborn 2011). Comparative research further suggests that political and cultural factors contribute to younger childbearing in the United States relative to European countries (Darroch, Singh, and Frost 2001). We extend this work by examining early childbearing through the lens of gender inequality, drawing on insights from a new line of inquiry into structural sexism, which is defined as systematic gender inequality in power and resources (Beccia et al. 2022; Everett et al. 2022; Homan 2019). Recent developments, perhaps most strikingly the overturning of Roe v. Wade, have called attention to wide variation in economic, cultural, and political environments for women in the United States and its implications for health and well-being (e.g., Pacia and Brown 2022).
Our study investigates how exposure to structural sexism at the critical developmental stage of adolescence is associated with young women’s early transitions to parenthood, defined as having a first birth before age 22. We posit that institutions and interactions that reproduce gender inequality increase the likelihood of early childbearing by limiting women’s economic prospects, reproductive autonomy, and power in different-gender romantic relationships. Given the intersectionality of structural sexism and anti-Black racism (Collins and Bilge 2020; Crenshaw 1991; Homan, Brown, and King 2021), we further explore differences in levels of exposure and the estimated effects of structural sexism among young women racialized as Black and White. This comparison is particularly important for understanding the matrix of oppression that uniquely shapes the lived experiences and reproductive health outcomes of women racialized as Black. We analyze data from the National Longitudinal Study of Adolescent to Adult Health (Add Health), which allows us to follow female-identified respondents from adolescence (ages 11 to 16) through their early childbearing years and match detailed individual-level survey data to contextual characteristics aggregated to their states, counties, and schools. Our study is innovative in mapping multiple layers of the gender system in the United States onto a critical aspect of reproductive health and well-being—the timing of first birth.
Our work contributes to the literatures on structural sexism and early childbearing in important ways. First, our focus on early childbearing extends the theoretical scope of the structural sexism perspective, which has predominately focused on physical health outcomes (Chae et al. 2018; Everett et al. 2022; Homan 2019; Nagle and Samari 2021; Philbin et al. 2021). Early childbearing is distinctive as a health outcome, notably because of the strong connection to relationship dynamics that are broadly understood to unfold in a larger gender structure (see Bell et al. 2018). Second, we evaluate structural sexism at multiple levels of the gender system. Following prior work (e.g., Homan 2019), we first measure institutional sexism with indicators of gender disparities in economic, cultural, and reproductive health resources. Extending the conceptual and empirical work to date, we then develop a novel measure of interactional sexism using reports of school peers to tap gendered aspects of everyday interactions, which may be particularly salient to adolescents looking to their classmates for social cues about gender roles and future opportunities (Fletcher 2007; Mollborn, Domingue, and Boardman 2014; Teitler and Weiss 2000). This measure is considered structural because it aggregates gendered dynamics that unfold within schools, in contrast to individual-level reports of personal attitudes or perceived cultural beliefs. Finally, our attention to variation across racialized groups in the links between structural sexism and early childbearing illustrates how the interdependence of sexism and other systems of oppression may have distinct consequences for women in marginalized social positions (Collins 2000; Collins and Bilge 2020; Crenshaw 1991; Homan et al. 2021), joining evidence on the consequences of intersectional structural inequalities for health and well-being (Homan et al. 2021). Taken together, our approach sheds new light on the structural forces that shape early entry into parenthood.
Background
Structural Sexism and Health
Structural sexism refers to “systematic gender inequality in power and resources manifest in a given gender system” (Homan 2019:490). This framework conceptualizes gender as a characteristic of the social context rather than a static individual trait (Ridgeway and Correll 2004; Risman 2004); it thus calls for an empirical approach that looks to broader structural forces. Risman (2004) argues that the gender system is equally important as economic and political structures, giving rise to social norms that shape women’s opportunities, interests, and choices. The gender system creates inequality on multiple levels: through resources and norms at the institutional level, via social relations and patterns of behavior at the interactional level, and by perceptions and identities at the individual level (Burt 1982; Homan 2019; Ridgeway and Correll 2004; Risman 2004). Like other systems of inequality, the gender system exposes girls and women (and other marginalized groups) to adverse experiences and expectations that “get under the skin” and harm health (Krieger 2001, 2012). A growing body of work has fleshed out the conceptualization and measurement of structural sexism and documented associations with poor health across the life course (Beccia et al. 2022; Everett et al. 2022; Homan 2019).
Research on structural sexism and health parallels a broader literature on structural inequalities, which includes racism and heteronormativity. Evidence has accumulated on how structural inequalities contribute to poor health (Homan and Brown 2022; Homan et al. 2021; Montez, Hayward, and Wolf 2017; Montez, Zajacova, and Hayward 2016; Phelan and Link 2015) and adverse birth outcomes, including Cesarean sections, preterm birth, and decreased birthweight (Chae et al. 2018; Everett et al. 2022; Nagle and Samari 2021). Related studies link structural stigma with riskier sexual behavior among sexual minority youth (Charlton, Hatzenbuehler,et al. 2019; Philbin et al. 2021). The idea is that structural inequality influences the distribution of both health-promoting factors (e.g., access to healthcare, material resources, and social support) and health-harming factors (e.g., stress, mistreatment, and poor health behaviors), putting marginalized groups at risk (Everett et al. 2022; Homan 2019; Link and Phelan 1995; McKetta et al. 2022; Rapp et al. 2022). This line of work has developed concrete measures of structural inequalities and provides scaffolding for our investigation of early childbearing.
Although structural sexism is a multilevel construct (Risman 2004), studies have largely operationalized the institutional level, with indicators aggregated to states or counties, including women’s economic and political positions relative to men, religious conservatism, and policies on abortion, gun ownership, and lesbian, gay, and bisexual rights (Everett et al. 2022; Homan 2019; Homan and Burdette 2021; Nagle and Samari 2021; Rapp et al. 2022). To our knowledge, only Homan’s (2019) study of chronic conditions, self-rated health, and physical functioning evaluated sexism on multiple levels. Homan assessed institutional sexism with state-level data linked to the 1979 National Longitudinal Survey of Youth; interactional sexism within marriages by evaluating spouses’ earnings, education, and age; and internalized sexism using respondents’ self-reported gender role attitudes. She found that institutional and interactional sexism were associated with worse health among women, but internalized sexism was not. This work demonstrates the value of more fully investigating multiple components of the gender system and emphasizes the need for greater attention to interactional settings as key sites for reproducing gender inequality (Homan 2019).
At the interactional level, gender operates in “social relational contexts” that involve patterns of behavior in which individuals define themselves in relation to others (Ridgeway and Correll 2004:514). Cultural beliefs about gender shape everyday interactions between women and men, setting taken-for-granted expectations, for example, about men’s competency and self-interest versus women’s altruism and orientation to others (England and Kilbourne 1990; Ridgeway and Correll 2004). Women and men enact these expectations—or “do gender”—in everyday contexts that reinforce gender inequalities (West and Zimmerman 1987). A robust line of research suggests the process of doing gender systematically advantages men in negotiating the division of household labor between romantic partners, perpetuating gender inequality in paid and unpaid work (Bittman et al. 2003; England and Kilbourne 1990; Evertsson and Nermo 2004; Gough and Killewald 2011; Musick and Jeong 2023; Sayer et al. 2009). Gendered power dynamics extend beyond household labor to sexual activity, contraceptive practices, and family-building, affecting women’s health and well-being (Bell et al. 2018; Manlove et al. 2011; Rosenbaum et al. 2012). Our study closely examines the extent to which interactional sexism is related to early childbearing.
