Relative Resources in Couples and Their Childbearing Behavior in the United States Relative Resources in Couples and Their Childbearing Behavior in the United States* Natalie Nitsche Abstract: A growing body of research indicates significant variation in the fertility- education relationship by partner education across high income countries. However, little is known on the education-fertility-couple nexus in the US context. The present study fills this gap. It investigates linkages between married couples’ relative socio-economic resources and their first and second birth transitions in the United States, using data from the National Longitudinal Survey of Youth 1979 (NLSY79) and a competing risk approach to model birth transitions and union dissolutions competing with first and second births independently. The study presents four findings. First, homogamous tertiary educated couples have the highest first and second birth rate, net of fertility preferences, indicating the relevance of resource pooling for family formation. Second, low-resource hypogamous and hypergamous couples have lower birth rates than most other pairings, underscoring that linkages between heterogamy and family formation may vary by the absolute level of the partners’ resources. Third, family income mediates first birth rate differences between homogamous highly educated couples and most other pairings. Lower first birth rates of hypogamous large distance couples, compared with homogamous tertiary educated couples, however, appear in part rooted in higher union dissolution rates. Fourth and finally, the higher second birth rate of homogamous highly educated couples was not mediated by any of the tested socio-economic mechanisms. More research is needed to investigate the mechanisms underlying this birth rate pattern found throughout high income societies. Keywords: Fertility · Couples · Education · Employment · Income · Union dissolution · United States Comparative Population Studies Vol. 49 (2024): 403-436 (Date of release: 19.11.2024) Federal Institute for Population Research 2024 URL: www.comparativepopulationstudies.de DOI: https://doi.org/10.12765/CPoS-2024-16 URN: urn:nbn:de:bib-cpos-2024-16en7 * This article belongs to a special issue on “Changes in Educational Homogamy and Its Consequences”. http://www.comparativepopulationstudies.de https://doi.org/10.12765/CPoS-2024-16 • Natalie Nitsche404 1 Introduction The relationship between fertility and education has received much attention in population science. The fertility postponement and sub-replacement fertility that emerged across high income countries in the second half of the 20th century coincided with an increase in women’s participation in higher education and the labor market. Fertility declines were subsequently attributed to female education expansion, and, in consequence, studies focused on understanding the fertility-education link mainly from the perspective of women (Gustafsson 2001; Kravdal/Rindfuss 2008; Rindfuss et al. 1996). Men or couples were much less frequently studied (for reviews, see Balbo et al. 2013; Vasireddy et al. 2023). However, around 95 percent of births occur to co-residential couples, with little change to this proportion over time (Andersson 2023; Perelli-Harris et al. 2012). This points to the relevance of the couple as unit of analysis for studying fertility dynamics and fertility-education linkages. Indeed, a growing body of literature documents meaningful variation in the link between one’s education and fertility tempo and quantum by the educational attainment of the partner. Studies on the education-fertility couple nexus largely examine transition rates to first, second, or subsequent births using individual level data. They indicate that the combination of both partners’ education is consequential for couples’ childbearing trajectories. For instance, couples with two tertiary educated partners postpone the first birth the longest and significantly longer than couples with one tertiary educated partner only, at least in some contexts (Bagavos 2017; Nitsche et al. 2018). Moreover, these homogamous tertiary educated couples tend to subsequently have higher second and third birth transition rates in many, but not all examined contexts (Bueno/ García-Román 2021; Dribe/Stanfors 2010; Nitsche et al. 2018; Trimarchi/Van Bavel 2020). Moreover, in Finland, homogamous tertiary educated couples contribute the largest share of births to the completed cohort fertility of the 1969-75 birth cohort (Andersson et al. 2024). However, educational combinations, or pooling, seems less relevant for childbearing among lower educated couples: across Europe, there seem to be no differences in birth rates between homogamous and heterogamous couples comprised of lower and medium educated individuals (Nitsche et al. 2021). This underscores the relevance of not only examining partners’ relative education (homogamous versus heterogamous) in the context of childbearing, but also of distinguishing between the levels of their combined education (high, medium, low). Despite these emerging patterns, various open questions on the fertility-education- couple nexus remain. First, studies have almost exclusively examined European societies; much less is known on the fertility-education-couple nexus in other world regions, including North America. One study using US National Survey of Family Growth (NSFG) data documents that fertility intentions of tertiary educated women who already have a child are higher when their partner is tertiary educated as well, compared to tertiary educated women with a lower educated partner and all other educational pairings (Morales 2020). While this finding is line with educational pairing differences in Relative Resources in Couples and Their Childbearing Behavior in the United States • 405 fertility behavior that were found in other contexts, birth rates by education pairing in the US context have not been investigated. Second, the mechanisms underlying the variation in birth transitions by educational pairings remain unclear. Education is closely linked to other socio- economic resources and one’s socio-cultural origin. However, no clear evidence has yet emerged on whether partners’ absolute and relative income, gendered work or care arrangements, or other factors underpin educational pairing differences in fertility. Studies which have investigated these mechanisms indicate complex relationships with employment, income, and educational fields that vary across social contexts (Osiewalska 2017; Trimarchi/Van Bavel 2020). Third, studies on birth rate differences often employ event-history methodology, censoring couples who separate before experiencing a birth. After a union dissolution, couples therefore no longer contribute to the denominator of the estimated hazard rates, implying that higher birth rates of certain educational pairings may in part be an artifact of some couples’ lower union separation rates and ensuing longer exposures to the risk of joint reproduction. The present study examines first and second birth transition rates by couples’ educational pairings in the United States, using data from the National Longitudinal Study of Youth (NLSY79), representing the 1957-65 US birth cohort, and Cox regression models. It extends the literature by offering evidence on the education- fertility-couple nexus in the United States, a populous high-income context with an early onset of female education expansion. Moreover, it investigates whether the emerging variation in first and second births by educational pairings in the US is mediated by couples’ labor market work arrangements, absolute or relative income resources, or the female partner’s attitudes toward gendered work divisions. Unlike many prior studies, it also controls for the woman’s desired family size. Furthermore, it employs a competing risk approach and considers the link between educational pairings and union dissolution risks that occur in competition to first and second birth transitions. Finally, it measures educational composition by employing an educational pairings approach, forming all combinations of low (defined as 0-12 years of schooling), medium (defined as 13-15 years of schooling), and high (defined as 16+ years of schooling) education. This educational pairing specification using all combinations of the partners’ low, medium, and high education has been commonly employed in the literature (Nitsche et al. 2018; Osiewalska 2017; Trimarchi/Van Bavel 2020). It is akin to a full interaction of both partners’ education. This allows for assessing couples’ joint education from various angles: their absolute joint education, their relative (gendered) education, and their educational distance among the subset of heterogamous couples. Furthermore, this specification is well suited to contrast homogamous tertiary educated couples to other educational pairings, particularly those comprised of one tertiary and one less educated partner. • Natalie Nitsche406 2 Couples and fertility: theoretical considerations 2.1 Absolute education ‒ pooling of resources One pathway through which couples’ educational pairings have been theorized to impact fertility behavior is via their joint absolute level of present and projected future resources. Oppenheimer’s work (1988, 1994, 1997) stressed the relevance of partners’ pooled resources for their family trajectories and production. She emphasized the individual importance of each spouse and their interactions within the marriage, arguing that the spouses’ “collaborative” roles in making contributions to the economic wealth of the marriage are essential to understanding couples’ family production, including their childbearing behaviors (Oppenheimer 1994: 333). Both partners’ resources contribute to the well-being of the family, and each partner’s socio-economic contributions can buffer the couple against unforeseen threats such as unemployment or ill health, thereby reducing risk and uncertainty and offering a more conducive environment for childbearing. The pooling argument implies higher birth transition rates with increases in the partners’ additive resources for either birth transition. It also implies that couples who acquire the highest level of education may postpone the first birth until they have completed their education. However, they may then catch up with and perhaps even exceed the first birth rates of couples with lower pooled resources over the course of their relationship. Hence, the higher the partners’ combined resources, the more conducive to childbearing their union is theorized to be. Accordingly, tertiary educated homogamous couples would be expected to have the highest birth rates. 