The long arm of childhood, with its wide-ranging influence on individuals’ life chances, highlights the importance of understanding the determinants of health in early life. Research has established that parents’ education is a major determinant of childhood health, but children across the globe increasingly experience their parents’ divorce and subsequent remarriage, raising questions of whether union instability alters these intergenerational processes. Does divorce and remarriage interfere with parents’ education benefiting their young children’s health? I explore this question in sub-Saharan Africa, a world region where parents’ education plays a major role in protecting children against severe health risks, and where young children commonly experience parental divorce and remarriage. Moreover, sub-Saharan Africa features distinct family lineage systems, affording an opportunity to explore this question in both majority matrilineal and patrilineal contexts. Analyses of Demographic and Health Survey data on 271,292 children in 30 sub-Saharan African countries offer no evidence that the high levels of union instability in the region will weaken the health benefits of parents’ education for future generations. Following divorce, children benefit from their biological parents’ education to the same degree as children with married parents—a finding that is consistent across lineage contexts. Moreover, stepfathers’ education corresponds with pronounced health benefits for their coresident stepchildren, particularly in patrilineal regions where these children benefit less dramatically from their mothers’ education. Together, the study results offer a renewed sense of the importance of parents’—including stepfathers’—education for early childhood health across diverse family structures.
Early childhood health has a lasting influence on individuals’ lives, with its innumerable consequences stretching across the life course (Conley and Bennett 2000; Palloni 2006). As such, early childhood health plays a key role in stratification (Palloni 2006), highlighting the need to understand its determinants.
Extensive social science has established that parents’ resources, including education, are strong determinants of their young children’s health and development (Elo 2009). Children born to highly educated, well-resourced parents are primed to experience healthy development, whereas children born to less-educated and socially disadvantaged parents are vulnerable to poor health and developmental delays (Prickett and Augustine 2016).
Across the globe, however, high rates of nonmarital childbearing, divorce, and remarriage mean that children often spend their early years residing apart from one parent, most commonly their biological father, and subsequently with a stepparent, typically a stepfather. Ample research has shown that parents’ divorce and remarriage can interfere with children’s health and development (Amato 2010; Fomby and Cherlin 2007), but little work has explored whether union instability can also alter the nature of other intergenerational processes of stratification (Kalmijn 2015), such as those pertaining to parents’ education. Does parents’ union instability disrupt the benefits of their education for their children’s early-life health?
Answering this question is of pressing importance in sub-Saharan Africa, where early-life health inequality is not only a precursor to inequality in adulthood but also a matter of who survives childhood and, among those who do, whose most basic physical needs (e.g., nutrition) are met. Like elsewhere, sub-Saharan African children with less-educated parents have worse health than those with highly educated parents (Cochrane et al. 1982; De Neve and Subramanian 2017; Desai and Alva 1998; Fuchs et al. 2010; Gakidou et al. 2010; Hobcraft 1993; LeVine 2012; Smith-Greenaway 2013a; Vollmer et al. 2016). In sub-Saharan Africa, children reside in diverse family structures (Assche and Adjiwanou 2009; Odimegwu et al. 2017), partly due to the high rates of divorce and remarriage in the region (Clark and Brauner-Otto 2015). Many children thus spend their earliest, most critical years with a single mother (Clark and Hamplová 2013) and subsequently a stepfather (Gaydosh 2015, 2017; Lopus 2017). These family structures generally correspond with child health disadvantage relative to children with married parents (Anderson et al. 1999; Clark and Hamplová 2013; DeRose et al. 2017; Izugbara 2016; Lopus 2017; Ntoimo and Odimegwu 2014; Thiombiano et al. 2013).
It is unclear, however, whether having a divorced and/or remarried mother can further disadvantage children by interfering with the intergenerational benefits of parents’ resources, including education. If so, this could help explain evidence that the child health benefits of parents’ education are attenuating in low-income countries (Karlsson et al. 2018). Alternatively, a parent’s education could take on elevated importance for a child’s health in the wake of divorce or remarriage, in which case union instability could amplify the intergenerational salience of education.
To explore these possibilities in sub-Saharan Africa, where approximately one-fourth of the world’s children reside,1 I analyzed Demographic and Health Survey (DHS) data from 30 countries. Specifically, I created a data set of 271,292 children under 5 years old coresiding with their mother, which features anthropometric data as well as information on parents’ union status and education. This multicountry data set allowed me to assess the relative influence of parents’ union status and education on children’s physical health as well as whether union status conditions the impact of parents’ education. Incorporating additional data on the lineage system in select contexts, the analyses also acknowledge variation in the concentration of matrilineal and patrilineal ethnic groups across the region—variation that corresponds with distinct norms surrounding marriage, union dissolution and remarriage, and parent-child relationships (Oheneba-Sakyi and Takyi 1997). I considered the lineage context for a subsample of children in 11 countries to assess its relevance to understanding how parents’ union instability conditions the child health benefits of parents’ education.