The Social Context of Early Childbearing
Teen childbearing (at age 19 or younger) has long been an area of scholarly and public concern (e.g., Furstenberg 2003). Early childbearing is a function not only of chronological age, but also social norms and sequences marking the transition to adulthood (Boonstra 2014; Rindfuss 1991). In recent decades, this transition has lengthened in the United States and other wealthy countries due to the growing importance of college education, delayed marriage, and extended independent living away from the parental home (Arnett 2000; Furstenberg 2010; Settersten, Furstenberg, and Rumbaut 2005). Births prior to age 22 can potentially truncate college attendance, with long-term effects on career trajectories and earnings. Reflecting these shifts, the average age of first birth for women has increased from 21.4 in 1970 to 27.1 in 2020 (Mathews and Hamilton 2002; Osterman et al. 2022). Given contemporary social contexts, we argue that childbearing at ages 20 and 21 should also be considered “early,” and for this study, we define early childbearing to be a first birth before age 22.
Pathways to early childbearing have become increasingly polarized by socioeconomic status (Guzzo and Hayford 2020). Women with higher education and family resources increasingly delay childbearing, widening the gap in age at first birth and contributing to growing inequality in the resources available to children (McLanahan 2004). The consequences of early childbearing for women’s educational attainment and career opportunities also increased over the last half of the twentieth century, with especially high educational penalties among women born in the early 1980s compared to previous decades (Wolfe et al. 2024). This cohort came of age when fertility delay was more common, there was widespread moral panic over the consequences of teen pregnancy, and college education became crucial for career opportunities and a middle-class income (Wolfe et al. 2024).
The drivers of early childbearing remain a critical topic of inquiry. Much prior literature on early childbearing focuses on individual- and couple-level determinants. One dominant account centers on the poor economic prospects of young mothers, who tend to come from socially disadvantaged backgrounds (Diaz and Fiel 2016; Guzzo and Hayford 2020). This body of research hypothesizes that poor economic prospects may lower the perceived opportunity costs of childbearing in terms of lost wages and human capital, relative to the cost of contracepting and benefits of parenthood, leading to earlier entry into motherhood (Edin and Kefalas 2005; Manlove 1998; Musick et al. 2009). Edin and Kefalas (2005) underscore the importance of meaning-making in the absence of promising educational and career pathways. They posit that early motherhood will be more common when children are the main route to fulfilling young adult roles, and they provide rich accounts of children as the greatest sources of meaning and identity in the lives of the poor, unmarried mothers in their study.
Relationship dynamics also shape how couples negotiate contraceptive use and fertility desires, and thus the risk of early, potentially undesired motherhood (Cha and Weitzman 2023; Kusunoki and Upchurch 2011; Manlove, Ryan, and Franzetta 2003, 2007; Manlove et al. 2011; Manning et al. 2009). In sexual relationships, partners with more perceived power—older male partners, or those providing money—are more likely to have their contraceptive preferences met (Manlove et al. 2003, 2007; Rosenbaum et al. 2012). Power differentials, conflict, and poor communication are associated with less consistent contraception, increasing the risk of undesired pregnancy (Manlove et al. 2011; Manning et al. 2009). Drawing on in-depth interviews, Bell and colleagues (2018) highlight the social meanings of contraceptive use, in particular the tendency to stop using condoms to signal greater relationship commitment and respect (see also Tavory and Swidler 2009). They also observe that disadvantaged young men often desired children sooner and placed less value on waiting to reach education or financial milestones than did disadvantaged young women, leading to unintended and ambivalently desired pregnancies. Women may engage in sex to satisfy partners’ desires, and in some cases, men use physical violence to coerce and manipulate their partners’ reproductive behavior (Barber et al. 2018; O’Sullivan and Allgeier 1998).
A structural perspective suggests the individual- and couple-level factors described above are responsive to broader social forces. The literature widely recognizes the importance of economic inequality and marginalization in shaping family-related behaviors, linking poor economic conditions at the neighborhood and state levels to early childbearing (Driscoll et al. 2005; Kearney and Levine 2012, 2014; Mollborn 2011). Kearney and Levine (2014), for example, examined state variation in income inequality and advanced a “culture of despair” hypothesis, suggesting that poor young women are more likely to embrace early motherhood due to limited economic opportunities. Weak economic conditions, measured by widespread job loss and rising local unemployment rates, have also been linked to higher teen birth rates, particularly among Black teens (Ananat, Gassman-Pines, and Gibson-Davis 2013; Kearney and Levine 2015). From a cross-national perspective, broader political and cultural factors are hypothesized to contribute to the young age at birth in the United States (Darroch et al. 2001; Furstenberg 2003). For example, in the United States, ambivalence and lack of openness around sexuality and transitions to sexual activity create barriers to contraceptive access and knowledge (Bearman and Brückner 2001; Sennott and Mollborn 2011).
Building on the structural sexism and health perspective, we posit that systemic gender inequality is a conceptually important—and understudied—factor in early transitions to motherhood. Although recognized as an overarching construct that shapes couple dynamics (e.g., Bell et al. 2018; Manlove et al. 2011), no empirical work, to our knowledge, has examined the association between systemic gender inequality and early childbearing. A partial exception is research linking early childbearing to the local or state-level reproductive health context, which captures a highly relevant aspect of gender inequality at the institutional level. On this count, early childbearing is more common when access to family planning services is more restrictive (Colen, Ramey, and Browning 2016; Guldi 2008; Kearney and Levine 2015; Myers 2023). We draw on prior research on structural sexism and health (e.g., Homan 2019) to flesh out the institutional features of gender inequality that potentially shape early childbearing, and we generate a novel measure of interactional sexism that we maintain may be especially important for adolescents. Applied to early childbearing, the structural sexism framework has potential to shed new light on the social forces that shape families.
Structural Sexism and Early Childbearing
Institutional sexism
At the institutional level, structural sexism is distinct from the literature on economic inequality and marginalization because it focuses on macro-level institutional structures—the economy, culture, policy—that specifically disadvantage women. The idea is that institutional sexism limits young women’s resources and opportunities for economic achievement and roles beyond the family. Fewer employment opportunities for women compared to men reduce their incentives to invest in human capital and lower the opportunity costs of early childbearing. Conservative religious and political ideologies that prioritize a traditional division of labor may also shift young women’s orientations from economic and career attainment to family formation (see Glass and Levchak 2014; Glass and Nath 2006; Moore and Vanneman 2003; Uecker 2014). Finally, restrictive reproductive health policies undermine women’s autonomy and sense of control, directly affecting pregnancy and childbirth through limited access to reproductive health services (Colen et al. 2016; Guldi 2008; Joyce and Kaestner 2001; Kearney and Levine 2015; de Londras et al. 2022; Myers 2023). Research shows substantial variation across states and counties in economic, cultural, and policy dimensions of institutional sexism that are associated with women’s health outcomes (Beccia et al. 2022; Everett et al. 2022; Homan 2019). These material and power inequalities in the macro-level structural domain create a gendered hierarchy and institutional barriers to women’s independence and autonomy (Risman 2004).
Interactional sexism
As noted earlier, interactional sexism is conceptually important but understudied in the literature on structural inequalities and health. Interactional sexism is enacted and reinforced through everyday interactions in contexts like schools, workplaces, and families, which can generate gender inequality even without individual-level intent to do so (Risman 2004). We focus on the school context, positing that what adolescents learn from classmates about gender roles should be a salient factor in shaping their behavior into young adulthood. Cues may come from gendered expectations, patterns of behavior, and power disparities that manifest in everyday interactions with peers.
Gender disparities in expectations around sexual behavior are particularly relevant for romantic relationships and early childbearing. A sexual double standard shows that men and women are judged differently for sexual behavior: men tend to be socially rewarded, whereas women risk stigma for the same behavior (England and Bearak 2014; Hamilton and Armstrong 2009; Kreager and Staff 2009; Soller and Haynie 2017). Soller and Haynie (2017) found that the strength of this sexual double standard varied across high schools. As expected, they found that in schools with a stronger double standard, young men had more sex partners, pointing to a positive association with early childbearing. They expected young women would have fewer sex partners in schools with a strong double standard, but they found no such association. They speculated that young women might still engage in sexual activity despite the double standard and bear the associated social costs. This speculation aligns with findings that link first sex to mental health detriments among young women, particularly women in less committed or stable relationships (Meier 2007).