2.2 Relative education ‒ specialization, bargaining, and gendered dynamics In contrast, the New Home Economic approach theorized a joint utility function of the family, which is maximized when both partners specialize in different roles, with one (traditionally the male partner) being the breadwinner in the labor market and the other (traditionally the female partner) being the primary caretaker of children and the household (Becker 1981). This approach assumes mutual agreement about gendered work divisions and fertility desires among partners and disregards potential bargaining dynamics within the couple. It implies higher birth rates among hypergamous couples because women’s lower earning potential and opportunity costs when interrupting labor market work to take care of offspring are assumed to stimulate family formation in a specialized role set-up. The same applies to couples with male breadwinner work configurations, where the male partner works and the female partner does not (regardless of the couples’ educational pairing), or couples who are supportive of traditional gendered work division attitudinally. However, empirical research shows that, on average, there is considerable disagreement in spousal fertility preferences (Testa 2012; Thomson et al. 1990; Voas 2003), suggesting that fertility is negotiated between partners. Blood and Wolfe (Blood Jr./Wolfe 1960) argued that the partner with more resources holds greater Relative Resources in Couples and Their Childbearing Behavior in the United States • 407 negotiation power. Several studies emphasize that bargaining over fertility may be a crucial element of the fertility decision making process (Bauer/Jacob 2010; Brodmann et al. 2007; Neyer et al. 2013). As more and more women enter university, gender ratios in higher education are shifting and hypogamous couples are becoming more common (Esteve et al. 2016; Van Bavel 2012). Despite women’s advancement in education and the labor market, they still perform most of the domestic work, with changes to this gender imbalance being rather slow (Bianchi et al. 2012; Mandel/ Lazarus 2021; Treas/Tai 2016). Several authors argued that fertility will be low as long as the “gender revolution” is “stalled,” because women will not be willing to engage in the second shift of domestic work (Goldscheider et al. 2015; McDonald 2000). According to these arguments, hypogamous couples (i.e., the female partner being more highly educated than the male partner) and female breadwinner couples may be a preferential set-up for first and second births, because the female partner may be able to successfully negotiate for fairer gendered work divisions. This argument also implies that couples’ more egalitarian domestic work divisions will stimulate first or second birth transitions. 2.3 Doing and un-doing gender Housework is still considered the woman’s responsibility, despite rapidly increasing gender equality in the public sphere, according to the doing and undoing gender approach (Lorber 2000; West/Zimmerman 1987). In practice, women in the US today indeed still do the majority of domestic work (Bianchi et al. 2012; Sullivan et al. 2018). In consequence, women can be expected to do the majority of domestic work as part of their gender display, even in couples where the female partner outearns the male partner or has higher earning potential, as is likely in educationally hypogamous couples. Moreover, gender structure theory suggests there is no gender neutrality of partners’ economic resources, including any potential bargaining power stemming from them (Dominguez-Folgueras 2022). Instead of bargaining for equal domestic work divisions on the basis of their higher resources, women in hypogamous couples may therefore engage in gender-deviance neutralization, by picking up more domestic work, rather than less (Sullivan/Gershuny 2016; Treas/Drobnič 2010). This may lead to a poor work-life balance and relationship dissatisfaction and imply lower rates of first and especially second births. 2.4 Educational distance ‒ cultural considerations Yet another perspective points out the relevance of the partners’ educational distance. The cultural matching perspective operationalizes education as a proxy for variation in social and cultural background, values, and habitus (Bourdieu 1984; Kalmijn 1994; Schwartz 2010). Not necessarily heterogamy per se, but especially large distances in heterogamous partners’ education (e.g., low-high education pairings) may thus signal heterogamy along socio-cultural dimensions, such as religion, nationality, social value orientation, or cultural practices. These “mismatches,” in turn, may destabilize the relationship and lead to (sooner) union dissolution or refraining • Natalie Nitsche408 from childbearing within this union due to lower relationship satisfaction, lack of joint social networks, etc. This distance hypothesis is gender neutral. Both first and second birth rates are expected to be lower in large-distance education unions, regardless of who is more and less educated. The lower fertility of these couples is theorized to operate in part but not only via higher union dissolution rates. 2.5 Prior studies Prior studies have tested these hypotheses across various societies. They offer empirical support for the pooling hypothesis in many contexts. Some support was also found for the specialization model. Higher second or third birth rates of homogamous tertiary educated couples were observed in Spain, Sweden, Belgium, Austria, Finland, Germany, and in a pooled sample from across Europe, compared with couples with one tertiary and one lower educated partner, and with couples in which neither partner has tertiary education (Bueno/García-Román 2021; Dribe/ Stanfors 2010; Nitsche 2017; Nitsche et al. 2018; Nitsche et al. 2022; Trimarchi/Van Bavel 2020). Moreover, parenthood occurs at later ages when both partners have tertiary education across a European sample and in the Netherlands (Corijn et al. 1996; Nitsche et al. 2018). In Finland and Flanders, the male partner’s education did not influence first birth timing beyond the female partner’s education (Corijn et al. 1996; Jalovaara/Miettinen 2013). There are exceptions to the additive educational pooling finding: A study on Greece finds the highest cumulative number of children among homogamous low educated couples, while homogamous highly educated and hypogamous couples had the lowest number of children (Bagavos 2017). In a sample of low educated people across 22 European countries, educational pairings did not significantly predict second or third birth transitions. Here, all couples involving at least one low educated partner had significantly lower birth rates compared with the highly educated homogamous couples, which supports the educational pooling hypothesis within highly educated but not within low educated couples (Nitsche et al. 2021). This finding points to the relevance of the absolute levels of each partner’s education for resource pooling, underscoring the importance of examining all educational pairings separately, including by educational level among homogamous couples. Some support for the specialization model was found across Eastern Europe, indicated by higher second birth rates among hypergamous couples (Trimarchi/Van Bavel 2020), whereas hypogamy was linked to lower second birth rates in Austria and Bulgaria, and across Europe generally (Nitsche et al. 2018; Osiewalska 2017). Couples with large distances in education have been examined less often. Studies using data from Finnish registers found that they had similar second birth rates as other heterogamous couples do (Nitsche et al. 2022), but also have especially high divorce risks (Mäenpää/Jalovaara 2014). Educational heterogamy has also been linked to higher risks of union dissolution in various other studies and contexts, including the United States (Schwartz 2010; Theunis et al. 2018). Relative Resources in Couples and Their Childbearing Behavior in the United States • 409 Absolute joint education, especially the joint education of couples involving one or two tertiary educated partners, has thus emerged as particularly relevant for couples’ childbearing. Heterogamy was linked with higher union dissolution risks. 2.6 The US context The US is characterized by an absence of comprehensive family policies such as maternity leave, parental leave, or comprehensive public childcare (Aisenbrey et al. 2009; Mandel/Shalev 2009). Furthermore, 75 percent of mothers return to the labor market within six months of the birth (Aisenbrey et al. 2009). In this relative absence of policies designed to support working families, the meso-level of partners’ own resources, childcare arrangements, and support of each other’s careers may be even more relevant for childbearing behaviors than in countries with more extensive public resources supporting families with young children. Thus, the US is an interesting case for examining fertility transitions through the lens of partners’ joint absolute and relative resources. 2.7 Hypotheses Based on the discussion above, the following four hypotheses are derived. H1 (Pooling): First and second birth transition rates will be higher the higher the partners’ additive resources are, all else equal. Education enrolment will depress first and second birth rates. Couples with the highest level of education may postpone the first birth due to longer education enrolment but will eventually catch up and perhaps surpass the birth rate of lower educated couples. Furthermore, controlling for the couple’s absolute financial resources will mediate the link between education pairings and birth rates. Absolute financial resources may, however, not fully mediate the link between higher additive education and birth rates because education may imply additional resources beyond current finances, e.g., future income potential, non-monetary resources such as health insurance, unemployment insurance, or social networks, including the resources of both partners’ families of origin. H2 (Specialization): First and second birth rates will be highest among educationally hypergamous couples and male breadwinner couples, all else equal. Educationally hypergamous couples will have the highest first and second birth rates. Higher birth rates of hypergamous couples may be mediated by these couples’ male-breadwinner-type work arrangements. Moreover, male breadwinner couples and couples with a male full-time and a female part-time worker are expected to have the highest birth rates, due to their family specialization, regardless of educational pairing effects. • Natalie Nitsche410 H3a (Bargaining): First and second birth rates will be the highest among educationally hypogamous couples and female breadwinner couples, all else equal. According to the bargaining argument, hypogamous unions may be conducive to childbearing, net of fertility preferences and economic resources. Potentially more egalitarian divisions of housework, negotiated via the woman’s higher relative earnings, may mediate such an effect. Ideally, gendered domestic work divisions would be measured directly to test this mechanism. However, housework time measures are not available in the NLSY79. Attitudes toward the gendered division of housework are therefore used as a proxy measure because they have been shown to predict actual domestic work divisions (Nitsche/Grunow 2016). Because educational hypogamy may not necessarily translate into higher income or bargaining power for the woman, female breadwinner arrangements or her relative income may instead be positively related to first and second birth rates. This positive effect of relative income may furthermore vary with the absolute level of family income – it may be stronger among couples in the upper third of the income distribution because more room for negotiation on how to arrange the family or spend excess income may be possible once basic needs are met. Increases in the female partner’s relative income may thus give her greater bargaining leverage in higher income couples specifically. H3b (Doing gender): First and second birth rates will be lowest among educationally hypogamous female breadwinner couples, all else equal. Competing with H3a, birth rates in couples with a woman who is more highly educated than her male partner may be the lowest. As discussed, women in hypogamous couples may have poorer work-life balance, which may lead to lower relationship satisfaction and imply lower rates of first and especially second births. Female breadwinning arrangements may mediate lower birth rates of hypogamous couples or be negatively associated with first and second birth transition rates in and of themselves. H4 (Distance): First and second birth rates will be lowest among couples with one low (0-12 years) and one highly (16+ years) educated partner, all else equal. These couples will also exhibit higher union dissolution rates. As discussed, couples with large educational distances are hypothesized to be less likely to transition to a first or second birth and more likely to separate due to a lack of shared social and cultural practices, background, and resources. Relative Resources in Couples and Their Childbearing Behavior in the United States • 411 3 Method 3.1 Data The data for the analyses come from the NLSY79, a panel study of individuals born between 1957 and 1965. The first wave was