Intergenerational Advantage: Parents’ Education and Child Health
In sub-Saharan Africa, even minimal education powerfully affects women’s and men’s lives and, in turn, advantages their children in various ways (Alemayehu et al. 2015; Asfaw et al. 2015; Balogun and Yakubu 2014; Desai and Alva 1998; Fuchs et al. 2010; Gakidou et al. 2010; LeVine 2012; Smith-Greenaway 2013a; Vollmer et al. 2016). With its effects on women’s and men’s behavior, knowledge, relationships, and resources, education bestows intergenerational advantages by enhancing children’s access to two parental resources: time and material support (Thomson and McLanahan 2012; Thomson et al. 1994).
In terms of the former, research has shown that parents’ education increases the quality—and, in the case of fathers, quantity—of time and direct care that they provide their children. Even minimal education equips parents with skills that can enhance their ability to prevent, identify, and treat child health problems, thereby improving the quality of childcare they provide (LeVine 2012). Moreover, for mothers, education corresponds with greater control over their time and more autonomy in making parenting decisions, further benefitting the quality of care their children receive (Smith-Greenaway 2013b). Among fathers, education also corresponds with their greater involvement in childcare (Engle and Breaux 1998; Hewlett 1987, 1992; Madhavan et al. 2015; Mercer 2015; Saraff and Srivastava 2010; Tronick et al. 1992), which facilitates their children’s healthy development.
Parents’ education also benefits their children by increasing children’s access to material resources. Education, particularly secondary education, increases individuals’ earning potential (Psacharopoulos and Patrinos 2004; Schultz 2004) and, in turn, their likelihood of residing in better-resourced homes and communities. In low-income countries, where poor households often lack amenities like running water or a flush toilet, and poor communities can lack infrastructure like electricity, living in a more-resourced household and community is a principal way that parents’ education benefits their children’s health (Desai and Alva 1998). Additionally, in a context where the infectious disease burden is high (i.e., high incidence of malaria, HIV infection, and tuberculosis), having more highly educated parents who are themselves healthier can contribute to a household’s overall economic security and, in turn, children’s well-being. Research has also shown that fathers’ education corresponds with their spending their income in ways that benefit their children (Gummerson and Schneider 2013; Madhavan et al. 2014), further promoting a strong, positive association between their education and their children’s health and development (Madhavan and Townsend 2007).
The extent to which a father’s, and a mother’s, education will correspond with their children’s health, however, may depend on the family structure, including the parents’ union status and biological relation to the child. Before discussing these possibilities in detail, I review research on the commonality and consequences of children’s exposure to diverse family structures in sub-Saharan Africa.
Parents’ Union Status and Child Health in Sub-Saharan Africa
A confluence of factors, including childbearing trends, union formation and dissolution patterns, migrant labor practices, and the HIV/AIDS epidemic, contribute to children living in diverse household structures in sub-Saharan Africa (Gaydosh 2015; Odimegwu et al. 2017), which was a focus of early demographic research on child well-being in the region. The childhood health implications of living in female-headed households was of particular interest: early studies reported that children residing in female-headed households have comparable, or even better, outcomes than children residing in male-headed households (Desai 1992; Kennedy and Haddad 1994; Kennedy and Peters 1992).
This somewhat puzzling finding helped pave the way for more recent scholarship, which has studied the specific family processes that drive distinct household structures in order to understand patterns in child health inequality better. For instance, more recent research has explored the specific child health and developmental implications of parents’ death (Case and Ardington 2006; Case et al. 2004), migration (Gaydosh 2017), and union status (Clark and Hamplová 2013; DeRose et al. 2017).
Research on parents’ union status has documented some of the most striking inequalities among children. For example, children with never-married mothers experience severe health disadvantages, including an elevated risk of premature death (Clark and Hamplová 2013). Children whose never-married mothers are cohabiting with their father are also disadvantaged (Ntoimo and Odimegwu 2014), a finding possibly driven by selection of lower-quality relationships into cohabitation (Bertrand-Dansereau and Clark 2016). Additionally, children whose parents were married but subsequently divorced experience poor health and high mortality (Chae et al. 2016; Clark and Hamplová 2013; Gaydosh 2017; Ntoimo and Odimegwu 2014; Thiombiano et al. 2013). Some evidence contended that a mother’s remarriage can reduce or undo this disadvantage: under select circumstances in Mozambique, living with a stepfather can benefit children’s well-being (Lopus 2017). Yet, other studies have suggested that children living with stepfathers are disadvantaged relative to children living with their biological fathers and, in some instances, even their peers whose divorced mothers remain unmarried (Chae 2013; Clark and Hamplová 2013; DeRose et al. 2017).
Does Union Status Alter the Child Health Benefits of Parents’ Education?
Parents’ union status may further influence children by altering whether their resources, including education, benefit their children’s health. For instance, if a child lives with a single, never-married mother, the child may reap minimal benefit from their nonresident father’s education. Additionally, a never-married mother’s education may not benefit her children to the same degree as it would if she were married, given the economic, social, and logistical difficulties associated with single parenting (Clark and Hamplová 2013).