Adolescents may also take cues about doing gender in romantic relationships from gender disparities in household and caretaking responsibilities of their peers, gender pay differentials among working peers, and age differences between romantic partners in their schools. Male advantage on these dimensions at the school level provides a blueprint for interactions at the couple level. Greater gender inequality among peers can weaken young women’s power in romantic relationships around, for example, sexual activity and contraceptive practices. Power differentials in relationships have been linked to less consistent contraception (Manlove et al. 2011; Manning et al. 2009). Furthermore, when men are more advantaged in negotiations with romantic partners and face lower social and opportunity costs of sex and children, the likelihood of early childbearing may be higher (Bell et al. 2018; Kreager and Staff 2009).
Finally, conservative religious beliefs among school peers can reinforce gendered power dynamics within couples. Prior research has found that the religiosity of friends and schoolmates is associated with family-related behavior, independent of individual religiosity (Adamczyk and Felson 2006; Uecker 2014). Religious fundamentalism promotes a traditional gendered division of labor within marriage and delegates final decision-making authority to men (Glass and Levchak 2014; Glass and Nath 2006). Peer environments with rigid gender role expectations may orient women away from career goals, complicate communication around sexuality, and create barriers to contraception and reproductive health services (Bearman and Brückner 2001; Kusunoki and Upchurch 2011; Manlove et al. 2011; Sennott and Mollborn 2011), again increasing the risk of early birth.
Interactional sexism is a component of structural sexism that operates at the meso level and is distinct from individual-level sexist attitudes, although interactions can shape ideologies about unequal gender roles that may eventually be internalized (Homan 2019; Risman 2004). This type of sexism is difficult to observe using individual-level survey data or interviews because interactional sexism does not depend on the experiences or sexist attitudes of individuals. For example, if a young man feels socially accepted and unembarrassed about having sex, he does not necessarily hold sexist attitudes. However, if men collectively have more positive social norms around sex than women, it creates a sexual double standard that shifts the balance of power. Similarly, a male student may not hold sexist attitudes about housework, but his parents may require him to do fewer chores than his sisters. If male students do less housework than female students in aggregate, the interactional context enacts gendered expectations about women’s roles in the domestic sphere.
Distinguishing between institutional and interactional sexism
Both institutional and interactional sexism tap into gender inequality in material resources and cultural expectations, but they do so at different levels (Risman 2004). Institutional sexism operates at the macro level of social institutions, and interactional sexism operates at the level of everyday peer interactions. These differences are conceptually important for understanding how exposure to sexism during adolescence may shape early childbearing. For example, at the institutional level, gender inequality in material resources has been operationalized in prior research by taking a county-level ratio of male to female earnings, which speaks to macro structural patterns of economic opportunity. By comparison, at the interactional level, gender disparities in hourly wages between male and female students within the same school tap into gendered material disparities and power dynamics that adolescents experience in their everyday interactions with peers.
Similarly, conservative religious ideology can be measured at both institutional and interactional levels. At the institutional level, studies have examined the percentage of conservative religious denomination adherents in a county. Conservative religious denominations reflect institutional belief systems and cultural identities that prescribe a gendered division of labor and prohibit nonmarital sex, birth control, and abortion (Glass and Nath 2006). At the interactional level, the percentage of peers in school who hold scriptural inerrancy beliefs (e.g., Genesis 1:28: “Be fruitful and multiply” and Ephesians 5:24b: “wives should submit to their husbands in everything”; Uecker 2014:398) reflects the normative context in which everyday interactions with peers occur. Even if an individual does not hold scriptural inerrancy beliefs for themselves, interacting with peers who do can shape their expectations about women’s roles and future opportunities.
Although we hypothesize that both institutional and interactional sexism will be independently associated with higher odds of early childbearing, interactional sexism may be especially important in thinking about adolescence as a distinct stage in the life course. During adolescence, teens separate from their parents, spend more time with classmates, and place greater value on acceptance and recognition from friends (Elder 1998; Guzzo, Hayford, and Lang 2019; Shanahan 2000; Steinberg and Morris 2001). Their romantic relationships also become more emotionally and sexually intimate and play a critical developmental role (Giordano, Longmore, and Manning 2001, 2006; Meier and Allen 2009). Peer interactions may be particularly salient at this developmental stage, and adolescents may have less direct exposure to the macro-level institutional inequalities that play out in the broader economic, policy, and cultural domains.
Variation by racialized status
Relationships between structural sexism and early childbearing may play out differently across racialized groups. Intersectionality theory highlights the interdependence of structural systems of inequality such as sexism and racism, and illustrates their differential effects based on individuals’ social locations (Collins and Bilge 2020; Crenshaw 1991). In particular, women racialized as Black face the intersecting and multiplicative effects of sexism and racism (Crenshaw 1989). We take an initial step at assessing intersectionality by examining racial differences in early childbearing, exposure to structural sexism, and structural sexism’s association with early childbearing (Ward et al. 2019).
Keeping in mind that structural sexism is a characteristic of a macro- or meso-level context and not individual-level experiences with discrimination, Black women may experience higher levels of institutional and interactional sexism compared to White women if they are more likely to be situated in these contexts due to historical patterns of racial migration and segregation. Research on structural inequality finds that structural racism tends to be lower in the American South, whereas structural sexism tends to be higher (Homan et al. 2021). 1 Because Black women are more likely to live in the South, they generally experience higher levels of structural sexism than racism. This research provides some initial evidence about differential exposure to macro-level institutional inequalities, but it is unclear whether this extends to interactional sexism.
In addition, the effects of sexism on early childbearing may vary by racialized status. Differential effects may operate through disparate access to reproductive healthcare and abortion, or through unequal opportunities that shape expectations for future family or economic trajectories. The reproductive justice framework developed by the SisterSong Women of Color Reproductive Health Collective calls attention to structural barriers that differentially limit reproductive choices for women based on their race and socioeconomic status, among other factors (Ross and Solinger 2017). Women with higher degrees of social privilege are less susceptible to institutional constraints on their reproductive lives. The weathering hypothesis also explores racial variation in early childbearing, and argues that it is a rational response to systemic disadvantage and health uncertainty experienced by Black women (Cohen 2016; Geronimus 1987, 1996; but see Meadows et al. 2013). For Black women with weak labor market prospects, early childbearing can be advantageous because it coincides with a period of expected unemployment and lowest earning potential (Geronimus 1987).
Questions about the role of intersectional inequalities in shaping health disparities are in an early stage of conceptual and methodological development (e.g., Hardeman et al. 2022; Homan et al. 2021). Limited empirical evidence suggests that young Black women may have greater exposure to structural sexism. The reproductive justice and weathering frameworks described above further suggest that structural constraints may be more binding for young Black women and other socially marginalized groups, that is, structural sexism may play a bigger role in early childbearing.
Conceptual Model and Hypotheses
Figures 1 and 2 map our conceptual and operational framework for structural sexism and its relationship to early childbearing, which to our knowledge has not been examined in prior research. Figure 1 summarizes our assessment of structural sexism at two levels, following prior literature on institutional sexism (Everett et al. 2022; Homan 2019), and expanding the construct of interactional sexism relative to existing research. At the institutional level, we posit that sexism plays into early childbearing by limiting young women’s resources and constraining their opportunities for economic achievement and roles beyond the family, as measured by indictors of economic inequality and cultural ideology at the county level, and by reproductive health policy at the state and county levels. At the interactional level, sexism reproduces gendered expectations and power dynamics in romantic relationships, as measured by indicators of gendered expectations about sexual behavior, power in relationships, and conservative religious beliefs enacted in everyday interactions among school peers. It is important to underscore here that our approach relies on research rooted in binary conceptualizations of gender and cisnormative ideas about childbearing, missing the experiences of people with different gender identities and sexual orientations (see Everett et al. 2022).