collected in 1979, when respondents were aged 14-22. The NLSY79 is a household panel. It provides (limited) information on all household members living with the main respondents in all waves. Since 1979, interviews were conducted annually between 1979 and 1994, and every two years thereafter. The survey is ongoing, and the waves until 2010 are included in the current analysis. Respondents roughly reached age 50 in 2010, hence the end of their reproductive lifespan. Most of the respondents’ first and second births took place during the 1980s and 1990s, though, so the more recent waves are of little relevance for the analyses. The strengths of the NLSY79 are its detailed information on the relationship history of primary respondents (and partly of those of their partners), its detailed fertility history information, and detailed employment histories. Additionally, the birth cohort design allows investigating one birth cohort in depth with a large enough sample size. Some aspects of the employment history were collected in a (retrospective) weekly format (employment status, work hours), but many other variables (education, income) are only available annually. For the present analyses, all covariates are used in their yearly format (rather than the weekly). The NLSY provides month/year dates for all major life events, such as marriage, divorce, and the birth of children. Socio-economic information on the spouses of main respondents was collected in detail from the beginning of the survey, but while educational attainment was always collected for all household members, there is no information on income, occupation, or hours spent in the labor market for cohabitating partners before 1994. Including cohabiting couples before 1994 is therefore not feasible due to the missing values on the cohabiting partner’s income and labor supply variables – a point in time at which the first birth had already occurred for many respondents. Besides the limited information on cohabiting partners, another drawback of the NLSY79 is that there is no information on the division of household labor, or on the number of children a spouse might have had from prior relationships. Variables on respondent’s attitudes toward gendered work divisions and responsibilities are, however, available, and were measured at three points in time (1979, 1982, 2006). One of these items measures if the respondent endorses a shared division of household labor between men and women, serving as a proxy for the gendered division of household labor in the present analysis. These value items are only available for respondents; hence, I can only measure attitudes of females, as the sample is restricted to female main respondents. 3.2 Sample The full NLSY sample consists of 12,686 respondents, alongside the information on their household members. The NLSY cautions that the fertility histories of • Natalie Nitsche412 male respondents are not as accurate as those of female respondents. I therefore restricted the sample to female respondents, and further to those women who were childless at the time of their first marriage (N = 2,768). Hence, only first marriages of female primary respondents who were without a birth at the time of their first marriage are included in the analytic sample. The male spouses, however, may have been married before. It is also possible for them to have fathered a child before entering the marriage under analysis, since this information is not available in the data. Furthermore, only white women are included in the sample. Initial results show that the relationship between relative resources and first and second birth transitions may differ between white, Black, and Hispanic Americans (models not shown), but the sample sizes for Black and Hispanic women are too small for meaningful analyses. Therefore, I estimated separate models for the three ethnic groups and will include only the results for white women (regardless of their partner’s race). This results in a sample size of 2,768 white women at risk for first birth within their first marriage. Missing values on the relevant covariates, including missing values that result from using lagged covariates (due to a loss of the first time period at risk) further reduce the sample. The final sample size is 1,067 couples during their first marriages in the risk set for the analysis of first birth, with 535 first birth events, and 158 union dissolution events. For the analysis of second birth, only couples with white female respondents who had their first birth in their first marriage (regardless of whether the birth was observed or occurred before the start of the panel) enter the risk set, resulting in a sample size of 1,295 couples at risk. There are 689 second birth events and 102 union dissolutions. Further descriptive statistics of analytic samples 1 (first birth) and 2 (second birth) can be found in Tables 1a and 1b. 3.3 Dependent process and estimation strategy In the analysis, time to event is measured in months, even though the covariates change in yearly intervals. Using detailed monthly information to measure union start and end dates and to estimate birth events has two advantages. First, it allows for modeling time to event in greater detail, and second, it results in a significantly reduced number of tied events (events occurring at the same point in time), which is important given that the Cox model uses the partial likelihood estimator. The Efron method for tied events is used. Given the precise monthly timescale of events in the data, the Cox proportional hazards model, which can operate on a continuous time scale, is preferrable over a piecewise constant or discrete time approach, because it can estimate the birth transitions by pairings with much greater precision. This is relevant in this sample, because birth transitions occur in quick succession because the sample is rather homogeneous. For modeling the transition to first birth, the dependent process is time measured in months, from the month of marriage to the occurrence of a first birth. In some couples, spouses do not yet cohabitate at the time of marriage. In these cases, the date of origin changes to the interview time at which the spouse was first present in the household. Couples can exit the risk set through three events. The first is the event of interest, the first birth, the second exit is via censoring (panel attrition or Relative Resources in Couples and Their Childbearing Behavior in the United States • 413 being in the first marriage but event-free until the last observation), and the third is the dissolution of the first marriage in competition to the first birth. Since a union dissolution is a competing event for leaving the risk set, a competing risk approach is used. For easier interpretation of the coefficients of the covariates on the birth Tab. 1a: First birth/union dissolution sample descriptives (couple-years) Note: F=female, M=male Source: National Longitudinal Study of Youth 1979 (NLSY79), own estimations Frequency Percent Min Max Mean S.D. Enrolment Respondent not enrolled 36,023 90.95 Respondent enrolled 3,583 9.05 Education Both 0-12 15,061 38.03 F 0-12/M 13-15 2,907 7.34 F 0-12/ M 16+ 1,509 3.81 F 13-15/M 0-12 3,428 8.66 Both 13-15 2,263 5.71 F 13-15/M 16+ 1,691 4.27 F 16+/M 0-12 2,111 5.33 F 16+/M 13-15 2,867 7.24 Both 16+ 7,769 19.62 Desired number of children 0 children desired 4,303 10.86 1 child desired 4,115 10.39 2+ children desired 31,188 78.75 Work arrangements Male breadwinner 3,441 8.69 Female breadwinner 1,107 2.80 F full time, M part time 1,684 4.25 M full time, F part time 6,184 15.61 Both full time 25,930 65.47 Both don’t work 464 1.17 Continuous variables F logged income 0.00 12.54 8.39 2.98 M logged income 0.00 12.64 9.29 2.45 Logged family income 6.99 13.87 10.44 0.78 F share of income 0.00 100.00 34.87 21.85 F age at union 13.00 44.00 22.99 4.89 F age at union 15.00 60.00 26.26 6.71 Housework attitude 1.00 4.00 3.15 0.95 Women’s place home attitude 1.00 4.00 1.84 0.99 Working wife useful attitude 1.00 4.00 2.61 0.86 Traditional roles best attitude 1.00 4.00 2.15 0.85 • Natalie Nitsche414 hazard, I estimate separate Cox proportional hazard models for the event of first birth (treating union dissolutions as right censored) and for the competing event of union dissolution (treating birth events as right censored), as suggested by Cleves et al. (2008) and Allison (2010). The same strategy is used for the models estimating Tab. 1b: Second birth/union dissolution sample descriptives (couple-years) Note: F=female, M=male Source: National Longitudinal Study of Youth 1979 (NLSY79), own estimations Frequency Percent Min Max Mean S.D. Enrolment Respondent not enrolled 50,273 94.70 Respondent enrolled 2,814 5.30 Education Both 0-12 25,396 47.84 F 0-12/M 13-15 3,811 7.18 F 0-12/ M 16+ 1,992 3.75 F 13-15/M 0-12 5,260 9.91 Both 13-15 2,815 5.30 F 13-15/M 16+ 2,498 4.71 F 16+/M 0-12 2,304 4.34 F 16+/M 13-15 1,957 3.69 Both 16+ 7,054 13.29 Desired number of children 0 children desired 2,022 3.81 1 child desired 7,555 14.23 2+ children desired 43,510 81.96 Work arrangements Male breadwinner 9,115 17.17 Female breadwinner 1,215 2.29 F full time, M part time 1,448 2.73 M full time, F part time 11,060 20.83 Both full time 28,993 54.61 Both don’t work 526 0.99 Continuous variables F logged income 0.00 12.49 7.48 3.68 M logged income 0.00 12.64 9.43 2.27 Logged family income 4.22 13.87 10.39 0.82 F share of income 0.00 100.00 27.73 21.85 F age at union 14.00 40.00 21.41 3.56 M age at union 15.00 48.00 24.44 5.01 Housework attitude 1.00 4.00 3.17 0.84 Women’s place home attitude 1.00 4.00 1.82 0.89 Working wife useful attitude 1.00 4.00 2.55 0.84 Traditional roles best attitude 1.00 4.00 2.17 0.81 Relative Resources in Couples and Their Childbearing Behavior in the United States • 415 second births. Here, observation time starts with the birth of the first child. Since the focus of the paper is on birth transitions, and not union dissolution, the model results for union results will be presented only briefly. A stepwise modeling strategy was employed. Educational pairings are modeled first and basic controls added thereafter. Next, work arrangements, absolute income, relative income, and gender attitudes are modeled separately. Finally, a full model with all covariates is estimated. 