Divorce could similarly disrupt the intergenerational benefits of parents’ education. In sub-Saharan Africa and elsewhere, children with divorced mothers tend to receive daily care from more caregivers and spend less time with their mothers, relative to children whose mothers remain married (Crosnoe et al. 2014, Grant and Yeatman 2014). Divorce is also tied to women’s worse health outcomes (Djuikom and van de Walle 2018), which likely spill over to negatively affect their children. Thus, the less time and attention mothers can provide their children, as well as the additional economic, social, physical, and logistical challenges they face (DeRose et al. 2017), may weaken the impact of divorced mothers’ education on their children’s well-being.
Children’s ability to benefit from their fathers’ education in the wake of divorce may be even more constrained, particularly when the child lives with their mother. Divorced fathers living elsewhere are often less financially and functionally involved in their children’s lives, and their absence typically reduces their involvement in day-to-day decisions (Manning et al. 2003). Thus, divorce may disrupt the extent to which a father’s education—and the social position, labor, and finances it affords—can benefit his children (Townsend 2002). Accordingly, the health-related advantages tied to having a highly educated father should be weaker, or even unobservable, among children whose parents are divorced.
The potential for divorce to disrupt the child health benefits of having an educated father may be especially high in sub-Saharan African communities where matrilineal customs prevail. Among matrilineal ethnic groups, children tend to “belong” to their mother’s kin (Mitchell 1962; Reniers 2003; Schneider and Gough 1961). Indeed, in some matrilineal communities, research has shown that children have more contact with their mothers (Kohler et al. 2012), and fathers are generally seen as less-significant figures (Cook et al. 1998). Additionally, women in matrilineal communities commonly reside with their matrilineal kin following divorce (Grant and Pike 2019), further suggesting that divorce in these settings may powerfully sever the father-child relationship.
Alternatively, in areas where patrilineal ethnic groups are concentrated, the cultural significance of fathers and their kin may mean that even divorced fathers remain highly involved in their children’s lives. Evidence from select patrilineal contexts has shown that nonresident fathers invest heavily in children (Clark et al. 2015, 2017; Madhavan 2010; Madhavan et al. 2008, 2012, 2014, 2016; Townsend et al. 2005). Thus, in such settings, processes of father-to-child transmission of advantage may not hinge on the continuation of the parents’ union.
Beyond divorce, remarriage could also alter the health implications of parents’ education for children. Little is known about stepmothers in sub-Saharan Africa, but emerging research has shown that it is common for sub-Saharan African children to reside with stepfathers (Adjiwanou 2017). One of the few studies directly studying stepfathers in an African context suggested that they may invest less in their stepchildren than in their biological children (Lopus 2017). Remarriage can also prevent a mother from channeling resources to her children from a previous marriage. In Malawi, a mother’s remarriage elevates her children’s likelihood of being out-fostered, attesting to the complexities of blended families and the potential for remarriage to interfere with mother-to-child transfers (Grant and Yeatman 2014).
Other work, however, has supported a more favorable view of stepfathers and the potential for their education to have an equivalent, or even pronounced, influence on children’s health relative to married biological fathers. Corresponding with studies that minimized the salience of biological relatedness (Hamilton et al. 2007; Hofferth and Anderson 2003), research on select African contexts has shown that stepfathers spend considerable time with their stepchildren (Anderson et al. 1999) and often fully adopt the role of a social father (Madhavan et al. 2016). Given that blended families are prone to family dynamics that can harm children’s health (Castro et al. 2017), stepfathers’ education could be of heightened importance for their stepchildren’s healthy growth and development.
To examine these competing possibilities, I offer a multicountry analysis of whether parents’ union status corresponds with modified associations between mothers’ and fathers’—and when relevant, stepfathers’—education and children’s nutritional status. Specifically, I examine whether parents’ union status corresponds with distinct associations between parents’/stepparents’ education and children’s likelihood of being stunted and underweight—two comprehensive indicators of children’s overall physical health. Then, in a subset of countries featuring diverse lineage systems, I explore whether the findings differ systematically across settings where patrilineal versus matrilineal ethnic groups are concentrated.
Data and Sample
I used data from Demographic and Health Surveys (DHS), the standard data source for studying childhood health inequalities in sub-Saharan Africa and other low-income regions. The DHS program is a nationally representative, cross-sectional survey fielded in participating countries at roughly five-year intervals. The DHS uses a stratified random sampling approach with clusters providing the primary sampling unit. These clusters reflect small areas within, or corresponding to, communities. Within selected clusters, the DHS randomly samples a group of households.
