Post on 09-Jul-2018
Parental Divorce and Child Mortality in Sub-Saharan Africa: Does Context Matter?
Emily Smith-Greenaway University of Michigan
University of Southern California mailto:smithgre@usc.edu
Shelley Clark McGill University
shelley.clark@mcgill.ca
Population Studies Center Research Report 15-843 July 2015
Paper first prepared for presentation in session 178 at the 2015 Population Association of America’s annual meeting in San Diego, California. Emily Smith-Greenaway gratefully acknowledges use of the services and facilities of the Population Studies Center at the University of Michigan, funded by NICHD Center Grant R24 HD041028 and the support of an NICHD training grant to the Population Studies Center at the University of Michigan (T32 HD007339).
Parental Divorce and Child Mortality in Sub-Saharan Africa: Does Context Matter? 2
Abstract Parental divorce is known to be associated with poor child health in sub-Saharan Africa. However, the great diversity in social and cultural practices, including marital ones, across the subcontinent likely yields substantial variation in the prevalence of divorce, which in turn could moderate its consequences for children. In this paper, we use Demographic and Health Survey data from 31 African countries to estimate variation in the prevalence of divorce across 290 subnational regions. We find that the prevalence of divorce varies significantly, ranging from less than 1 percent of women in some subnational regions to nearly 20 percent in others. We then estimate a series of multilevel discrete-time hazard models to analyze the association between parental divorce and child mortality, and to test whether the mortality risk associated with divorce varies significantly in subnational regions where divorce is rare compared to those where it is common. We find that children with divorced parents experience significantly amplified mortality risk where divorce is rare, even though their mothers are positively selected on socioeconomic traits. We argue that the stigmatization and discrimination associated with divorce in subnational regions where it is rare, and the lack of social and cultural support, leave divorced mothers and their children particularly vulnerable to poor health outcomes in these settings.
Parental Divorce and Child Mortality in Sub-Saharan Africa: Does Context Matter? 3
INTRODUCTION
Extensive evidence shows that children with divorced parents have worse outcomes
compared to their peers whose parents stay married (Amato, 2010). The vast majority of
evidence comes from studies on North American and European families; however, recent
research on African families has linked divorce to especially severe childhood disadvantages.
African children with divorced parents are more likely to be malnourished (Gage, 1997),
miss out on vaccinations (Gage, 1997), attend less school (Chen, 2012; Thiombiano,
LeGrand, & Kobiané, 2013), and die before age five (Clark & Hamplová, 2013;
Thiombiano et al., 2013) than their peers whose parents remain married.
Like most research on the consequences of family processes for children’s wellbeing,
studies on the implications of parental divorce in Africa have focused exclusively on the
family-level, giving no recognition to the broader cultural, social, and marital context in
which divorce unfolds. Nonetheless, recent research suggests that one contextual feature that
can powerfully moderate the consequences of non-traditional family forms for children is the
extent to which they are prevalent (Amato & Keith, 1991; Kalmijn, 2009; Liefbroer &
Dourleijn, 2006; Pong, Dronkers, & Hampden‐Thompson, 2003; Smith-Greenaway &
Trinitapoli, 2014). Based on this research, we hypothesize that having divorced parents in a
setting where divorce is rare could bear distinct consequences for children than having
divorced parents in a setting where divorce is more prevalent.
Although demographers often characterize African marriages as having high divorce
rates (Caldwell, Caldwell, & Quiggin, 1989; Therborn, 2004), in reality, the subcontinent’s
diverse nuptial, cultural, and kinship practices—matched with socioeconomic disparities
across contexts—produce considerable variation in the prevalence and corresponding social
acceptability of divorce. For instance, a recent country-level study shows that whereas as few
as 6.5 percent of women ever divorced (by age 50) in Lesotho, nearly 40 percent of women
divorce in Liberia (Clark & Brauner-Otto, 2015). Evidence from a handful of country-
specific studies suggest that there may be further variation at the subnational-level due to
the ethnic, kinship, and socioeconomic diversity within African countries. For instance,
research from Malawi concludes that first marriages are more than twice as likely to end in
divorce in the southern district compared to marriages in the central and northern districts
(Reniers, 2003, 2008). Similarly, ethnographic work in Nigeria confirms divorce is
extraordinarily common in the northern Hausa region (Solivetti, 1994) but virtually non-
existent in the country’s southeastern region (Smith, 2009).
Parental Divorce and Child Mortality in Sub-Saharan Africa: Does Context Matter? 4
This substantial variation in the prevalence of divorce could alter the impact of
divorce on children’s wellbeing. On the one hand, the childhood health disadvantages
associated with divorce could be dampened where divorce is less common and exacerbated
where divorce is more prevalent. Multiple studies have shown that individuals who select
into divorce in contexts (or historical periods) where it is rare tend to be positively selected
on socioeconomic traits (Chen, 2012; Goode, 1963; Park & Raymo, 2013; Raymo,
Iwasawa, & Bumpass, 2004). Because socioeconomic resources are known to be protective
of child mortality (Boyle et al., 2006), if divorced women are socioeconomically advantaged
in settings where divorce is less common, divorce may be unassociated—or only weakly
associated—with poor child health in these settings.
On the other hand, the opposite could be true: the child health disadvantages
associated with divorce may be amplified in settings where divorce is rare versus those where
it is more prevalent. Despite evidence of pronounced positive selection of the
socioeconomically advantaged into divorce in low divorce settings, there is evidence that
divorce in these contexts tend to be negatively selected in terms of marital conflict. That is,
marriages that end in divorce where the barriers to getting divorced are greater tend to have
worse marital problems compared to those who opt for divorce in settings where divorce is
more common and accessible (De Graaf & Kalmijn, 2006). Because high-conflict, violent
home environments are known to be associated with worse child health outcomes (Åsling-
Monemi, Pena, Ellsberg, & Persson, 2003), the selection of high-conflict situations into
divorce where it is uncommon could drive the association between divorce and poor child
health to be larger in these settings. The unique social and psychological challenges divorcées
face in settings where it is rare versus widespread could also operate to more directly elevate
its negative consequences for children. Individuals—particularly women—who are divorced
where it remains uncommon are likely to face considerable social stigma and to lack
institutional, cultural, and social support when navigating the divorce process, thereby
leaving them and their children particularly vulnerable to social isolation, psychological
distress, and ultimately poor health.
In this paper, we extend the literature on divorce and child health through three
objectives. First, building on evidence of variation in the prevalence of divorce both between
(Clark & Brauner-Otto, 2015) and within (Reniers, 2003, 2008; Smith, 2009; Solivetti, 1994)
countries, we use Demographic and Health Survey (DHS) data from 31 countries to provide
the first comprehensive estimates of divorce prevalence across 290 of Africa’s subnational
regions. Second, leveraging variation in the prevalence of divorce, we estimate a series of
multilevel discrete-time hazard models among 357,474 children to assess whether the child
mortality risk associated with divorce varies by the prevalence of divorce in the subnational
region. Our modeling approach is guided by competing hypotheses of whether (1) children
Parental Divorce and Child Mortality in Sub-Saharan Africa: Does Context Matter? 5
with divorced parents fare better in settings where divorce is rare due to the pronounced,
positive selection of socioeconomically advantaged women into divorce in these settings or if
(2) children with divorced parents fare worse where divorce is rare due to either (a) negative
selection of high-conflict, violent relationships into divorce in these settings or (b) the distinct
social and psychological challenges divorced women face where it is uncommon and taboo.
Divorce in sub-Saharan Africa and its Consequences for Children
A rich literature from diverse contexts concludes that parental divorce produces
negative consequences for children across multiple domains of wellbeing, and that these
disadvantages persist over their life course (for a review, see Amato 2010). Historically, the
literature has focused heavily on North American and European families, with little known
about the consequences of divorce for children in low-income regions of the world.
More recently, a handful of studies have confirmed that the consequences of divorce
for African children are even more severe than those documented in high-income world
regions. Research on Kenyan families shows that children of formerly married mothers are
significantly more likely to be malnourished and to miss out on key vaccinations compared to
their peers with married mothers (Gage, 1997). In Burkina Faso, having divorced parents is
associated with significantly higher risk of under-five mortality compared to having married
parents (Thiombiano et al., 2013). A recent multinational study corroborates these country-
specific findings: children with divorced parents have a 36 percent increase in the risk of
dying in the Democratic Republic of the Congo, a 52 percent increase in Liberia, a 57 percent
increase in Nigeria, and an almost two-fold increase in Sierra Leone compared to their peers
with married parents (Clark & Hamplová, 2013). African children with divorced parents who
survive early childhood go on to face sizeable educational setbacks, including delayed school
entry (Thiombiano et al., 2013) and lower educational attainment compared to their peers
with married parents (Chae, 2013).
