Artículo de revista: Abufhele, Alejandra; Pesando, Luca Maria; Castro, Andrés. 2022. “Parental educational similarity and inequality implications for infant health in Chile: Evidence from Administrative Records, 1990–2015.” Research in Social Stratification and Mobility, 82 100736 (ISSN 0276-5624) https://doi.org/10.1016/j.rssm.2022.100736. Research in Social Stratification and Mobility 82 (2022) 100736 Contents lists available at ScienceDirect Research in Social Stratification and Mobility journal homepage: www.elsevier.com/locate/rssm Parental educational similarity and inequality implications for infant health in Chile: Evidence from administrative records, 1990–2015 Alejandra Abufhele a, 1, Luca Maria Pesando b, c, *, 2, Andrés F. Castro T. d, 3 a Escuela de Gobierno, Universidad Adolfo Ibáñez, Chile b Division of Social Science, New York University (AD), 1737 Cambridge Street, Cambridge, MA 02138, USA c Weatherhead Center for International Affairs, Harvard University, United States d Center for Demographic Studies, Universitat Autònoma de Barcelona, Spain A R T I C L E I N F O A B S T R A C T Keywords: This study expands existing scholarship on the relationship between parental educational similarity and infant Educational similarity health using rich administrative data from Chile covering births that occurred between 1990 and 2015. We test Parental influences the relationship between parental educational similarity (homogamy) or dissimilarity (heterogamy) and two Infant health measures of infant health, namely low birth weight (LBW) and preterm birth (PB). We show that parental Birth records Gender educational homogamy is associated with a reduced probability of low birth weight and preterm birth – Social stratification particularly at the high end of the educational distribution – and the observed association is only partly driven by Inequality selection into homogamous couples, as demonstrated by complementary quasi-experimental analyses conducted Chile on a subsample of matched step-siblings from same mothers but different fathers. We further show that couples where women outrank men in educational attainment (hypogamy) exhibit worse birth outcomes relative to their homogamous and hypergamous counterparts. Municipality-level analyses merging external information on fe­ male labor force participation (FLFP) prior to childbirth reveal that the association between hypogamy and children’s outcomes is increasingly negative as FLFP increases, highlighting a strong work-life balance tension for educated women who are actively engaged in the labor force. Insights from this study contribute to a better understanding of the inequality debate surrounding the intergenerational transmission of advantage and disadvantage – a topical issue in a country that has recently joined the rank of the world’s wealthiest nations yet maintains extreme levels of socioeconomic inequality. 1. Introduction 1993) and, more recently, in low- and middle-income societies (Esteve, García-Román, & Permanyer, 2012; Gullickson & Torche, 2014; Hu & Educational assortative mating – the non-random matching of part­ Qian, 2015; Pesando, 2021; Smits & Park, 2009). ners with respect to education – represents a critical interplay between Documenting patterns of spousal choice and examining their impli­ the growing importance of education in contemporary societies and the cations are two rather distinct analytic endeavors. In this respect, most role of the family in shaping children’s life chances (Carlson, Mclana­ studies to date have attempted to measure the contribution of educa­ han, & England, 2004; McLanahan, 2004; Schwartz, 2013). Traditional tional homogamy – a scenario in which partners share the same level of assortative mating scholarship has focused on analyzing the patterns of education – to economic inequality (Breen & Salazar, 2011; Eika, educational sorting and identifying the processes that generate such Mogstad, & Zafar, 2019; Gottschalk & Danziger, 2005; Pesando, 2021), patterns. These lines of inquiry have documented an increase in the or the contribution of homogamy to relationship status and transitions propensity of partners to resemble each other in educational attainment (Goldstein & Harknett, 2006), in a predominantly within-generation in high-income societies (Blossfeld & Timm, 2003; Qian & Preston, perspective. Attempts to explore the relationship between parental * Corresponding author at: Division of Social Science, New York University (AD), 1737 Cambridge Street, Cambridge, MA 02138, USA. E-mail addresses: alejandra.abufhele@uai.cl (A. Abufhele), lucamaria.pesando@nyu.edu (L.M. Pesando), acastro@ced.uab.es (A.F. Castro T.). 1 Avenida Diagonal las Torres 2640, Peñalolén. Oficina 255-A, edificio de Pregrado, Chile 2 NYUAD, Saadiyat Campus, Social Science Building A5–1207, P.O. Box 129188, Abu Dhabi, United Arab Emirates 3 Carrer de Ca n’Altayó, Edifici E2, 08193, Barcelona, Spain https://doi.org/10.1016/j.rssm.2022.100736 Received 15 December 2021; Received in revised form 7 October 2022; Accepted 14 October 2022 Available online 21 October 2022 0276-5624/© 2022 The Authors. Published by Elsevier Ltd. This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by- nc-nd/4.0/). A. Abufhele et al. Research in Social Stratification and Mobility 82 (2022) 100736 educational homogamy and children’s outcomes – i.e., cross-genera­ above-stated research question focusing on two measures of infant tionally – have been rare or very much focused on the effects that a health, namely, low birth weight and preterm birth. We document the parent’s status (mostly, the mother) has on the demographic mecha­ intergenerational implications of parental educational similarity in nisms of assortative mating and differential fertility in the US (Maralani, Chile and advance some speculations on the complex interplay between 2013; Mare & Maralani, 2006). This is surprising, as one of the concerns couple and societal-level dynamics that might underlie the heteroge­ behind couples’ educational homogamy is its potential to widen dis­ neity of our results. Drawing on a combination of family systems and parities in the ability of families to invest in their children’s develop­ gender relations perspectives, we stress that birth outcomes may be ment, health, and wellbeing, i.e., the potential to perpetuate social influenced not only by complex economic, social, and emotional in­ inequalities across generations (Fernandez & Rogerson, 2001). In simple teractions of the parents, but also by different micro-level household terms, provided that both men and women have access to schooling dynamics stemming from couples’ relative similarities and the societal opportunities, a society in which high-educated individuals marry context in which these are embedded. high-educated individuals and low-educated marry low-educated will Our focus on birth outcomes in the Chilean context relies on the be more unequal than a society in which high-educated marry general premise that health at birth is a strong predictor of later-life low-educated (or vice versa). This is rather intuitive within generations outcomes such as later-life health, education, and labor market out­ (e.g., societies become more polarized in terms of income and wealth), comes (J. R. Behrman & Rosenzweig, 2004; Bharadwaj, Løken, & Neil­ yet it might well be true across generations thinking, for instance, about son, 2013; Figlio, Guryan, Karbownik, & Roth, 2014; Torche & how heterogeneity in parental resources translates into heterogeneity of Echevarría, 2011). Most importantly, despite the rapid economic growth outcomes of children born to different couple types, thus shaping their that Chile experienced over the last decades, the share of children born later-life outcomes (Bratsberg, Markussen, Raaum, Røed, & Røgeberg, with less than 2500 g – the technical definition of low weight at birth 2018). (LBW) – and children born before 37 weeks of gestation – the technical Both the absolute levels of individual educational attainment and the definition of preterm birth (PB) – has increased over time, echoing relative disparity between parents’ education can be relevant for the recent worrisome evidence from the US (Rauscher & Rangel, 2020). organization of family life and investments in children. First, an accu­ Trends in Fig. 1 show that since the 1990 s the prevalence of these mulation logic suggests that the total level of resources available for negative birth outcomes has increased significantly. In 1990, the pro­ such investments reflects each partner’s contributions – or lack thereof – portion of LBW was 5.7 %, while this rose to 6.3 % in 2015. For PB the of economic, cultural, and social inputs. Second, sorting on education increase was even more marked: from 5.5 % in 1990 to 8.1 % in 2015. can be taken as an indicator of homogeneity in partners’ preferences, Although this pattern is not unique to Chile and partly a reflection of and couples in which both partners have attained similar levels of ed­ increased maternal age and attributable to the fact that very preterm ucation can be expected to suffer fewer frictions; that is, partners’ babies are increasingly likely to survive (Chawanpaiboon et al., 2019; relative similarity on these dimensions may interact positively with the Zeitlin et al., 2013), these trends – inserted within a context of declining level of household resources available and lead to less conflicting educational hypergamy and reversals in gender gaps in education (De decision-making processes, in turn translating into higher or more effi­ Hauw et al., 2017; Esteve et al., 2016; Van Bavel et al., 2018) – make the cient investments in children’s health, schooling, and wellbeing. While study of the intergenerational implications of parental educational accounting for the first channel, our interest here centers on this second similarity in Chile particularly compelling and policy-relevant. dimension of relative parental similarity in education and its intergen­ The causes for a baby to be born low weight or preterm are complex erational implications. and likely the result of the interplay of biological, psychosocial, An incipient line of research has attempted to link child outcomes to behavioral, sociodemographic, and environmental factors (Torche & partners’ similarity in parenting styles (Martin, Ryan, & Brooks-Gunn, Abufhele, 2021). Some of these include prenatal care, uterine infection, 2007) or to the concordance between parenting and marital quality nutrition, economic resources, maternal stress and anxiety during (Belsky & Fearon, 2004), and tends to find beneficial effects of parental pregnancy, lifestyle, diet, tobacco and/or alcohol consumption. All of harmony and concordance. Yet the evidence of interactive influences these factors are in turn influenced by parents’ individual educational between mothers’ and fathers’ characteristics remains limited. Recently, attainment, alongside their combination and complex interplay. Rauscher (2020) expanded the sociological scholarship on the topic by Furthermore, these individual-level factors are just one source of risk estimating the effects of parental educational similarity on infant health during pregnancy and childbirth, as system-level factors such as struc­ using birth records from the United States (US). Hypothesizing that tural inequalities and biases, the social determinants of health, and educational similarity affects infant health through its influence on maternal stress and characteristics of the prenatal context, Rauscher’s results suggest that parental educational similarity is beneficial for in­ fant health (homogamy-benefit hypothesis), with significant variations by birth cohort and maternal education. Focusing on contexts other than the US, Pesando (2022) examined a similar research question using longitudinal data from Ethiopia, India, Peru, and Vietnam, and found evidence in favor of the homogamy-benefit hypothesis, yet only in the more developed and less gender unequal ones, namely Peru and Viet­ nam. Conversely, Behrman (2020) focused on a similar research ques­ tion in Malawi, finding results similar to Pesando (2022) in Ethiopia and India, i.e., that children may be better off in non-homogamous (mostly, hypergamous) unions. The present study seeks to expand the existing scholarship focusing on the relationship between parental educational similarity and chil­ dren’s outcomes in Chile. Chile has experienced rapid economic growth over the past 30 years, accompanied by changes in union formation practices, massive educational expansion, yet persistent gender dispar­ ities and sustained levels of inequality (Celhay & Gallegos, 2015; García & de Oliveira, 2011). Using rich and high-quality administrative data on Fig. 1. Time trends in birth outcomes in Chile between 1990 and 2015 births that occurred between 1990 and 2015, we address the Source: 1990–2015 Birth records, Chilean Ministry of Health. 