I used DHS data from the 30 sub-Saharan African countries highlighted in Fig. 1 and listed in Table A1 in the online appendix. I classified the predominant family lineage pattern in each subnational region using ethnicity data, but the DHS does not collect these data for all 30 countries. Thus, to clarify whether the findings are consistent across matrilineal and patrilineal contexts in the region, as depicted in Fig. 1, I focused on a subsample of 11 countries where the DHS collects ethnicity information and where there is subnational regional variation in the dominant lineage system (which I identify through the process described shortly).
Because the DHS collects nearly identical data across sub-Saharan Africa, I pooled the surveys for analysis, as researchers commonly do (Kravdal 2002; Molitoris et al. 2019). I focused on children under 5 years old at the time of the survey, for whom there is complete anthropometric data, resulting in a sample of children aged 0–59 months who were born between 1998 and 2017. In supplementary analyses, I explored whether the key findings varied systematically over the 20-year period; I found no evidence that they do.
Because the DHS asks mothers to report on their own education and that of their current husband (or previous husband if a woman is divorced and not remarried), I focused on the approximately 90% of children in the 30 countries whose mothers were alive, residing in the same household as them, and had completed an interview. To identify this target sample of children, I began with data from household heads. DHS interviewers ask household heads to provide a full roster of household members, including young children. They also ask household heads to report whether each child’s mother and/or father is alive and, if so, whether the parent is residing in the household.
Note that the analytic sample excludes the 10% of children who live away from their mother (or who have a deceased mother), given that there is no parent education data for these children. This exclusion narrows the study’s ability to analyze variation in parents’ education as a function of union status (and living arrangements). Specifically, the study cannot speak to the experiences of children residing apart from their divorced mother or with a remarried father and stepmother.
Additional data limitations required that I exclude the approximately 5% of children born to mothers who had never been married and who were not cohabiting with a partner at the time of the survey. The DHS does not ask these women to provide information about their child’s father, which prevented me from exploring the salience of fathers’ education in the case of never-married parents who are not currently cohabiting. In supplementary analyses (see Table A2, online appendix), however, I confirmed that the impact of mothers’ education for these children’s health outcomes does not differ from that of children with married mothers. Finally, I excluded the 2% of children with deceased fathers (and still unmarried mothers). In some of these cases, I had data on deceased fathers’ education; I combined these children with their peers with divorced fathers to confirm that doing so did not alter the key findings. After excluding the less than 3% of children with missing data, the full sample consisted of 271,292 children under age 5 residing with their mother in 311 regions in 30 sub-Saharan African countries.
Child Health Outcomes
I assessed children’s nutritional status by analyzing the anthropometric data that the DHS collects on children’s height and weight (Mei and Grummer-Strawn 2007). DHS interviewers receive training on collecting anthropometric data to help ensure its accuracy and completeness. Following standard protocol, DHS interviewers measure children’s height (or length, using a measuring board, if children are younger than 24 months) to the nearest 0.1 cm. DHS interviewers measure children’s weight on a pediatric scale (or beam balance scale) to the nearest 0.1 kg. All interviews require two measurements of each child’s height and weight to ensure agreement within 0.1 cm and 0.1 kg, respectively (Mei and Grummer-Strawn 2007). The DHS then codes children according to the number of standard deviations they are below/above the World Health Organization’s standard for healthy growth.
Using these data, I created two binary dependent variables for whether children are stunted and/or underweight, which are correlated and share similar risk factors (Brennan et al. 2003). Following convention, I coded children as stunted if they fell 2 or more standard deviations below the recommended height for their age and sex. I coded children as underweight if they fell 2 or more standard deviations below the recommended weight for their age and sex. The cross-sectional nature of the data is a limitation of analyzing child growth. Children had experienced a growth trajectory prior to their parents’ current union status at the time of the survey. This is arguably most concerning in the analysis of stunting; however, even stunting is sensitive to recent changes in children’s environments (Desmond and Casale 2017). Nonetheless, it is important to keep this limitation in mind when interpreting the results pertaining to stepfathers, given that the analyses do not account for how long children have resided with them.
Mothers’ and Fathers’/Stepfathers’ Education
I classified children according to their mothers’ self-reports of their highest level of education, including having attended no school, some primary school, complete primary school, some secondary school, complete secondary school, or higher education (coded as 0–5, respectively).2 The DHS also interviews men, including some of the fathers and stepfathers of the children in the sample. If the DHS interviewed a child’s father/stepfather, I created an analogous indicator of his highest educational level based on his self-report. If a child’s father/stepfather did not participate in an interview because he was away from the household at the time of the interview, I relied on the mother’s report of his education. For children with divorced parents whose mothers had not remarried, the mother reported on the father’s education.
Given the cross-sectional nature of the data and the lack of information on when mothers and fathers completed their education, it is possible that some (particularly young) parents returned to school after their child’s birth. This could introduce error into the analysis of their education and their children’s health outcomes. Most young adults in the region leave school by age 20 (Lloyd and Mensch 2008), so in supplementary analyses, I excluded all children born to a teenage parent who may not have completed school before their child’s birth. Results were consistent, so I maintained them in the analyses shown here.