Variation in the Consequences of Parental Divorce for Child Wellbeing
These studies provide strong evidence that divorce is associated with significant
health and educational disadvantages in Africa. The severity and size of the childhood
disadvantages associated with divorce raises the question of whether the magnitude of the
divorce penalty is uniform across Africa’s diverse social, cultural, and marital contexts: do
children with divorced parents living in African settings where divorce is prevalent
experience the same health risks as children with divorced parents living in contexts divorce
is exceptionally rare?
Parental Divorce and Child Mortality in Sub-Saharan Africa: Does Context Matter? 6
Dampened Childhood Consequences of Divorce in Low vs High Divorce Settings
Children with divorced parents may be less disadvantaged in places where divorce is
rare (versus more common) namely due to the pronounced positive selection of
socioeconomically advantaged women into divorce in these settings. Multiple studies from
diverse contexts have shown that individuals who select into divorce in contexts (or historical
periods) where it is rare tend to be positively selected on several socioeconomic traits (Chen,
2012; Goode, 1963; Park & Raymo, 2013; Raymo et al., 2004). Researchers hypothesize
that socioeconomically privileged individuals—especially women—are more likely to opt for
divorce where it is uncommon (Bernardi & Martínez-Pastor, 2011; Chen, 2012; Martin,
2006; Park & Raymo, 2013; Raymo et al., 2004) because they have sufficient resources to
independently support themselves and their children, thereby making divorce a viable option.
Moreover, because divorce is socially “innovative” in these settings(Goode 1963), it is likely
to be more readily adopted by those who are more educated and have been exposed to new
ideas and perspectives on the traditional family form. Conversely, in settings and historical
periods where there are few legal, social, and cultural barriers to divorce, the patterns of
selection have been shown to “cross-over”, resulting in divorce being concentrated among the
socioeconomically disadvantaged (Chen 2012).
Combining extensive evidence that the socioeconomic profiles of divorced
individuals are advantaged in low divorce settings (versus high divorce ones) with evidence
that socioeconomic factors drive most of the child health disparities associated with single
motherhood (Thomson, Hanson, & McLanahan, 1994), if divorced women are
socioeconomically advantaged in low-divorce settings, their children’s health experiences
may be comparable to—or even superior than—their peers with married parents. Put
differently, if divorced women in high divorce settings in Africa socioeconomically
disadvantaged, their children will have considerably worse health profiles than their peers
with married parents. Together, this leads to the following hypothesis:
Hypothesis 1: The selection of socioeconomically advantaged women into divorce in settings where it is rare will drive children with divorced parents to have minimal (or no) health disadvantages (compared to their married peers) in these settings compared to contexts where divorce is more common and either less-selective or concentrated among socioeconomically disadvantaged women.
Amplified Childhood Consequences of Divorce in Low vs High Divorce Settings
Despite the pronounced, positive selection of socioeconomically advantaged women
into divorce in settings where it is rare, some evidence suggests that children with divorced
parents may fare worse in these settings. Amato and Keith’s (1991) meta-analysis shows that
the consequences of parental divorce on a host of psychological, social, economic,
educational, and relational factors were far greater when divorce was less common in the
Parental Divorce and Child Mortality in Sub-Saharan Africa: Does Context Matter? 7
United States and lessened as it became more prevalent. Using a cross-contextual framework,
Kalmijn’s (2009) multinational study similarly found that the negative consequences of
divorce on individuals’ wellbeing are amplified in European settings where divorce is less
common and muted in settings where it is more prevalent.
Why would the child health disadvantage associated with divorce be greater in
settings where it is rare? One explanation is negative selection: despite positive selection on
socioeconomic factors, other evidence suggests marriages that end in divorce in low divorce
contexts highlight negative selection in terms of marital factors. A low divorce rate in a
society is, at least in part, the result of high legal, social, and cultural barriers to divorce
(González & Viitanen, 2009; Goode, 1963). As a result, the threshold for leaving a bad
marriage is higher, meaning that only the most conflict-ridden, violent marriages end in
divorce. As a result, marriages that end in divorce where divorce is uncommon tend to have
higher levels of severe marital conflict compared to marriages that dissolve under high
divorce conditions (De Graaf & Kalmijn, 2006). In Africa, marital conflict often takes the
form of physical and sexual violence, as evidenced by some of the world’s highest rates of
intimate partner violence (Garcia-Moreno, Jansen, Ellsberg, Heise, & Watts, 2006; Jewkes,
Levin, & Penn-Kekana, 2002; Koenig et al., 2003). Based on the fact that mothers’ exposure
to intimate partner violence is strongly correlated with poor health among their children
(Åsling-Monemi et al., 2003), if divorce is concentrated among a select group of violent
relationships, especially where divorce is uncommon, children with divorced parents in these
settings may experience worse health simply because of the corresponding exposure to
violence. This perspective leads to the following hypothesis:
Hypothesis 2a: Children with divorced parents will experience the greatest disadvantage where divorce is less common due to the selection of especially violent marriages into divorce in these settings.
Even after accounting for marital conflict, another explanation is that divorce bears
higher social and psychological costs where it is rare, which may operate through multiple
pathways to more directly exacerbate the disadvantages children with divorced parents
experience in these settings. Family sociologists agree that the diffusion of a family
process—like divorce—in a society is closely associated with its level of social acceptability
and institutionalization (Cherlin, 2004). In other words, the social acceptability of divorce
and its prevalence tend to go hand-in-hand. For instance, the historical rise in divorce in the
United States and Europe was accompanied by a parallel increase in its social acceptability
(DiFonzo, 1997; Spanier & Thompson, 1987; Thornton, 1985; Thornton & Young‐
DeMarco, 2001; White & Booth, 1991).
Parental Divorce and Child Mortality in Sub-Saharan Africa: Does Context Matter? 8
Although we are not aware of any studies that have tracked the association between
the prevalence and social acceptability of divorce in Africa, a handful of ethnographies and
community studies suggest the two are as closely related. For instance, divorce is generally
seen as a socially acceptable outcome where it is common among the Hausa in northern
Nigeria (Solivetti, 1994) and matrilineal groups in central and southern Malawi (Kaler,
2001). In fact, in central Malawi, even religious leaders support divorce as a response to
particular marital issues, such as sexual infidelity (Trinitapoli, 2011). Conversely,
ethnographers have noted that divorce remains culturally, religiously, and socially
unacceptable in contemporary African societies where it is rare. For example, Smith (2009)
reports that divorce is viewed as immoral behavior in southeastern Nigeria, regardless of
the circumstances surrounding it.
In settings where divorce is rare and taboo, ethnographic research confirms that
divorcées, particularly women, experience intense stigma (Al-Krenawi & Graham, 1998).
For instance, divorced women in Israeli communities are perceived as distrustful, sexually
seductive, and a mortal threat to family life (Al-Krenawi & Graham, 1998). Qualitative
evidence from Kenya confirms that the stigma associated with divorce drives community
members to sever relationships, socially isolating divorced mothers and their children (Clark,
Beguy, Boco, and Cotton 2013). Cross-contextual research from Europe corroborates this
qualitative evidence, demonstrating that divorcées experience steeper declines in social
contacts after divorce in European regions where divorce is socially disapproved versus their
peers who divorce in European regions where it has greater acceptance (Kalmijn & Uunk,
2007).1 Because children of mothers with smaller, weaker social networks are known to have
worse health outcomes (Adams, Madhavan, & Simon, 2002), this suggests that if divorced
mothers have the least social support in African settings where divorce is rare, their children
may especially vulnerable to adverse health.
Related to the more dramatic social fallout associated with divorce where it is
uncommon, the psychological cost of divorce may also be more severe in these
contexts. Qualitative evidence from Kenya suggests that divorced women internalize the
social ostracizing that they encounter, with one woman stating: “I am disrespected
because I don’t have a husband. The women around here feel that since I am not
married then I am not a good member of the community” (Clark, Beguy, & Cotton,
1 Of course, even where divorce is relatively common and socially acceptable, divorced women likely experience some social losses after the dissolution of their marriage. However, in these contexts divorcées can turn to other divorced persons as experienced “veterans” (Gerstel, 1987). Thus, in settings where divorce is more prevalent, divorcées will have a naturally larger support system of other divorced women available in their community.
Parental Divorce and Child Mortality in Sub-Saharan Africa: Does Context Matter? 9
2013). Based on large literatures showing that internalizing negative self-perceptions
interferes with psychological well-being (Yang et al., 2007), which in turn, is closely
linked to a host of child health outcomes (Engle et al., 2007; Patel, Rahman, Jacob, &
Hughes, 2004), children with divorced mothers may have especially poor health
outcomes where divorce is uncommon and imposes the greatest psychological stress.
Together, this perspective leads to the following hypothesis:
Hypothesis 2b: Even after accounting for selection, the social and psychological challenges that accompany divorce in settings where it is uncommon will lead children to have worse health outcomes (compared to their married peers) than in settings where divorce is more prevalent.