Conceptual Framework for Institutional and Interactional Structural Sexism
Conceptual Model for the Relationship between Structural Sexism and Early Childbearing
Figure 2 represents our conceptual model of the relationship between structural sexism and early childbearing (see, e.g., Kohler, Class, and Sawert 2024; Lundberg, Johnson, and Stewart 2021), which we estimate as a discrete-time event history model. Our model includes a direct link between structural sexism and early childbearing, as well as the potential for moderation by an individual’s racialized status. To increase the precision of our key estimates, we control for individual and family characteristics that are associated with early childbearing (racialized status, foreign-born status, family structure, and parental education). We also account for a set of confounders that potentially shape both structural gender inequality and early childbearing, including calendar time and sociodemographic contextual characteristics (percent female at the school and county-level poverty quintiles and racial/ethnic composition). To further address the threat of bias due to unobserved or omitted regional or state characteristics, we conduct robustness checks with models that include region-by-year and state-by-year fixed effects. We exclude fixed effects from our main models so we can directly measure aspects of institutional sexism that would be (or would potentially be) accounted for in a fixed-effect specification, but we show that our key findings are not sensitive to this choice.
Based on our review of existing theory and empirical evidence, we hypothesize that structural gender inequality is associated with an increase in early, potentially undesired motherhood at both the institutional (Hypothesis 1) and interactional (Hypothesis 2) levels. We further hypothesize that exposure to interactional sexism during adolescence will be more salient than institutional sexism in shaping early childbearing, given the strong orientation to peer and romantic relationships during this developmental stage (Hypothesis 3). We draw on intersectional frameworks to develop our last two hypotheses. Based on the (limited) empirical evidence, we hypothesize that women racialized as Black will be exposed to higher levels of structural sexism compared to women racialized as White (Hypothesis 4). Finally, based on our reading of reproductive justice and weathering frameworks, we hypothesize that the relationship between structural sexism and early childbearing will be stronger for women racialized as Black compared to women racialized as White (Hypothesis 5).
Data and Method
Add Health
The Add Health dataset is uniquely structured to address our research questions, with prospective information on young women’s family and sociodemographic backgrounds, first birth timing, and characteristics of their states, counties, and schools (Brown 2022). Add Health is a nationally representative longitudinal cohort study of adolescents who were in grades 7 to 12 in 1994–95 (Harris 2013). These respondents were born in the early 1980s and are part of the cohort that experienced the strongest educational penalties for early childbearing (Wolfe et al. 2024). The study used a school-based sampling design (132 schools), with stratification of schools by region, urbanicity, type (public, private), racial composition, and size. A sample of 20,745 students completed a detailed Wave I in-home survey in 1994–95, and follow-up surveys were conducted in 1996 (Wave II), 2001–02 (Wave III), 2008–09 (Wave IV), and 2016–18 (Wave V). We use the Wave I in-home survey to measure individual- and school-level characteristics during adolescence, and we use fertility histories collected in Waves IV and V to assess the timing of first birth. We pull state and county characteristics from the Add Health contextual file, which includes hundreds of variables from various sources assessed around the time of the Wave I survey and linked to individual respondent records by the Add Health team using GPS coordinates of respondents’ households (Harris 2013).
Our main analytic sample includes female-identified respondents in the nationally representative sample. We often refer to respondents as young women, although we cannot determine the extent to which they are cisgender due to data limitations (at Wave I, only female and male options were offered, and it is unclear how respondents with other gender identities answered this question). We make no further restrictions based on gender or sexual orientation. Of the 9,642 potential female-identified respondents, we exclude those who did not have a geocode match for contextual data (n = 51), did not participate in either Wave IV or Wave V when pregnancy and birth data were collected (n = 1,091), or did not have valid information on first birth (n = 286). Because we are interested in the association between exposure to sexism in adolescence and subsequent childbearing, we exclude respondents who had a child prior to the Wave I interview (n = 172) to maintain temporal ordering. We further exclude respondents who were 17 or older at Wave I (n = 2,619) to address concerns about left truncation at ages when fertility begins to rise, and again to preserve the temporal ordering of exposure to sexism and subsequent childbearing; our results are not sensitive to this exclusion (results available upon request). Finally, we exclude respondents with missing data on interactional sexism (n = 270, primarily in same-gender schools or schools with insufficient samples disaggregated by gender) and individual-level covariates (n = 101). Our final sample includes 5,052 female-identified respondents ages 11 to 16 at Wave I.
We transformed the data into person-year observations for the purposes of our discrete-time event history analysis, with respondents contributing one observation for each year at risk of first birth from ages 15 to 21 (N = 32,595 person-year observations). To examine racial variation, we analyze a subset of 3,990 respondents (N = 25,799 person-year observations) who identify as non-Hispanic Black or non-Hispanic White. We exclude other racialized groups because of insufficient sample size and geographic variation in our sexism measures.
Measures
Early childbearing
Our outcome of interest is the transition to early childbearing. We created an indicator for whether a respondent had a first birth in a given year between ages 15 and 21. We calculate age at first birth using a respondent’s own birth date and the date of first birth (if any), which was collected in the live birth roster at the Wave IV follow-up in 2008, when respondents were 24 to 32 years old. We use the Wave V birth roster to fill in missing data from Wave IV (91 percent of early births were reported at Wave IV). Figure 3 plots the unweighted cumulative percentage of respondents transitioning to first birth by age within our analytic sample; 27 percent had a birth by age 22. Dashed lines show estimates separately by racialized status: 23 percent of non-Hispanic White respondents and 36 percent of non-Hispanic Black respondents had an early first birth.
Cumulative Percentage of Female Respondents with First Birth by Age
Institutional structural sexism
Following Homan’s (2019) approach, we combine indicators of gender inequality in economic, cultural, and reproductive health resources at the societal level to generate an index of institutional sexism. All measures come from the Add Health contextual file. Our indicator of economic inequality is the ratio of men’s to women’s labor force participation at the county level among residents age 16 and older from the 1990 Census of Population and Housing. A ratio above 1 indicates stronger economic conditions for men relative to women. Analogously, we assess conservative cultural ideology with two indicators at the county level: the proportion that voted Republican in the 1992 presidential election and the proportion of conservative religious denomination adherents in 1990. Higher proportions indicate more traditional orientations to gendered work and family roles (Glass and Levchak 2014; Glass and Nath 2006).
We include four indicators of restricted reproductive health resources. Abortion law variables are measured at the state level and come from the 1995 NARAL Abortion Access Data. Limited public state funding has a value of 1 if abortion funds are limited to cases of rape, incest, or life endangerment, and 0 if they are available in all or most circumstances. Waiting period after informed consent for abortion is coded 1 for states with mandatory waiting periods and 0 otherwise. Waiting periods are a clear signal of political intervention in women’s healthcare; they are not medically indicated for abortion and impose burdens, particularly for women traveling for abortion care (de Londras et al. 2022; Myers 2021). Parental consent is coded 2 for states with enforced laws that require minors seeking abortion to obtain parental consent or give notice, 1 for states with unenforced parental consent laws, and 0 for states with no parental consent laws. Our fourth indicator of restricted reproductive health resources is measured at the county level and comes from the 1992 Alan Guttmacher Institute Files: no abortion providers in a respondent’s county is coded 1 if there are no providers and 0 if there is at least one provider. Values of 1 or higher on these measures indicate greater restrictions to women’s autonomy and reproductive health (Guldi 2008; Joyce and Kaestner 2001).