Parents’ Union Status
I used data from multiple DHS data files to classify children according to their parents’ union status. Again, I focused on a target sample of children under 5 years old (born between 1998 and 2017) who were residing with their mother at the time of the survey (given that these are the only children for whom I have information on parental education). Then, combining data from household heads on parental presence with data from mothers on their current and former marital status, I classified children accordingly: (1) cohabiting biological parents, (2) married biological parents (reference group), (3) divorced parents, or (4) a remarried mother and stepfather.3
Interaction Between Parents’ Education and Union Status
I created a series of interaction terms between parents’ education and union status. These two sets of interactions (for mothers and fathers/stepfathers) enable me to identify any distinct associations between parents’ education and children’s health as a function of union status.
Because the lineage system in which families reside influences norms surrounding marriage, divorce, and parent-child relationships, I classified, when possible, children’s local region of residence as either matrilineal or patrilineal/other. To do so, I drew on the Ethnographic Atlas, which Murdock, a prominent anthropologist, published in 29 successive installments over an approximately 20-year period. Murdock (1967) first published data for 862 societies, which was later expanded to cover 1,167 societies. Gray (1999) compiled the final version, which includes information on whether each society is matrilineal (vs. patrilineal or other).4 In line with past studies (Rossi and Rouanet 2015), I linked these data to DHS data (see Table A1 in the online appendix for the list of specific countries). Specifically, for the subsample of countries in which the DHS asked women their ethnicity, I created an indicator that denotes the majority ethnic group for each region. I coded regions as matrilineal if the Ethnographic Atlas lists the majority ethnic group as such. In the 11 countries, 46 regions were coded as matrilineal (= 1); the remaining 73 regions were coded as patrilineal/other (= 0) (see Fig. 1).5
Individual, Household, Regional, and Country Controls
All models account for several individual, household, and regional covariates that influence child health. In terms of individual characteristics, which I derived from interviews with children’s mothers, I included the child’s gender and the mother’s age at the time of the child’s birth.6 I did not include data on fathers’ age because of excessive missing data on divorced fathers. I also included an indicator for duration of the preceding birth interval (none: child is firstborn; ≤17 months; 18–36 months; and >37 months) and whether the child is a multiple.7 I controlled for the child’s birth order and the child’s age (in months).8 I also accounted for the DHS household wealth index, which captures inequality in material resources across households. A principal component factor analysis of a household’s ownership of various assets (e.g., radio, television, refrigerator, bicycle, car) and housing characteristics (e.g., availability of electricity, source of drinking water, type of toilet facility, number of rooms) is the basis for categorizing households into five wealth quintiles, ranging from poorest to least poor.
I created covariates to account for differences in the subnational regional contexts in which children live. To account for differences in regions’ educational profiles (i.e., the adult population’s educational attainment), I leveraged full DHS data on all participating men and women, regardless of their marital or parental status, to quantify the percentage of adults who have ever attended secondary school. I further differentiated regions according to the proportion of households that are in a rural versus urban area, as well as the proportion that have piped water into the dwelling.
Because country-level factors are associated with child health (Swiss et al. 2012), and specifically children’s risk of being stunted and underweight (Stevens et al. 2012), I included a set of dummy variables to represent each of the 30 countries in the sample (11 countries in the subanalysis focused on lineage). This country-level fixed-effects approach allowed me to control for constant, unobserved, between-country factors. I also accounted for the year of the survey.
After offering a descriptive overview of the children in the sample, including their health, parents’ education, and parents’ union status, I present results from multilevel logistic regression models of the relationships between parents’ education and children’s likelihood of being stunted and underweight, and whether they vary by parents’ union status.
The multilevel logistic regression models address the binary nature of the dependent variables and the data’s hierarchical structure. In the models of stunting and underweight, I first show the individual associations between children’s outcomes and their parents’ education and union status, net of covariates. I then model the interactions between parents’ education and union status on children’s likelihood of being stunted or underweight. Next, I show the same models reestimated among the subsample of children in the 11 countries where I can disaggregate children by the majority lineage system in their local region. Because interpreting interaction terms is difficult in nonlinear regression (Ai and Norton 2003), I graph the predicted probabilities for significant results.
Table 1 presents descriptive statistics. Just over one-third of the children in the analytic sample have stunted growth, and approximately one-fourth are underweight for their age, highlighting the compromised nutritional standing of many children in sub-Saharan Africa. Most of the sample reside with married biological parents, yet more than 1 in 10 live with their cohabiting biological parents. Approximately 6% of the sample live with a divorced single mother, and 12% are currently residing with their remarried mother and stepfather.
The second and third columns of Table 1 offer descriptive statistics for the subsamples of children residing in the 11 countries included in the lineage analyses (specified in Table A1, online appendix), disaggregated by the lineage of the majority ethnic group in their subnational region. The prevalence of stunted or underweight children is comparable in majority matrilineal and patrilineal regions. Notably, a higher percentage of children reside with a currently divorced mother in majority matrilineal regions, which could be due to higher divorce and lower remarriage rates in these settings. This distinction could also be due to lower levels of coresidence between children and divorced mothers in patrilineal settings, or divorced women’s slower transition to a new marriage in these settings. Parents’ educational attainment also tends to be higher in matrilineal settings relative to patrilineal ones.