DATA
Data for this study come from 31 Demographic and Health Surveys (DHS)
administered in sub-Saharan Africa since 2000. We use the most recent survey for the
following countries: Burkina Faso, Benin, Burundi, Cameroon, Chad, Congo (Brazzaville),
Democratic Republic of the Congo, Ethiopia, Gabon, Ghana, Guinea, Ivory Coast, Kenya,
Liberia, Lesotho, Madagascar, Malawi, Mali, Mozambique, Namibia, Niger, Nigeria,
Rwanda, Sao Tome Principe, Sierra Leone, Senegal, Swaziland, Tanzania, Uganda, Zambia,
and Zimbabwe.2
The DHS program is a nationally representative, cross-sectional survey fielded in
participating countries, typically every five years. The DHS uses a stratified random
sampling approach, with clusters providing the primary sampling unit. Within each selected
cluster, the DHS randomly samples families. Household heads complete a full roster of
members, from which the DHS identifies eligible men and women. Women are asked
whether they have children and, if so, to provide detailed information on each live birth (e.g.,
date of birth, whether the child is still alive, and the month and year of death for all deceased
children) from which the DHS generates the child datasets that we use here.
Analytic Sample
We focus our analysis on births that occurred within seven years of the survey. This
restriction reduces recall error and ensures that current family characteristics correspond
2 In ancillary analyses, we used the preceding round of each survey to ensure the stability of our findings in terms of the levels of divorce and multivariate results. In all instances, our results were consistent regardless of the year of survey, with the exception of Mozambique. In Mozambique, women’s reports of divorce/separation are significantly lower in 2011 versus 2003/04 and 1997. Because southern African countries contiguous to Mozambique show no downward trends in the prevalence of women who are currently divorced, we use the 2003/04 DHS survey in Mozambique.
Parental Divorce and Child Mortality in Sub-Saharan Africa: Does Context Matter? 10
reasonably well with children’s environment since birth, but also reduces excessive censoring
by ensuring the majority of the sample has reached their fifth birthday.3 In ancillary analyses,
we loosened the restriction to children born in the last ten years and reproduced the study’s
key findings. We further restricted the sample to children whose mothers have been married
no more than once because we cannot determine whether the first union of a remarried
mother ended in divorce or widowhood. Appendix A demonstrates that the profiles of
children whose mothers have been married more than once and are thus excluded from the
study differ from their peers, including higher levels of mortality and greater family
instability. Despite these differences, ancillary analyses (not shown but available upon
request) confirm that including these children in our hazard analysis and categorizing their
mothers as “remarried” produces results that are consistent with those shown in the main
analyses. Finally, we exclude approximately one percent of children from our analyses due to
missing data. Our final analytic sample consists of 357,474 births. Appendix B shows the
survey years and sample sizes for children in each of the 31 countries.
To better capture the within country variance in the prevalence of divorce, our
aggregate units of focus are subnational regions. Subnational regions have been used in
previous research to approximate cultural, ethnic, geographic, and political boundaries in
Africa (Smith-Greenaway & Heckert, 2013; Smith-Greenaway & Trinitapoli, 2014).
Because the DHS does not collect contextual data, to characterize the 290 subnational regions
in our sample, we aggregate data from interviews with household heads and reproductive-age
women. In some instances, the household head or mother of the children in our study are
included in the samples from which we create our aggregate measures. However, because
there are an average of 1,894 children in each subnational region, and because excluding
them does not alter our results, we include the index-family when creating the contextual-
level measures.
MEASURES
Child Mortality
The outcome variable is the hazard of mortality before age five. Children born less
than five years prior to the survey are censored at the time of the survey, which we
appropriately address through our multilevel discrete-time hazard modeling approach.
3 The DHS collects additional health related data for children born in the last five years. The additional survey items lead respondents to omit or displace young children’s birth dates (Schoumaker 2011), encouraging us to deliberately sample children older than five years to avoid this displacement (see Clark and Hamplová 2013).
Parental Divorce and Child Mortality in Sub-Saharan Africa: Does Context Matter? 11
Family Structure
The DHS asks all mothers their marital status at the time of the survey, including
whether they are never married, married, widowed, or divorced. For the purposes of our
analysis, we modify this measure in several ways. First, as we described in the analytic
sample section, because we limited our sample to children whose mothers have never
remarried, all married mothers in our sample have been continuously married. Similarly,
divorced and widowed mothers have been continuously divorced or widowed since their first
marriage ended. Second, we use information on the timing of mothers’ first marriage and
their child’s birth— precise to the month—to classify women who were married long after
their child’s birth as “never married” in order to reflect their marital status at the time of
birth. Recognizing that entry into marriage is a prolonged process in Africa, if married within
six months of their child’s birth, we classify women as having been “married”. Thus, our
“never married” category includes women who never married and those not married at (or
soon after) their child’s birth. In detailed ancillary analyses, we eliminated this “grace period”
and confirmed that it does not significantly influence the results shown here. Third, we
leverage data on whether there are co-wives present in the union. Based on the extensive
literature linking polygynous marriage to poor child health outcomes (Omariba & Boyle,
2007; Smith- Greenaway & Trinitapoli, 2014), we use this information to differentiate
between mothers who are in a monogamous versus polygynous union.
Our ultimate measure is a five categorical variable: divorced, widowed, never
married, married monogamously (reference group), and married polygynously.4 A limitation
of our measurement is that we do not have data on the timing of mothers’ divorce (or
widowhood), making it possible that some children died prior to divorce or widowhood.
However, our estimates of the association between parental divorce and child mortality are
strikingly similar to those produced by other studies that had more detailed data on the timing
of divorce (see, for example, Thiombiano et al 2013 and Clark and Hamplová 2013).
Prevalence of Divorce
We use data from all reproductive-age women interviewed—including women without
children—in each subnational region to calculate the percentage of women who are currently
divorced. We specifically estimate the prevalence of reproductive-age women who are currently
divorced in each subnational region, not the prevalence of women who have ever divorced,
because it provides a valuable snapshot of the subnational region’s marital landscape and offers a
4 In line with existing research, given the high number of informal and unregistered unions in sub-Saharan Africa, we use the term “married” to refer to both married and cohabiting unions and “divorce” to refer to women who are both divorced or separated (see, for example, Porter et al 2004).
Parental Divorce and Child Mortality in Sub-Saharan Africa: Does Context Matter? 12
better sense of the extent to which divorced women in a community remain single versus
remarry. Because stigma acts as a powerful motivation for divorced women to quickly remarry
(Grover, 2011), which keeps the prevalence of divorce low at any given time, the percentage of
currently divorced arguably offers a better sense of the extent of its social acceptability
compared to the percentage of women who have ever divorced.
Cross-level Interaction between Prevalence of Divorce and Family Structure
To assess whether the child mortality risk associated with divorce varies across low
versus high divorce settings, we interact the subnational region-level prevalence of currently
divorced women with the marital status of individual mothers. A negative coefficient for the
cross-level interaction term will confirm that the child mortality risk associated with divorce
is larger in settings where divorce is more prevalent (and thus smaller where divorce is rare)
(hypothesis 1), whereas a positive coefficient will confirm that the child mortality risk
associated with divorce is significantly smaller in settings where divorce is more prevalent
(and thus larger where divorce is rare) (hypothesis 2a and 2b).
Socioeconomic Resources
Because prior research has suggested that the socioeconomic profiles of divorced
women may differ by the prevalence of divorce, we account for two aspects of mothers’
socioeconomic profiles. First, we include a continuous measure of the mothers’ highest level
of formal schooling. Second, we make use of the DHS-constructed wealth index to account
for inequality in material resources. The DHS collects information on household’s ownership
of various assets (e.g., radio, television, refrigerator, bicycle, car) and about the
characteristics of the house (e.g., availability of electricity, source of drinking water, type of
toilet facility, number of rooms) and aggregates these variables into a principal component
factor analysis. The factor scores are used to categorize households into the five quintiles that
we make use of here: poorest, poor, middle, rich, or richest (see Bollen, Glanville and
Stecklov 2007; Houweling, Kunst and Mackenbach 2003; Filmer and Pritchett 1998 for more
detailed discussion).