Weighted descriptives of our seven indicators of economic, cultural, and reproductive health resources are shown in Table 1, Panel A, by early birth status. To create our institutional sexism index, we standardize all seven measures, sum them, and divide by the standard deviation (unweighted mean = 0, std. dev. = 1, Cronbach’s alpha = 0.71). Higher values on the indicators and index indicate a higher degree of sexism, and point to higher levels of exposure to institutional sexism during adolescence among respondents with an early birth relative to those without.
Weighted Descriptive Statistics for Structural Sexism Variables by Birth status
Source: Add Health.
Parental consent law for abortion is based on a nominal variable with categories coded as (0) no parental consent law, (1) unenforced parental consent law, and (2) enforced parental consent law.
p < 0.05; **p < 0.01; ***p < 0.001 for differences between respondents with early first birth versus no early first birth (two-tailed tests).
Interactional structural sexism
We combine indicators of gendered expectations, power dynamics, and conservative religious beliefs that manifest in everyday interactions with peers at school to create an index of interactional sexism. All indicators are drawn from the Wave I in-home survey and aggregated by school to capture the interactional peer context that respondents experienced during adolescence. This approach avoids the social desirability bias inherent in asking people to self-report their own perceptions of sexism or individual-level sexist beliefs.
Gendered expectations about sex are captured with three measures. First, we created an aggregate measure of positive norms about sex and pregnancy by averaging responses to five questions: (1) if you had sexual intercourse, your partner would lose respect for you; (2) if you had sexual intercourse, afterward, you would feel guilty; (3) if you had sexual intercourse, it would upset [name of mother]; (4) if you got (someone) pregnant, it would be embarrassing for your family; (5) if you got (someone) pregnant, it would be embarrassing for you. Respondents rated their agreement with each statement on a 1 to 5 scale (1 = strongly agree, 5 = strongly disagree), such that higher values indicate more positive norms. Second, respondents were asked about activities that would be part of their ideal hypothetical relationship, and we created a dichotomous variable for whether this included sex. Finally, respondents were asked how their residential mother would feel about them having sex at this time in their life, using a scale of 1 to 5 (1 = strongly disapprove, 5 = strongly approve). For these three measures, we calculated average responses for male and female respondents within schools, and then took the male/female ratio such that values over 1 indicate more positive responses about sex among young men in schools (we top-coded a small share of outlying values on the ideal relationship to the 99th and 1st percentiles).
Gendered power dynamics are captured by disparities in domestic labor, wages, and ages of romantic partners. Frequency of housework such as cleaning, cooking, or yardwork is measured with a 0 to 3 scale (0 = not at all, 1 = one or two times, 2 = three or four times, 3 = five or more times). Hourly wage for a typical non-summer week is based on respondent’s self-reported work hours and pay. We calculated average housework and pay for male and female respondents within a school, and then took the male/female ratios such that values over 1 indicate male advantage (i.e., less housework, higher pay; we top-coded a small share of outlying ratios to the 99th and 1st percentiles). Age difference between male and female romantic partners is based on respondent reports of their romantic partner’s age at the start of the relationship. We calculate the ratio of male/female ages in the respondent’s first reported different-gender romantic relationship, then take the average within each school.
Finally, we include an indicator of conservative religious beliefs among school peers, measured by the proportion of school peers who agree that the sacred scriptures of their religion are the word of God and completely without mistakes. Higher shares indicate more rigid gender roles and stronger support for men’s decision-making authority (Glass and Levchak 2014; Glass and Nath 2006).
Weighted descriptives of our seven indicators of gendered expectations, power dynamics, and conservative religious beliefs are in Table 1, Panel B, by early birth status. We standardize all seven interactional measures, sum them, and divide by the standard deviation to create our interactional sexism index (unweighted mean = 0, std. dev. = 1; Cronbach’s alpha = 0.37). Higher values on the indicators and index reflect a higher degree of sexism. Respondents with an early birth experience higher levels of interactional sexism during adolescence compared to those who did not have an early birth.
In summary, we measure institutional and interactional sexism at different levels of the ecological context. Institutional sexism is measured at the macro level, using state and county characteristics of the economy, cultural ideology, and public policy. Interactional sexism is measured at the meso level and taps into peer attitudes and behaviors within schools. In the Robustness Checks section, we assess the sensitivity of our institutional and interactional structural sexism indices to the exclusion of particular indicators.
Potential confounders
As noted earlier (see also Figure 2), our multivariable models include a set of covariates that potentially confound the relationship between structural sexism and early childbearing. These covariates include a (time-varying) continuous variable for calendar time to account for period changes jointly affecting the broad cultural context and transitions to parenthood. At the county level, we control for Wave I percent in poverty (in quintiles) and racial/ethnic composition (percent Black and percent Hispanic). At the school level, we include the proportion of students who are female. We experimented with additional controls for school characteristics but removed them due to collinearity with county and respondent characteristics.
Respondent and family controls
Our multivariable analyses account for respondent and family characteristics measured at Wave I that are plausibly exogenous to structural sexism, including racialized status (non-Hispanic White, non-Hispanic Black, Hispanic, and non-Hispanic “other”), nativity (1 = born in the United States), family structure (married biological parents, step-family, single parent, and other, e.g., living with a grandparent), and highest parental education (less than high school, high school/GED, some college or technical school, and bachelor’s degree or more).
Weighted descriptive statistics for respondent and family characteristics and regional, county, and school controls are presented in Appendix Table A1 by birth status. Consistent with prior research (Kost, Maddow-Zimet, and Arpaia 2017; Sick, Spaulding, and Park 2018), respondents with an early first birth are more likely to be non-Hispanic Black, their parents are less advantaged, and their counties are poorer.
Discrete-Time Event History
We evaluate the relationship between sexism and early childbearing using discrete-time logistic regression event history models with random effects to account for the nesting of students within schools. Event history is designed to study transitions by estimating both their occurrence and timing and allowing for careful accounting of exposure to risk (Allison 2014; Singer and Willett 1993). This approach allows respondents to contribute an observation for each year, from age 15 to 21, that they are at “risk” of an early first birth, removes respondents from the exposure to risk after the transition to parenthood, and includes respondents who never have a child. We use a logit specification because it provides a very straightforward interpretation in terms of age-specific odds or probabilities of first birth. In supplementary analyses, we estimated the same models using a complementary log-log specification, which can be a better option for rare outcomes; our results are not sensitive to this choice, and we use logistic regression for ease of interpretation.
We estimate the following model on our person-year data file:
where P is the probability of an early first birth, j indexes schools, and t indexes single years of age from 15 to 21. The log odds of birth at age t nested in school j is an additive function of institutional and interactional sexism measured in adolescence, calendar year at age t, county controls at Wave I, and controls for individual, family, and school characteristics at Wave I. The parameters α1, α2, and α3 represent the baseline hazard, modeled as a quadratic function of age, and µ is a school-level random effect. Models are not sensitive to the functional form of age or calendar year. All models are weighted to account for Add Health’s complex sampling design (Chen and Harris 2020).
We further explore racial variation in structural sexism and early childbearing within our subsample of non-Hispanic Black and non-Hispanic White respondents. We first evaluate racial differences in average levels of exposure to sexism, and we predict age-specific probabilities under different assumptions about the level of exposure to sexism. Next, we estimate racial variation in the association between structural sexism and early birth in our regression models by including interactions between structural sexism and an indicator for non-Hispanic Black (relative to non-Hispanic White).