2 A. Abufhele et al. Research in Social Stratification and Mobility 82 (2022) 100736 financing and policy decisions in the health system may exacerbate shown to be important determinants of birth outcomes (Torche & existing individual-level risk factors (National Academies of Sciences Abufhele, 2021; Zeitlin, Saurel-Cubizolles, & Ancel, 2002). Additionally, Engineering & Medicine, 2020). prenatal care and unhealthy behaviors such as smoking or drinking Estimates from birth records in Chile support the idea that parental alcohol during pregnancy could also be mechanisms that impact birth educational similarity is beneficial for children’s outcomes, particularly outcomes through partners’ interaction; for instance, couples could at the high end of partners’ joint educational distribution. Although the provide emotional support, monitor, and promote attitudes and be­ type of administrative data used does not allow to test explicit mecha­ haviors that translate into better child health outcomes at birth (Chris­ nisms at the individual level, we hypothesize that these benefits are due takis & Fowler, 2011; Duncan, Wilkerson, & England, 2006; Torche & to reduced maternal stress, enhanced complementarity in parental in­ Abufhele, 2021). In what follows, we rely on a combination of theo­ puts towards child production (e.g., more and higher-quality in­ retical perspectives to conceptualize how parents’ interacting charac­ vestments in children), better relationship quality including reduced teristics may play a role in shaping birth-related outcomes. conflict and frictions (resulting, ultimately, in less marital disruptions), and shared lifestyles (including, for instance, healthy habits such as 2.2. Parental educational similarity and health outcomes at birth: physical activity, less smoking, drinking, substance use, etc.). None­ Theoretical perspectives and potential mechanisms theless, the above idea is only partially supported, as we also show that couples where women outrank men in educational attainment (hypog­ Family systems and gender relations perspectives complement each amy) exhibit worse birth outcomes relative to their homogamous other as they conceptualize women-to-men relations at two different counterparts, while couples where men outrank women (hypergamy) levels. First, at the couple level, family systems theory describes how men exhibit similar birth outcomes relative to a homogamy scenario. As such, and women interact when they live together. This perspective focuses the homogamy-benefit hypothesis only holds relative to an educational primarily on within-couple dynamics, including the degree of in­ hypogamy scenario. terdependencies between partners and partners’ characteristics (Kerr, To compensate for the lack of identification of individual-level 2000). Second, at the society level, gender relations theory describes how mechanisms, we push the analysis forward by resorting to intra-couple dynamics (including within-couple bargaining and decision municipality-level variables on female labor force participation – in making processes) are influenced by social norms and perceptions that years prior to childbirth – obtained from an ancillary Chilean survey. hinder gender equality (Acker, 1992). Gendered institutions and social Taking municipality-level female labor force participation (FLFP) as a norms, and the differential perceptions on women and men’s roles distal measure of the social role that women play in society, we conduct within society, manifest in sex gaps in educational attainment, income, heterogeneity analyses to explore whether our results largely reflect distribution of care work, and labor force participation. These differential gender norms. These additional analyses show that the as­ societal-level disparities exacerbate within-couple inequalities as they sociation between hypogamy and children’s outcomes is more negative undermine women’s conditions, for instance, in terms of bargaining as municipality-level FLFP increases (what we label as “double burden” of power and income vis-à-vis men (Agarwal, 1997). We capitalize on the hypogamy). intersection of these two theories to elaborate a micro-meso compre­ Insights from this study contribute to the inequality debate on the hensive view of couples as complex units, embedded in specific intergenerational transmission of advantage and disadvantage (persis­ gendered contexts. tence of inequalities versus fading) and shed additional light on the From a family systems perspective, families function based on com­ relationship between joint parental characteristics and children’s out­ plex economic and emotional interactions, e.g., for pooling resources, comes in a country that has recently joined the rank of the world’s sharing credit, deciding about purchases, or jointly organizing parental wealthiest nations yet maintains extreme levels of inequality and limited practices (Furstenberg, 2005; Kerr, 2000; Minuchin, 1985). Differences multigenerational educational mobility (Celhay & Gallegos, 2015; in the socioeconomic background between spouses could be a source of Daude & Robano, 2015; Torche, 2014). disagreement and conflict due to knowledge and information asymme­ tries and reliance on different systems of values and beliefs. Simple ex­ 2. Background amples could be spouses who observe different religions, or spouses with very different levels of education (Rauscher, 2020). Conversely, 2.1. Parental education and children’s outcomes between-partner resemblance could favor agreement, comfort, and more aligned decision-making concerning parental practices (Garfinkel, The quest for understanding the role of parental education on chil­ Glei, & McLanahan, 2002; Goldstein & Harknett, 2006; Schwartz, 2013), dren’s health outcomes has been prolific (Case & Paxson, 2002; Desai & and therefore potentially affect daily behavior during the prenatal Alva, 1998; Kemptner & Marcus, 2013; Lindeboom, Llena-Nozal, & van period (e.g., smoking, diet, physical activity, etc.). Although not focused der Klaauw, 2009). Studies have documented a positive association on health outcomes, Beck and González-Sancho (2009) found evidence between mother’s educational attainment and birth-related outcomes in support of these mechanisms using US data from the Fragile Families including neonatal, post-neonatal, and infant mortality (Chou, Liu, and Child Wellbeing Study. They documented positive associations be­ Grossman, & Joyce, 2010), birth weight (Chevalier & O’Sullivan, 2007; tween parental educational similarity and children’s school readiness at Currie & Moretti, 2003; Güneş, 2015), antenatal, postnatal care, and age five, postulating enhanced levels of parental agreement about the gestational age (Cantarutti, Franchi, Monzio Compagnoni, Merlino, & organization of family life and symmetry in the allocation of time Corrao, 2017; Ruiz et al., 2015). Although the importance of father’s devoted to childcare as underlying mechanisms. education is more often neglected in the literature, evidence also sug­ The gender relations perspective recognizes that micro-level house­ gests that father’s education matters for children’s health (Chen & Li, hold dynamics are gendered, i.e., that the role of partners’ (dis)simi­ 2009), yet to a slightly smaller degree than mother’s (Cochrane, Leslie, larities on intra-household dynamics interacts with the power & O’Hara, 1982). imbalances between men and women already at play in society due to However, the sole influence of individual characteristics does not tell rooted social norms and perceptions (Agarwal, 1997). For instance, this the whole story. The familial and societal contexts can also influence theory suggests that in patriarchal contexts women’s intrahousehold birth outcomes. The association between maternal stress and birth bargaining power is undermined by societal norms that confine their outcomes has been well established in many and diverse contexts role to care work. Where male breadwinner models are prevalent, men (Beijers, Jansen, Riksen-Walraven, & De Weerth, 2010; Dancause et al., are expected to financially provide for the family; therefore, they are 2011; Torche & Kleinhaus, 2012; Torche, 2011). The marital status and expected to have access to wealth-generating property including pro­ its changing nature in the context surrounding the family have also been ductive assets, land, educational credentials, social/professional 3 A. Abufhele et al. Research in Social Stratification and Mobility 82 (2022) 100736 networks, government support, and labor-market skills (Agarwal, To summarize, the interaction between intrahousehold dynamics 1997). Conformity to these expectations is socially rewarded as several and broader societal forces creates diverse configurations of resource- institutions act upon gendered premises (Acker, 1992). In contrast, availability, union (in)stability, and exposure to stress/discomfort couples that do not fit into this pattern “threaten” gender norms, are across couples with different educational composition (homogamous vs. more often subject to social accountability, and risk incurring into heterogamous). These diverse configurations are relevant for under­ negative societal judgments (Brines, 1994). standing variability and inequality in infant health outcomes because, In addition, although in high-income societies differences between during the prenatal period, partners’ differences in decision-making men and women in socioeconomic outcomes such as educational processes likely accentuate, and social accountability increases attainment have narrowed and even reversed – including in Chile (Rauscher, 2020). Pregnancies imply changes in couples’ daily lives and (OECD, 2021) – sex-differences in salaries, labor force participation behaviors (e.g., dietary restrictions, less physical activity, higher hor­ rates, and the propensity of having part-time jobs persist, suggesting that mone levels, etc.), as well as in the degree of attention that partners traditional gender roles and expectations are still present and pressing in receive from family members, close relatives, friends, and society. contemporary societies (Bittman, England, Sayer, Folbre, & Matheson, 2003; England, Levine, & Mishel, 2020; Litman et al., 2020). These may 2.3. Geographical context diminish both the overall level of resources and the societal acceptance of couples where women perform traditionally ‘male’ roles, and may Socioeconomic and demographic transformations, along with further translate into negative societal inputs for mothers in these po­ persistent levels of inequality and slowly-changing gender roles, are sitions (Blossfeld, 2009). In other words, there might be a tension – or at likely to play a role in the extent to which parental educational similarity least a lag – between increasing gender equality in institutions – what or dissimilarity are associated with children’s outcomes in the Chilean Goldscheider, Bernhardt, and Lappegård (2015) define as the “first context. Cross-sectional evidence indicates that Chile features very phase” of the gender revolution – and shifting gender roles within the strong barriers to intermarriage at the top of the educational distribution household – the “second phase” of such revolution. but a more fluid exchange elsewhere (Torche, 2010). Evidence Hence, family systems and gender relations theoretical perspectives regarding the evolution of assortative mating is, however, scant. Using combined suggest that birth outcomes may be influenced not only by data from the Chilean National Socioeconomic Characterization Survey individual characteristics of the parents, but also by the different micro- (CASEN), Bucca and Urbina (2016) found that educational homogamy level household dynamics stemming from couples’ relative similarities decreased between 1990 and 2013, and the combination of college and the societal context in which these are embedded. Take educational expansion and higher labor-force participation of women favored the attainment as an example – the wealth-generating characteristic we formation of highly educated and high-earner couples. Esteve, McCaa, measure in our analysis due to data availability and to its relevance for and López (2013) used census data from the Integrated Public Use the Chilean context. Under this perspective, educational attainment is a Microdata Series (IPUMS) and showed that educational homogamy in fundamentally different resource for men and women within a couple. Chile increased since the 2000 s, being highest among college graduates, For the former, high educational attainment is somewhat “expected,” yet it did not increase among individuals with less than primary edu­ and its lack may be viewed as a personal failure. For the latter, instead, cation. These studies provide a good background, yet assortative mating high educational attainment could be seen as an “achievement,” as well patterns for couples with children – the analytical focus of this paper – as a feature that threatens gender norms and creates strong status in­ may differ from the above-documented trends. Table 1 suggests that consistencies within the household. As such, high educational attain­ births from homogamous parents modestly increased over the period ment for women may be conducive to sources of distress that accentuate (69.6 % in 1990 and 71.6 % in 2015). Births from hypogamous couples the gendered nature of intrahousehold dynamics. Among couples with also increased over time, from 11. 