Marital Violence
In order to assess whether selection into divorce is concentrated among high-conflict
marriage in low divorce settings, we make use of data on women’s experience with both
physical and sexual intimate-partner violence. The DHS recently incorporated a special
module into select country surveys to track women’s experience with violence. The DHS
Parental Divorce and Child Mortality in Sub-Saharan Africa: Does Context Matter? 13
asks women whether their current or— in the instance that she is divorced or widowed—
most recent partner ever (1) pushed, shook, threw something at them, (2) slapped, (3)
punched with fist or something hard, (4) kicked or dragged, (5) tried to strangle or burn, (6)
threatened with a knife/gun or other weapon, (7) ever attacked with a knife/gun or other
weapon, (8) physically forced sex when not wanted, or (9) ever forced sexual acts when not
wanted. From these nine items, we create a binary indicator of whether children’s mother
ever experienced physical or sexual violence (no=0, yes=1).5
These data afford a rare opportunity to compare the prevalence of physical and sexual
violence among women by marital status and marital context; however, the data are available
in only 17 of the 31 countries in our study: Burkina Faso, DRC, Cameroon, Gabon, Ghana,
Kenya, Ivory Coast, Liberia, Mali, Malawi, Nigeria, Rwanda, Sao Tome Principe, Tanzania,
Uganda, Zambia, and Zimbabwe. Narrowing our sample to these countries would
compromise the extent of our study’s representation of African families and subnational
regions. Thus, in our multivariate results we include a category representing “no data” (=2)
(not shown in tables) in order to maintain the full sample while exploring whether divorce is
more closely associated with violent relationships in low divorce settings, and if this selection
drives divorce to be associated with especially poor child health in these settings.
Child Controls
In all models, we include a set of demographic variables known to influence child
survival: maternal age at the time of the child’s birth, birth order, birth spacing (first born,
fewer than 36 months, 36 months or more), child’s sex, and whether the child is from a
singleton or multiple birth. To accommodate our modeling strategy, we restructured our data
to a time-based dataset in which observations refer to time (i.e., months) and each birth
contributes the number of months the child was observed. In addition to explicitly handling
time through our event-history modeling approach, all models control for the child’s age (0 to
11 months, 12 to 23 months, 24 to 35 months, and 36+ months).
Subnational Region-level Controls
Although little is known about the contextual correlates of divorce in Africa,
extrapolating from individual-level findings and from contextual research in the United
States, we anticipate that a subnational region’s prevalence of divorce will be correlated with
the socioeconomic profile of its residence. It is important to account for these factors to
5 See the World Health Organization Report for a similar approach of estimating physical and sexual violence, and for country comparisons ( http://apps.who.int/iris/bitstream/10665/77432/1/WHO_RHR_12.36_eng.pdf?ua=1).
Parental Divorce and Child Mortality in Sub-Saharan Africa: Does Context Matter? 14
ensure that they do not drive any observed difference in the child health consequences of
divorce by its prevalence. In general, more highly-educated African women are more likely
to divorce (DeRose, Garcia, Salazar, & Tarud, 2014; Takyi & Broughton, 2006), which
means that divorce may be concentrated in settings with the highest female education.
Because the children living among more highly-educated women are known to have lower
mortality (Kravdal, 2004), we include an aggregate measure of the highest level of education
of the average woman in each subnational region. In addition to female education, extending
evidence from the United States that the extent of female employment is highly correlated
with the prevalence of divorce (Ruggles, 1997), we include a continuous measure of the
prevalence of women in the subnational region who have been employed in the prior year.
Third, because of evidence that urbanization is also associated with higher divorce rates
(Clark & Brauner-Otto, 2015), and because of known disparities in child mortality in rural
versus urban settings (Van de Poel, O'donnell, & Van Doorslaer, 2009), we use the DHS’s
classification of households as “rural” versus “urban” to generate a continuous measure of the
percentage of households that are classified as rural in each subnational region.
Country-level Controls
A country’s political, cultural, and economic climate is likely associated with both the
level of divorce and child mortality across its subnational regions. To address the possible
confounding nature of these factors in our multinational study design, our multilevel discrete-
time hazard models conservatively account for country-level factors by taking a fixed-effects
approach that includes a set of dummy variables representing each of the 31 countries in our
sample. This modeling strategy enables us to conservatively account for constant, unobserved
country-level factors that may confound the associations of interest. See Smith-Greenaway
and Trinitapoli 2014 and Liefbroer and Dourleijn 2006 for examples of multinational studies
that take comparable approaches.
ANALYTIC PLAN
We begin the results with descriptive statistics in order to characterize our sample.
We then provide subnational region-level estimates of the prevalence of divorce to highlight
its variability across the 290 regions in our sample. Next, because we are interested in the
distinct selection patterns of women into divorce in low versus high divorce settings, we
show bivariate analyses that compare divorced and married mothers (1) socioeconomic
resources (education, household wealth) and (2) experience with intimate partner violence
(physical or sexual violence) by the prevalence of divorce in their subnational region. For the
Parental Divorce and Child Mortality in Sub-Saharan Africa: Does Context Matter? 15
purposes of these analyses, we categorize our continuous measure of the prevalence of
divorce in each subnational region according to whether the level of divorce falls within one
standard deviation (3.74 percent) below, around, or above the mean value of 6.37 percent.
Next, we estimate a series of multilevel discrete-time hazard models to test whether
the mortality risk of having a divorced mother varies in contexts where divorce is rare versus
those where it is more prevalent. This modeling strategy allows us to handle two complexities
of our data: censoring and its hierarchical structure. In terms of the former, our hazard
modeling approach addresses the fact that children who are still alive at the time of the
observation period (i.e., at the survey), or who have not yet reached their 5th birthday, are
right-censored. The multilevel nature of our models appropriately estimates standard errors to
address the fact that children (level-1) are nested within subnational regions (level-2), which
are in turn nested within countries (level-3).6
We present a series of four multilevel hazard models:
(1) logit(htijk) = αt + ß1FamStructureijk + ß2Xijk + ß3Ck + uj (2) logit(htijk) = αt + ß1FamStructureijk + ß2PrevDivorcejk + ß3SocioeconomicContextjk +
ß4Xijk + ß5Ck + uj (3) logit(htijk) = αt + ß1FamStructureijk + ß2(PrevDivorcejk*FamStructureijk) +
ß3PrevDivorcejk + ß4SocioeconomicContextjk + ß5Xijk + ß6Ck + uj (4) logit(htijk) = αt + ß1FamStructureijk + ß2FamilySocioeconomicijk +
ß3IntimatePartnerViolenceijk + ß4(PrevDivorcejk*FamStructureijk) + ß5PrevDivorcejk + ß6SocioeconomicContextjk + ß7Xijk + ß8Ck + uj
We begin with a base model (1) that aims to replicate the association between
parental divorce and child mortality that has been documented in previous research on
specific countries (Clark & Hamplová, 2013; Thiombiano et al., 2013). In this model, and
all subsequent models, we include controls for child i in subnational region j in country k;
where Xijk is a vector of standard demographic controls that are known to influence child
survival; Ck are country dummy variables; uj is the subnational region-level random effect;
and the ßs represent the corresponding coefficients. In Model (2) we include the subnational-
region level indicator of the prevalence of currently divorced women, as well as indicators of
the socioeconomic context (average level of women’s education, women’s employment, and
urbanization). In model (3) we interact the prevalence of divorce at the subnational region-
level with family structure to test our competing hypotheses of whether the child mortality 6 We use the clustering feature in Stata to account for the fact that some children share the same mothers in our sample. In supplementary analyses, we randomly sampled one child per mother among women with multiple children in the dataset and re-estimated each model. Results from these analyses confirm that the findings are consistent with those shown here, suggesting the presence of siblings is not influencing the findings.
Parental Divorce and Child Mortality in Sub-Saharan Africa: Does Context Matter? 16
risk associated with divorce is either muted (hypothesis 1) or exacerbated (hypothesis 2a/2b)
in low divorce settings. In the final model (4), we include measures of both women’s
socioeconomic status (education and household wealth) and marital conflict (physical or
sexual violence) in order to test whether the differential selection into divorce across low
versus high divorce settings drives children to have distinct mortality risks. We also graph
Kaplan-Meier hazard functions in order to depict the study’s central findings.
RESULTS
Descriptive Findings
Table 1 provides descriptive statistics to characterize the sample of children and the
subnational regions in which they live. As shown, over nine percent of the children in our
sample have died either before their fifth birthday or the time of the survey (whichever came
first). The results also highlight that African children experience diverse family structures.
Although the vast majority of children (63%) have mothers who have been continuously
married (monogamously) since their birth, approximately one-fourth have polygynously
married mothers, nearly five percent have divorced mothers,7 and small percentages have
widowed mothers or mothers who have remain unmarried since their birth (1.7% and 7.1%
respectively). Turning to the subnational regional prevalence of divorce, the results show that
the average child in our sample lives in a context where more than six percent of mothers are
currently divorced. However, as shown by the standard deviation, the prevalence of divorce
varies considerable across subnational regions—motivating our investigation of the potential
cross-contextual variability in its implications for children.