3.4 Covariates Most covariates are time-varying measures. While birth and relationship histories are available on a monthly time scale, covariates are only measured yearly (or bi- yearly) on interview dates. All time-varying covariates, including the educational variables, are lagged by one period (i.e., one year), and therefore measured at or before conception. Education is time-varying and measured as highest year completed. I settled on a three-category specification for both the respondents’ and their spouses’ educational attainment. It follows the standard educational US groupings: (1) less than high school education or high school education (0-12 years), (2) some college education (13-15), or (3) completed college education or higher (16+ years). The models contain a full set of interactions between the respondent's (i.e., female partner) and their spouse's (i.e., male partner) education. Hence, each possible pairing of education is reflected in a separate indicator variable. Couples in which both spouses have 0-12 years of education serve as the reference group. Enrolment in education is time- varying and unfortunately only available as a variable for respondents, not spouses. Since enrolment can be expected to lead to a postponement of pregnancy and birth, it is included in all models as a dummy variable and indicates respondents’ enrolment in either college or high school. Work status is time-varying and coded as a dummy variable, indicating current employment in the labor market versus non-employment, including inactivity and unemployment. The partners’ work statuses are interacted, resulting in four possible working status indicator variables: dual earner couples (the largest group), male breadwinner couples, female breadwinner couples, and couples with two non-working spouses. Dual earner couples serve as the reference category. Hours worked is time-varying and measured differently for respondents and spouses. For respondents (i.e., women), a variable indicates the number of hours worked in the last calendar year. For spouses (i.e., men), average hours worked per week in the last calendar year is used. I multiply this value by 52 to have a roughly comparable indicator for men’s and women’s hours worked. Due to issues with multicollinearity, it was not feasible to include the work-hours variable in a continuous format. Therefore, I use four indicators, indicating whether (1) both work full time (defined as working 35 hours per week or more), (2) the woman works full time and the man part time (part time is here defined as working less than 35 hours per week), (3) the man working full time and the woman part time, or (4) both work part time. Both partners working full time serves as the reference category. • Natalie Nitsche416 Income measures are also time-varying. The models contain logged measures for female yearly income from wages and salaries, male yearly income from wages and salaries, and the yearly joint family income. There is also a measure for the woman’s share of the total family income, which is measured as the percentage which the woman contributes to the total (non-logged) yearly family income, ranging from 0-100. There were some instances in which the wages of the woman were larger than the yearly family income; here, her share was set to 100 percent. In addition, the woman’s income share was interacted with tertiles of family income, which designate the lower, middle, and upper third of the log-family income distribution. The partners’ ages at marriage are the only non-time-varying covariates included in the models as mean-centered variables. The number of children desired is included as a time-varying categorical covariate and is only available for respondents (women), not for spouses (men). However, this question was asked in only two surveys; in 1979 and 1982. The 1979 value was filled in for the 1980/81 surveys and the 1982 value for all subsequent surveys. It is coded in three categories: no children desired, one child desired, two or more children desired, which serves as reference category. Finally, the NLSY asked respondents for their values regarding gender roles. They are measured on a scale of 1-4 (strongly disagree, disagree, agree, strongly agree). Four measures are included into the analysis: (1) “a woman’s place is in the home,” (2) “a working wife feels more useful,” (3) “traditional husband/wife roles are best,” and (4) “men should share housework.” These variables were used as linear predictors. Year fixed effects (i.e., calendar year dummies) were included in all models but are not shown in the model output due to space considerations. 4 Results 4.1 Cause specific hazards Panels A-D in Figure 1 show cause-specific hazard functions for first and second birth events and for the competing risks events of union dissolutions among the couples. These hazards are estimated based on the full analytical sample, and include all cases from the sample, including those with missing values on covariates which are omitted from the models. For both risk sets, birth events are much more likely to occur than union dissolution events. Please note, though, that those union dissolutions only refer to dissolutions which occur in competition to birth events, hence after the marriage but before the first birth (Panel B) or after the first birth but before the second birth (Panel D). Union dissolutions which occur after the second birth are not reflected. The first birth hazard (Panel A) is highest (approx. .018) within 18-48 months after the first marriage and declines rapidly thereafter. The competing hazard of union dissolution is distributed more evenly across time; it hovers around .005 in the first twelve years (144 months) of marriage and declines even further thereafter. The hazard of the second birth is distributed somewhat differently; it remains at zero during the first ten months after the first birth due to the duration of a potential Relative Resources in Couples and Their Childbearing Behavior in the United States • 417 second pregnancy. After ten months, the second birth hazard increases rapidly and peaks at approx. 36 months at .025. It rapidly declines after three years, to under .005 nine to ten years after the first birth. The hazard for the competing risk of union dissolution is highest at about four years after the first birth (.005) but generally low and remains well below .005 for most of the time. In sum, union dissolutions after marriage but before a birth in competition to birth events are rather rare, specifically after a first birth has occurred. While the birth hazard varies strongly over time, the dissolution hazard remains rather flat over the years. The Cox model assumes proportional hazards among relevant groups (here: educational pairings). I tested the proportionality assumption by including the pairings as time-varying covariates in the models (“tvc” command in Stata 18). All time-varying covariates were insignificant when controlling for year fixed effects. This indicates proportionality of hazards among the sub-groups, thus, that the proportional hazards assumption is not violated. Plotted hazards by subgroup Fig. 1: Hazard functions 0 .005 .01 .015 .02 .025 0 24 48 72 96 120 144 168 192 216 months since first marriage Smoothed hazard function for first birth Panel A ‒ first births since first marriage, per month Panel B ‒ union dissolution since first marriage (before first birth event), per month 0 24 48 72 96 120 144 168 192 216 240 264 288 months since first marriage Smoothed hazard function for comp. risk union dissolution 0 .005 .01 .015 .02 .025 Source: NLSY79, first birth sample, own estimations 0 24 48 72 96 120 144 168 192 216 months since first birth Smoothed hazard function for second birth 0 .005 .01 .015 .02 .025 0 24 48 72 96 120 144 168 192 216 240 months since first birth Smoothed hazard function for comp. risk union dissolution 0 .005 .01 .015 .02 .025 Source: NLSY79, second birth sample, own estimations Panel C ‒ second births (since first birth), per month Panel D ‒ union dissolution after a first birth (before second birth event), per month • Natalie Nitsche418 indicate some variation, especially for the first birth. However, the overall trend is very similar across groups (figures not shown but available upon request), further supporting proportionality of first birth hazards after marriage and second birth hazards after the first birth across the educational pairings. 4.2 First births Table 2 shows stepwise results for the Cox proportional hazards model estimating first birth transitions. Model 1.1 shows several significant contrasts between the educational pairings and the reference group (both 0-12 years of education). Couples with two college educated spouses and with a college educated man married to a woman with some college education (13-15 years) have significantly higher first birth hazards than the reference group (both 0-12 years) (p <= 0.05). Indeed, birth hazards increase almost linearly as the couples’ pooled educational capital increases. This finding offers support for the pooling hypothesis (H1), and for the specialization hypothesis (H2), but only among hypergamous couples with high joint human capital (man: college / woman: 13-15 years). Both large distance type couples (one partner 0-12, the other 16+ years of education) stand out, as they are the only ones who have lower first birth hazards than the reference group, although these contrasts remain statistically insignificant. However, further testing reveals significant first birth rate differences between the pairing of two college educated spouses and either of the large distance couples (p <= 0.05). This offers some support for the large distance hypothesis (H4). Stepwise models add further evidence in support of these findings. Increases in the logged family income (Model 1.4) accelerate the transition to first birth and render the contrasts between the reference group and the homogamous highly educated pairing and that of a college educated man and a woman with some college education insignificant. However, contrasts between the highly educated homogamous pairing and the large distance couples remain significant, at least marginally (he col/she 0-12 p <=. 01, she col/he 0-12 p <=.10). Thus, the difference in first birth rates between homogamous college educated couples and the reference group (or the hypogamous large distance pairing) seems partly mediated by the higher pooled financial resources these couples hold. However, this does not apply to the contrast between the highly educated homogamous pairing and the hypergamous large distance couple; their birth rate remains significantly lower. This goes against the specialization hypothesis (H2) among large distance couples. Model 1.5 shows a significant association between the woman’s income share and the first birth hazard. Increases in the income share accelerate the first birth among the two upper family income tertiles, but depress it among couples at the lower end of the income spectrum. This finding offers some support for H3a with respect to relative income, which states that bargaining or more gender-egalitarian domestic work sharing may occur as the woman’s income share rises, leading to higher birth rates. However, in this case, increases in the woman’s income share are associated with increases in first birth rates only among couples in the middle and upper income tertile. In contrast, among couples in the lowest income tertile, Relative Resources in Couples and Their Childbearing Behavior in the United States • 419 Ta b. 2 : Co x H az ar d m od el s o f fi rs t b irt h Va ria bl es H az ar d ra tio (S E) 1 2 3 4 5 6 7 Ba se lin e Ba sic c on tr ol W or ks s ta tu s Ab so lu te in co m e Re la tiv e in co m e G en de r a tt itu de s Fu ll m od el Ed uc at io n En ro lle d 0. 72 8* 0. 75 9^ 0. 76 2^ 0. 75 8^ 0. 74 5^ 0. 76 6^ 0. 76 4^ (0 .11 4) (0 .11 9) (0 .12 0) (0 .12 0) (0 .11 8) (0 .12 1) (0 .12 2) Sh F 0- 12 /M 1 3- 15 1. 23 3 1. 24 0 1. 25 9 1. 20 6 1. 25 7 1. 23 9 1. 24 0 (0 .2 07 ) (0 .2 09 ) (0 .2 13 ) (0 .2 04 ) (0 .2 12 ) (0 .2 10 ) (0 .2 11 ) F 0- 12 / M 1 6+ 0. 64 7 0. 63 3 0. 64 2 0. 59 8^ 0. 61 3 0. 62 9 0. 59 8^ (0 .19 3) (0 .19 0) (0 .19 3) (0 .18 0) (0 .18 4) (0 .18 9) (0 .18 0) F 13 -1 5/ M 0 -1 2 1. 07 5 1. 01 5 1. 02 2 1. 01 5 1. 01 5 1. 02 6 1. 04 0 (0 .19 1) (0 .18 3) (0 .18 4) (0 .18 3) (0 .18 3) (0 .18 6) (0 .18 9) Re fe re nc e: b ot h 0- 12 Bo th 1 3- 15 1. 29 2 1.1 79 1.1 85 1.1 70 1.1 90 1.1 87 1. 20 2 (0 .2 30 ) (0 .2 13 ) (0 .2 15 ) (0 .2 12 ) (0 .2 16 ) (0 .2 16 ) (0 .2 19 ) F 13 -1 5/ M 1 6+ 1. 56 0* 1. 44 6^ 1. 47 3* 1. 35 9 1. 43 8^ 1. 44 9^ 1. 41 3^ (0 .2 93 ) (0 .2 80 ) (0 .2 86 ) (0 .2 65 ) (0 .2 79 ) (0 .2 81 ) (0 .2 80 ) F 16 +/ M 0 -1 2 0. 90 0 0. 83 9 0. 83 5 0. 81 3 0. 80 3 0. 84 8 0. 81 7 (0 .2 03 ) (0 .19 6) (0 .19 5) (0 .19 0) (0 .18 9) (0 .19 9) (0 .19 3) F 16 +/ M 1 3- 15 1. 32 7 1.1 66 1.1 94 1.1 13 1.1 48 1.1 75 1.1 39 (0 .2 50 ) (0 .2 26 ) (0 .2 33 ) (0 .2 17 ) (0 .2 24 ) (0 .2 30 ) (0 .2 25 ) Bo th 1 6+ 1. 43 0* * 1. 33 5* 1. 34 1* 1. 24 3 1. 31 1^ 1. 35 0* 1. 28 8^ (0 .18 1) (0 .18 4) (0 .18 6) (0 .17 4) (0 .18 3) (0 .19 0) (0 .18 6) Ba sic c on tr ol s F ag e at u ni on 1. 04 1 1. 04 0 1. 02 4 1. 03 1 1. 04 2 1. 02 1 (0 .0 29 ) (0 .0 29 ) (0 .0 29 ) (0 .0 29 ) (0 .0 29 ) (0 .0 29 ) M a ge a t u ni on 0. 94 5* ** 0. 94 5* ** 0. 94 1* ** 0. 94 3* ** 0. 94 5* ** 0. 94 0* ** (0 .0 13 ) (0 .0 13 ) (0 .0 13 ) (0 .0 13 ) (0 .0 13 ) (0 .0 13 ) 0 ch ild re n de sir ed 0. 35 0* ** 0. 35 1* ** 0. 36 0* ** 0. 35 0* ** 0. 35 4* ** 0. 35 8* ** (0 .0 87 ) (0 .0 87 ) (0 .0 89 ) (0 .0 87 ) (0 .0 88 ) (0 .0 89 ) 1 ch ild d es ire d 0. 53 0* * 0. 53 4* * 0. 52 5* ** 0. 52 5* ** 0. 53 5* * 0. 