7 % in 1990 to 15.2 % in 2015, while different educational backgrounds, women’s educational attainment births from hypergamous couples decreased from 18.7 % in 1990 to 13.2 relative to men’s substantially affects the resources, knowledge, and % in 2015. If, as hypothesized, children from these different couple practices that may eventually boost or hinder infant health, including configurations (homogamous, hypogamous, hypergamous) feature dif­ birth outcomes via the prenatal context. Sex-differences in labor-market ferential outcomes at birth, the above trends could directly affect the returns to education – higher for men, on average, including in Chile intergenerational transmission of disadvantage. For instance, as the (Blagrave & Santoro, 2017; Sánchez, Finot, & Villena, 2020) – suggest prevalence of hypogamy is increasing (Table 1), if children from that the availability of resources differs between couples of hypogamous couples feature worse infant heallth relative to their ho­ highly-educated women and less-educated men (hypogamy) and cou­ mogamous counterparts, this could amplify the intergenerational ples of highly-educated men and less-educated women (hypergamy). If transmission of inequalities. women have, on average, lower salaries than men conditional on Even though over the last 30 years Chile has excelled for its high educational attainment and type of job, hypogamous couples will have levels of socioeconomic development accompanied by dramatic educa­ fewer resources than hypergamous ones. tional expansion (Torche, 2005), gender inequality is still a pressing When thinking about status inconsistencies between partners, issue in the Chilean society, where traditional gender-role beliefs persist research has shown how higher socioeconomic status among female (Center for Reproductive Rights, 2010; Contreras & Plaza, 2010). A partners, measured in terms of relative educational attainment, 2010 study by the United Nations Development Programme (UNDP) employment status, and share of earnings, may be associated with reported that 62 % of Chileans were opposed to full gender equality. higher instances of intimate partner violence (Behrman, 2019; Weitz­ man, 2014). In these two cases, the central idea is that men feel Table 1 threatened by women with relatively higher levels of achievement and Distribution of births according to couples’ educational similarity in Chile, they try to regain status within the couple by, for instance, exerting percentage of couples with children in each group by selected years. violence, a series of phenomena referred to as ‘backlash effects’ or Year Homogamy Hypogamy Hypergamy ‘gender deviance neutralization’ (Brines, 1994; Weitzman, 2014). 1990 69.6 11.7 18.7 Although evidence of this specific phenomenon for Latin America is 1995 69.8 12.4 17.8 scant, the high levels of intimate partner violence in the region suggest 2000 69.7 13.1 17.2 that intrahousehold dynamics in Chile are likely to be gendered, and 2005 69.7 13.5 16.8 that a higher relative socioeconomic status among female partners may 2010 71.2 13.8 15.0 trigger unexpected reactions on the part of their male partners (WHO, 2015 71.6 15.2 13.2 2013). Source: 1990–2015 Birth records, Chilean Ministry of Health. 4 A. Abufhele et al. Research in Social Stratification and Mobility 82 (2022) 100736 Many of those surveyed – around 18 % – expressed the belief that 2.4. Hypotheses women should limit themselves to the traditional roles of mother and wife (Estrada, 2010). Also, Boncompte and Paredes (2020) showed that Exploiting (i) heterogeneity across parents’ average level of educa­ while Chilean men and women positively value their partner’s income, tion, (ii) variation across different couple configurations in terms of women give a far higher valuation to their partner’s earnings than men educational attainment (hypogamy/homogamy/hypergamy), and (iii) do. As far as educational expansion is concerned, efforts have been made municipality-level information on female labor force participation to reduce the gender gap, which has been narrowing between 1990 and (FLFP), we empirically evaluate the following hypotheses: 2015 (Ministerio de Desarrollo Social, 2018). Attendance rates increased H1. (Homogamy-benefit): Educational homogamy is positively associ­ for all levels of education, and existing gender differences in attendance ated with desirable birth outcomes relative to educational heterogamy. tend to favor – albeit marginally – women (ComunidadMujer, 2018; Specifically, parental educational homogamy is negatively associated OECD, 2021). When focusing on women with children only – the group with the probability of having low-weight and preterm births. If this is of interest in the current study – we observe that educational attainment confirmed by the data, we speculate that the positive association may among mothers in Chile has improved over the last 25 years. In 1990, ensue from enhanced complementarity in parental inputs towards child 36.5 % of mothers had primary education, and only 9.5 % had tertiary production, better relationship quality and stability, and less conflicting education, while these same percentages were 9.3 % and 37.5 % in decision-making processes among partners with similar educational 2015, respectively (Fig. 2). backgrounds. Moreover, gender inequalities within the household are starker (Contreras & Plaza, 2010). Since childhood, women disproportionately H2. (Homogamy-heterogeneity): As the existing literature suggests that suffer from the unpaid care work burden. Limited existing data suggest higher socioeconomic status (SES) couples in Chile are increasingly that between the ages of 5 and 17, girls spend 50 % more hours per week homogamous and higher-SES couples hold a higher level of (pooled) than boys doing housework and that this pattern does not improve over resources, we hypothesize that couples’ educational similarity may time. Work-inactive women continue to be much more numerous than matter differently for people at different places in the educational dis­ their male counterparts, and the main reason women report to explain tribution. Specifically, educational homogamy is more positively asso­ their inability to study or hold a paid job is related to domestic chores ciated with desirable birth outcomes at the higher end of the educational (ComunidadMujer, 2018). ladder. The reduction in the gender gap in education and labor force H3. (Heterogamy-heterogeneity): As the group of educationally- participation, combined with the persistence of gender inequalities heterogamous couples is heterogeneous, we hypothesize that the asso­ within and outside of households make Chile an interesting case study ciation between parental educational similarity and birth outcomes may for this analysis. This confluence of circumstances allows us to examine differ depending on the couple configuration, with a relationship that is whether parental educational similarity is associated with children’s contingent on the sex of the most-educated parent: outcomes, and whether the evidence is consistent with the mechanisms outlined through the combination of theoretical perspectives outlined H3a. : If the mother is the more-educated parent (hypogamy), we above. Slowly-changing gender roles make Chile a peculiar setting expect a negative association between parental educational dissimilarity compared to, for instance, contexts in which narrowing gender gaps in and infant health. In line with the aforementioned theoretical perspec­ education and labor force participation have also been accompanied by tives, this negative association may be due to lower combined household gradual changes towards more equal social norms regarding the status of resources (driven by lower labor-market returns for women) and less women in society, such as the United States. As such, the study by conformity to traditional gender norms. Rauscher (2020) constitutes a key point of reference for our work, yet there is no a priori reason to expect our findings to align with those H3b. : In line with dose-response and boundary-crossing approaches documented in the US context. Given that Chile still features very high (Mare, 1991), we expect a stronger association (in absolute value) be­ gender inequalities both within households and outside of households, tween parental educational dissimilarity and infant health the wider the which suggests that Chile is far from an “ideal” of gender equality, we differences in spouses’ educational levels. expect social norms to play a stronger role in the Chilean context, thus leading to an array of different potential outcomes relative to evidence H4. (Heterogamy heterogeneity, by municipality): Lastly, we complement from other contexts. hypotheses at the individual level using a meso-level indicator of FLFP at the municipality level proxying for women’s social roles and position within society. As further discussed below, previous research has shown that differences in FLFP across contexts constitute a reasonable proxy for gender norms in the Chilean context (Contreras & Plaza, 2010; Ramírez & Ruben, 2015). If social norms surrounding gender are the driving factor underlying the negative association between hypogamy and chil­ dren’s outcomes, we might expect to observe worse outcomes the higher the FLFP in the municipality before childbirth. A negative gradient over the FLFP distribution would corroborate the idea of a “double burden” for women that are more educated than their husbands, facing increasing access to labor-market opportunities unaccompanied by more equitable opportunities within the household. A scenario of this kind would be consistent with the idea of breadwinner mothers-to-be expe­ riencing more stress, intra-household conflict, and relationship insta­ bility (“double burden” of hypogamy, henceforth). 3. Data and methods Fig. 2. Time trends in mothers’ educational attainment in Chile between 1990 3.1. Data and 2015 Source: 1990–2015 Birth records, Chilean Ministry of Health. For this study, we use the birth registry database available through 5 A. Abufhele et al. Research in Social Stratification and Mobility 82 (2022) 100736 the Chilean Ministry of Health. The data contain all registered births in high school degree (i.e., two-level difference favoring the father) is Chile from 1990 to 2015 from mothers born between 1950 and 1990. classified as ‘strongly hypergamous.’ Conversely, a couple where the This dataset includes information on children’s date of birth, sex, birth mother has a college education and the father has secondary education weight, birth length, and weeks of gestation. The database also provides is classified as ‘strongly hypogamous.’ Cases where the difference in information on parents, including age, educational attainment, marital educational attainment is just one level (primary vs. some high school, status, municipality of residence, and area of residence (urban versus or high school graduate vs. college) are classified as hypogamous and rural). We focus on singleton births (127,698 out of the 6,583,493 births hypergamous. This classification permits to better capture the are multiple births, which corresponds to 2.0 %) to mothers between 25 complexity of crossing educational boundaries in a highly stratified and 40 years old (2,734,105 are not within this age range, which cor­ society like Chile (Torche, 2010). In line with H3b, we expect the as­ responds to 41.5 % of the total), and we restrict the analysis to cases sociation of hypogamy and strong hypogamy (and hypergamy and where information on father’s educational attainment is complete strong hypergamy) with birth outcomes to be in the same direction. Yet, (738,510 fathers without information, which corresponds to 11.2 % of we expect the associations to be magnified following a gradient, i.e., the total). exhibiting larger coefficients in absolute values for the groups defined as This analytical sample provides an appropriate set of births to test strong hypogamy (relative to hypogamy) and strong hypergamy (rela­ our hypotheses. We restrict the analysis to singleton births because the tive to hypergamy). aetiology of birth outcomes is different for multiple births – e.g., mul­ To assess whether parental educational similarity matters similarly tiple births are more likely to have adverse outcomes such as perinatal for people at different places in the educational distribution (in line with mortality (Payne, Campbell, DaSilva, & Koval, 2002). The range for H2), we also construct a measure of average SES at the couple level using mothers’ age allows us to build consistent measures of educational completed years of education. We first take the average between spouses attainment, as by age 25 most women have reached the completion of and then build terciles by average parental education, or average their educational careers. After age 40, births are uncommon (2 % of the parental SES. We take this variable as a proxy for the pooled earnings- sample). Additionally, fathers’ information is necessary to measure potential of the couple. couples’ educational composition. Children of couples where informa­ We control for several factors that could be associated with both tion on the father was missing (7.3 % of the analytical sample) have a parental educational similarity and infant health. These include: higher prevalence of low birth weight and preterm births (as shown in mother’s age, grouped in categories (25–29 years old, 30–34 and Appendix Table A1), yet these births are out of the scope of the paper 35–40), average parental education using completed years of education since our analytical focus is on parental educational similarity – a var­ and the squared term of this variable to capture potential nonlinearities, iable which requires both parents’ education. parity of the birth (0, 1, 2 and 3 or more), age difference between father For the last part of the analysis (i.e., to test H4), we augment birth and mother, rural/urban residence, region of residence, infant’s sex, and records from the Chilean Ministry of Health with data from the National year of birth. The region fixed effects account for time-constant region Socioeconomic Characterization Survey (CASEN), a multi-purpose sur­ differences in infant health. For instance, if there are regions that have vey providing information about the socioeconomic conditions of the better prenatal programs or regions that have better hospitals, this will country’s different social sectors, its most essential deficiencies, the di­ be accounted for in the models. Similarly, the year of birth dummies mensions and characteristics of poverty, and income distribution of address potential changes over time in infant health; for example, households (Bravo & Valderrama Torres, 2011). The survey is con­ public-policy efforts to reduce smoking during pregnancy could improve ducted at non-regular intervals (approximately every two years). We birth weight over time. obtain information on FLFP and merge it with birth records at the mu­ Table 2 presents mean and percentages of the variables of interest. nicipality level using the year – preceding the birth – closest to the birth The first column displays descriptive statistics for the analytical sample. year of the child. 3.2. Measures and summary statistics Table 2 Descriptive statistics – analytical sample and by educational similarity groups. We examine two infant health measures, namely LBW, defined as Mean or Mean or Percent births below 2500 g, and PB, defined as births occurring before 37 weeks Percent of gestation. Analogous outcomes have been investigated in related Homogamy Hypogamy Hypergamy literature on the topic (Rauscher, 2020; Torche, 2011). We measure Low birth weight 4.3 4.3 4.5 4.2 parents’ educational attainment using a five-category variable based on Preterm birth 5.4 5.4 5.6 5.4 parents’ completed years of education: primary (1–8 years), some high Average parental 11.8 11.8 11.6 12.0 school (9–11 years), high school graduate (12 years), some college education Married 63.3 65.6 59.9 61.3 (13–16 years) and college graduate (17 years or more). Analogous Mother’s age classifications have been previously adopted and have proven to well 25–29 42.9 42.4 44.3 42.9 adapt to the Chilean society (Bucca & Urbina, 2019). 30–34 35.5 36.0 34.7 35.2 To measure couples’ educational similarity, we construct three 35–40 21.6 21.6 20.9 21.9 related variables, all built through the five-category educational vari­ Parity ables for both mothers and fathers. The first one distinguishes educa­ 0 24.3 25.1 28.0 19.4 tional homogamy from educational heterogamy, i.e., partners who have 1 37.7 37.5 39.9 36.2 the same level of education from partners who have different levels of 2 23.9 23.1 21.8 27.2 education – irrespective of the specific levels. The second variable has 3 or more 14.1 14.2 10.3 17.3 Father - mother age 2.4 2.4 2.1 2.9 three categories: homogamy (partners who have the same educational difference (years) attainment), hypogamy (mothers having higher educational attainment Male infant 52.1 51.2 51.1 51.1 than fathers), and hypergamy (fathers having higher educational Urban residence 89.8 88.5 89.9 92.8 attainment than mothers). The third variable further splits hypergamous Obs. 3,481,584 1,884,290 741,018 856,276 and hypogamous couples into two groups. When the difference in Note: Mother’s education in this table is reported in three categories, as that is educational attainment is two levels or more, we classify the couple as how we use it as control. However, the main variable in the analysis – parental ‘strongly hypogamous’ and ‘strongly hypergamous.’ For example, a educational similarity – is built using a five-category education variable for both couple where the mother has primary education and the father has a mothers and fathers. 6 A. Abufhele et al. Research in Social Stratification and Mobility 82 (2022) 100736 The other three columns disaggregate descriptive statistics according to three- and five-category variables); xi is a vector of observed charac­ couples’ educational similarity. On average, there were 134 thousand teristics of the child and the mother at the time of delivery (including the births per year to couples where information for fathers and mothers is region of residence), and yt is a year-of-birth fixed effect. We estimate available. Among these total births, 54.1 % occurred to homogamous robust standard errors, yet note that given the use of data for the entire couples, 21.3 % to hypogamous couples, and the remaining 24.6 % to population of Chilean births, significance tests are used mostly heu­ hypergamous couples. The prevalence of low weight and preterm births ristically (in line with the discussion included in Torche, 2011). in the overall sample is 4.3 % and 5.4 %, respectively. These low For the models interacting couples’ educational composition with prevalences are in line with Chile’s level of development and improve­ municipality-level FLFP, besides the controls mentioned above we ments over time in prenatal and maternal care (Gonzalez et al., 2009; added a variable for the poverty level in the municipality (to capture Lopez & Bréart, 2012). contextual economic conditions) and the female-to-male ratio in earn­ Table 2 also provides the first indication that children from hypog­ ings in the municipality (to capture gender inequalities in pay). To show amous couples exhibit worse health outcomes, i.e., higher prevalence of these results, we display predicted probabilities of low birth weight and LBW and PB, compared to homogamous and hypergamous couples. pre-term birth at different levels of municipality FLFP, while full ana­ Homogamous couples are more likely to be married and to live in rural lyses and coefficient estimates are provided in the Appendix. areas than couples with unequal levels of education. In line with what To test the sensitivity of our results to potential confounding and/or we might expect on the relationship between educational homogamy selection concerns, we also (i) rerun analyses restricted to first births, and age homogamy, the group with the biggest age difference between and (ii) run analyses on a restricted sample of step-siblings to account for the father and the mother is the hypergamous group. In contrast, the the fact that maternal preferences for particular characteristics in a hypogamous group presents the smallest age gap. partner could influence educational similarity and infant health simul­ Although the overall percentages of LBW and PB and the differences taneously, thus biasing our estimates. The sub-sample of step-siblings is across couples’ educational similarity groups are low, the significance of a matched sample of infants born to the same mother but a different our results should be assessed in light of the total number of births. For father, meant to isolate maternal characteristics and allowing to esti­ example, for the overall sample, the 0.02 %-point difference in the mate the educational-similarity coefficient more accurately, arguably prevalence of LBW between births to homogamous and hypogamous reducing selection effects.4 To identify changes in father, we use the couples (4.5–4.3 % = 0.02) represents, on average, 570 births under father’s age and mothers’ identification number. If the father is the same 2500 g per year. Over 15 years, we see this difference as constituting a for two (or more) records with the same mother’s id number, his age source of concern for health-related public policies, as it involves more must be consistent at the time of each birth (e.g., if two children were than eight thousand births. born three years apart, the father must be three years older by the time of the second birth). Inconsistencies in father’s age may indicate that the father is a different person. We acknowledge that age inconsistencies 3.3. Analytical approach may also be due to misreporting and lack of precision when reporting fathers’ age. To minimize the influence of misreporting, we only We estimate a series of logistic regression models predicting (1) the consider different fathers in cases where the inconsistency is larger than probability of LBW and (2) the probability of PB, separately. We run five years. three different specifications depending on the measure of educational similarity used. The first specification includes a two-category variable 4. Results of educational similarity (homogamy vs. heterogamy). This specifica­ tion provides a first simple test of the homogamy-benefit hypothesis (H1) 4.1. Educational homogamy and the homogamy-benefit hypothesis by average parental SES (H2). In the second specification, we use a the three-category educational-simi­ Table 3 presents odds ratios from logistic regressions predicting LBW larity variable (homogamy, hypogamy, and hypergamy) to examine (1) and PB (2) as a function of parental educational similarity (homog­ further the role of couples’ educational composition and within-couple amy vs. heterogamy). Our results provide evidence aligned with the gender dynamics (H3). This specification tells us whether and, if so, homogamy-benefit hypothesis (H1) when the whole heterogamy category how the direction of the difference in educational attainment – whether is taken as reference. Parental educational homogamy is associated with it favors the mother or the father – is related to birth outcomes (H3a). In a reduced probability of LBW and PB, i.e., positive birth-related out­ the third specification, we test the sensitivity of the results obtained comes. Specifically, the odds of being LBW and PB for an infant born to a using a five-category variable of educational similarity to compare homogamous couple are 2.0 % and 2.2 % lower than those of an infant health outcomes across homogamous, hypogamous and strongly born to an heterogamous couple. Coefficients on control variables hypogamous, and hypergamous and strongly hypergamous couples display expected associations with birth outcomes in terms of magnitude (H3b). To test whether the association between hypogamy and chil­ and direction. For instance, LBW is less prevalent among more educated dren’s outcomes is more negative as FLFP increases, we rely on the parents and the relastionship is not linear (for PB average parental ed­ second specification – homogamy/hypogamy/hypergamy – and interact ucation is not statistically significant). Maternal age is negatively asso­ this couple-level educational composition with municipality-level FLFP ciated with birth outcomes. Compared to first births, being in higher (H4). parity groups is associated with a lower probability of having LBW and We present results from models controlling for average parental PB. In contrast, a wider age difference between parents is associated education, average parental education squared, mother’s age at birth, with a higher probability of experiencing adverse birth outcomes. Lastly, marital status, age difference between father and mother, parity, child boys face a lower probability of being low birth weight but a higher sex, year of birth of the child, area (urban/rural), and region of residence probability of being born before 37 weeks of gestation, in line with the of the mother. The following equation presents the generic form of the literature (Eriksson, Kajantie, Osmond, Thornburg, & Barker, 2010). models we estimate: Appendix Table A2 reports the same specification separating mothers by pit birth cohort and shows that the homogamy-benefit hypothesis holds more logit(pit ) = ln( ) = α + βeduc simi + γxi + yt (1) 1 − pit where pit = Pr(Dit = 1|Xi ) and Dit is the dichotomous birth outcome 4 We nonetheless acknowledge that the stepsibling analyses cannot fully (low-weight birth, preterm birth) for the child born to couple i in year t, address selection into relationships, as maternal preferences for particular educ simi is the couple´s educational similarity at the time of birth (two-, characteristics also shape the choice of a second partner. 