The descriptive results further show that nearly 20 percent of women report having
been the victim of physical or sexual violence perpetrated by either their current or previous
partner (among women in the 17 countries where data on violence were collected). In
general, children’s mothers have fewer than three years of school. Furthermore, the sample
varies according to key proximate determinants of mortality, making it important to account
for these factors. The subnational estimates align with the characteristics of the mothers in
our sample. On average, nearly 70% of households in the average subnational region are
rural, about 65% of women living in these areas worked in the prior year, and the average
education is only a bit over 4 years; however, the large standard deviations point to the
diverse socioeconomic contexts across the subnational regions.
7 It is important to bear in mind that our estimates of divorce pertain to children whose mothers are currently divorced and have not remarried, and thus are lower than the prevalence of children whose mothers have ever divorced.
Parental Divorce and Child Mortality in Sub-Saharan Africa: Does Context Matter? 17
Variation in Prevalence of Divorce across Africa’s Subnational Regions
To provide a better sense of the variation in the prevalence of divorce across the 290
subnational regions in our sample, and how it varies both between and within the 31
countries in our study, Figure 1 plots each of the 290 subnational regions according to its
prevalence of divorce, by country.8 For reference, the black line indicates the mean
prevalence of divorce for the entire sample and the gray lines show one standard deviation
below/above the mean. The figure demonstrates two important findings. First, there are clear
country and regional patterns in the level of divorce across subnational regions. In general,
subnational regions in West African countries have some of the lowest levels of divorce (e.g.,
Burkina Faso, Guinea, Mali, Benin, Nigeria, Niger) whereas subnational regions in Southern
African countries tend to have notably higher levels (e.g., Mozambique, Malawi, Zambia).
Despite these overall trends, there are important exceptions. For example, Liberia, a West
African country, has generally high levels of divorce across its subnational regions whereas
some regions in Namibia, a Southern African country, have exceptionally low levels of
divorce.
Aside from patterns across countries, the figure shows remarkable within-country
variation in the prevalence of currently divorced women. For instance, in several countries—
Ethiopia, Madagascar, Mozambique, Tanzania, and Uganda—the prevalence of currently
divorced women varies by approximately a ten percentage point difference across subnational
regions. In total, we capture a nearly 20-point percentage difference in the prevalence of
divorce—a range comparable to the change in divorce rates that unfolded in the United States
over an entire century (Cruz, 2013).
Differential Selection into Divorce in Low versus High Divorce Settings
As prior research suggests, it is possible that women who divorce in low divorce
settings have distinct socioeconomic and marital profiles than those who divorce in high
divorce areas. In Table 2, we first explore the possibility that divorce is concentrated among
socioeconomically advantaged women in settings where it is rare. As shown, we find
evidence that it is: divorced women in low divorce settings have about one more year of
education compared to their married peers, but the difference is only 0.18 years in high
divorce settings. Furthermore, divorced women have a 0.21 advantage in their wealth ranking
in low divorce settings, but are slightly disadvantaged in settings where divorce is
8 Appendix C provides the full list of countries and its corresponding number of regions, and the average, standard deviation, minimum, and maximum values of divorce across subnational regions.
Parental Divorce and Child Mortality in Sub-Saharan Africa: Does Context Matter? 18
widespread. These descriptive statistics provide preliminary support for hypothesis 1, that the
pronounced socioeconomic advantage among divorced women where it is rare could
dampen—or even reverse—its negative consequences for children.
Despite socioeconomic advantage, however, the results also provide evidence of
pronounced negative selection in terms of marital conflict. The first column demonstrates
that, in general, divorced women are more likely to experience physical or sexual violence
compared to their married peers. But the subsequent columns confirm that the discrepancy in
exposure to violence is much larger in low divorce settings. In low divorce settings, 15
percent more divorced women report violence than do married women. In contrast, the size
of the disparity is halved in low divorce settings, with the prevalence of reported violence
only 7.8 percentage points higher among divorced mothers. This aligns with the possibility
that children with divorced parents may actually have the worst health outcomes where
divorce is rare, and that their greater exposure to violent environments could be why
(hypothesis 2a).
Cross-contextual Variation in Association between Parental Divorce and Child
Mortality
Table 3 shows estimates from multilevel discrete-time hazard models analyzing the
interrelationships between the prevalence of divorce, family structure, and children’s
likelihood of dying before age five. Corroborating previous research, the results in Model 1
demonstrate a strong association between parental divorce and children’s risk of death
(before age five) in sub-Saharan Africa. A child whose mother is divorced has 37 percent
higher odds of dying compared to a peer child whose mother is monogamously married. The
results further show children’s whose mothers are single either because they have never
married or are currently widowed have similarly high risk of mortality, experiencing a 22 and
46 percent higher odds of mortality, respectively. The results further confirm that children
whose mothers are in a polygynous union experience a higher likelihood of death compared
to their peers whose mothers are in a monogamous union.
In addition to distinct mortality risk by family structure, the results show that several
of the known proximate determinants of mortality are associated with children’s risk of death
as anticipated. Most notably, the results show that a lengthier birth interval preceding the
child’s birth lowers their risk of mortality and that female children experience lower mortality
than males. Furthermore, the results show that children of multiples (rather than from
singleton births) experience substantially higher mortality (odds ratio: 3.81).
Parental Divorce and Child Mortality in Sub-Saharan Africa: Does Context Matter? 19
To characterize the subnational regions in which children live, model 2 includes an
indicator for the prevalence of currently divorced women as well as socioeconomic indicators
that are likely to be correlated with both the level of divorce and child mortality. The results
show that the prevalence of currently divorced women in the subnational region is not
directly associated with child mortality, suggesting that for the average child, there is no
direct disadvantage associated with living among greater family instability. Moreover, as we
would expect, living in a subnational region wherein the average woman has more education
is associated with lower mortality risk: a one unit increase in the average schooling for
women is associated with a six percent lower odd of child death. Interestingly, the results
show that children living in settings wherein more women are employed have higher
mortality. Although this merits additional analysis, because most female employment
consists of agricultural labor in rural settings, it is likely that this association is in part a
reflection of the general lack of infrastructural provisions in settings wherein most women are
engaged in agricultural work. Net of these indicators of socioeconomic status, we find no
significant differences in the risk of mortality by level of urbanization.
Model 3 includes a cross-level interaction between the prevalence of divorce in the
subnational region with the child’s individual family structure to test our study’s central
hypotheses: whether (and if so, why) the magnitude of the mortality risk associated with
divorce varies systematically by its prevalence. The significant, positive interaction term
refutes Hypothesis 1 and instead confirms that the child health disadvantage associated with
divorce is significantly smaller in settings where divorce is more prevalent and thus larger in
settings where divorce is rare. In fact, a 10 percent higher prevalence of divorce corresponds
with a 10 percent decrease in the likelihood of child mortality for children with divorced
mothers (compared to children with married mothers). Furthermore the magnitude of the
odds ratio for having a divorced mother increases (from 1.38 in model 2 to 1.53 in model 3),
suggesting that failure to account for these cross-level interactions underestimates the true
size of the child mortality risk associated with divorce. Put differently, although the odds of
dying are 50 percent higher for children with divorced parents, this negative effect of divorce
is at least partially offset if the child is born in an area were divorce is common. Figure 2
illustrates this finding graphically with Kaplan-Meier survival estimates (holing all covariates
in model 3 at the mean value). In addition to differences in the size of the magnitude of the
disadvantage associated with having a divorced mother versus a married one, the Kaplan-
Meier analysis also confirms that children with divorced mothers where divorce is rare have
significantly higher mortality compared to their peers with divorced mothers in settings
where divorce is more common (log rank test, p<.001).
Parental Divorce and Child Mortality in Sub-Saharan Africa: Does Context Matter? 20
Why is the child health disadvantage associated with divorce significantly greater in
settings where divorce is rare? Given that divorce is more concentrated among violent
marriages in these sub-national regions (Table 2), it is important to assess whether the
elevated mortality risk associated with divorce in low divorce settings is merely a function of
this pronounced negative selection (Hypothesis 2a). In Model 4 we include an indicator for
mothers’ exposure to physical and/or sexual violence as well as mother’s education and
wealth. As shown, we find no support that negative selection is driving the finding: the cross-
level interaction remains significant net of selection in terms of marital violence. Although
the results show that children’s risk of mortality is increased by three percent if their mother
has experienced physical or sexual violence, accounting for exposure to violence, as well as
the socioeconomic factors, actually increases the size of the cross-level interaction: each unit
increase in the prevalence of divorce corresponds with a two percentage point reduction in
the risk associated with divorce.
The fact that the subnational regional effect of divorce persists even after controlling
for selection lends indirect support to our theoretical argument that the greater social and
psychological cost of divorce in low divorce settings drive the finding (hypothesis 2b). In
particular, in settings were divorce is uncommon, divorced mothers are likely to experience
social isolation and significant psychological strain which likely compromises their health-
seeking behaviors, children’s receipt of care, and ultimately their children’s health and
survival.