53 2* * -0 .10 3 -0 .10 4 -0 .10 2 -0 .10 2 -0 .10 4 -0 .10 4 Re fe re nc e: 2 + ch ild re n de sir ed W or k st at us M al e br ea dw in ne r 0. 96 8 1. 52 4 (0 .16 9) (0 .5 76 ) Fe m al e br ea dw in ne r 0. 70 0 0. 81 5 (0 .2 39 ) (0 .3 22 ) • Natalie Nitsche420 Va ria bl es H az ar d ra tio (S E) 1 2 3 4 5 6 7 Ba se lin e Ba sic c on tr ol W or ks s ta tu s Ab so lu te in co m e Re la tiv e in co m e G en de r a tt itu de s Fu ll m od el Re fe re nc e: b ot h fu ll tim e Bo th d on ’t w or k 0. 43 1 (0 .2 59 ) F fu ll tim e, M p ar t t im e 0. 76 4 0. 87 2 (0 .17 3) (0 .2 02 ) M fu ll tim e, F p ar t t im e 0. 99 4 1. 06 4 (0 .12 0) (0 .13 6) Ab so lu te in co m e F in co m e 1. 00 0 1. 06 8 (0 .0 19 ) (0 .0 56 ) M in co m e 1. 00 4 0. 98 5 (0 .0 23 ) (0 .0 27 ) Fa m ily in co m e 1. 32 9* * 1.1 97 (0 .11 7) (0 .13 4) Re la tiv e in co m e F sh ar e of fa m ily in co m e 0. 99 3* 0. 99 3 (0 .0 03 ) (0 .0 05 ) Re fe re nc e: F s ha re *3 3% F sh ar e* 66 % 1. 00 7* 1. 00 4 (0 .0 03 ) (0 .0 03 ) F sh ar e* 10 0% 1. 01 1* * 1. 00 6 (0 .0 04 ) (0 .0 05 ) G en de r a tti tu de s M en s ho ul d sh ar e ho us ew or k 0. 97 5 0. 95 6 -0 .0 74 (0 .0 73 ) W om en 's pl ac e is in th e ho m e 0. 99 6 1. 00 5 (0 .0 79 ) (0 .0 80 ) W or ki ng w ife fe el s m or e us ef ul 0. 99 8 1. 00 4 (0 .0 52 ) (0 .0 53 ) Tr ad iti on al h us ba nd /w ife ro le s b es t 1. 02 6 1. 02 8 (0 .0 72 ) (0 .0 72 ) N um be r o f o bs er va tio ns 39 66 5 39 66 5 39 66 5 39 66 5 39 66 5 39 66 5 39 66 5 Ta b. 2 : Co nt in ua tio n ** * p <. 00 1, ** p <. 01 , * p <. 05 , ^ p <. 1, F =f em al e, M =m al e So ur ce : N at io na l L on gi tu di na l S tu dy o f Y ou th 1 97 9 (N LS Y7 9) , o w n es tim at io ns Relative Resources in Couples and Their Childbearing Behavior in the United States • 421 increases in the woman’s relative income are associated with a lower first birth rate, offering some support for H3b. 4.3 Union dissolutions as a competing risk to first births Table 3 shows the stepwise model results for union dissolutions as a competing risk to a first birth. In Model 2.1, union dissolutions hazards are lower among couples with a college educated woman and a spouse with some college education compared to the reference group (p < 0.10). This marginal significance disappears once basic controls for the spouses’ ages at union formation are added. Nonetheless, hypogamous large distance couples have significantly higher union dissolution hazards compared with couples with a college educated woman and a man with some college education (woman: 16+, man: 13-15) (p <= 0.05). This contrast remains significant regardless of which additional variables are added in the stepwise models. Hypogamous large distance couples also have a higher union dissolution risk than homogamous college educated couples do, but the difference is not statistically significant. All other educational pairing contrasts remain insignificant. In sum, hypogamous large distance couples indeed have a higher union dissolution hazard before the first birth than other types of couples involving a highly educated woman. These differences in union dissolution hazards may be partly responsible for the differences in first birth hazard between the educational pairings involving a highly educated woman. Note that increases in the woman’s absolute and relative income accelerate union dissolution before a first birth occurs independently of the educational pairings. 4.4 Second birth Stepwise model results for second births hazards are presented in Table 4. Educational pairings significantly relate to second birth hazards. First, couples with more pooled resources – i.e., all pairings with combinations of college and some college education – have significantly higher second birth hazards (p <= 0.10 or 0.05) than the reference group does (both no college education). This finding lends support for the pooling hypothesis (H1). Second, birth hazards of college educated men are significantly higher when their spouse is highly educated as well, compared with hypergamous large distance couples (p <= .05). Women with some college education have higher birth rates with a spouse with at least the same educational level compared with a less educated spouse (p <= 0.01). This finding suggests a depressed second birth rate among large distance couples, offering further support for H4. College educated women married to a man with some college education have higher second birth rates than the reference group and than hypogamous large distance couples do. This offers some support for H3a, but only when the hypogamous couple has a high level of joint education. In contrast, the significantly lower second birth rate of hypogamous couples with lower joint educational resources (e.g., the woman having some and the man no college education) compared to homogamous and hypergamous couples involving a woman with 13-15 years of education offer some • Natalie Nitsche422 Ta b. 3 : Co x H az ar d m od el s o f fi rs t u ni on d iss ol ut io n Va ria bl es H az . r at io (S td . e rr. ) 1 2 3 4 5 6 7 Ba se lin e Ba sic c on tr ol W or ks s ta tu s Ab so lu te in co m e Re la tiv e in co m e G en de r a tt itu de s Fu ll m od el Ed uc at io n En ro lle d 0. 88 6 0. 80 8 0. 84 4 0. 85 1 0. 88 1 0. 71 3 0. 77 5 (0 .2 51 ) (0 .2 32 ) (0 .2 43 ) (0 .2 45 ) (0 .2 54 ) (0 .2 08 ) (0 .2 29 ) F 0- 12 /M 1 3- 15 1.1 13 1. 06 0 1. 02 2 1. 02 3 1. 00 0 1.1 04 1. 03 1 (0 .3 29 ) (0 .3 19 ) (0 .3 08 ) (0 .3 08 ) (0 .3 02 ) (0 .3 33 ) (0 .3 15 ) F 0- 12 / M 1 6+ 0. 91 7 0. 87 9 0. 84 2 0. 78 9 0. 80 3 0. 84 8 0. 76 9 (0 .3 95 ) (0 .3 80 ) (0 .3 64 ) (0 .3 43 ) (0 .3 50 ) (0 .3 69 ) (0 .3 39 ) F 13 -1 5/ M 0 -1 2 1. 00 6 1. 00 7 1. 01 2 1. 00 9 1. 00 1 0. 94 8 0. 95 4 (0 .3 10 ) (0 .3 19 ) (0 .3 21 ) (0 .3 19 ) (0 .3 19 ) (0 .3 05 ) (0 .3 09 ) Re fe re nc e: b ot h 0- 12 Bo th 1 3- 15 1.1 27 1. 33 6 1. 28 0 1. 29 1 1. 23 6 1. 28 5 1.1 79 (0 .3 61 ) (0 .4 34 ) (0 .4 17 ) (0 .4 19 ) (0 .4 03 ) (0 .4 20 ) (0 .3 88 ) F 13 -1 5/ M 1 6+ 0. 89 8 1. 01 6 1. 08 3 0. 94 7 0. 91 8 0. 97 5 0. 91 4 (0 .3 67 ) (0 .4 27 ) (0 .4 56 ) (0 .4 00 ) (0 .3 87 ) (0 .4 11 ) (0 .3 93 ) F 16 +/ M 0 -1 2 1. 32 6 1. 60 0 1. 63 3 1. 59 6 1. 36 9 1. 52 4 1. 36 2 (0 .4 51 ) (0 .5 62 ) (0 .5 76 ) (0 .5 63 ) (0 .4 93 ) (0 .5 38 ) (0 .4 99 ) F 16 +/ M 1 3- 15 0. 42 2^ 0. 54 1 0. 52 4 0. 47 4 0. 45 7 0. 53 7 0. 45 3 (0 .2 20 ) (0 .2 86 ) (0 .2 77 ) (0 .2 52 ) (0 .2 44 ) (0 .2 84 ) (0 .2 42 ) Bo th 1 6+ 0. 81 3 0. 98 0 0. 98 2 0. 88 5 0. 86 4 0. 94 7 0. 77 3 (0 .2 12 ) (0 .2 73 ) (0 .2 73 ) (0 .2 51 ) (0 .2 44 ) (0 .2 66 ) (0 .2 27 ) Ba sic c on tr ol s F ag e at u ni on 0. 88 2* * 0. 86 8* * 0. 85 1* * 0. 86 4* * 0. 87 2* * 0. 83 8* ** (0 .0 41 ) (0 .0 42 ) (0 .0 42 ) (0 .0 41 ) (0 .0 42 ) (0 .0 42 ) M a ge a t u ni on 1. 08 7* ** 1. 09 2* ** 1. 08 9* ** 1. 09 5* ** 1. 09 1* ** 1. 09 6* ** (0 .0 15 ) (0 .0 15 ) (0 .0 15 ) (0 .0 15 ) (0 .0 15 ) (0 .0 15 ) 0 ch ild re n de sir ed 0. 98 6 0. 97 7 0. 98 9 1. 00 6 0. 92 4 0. 96 1 (0 .2 65 ) (0 .2 63 ) (0 .2 66 ) (0 .2 72 ) (0 .2 54 ) (0 .2 64 ) 1 ch ild d es ire d 0. 69 9 0. 73 0 0. 74 3 0. 72 2 0. 63 8 0. 65 6 -0 .2 13 -0 .2 23 -0 .2 27 -0 .2 20 -0 .19 7 -0 .2 04 Re fe re nc e: 2 + ch ild re n de sir ed W or k st at us M al e br ea dw in ne r 0. 52 3^ 1. 06 2 (0 .18 0) (0 .7 60 ) Fe m al e br ea dw in ne r 0. 26 2 0. 12 8^ (0 .2 65 ) (0 .14 0) Relative Resources in Couples and Their Childbearing Behavior in the United States • 423 Ta b. 3 : Co nt in ua tio n Va ria bl es H az . r at io (S td . e rr. ) 1 2 3 4 5 6 7 Ba se lin e Ba sic c on tr ol W or ks s ta tu s Ab so lu te in co m e Re la tiv e in co m e G en de r a tt itu de s Fu ll m od el Re fe re nc e: b ot h fu ll tim e Bo th d on ’t w or k 0. 38 6 (0 .3 97 ) F fu ll tim e, M p ar t t im e 1. 41 2 1. 27 9 (0 .4 74 ) (0 .4 48 ) M fu ll tim e, F p ar t t im e 0. 73 1 0. 92 9 (0 .17 5) (0 .2 38 ) Ab so lu te in co m e F in co m e 1. 09 4* 1. 04 6 (0 .0 42 ) (0 .10 4) M in co m e 0. 95 6 0. 93 4 (0 .0 36 ) (0 .0 45 ) Fa m ily in co m e 1. 27 1 1. 40 6 (0 .2 21 ) (0 .3 25 ) Re la tiv e in co m e F sh ar e of fa m ily in co m e 1. 01 1* 1. 01 4^ (0 .0 05 ) (0 .0 08 ) Re fe re nc e: F s ha re *3 3% F sh ar e* 66 % 0. 99 8 0. 99 3 (0 .0 05 ) (0 .0 06 ) F sh ar e* 10 0% 1. 00 9 1. 00 0 (0 .0 06 ) (0 .0 08 ) G en de r a tti tu de s M en s ho ul d sh ar e ho us ew or k 0. 91 3 0. 89 0 -0 .13 1 (0 .12 8) W om en 's pl ac e is in th e ho m e 1. 06 8 1. 09 5 (0 .15 2) (0 .15 7) W or ki ng w ife fe el s m or e us ef ul 0. 93 7 0. 94 0 (0 .0 87 ) (0 .0 88 ) Tr ad iti on al h us ba nd /w ife ro le s b es t 0. 64 8* * 0. 65 8* * (0 .0 89 ) (0 .0 90 ) N um be r o f o bs er va tio ns 39 60 6 39 60 6 39 60 6 39 60 6 39 60 6 39 60 6 39 60 6 ** * p <. 00 1, ** p <. 01 , * p <. 05 , ^ p <. 1, F =f em al e, M =m al e So ur ce : N at io na l L on gi tu di na l S tu dy o f Y ou th 1 97 9 (N LS Y7 9) , o w n es tim at io ns • Natalie Nitsche424 Ta b. 4 : Co x H az ar d m od el s o f s ec on d bi rt h Va ria bl es H az . r at io (S td . e rr. ) 1 2 3 4 5 6 7 Ba se lin e Ba sic c on tr ol W or ks s ta tu s Ab so lu te in co m e Re la tiv e in co m e G en de r a tt itu de s Fu ll m od el Ed uc at io n En ro lle d 0. 84 9 0. 82 8 0. 83 5 0. 83 1 0. 80 9 0. 82 9 0. 82 4 (0 .16 0) (0 .15 6) (0 .15 8) (0 .15 7) (0 .15 3) (0 .15 7) (0 .15 7) F 0- 12 /M 1 3- 15 1.1 76 1.1 93 1.1 69 1.1 80 1. 20 0 1.1 61 1.1 48 (0 .17 7) (0 .18 0) (0 .17 7) (0 .17 9) (0 .18 2) (0 .17 6) (0 .17 5) F 0- 12 / M 1 6+ 0. 90 3 0. 94 6 0. 93 3 0. 92 4 0. 92 0 0. 96 9 0. 94 9 (0 .2 28 ) (0 .2 39 ) (0 .2 36 ) (0 .2 34 ) (0 .2 33 ) (0 .2 45 ) (0 .2 41 ) F 13 -1 5/ M 0 -1 2 0. 78 0 0. 75 0^ 0. 73 2^ 0. 74 0^ 0. 75 9^ 0. 72 7^ 0. 72 2* (0 .12 6) (0 .12 2) (0 .11 9) (0 .12 1) (0 .12 4) (0 .11 9) (0 .11 9) Re fe re nc e: b ot h 0– 12 Bo th 1 3- 15 1. 40 5* 1. 36 4* 1. 33 3^ 1. 36 0* 1. 37 0* 1. 32 8^ 1. 31 5^ (0 .2 14 ) (0 .2 11 ) (0 .2 07 ) (0 .2 11 ) (0 .2 13 ) (0 .2 06 ) (0 .2 05 ) F 13 -1 5/ M 1 6+ 1. 38 6^ 1. 28 4 1. 29 2 1. 26 7 1. 28 4 1. 21 8 1. 21 7 (0 .2 72 ) (0 .2 55 ) (0 .2 57 ) (0 .2 53 ) (0 .2 56 ) (0 .2 43 ) (0 .2 45 ) F 16 +/ M 0 -1 2 1. 32 4 1.1 49 1. 08 0 1.1 28 1.1 34 1.1 25 1. 05 0 (0 .2 81 ) (0 .2 48 ) (0 .2 34 ) (0 .2 45 ) (0 .2 48 ) (0 .2 43 ) (0 .2 31 ) F 16 +/ M 1 3- 15 1. 64 7* * 1. 55 8* 1. 53 6* 1. 53 1* 1. 52 8* 1. 55 1* 1. 49 1* (0 .3 15 ) (0 .3 03 ) (0 .3 00 ) (0 .3 00 ) (0 .3 00 ) (0 .3 04 ) (0 .2 96 ) Bo th 1 6+ 1. 64 0* ** 1. 53 5* ** 1. 47 7* * 1. 49 1* * 1. 47 8* * 1. 49 5* * 1. 41 1* (0 .2 06 ) (0 .19 8) (0 .19 2) (0 .2 01 ) (0 .19 6) (0 .19 7) (0 .19 7) Ba sic c on tr ol s F ag e at u ni on 1. 06 7* * 1. 06 8* * 1. 06 8* * 1. 06 8* * 1. 06 9* * 1. 07 1* * (0 .0 23 ) (0 .0 23 ) (0 .0 23 ) (0 .0 23 ) (0 .0 23 ) (0 .0 24 ) M a ge a t u ni on 0. 95 2* ** 0. 95 3* ** 0. 95 1* ** 0. 95 0* ** 0. 95 5* ** 0. 95 4* ** (0 .0 12 ) (0 .0 12 ) (0 .0 12 ) (0 .0 12 ) (0 .0 12 ) (0 .0 12 ) 0 ch ild re n de sir ed 0. 58 9* 0. 61 6^ 0. 59 5* 0. 59 9* 0. 63 8^ 0. 67 2 (0 .15 2) (0 .15 9) (0 .15 3) (0 .15 4) (0 .16 6) (0 .17 5) 1 ch ild d es ire d 0. 30 9* ** 0. 73 0 0. 30 8* ** 0. 30 5* ** 0. 31 2* ** 0. 31 1* ** (0 .0 57 ) (0 .2 23 ) (0 .0 57 ) (0 .0 56 ) (0 .0 57 ) (0 .0 57 ) Re fe re nc e: 2 + ki ds d es ire d W or k st at us M al e br ea dw in ne r 0. 87 6 0. 63 5^ (0 .0 97 ) (0 .15 1) Fe m al e br ea dw in ne r 0. 29 2* 0. 32 4* (0 .14 7) (0 .17 1) Relative Resources in Couples and Their Childbearing Behavior in the United States • 425 Ta b. 4 : Co nt in ua tio n Va ria bl es H az . r at io (S td . e rr. ) 1 2 3 4 5 6 7 Ba se lin e Ba sic c on tr ol W or ks s ta tu s Ab so lu te in co m e Re la tiv e in co m e G en de r a tt itu de s Fu ll m od el Re fe re nc e: b ot h fu ll tim e F fu ll tim e, M p ar t t im e 0. 55 0^ 0. 62 2 (0 .17 0) (0 .19 4) M fu ll tim e, F p ar t t im e 1. 09 6 1. 08 1 (0 .10 0) (0 .11 1) Ab so lu te in co m e F in co m e 1. 00 1 0. 96 9 (0 .0 12 ) (0 .0 33 ) M in co m e 1. 03 3 1. 00 2 (0 .0 24 ) (0 .0 25 ) Fa m ily in co m e 1. 00 7 0. 96 5 (0 .0 77 ) (0 .0 86 ) Re la tiv e in co m e F sh ar e of fa m ily in co m e 0. 99 3* 0. 99 5 (0 .0 03 ) (0 .0 04 ) Re fe re nc e: F sh ar e* 33 % F sh ar e* 66 % 1. 00 4 1. 