7 A. Abufhele et al. Research in Social Stratification and Mobility 82 (2022) 100736 Table 3 to a hypogamous and strongly hypogamous couple are 3.1 % and 4.9 % Logit models on the association between parental educational homogamy and higher, respectively, than those of an infant born to a homogamous birth-related outcomes (odds ratios reported). couple. For PB, the negative hypogamy gradient is such that the odds of (1) (2) being PB for an infant born to a hypogamous and strongly hypogamous Low birth weight Preterm birth couple are 2.7 % and 6.6 % higher, respectively than those of an infant Homogamy (ref.: Heterogamy) 0.980 *** 0.978 *** born to a homogamous couple. For full-model estimates reporting all (0.005) (0.005) controls, see Appendix Table A4. Average parental education 0.958 *** 1.005 In light of the well-established positive relationship between (0.004) (0.004) parental education and child health (Chou et al., 2010; Currie & Moretti, Average parental education squared 1.000 * 0.999 *** (0.000) (0.000) 2003), the finding that when mothers outrank their partners in terms of Married 0.998 1.033 *** educational attainment their babies are more likely to face adverse (0.006) (0.005) health conditions at birth is worth some reflections. One mechanism that Age mother 30–34 (ref.: 25–29) 1.165 *** 1.155 *** could be driving this negative association – note that with administra­ (0.007) (0.007) tive records we have no way to explicitly test this mechanism – is the Age mother 35–40 (ref.: 25–29) 1.523 *** 1.481 *** (0.011) (0.010) high stress that breadwinner mothers in this group might face during the Parity 1 (ref.: 0) 0.622 *** 0.760 *** prenatal period. Higher stress levels could eventually translate into (0.004) (0.005) worse outcomes for newborns. There is reason to suspect that in Chile Parity 2 (ref.: 0) 0.634 *** 0.852 *** women in hypogamous relationships might face challenges that are (0.005) (0.006) Parity 3 or more (ref.: 0) 0.652 *** 0.938 *** unique to their position within the couple. These challenges exacerbate (0.006) (0.008) in a context where hypergamy and homogamy have been more preva­ Father - mother age difference 1.002 *** 1.003 *** lent couple configurations. At the couple-level, a hypogamous setting (0.000) (0.000) may imply a double burden for the woman. Being more educated makes Sex child: Boy (ref.: Girl) 0.942 *** 1.181 *** women more likely to be the leading financial providers of the house­ (0.005) (0.006) Urban (ref.: Rural) 1.041 *** 1.093 *** hold, while gender roles and expectations are still such that a high share (0.010) (0.010) of housework and childcare remains under women’s responsibility. Constant 0.078 *** 0.037 *** (0.003) (0.001) Obs. 3481,584 3481,584 4.3. Hypogamy/hypergamy and female labor force participation Notes: Odds ratios reported. Robust standard errors in parentheses. To compensate for the lack of adequate data enabling to test mech­ *** p < 0.01, ** p < 0.05, * p < 0.1. anisms at the individual level using administrative data, we provide some empirical tests of the idea that women having children in hypog­ strongly for more recent birth cohorts. amous couples might be facing a double burden driven by increased Results in Table 4 explore heterogeneity in the applicability of the participation in the labor market unaccompanied by equitable gender homogamy-benefit hypothesis by average parental SES as measured by dynamics within the household. In a gender-unequal society such as average education grouped in terciles. Findings partially align with the Chile, stringent social norms around gender may be such that women do idea that educational homogamy is more positively associated with not participate in the labor force at all or actively participate in the labor desirable birth outcomes at the higher end of the educational ladder force yet maintain high housework and childcare responsibilities. If this (H2), with a clear gradient across terciles for PB. Coefficients for the tension exists, we might expect the outcomes of children born to third tercile suggest that the odds of being PB for an infant born to a hypogamous couples to be even worse in those municipalities where homogamous couple are 4.1 % lower than those of an infant born to a FLFP before childbirth is higher (H4). heterogamous couple, compared to the 2.6 % in the first tercile. Ideally, we would like to have a more precise indicator of societal gender beliefs (e.g., information on norms and values surrounding gender). We resorted to municipality-level FLFP as the former are not 4.2. Educational hypergamy and hypogamy available in this data setting.5 Existing evidence has shown that municipality-level variables such as FLFP are well correlated with so­ Recognizing that the group of heterogamous couples is heteroge­ cietal beliefs and have been used to proxy for gender norms both in Chile neous, models presented in Table 5 (Panel A) separate heterogamous (Contreras & Plaza, 2010; Ramírez & Ruben, 2015) and elsewhere couples into two groups, hypogamous and hypergamous. These models (Jayachandran, 2020). Municipalities are geographic and administra­ show that, relative to homogamy, hypogamy is negatively associated tive units that act like socio-spatial settings where the main social in­ with measures of infant health (H3a). Specifically, the odds of being teractions occur and that in Chile are recognized as important markers of LBW and PB for an infant born to a hypogamous couple are 3.7 % and socioeconomic status (Torche & Abufhele, 2021). Nonetheless, to 3.8 % higher than those of an infant born to a homogamous couple, further show that FLFP is a good-enough proxy for social norms around respectively. Relative to homogamy, hypergamy is not statistically sig­ gender, we used external data from the Encuesta Longitudinal de la nificant for LBW or PB, suggesting that homogamous and hypergamous Primera Infancia (ELPI) from the 2012 and 2017 waves to show that as couples are statistically similar when it comes to infant health. For full- FLFP (coded in quintiles) increases, views and beliefs around the role of model estimates reporting all controls, see Appendix Table A3. women and mothers in childcare and housework responsibilities become Estimates in Figure 5, Panel B show that these findings are consistent less conservative, following a rather clear gradient between the quintile under a stricter definition of hypergamy and hypogamy that separates strong hypogamy from hypogamy, and strong hypergamy from hyper­ gamy. According to these latter results, hypogamy is negatively associ­ 5 We do acknowledge that many other drivers/factors may play a relevant ated with both infant health outcomes, while hypergamy is not role in explaining the level of heterogeneity we document, such as the role of statistically significant. As expected, the larger the differences in the the local economy and governance, local differences in fertility, culture and educational levels of the parents for hypogamous couples, the stronger norms, differential access to hospitals and specialty care, pollution, travel times, the association in absolute value between couple educational composi­ etc. However, the nature of our data (administrative records) does not allow us tion and birth-related outcomes (H3b). For instance, the negative to test for these channels, which we hope to explore in future research on the hypogamy gradient is such that the odds of being LBW for an infant born topic. 8 A. Abufhele et al. Research in Social Stratification and Mobility 82 (2022) 100736 Table 4 Logit models on the association between parental educational similarity and birth-related outcomes, by average parental education terciles. Average parental education 1st Tercile 2nd Tercile 3rd Tercile Low birth weight Preterm birth Low birth weight Preterm birth Low birth weight Preterm birth Homogamy (ref.: Heterogamy) 0.970 *** 0.974 *** 0.994 1.002 0.980 * 0.959 *** (0.009) (0.008) (0.011) (0.010) (0.011) (0.009) Average parental education 0.950 *** 0.986 0.884 1.043 0.966 1.031 (0.009) (0.010) (0.073) (0.078) (0.063) (0.060) Average parental education squared 1.001 1.000 1.003 0.998 1.000 0.999 (0.001) (0.001) (0.003) (0.003) (0.002) (0.002) Married 1.023 *** 1.052 *** 0.994 1.041 *** 0.967 *** 1.007 (0.009) (0.008) (0.011) (0.010) (0.011) (0.010) Age mother 30–34 (ref.: 25–29) 1.170 *** 1.160 *** 1.178 *** 1.175 *** 1.141 *** 1.124 * ** (0.012) (0.011) (0.014) (0.012) (0.013) (0.012) Age mother 35–40 (ref.: 25–29) 1.523 *** 1.494 *** 1.546 *** 1.511 *** 1.480 *** 1.405 * ** (0.017) (0.015) (0.021) (0.018) (0.021) (0.017) Parity 1 (ref.: 0) 0.593 *** 0.686 *** 0.619 *** 0.737 *** 0.644 *** 0.820 * ** (0.007) (0.008) (0.008) (0.008) (0.007) (0.008) Parity 2 (ref.: 0) 0.589 *** 0.743 *** 0.644 *** 0.848 *** 0.675 *** 0.955 * ** (0.008) (0.009) (0.009) (0.011) (0.010) (0.012) Parity 3 or more (ref.: 0) 0.600 *** 0.811 *** 0.701 *** 0.980 0.699 *** 1.084 * ** (0.008) (0.011) (0.013) (0.016) (0.016) (0.020) Father-mother age difference 1.004 *** 1.004 *** 1.002 ** 1.003 *** 1.000 1.000 (0.001) (0.001) (0.001) (0.001) (0.001) (0.001) Sex child: Boy (ref.: Girl) 0.953 *** 1.170 *** 0.959 *** 1.195 *** 0.909 *** 1.182 * ** (0.008) (0.009) (0.009) (0.010) (0.009) (0.010) Urban (ref.: Rural) 1.038 *** 1.091 *** 1.033 1.071 *** 1.031 1.014 (0.012) (0.012) (0.025) (0.024) (0.028) (0.024) Constant 0.088 *** 0.047 *** 0.137 *** 0.030 *** 0.067 *** 0.029 *** (0.005) (0.003) (0.067) (0.013) (0.033) (0.013) Obs. 1,339,827 1,339,827 1,069,429 1,069,429 1,072,328 1,072,328 Notes: Odds ratios reported. Robust standard errors in parentheses. *** p < 0.01, ** p < 0.05, * p < 0.1. and pre-term birth (panel b) resulting from the interaction of couple Table 5 educational composition and municipality-level FLFP provides evidence Logit models on the association between parental educational dissimilarity and in line with our expectations.6 Focusing on panel a, we observe that birth-related outcomes (odds ratios reported). children born in hypogamous unions have worse LBW than homoga­ Panel A mous couples and these outcome worsen as FLFP increases. As FLFP Low birth weight Preterm birth increases from 0.27 to 0.52, the risk of a hypogamous couple to have a Hypogamy (ref. homogamy) 1.037 *** 1.038 *** LBW infant increases from 0.045 to 0.047. Focusing on panel b, children (0.007) (0.006) born in hypogamous unions have worse PB than homogamous couples Hypergamy (ref. homogamy) 1.005 1.008 and these outcomes worsen as FLFP increases (from 0.055 to 0.059), yet (0.007) (0.006) these differences are not statistically significant as demonstrated by Constant 0.076 *** 0.036 *** (0.003) (0.001) overlapping confidence intervals. Overall, heterogeneous analyses by FLFP would seem to reflect a gender-norms story that applies to Obs. 3,481,584 3,481,584 hypogamous couples, confirming the idea of a double burden faced by Panel B women. For full-model estimates reporting all controls, see Appendix Low birth weight Preterm birth Table A6 (Table A7 also provides results for hypergamy). Hypogamy (ref. homogamy) 1.031 *** 1.027 *** (0.008) (0.007) 5. Sensitivity analyses Hypogamy strong (ref. homogamy) 1.049 *** 1.066 *** (0.011) (0.011) 5.1. First births, multi-partnership, and educational upgrading Hypergamy (ref. homogamy) 1.009 1.008 (0.007) (0.007) Hypergamy strong (ref. homogamy) 0.995 1.008 First births are unique in that, compared to subsequent births, cou­ (0.011) (0.010) ples face an important life-course milestone with the first child, having Constant 0.076 *** 0.036 *** had no previous direct experience with childbearing. This uniqueness (0.003) (0.001) makes the prenatal context of first births more likely to be influenced by Obs. 3,481,584 3,481,584 the couple- and community-level mechanisms discussed in this paper. Notes: Robust standard errors in parentheses. Odds ratios reported. Controls for This is also empirically confirmed by the substantially higher prevalence average parental education, average parental education squared, mother’s age, of LBW and PB among first births compared to higher parities (see, for married, parity, father-mother age difference, infant’s sex, urban residence, re­ instance, Table 3). gion of residence, and year of birth included but not shown. Full models reported These considerations suggest that the experience gained through a in Appendix Table A3 and Table A4. first birth might translate in couples’ ability to deliver better health *** p < 0.01, ** p < 0.05, * p < 0.1 with more FLFP and the quintile with less FLFP (see Table A5 for these 6 Note that these are municipality-level averages, hence there are lots of additional analyses). women who do not report any work outside of the household in these estimates Fig. 3 plotting predicted probabilities of low birth weight (panel a) (out of the labor force). 