Before discussing the results, another finding in Table 3 merits further explanation:
the child mortality risk associated with having a widowed mother also varies by the
prevalence of divorce. Why do African children with widowed mothers have a higher
likelihood of dying in settings where divorce is more prevalent? Our supplemental
investigation suggests it is due to the confounding effect of HIV/AIDS prevalence and
concentration of AIDS-related widowhood in more urbanized, socioeconomically advantaged
settings (which also have higher levels of divorce). Because HIV/AIDS prevalence data are
not available for all countries/subnational regions, in supplementary analyses (not shown) we
attempt to reduce its confounding effect by excluding countries where HIV/AIDS prevalence
exceeds 10 percent in at least one subnational region (Cameroon, Namibia, Kenya, Malawi,
Mozambique, Swaziland, Tanzania, Zambia, and Zimbabwe). In these analyses, the
coefficient on the cross-level interaction between widowhood and divorce prevalence
becomes non-significant.
Parental Divorce and Child Mortality in Sub-Saharan Africa: Does Context Matter? 21
DISCUSSION
In this paper, we extend the literature on parental divorce and child wellbeing in
Africa in three unique ways. First, we leverage data from 290 subnational regions in Africa to
highlight the variability in the prevalence of currently divorced women, and as a result, the
distinct conditions under which African families experience divorce. Corroborating emerging
research documenting the variability in the level of divorce across African countries (Clark
and Brauner-Otto 2015), we find important differences in the level of divorce across
subnational regions. In general, subnational regions in West African countries have the
lowest levels of divorce whereas those in Southern Africa have markedly higher levels of
divorce. However, we find important within-country variation in the prevalence of currently
divorced women across subnational regions. For instance, in several countries—Ethiopia,
Madagascar, Mozambique, Tanzania, and Uganda—the prevalence of divorced women varies
by approximately a ten percentage point difference between subnational regions. In total, we
capture a nearly 20-point percentage difference in the number of women who are currently
divorced—a range comparable to the change in divorce rates that unfolded in the United
States over an entire century (Cruz, 2013).
Second, building on recent evidence that the prevalence of a family form can
powerfully moderate its consequences for individuals and their children (Amato & Keith,
1991; Kalmijn, 2009; Liefbroer & Dourleijn, 2006; Pong et al., 2003; Smith-Greenaway &
Trinitapoli, 2014), we show that context matters a great deal for understanding the child
health consequences of divorce in sub-Saharan Africa. Despite the more pronounced positive
selection of the socioeconomically advantaged into divorce in settings where it is rare,
children with divorced parents in these settings experience exceptionally high mortality risk.
Although their peers with divorced parents in settings where it is more common are still at
higher risk of mortality compared to their peers with married mothers, a one percentage
increase in the prevalence of divorce corresponds with an approximately two percentage
point reduction in the size of the divorce penalty. Although the size of the association is
small, multiplying the coefficient by the nearly 20 percentage point difference in the
prevalence of divorce across subnational regions corresponds to a 40 percent difference in the
odds of child mortality associated with divorce.
Third, building on past research, we test whether the pronounced selection of high-
conflict, violent marriages into divorce in settings where it is rare drives the additional
mortality risk associated with divorce in these settings (hypothesis 2a). Although we find
evidence that divorce is, in fact, more tightly connected to violence in low divorce settings,
this does not explain the observed mortality differentials. Therefore, we contend that at least
Parental Divorce and Child Mortality in Sub-Saharan Africa: Does Context Matter? 22
part of the explanation lies in the distinctly harsh experience of being a divorced mother in
settings where divorce is uncommon. Building on evidence that divorcées —particularly
women—experience intense stigma and social penalties in cultural context where divorce is
rare and socially unacceptable, we argue that divorced women’s social isolation,
psychological challenges, and general lack of support are what drive their children’s
especially poor health in these settings.
The study findings confirm the importance of ensuring that children with divorced
parents have equal access to health, nutrition, and educational services, especially in
resource-scarce settings like Africa. However, our results suggest that doing so successfully,
at least in settings where divorce is rare, will require community-wide initiatives to
destigmatize divorce and reach divorced women who, despite their socioeconomic advantage,
face steep social and psychological barriers to good health. More generally, the study adds to
the growing evidence that it is imperative to consider the broader social, cultural, marital, and
socioeconomic context in which families live order to develop a more nuanced understanding
of how family processes like divorce influence adults and their children.
Parental Divorce and Child Mortality in Sub-Saharan Africa: Does Context Matter? 23
Appendix A. Sample characteristics of children excluded because mother has been in more than one union *
Mean/% (SD) Key variables Child mortality (% ever died) 14.12*
Family structure Divorced mother 9.08* Never married mother NA Widowed mother 2.39* Polygynously married mother 33.73* Monogamously married (ref) 54.8*
Subnational region-level prevalence of divorced women (%) 6.92 (4.15)*
Selection indicators Socioeconomic resources Mother's years of formal schooling 2.78 (3.42)* Household wealth
Poorest 28.50* Poor 23.60* Average 20.00 Rich 16.86* Richest 11.04*
Mother experienced violence1 19.96*
Child-level controls Female 49.50 Multiple birth 3.87 Maternal age at birth 27.92 (6.95)* Birth order 4.27 (2.46)* Length of preceding birth interval
First born 12.38* Short interval 47.97* Long interval 39.65*
Subnational region-level controls Average years of schooling among adult women 4.02 (2.28)* Adult women's employment (%) 66.43 (16.43)* Rural (% households rural) 71.75 (21.31)* N=94,209 children *Indicates that the value is significantly different from value in Table 1 1Among subsample of countries with intimate-partner violence module
Parental Divorce and Child Mortality in Sub-Saharan Africa: Does Context Matter? 24
Appendix B. List of Countries, Survey Year, and Sample Size Survey year N
Benin 2011/12 17,187 Burkina Faso 2010 20,868 Burundi 2010/2011 10,230 Cameroon 2011 13,663 Chad 2004 7,177 Congo (Brazzaville) 2011/12 10,314 Democratic Republic of the Congo 2007 10,683 Ethiopia 2003 14,846 Gabon 2012 6,602 Ghana 2008 3,560 Guinea 2012 9,680 Ivory Coast 2011/12 9,508 Kenya 2008/09 8,308 Lesotho 2009/10 5,783 Liberia 2006/07 6,165 Madagascar 2008/09 13,901 Malawi 2010 23,386 Mali 2006 18,931 Mozambique 2003 12,220 Namibia 2006/07 6,868 Niger 2012 16,064 Nigeria 2008 37,659 Rwanda 2010/11 11,873 Sao Tome Principe 2008/09 1,725 Sierra Leone 2008 6,934 Senegal 2010/11 15,651 Swaziland 2006/07 3,992 Tanzania 2009/10 9,284 Uganda 2011 9,836 Zambia 2007 7,763 Zimbabwe 2010/11 6,813
N 357,474 Source: Demographic and Health Survey
Parental Divorce and Child Mortality in Sub-Saharan Africa: Does Context Matter? 25
Appendix C. Variation in Prevalence of Divorce within and across African Countries Average % divorced Number women of across Regions regions SD Min. Max.
Benin 12 3.40 1.59 0.98 5.75 Burkina Faso 13 1.41 0.89 0.17 3.05 Burundi 5 4.64 1.07 3.57 6.30 Cameroon 12 5.69 1.91 3.31 9.84 Chad 9 6.42 2.22 4.11 10.93 Congo (Brazzaville) 12 12.51 1.97 10.07 16.01 Democratic Republic of the Congo 11 7.20 1.97 4.50 11.52 Ethiopia 11 7.96 2.75 3.93 13.02 Gabon 10 9.42 2.48 4.55 11.71 Ghana 10 6.21 2.67 2.41 9.43 Guinea 8 2.08 1.08 0.49 4.10 Ivory Coast 11 4.19 1.99 1.41 8.57 Kenya 8 6.10 1.53 3.50 8.74 Lesotho 10 4.43 1.17 2.62 6.28 Liberia 6 7.20 1.33 5.31 8.50 Madagascar 6 11.72 3.09 6.69 15.10 Malawi 3 9.14 1.30 8.07 10.59 Mali 9 2.56 1.89 0.58 5.88 Mozambique 11 13.31 3.10 9.42 18.31 Namibia 13 4.49 2.24 1.36 9.00 Niger 8 3.18 2.05 1.34 7.18 Nigeria 6 1.96 0.71 1.31 3.23 Rwanda 5 5.41 1.18 4.07 7.12 Sao Tome Principe 4 11.78 2.12 9.90 14.59 Senegal 14 3.28 1.16 1.98 5.90 Sierra Leone 4 3.70 0.70 2.95 4.64 Swaziland 4 3.15 0.43 2.57 3.66 Tanzania 26 8.51 2.99 4.86 17.69 Uganda 10 8.92 3.23 2.12 12.65 Zambia 9 8.13 1.52 6.69 10.94 Zimbabwe 10 7.33 0.88 6.12 8.95
Full Sample 290 6.37 3.74 0.17 18.31 Source: Demographic and Health Survey
Parental Divorce and Child Mortality in Sub-Saharan Africa: Does Context Matter? 26
References Adams, A. M., Madhavan, S., & Simon, D. (2002). Women's social networks and child survival in
Mali. Social science & medicine, 54(2), 165-178. Al-Krenawi, A., & Graham, J. R. (1998). Divorce among Muslim Arab women in Israel. Journal of
Divorce & Remarriage, 29(3-4), 103-119. Amato, P. R. (2010). Research on divorce: Continuing trends and new developments. Journal of
Marriage and Family, 72(3), 650-666. Amato, P. R., & Keith, B. (1991). Parental divorce and adult well-being: A meta-analysis. Journal of
Marriage and the Family, 43-58. Åsling-Monemi, K., Pena, R., Ellsberg, M. C., & Persson, L. Å. (2003). Violence against women
increases the risk of infant and child mortality: a case-referent study in Nicaragua. Bulletin of the World Health Organization, 81(1), 10-16.