Results
Sexism and Transitions to Early Childbearing
Summary results from our event history models with all controls are presented in Table 2. This truncated table presents a summary of coefficients on our structural sexism indices; Appendix Table A2 shows full results for sequential models that iteratively add controls. In addition to controls, Model 1 includes our institutional sexism index only, Model 2 includes our interactional sexism index only, and Model 3 includes both sexism indices. The coefficient on institutional sexism in Model 1 is not statistically significant (B = 0.158, p > 0.05). A conservative interpretation is that our first hypothesis is not supported, that is, there is insufficient evidence for a direct relationship between higher institutional sexism and early first birth. We will later demonstrate, however, that the association between institutional sexism and early first birth is moderated by racialized status in our subsample of non-Hispanic Black and White respondents.
Discrete-Time Logistic Regression Event History Models with Random Effects Predicting Early Childbearing (Age 15 to 21)
Source: Add Health.
Note: Standard errors are in parentheses. Models control for time-varying age, age2, calendar year, and Wave 1 measures of adolescents’ race/ethnicity, nativity, family structure, parental education, county demographics (poverty quintiles, percent Black, percent Hispanic), and percent female in school.
p < 0.05; **p < 0.01; ***p < 0.001 (two-tailed tests).
Model 2 indicates that a one standard deviation increase in interactional sexism is associated with a 37 percent (B = 0.312, eB = 1.37, p < 0.01) increase in the age-specific probability of early birth. These results support our second hypothesis, predicting a positive relationship between interactional sexism and early childbearing. The coefficient on interactional sexism is slightly attenuated and remains statistically significant in Model 3 (B = 0.302, eB = 1.35, p < 0.01), which includes both institutional and interactional sexism. These results support our third hypothesis, which posited that exposure to interactional sexism in adolescence would be more salient than institutional sexism for predicting early childbearing.
To provide a more meaningful interpretation of this association, Figure 4 shows the average adjusted predicted probabilities of fist birth at any age by interactional sexism. These probabilities are generated from coefficients in Table 2, Model 3, across the distribution of interactional sexism (mean, ±1 std. dev., ± 2 std. dev.), holding all other variables at their means. When exposed to interactional sexism that is 1 standard deviation below the mean, an average of 2.5 percent of young women are predicted to have a child in any given year between ages 15 to 21, compared to 4.5 percent when exposed to interactional sexism that is 1 standard deviation above the mean.
Predicted Probabilities of Early Childbearing by Interactional Sexism
To put these results into further context, we calculated predicted probabilities of early birth by age p15 – p21, and then calculated a cumulative predicted probability of having a first birth by age 22 by multiplying the conditional yearly predicted probabilities: 1 – (1 – p15)(1 – p16) . . . (1 – p21). At the average level of interactional sexism, 22 percent of women are predicted to have a birth by age 22, compared to 17 percent of women exposed to 1 standard deviation lower and 29 percent of women exposed to 1 standard deviation higher. This amounts to a 12 percentage-point difference in the cumulative probability of early birth for young women exposed to relatively low (–1 std. dev.) and high (+1 std. dev.) levels of interactional sexism.
Our statistical models are set up to estimate the independent and additive effects of institutional and interactional sexism. It is possible, however, that the effects of institutional sexism are amplified or diminished at different levels of interactional sexism, and vice versa. For example, exposure to high levels of institutional sexism may be buffered by low levels of interactional sexism within schools. We did not find evidence to support this hypothesis when we included a statistical interaction between institutional and interactional sexism. This result is consistent with prior research, which also found that macro, meso, and micro levels of structural sexism have additive effects on health (Homan 2019).
Differences across Racialized Groups
To evaluate how sexism may be experienced differently by adolescents racialized as non-Hispanic White and non-Hispanic Black, we first describe differences in exposure to structural sexism. Table 3 shows mean values for our indicators of institutional structural sexism (Panel A) and interactional structural sexism (Panel B), separately for non-Hispanic White and non-Hispanic Black respondents. Panel A shows that both groups are exposed to similar levels of institutional sexism in adolescence. In contrast, Panel B shows that Black adolescents are exposed to significantly higher levels of interactional sexism. Black adolescents face greater gender disparities in expectations about sex (in particular, perceiving that their mothers are more disapproving of sex) compared to White adolescents, there are larger male/female age differences in peers’ romantic relationships, and more of their school peers have conservative religious beliefs. Combining all indicators, differences are striking, with a mean interactional sexism index of 0.59 among Black adolescents (more than half a standard deviation above the overall sample mean), relative to –0.18 among White adolescents. These results provide partial support for our fourth hypothesis, which predicted that Black adolescents would have higher exposure to structural sexism compared to White adolescents.
Weighted Descriptive Statistics for Structural Sexism Variables, by Racialized Group
Source: Add Health.
Parental consent law for abortion is based on a nominal variable with categories coded as (0) no parental consent law, (1) unenforced parental consent law, and (2) enforced parental consent law.
p < 0.05; **p < 0.01; ***p < 0.001 for differences between non-Hispanic White and non-Hispanic Black respondents (two-tailed tests).
We next assess how the association between structural sexism and early childbearing differs for non-Hispanic White and non-Hispanic Black women. Results are shown in Table 4. Models 1 and 2 show results from our baseline event history model (Equation 1) on the additive estimated effects of institutional and interactional sexism. Estimates for this subsample of non-Hispanic Black and White young women are very similar to estimates for the full sample (Table 2, Models 1 and 2). Notably, we again see a non-significant coefficient on institutional sexism, and a statistically significant and meaningful coefficient on interactional sexism. Model 3 adds the statistical interaction between institutional sexism and an indicator for non-Hispanic Black (ref.: non-Hispanic White). Institutional sexism is a statistically significant predictor of early childbearing among young Black women (B = 0.095 + 0.282, p < 0.05), but not among young White women (B = 0.095, p > 0.05). By contrast, interactional sexism is statistically significantly and positively associated with early childbearing for both groups (Table 4, Model 4). Model 5 includes both institutional sexism and interactional sexism, as well as their statistical interactions with race. Again, this model provides evidence that institutional sexism is linked with early childbearing for Black women but not for White women, providing partial support for our fifth hypothesis, predicting a stronger relationship between structural sexism and early childbearing among women racialized as Black compared to women racialized as White.
Discrete-Time Logistic Regression Event History Models with Random Effects Predicting Early Childbearing (before age 22), with Moderation by Racialized Group
Source: Add Health.
Note: Standard errors are in parentheses. Models control for time-varying age, age2, calendar year, and Wave 1 measures of adolescents’ race/ethnicity, nativity, family structure, parental education, county demographics (poverty quintiles, percent Black, percent Hispanic), and percent female in school.
p < 0.05; **p < 0.01; ***p < 0.001 (two-tailed tests).
Although the statistically significant coefficients on the interaction terms between institutional sexism and race indicate nonlinearity in the log odds, it is not a definitive test of nonlinearity in the predicted probabilities (Mize 2019; Mustillo, Lizardo, and McVeigh 2018). To evaluate the statistical significance of differences in the predicted probabilities by racialized status, we calculated the average marginal effects of race (pBlack – pWhite) across the range of structural sexism values. We again find evidence that the association between institutional sexism and early childbearing varies by race, but interactional sexism has similar associations among both White and Black respondents. Figure 5 plots the adjusted predicted probabilities of early childbearing by institutional sexism and race (Mize 2019). It illustrates a steeper slope for institutional sexism among Black women and shows that differences in the predicted probabilities of early childbearing between White and Black women become statistically significant at higher levels of institutional sexism (at or above the mean). When exposed to institutional sexism that is 1 standard deviation above the mean, an average of 4 percent of White females are predicted to have a child in any given year between ages 15 to 21, compared to 7 percent of Black females.