9 A. Abufhele et al. Research in Social Stratification and Mobility 82 (2022) 100736 Table 6 Panel A shows the homogamy versus heterogamy models (same as Table 3) for first births only, while Panel B shows the ho­ mogamy vs. hypogamy vs. hypergamy models (same as Table 4) limited to first births. Results are entirely in line with Tables 3 and 5. The co­ efficients have the same sign and magnitudes and are further strength­ ened. Educational homogamy is positively associated with health outcomes (OR<1). Specifically, the odds of being LBW and PB for an infant born to a homogamous couple are 3.0 % and 4.3 % lower than those of an infant born to a heterogamous couple. Within the group of couples with different levels of education (Panel B), we observe similar differences as those documented above: hypog­ amy is negatively associated with LBW and PB (OR>1). Specifically, the odds of being LBW and PB for an infant born to a hypogamous couple are, respectively, 5.6 % and 6.8 % higher than those of an infant born to a homogamous couple. For full-model estimates reporting all controls, see Appendix Table A8. 5.2. Selection into couples How couples partner is not random. Maternal preferences for particular characteristics in a partner could influence educational sim­ ilarity and infant health simultaneously. If this is the case – and since these preferences are ultimately unobserved factors – our results can be biased. To partly address the potentially endogenous relationship be­ tween parental educational similarity and infant health, we use a sub­ sample of step-siblings to rule out the possibility that some unobserved maternal characteristics might explain both partner selection and infant health. The step-sibling sample corresponds to a matched sample of children born to the same mother but different father. Using this subsample, we compare the prevalence of low weight and preterm births among ho­ mogamous vs. heterogamous, homogamous vs. hypogamous, and ho­ mogamous vs. hypergamous couples, only in cases where one child had a bad outcome (dummy=0) and the other a good outcome (dummy=1). Results are reported in Table 7, with full-model estimates including all controls in Appendix Table A9. If our previous results were solely driven by mothers’ selection into homogamous and heterogamous couples, we would not observe significant associations between couples’ educational Fig. 3. Interaction effects between hypogamy - homogamy and municipality- similarity and birth outcomes in the step-sibling sample. Results from level FLFP, predicted health outcomes Notes: Predicted probabilities of low Table 7 suggest that selection does not seem to be the driving factor birth weight (panel a) and pre-term birth (panel b) resulting from the inter­ behind the associations reported in Tables 3, 4, 5, and 6. Homogamy is action of couples’ educational composition and municipality-level FLFP, for positively associated with infant health relative to heterogamy (Panel different levels of municipality female labor force participation. Models control for average parental education, average parental education squared, mother’s age at birth, marital status, age difference between father and mother, parity, Table 6 child sex, year of birth of the child, area (urban/rural), region of residence of Models on the association between parental educational similarity and dissim­ the mother, poverty level of the municipality, and female-to-male ratio in ilarity and birth-related outcomes, first births only (odds ratios reported). earnings in the municipality. Statistically significant differences are shown by A. Homo/hetero A. Homo/hypo/hyper (non)overlapping confidence intervals in the figures. Corresponding full esti­ Low birth Preterm Low birth Preterm mates provided in Appendix Table A6. Source: 1990–2015 Birth records, Chil­ weight birth weight birth ean Ministry of Health, merged with municipality-level information on female labor force participation from the Chile National Socioeconomic Characteriza­ Homogamy (ref. 0.970 *** 0.957 *** tion Survey (CASEN). heterogamy) (0.010) (0.009) Hypogamy (ref. 1.056 *** 1.068 *** outcomes for their subsequent children at birth (thus, focusing on first homogamy) births would represent an “extreme-case scenario”). After first birth, (0.012) (0.012) Hypergamy (ref. 1.001 1.017 couples are better equipped to deal with differential preferences and homogamy) informational asymmetries when they move on to have a second or third (0.013) (0.013) child. Indeed, gaps in preferences and information may disappear as Constant 0.091 *** 0.047 *** 0.088 *** 0.046 *** couples continue to live together and decide to have more children. In (0.007) (0.004) (0.006) (0.003) short, higher-order births take place in substantially different contexts Obs. 846,861 846,861 846,861 846,861 compared to first, thus making them less comparable to first births Notes: Robust standard errors in parentheses. Odds ratios reported. Controls for (Khan & Raeside, 1998; Zhang et al., 2019). In addition, when analyzing average parental education, average parental education squared, and mother’s births of all orders combined, there is the potential influence of educa­ age, married, father-mother age difference, infant’s sex, urban residence, region tional upgrading (i.e., obtaining more education after having had the of residence, and year of birth included but not shown. Full models reported in first child) and change in partners after first birth. These influences can Appendix Table A6. be minimized by focusing the analysis on first births. *** p < 0.01, ** p < 0.05, * p < 0.1 10 A. Abufhele et al. Research in Social Stratification and Mobility 82 (2022) 100736 Table 7 (administrative records), our results suggest that educational homoga­ Fixed effects models on the association between parental educational similarity my is positively associated with desirable birth outcomes. Compared to and birth-related outcomes (odds ratios reported). parents with different levels of education, parents with the same level of A. Homo/hetero A. Homo/hypo/hyper education face a lower probability of having low-weight and preterm Low birth Preterm Low birth Preterm births (H1). This result aligns with recent evidence from the US weight birth weight birth (Rauscher, 2020), and the specific result for Chile adds to the existing literature by showing that the homogamy-benefit hypothesis also holds in Homogamy (ref. 0.884 * 0.894 * heterogamy) a country that has only recently joined the rank of high-income societies. (0.057) (0.051) Furthermore, our analysis suggests that the homogamy-benefit hypothesis Hypogamy (ref. 1.205 ** 1.130 more strongly holds at the high end of the educational distribution for homogamy) PB (H2), a finding which is consistent with the observation that (0.105) (0.087) Hypergamy (ref. 1.067 1.107 higher-SES couples in Chile are increasingly homogamous (Bucca & homogamy) Urbina, 2019; Esteve et al., 2013). (0.089) (0.083) The coefficients that support these claims are small compared to Obs. 4992 6476 4992 6476 other determinants of birth outcomes (e.g., parity, mother’s age); however, they are comparable in size to other meaningful social markers Notes: Controls for average parental education, average parental education in the Chilean society, namely, being married or in a union. Also, evi­ squared, married, parity, father–mother age difference, infant’s sex, urban dence suggests that the prevalence and relative importance of marriage residence, region of residence, and year of birth included but not shown. Full models reported in Appendix Table A7. for children’s health outcomes is declining over time (Torche & Abuf­ *** p < 0.01, ** p < 0.05, * p < 0.1 hele, 2021), while the prevalence of hypogamous couples and female labor force participation are increasing. Hence, the assessment of these relatively small associations remains an important endeavor in light of a A). Disentangling the heterogamous group suggests that hypogamy is rapidly-changing society. negatively associated with birth outcomes, while hypergamous and Two important observations follow from the homogamy-benefit homogamous couples are not statistically different (Panel B). Estimated result. First, benefits related to parental educational homogamy may coefficients from the matched sample of step-siblings are far higher in have far-reaching implications due to the well-established associations magnitude, yet we do not comment on the latter in order to remain between birth outcomes and developmental and socioeconomic condi­ conservative. It is also worth noting that variation across different fa­ tions throughout the life course (Case & Paxson, 2010; Gluckman, thers’ education requires either a separation or bereavement, complex Hanson, Cooper, & Thornburg, 2008; Pakpahan, Hoffmann, & Kröger, phenomena which may limit the generalizability of these results. Cor­ 2017). Second, to the extent that educational homogamy is more prev­ responding OLS estimates including all children are reported in Ap­ alent among highly-educated groups, homogamy-related benefits may pendix Table A10 and show comparable results. contribute to perpetuating socioeconomic inequalities both within and across generations. 6. Discussion and Conclusions Relatedly, we also showed in an incremental manner that testing the homogamy-benefit hypothesis does not tell the whole story and that the Social scientists have had a long-standing interest in how in­ benefit of homogamy is only apparent when the reference group is made teractions among family members and their characteristics have impli­ up of hypogamous couples, which feature the worst infant-health out­ cations for the health and wellbeing of the family and its members comes. In unequal societies where gender roles and expectations still (Brown, Manning, & Stykes, 2015; Case & Paxson, 2002; Rauscher, draw heavily on the male-breadwinner family model, the educational 2020). Understanding parents’ interacting characteristics is vital for composition of heterogamous couples may imply different levels of illuminating a whole range of dynamics in the demographic makeup of maternal stress or conflict between partners. Our second set of results is households, including children’s outcomes. In this study, we have in line with these predictions, and our test for selection using the placed an exclusive focus on the relationship between parental educa­ matched sample of step-siblings suggests that our findings are not only tional similarity and children’s health at birth. From a theoretical driven by selection issues. Moreover, stronger results among first births standpoint, parental educational similarity could be beneficial for chil­ (a set of births with higher risk of negative outcomes) further support dren’s outcomes to the extent that it implies complementarity in these claims. The relationship between heterogamy and birth outcomes parental inputs towards childbearing and childrearing, reduced depends on who in the couple is the more educated partner (H3). When maternal stress, enhanced relationship quality, and reduced conflict. the woman is the more-educated one (hypogamy), educational heter­ Higher resources ought to be associated with better child outcomes, as ogamy is negatively associated with birth outcomes; when the man is the well as births to better-educated individuals should be more likely to more-educated one (hypergamy), results between homogamy and hy­ exhibit positive outcomes. The reality is more complex among couples pergamy are not statistically different. with diverging socioeconomic characteristics than among couples with Why would hypogamous couples exhibit worse infant health than similar ones. Couple composition matters for birth outcomes because homogamous or hypergamous couples? We speculated on two potential within-couple and parent-to-child interactions occur in non-neutral explanations and provided indirect empirical support for the latter using contexts. By non-neutral contexts we refer to situations in which men additional data on FLFP at the municipality level. First, and in line with and women’s returns to education favor men, and where specific couple the above discussion, hypogamy is a less traditional couple configura­ configurations fit better with established social norms than other – e.g., tion as men have historically outpaced women in educational attain­ rising educational hypogamy that challenges the idea of male economic ment; it is only in relatively recent cohorts – 1972 is estimated to be the dominance. Consequently, in non-neutral contexts we may observe un­ birth year of the first cohort to close the gender gap in education in Chile expected relationships between joint parental characteristics and birth – that women started reaching similar or higher schooling attainment