Demography

, Volume 50, Issue 5, pp 1521–1549

Single Motherhood and Child Mortality in Sub-Saharan Africa: A Life Course Perspective

Authors

    • Department of SociologyMcGill University
  • Dana Hamplová
    • Charles University and Institute of Sociology, ASCR
Article

DOI: 10.1007/s13524-013-0220-6

Cite this article as:
Clark, S. & Hamplová, D. Demography (2013) 50: 1521. doi:10.1007/s13524-013-0220-6

Abstract

Single motherhood in sub-Saharan Africa has received surprisingly little attention, although it is widespread and has critical implications for children’s well-being. Using survival analysis techniques, we estimate the probability of becoming a single mother over women’s life course and investigate the relationship between single motherhood and child mortality in 11 countries in sub-Saharan Africa. Although a mere 5 % of women in Ethiopia have a premarital birth, one in three women in Liberia will become mothers before first marriage. Compared with children whose parents were married, children born to never-married single mothers were significantly more likely to die before age 5 in six countries (odds ratios range from 1.36 in Nigeria to 2.61 in Zimbabwe). In addition, up to 50 % of women will become single mothers as a consequence of divorce or widowhood. In nine countries, having a formerly married mother was associated with a significantly higher risk of dying (odds ratios range from 1.29 in Zambia to 1.75 in Kenya) relative to having married parents. Children of divorced women typically had the poorest outcomes. These results highlight the vulnerability of children with single mothers and suggest that policies aimed at supporting single mothers could help to further reduce child mortality in sub-Saharan Africa.

Keywords

Child mortalityFamily structuresSingle motherhoodSub-Saharan Africa

Introduction

Sub-Saharan Africa bears the highest rates of child mortality worldwide, suffering nearly twice the average mortality rate for developing countries and 18 times the rate of developed countries (United Nations 2011). The bulk of previous research has focused on access to health care, environmental conditions, poverty, and women’s education as key determinants of child health and survival (Gakidou et al. 2010). The past two decades, however, have seen a comparatively small but growing interest in the role that family structures play in protecting children’s health. Part of this interest has been triggered by high rates of internal migration and the AIDS crisis, which disrupt existing family structures by decreasing child and parent coresidence and increasing parental deaths. This line of research has typically highlighted the beneficial effects of having both parents alive and present in the household on children’s education, health, and survival, although some studies have found a positive effect of having a nonresidential migrant father (Sear et al. 2002; Townsend et al. 2002; Zaba et al. 2005). Other research has focused on traditional features of African family structures, such as polygyny and child fostering. These studies generally have found that children living in polygynous households are less likely to survive than children living in monogamous households (Gyimah 2009; Omariba and Boyle 2007) and that children closely related to the household head tend to fare better than children fostered by more-distant relatives (Case et al. 2004).

Interestingly, however, few studies have examined the effects of women’s marital status on children’s well-being in sub-Saharan Africa. This dearth of research contrasts sharply with the abundance of studies concentrating on North America and Europe (Amato 2000). In general, these studies have shown that children of single mothers fare worse with respect to cognitive development (Gennetian 2005; Kim 2011), educational outcomes (Steele et al. 2009), behavioral adjustment (Magnuson and Berger 2009), and health outcomes (Luo et al. 2004). However, the effects of single motherhood may be considerably more severe in poor countries, where families face greater resource constraints. One study in Bangladesh, for example, found that children of divorced parents suffered significantly higher rates of mortality (Bhuiya and Chowdhury 1997). Whether single motherhood has similarly negative implications for child survival in sub-Saharan Africa is largely unknown, although one study in Kenya found that children of never-married and formerly married women had higher rates of wasting and were less likely to receive a complete course of polio vaccination (Gage 1997).

Measures of Female Headship

One of the main reasons for this dearth of research on single motherhood is that measuring marital unions in sub-Saharan Africa is notoriously difficult. Marriages are often informal, and the process of union formation is typically elongated and may involve an incremental transfer of bridewealth. Furthermore, standard demographic and health surveys gather limited data on women’s marital histories, typically asking only their age at first marriage and current marital status. As a result, rather than examining single motherhood, studies in sub-Saharan Africa have tended to focus on female-headed households. Although these categories may overlap, female household heads and single mothers differ in important ways. Specifically, measures of female-headed households generally fail to capture what Buvinic and Gupta called “subfamilies” (1997), which are made up of single mothers and their children who live within male-headed households. In addition, in some countries, a large portion of female-headed households consist of married women with migrant husbands who send their families substantial remittances (Posel 2001; Villarreal and Shin 2008). Last, because headship is measured at the time of the survey, female headship overrepresents the “survivors,” given that the poorest female-headed households may simply collapse or be absorbed into male-headed households.