Predicted Probabilities of Early Childbearing by Institutional Sexism and Racialized Status with 95 Percent CIs, Calculated from Coefficients in Table 4, Model 5
Although the association between interactional sexism and early childbearing is similar by race, Table 3 shows that exposure to interactional sexism was higher among young Black women compared to White women. Inequality in exposure to sexism may help us understand higher rates of early childbearing among Black women (Ward et al. 2019). To flesh out the implications of this inequality, we used results from Table 4, Model 5, to calculate the cumulative predicted probabilities of early childbearing at different levels of interactional sexism—the average level experienced by White adolescents in our sample (–0.18), and the average level experienced by Black adolescents in our sample (0.59), from Table 3—while holding other variables at their means. We estimate that if Black adolescents were exposed to the level of interactional sexism experienced by White adolescents in our sample, their adjusted cumulative probability of early birth would be 17 percent lower by age 22 (24 versus 29 percent, respectively). This simulation is not a causal prediction, but a descriptive exercise that underscores the magnitude of differences in exposure to interactional sexism by racialized status.
Robustness Checks
We conducted a set of robustness checks to assess the possibility of confounding due to unobserved or omitted regional and state characteristics that may vary in their effects on early childbearing over time. As noted earlier, our main models exclude region and state fixed effects so that we can directly measure aspects of sexism that would (potentially) be accounted for in a fixed-effect specification. One concern with this approach is that the association between institutional sexism and early childbearing is confounded by regional characteristics that co-vary with but are distinct from sexism. In particular, the Southern region of the United States has the highest levels of institutional sexism and early childbearing, and it is possible that the association between the two is due to some other cultural or economic characteristic. We evaluated the sensitivity of our results to the inclusion of region-by-year fixed effects, and results in Appendix Table A3 (Models 1 and 3) illustrate that our findings are robust. We continue to find racial variation in the association between institutional sexism and early childbearing, and our estimated effects of interactional sexism differ little from our main models, suggesting that unobserved regional differences do not drive our results.
A related concern is that our conclusions about the salience of interactional sexism are confounded by unobserved characteristics of the macro environments in which schools are situated. To address this possibility, we estimated models with state-by-year fixed effects (Appendix Table A3, Models 2 and 4). 2 Estimated effects of interactional sexism differ little from our main models, lending further support to the idea that school context is particularly relevant for adolescents, who may have less direct connection with the institutional context. We note that state-by-year fixed effects also account for state-level components of our institutional sexism index (namely state abortion laws), and thus estimates of institutional sexism in these models are identified only from variation in county-level gender inequality.
We examined alternative specifications of the institutional and interactional sexism indices to assess whether our measures are tapping widespread structural phenomena or specific domains of sexism. We first conducted a “leave-one-out” analysis (Homan 2019), in which we re-calculated our sexism indices, omitting a single item each time and re-estimating our main early childbirth models in Table 2, Models 1 and 2. We show that the coefficients on institutional sexism across specifications are not statistically significant at conventional levels, consistent with results using the full index (Appendix Table A4). The coefficients on interactional sexism are of a similar magnitude and statistically significant across specifications (Appendix Table A5), indicating that no one single item or dimension of interactional sexism is disproportionately driving our results, consistent with the idea that the gender system is simultaneously created by several complementary processes (Ridgeway and Correll 2004:512). Second, we performed a single item analysis using each individual standardized component of institutional sexism (Appendix Table A6) and interactional sexism (Appendix Table A7) as separate predictors of early childbearing. Across seven individual indicators of institutional sexism, only two are not signed in the expected (positive) direction, and these are not statistically significant. All interactional sexism items are related to early childbearing in the expected (positive) direction. Not all are statistically significant as independent predictors, but these components collectively contribute to the measurement of structural sexism.
Discussion
Childbearing decisions seem intimate and personal, yet our study provides evidence they are shaped by the social structures in which individuals are embedded. We evaluated structural sexism in the United States and its relationship with a critical aspect of reproductive health and well-being—the timing of first birth. We examined institutional structural sexism, which encompasses gender inequality in economic, political, religious, and family planning policy. We also generated a novel measure of interactional structural sexism to capture gendered expectations, power dynamics, and conservative beliefs that are reified through everyday interactions with classmates. Broadly, we found that when young women are exposed to structural sexism in adolescence, they are more likely to become young parents. This work sheds light on the structural forces that shape young women’s opportunities and constraints, and that potentially exacerbate inequality via fertility and family formation patterns.
Highlighting the inherently intersectional nature of sexism and other structural inequalities, we uncovered significant variation in the relationship between institutional sexism and early childbearing across racialized groups. Namely, we found that institutional sexism is a significant predictor of early childbearing among young non-Hispanic Black women, but not among their non-Hispanic White counterparts. This finding is consistent with a reproductive justice framework, which highlights structural barriers that limit reproductive choices for women from minoritized racial backgrounds (Ross and Solinger 2017). White women, by contrast, may be better positioned to leverage resources that mitigate institutional constraints on their reproductive lives. For example, there is evidence that restrictive abortion laws disproportionately affect women of color because of poorer access to and quality of reproductive care (Becker and Tsui 2008; Fuentes 2023; Solomon 2021). Relatedly, the weathering hypothesis posits that early childbearing is a rational response to systemic disadvantage and health uncertainty (Geronimus 1987, 1996). The link between structural sexism and poor health among women is layered onto inequities in other social and economic structures that cumulatively disadvantage Black women. In this context, early childbearing can be advantageous because it coincides with a period of better health and relatively low earnings penalties (Geronimus 1987). Although our analysis cannot identify the specific mechanisms, our findings suggest that institutional gender inequality is more germane to young Black women’s transitions to parenthood.
We also find a significant positive association between interactional sexism and early childbearing among both White and Black young women. The interactional context is especially salient for adolescents, who are increasingly taking cues about gender roles from their peers but may have less direct exposure to institutional structures such as the economy or social policy. Although the estimated effects of interactional sexism operate similarly across racialized groups, we highlight significantly higher exposure among Black adolescents that is, on average, fully half a standard deviation higher than the overall sample. This racial disparity in interactional sexism could reflect the combined effects of both racism and sexism, which manifests more readily in the interactional context within Black students’ schools. We use this differential exposure to probe one potential pathway that produces racial disparities in early childbearing. We find that if Black adolescents were exposed to the same level of interactional sexism as White adolescents, their prevalence of early childbearing would be 17 percent lower. This comprehensive approach to evaluating racial variation in both levels of exposure and the effects of sexism contributes to our understanding of how structural inequalities shape patterns of family formation.
Our study has some important limitations. We draw on empirical and theoretical models of structural sexism and fertility that are heteronormative and conceptualize gender as a binary construct (men and women). Although early childbearing in the United States largely arises from sexual activity between males and females, and sexism is typically understood as discrimination against women on the basis of sex, this lens misses key aspects of gender identity, sexuality, and discriminatory gender systems. An increasing number of individuals identify as non-binary or transgender (Carpenter, Eppink, and Gonzales 2020), and many sexual and gender minority families have children (Reczek 2020). Furthermore, sexual minority women are at greater risk of teen pregnancy than their heterosexual peers, and many pregnancies among sexual minority youth are due to sexual violence (Charlton, Nava-Coulter, et al. 2019; Goldberg, Reese, and Halpern 2016). These dynamics are particularly salient given evidence that both sexist and heterosexist structural inequality are linked with poor birth outcomes (Everett et al. 2022). Future work should consider how different types of identity-based discrimination work together to create a larger system of structural disadvantage for women, people of color, and sexual and gender minorities.