00 3 (0 .0 03 ) (0 .0 04 ) F sh ar e* 10 0% 1. 00 9* 1. 00 8^ (0 .0 04 ) (0 .0 05 ) G en de r a tti tu de s M en s ho ul d sh ar e ho us ew or k 1. 21 8* * 1. 21 3* * (0 .0 82 ) (0 .0 82 ) W om en 's pl ac e is in th e ho m e 1. 06 2 1. 05 8 (0 .0 70 ) (0 .0 70 ) W or ki ng w ife fe el s m or e us ef ul 0. 88 2* * 0. 88 3* * (0 .0 40 ) (0 .0 41 ) Tr ad iti on al h us ba nd /w ife ro le s b es t 1.1 26 * 1.1 22 ^ (0 .0 68 ) (0 .0 69 ) N um be r o f o bs er va tio ns 53 08 7 53 08 7 53 08 7 53 08 7 53 08 7 53 08 7 53 08 7 ** * p <. 00 1, ** p <. 01 , * p <. 05 , ^ p <. 1, F =f em al e, M =m al e So ur ce : N at io na l L on gi tu di na l S tu dy o f Y ou th 1 97 9 (N LS Y7 9) , o w n es tim at io ns • Natalie Nitsche426 support for a lower birth rate of hypogamous couples (H3b) among low-resourced couples. Third, various significant relationships between the other covariates and second birth hazards emerge in the stepwise models. However, none of these mechanisms mediate the educational pairing links with second births hazards. Female breadwinner couples and couples in which the woman works full time and the man part time have significantly lower second birth hazards, further rejecting the bargaining hypothesis (H3a) and offering support for the “Doing Gender” hypothesis (H3b). Increases in family income positively predict the second birth hazard, offering more support for the pooling hypothesis (H1). Increases in the woman’s relative income have a small but significantly positive association with the second birth rate among the highest income tertile, while depressing the second birth rate in the lowest income tertile. This further lends some support for the bargaining hypothesis (H3a) among high- resource couples and the “Doing Gender” hypothesis (H3b) among low-resource couples. Finally, the wife’s egalitarian attitude toward housework sharing positively predicts second births, offering further support for H3a. Conversely, egalitarian attitudes toward women’s labor market work lowers the second birth rate, and traditional attitudes towards gender roles increase the second birth rate, giving some support to H2. 4.5 Union dissolution in competition to second births Estimates for the competing risk model for union dissolutions which occur after the first birth are shown in Table 5. Homogamous college educated couples display a marginally significantly lower dissolution rate than the reference group. Again, this contrast becomes insignificant as control variables are added. No other educational pairing contrast or other predictors of interest significantly predict union dissolutions that occur in competition to a second birth. Thus, educational pairings emerge as relevant predictors of union dissolution that occur before parenthood occurs, but no longer after a couple became parents. 5 Discussion and conclusion Using data from the NLSY79 and Cox proportional hazard models, this paper aimed at understanding the relationship between educational pairings in couples and their first and second birth hazards among married white couples in the US. Four major conclusions can be drawn. First, I find strong support for the pooling hypothesis (H1). The first birth rate increases with increasing joint education; it is highest among homogamous tertiary educated couples and couples with a man with completed college education and a woman with some college education. Couples were at risk for childbirth starting at the time of their union formation, and models controlled for both partners’ ages at union formation. These results therefore do not focus on the age at first birth and its postponement from a life course perspective, as prior studies did, and the Relative Resources in Couples and Their Childbearing Behavior in the United States • 427 Ta b. 5 : Co x H az ar d m od el s o f s ec on d un io n di ss ol ut io n Va ria bl es H az . r at io (S td . e rr. ) 1 2 3 4 5 6 7 Ba se lin e Ba sic c on tr ol W or ks s ta tu s Ab so lu te in co m e Re la tiv e in co m e G en de r a tt itu de s Fu ll m od el Ed uc at io n En ro lle d 0. 35 3 0. 34 6 0. 34 8 0. 35 5 0. 35 4 0. 35 2 0. 36 0 (0 .2 56 ) (0 .2 50 ) (0 .2 52 ) (0 .2 57 ) (0 .2 57 ) (0 .2 55 ) (0 .2 62 ) F 0- 12 /M 1 3- 15 1. 49 2 1. 68 8 1. 68 3 1. 68 9 1. 67 4 1. 72 6 1. 72 8 (0 .4 96 ) (0 .5 71 ) (0 .5 72 ) (0 .5 74 ) (0 .5 67 ) (0 .5 86 ) (0 .5 90 ) F 0- 12 / M 1 6+ 0. 87 3 0. 95 9 0. 96 0 0. 98 6 0. 97 4 0. 94 6 0. 98 6 (0 .5 27 ) (0 .5 86 ) (0 .5 87 ) (0 .6 04 ) (0 .5 95 ) (0 .5 77 ) (0 .6 05 ) F 13 -1 5/ M 0 -1 2 0. 83 3 0. 93 4 0. 93 3 0. 92 8 0. 89 6 0. 95 9 0. 93 1 (0 .3 19 ) (0 .3 61 ) (0 .3 62 ) (0 .3 61 ) (0 .3 52 ) (0 .3 73 ) (0 .3 68 ) Re fe re nc e: b ot h 0- 12 Bo th 1 3- 15 1. 63 4 1. 82 9 1. 86 0^ 1. 81 3 1. 80 3 1. 82 5 1. 84 3 (0 .5 94 ) (0 .6 76 ) (0 .6 92 ) (0 .6 73 ) (0 .6 68 ) (0 .6 78 ) (0 .6 92 ) F 13 -1 5/ M 1 6+ 1. 45 0 1. 67 5 1. 68 8 1. 67 9 1. 66 9 1. 76 3 1. 82 7 (0 .7 13 ) (0 .8 34 ) (0 .8 44 ) (0 .8 38 ) (0 .8 33 ) (0 .8 85 ) (0 .9 24 ) F 16 +/ M 0 -1 2 0. 29 0 0. 33 3 0. 32 9 0. 33 0 0. 31 2 0. 34 2 0. 32 0 (0 .2 98 ) (0 .3 44 ) (0 .3 40 ) (0 .3 42 ) (0 .3 25 ) (0 .3 54 ) (0 .3 34 ) F 16 +/ M 1 3- 15 0. 85 5 1. 04 0 1. 05 3 1. 04 5 1. 00 6 1. 04 8 1. 07 4 (0 .6 31 ) (0 .7 76 ) (0 .7 89 ) (0 .7 85 ) (0 .7 57 ) (0 .7 84 ) (0 .8 14 ) Bo th 1 6+ 0. 35 1^ 0. 44 3 0. 44 4 0. 45 5 0. 43 2 0. 45 4 0. 46 9 (0 .2 15 ) (0 .2 77 ) (0 .2 79 ) (0 .2 89 ) (0 .2 74 ) (0 .2 85 ) (0 .3 01 ) Ba sic c on tr ol s F ag e at u ni on 0. 93 1 0. 93 1 0. 93 2 0. 92 9 0. 93 2 0. 93 2 (0 .0 55 ) (0 .0 55 ) (0 .0 55 ) (0 .0 55 ) (0 .0 55 ) (0 .0 55 ) M a ge a t u ni on 0. 96 5 0. 96 5 0. 96 6 0. 96 6 0. 96 4 0. 96 7 (0 .0 29 ) (0 .0 29 ) (0 .0 29 ) (0 .0 29 ) (0 .0 29 ) (0 .0 29 ) 0 ch ild re n de sir ed 0. 93 6 0. 93 4 0. 92 9 0. 94 2 0. 86 4 0. 86 7 (0 .5 72 ) (0 .5 73 ) (0 .5 69 ) (0 .5 81 ) (0 .5 33 ) (0 .5 41 ) 1 ch ild d es ire d 1.1 84 1.1 92 1. 20 6 1. 20 2 1.1 83 1. 22 1 (0 .3 04 ) (0 .3 08 ) (0 .3 12 ) (0 .3 10 ) (0 .3 05 ) (0 .3 18 ) Re fe re nc e: 2 + ki ds d es ire d W or k st at us M al e br ea dw in ne r 0. 92 8 1. 40 3 (0 .2 45 ) (0 .8 70 ) Fe m al e br ea dw in ne r 0. 97 0 0. 86 1 (0 .7 07 ) (0 .7 25 ) • Natalie Nitsche428 Ta b. 5 : Co nt in ua tio n Va ria bl es H az . r at io (S td . e rr. ) 1 2 3 4 5 6 7 Ba se lin e Ba sic c on tr ol W or ks s ta tu s Ab so lu te in co m e Re la tiv e in co m e G en de r a tt itu de s Fu ll m od el Re fe re nc e: b ot h fu ll tim e F fu ll tim e M p ar t t im e 0. 63 1 0. 60 1 (0 .4 61 ) (0 .4 45 ) M fu ll tim e, F p ar t t im e 0. 94 3 1. 00 2 (0 .2 40 ) (0 .2 81 ) Ab so lu te in co m e F in co m e 1. 01 5 1. 04 9 (0 .0 30 ) (0 .0 94 ) M in co m e 0. 98 9 0. 98 6 (0 .0 45 ) (0 .0 52 ) Fa m ily in co m e 0. 93 2 0. 86 3 (0 .16 1) (0 .18 0) Re la tiv e in co m e F sh ar e of fa m ily in co m e 1. 00 1 0. 99 8 (0 .0 06 ) (0 .0 09 ) Re fe re nc e: F sh ar e* 33 % F sh ar e* 66 % 1. 00 3 1. 00 5 (0 .0 08 ) (0 .0 08 ) F sh ar e* 10 0% 1. 00 3 1. 00 5 (0 .0 10 ) (0 .0 11 ) G en de r a tti tu de s M en s ho ul d sh ar e ho us ew or k 0. 89 1 0. 89 3 (0 .14 5) (0 .14 6) W om en 's pl ac e is in th e ho m e 0. 80 7 0. 80 3 (0 .13 8) (0 .13 9) W or ki ng w ife fe el s m or e us ef ul 1. 09 5 1. 09 3 (0 .13 5) (0 .13 6) Tr ad iti on al h us ba nd /w ife ro le s b es t 1. 09 8 1.1 06 (0 .17 2) (0 .17 5) N um be r o f o bs er va tio ns 53 08 7 53 08 7 53 08 7 53 08 7 53 08 7 53 08 7 53 08 7 ** * p <. 00 1, ** p <. 01 , * p <. 05 , ^ p <. 1, F =f em al e, M =m al e So ur ce : N at io na l L on gi tu di na l S tu dy o f Y ou th 1 97 9 (N LS Y7 9) , o w n es tim at io ns Relative Resources in Couples and Their Childbearing Behavior in the United States • 429 results are not directly comparable to them. Rather, they offer evidence on the first birth hazard over the course of the first marriage, which in this US sample is highest when both partners have at least some college education. Second birth rates were also highest among couples with two college educated partners. Taken together, these findings corroborate results from prior studies in several other contexts: Resource pooling of two tertiary educated partners seems to provide a conducive environment for (continued) childbearing in highest-income countries, including in the US (Bagavos 2017; Bueno/García-Román 2021; Dribe/Stanfors 2010; Nitsche et al. 2018; Trimarchi/Van Bavel 2020). A strength of the present study is that, unlike most prior research, it controls for the respondent’s number of desired children. Hence, higher birth rates of couples with high pooled educational resources likely don’t hinge on different preferences regarding envisioned family size. Rather, the partners’ combined education may impact couples’ decision-making on whether they transform their fertility desires into concrete pregnancy plans, as fertility intentions in the US have been shown to be highest among homogamous tertiary educated couples (Morales 2020). This suggests that tertiary educated individuals feel most inclined to realize their fertility desires with a partner who is also college educated, or that homogamous highly educated couples perceive themselves to be in a more favorable position to form or expand their family than heterogamous couples with only one tertiary educated or jointly lower educated partner(s) do. Interestingly, respondents in the study on fertility intentions were born between 1965 and 2000, hence, after the 1957-64 NLSY79 birth cohort examined in this study (Morales 2020). Further research on cohort differences in the educational pairing- fertility link are needed to examine whether tertiary educated homogamous couples indeed consistently have, or intend to have, higher second birth rates across various birth cohorts. Second, the link between educational heterogamy and birth rates seems to depend on the partners’ absolute levels of education, more specifically, if the heterogamous couple included a tertiary education partner or not. Little support was found for the specialization hypothesis (H2). If present, it only applied to highly resourced hypergamous couples. Jointly college educated hypergamous couples (men with 16+ years, women with 13-15 years) had higher first birth rates than the reference group of homogamous high school educated couples, and than the lower resourced hypergamous large distance couples (men with 16+ years, women with 0-12 years). This offers some support for the specialization hypothesis (H2) regarding first births. Some support was also found for the bargaining hypothesis (H3a), but again only among highly resourced couples, in terms of joint education levels and household income tertile. Jointly college educated hypogamous couples (men with 16+ years, women with 13-15 years) had higher second birth rates than the reference group and than the lower resourced hypogamous large distance couples (men with 16+ years, women with 0-12 years). This offers some support for the bargaining hypothesis (H3a) regarding second births, but only among hypogamous couples with a woman who completed college. In contrast, hypogamous couples with lower joint educational resources (women with 13-15 years, men with 0-12 years) had the lowest second birth rates of all • Natalie Nitsche430 couples. This offers some support for the “Doing Gender” hypothesis (H3b), which stated that hypogamous couples would have the lowest second birth rates due to their mismatch with prevailing gendered norms, but this applied only among low resourced hypogamous couples. Taken together, these findings on heterogamous couples underscore that associations between heterogamy and birth rates seem to depend on a couple’s joint absolute level of resources. Higher resourced heterogamy appears conducive to first births, while low resourced heterogamy does not. Hence, future work should consider absolute joint resources when studying the link between heterogamy and childbearing. This finding contributes to an ongoing debate on how partners’ absolute and relative resources can be best operationalized in research on homogamy and heterogamy and its linkages with family dynamics (for an overview of this