These measurement issues could help explain the surprisingly weak and inconsistent evidence linking female-headed households to poorer health outcomes for children (Hargreaves et al. 2004; Onyango et al. 1994). In fact, several studies even found a positive association between female-headed households and child health in developing countries (Adhikari and Podhisita 2010; Kennedy and Haddad 1994; Kennedy and Peters 1992). One explanation for these puzzling findings is that female-headed households are not always poorer, particularly if the women are married to migrant men (Chant 2008; Katapa 2006; Quisumbing et al. 2001; Villarreal and Shin 2008). Consequently, researchers have sharply questioned the continued use of measures of female-headed households in developing countries (Budlender 2003; Chant 2008; Dungumaro 2008; Posel 2001), and researchers in the United States have largely abandoned this measure in favor of indicators of single parenthood (London 1998).

Single Motherhood in Sub-Saharan Africa

Although there are no systematic estimates of the rates of single motherhood in sub-Saharan Africa, several relevant indicators suggest that single motherhood is probably widespread and quite common in many countries. The two main pathways into single motherhood are (1) giving birth before marriage, and (2) experiencing a union dissolution through divorce1 or widowhood and having at least one dependent child. In this article, we refer to these pathways as premarital and postmarital single motherhood, respectively, and define dependent children as those under the age of 15. Because of its effect on women’s education, premarital childbearing has attracted both research and policy attention in sub-Saharan Africa (Lloyd and Mensch 2008). The rates of premarital childbearing show considerable variation across sub-Saharan Africa (Harwood-Lejeune 2001), but in some areas with late age of first marriage, never-married women contribute nearly one-half of all births among women aged 12–26 (Garenne et al. 2000).

Although it has received less attention, marital instability is also common in some parts of sub-Saharan Africa, potentially resulting in a large number of postmarital single mothers. A recent analysis covering 13 countries across sub-Saharan Africa found that for seven of them, more than 20 % of women aged 14–49 had already experienced a union dissolution (de Walque and Kline 2012). However, because most surveys fail to record the dates when previous unions ended, lifetime risks of union dissolution are difficult to estimate, and statistical agencies do not report divorce rates for this region (United Nations 2010). Data from a handful of studies over the past 25 years, however, indicate that roughly one-half of all first marriages end in divorce in Ethiopia, Ghana, Togo, and Malawi (Bracher et al. 2003; Gage and Njogu 1994; Locoh and Thiriat 1995; Tilson and Larsen 2000). Rates of widowhood are similarly rarely published, although adult mortality rates are high, particularly in those countries hit hardest by the AIDS epidemic (Luginaah et al. 2005). This combination of high rates of premarital fertility, divorce, and widowhood suggest that many children in sub-Saharan Africa will spend at least a portion of their childhood being raised by a single mother.

Mechanisms Linking Single Motherhood and Child Mortality

Whether children of single mothers in Africa fare worse than those with married mothers is an open empirical question. Anthropologists have long noted that consanguineous ties are more important than conjugal ties in sub-Saharan Africa (Fortes 1958). Consequently, it is plausible that as long as children are recognized as belonging to a particular kinship lineage, the marital status of their parents is relatively unimportant (Gage 1997). However, the bonds between extended kin may be weakening, particularly in response to rapid urbanization, widespread migration, and the AIDS epidemic (Mtika 2001). If so, the negative association between single motherhood and child well-being found in Western societies might also apply in the African context. The extensive literature from Western cultures proposes three main mechanisms to explain why children of single mothers fare poorly: (1) lower economic status, (2) less parental supervision and care, and (3) selection effects into and out of single motherhood (Amato 2000). Variation across these three mechanisms may also help to explain whether some types of single motherhood (premarital vs. postmarital) are more detrimental in different cultural contexts and whether new marriages will improve child outcomes.

For example, financial hardship may be especially acute among widows (van de Walle 2011), whereas the economic consequences of single motherhood may be attenuated for never-married adolescent mothers who continue to live with their parents. Furthermore, although new marriages are likely to be associated with greater household wealth, fewer household resources may be directed toward children who are not biologically related to the new spouse (Bledsoe 1995). Parental care may also vary by women’s marital status. Children of unmarried mothers may receive less care not only from their fathers but also from their mothers. Children born out of wedlock are more likely to be fostered (Parr 1995), and children from previous unions may not follow their mother into her new household when she marries (Bledsoe 1995; Grant and Yeatman forthcoming). Single mothers may also be more likely to work outside the home and, hence, spend less time breast-feeding or directly supervising their coresidential children (Oya and Sender 2009).