Table A3 in the online appendix offers additional descriptive statistics for each group of children, disaggregated further by parents’ union status. These additional statistics demonstrate that in general, children with cohabiting or remarried mothers have lower levels of stunting and wasting, and this pattern is consistent across regions with distinct lineage systems. The findings further show important variation in average educational attainment across family structures.
The strong relationship between parents’ education and child health shown in Fig. 2 further attests to the need to understand whether children in distinct family structures benefit from parents’ education equally. Figure 2 shows a strong linear reduction in the odds of poor physical health with each level of education that a parent attained. As shown in the left panel of the figure, nearly twice as many children whose fathers/stepfathers have never been to school are stunted or underweight relative to those whose fathers/stepfathers completed higher education. As shown in the right panel, mothers’ education also corresponds with children’s better health: nearly 30% and 40% of children whose mothers have never been to school are underweight and stunted, respectively, compared with less than 10% of children whose mothers have higher education.
Parents’ Education, Union Status, and Child Health
Table 2 begins with multilevel logistic regression models of children’s likelihood of stunting by their parents’ union status and education, net of a host of individual, family, and contextual controls. Results in Models 1 and 2 confirm previously identified inequalities across family structures: after confounding indicators are accounted for, children residing with cohabiting versus married biological parents have a 5% increase in the odds of being stunted and a 7% increase in the odds of being underweight. Children with divorced parents experience a 15% increase in the odds of stunting and a 20% increase in the odds of being underweight relative to their peers whose parents are married. Children living with their remarried mothers and stepfathers, however, have no meaningful elevation in the odds of being stunted or underweight.
Net of inequalities tied to parents’ union status, Models 1 and 2 show that parents’ education also has strong independent associations with children’s physical health. An increase in a father’s/stepfather’s level of education corresponds with a 5% reduction in the odds of stunting and a 6% reduction in the odds of being underweight (p < .001). An increase in a mother’s level of education corresponds with an additional 10% reduction in the odds that a child is stunted and an 11% reduction in the odds of being underweight (p < .001). This highlights the additive and large associations between parents’ education and child health.
Do the strong, intergenerational health advantages tied to parents’ education vary by parents’ union status? Models 3 and 4 in Table 2 include interactions between parents’ union status and education. The intergenerational advantages associated with fathers’/stepfathers’ education are no smaller for children whose parents have not formalized their union, nor are they distinct for children whose parents are divorced. Moreover, results pertaining to stepfathers’ education demonstrate a pronounced protective association between a stepfather’s education and a child’s odds of being stunted or underweight. That is, children with more highly educated stepfathers experience a more dramatic reduction in the risk of being stunted or underweight than do their peers who live with their comparably educated biological fathers. This heightened benefit tied to stepfathers’ education operates net of any possible—although nonsignificant—distinctions in the impact of mothers’ education. As shown by the nonsignificant interaction terms, the results offer no evidence that divorced mothers’, or remarried mothers’, education has distinct associations with children’s physical health. To depict this graphically, Figs. 3 and 4 show the more dramatic decline in children’s probability of stunting (Fig. 3) and being underweight (Fig. 4) by stepfathers’ educational attainment relative to the decline experienced among children living with their equally educated, married biological fathers.
Variation Across Lineage Systems?
I reestimated the preceding models among a subsample of children to assess whether these findings are consistent across predominately matrilineal and patrilineal subnational regions. As shown in the first two columns of Table 3, there are no meaningful distinctions for married biological fathers’ versus stepfathers’ education in majority matrilineal regions. In line with the main study findings, there is also no evidence that mothers’ education varies by parents’ union status in majority matrilineal settings.
The last two columns of Table 3, however, show that the results documented in the full sample are only evident in patrilineal regions, especially in terms of stunting. As depicted in Fig. 5, for children residing in patrilineal settings, a stepfather’s education has a notably stronger influence on lowering their likelihood of stunting than that of a married biological father. Additionally, Fig. 5 graphs the documented variation in the impact of remarried mothers’ education versus mothers still married to their child’s father (see also Table 3). Children whose stepfathers have more education experience more protection from stunting, but their mothers’ education provides less protection relative to their peers with equally educated, married parents.
The long arm of childhood health, and its wide-ranging influence on individuals’ lives, highlights the critical need to understand the determinants of well-being in early life. Parents’ education continues to powerfully pattern young children’s health (Karlsson et al. 2018), yet we know little about whether parents’ union instability could disrupt this intergenerational process (Kalmijn 2015). This issue is of prime importance in the sub-Saharan African context, given the region’s emphasis on educational expansion as a strategy for improving young adults’, and their children’s, lives. Thus, against the backdrop of educational change (Grant and Behrman 2010), this study asks whether the high levels of union instability and corresponding diversity in family structures could prevent young children from benefiting from their parents’ education.