outcomes. than men (Ganguli, Hausmann, & Viarengo, 2014). The socio-cultural We have used rich data on births occurring in Chile between 1990 change towards the acceptance of new roles for cohorts of and 2015 – combined with municipality-level information obtained recently-educated women might take longer (Cha & Thébaud, 2009). from an ancillary survey – to assess whether, empirically, the associa­ Educational hypergamy is instead a more socially-accepted couple tions observed are in line with some of the above expectations. Although configuration, and hypergamous couples might suffer less from social we could not test for detailed mechanisms due to the nature of the data stigma. Also, hypergamous couples might have more financial resources 11 A. Abufhele et al. Research in Social Stratification and Mobility 82 (2022) 100736 relative to hypogamous couples due to higher returns to education for resources. Nonetheless, we have carefully assessed the sensitivity of the men in the labor market. These conditions may translate into less results to alternative model specifications. within-household conflict, reduced maternal stress, and higher and Despite the above limitations, the economic, social, and gender better investments in children. norms prevailing in Chile, combined with the sustained inequality and Second, a family where the mother has more education than the uneven educational expansion that has occurred over the last 30 years father likely sees the mother as the primary provider of the household. In (Celhay & Gallegos, 2015; Daude & Robano, 2015; Torche, 2010), a country like Chile, where an egalitarian division of childcare or provided a stimulating scenario to test the homogamy-benefit hypoth­ housework is far from established (Yopo Diaz, 2016; Matear, 1997), a esis and its variations in relation to children’s outcomes at birth. breadwinner mother-to-be might face a double burden, which might Moreover, exploring differences between parental-SES groups and het­ again translate into more stress, conflict, and relationship instability. By erogamous couples allowed us to understand better the patterns and showing that outcomes of children born to hypogamous couples are associations of marital sorting and children’s health outcomes. Ulti­ more negative in municipalities where women participate more in the mately, a proper understanding of assortative mating patterns allows to labor force, we provided indirect empirical support for the existence of shed light on fundamental changes underlying the demography of the such conflict (H4). Using Goldscheider, Bernhardt, and Lappegård’s population, the characteristics of family formation, and the reproduc­ terminology (2015), the Chilean society might still be far from achieving tion of intra- and inter-generational inequalities. This study was an the “second phase” of the gender revolution, and social norms sur­ attempt to move this scholarship forward in the context of Chile, an rounding the role of women might well underlie the associations we interesting yet rarely investigated country that has recently joined the document. rank of the world’s wealthiest nations yet keeps maintaining high levels Despite several robustness checks, our results should be interpreted of inequality. with caution. First, we are not claiming that the association between parental educational similarity and infant health is causal. The step- Acknowledgments sibling sample allows us to control for unobservable maternal charac­ teristics, but we could not identify an exogenous variation to address all Abufhele acknowledges financial support from the Agencia Nacional selection issues fully. Relatedly, the step-sibling analyses cannot fully de Investigacion y Desarrollo (ANID) Millennium Nucleus on Intergen­ address selection into relationships, as maternal preferences for partic­ erational Mobility: From Modelling to Policy (MOVI) [NCS2021072] ular characteristics also shape the choice of a second partner. Second, as and the Center for Social Conflict and Cohesion Studies (grant ANID/ mentioned several times, our results do not provide detailed evidence on FONDAP/15130009). Pesando acknowledges financial support from the individual-level mechanisms driving the observed associations. Future School of Arts at McGill University, the Division of Social Science at New research drawing on alternative suitable data sources should investigate York University (AD) (grant # 76-71240-ADHPG-AD405), and the Ja­ behaviors during the prenatal period that could shed light on potential cobs Foundation (grant # 2021-1417-00). Andrés F. Castro T. ac­ mechanisms. Third, the municipality-level analyses provide a first knowledges funding from the European Research Council (ERC-2020- approximation towards understanding the importance of women’s so­ STG-948557-MINEQ), and the Spanish Ministry of Science and Innova­ cial roles in the Chilean society in explaining the association between tion (FJC2020-042965-I). The authors are grateful for useful comments educational (dis)similarity and children’s outcomes, yet we acknowl­ from seminar participants at the 2019 Population Association of edge that more and better meso- and macro-level indicators of gender America (PAA). The authors are also grateful for useful comments and norms – alongside a whole other range of institutional factors – should suggestions provided by Dr. Emily Rauscher. be explored in further research on the topic. Fourth, the dataset only provides the partnership status – married or unmarried – and it does not Appendix A. Supporting information contain information on cohabitation. Relatedly, the sex of the partner is not recorded (birth-records questions are directly asked about the “fa­ Supplementary data associated with this article can be found in the ther”), hence we do not have clear information on births from same-sex online version at doi:10.1016/j.rssm.2022.100736. couples and we cannot generalize results beyond heterosexual couples. Lastly, the birth-records information is entirely provided by the mother, References hence the educational level of the father could be measured with error, or be missing altogether. A separate investigation should be devoted to Acker, J. (1992). From sex roles to gendered institutions. Contemporary Sociology, 21(5), 565–569. studying the outcomes of children born to single mothers (or “fathers Agarwal, B. (1997). ‘Bargaining’” and gender relations: Within and beyond the missing”) in the Chilean context. household. Feminist Economics, 3(1), 1–51. There is an open methodological question in the literature on how to Beck, A., & González-Sancho, C. (2009). Educational assortative mating and children’s school readiness. Center for Research on Child Wellbeing Working Paper. deal with the issue of disentangling the homogamy/heterogamy co­ Behrman, J. A. (2019). Contextual declines in educational hypergamy and intimate efficients from the additive coefficients of mother’s and father’s edu­ partner violence. Social Forces, 97(3), 1257–1282. cation. Identifying the former is made difficult by the collinearity Behrman, J. A. (2020). Mother’s relative educational status and early childhood height- for-age z scores: A decomposition of change over time. Population Research and Policy between mother’s education, father’s education, and the difference be­ Review, 39(1), 147–173. tween the two (Eeckhaut, Van De Putte, Gerris, & Vermulst, 2013), in a Behrman, J. R., & Rosenzweig, M. R. (2004). Returns to birthweight. Review of Economics spirit similar to the Age-Period-Cohort (APC) dilemma. Social mobility and Statistics, 86(2), 586–601. Beijers, R., Jansen, J., Riksen-Walraven, M., & De Weerth, C. (2010). Maternal prenatal scholars have dealt with similar issues for decades, and there is an anxiety and stress predict infant illnesses and health complaints. Pediatrics, 126, 2. ongoing yet unresolved debate as to what method should be preferred, Belsky, J., & Fearon, R. M. P. (2004). Exploring marriage-parenting typologies and their including alternatives such as Duncan’s Square Additive Model (SAM) contextual antecedents and developmental sequelae. Development and (O. D. Duncan, 1966), Sobel’s Diagonal Reference Model (DRM) Psychopathology, 16(3), 501–523. Bharadwaj, P., Løken, K. V., & Neilson, C. (2013). Early life health interventions and (Eeckhaut et al., 2013; Sobel, 1981, 1983), or more recent and elaborate academic achievement. American Economic Review, 103(5), 1862–1891. options such as the Mobility Contrast Model (MCM) suitable to analyze Bittman, M., England, P., Sayer, L., Folbre, N., & Matheson, G. (2003). When does gender heterogeneous effects of mobility (Luo, 2022). In line with related trump money? Bargaining and time in household work. American Journal of Sociology, 109(1), 186–214. research by Rauscher (2020) and Pesando (2022), in this paper we have Blagrave, P., & Santoro, M. (2017). Labor force participation in Chile: Recent trends, opted for simpler model specifications, acknowledging their possible drivers, and prospects. IMF Working Paper 17/54. limitations. We chose a specification that – on top of the assortative Blossfeld, H.-P. (2009). Educational assortative marriage in comparative perspective. Annual Review of Sociology, 35(1), 513–530. mating coefficient – includes the average of mother and father’s edu­ cation and its squared term as a way to measure pooled household 12 A. Abufhele et al. Research in Social Stratification and Mobility 82 (2022) 100736 Blossfeld, H.-P., & Timm, A. (2003). Who marries whom? Educational systems as marriage Esteve, A., García-Román, J., & Permanyer, I. (2012). The gender-gap reversal in markets in modern societies. Dordrecht: Kluwer Acad (Ed.). education and its effect on union formation: The end of hypergamy? Population and Boncompte, J. G., & Paredes, R. D. (2020). Human capital endowments and gender Development Review, 38(3), 535–546. differences in subjective well-being in Chile. Journal of Happiness Studies, 21(1), Esteve, A., McCaa, R., & López, L.Á. (2013). The educational homogamy gap between 241–269. married and cohabiting couples in Latin America. Population Research and Policy Bratsberg, B., Markussen, S., Raaum, O., Røed, K., & Røgeberg, O. (2018). Trends in Review, 32(1), 81–102. assortative mating and offspring outcomes. IZA Discussion Paper Series, (11753). Esteve, A., Schwartz, C. R., Bavel, J., Van, Permanyer, I., Klesment, M., & Garcia, J. Bravo, D., & Valderrama Torres, J. A. (2011). The impact of income adjustments in the (2016). The end of hypergamy: Global trends and implications. Population and Casen Survey on the measurement of inequality in Chile. Estudios Délelő tt Economía, Development Review, 42(4), 615–625. 38(1), 43–65. Estrada, D. (2010). CHILE: Gender Equity Progress Blocked by Hard-Core “Machismo.” Breen, R., & Salazar, L. (2011). Educational assortative mating and earnings inequality in Retrieved December 2, 2019, from 〈http://www.ipsnews.net/2010/05/chile-gender the United States. American Journal of Sociology, 117(3), 808–843. -equity-progress-blocked-by-hard-core-machismo/〉. Brines, J. (1994). Economic dependency, gender, and the division of labor at home. Fernandez, R., & Rogerson, R. (2001). Sorting and long-run inequality. The Quarterly American Journal of Sociology, 100(3), 652–688. Journal of Economics, 116(4), 1305–1341. Brown, S. L., Manning, W. D., & Stykes, J. B. (2015). Family structure and child well- Figlio, D., Guryan, J., Karbownik, K., & Roth, J. (2014). The effects of poor neonatal being: Integrating family complexity. Journal of Marriage and Family, 77(1), health on children’s cognitive development? American Economic Review, 104(12), 177–190. 