Bernardi, F., & Martínez-Pastor, J.-I. (2011). Divorce risk factors and their variation over time in Spain. Demographic Research, 24(31), 771-800.
Boyle, M. H., Racine, Y., Georgiades, K., Snelling, D., Hong, S., Omariba, W., . . . Rao-Melacini, P. (2006). The influence of economic development level, household wealth and maternal education on child health in the developing world. Social science & medicine, 63(8), 2242-2254.
Caldwell, J. C., Caldwell, P., & Quiggin, P. (1989). The social context of AIDS in sub-Saharan Africa. Population and development review, 185-234.
Chae, S. (2013). Divorce, Remarriage, and Children’s Outcomes in Rural Malawi. Demographic and Institutional Change in Global Families, 28-30.
Chen, W.-C. (2012). The Changing Pattern of Educational Differentials in Divorce in the Context of Gender Egalitarianization: The Case of Taiwan. Population Research and Policy Review, 31(6), 831-853.
Cherlin, A. J. (2004). The deinstitutionalization of American marriage. Journal of Marriage and Family, 66(4), 848-861.
Clark, S., Beguy, D., & Cotton, C. (2013). Single Motherhood and Child Mortality: Is Poverty the Link Paper presented at the International Union for the Scientific Study of Population, XXVII International Population Conference Busan, Korea.
Clark, S., & Brauner-Otto, S. (2015). Divorce in sub-Saharan Africa: Are Unions Becoming More Stable? . Paper presented at the Population Association of America Annual Meeting, San Diego, CA.
Clark, S., & Hamplová, D. (2013). Single motherhood and child mortality in sub-Saharan Africa: A life course perspective. Demography, 50(5), 1521-1549.
Cruz, J. (2013). Marriage: More than a Century of Change (pp. 2): National Center for Family and Marriage Research
De Graaf, P. M., & Kalmijn, M. (2006). Divorce Motives in a Period of Rising Divorce Evidence From a Dutch Life-History Survey. Journal of Family Issues, 27(4), 483-505.
DeRose, L., Garcia, P. C., Salazar, A., & Tarud, C. (2014). Household Structure and School Attendance in 67 Countries: Why Children with Absent Fathers Do Better in Some Places. Paper presented at the Population Association of America Annual Meeting, Boston, MA.
DiFonzo, J. H. (1997). Beneath the fault line: The popular and legal culture of divorce in twentieth-century America: University of Virginia Press.
Engle, P. L., Black, M. M., Behrman, J. R., Cabral de Mello, M., Gertler, P. J., Kapiriri, L., Young, M. E. (2007). Strategies to avoid the loss of developmental potential in more than 200 million children in the developing world. The Lancet, 369(9557), 229-242.
Filmer, D., & Pritchett, L. (1999). The impact of public spending on health: does money matter? Social science & medicine, 49(10), 1309-1323.
Parental Divorce and Child Mortality in Sub-Saharan Africa: Does Context Matter? 27
Gage, A. J. (1997). Familial and socioeconomic influences on children's well-being: an examination of preschool children in Kenya. Social science & medicine, 45(12), 1811-1828.
Garcia-Moreno, C., Jansen, H. A., Ellsberg, M., Heise, L., & Watts, C. H. (2006). Prevalence of intimate partner violence: findings from the WHO multi-country study on women's health and domestic violence. The Lancet, 368(9543), 1260-1269.
Gerstel, N. (1987). Divorce and stigma. Social problems, 172-186. González, L., & Viitanen, T. K. (2009). The effect of divorce laws on divorce rates in Europe.
European Economic Review, 53(2), 127-138. Goode, W. J. (1963). World revolution and family patterns. Grover, S. (2011). ‘Purani aur nai shaadi:’Separation, Divorce, and Remarriage in the Lives of the
Urban Poor in New Delhi. Asian Journal of Women's Studies, 17(1), 67-99. Houweling, T. A., Kunst, A. E., Looman, C. W., & Mackenbach, J. P. (2005). Determinants of under-
5 mortality among the poor and the rich: a cross-national analysis of 43 developing countries. International journal of epidemiology, 34(6), 1257-1265.
Jewkes, R., Levin, J., & Penn-Kekana, L. (2002). Risk factors for domestic violence: findings from a South African cross-sectional study. Social science & medicine, 55(9), 1603-1617.
Kaler, A. (2001). “Many Divorces and Many Spinsters”: Marriage as an Invented Tradition in Southern Malawi, 1946-1999. Journal of Family History, 26(4), 529-556.
Kalmijn, M. (2009). Country differences in the effects of divorce on well-being: The role of norms, support, and selectivity. European Sociological Review, jcp035.
Kalmijn, M., & Uunk, W. (2007). Regional value differences in Europe and the social consequences of divorce: A test of the stigmatization hypothesis. Social Science Research, 36(2), 447-468.
Koenig, M. A., Lutalo, T., Zhao, F., Nalugoda, F., Wabwire-Mangen, F., Kiwanuka, N., Gray, R. (2003). Domestic violence in rural Uganda: evidence from a community-based study. Bulletin of the World Health Organization, 81(1), 53-60.
Kravdal, Ø. (2004). Child mortality in India: the community-level effect of education. Population studies, 58(2), 177-192.
Liefbroer, A. C., & Dourleijn, E. (2006). Unmarried cohabitation and union stability: Testing the role of diffusion using data from 16 European countries. Demography, 43(2), 203-221.
Martin, M. A. (2006). Family structure and income inequality in families with children, 1976 to 2000. Demography, 43(3), 421-445.
Omariba, D., & Boyle, M. H. (2007). Family structure and child mortality in Sub‐Saharan Africa: cross‐national effects of polygyny. Journal of Marriage and Family, 69(2), 528-543.
Park, H., & Raymo, J. M. (2013). Divorce in Korea: Trends and educational differentials. Journal of Marriage and Family, 75(1), 110-126.
Patel, V., Rahman, A., Jacob, K., & Hughes, M. (2004). Effect of maternal mental health on infant growth in low income countries: new evidence from South Asia. Bmj, 328(7443), 820-823.
Pong, S. l., Dronkers, J., & Hampden‐Thompson, G. (2003). Family policies and children's school achievement in single‐versus two‐parent families. Journal of Marriage and Family, 65(3), 681-699.
Porter, L., Hao, L., Bishai, D., Serwadda, D., Wawer, M. J., Lutalo, T., & Gray, R. (2004). HIV status and union dissolution in sub-Saharan Africa: the case of Rakai, Uganda. Demography, 41(3), 465-482.
Raymo, J. M., Iwasawa, M., & Bumpass, L. (2004). Marital dissolution in Japan: Recent trends and patterns. Demographic Research, 11(14), 395-420.
Reniers, G. (2003). Divorce and remarriage in rural Malawi. Reniers, G. (2008). Marital strategies for regulating exposure to HIV. Demography, 45(2), 417-438. Ruggles, S. (1997). The rise of divorce and separation in the United States, 1880–1990. Demography,
34(4), 455-466.
Parental Divorce and Child Mortality in Sub-Saharan Africa: Does Context Matter? 28
Smith-Greenaway, E., & Heckert, J. (2013). Does the Orphan Disadvantage “Spill Over?” An analysis of whether living in an area with a higher concentration of orphans is associated with children’s school enrollment in sub-Saharan Africa. Demographic research, 28, 1167.
Smith-Greenaway, E., & Trinitapoli, J. (2014). Polygynous Contexts, Family Structure, and Infant Mortality in Sub-Saharan Africa. Demography, 51(2), 341-366.
Smith, D. J. (2009). Managing men, marriage, and modern love: Women's perspectives on intimacy and male infidelity in southeastern Nigeria. Love in Africa, 157-180.