The study results offer no indication that this is the case: I find no evidence that being born to unmarried parents, nor experiencing parents’ divorce, disrupts the health advantages tied to parents’ education. This is true in subnational regions where patrilineal practices are predominant as well as in regions where matrilineal groups are concentrated. This finding also appears to be stable across time. Indeed, children with highly educated stepfathers experience slightly better health outcomes than do children living with their equally educated, married biological fathers. Additional analyses of a subsample of 11 countries suggest this finding is mostly concentrated among children residing in majority patrilineal settings: in these regions, the association between stepfathers’ education and child health is pronounced, whereas the relationship between mothers’ education and child health is modestly flattened.
The data do not include information on parental behaviors and family dynamics that would facilitate analysis of how fathers’ education continues to have such a strong positive influence on their children’s lives even when they are divorced and residing elsewhere. Moreover, the data lack information that would offer insight into why stepfathers’ education has heightened benefits for stepchildren’s health, particularly in majority patrilineal areas. Yet, past work points to a few plausible explanations. One possibility is that fatherhood is such a major element of men’s identity in patrilineal settings that children with stepfathers are truly experiencing double paternal investments. Although Madhavan et al. (2015) showed that in some settings, nonresident biological fathers assume a more minor role in their children’s lives in the presence of a stepfather—and perhaps particularly so if the stepfather is investing in the child’s well-being—often these nonresident fathers are still highly involved in their children’s lives. Thus, the heightened benefit of stepfathers’ education that I document could be an artifact of the child simultaneously experiencing investments from the biological father and the stepfather—a sort of doubling-up of advantage. This possibility is particularly intriguing given that it defies the typical narrative of divorced and biologically unrelated fathers as less invested in children.
The possibility that divorced fathers and stepfathers are simultaneously investing in children attests to the value of collecting data on the multiple adults in children’s lives—a conclusion stressed by recent innovative data collection on kin and child well-being (Clark et al. 2018; Hosegood and Madhavan 2012; Madhavan et al. 2008, 2015). The scope of the DHS prohibits it from collecting detailed data on all potentially relevant kin, but at the very least, the DHS could incorporate follow-up questions that ask never-married, divorced, and remarried mothers about their children’s biological fathers’ and stepfathers’ basic characteristics and involvement in the children’s lives. Such minor additions to the questionnaire could significantly extend our understanding of men’s role in young children’s lives in the region.
It is also possible that the heightened health benefits tied to stepfathers’ education are specific to the select stepfather-child pairs I capture with DHS data. I am able to capture only those children residing with their stepfathers, and these stepfathers may be positively selected on their willingness to invest in the child, as evidenced by the fact that upon remarriage, the mother did not out-foster the child (Grant and Yeatman 2014). Moreover, it is notable that fewer children in the sample are residing with stepfathers in patrilineal settings than in matrilineal ones, potentially signaling that such coresidence arrangements are rare. Thus, it is possible that these findings are due to selection and are not necessarily symbolic of stepfathers’ generally high investments in stepchildren. Even so, this finding suggests the presence of a stepfather in a child’s home is not a risk factor, at least in terms of early childhood nutrition, but instead a protective one, particularly if the stepfather is socioeconomically advantaged.
The findings also demonstrate the growing need for research that explicitly studies distinctions in family-based inequalities across lineage contexts. Despite recognition that marriage and childbearing norms and practices vary dramatically across matrilineal and patrilineal communities in sub-Saharan Africa, research on child health inequalities by parents’ union status has rarely considered the lineage setting in which these processes unfold. This study demonstrates the ease with which one can synchronize DHS data with the Ethnographic Atlas to characterize subnational regions according to the majority lineage system. Future research that considers the majority lineage system in different regions may reveal important new insights for understanding child health inequality across families.
Overall, this study adds more evidence to the large literature on the stratifying nature of educational participation in Africa as well as education’s role in determining which parents have the means to ensure their children’s basic physical growth. The results demonstrate the additive benefit of having a more highly educated mother and father, and they attest to the importance of each parent across diverse family structures. There is no indication that divorce disrupts parents’ ability to use their education to transmit advantage. Rather, the results show that education is an even more powerful stratifying force in the presence of other family structures, specifically coresidence with a stepfather. Thus, even as high proportions of children experience their parents’ divorce in sub-Saharan Africa, their parents’, including their stepfathers’, education will continue to pattern their health and well-being.
I thank the three anonymous reviewers and the Editors of Demography, whose comments have greatly improved this manuscript. I also thank Vissého Adjiwanou, Kieron Barclay, Abdallah Chilungo, Sara Lowes, Synab Njerenga, and Abby Weitzman for helpful comments and suggestions on both technical and conceptual components of this research.
Author’s calculation based on United Nations World Population Prospects database: https://esa.un.org/unpd/wpp/dataquery.