4205–4230. Bucca, M., & Urbina, D.R. (2016). Changing Patterns of Educational Assortative Mating Furstenberg, F. F. (2005). Banking on families: How families generate and distribute and Income Inequality: The case of Chile 1990–2013. Paper Presented at the 2016 social capital. Journal of Marriage and Family, 64(4), 809–821. Population Association of America, 1–10. Ganguli, I., Hausmann, R., & Viarengo, M. (2014). Closing the gender gap in education: Bucca, M., & Urbina, D. R. (2019). Lasso regularization for selection of log-linear models: What is the state of gaps in labour force participation for women, wives and An application to educational assortative mating. Sociological Methods and Research. mothers? International Labor Review, 153, 2. Cantarutti, A., Franchi, M., Monzio Compagnoni, M., Merlino, L., & Corrao, G. (2017). García, B., & de Oliveira, O. (2011). Family changes and public policies in Latin America. Mother’s education and the risk of several neonatal outcomes: An evidence from an Annual Review of Sociology, 37(1), 593–611. Italian population-based study. BMC Pregnancy and Childbirth, 17(1), 1–10. Garfinkel, I., Glei, D. A., & McLanahan, S. (2002). Assortative mating among unmarried Carlson, M., Mclanahan, S., & England, P. (2004). Union formation in fragile families. parents. Journal of Population Economics, 15, 417–432. Demography, 41(2), 237–261. Gluckman, P. D., Hanson, M. A., Cooper, C., & Thornburg, K. L. (2008). Effect of in utero Case, A., & Paxson, C. (2002). Parental behavior and child health. Health Affairs, 21(2), and early-life conditions on adult health and disease. New England Journal of 164–178. Medicine, 359(1), 61. Case, A., & Paxson, C. (2010). Causes and consequences of early-life health. Demography, Goldscheider, F., Bernhardt, E., & Lappegård, T. (2015). The gender revolution: A 47, 65–85. framework for understanding changing family and demographic behavior. Celhay, P., & Gallegos, S. (2015). Persistence in the transmission of education: Evidence Population and Development Review, 41(2), 207–239. across three generations for Chile. Journal of Human Development and Capabilities, 16 Goldstein, J. R., & Harknett, K. (2006). Parenting across racial and class lines: Assortative (3), 420–451. mating patterns of new parents who are married, cohabiting, dating or no longer Center for Reproductive Rights. (2010). Dignity Denied. Violations of the Rights of HIV- romantically involved. Social Forces, 85(1), 121–143. Positive Women in Chilean Health Facilities. Santiago de Chile. Gonzalez, R., Requejo, J. H., Jyh, K. N., Merialdi, M., Bustreo, F., & Betran, A. P. (2009). Cha, Y., & Thébaud, S. (2009). Labor markets, breadwinning, and beliefs: How economic Tackling health inequities in Chile: Maternal, newborn, infant, and child mortality context shapes men’s gender ideology. Gender and Society, 23(2), 215–243. between 1990 and 2004. American Journal of Public Health, 99(7), 1220–1226. Chawanpaiboon, S., Vogel, J. P., Moller, A. B., Lumbiganon, P., Petzold, M., Hogan, D., & Gottschalk, P., & Danziger, S. (2005). Inequality of wage rates, earnings and family Gülmezoglu, A. M. (2019). Global, regional, and national estimates of levels of income in the United States, 1975-2002. Review of Income and Wealth, 51(2), preterm birth in 2014: A systematic review and modelling analysis. The Lancet Global 231–254. Health, 7(1), e37–e46. Gullickson, A., & Torche, F. (2014). Patterns of racial and educational assortative mating Chen, Y., & Li, H. (2009). Mother’s education and child health: Is there a nurturing in Brazil. Demography, 51(3), 835–856. effect? Journal of Health Economics, 28(2), 413–426. Güneş, P. M. (2015). The role of maternal education in child health: Evidence from a Chevalier, A., & O’Sullivan, V. (2007). Mother’s education and birth weight. IZA compulsory schooling law. Economics of Education Review, 47, 1–16. Discussion Paper Series, (2640). Hu, A., & Qian, Z. (2015). Educational homogamy and earnings inequality of married Chou, S. Y., Liu, J. T., Grossman, M., & Joyce, T. (2010). Parental education and child couples: Urban China 1988-2007. Research in Social Stratification and Mobility, 40, health: Evidence from a natural experiment in Taiwan. American Economic Journal of 1–15. Applied Economics, 2(1), 33–61. Jayachandran, S. (2020). Social Norms as a Barrier To Women’s Employment in Christakis, N. A., & Fowler, J. H. (2011). Connected: The surprising power of our social Developing Countries. Working Paper. networks and how they shape our lives. New York: Back Bay. Kemptner, D., & Marcus, J. (2013). Spillover effects of maternal education on child’s Cochrane, S. H., Leslie, J., & O’Hara, D. J. (1982). Parental education and child health: health and health behavior. Review of Economics of the Household, 11(1), 29–52. Intracountry evidence. Health Policy and Education, 2, 213–250. Kerr, M. E. (2000). One’s family story: A primer on bowen theory. Washington, DC: The ComunidadMujer. (2018). Género, Educación y Trabajo: Avances, contrastes y retos de Bowen Center for the Study of the Family. tres generaciones. Santiago de Chile. Khan, H. T. A., & Raeside, R. (1998). The determinants of first and subsequent births in Contreras, D., & Plaza, G. (2010). Cultural factors in women’s labor force participation in urban and rural areas of Bangladesh. Asia-Pacific Population Journal, 13(2), 39–72. Chile. Feminist Economics, 16(2), 27–46. https://doi.org/Bravo_2011.pdf. Lindeboom, M., Llena-Nozal, A., & van der Klaauw, B. (2009). Parental education and Currie, J., & Moretti, E. (2003). Mother’s education and the intergenerational child health: Evidence from a schooling reform. Journal of Health Economics, 28(1), transmission of human capital: Evidence from college openings. Quarterly Journal of 109–131. Economics, 118(4), 1495–1532. Litman, L., Robinson, J., Rosen, Z., Rosenzweig, C., Waxman, J., & Bates, L. M. (2020). Dancause, K. N., Laplante, D. P., Oremus, C., Fraser, S., Brunet, A., & King, S. (2011). The persistence of pay inequality: The gender pay gap in an anonymous online labor Disaster-related prenatal maternal stress influences birth outcomes: Project Ice market. PLoS One, 15(2), 1–19. Storm. Early Human Development, 87(12), 813–820. Lopez, P. O., & Bréart, G. (2012). Trends in gestational age and birth weight in Chile, Daude, C., & Robano, V. (2015). On intergenerational (im)mobility in Latin America. 1991-2008. A descriptive epidemiological study. BMC Pregnancy and Childbirth, Latin American Economic Review, 24(1), 1–29. 2009. De Hauw, Y., Grow, A., & Van Bavel, J. (2017). The reversed gender gap in education and Luo, L. (2022). Heterogeneous effects of intergenerational social mobility: An improved assortative mating in Europe. European Journal of Population, 33(4), 445–474. method and new evidence. American Sociological Review, 87(1), 143–173. Desai, S., & Alva, S. (1998). Maternal education and child health: Is there a strong causal Maralani, V. (2013). The demography of social mobility: Black-White differences in the relationship? Demography, 35(1), 71–81. process of educational reproduction. American Journal of Sociology, 118(6), Duncan, G. J., Wilkerson, B., & England, P. (2006). Cleaning up their act: The effects of 1509–1558. marriage and cohabitation on licit and illicit drug use. Demography, 43(4), 691–710. Mare, R. D. (1991). Five decades of educational assortative mating. American Sociological Duncan, O. D. (1966). Methodological issues in the analysis of social mobility. In Review, 56(1), 15–32. N. L. Smelser, & S. M. Lipset (Eds.), Social structure and mobility in economic Mare, R. D., & Maralani, V. (2006). The intergenerational effects of changes in women’s development (pp. 51–97). Chicago: Aldine Publishing Company. educational attainments. American Sociological Review, 71, 542–564. Rauscher, E., & Rangel, D.E. (2020). Rising Inequality of Infant Health in the U.S. SSM - Martin, A., Ryan, R. M., & Brooks-Gunn, J. (2007). The joint influence of mother and Population Health, 12, 100698. father parenting on child cognitive outcomes at age 5. Early Childhood Research Eeckhaut, M. C. W., Van De Putte, B., Gerris, J. R. M., & Vermulst, A. A. (2013). Quarterly, 22(4), 423–439. Analysing the effect of educational differences between partners: A methodological/ Matear, A. (1997). Gender and the state in Rural Chile. Bulletin of Latin American theoretical comparison. European Sociological Review, 29(1), 60–73. Research, 16(1), 97–105. Eika, L., Mogstad, M., & Zafar, B. (2019). Educational assortative mating and household McLanahan, S. (2004). Diverging destinies: How children are faring under the second income inequality. Journal of Political Economy, 127(6), 000–000. demographic transition. Demography, 41(4), 607–627. England, P., Levine, A., & Mishel, E. (2020). Progress toward gender equality in the Ministerio de Desarrollo Social, C. (2018). Caracterización Educacional de la Población. United States has slowed or stalled. Proceedings of the National Academy of Sciences, CASEN, 2017, 1–129. Online. https://doi.org/10.1073/pnas.1918891117 Minuchin, P. (1985). Families and individual development: Provocations from the field Eriksson, J. G., Kajantie, E., Osmond, C., Thornburg, K., & Barker, D. J. P. (2010). Boys of family therapy. Child Development, 56, 289–302, 2Case_2010.pdf. live dangerously in the womb. American Journal of Human Biology, 22(3), 330–335. 13 A. Abufhele et al. Research in Social Stratification and Mobility 82 (2022) 100736 National Academies of Sciences Engineering and Medicine. (2020). Birth settings in Sobel, M. E. (1983). Structural mobility, circulation mobility and the analysis of America: Outcomes, quality, access, and choice. Washington, D.C.: The National occupational mobility: A conceptual mismatch. American Sociological Review, 48(5), Academies Press. 721–727. OECD. (2021). Gender equality in Chile: Towards a better sharing of paid and unpaid work. Torche, F. (2005). Privatization reform and inequality of educational opportunity: The Paris: OECD Publishing. https://doi.org/10.1787/6cc8ea3e-en case of Chile. Sociology of Education, 78(4), 316–343. Pakpahan, E., Hoffmann, R., & Kröger, H. (2017). The long arm of childhood Torche, F. (2010). Educational assortative mating and economic inequality: A circumstances on health in old age: Evidence from SHARELIFE. Advances in Life comparative analysis of three Latin American countries. Demography, 47(2), Course Research, 31, 1–10. 481–502. Payne, J. C., Campbell, M. K., DaSilva, O., & Koval, J. (2002). Perinatal mortality in term Torche, F. (2011). The effect of maternal stress on birth outcomes: Exploiting a natural and preterm twin and singleton births. Twin Research, 5(4), 260–264. experiment. Demography, 48(4), 1473–1491. Pesando, L. M. (2021). Educational assortative mating in sub-Saharan Africa: Torche, F. (2014). Intergenerational mobility and inequality: The Latin American case. Compositional changes and implications for household wealth inequality. Annual Review of Sociology, 40(1), 619–642. Demography, 58(2), 571–602. Torche, F., & Abufhele, A. (2021). The normativity of marriage and the marriage Pesando, L. M. (2022). A four-country study on the relationship between parental premium for children’s outcomes. American Journal of Sociology, 126(4), 931–968. educational homogamy and children’s health from infancy to adolescence. Torche, F., & Echevarría, G. (2011). The effect of birthweight on childhood cognitive Population Research and Policy Review, 41, 251–284. development in a middle-income country. International Journal of Epidemiology, 40 Qian, Z., & Preston, S. H. (1993). Changes in American marriage, 1972 to 1987: (4), 1008–1018. Availability and forces of attraction by age and education. American Sociological Torche, F., & Kleinhaus, K. (2012). Prenatal stress, gestational age and secondary sex Review, 58(4), 482–495. ratio: The sex-specific effects of exposure to a natural disaster in early pregnancy. Ramírez, E., & Ruben, R. (2015). Gender Systems and Women’s Labor Force Participation Human Reproduction, 27(2), 558–567. in the Salmon Industry in Chiloé, Chile. World Development, 73, 96–104. Van Bavel, J., Schwartz, C. R., & Esteve, A. (2018). The reversal of the gender gap in Rauscher, E. (2020). Why who marries whom matters: Effects of educational assortative education and its consequences for family life. Annual Review of Sociology, 44, mating on infant health in the United States 1969–1994. Social Forces, 98(3), 341–360. 1143–1173. Weitzman, A. (2014). Women’s and men’s relative status and intimate partner violence Ruiz, M., Goldblatt, P., Morrison, J., Kukla, L., Švancara, J., Riitta-Järvelin, M., & in India. Population and Development Review, 40(1), 55–75. Pikhart, H. (2015). Mother’s education and the risk of preterm and small for Yopo Diaz, M. (2016). El tiempo de las mujeres en Chile: Repensar la agencia. Revista de. gestational age birth: A DRIVERS meta-analysis of 12 European cohorts. Journal of Estudios Sociales, 57, 100–109. Epidemiology and Community Health, 69(9), 826–833. Zeitlin, J., Szamotulska, K., Drewniak, N., Mohangoo, A. D., Chalmers, J., Sakkeus, L., & Sánchez, R., Finot, J., & Villena, M.G. (2020). Gender wage gap and firm market power: Blondel, B. (2013). Preterm birth time trends in Europe: A study of 19 countries. Evidence from Chile. IZA Discussion Paper Series, (13856). BJOG: An International Journal of Obstetrics and Gynaecology, 120(11), 1356–1365. Schwartz, C. R. (2013). Trends and variation in assortative mating: Causes and Zeitlin, J. A., Saurel-Cubizolles, M. J., & Ancel, P. Y. (2002). Marital status, cohabitation, consequences. Annual Review of Sociology, 39(1), 451–470. and the risk of preterm birth in Europe: Where births outside marriage are common Smits, J., & Park, H. (2009). Five decades of educational assortative mating in 10 east and uncommon. Paediatric and Perinatal Epidemiology, 16(2), 124–130. asian societies. Social Forces, 88(1), 227–255. Zhang, Y., Zhou, J., Ma, Y., Liu, L., Xia, Q., Fan, D., & Ai, W. (2019). Mode of delivery and Sobel, M. E. (1981). Diagonal mobility models: A substantively motivated class of designs preterm birth in subsequent births: A systematic review and meta-analysis. PLoS One, for the analysis of mobility effects. American Sociological Review, 46(6), 893–906. 14(3), 1–12. 14