debate, see Blossfeld et. al 2024 in this special issue). The finding could also be interpreted as further evidence for the pooling hypothesis (H1) – among heterogamous couples, the higher the joint pooled educational resources, the higher the birth rate, especially with respect to second births. Third, both hypogamous and hypergamous large distance couples had the lowest first birth rates, and significantly lower second birth rates than homogamous highly educated couples, offering support for H4. Among tertiary educated people, the low education of a spouse seems to lower the chance of a joint first birth, regardless of gender. Competing risk models show that hypogamous large distance couples had higher union dissolution rates compared with other couple types involving a college educated woman. The low birth rate of hypogamous large distance couples seems thus partly driven by higher separation rates of this type of couple before having a first child. Note that once a first child is born to large distance couples, their union dissolution rate decreases. Prior research has demonstrated higher union dissolution and lower childbearing rates among heterogamous couples, as well as lower union dissolution rates after parenthood across high income countries (Hart et al. 2017; Saarela/Finnäs 2014; Schwartz 2010; Theunis et al. 2018). Mazzeo et al. (2024) in this special issue, for instance, find that marriages in which wives have more education than their husbands have higher union dissolution rates in the West German birth cohorts 1951-60 and 1961-70. Hence, their parallel finding pertains to roughly the same birth cohort as the US NLSY79 cohort examined in this study. However, they do not investigate the hypogamy-union dissolution link by couples’ number of children. It may therefore be fruitful to pay more attention to the intersection of educational pairings and parenthood status when studying union dissolutions in the future. Fourth and finally, this study also tested the mechanisms through which educational pairings and birth rates may be linked. For first births, high family income emerged as a relevant mechanism underlying the higher birth rate of homogamous highly educated couples. However, it did not fully explain why college educated individuals with a low educated spouse had a lower birth rate, especially in the case of hypergamous large distance couples. The income variables also did not mediate second birth rate differences between highly educated homogamous couples and other couples. Mechanisms underlying first and second birth rate Relative Resources in Couples and Their Childbearing Behavior in the United States • 431 differences between educational pairings thus seem to vary, and they may even vary between educational pairings concerning the same birth transition. While economic aspects and “socio-cultural” distance appear to play a role for couples’ differences by educational pairing to start a family together, second birth rate differences in educational pairings seem to be underpinned by other aspects. Further research is needed to investigate underlying mechanisms beyond the socio-economic variables tested here. This study indicates that neither absolute or relative income, nor labor market work divisions, nor gender attitudes can help us understand why second birth rates of homogamous highly educated couples are higher compared to other educational pairings. Further factors that could be worth considering include relationship satisfaction, the partners’ social networks and support, resources accessible via families of origin, housing conditions, health and health care access, or psychological factors such as stress resilience and conflict resolution skills. This study has limitations. It focuses on a very specific sample, namely first marriages involving a white woman in the US NLSY79 birth cohort. It remains unclear whether similar patterns apply to other ethnicities, other birth cohorts, cohabitations that do not result in marriage, and second or subsequent unions in the United States. Moreover, births that occurred before the start of the panel in 1979 among this cohort of women were excluded from the sample because the partners’ covariates were not recorded. This excludes about 25 percent of first and 10 percent of second births to NLYS79 women. Hence, this analysis may systematically exclude women who had their first child at very young ages, and therefore potentially underestimate first birth rates of lower educated couples, who tend to have children earlier in the life course. Furthermore, this study does not investigate lifetime fertility. This would involve measuring fertility across multiple unions and their educational pairings and how these unions’ fertility outcomes contribute to completed fertility. One recent study on Finland indicates that educational pairings are also important for a deeper understanding of completed cohort fertility. Here, the highest proportion of births was attributed to homogamous tertiary educated couples, further underscoring the relevance of understanding why this educational couple type appears to reproduce at higher rates than others (Andersson et al. 2024). Finally, hazard rates do not distinguish between tempo and quantum effects. More research is needed to differentiate timing versus quantum of the educational pairings on all birth orders. In summary, considering the educational composition of a couple adds important insights to the fertility-education link. This study indicates that, first, homogamous highly educated couples have the highest first and second birth rates, providing evidence for the relevance of resource pooling for family formation. Second, the link between educational heterogamy and birth rates hinges on the partners’ absolute education levels, while hypogamous and hypergamous couples with large educational distances have lower birth rates than most other pairings do. Third, family income mediated first birth rate differences between homogamous highly educated couples and most other pairings. Lower first birth rates of hypogamous large distance couples, however, appear in part rooted in higher union dissolution rates of this couple type. Fourth, the higher second birth rate of homogamous • Natalie Nitsche432 highly educated couples was not mediated by any of the tested socio-economic mechanisms. Finally, there are four aspects to consider when operationalizing couples’ joint education: their absolute joint (pooled) education, their relative education (homogamy vs. heterogamy), the gender composition among heterogamous couples (hypergamy vs. hypogamy), and the distance of the educational attainment among heterogamous couples. All four aspects were relevant for couples’ birth transitions in this study. While this study adds further evidence to the emerging finding of higher birth rates among homogamous tertiary educated couples in high income countries (especially concerning second births), it may be premature to interpret the findings as a causal effect of a couples’ education composition on fertility. This study offers comprehensive life history data, a clear temporal ordering of potential cause and outcome, and the results are consistent with prior research. The underlying mechanisms of the educational pairing findings regarding first births were successfully examined (indicating that educational “effects” are rooted in economic circumstances and differences in union dissolution rates) but remained in large part elusive regarding second births. Furthermore, due to data limitations, not all births and unions of this birth cohort were observable. The results therefore may be rooted in unobserved heterogeneity, for instance regarding the selection into partnerships or into tertiary education. More research is needed to understand which specific characteristics of tertiary educated homogamy protect couples from union dissolution and stimulate their fertility. While causal effects of educational attainment and its pooling in couples may indeed emerge, especially concerning second births, this study can only document associations and does not speak to causality. Acknowledgements This study was supported with funding from the German Research Foundation (DFG) under grant agreement 459056529. Additional support was provided from the Australian National University via the ANU Future Scheme Grant to Natalie Nitsche. I would like to thank Hannah Brückner for her invaluable feedback on earlier versions of this study. References Aisenbrey, Silke; Evertsson, Marie; Grunow, Daniela 2009: Is There a Career Penalty for Mothers’ Time Out? A Comparison of Germany, Sweden and the United States. In: Social Forces 88,2: 573-605. https://doi.org/10.1353/sof.0.0252 Allison, Paul D. 2010: Survival Analysis Using SAS: A Practical Guide. 2nd edition. Cary: SAS Institute. https://doi.org/10.1353/sof.0.0252 Relative Resources in Couples and Their Childbearing Behavior in the United States • 433 Andersson, Linus 2023: A Novel Macro Perspective on Family Dynamics: The Contribution of Partnership Contexts of Births to Cohort Fertility Rates. In: Population and Development Review 49,3: 617-649. https://doi.org/10.1111/padr.12579 Andersson, Linus; Nitsche, Natalie; Jalovaara, Marika 2024: Cohort Fertility and Couple Educational Pairing. In: Stockholm Research Reports in Demography 31. https://doi.org/10.17045/sthlmuni.26123728.v1 Bagavos, Christos 2017: Do different educational pairings lead to different fertility outcomes? A cohort perspective for the Greek case. In: Vienna Yearbook of Population Research 15: 215-237. https://doi.org/10.1553/populationyearbook2017s215 Balbo, Nicoletta; Billari, Francesco C.; Mills, Melinda 2013: Fertility in Advanced Societies: A Review of Research. In: European Journal of Population / Revue Européenne de Démographie 29,1: 1-38. https://doi.org/10.1007/s10680-012-9277-y Bauer, Gerrit; Jacob, Marita 2010: Fertilitätsentscheidungen im Partnerschaftskontext: Eine Analyse der Bedeutung der Bildungskonstellation von Paaren für die Familiengründung anhand des Mikrozensus 1996-2004. In: Kölner Zeitschrift für Soziologie und Sozialpsychologie 62,1: 31-60. https://doi.org/10.1007/s11577-010-0089-y Becker, Gary 1981: A Treatise on the Family. NBER Books. National Bureau of Economic Research, Inc. Bianchi, Suzanne M. et al. 2012: Housework: Who Did, Does or Will Do It, and How Much Does It Matter? In: Social Forces 91,1: 55-63. https://doi.org/10.1093/sf/sos120 Blood Jr., Robert O.; Wolfe, Donald M. 1960: Husbands and wives: The dynamics of family living. Oxford: Free Press Glencoe. Blossfeld, Pia; Scherer, Stefani; Uunk, Wilfred 2024: Editorial on the Special Issue “Changes in educaitonal homogamy and its consequences”. In: Comparative Population Studies 49. Forthcoming Bourdieu, Pierre 1984: Distinction: A Social Critique of the Judgement of Taste. Cambridge: Harvard University Press. Brodmann, Stefanie; Esping-Andersen, Gosta; Güell, Maia 2007: When Fertility is Bargained: Second Births in Denmark and Spain. In: European Sociological Review 23,5: 599-613. https://doi.org/10.1093/esr/jcm025 Bueno, Xiana; García-Román, Joan 2021: Rethinking Couples’ Fertility in Spain: Do Partners’ Relative Education, Employment, and Job Stability Matter? In: European Sociological Review 37,4: 571-587. https://doi.org/10.1093/esr/jcaa070 Cleves, Mario A. et al. 2008: An Introduction to Survival Analysis Using Stata. 