I used this measure, rather than a measure of the highest year of school completed, given differences in educational systems across sub-Saharan African countries (and within countries at different points in time) that mean primary and secondary school constitute a variable number of years. Because of the study’s multicountry scope, this measure offers a better standard for differentiating parents by their educational credentials. Using a six-categorical variable as a continuous variable is appropriate when categorical thresholds are symmetrical (Rhemtulla et al. 2012), and the predicted probabilities confirm a linear association between each level of education and children’s odds of poor physical health.
Seven percent of mothers of children coded as having a “stepfather” were cohabiting with their partner but were not married. Results were stable when these children were disaggregated, so I combine them into a single category and, for ease of discussion, reference these men as children’s “stepfathers.” Among children with married biological parents or remarried mothers, some of the fathers/stepfathers were temporarily staying elsewhere at the time of the survey. Paternal migration can benefit child health (DeRose et al. 2017; Townsend 2002; Yabiku et al. 2012) and could alter intergenerational processes. Thus, in supplementary analyses, I disaggregated children according to whether their biological fathers and stepfathers were staying elsewhere at the time of the survey. I found no evidence that doing so alters the findings. In additional analyses, I omitted these children, and the results were stable to their exclusion.
Data are publicly available at http://intersci.ss.uci.edu/wiki/index.php/Ethnographic_Atlas#Electronic_versions_of_the_Atlas_and_SCCS. “Other” lineage includes duolateral, quasi-lineages, ambilineal, bilateral, and mixed.
Regions coded as matrilineal include Sud-Ouest in Burkina Faso; Kinsaha, Bans-Congo, Bandundu, Nord-Kivu, Maniema, Sud-Kivu, Katanga, Kasa-Oriental, Kasa-Occident, Equateur, and Orientale in the Democratic Republic of the Congo; Kouilou, Niari, Bouenza, Pool, Sangha, Brazzaville, and Pointe-Noire in the Congo; Centre, Centre-Nord, Sud Sans Abidjan, and Ville d’Abidjan in Ivory Coast; Centre, Sud, and Yaoundo in Cameroon; Haut-Ogoou and Ogoou-Lolo in Gabon; Western, Central, Greater Accra, Eastern, Ashanti, and Brong-Ahfo in Ghana; Central and Southern in Malawi; Fatick in Senegal; Savanes in Togo; and Central, Cobberbelt, Eastern, Luapula, Lusaka, Northern, Northwestern, and Southern in Zambia. The majority ethnic groups that are matrilineal in these listed regions include the Akan, Akye Out Attie, Bas-Kasai and Kwilu-Kwngo, Bakongo, Baoule, Bemba, Beti/Bassa/Mbam, Basele-K, Luvale, Chewa, Kasawi, Katanga, Kongo, Lala, Lobi, Lunda, Mbede-Teke, Sangha, Serer, Teke, Tonga, Tanganika, and Yao.
In supplementary analyses, I used data on mothers’ HIV status for the subsample of 100,454 children in 23 countries where such data were available (Burkina Faso, Burundi, Congo, Democratic Republic of the Congo, Cameroon, Ethiopia, Gabon, Ghana, The Gambia, Guinea, Ivory Coast, Kenya, Liberia, Lesotho, Niger, Namibia, Rwanda, Sierra Leone, Senegal, Togo, Tanzania, Zambia, and Zimbabwe). Supplemental models confirmed that after accounting for mothers’ HIV status, the coefficients and standard errors vary in minor ways. Yet the main findings, in terms of the direct impact of family composition and parental education, were consistent. Full results are available upon request.
Some of the children in the sample are siblings; however, because the majority of children are the only focal child of their mothers included in the study (because the sample is restricted to living children under age 5), I did not include household as a level in the analysis. In supplementary models, I used the random sampling feature in Stata to select one child per mother to assess whether the findings were consistent (N = 196,078). Full results are available upon request.
A child’s gender can consciously and subconsciously influence intergenerational processes (Raley and Bianchi 2006), including fathers’ and stepfathers’ investments (Anderson et al. 1999; Harris et al. 1998), which can manifest as distinctions in children’s nutritional standing (Wamani et al. 2007). In supplementary analyses, I explored whether these key findings vary for girls versus boys. To do so, I first analyzed three-way interactions by child gender. I then disaggregated children by their family composition and included two-way interactions between fathers’/stepfathers’ education and child gender as well as three-way interactions that added birth order. None of these models showed evidence that the findings differ significantly for girls versus boys, nor for children of higher or lower birth order. The same was true of models exploring the variable effect of mothers’ education on child health: considering child gender did not alter the nonsignificant variation in mothers’ education across distinct family compositions, nor did I find evidence that within specific family compositions, mothers’ education differentially benefitted girls versus boys, even when considering birth order. Evidence of important distinctions in child gender across matrilineal and patrilineal regions, however, merits further research on how gender alters parent-child relationships across distinct lineage contexts.
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.