Solivetti, L. M. (1994). Family, marriage and divorce in a Hausa community: A sociological model. Africa, 64(02), 252-271.
Spanier, G. B., & Thompson, L. (1987). Parting: The aftermath of separation and divorce: Sage publications.
Takyi, B. K., & Broughton, C. L. (2006). Marital stability in sub-Saharan Africa: Do women’s autonomy and socioeconomic situation matter? Journal of family and economic issues, 27(1), 113-132.
Therborn, G. (2004). African families in a global context (Vol. 131): Nordic Africa Institute. Thiombiano, B. G., LeGrand, T. K., & Kobiané, J.-F. (2013). Effects of parental union dissolution on child mortality and schooling in Burkina Faso. Demographic Research 29(29), 797-816.
Thomson, E., Hanson, T. L., & McLanahan, S. S. (1994). Family structure and child well-being: Economic resources vs. parental behaviors. Social Forces, 73(1), 221-242.
Thornton, A. (1985). Changing attitudes toward separation and divorce: Causes and consequences. American Journal of Sociology, 856-872.
Thornton, A., & Young‐DeMarco, L. (2001). Four decades of trends in attitudes toward family issues in the United States: The 1960s through the 1990s. Journal of marriage and family, 63(4), 1009-1037.
Trinitapoli, J. (2011). The AIDS-related activities of religious leaders in Malawi. Global public health, 6(1), 41-55.
Van de Poel, E., O'donnell, O., & Van Doorslaer, E. (2009). What explains the rural-urban gap in infant mortality: household or community characteristics? Demography, 46(4), 827-850.
White, L. K., & Booth, A. (1991). Divorce Over the Life Course The Role of Marital Happiness. Journal of Family Issues, 12(1), 5-21.
Yang, L. H., Kleinman, A., Link, B. G., Phelan, J. C., Lee, S., & Good, B. (2007). Culture and stigma: adding moral experience to stigma theory. Social science & medicine, 64(7), 1524-1535.
Parental Divorce and Child Mortality in Sub-Saharan Africa: Does Context Matter? 29 Table 1. Descriptive Statistics Mean/% Key variables Child mortality (% ever died) 9.41
Family structure Divorced mother 4.39 Widowed mother 1.73 Never married mother 7.07 Monogamously married (ref) 63.07 Polygynously married mother 23.74
Subnational region-level prevalence of divorced women (%) 6.36 (3.75)
Selection indicators Socioeconomic resources Mother's years of formal schooling 3.49 (4.09) Household wealth
Poorest 24.43 Poor 21.15 Average 19.64 Rich 18.35 Richest 16.43
Marital conflict Mother experienced violence1 18.43 Controls Child-level Maternal age at birth 25.91 (6.76) Birth order 3.57 (2.43) Length of preceding birth interval
First born 23.28 Short interval 45.48 Long interval 31.24
Female 49.33 Multiple birth 3.32 Subnational region-level Average years of schooling among adult women 4.15 (2.54) Adult women's employment (%) 64.82 (17.51) Rural (% households rural) 69.89 (22.19) N=357,474 children 1Among subsample of countries with intimate-partner violence module
Parental Divorce and Child Mortality in Sub-Saharan Africa: Does Context Matter? 30 Table 2. Selection into divorce: Comparison of divorced versus married mothers among full sample and stratified by prevalence of divorce Full Sample Low Divorce Medium Divorce High Divorce
Divorced Married Divorced Married Divorced Married Divorced Married %/Mean %/Mean Difference %/Mean %/Mean Difference %/Mean %/Mean Difference %/Mean %/Mean Difference Socioeconomic Factors Mothers' Education 4.89 4.03 0.86 * 3.97 2.99 0.98 * 4.98 4.15 0.83 * 4.83 4.65 0.18 * Wealth index (1-5 scale) 2.98 2.97 0.01 3.14 2.93 0.21 * 3.04 3.02 0.02 * 2.77 2.85 -0.08 *
Marital Conflict Physical or sexual violence1 26.41 19.17 7.24 * 28.89 13.80 15.09 * 27.09 21.86 5.23 * 23.76 15.98 7.78 * Source: Demographic and Health Survey 1Among subsample of countries with intimate-partner violence module
Parental Divorce and Child Mortality in Sub-Saharan Africa: Does Context Matter? 31 Table 3. Multilevel discrete time hazard models analyzing whether the child mortality risk associated with divorce vary by its subnational regional prevalence in sub-Saharan Africa Model 1 Model 2 Model 3 Model 4 OR Coeff S.E. Sig OR Coeff S.E. Sig OR Coeff S.E. Sig OR Coeff S.E. Sig Key Variables Cross-level Interaction % Divorce * Divorced 0.99 -0.01 0.01 * 0.98 -0.02 0.01 * % Divorce * Widowed 1.03 0.03 0.01 * 1.03 0.03 0.01 * % Divorce * Never married 1.01 0.01 0.01 1.01 0.01 0.01 % Divorce * Married (ref) -- -- % Divorce *Polygynously married 0.99 -0.01 0.00 0.99 -0.01 0.00
Family structure Divorced Mother 1.37 0.32 0.03 *** 1.38 0.32 0.03 *** 1.52 0.42 0.06 *** 1.54 0.43 0.06 *** Widowed mother 1.46 0.38 0.04 *** 1.46 0.38 0.04 *** 1.25 0.23 0.07 ** 1.22 0.20 0.07 ** Never married mother 1.22 0.20 0.03 *** 1.22 0.20 0.03 *** 1.12 0.11 0.05 * 1.18 0.16 0.05 *** Monogamously married (ref) -- -- -- -- Polygynously married mother 1.17 0.16 0.01 *** 1.17 0.16 0.01 *** 1.19 0.18 0.02 *** 1.17 0.16 0.02 ***
Subnational region % divorce 0.99 -0.01 0.01 0.99 -0.02 0.01 0.98 -0.02 0.01
Selection indicators Socioeconomic resources Mother's years of schooling 0.97 -0.03 0.00 *** Household wealth
Poorest Poor 1.03 0.02 0.02 Average 0.96 -0.04 0.02 ** Rich 0.91 -0.10 0.02 *** Richest 0.78 -0.25 0.02 ***
Marital conflict Physical/sexual violence1 1.03 0.03 0.02
Controls Child-level Maternal age at birth 0.98 -0.02 0.00 *** 0.98 -0.02 0.00 *** 0.98 -0.02 0.00 *** 0.98 -0.02 0.00 *** Birth order 1.09 0.09 0.00 *** 1.09 0.09 0.00 *** 1.09 0.09 0.00 *** 1.08 0.07 0.00 *** Length of preceding birth interval
Parental Divorce and Child Mortality in Sub-Saharan Africa: Does Context Matter? 32
First born Short interval 0.91 -0.10 0.02 *** 0.91 -0.10 0.02 *** 0.91 -0.10 0.02 *** 0.89 -0.12 0.02 *** Long interval 0.55 -0.60 0.02 *** 0.55 -0.60 0.02 *** 0.55 -0.60 0.02 *** 0.54 -0.61 0.02 ***
Female 0.86 -0.16 0.01 *** 0.86 -0.16 0.01 *** 0.86 -0.16 0.01 *** 0.86 -0.16 0.01 *** Multiple birth 3.81 1.34 0.02 *** 3.82 1.34 0.02 *** 3.82 1.34 0.02 *** 3.86 1.35 0.02 *** Age dummies (not shown) Subnational region-level controls Average female schooling 0.94 -0.06 0.02 *** 0.94 -0.06 0.02 *** 0.98 -0.02 0.02 Adult women's employment (%) 1.65 0.50 0.13 *** 1.64 0.50 0.13 *** 1.62 0.48 0.14 *** Rural (% households rural) 1.05 0.05 0.10 1.05 0.05 0.10 0.92 -0.08 0.11 Country-level Country dummies (not shown)
Variance components Between subnational region variance 0.02 0.01 0.01 0.01
Model Fit Likelihood-ratio test 1146.49*** 752.82*** 755.03*** 768.30*** †p<.1 ; *p<.05 ; **p<.01 ; ***p<.001 Source: Demographic and Health Survey
Parental Divorce and Child Mortality in Sub-Saharan Africa: Does Context Matter? 33
Figure 1. Variation in the prevalence of currently divorced women across 290 subnational-regions in Africa, by country
Congo (Brazzaville) Sao Tome Principe
Mozambique Malawi Zambia
Madagascar Zimbabwe
Liberia Tanzania
Gabon DRC Chad
Rwanda Ethiopia Burundi
Kenya Cameroon
Sierra Leone Lesotho
Swaziland Ghana
Uganda Senegal
Ivory Coast Namibia
Niger Nigeria
Benin Mali
Guinea Burkina Faso
0 5 10 15 20
% Women Divorced in Subnational Region