Relatedly, our initial steps to evaluate racial differences in exposure to structural sexism and linkages with early childbearing do not directly examine structural sources of racial inequality or racism. Intersectionality theory emphasizes that intersecting systems of structural inequality, such as racism and sexism, create vulnerabilities for people in marginalized social positions (Collins and Bilge 2020; Crenshaw 1991). Although we examine differential experiences with sexism based on individual-level intersecting identities (i.e., White woman, Black woman), we want to be clear that the structural roots of racism and sexism shape these identities as social positions, and their manifestation of racial disparities does not reflect biological or innate characteristics (Williams 2024). We were limited in operationalizing a fully intersectional approach by our reliance on within-school variation to tap interactional contexts; given patterns of educational and residential segregation, we did not have sufficient racial variation within schools to create a measure of interactional racism. Future research could potentially leverage racial inequality between schools (e.g., Polos et al. 2022) to gain traction on this issue. We were also limited to examining racial disparities among Black and White respondents because of insufficient geographic variation among Hispanic and Asian adolescents in our sample. This limitation represents another potential avenue for future research.
Our study focuses on early childbearing among female-identified respondents because they are more likely to accurately report an early birth than males (Joyner et al. 2011), and they disproportionately bear the costs of raising children (Budig and England 2001). Nevertheless, men’s roles in families are shifting and they are assuming related but unique costs, such as increased expectations for breadwinning and economic support (Forste 2002; Goldscheider, Bernhardt, and Lappegård 2015; Goldscheider and Kaufman 1996; Townsend 2002). Existing research on structural sexism is compelling, in part, because it demonstrates differential effects on both women’s and men’s health outcomes (Homan 2019; Homan and Burdette 2021). Future research could thus explore how structural sexism affects the family formation patterns of both men and women.
Finally, it is worth reiterating the limitations of our data and approach to identify causal relationships. Our empirical strategy relies on cross-sectional geographic variation and observable characteristics to estimate the relationship between structural sexism and early childbearing. We show that our estimated effects are robust to various measurement strategies and fixed effects for region and state, but we cannot rule out potential confounding due to omitted or unmeasured variables within counties, schools, or individuals. Measurement error also remains a source of potential bias. The structural features of inequality and how they shape health, family, and well-being raise critical questions with limited possibilities for experimental manipulation.
Despite these limitations, our study makes several contributions to our understanding of gender inequality and early childbearing in the United States. First, we extend the theoretical scope of the structural sexism perspective, which has been predominately applied to disparities in physical health, by demonstrating its relevance for early transitions to parenthood. Existing research on early childbearing in the United States provides a solid foundation for understanding the contributing roles of individual characteristics, couple dynamics, and macro social structural contexts. The structural sexism perspective highlights gender inequality as a structural characteristic that shapes women’s opportunities across these domains. When women have fewer economic opportunities than men, it decreases the opportunity costs of early childbearing. Restrictions on reproductive healthcare and abortion limit women’s ability to control the timing and spacing of their births, and religious and political conservatism is related to poor health and birth outcomes for women. Gendered expectations for future opportunities and unequal power dynamics are further developed through interactions with peers.
We examined exposure to structural sexism during the critical developmental stage of adolescence and evaluated its relevance for the timing of first birth. Prior research has explored structural sexism’s role for adult health outcomes but has generally been agnostic about the timing of exposure within the life course. Adolescence is a crucial period for developing expectations for future fertility and family formation (see Guzzo et al. 2019). Although sexism can be harmful at any age, gendered norms and power dynamics reified in the interactional context during adolescence can serve as a blueprint for future relationships. Exposure during adolescence can be particularly harmful if the consequences accumulate over an individual’s lifetime (Ridgeway and Correll 2004). Indeed, there is some evidence that exposure to sexism in childhood has enduring effects on women’s labor market outcomes and childbearing patterns into adulthood (Charles, Guryan, and Pan 2022). Alternatively, it is possible that interactional sexism is relevant during adolescence but recedes in importance as individuals age. As young adults enter the workforce, institutional sexism may become a more central consideration for fertility decisions. Future research could further investigate whether institutional and interactional sexism have age-specific effects on childbearing and family formation.
Our study demonstrates the relevance of schools for structural sexism research and answers the call to investigate a wider variety of settings that produce interactional sexism (Homan 2019). Everyday interactions with peers in school can create or reinforce gendered expectations for women’s roles in the labor force, the home, and romantic relationships. Future research should explore additional interactional spaces that potentially manifest structural sexism for adolescents, such as neighborhoods, families, and extracurricular organizations. In addition, studies on early childbearing should pursue measures that tap into individual manifestations of structural sexism, such as internalized gender role attitudes and perceived gender discrimination or sexual harassment.
This line of research is particularly relevant as U.S. women face very different economic, cultural, and political environments based on the states in which they live. In June 2022, the Supreme Court’s decision in Dobbs v. Jackson overturned Roe v. Wade, giving states the ability to restrict or ban abortion and calling into question other rights predicated on the right to privacy, including contraception, same-sex consensual sex, and same-sex marriage (Stolberg 2022). In addition to creating obstacles to reproductive healthcare, these policies also contribute to economic inequality and gendered norms that constrain women’s roles and future opportunities. The socio-legal context sends a strong message about women’s roles in society even for those who are not seeking abortion. The rolling back of abortion rights has called attention to the wide variation in the gender climate in the United States and its potential implications for women’s health and well-being. Our results underscore the importance of structural forces for shaping family formation—and raise questions about increasing polarization in experiences of structural sexism in a post-Dobbs world.
Footnotes
Appendix
Summary of Standardized Interactional Sexism Components as Independent Predictors of Early First Birth
| B(SE) | B(SE) | B(SE) | B(SE) | B(SE) | B(SE) | B(SE) | B(SE) | |
|---|---|---|---|---|---|---|---|---|
| Interactional sexism scale | .312** (.104) |
|||||||
| M/F positive norms about sex/pregnancy | .048 |
|||||||
| M/F ideal relationship includes sex | .092 |
|||||||
| M/F perceived mother approval of sex | .111 |
|||||||
| F/M housework frequency | .058 |
|||||||
| M/F hourly wages | .096 |
|||||||
| M/F age in relationships | .227** |
|||||||
| Proportion with scriptural inerrancy beliefs | .174 |
|||||||
| Unweighted unique observations | 5,052 | 5,052 | 5,052 | 5,052 | 5,052 | 5,052 | 5,052 | 5,052 |
| Unweighted person-year observations | 32,595 | 32,595 | 32,595 | 32,595 | 32,595 | 32,595 | 32,595 | 32,595 |
Source: Add Health.
Note: Models control for time-varying age, age2, calendar year, and Wave 1 measures of adolescents’ race/ethnicity, nativity, family structure, parental education, county demographics (poverty quintiles, percent Black, percent Hispanic), and percent female in school. Standard errors are in parentheses.
p < 0.05; **p < 0.01; ***p < 0.001 (two-tailed tests).
Acknowledgements
The authors thank Ken Bollen, Jason Fletcher, Taylor Hargrove, Kathie Harris, Bob Hummer, Patrick Ishizuka, and Megan Sweeney for helpful suggestions on this work and comments on previous drafts. We thank Rebecca Bielamowicz for excellent research assistance. We are also grateful to seminar participants at the Census Bureau’s Center for Economic Studies and Bowling Green State University’s Center for Family and Demographic Research.
Funding
Research reported in this publication was supported by NICHD of the National Institutes of Health under award number P2C HD050924.
Data Note
This research uses data from Add Health, funded by grant P01 HD31921 (Harris) from the Eunice Kennedy Shriver National Institute of Child Health and Human Development (NICHD), with cooperative funding from 23 other federal agencies and foundations. Add Health is currently directed by Robert A. Hummer and funded by the National Institute on Aging cooperative agreements U01 AG071448 (Hummer) and U01AG071450 (Aiello and Hummer) at the University of North Carolina at Chapel Hill. Add Health was designed by J. Richard Udry, Peter S. Bearman, and Kathleen Mullan Harris at the University of North Carolina at Chapel Hill. Replication materials are available at 
.