2nd edition. Stata Press. Corijn, Martine; Liefbroer, Aart C.; Gierveld, Jenny de Jong 1996: It Takes Two to Tango, Doesn’t It? The Influence of Couple Characteristics on the Timing of the Birth of the First Child. In: Journal of Marriage and Family 58,1: 117-126. https://doi.org/10.2307/353381 Dominguez-Folgueras, Marta 2022: It's about gender: A critical review of the literature on the domestic division of work. In: Journal of Family Theory & Review 14,1: 79-96. https://doi.org/10.1111/jftr.12447 Dribe, Martin; Stanfors, Maria 2010: Family life in power couples.: Continued childbearing and union stability among the educational elite in Sweden, 1991-2005. In: Demographic Research 23: 847-878. https://doi.org/10.4054/DemRes.2010.23.30 Esteve, Albert et al. 2016: The End of Hypergamy: Global Trends and Implications. In: Population and Development Review 42,4: 615-625. https://doi.org/10.1111/padr.12012 https://doi.org/10.1111/padr.12579 https://doi.org/10.17045/sthlmuni.26123728.v1 https://doi.org/10.1553/populationyearbook2017s215 https://doi.org/10.1007/s10680-012-9277-y https://doi.org/10.1007/s11577-010-0089-y https://doi.org/10.1093/sf/sos120 https://doi.org/10.1093/esr/jcm025 https://doi.org/10.1093/esr/jcaa070 https://doi.org/10.2307/353381 https://doi.org/10.1111/jftr.12447 https://doi.org/10.4054/DemRes.2010.23.30 https://doi.org/10.1111/padr.12012 • Natalie Nitsche434 Goldscheider, Frances; Bernhardt, Eva; Lappegård, Trude 2015: The Gender Revolution: A Framework for Understanding Changing Family and Demographic Behavior. In: Population and Development Review 41,2: 207-239. https://doi.org/10.1111/j.1728-4457.2015.00045.x Gustafsson, Siv 2001: Optimal age at motherhood. Theoretical and empirical considerations on postponement of maternity in Europe. In: Journal of Population Economics 14,2: 225- 247. https://doi.org/10.1007/s001480000051 Hart, Rannveig Kaldager; Lyngstad, Torkild Hovde; Vinberg, Elina 2017: Children and Union Dissolution across Four Decades: Evidence from Norway. In: European Sociological Review 33,2: 317-331. https://doi.org/10.1093/esr/jcx039 Jalovaara, Marika; Miettinen, Anneli 2013: Does his paycheck also matter? The socioeconomic resources of co-residential partners and entry into parenthood in Finland. In: Demographic Research 28: 881-916. https://doi.org/10.4054/DemRes.2013.28.31 Kalmijn, Matthijs 1994: Assortative Mating by Cultural and Economic Occupational Status. In: American Journal of Sociology 100,2: 422-452. https://doi.org/10.1086/230542 Kravdal, Øystein; Rindfuss, Ronald R. 2008: Changing Relationships between Education and Fertility: A Study of Women and Men Born 1940 to 1964. In: American Sociological Review 73,5: 854-873. https://doi.org/10.1177/000312240807300508 Lorber, Judith 2000: Using gender to undo gender: A feminist degendering movement. In: Feminist Theory 1,1: 79-95. https://doi.org/10.1177/14647000022229074. Mäenpää, Elina; Jalovaara, Marika 2014: Homogamy in socio-economic background and education, and the dissolution of cohabiting unions. In: Demographic Research 30: 1769- 1792. https://doi.org/10.4054/DemRes.2014.30.65 Mandel, Hadas; Lazarus, Amit 2021: Contextual Effects on the Gendered Division of Housework: A Cross-Country and Cross-Time Analysis. In: Sex Roles 85,3: 205-220. https://doi.org/10.1007/s11199-020-01215-0 Mandel, Hadas; Shalev, Michael 2009: How Welfare States Shape the Gender Pay Gap: A Theoretical and Comparative Analysis. In: Social Forces 87,4: 1873-1911. https://doi.org/10.1353/sof.0.0187 Mazzeo, Flavia et al. 2024: Trends in Women’s Educational Advantage and Divorce in East and West Germany. In: Comparative Population Studies 49: 317-336. https://doi.org/10.12765/CPoS-2024-13 McDonald, Peter 2000: Gender Equity in Theories of Fertility Transition. In: Population and Development Review 26,3: 427-439. https://doi.org/10.1111/j.1728-4457.2000.00427.x Morales, Danielle Xiaodan 2020: Partners’ Educational Pairings and Fertility Intentions in the United States: Evidence from 2015-2017 National Survey of Family Growth. In: Social Currents 7,4: 299-309. https://doi.org/10.1177/2329496520906515 Neyer, Gerda; Lappegård, Trude; Vignoli, Daniele 2013: Gender Equality and Fertility: Which Equality Matters? In: European Journal of Population / Revue Européenne de Démographie 29,3: 245-272. https://doi.org/10.1007/s10680-013-9292-7 Nitsche, Natalie 2017: Partners’ Educational Pairings, Work Divisions, and Fertility: Evidence from Germany. In: Vienna Institute of Demography Working Papers 19: 1-30. https://doi.org/10.1553/0x003cceca Nitsche, Natalie; Grunow, Daniela 2016: Housework over the course of relationships: Gender ideology, resources, and the division of housework from a growth curve perspective. In: Advances in Life Course Research 29: 80-94. https://doi.org/10.1016/j.alcr.2016.02.001 Nitsche, Natalie et al. 2018: Partners’ Educational Pairings and Fertility Across Europe. In: Demography 55,4: 1195-1232. https://doi.org/10.1007/s13524-018-0681-8 https://doi.org/10.1111/j.1728-4457.2015.00045.x https://doi.org/10.1007/s001480000051 https://doi.org/10.1093/esr/jcx039 https://doi.org/10.4054/DemRes.2013.28.31 https://doi.org/10.1086/230542 https://doi.org/10.1177/000312240807300508 https://doi.org/10.1177/14647000022229074 https://doi.org/10.4054/DemRes.2014.30.65 https://doi.org/10.1007/s11199-020-01215-0 https://doi.org/10.1353/sof.0.0187 https://doi.org/10.12765/CPoS-2024-13 https://doi.org/10.1111/j.1728-4457.2000.00427.x https://doi.org/10.1177/2329496520906515 https://doi.org/10.1007/s10680-013-9292-7 https://doi.org/10.1553/0x003cceca https://doi.org/10.1016/j.alcr.2016.02.001 https://doi.org/10.1007/s13524-018-0681-8 Relative Resources in Couples and Their Childbearing Behavior in the United States • 435 Nitsche, Natalie et al. 2021: Educational Pairings and Fertility Across Europe: How Do the Low- Educated Fare? In: Comparative Population Studies 46: 533-554. https://doi.org/10.12765/CPoS-2021-19 Nitsche, Natalie; Trimarchi, Alessandra; Jalovaara, Marika 2022: The Power of Two: Second Birth Rate Differences between Couples with Homogamous and Heterogamous Educational Pairings'. MPIDR Working Paper WP 2020-029. Max Planck Institute for Demographic Research. Rostock, Germany. Oppenheimer, Valerie Kincade 1988: A Theory of Marriage Timing. In: American Journal of Sociology 94,3: 563-591. https://doi.org/10.1086/229030 Oppenheimer, Valerie Kincade 1994: Women’s Rising Employment and the Future of the Family in Industrial Societies. In: Population and Development Review 20,2: 293-342. https://doi.org/10.2307/2137521 Oppenheimer, Valerie Kincade 1997: Women’s Employment and the Gain to Marriage: The Specialization and Trading Model. In: Annual Review of Sociology 23: 431-453. https://doi.org/10.1146/annurev.soc.23.1.431 Osiewalska, Beata 2017: Childlessness and fertility by couples’ educational gender (in)equality in Austria, Bulgaria, and France. In: Demographic Research 37: 325-362. https://doi.org/10.4054/DemRes.2017.37.12 Perelli-Harris, Brienna et al. 2012: Changes in union status during the transition to parenthood in eleven European countries, 1970s to early 2000s. In: Population Studies 66,2: 167-182. https://doi.org/10.1080/00324728.2012.673004 Rindfuss, Ronald R.; Morgan, S. Philip; Offutt, Kate 1996: Education and the changing age pattern of American fertility: 1963-1989. In: Demography 33,3: 277-290. https://doi.org/10.2307/2061761 Saarela, Jan; Finnäs, Fjalar 2014: Transitions Within and From First Unions: Educational Effects in an Extended Winnowing Model. In: Marriage & Family Review 50,1: 35-54. https://doi.org/10.1080/01494929.2013.851053 Schwartz, Christine R. 2010: Pathways to educational homogamy in marital and cohabiting unions. In: Demography 47,3: 735-753. https://doi.org/10.1353/dem.0.0124 Sullivan, Oriel; Gershuny, Jonathan 2016: Change in Spousal Human Capital and Housework: A Longitudinal Analysis. In: European Sociological Review 32,6: 864-880. https://doi.org/10.1093/esr/jcw043 Sullivan, Oriel; Gershuny, Jonathan; Robinson, John P. 2018: Stalled or Uneven Gender Revolution? A Long-Term Processual Framework for Understanding Why Change Is Slow. In: Journal of Family Theory & Review 10,1: 263-279. https://doi.org/10.1111/jftr.12248 Testa, Maria Rita 2012: Couple disagreement about short-term fertility desires in Austria: Effects on intentions and contraceptive behaviour. In: Demographic Research 26: 63-98. https://doi.org/10.4054/DemRes.2012.26.3 Theunis, Lindsay et al. 2018: His and Her Education and Marital Dissolution: Adding a Contextual Dimension. In: European Journal of Population 34,4: 663-687. https://doi.org/10.1007/s10680-017-9448-y Thomson, Elizabeth; McDonald, Elaine; Bumpass, Larry L. 1990: Fertility desires and fertility: Hers, his, and theirs. In: Demography 27,4: 579-588. https://doi.org/10.2307/2061571 Treas, Judith; Drobnič, Sonja (Eds.) 2010: Dividing the domestic: Men, women, and household work in cross-national perspective. Redwood City: Stanford University Press. Treas, Judith; Tai, Tsuio 2016: Gender Inequality in Housework Across 20 European Nations: Lessons from Gender Stratification Theories. In: Sex Roles 74,11-12: 495-511. https://doi.org/10.1007/s11199-015-0575-9 https://doi.org/10.12765/CPoS-2021-19 https://doi.org/10.1086/229030 https://doi.org/10.2307/2137521 https://doi.org/10.1146/annurev.soc.23.1.431 https://doi.org/10.4054/DemRes.2017.37.12 https://doi.org/10.1080/00324728.2012.673004 https://doi.org/10.2307/2061761 https://doi.org/10.1080/01494929.2013.851053 https://doi.org/10.1353/dem.0.0124 https://doi.org/10.1093/esr/jcw043 https://doi.org/10.1111/jftr.12248 https://doi.org/10.4054/DemRes.2012.26.3 https://doi.org/10.1007/s10680-017-9448-y https://doi.org/10.2307/2061571 https://doi.org/10.1007/s11199-015-0575-9 • Natalie Nitsche436 Trimarchi, Alessandra; Van Bavel, Jan 2020: Partners’ Educational Characteristics and Fertility: Disentangling the Effects of Earning Potential and Unemployment Risk on Second Births. In: European Journal of Population 36,3: 439-464. https://doi.org/10.1007/s10680-019-09537-w Van Bavel, Jan 2012: The reversal of gender inequality in education, union formation and fertility in Europe. In: Vienna Yearbook of Population Research 10,1: 127-154. https://doi.org/10.1553/populationyearbook2012s127 Vasireddy, Sindhu et al. 2023: Education and fertility in Europe in the last decade: A review of the literature. In: Comparative Population Studies 48: 553-588. https://doi.org/10.12765/CPoS-2023-21 Voas, David 2003: Conflicting preferences: A reason fertility tends to be too high or too low. In: Population and Development Review 29,4: 627-646. https://doi.org/10.1111/j.1728-4457.2003.00627.x West, Candace; Zimmerman, Don H. 1987: Doing gender. In: Gender & Society 1,2: 125-151. https://doi.org/10.1177/0891243287001002002 Date of submission: 28.05.2023 Date of acceptance: 12.09.2024 Prof. Dr. Natalie Nitsche (*). Australian National University, School of Demography. Canberra, Australia. E-mail: Natalie.Nitsche@anu.edu.au URL: https://demography.cass.anu.edu.au/people/associate-professor-natalie-nitsche https://doi.org/10.1007/s10680-019-09537-w https://doi.org/10.1553/populationyearbook2012s127 https://doi.org/10.12765/CPoS-2023-21 https://doi.org/10.1111/j.1728-4457.2003.00627.x https://doi.org/10.1177/0891243287001002002 mailto:Natalie.Nitsche@anu.edu.au https://demography.cass.anu.edu.au/people/associate-professor-natalie-nitsche Published by Federal Institute for Population Research (BiB) 65180 Wiesbaden / Germany Managing Publisher Dr. Nikola Sander 2024 Editor Prof. Dr. Roland Rau Prof. Dr. Heike Trappe Managing Editor Dr. Katrin Schiefer Editorial Assistant Beatriz Feiler-Fuchs Wiebke Hamann Layout Beatriz Feiler-Fuchs E-mail: cpos@bib.bund.de Scientific Advisory Board Kieron Barclay (Stockholm) Karsten Hank (Cologne) Ridhi Kashyap (Oxford) Natalie Nitsche (Canberra) Alyson van Raalte (Rostock) Pia S. Schober (Tübingen) Rainer Wehrhahn (Kiel) Comparative Population Studies www.comparativepopulationstudies.de ISSN: 1869-8980 (Print) – 1869-8999 (Internet) Board of Reviewers Bruno Arpino (Barcelona) Laura Bernardi (Lausanne) Gabriele Doblhammer (Rostock) Anette Eva Fasang (Berlin) Michael Feldhaus (Oldenburg) Alexia Fürnkranz-Prskawetz (Vienna) Birgit Glorius (Chemnitz) Fanny Janssen (Groningen) Frank Kalter (Mannheim) Stefanie Kley (Hamburg) Bernhard Köppen (Koblenz) Anne-Kristin Kuhnt (Rostock) Hill Kulu (St Andrews) Nadja Milewski (Wiesbaden) Thorsten Schneider (Leipzig) Tomas Sobotka (Vienna) Jeroen J. A. Spijker (Barcelona) Helga de Valk (The Hague) Sergi Vidal (Barcelona) Michael Wagner (Cologne) mailto:cpos@bib.bund.de http://www.comparativepopulationstudies.de