Last, there is selection into and out of single motherhood. For example, while being in a polygynous union may directly affect child survival chances (Gyimah 2009; Omariba and Boyle 2007), unmarried mothers with young children may be more willing to enter into a polygynous union and women with co-wives may be more likely to get divorced. Perhaps most importantly, widowhood and child’s mortality may be correlated. In the case of HIV, children of HIV-positive mothers suffer from higher rates of mortality both because these children are more likely to be HIV-positive and because their mothers may be ill (Zaba et al. 2005). Moreover, women who are HIV-positive are more likely to be widowed and divorced (Reniers 2003).

In this article, we investigate the relationship between single motherhood and child survival in 11 countries in sub-Saharan Africa.2 We begin by estimating women’s lifetime risk of becoming a single mother through either a premarital birth or union dissolution. We then examine whether children of single mothers are more likely to die before the age of 5 compared with children whose parents are married. These analyses explore differences in child mortality by types of single motherhood, including mothers who are never married, divorced, widowed, and newly married. Last, we investigate some plausible mechanisms through which single motherhood may lead to higher levels of child mortality. Our findings draw attention to the importance of mother’s marital status for children’s survival in sub-Saharan Africa.

Data and Methods

We analyze data from 11 Demographic and Health Surveys (DHS), which are designed to generate nationally representative population samples using a two-stage sampling design. In the first stage, urban and rural enumeration areas (EAs) were selected within each region using the most recent Population and Housing Census sample. In the second stage, a systematic sample of households was drawn from each EA, and all women ages 15–49 in the selected households were eligible to be interviewed. We first selected all DHSs conducted in the last five years in countries where the under-5 mortality rate (5q0) exceeded 80. These include the Democratic Republic of the Congo (DRC) (2007), Ghana (2008), Liberia (2007), Nigeria (2008), Sierra Leone (2008), and Zambia (2007). These surveys contain standard albeit limited questions about women’s marital status, including (1) her marital status at the time of the survey (never married, married/living together, divorced/separated, or widowed); (2) number of previous unions (one or more than one); and (3) age at first marriage (including both month and year). To obtain more detailed information about women’s union histories, we also selected the five DHSs conducted in the last 10 years that contained monthly retrospective marital history calendars: Ethiopia (2005), Kenya (2003), Malawi (2004), Tanzania (2004–2005), and Zimbabwe (2005–2006). These calendars began in January five years before the start of the survey and recorded whether the respondent was married or living as married for each subsequent month until the time of the survey.3 Thus, all women provided monthly marital history data for between five and six years preceding the survey. Although these data represent a clear improvement over the limited questions in the standard DHS, they have some serious limitations. Most notably, marital history data before the calendar are incomplete. Specifically, these surveys did not collect the dates of marital dissolution or second unions if these events occurred before the start of the calendar. Moreover, the calendar data do not provide the reason for the marital dissolution. As a consequence, if a formerly married woman remains unmarried until the time of the survey, we can determine whether she was divorced or widowed on the basis of her current marital status. However, if she has remarried, we do not know whether her previous union ended through divorce or widowhood.

All 11 DHSs collected full birth histories, including the month and year each child was born and the date the child died (if the child is no longer alive). Despite efforts to ensure the accuracy of these birth histories, these records are imperfect. For children born in the last five years, additional health-related data are collected. The collection of these additional health measures has been shown to lead to the omission or displacement of young children’s birth dates (Schoumaker 2011). In addition, children who have died are less likely to be reported by their mothers (Schoumaker 2011), and children whose mothers have died are not reported at all (Hallett et al. 2010), leading to underestimates of child mortality.

Single Motherhood Over the Life Course

Using standard DHS questions, we estimate the cumulative probability that a woman will become a single mother before first marriage from ages 10 to 45 for all 11 countries. We define women as “premarital single mothers” if the date of birth for their first child precedes their date of first marriage.4 Because women who marry before the birth of their first child are no longer at risk of becoming premarital single mothers, we treat marriage as a competing risk to premarital childbirth. To estimate the competing risk model, we apply the cumulative incidence approach proposed by Coviello and Boggess (2004). Women who neither married nor gave birth by the time of the survey are treated as censored. To calculate this premarital single motherhood curve, we rely on data from all women aged 15–49 surveyed in DRC (n = 9,995), Ghana (n = 4,916), Liberia (n = 7,092), Nigeria (n = 33,385), Sierra Leone (n = 7,374), Zambia (n = 7,146), Ethiopia (n = 14,070), Kenya (n = 8,195), Malawi (n = 11,698), Tanzania (n = 10,329), and Zimbabwe (n = 8,907).5

For the five countries with marital history calendar data, we also calculate the risk of becoming a single mother after first marriage. This group of “postmarital single mothers” consists of (1) formerly married women who had at least one living child younger than age 15 at the time of their marital dissolution, and (2) formerly married women who gave birth after their union ended. The latter group represents only a small fraction of postmarital single mothers. Owing to limited marital history information before the calendar began (described earlier), we are unable to reconstruct the full marital histories for 19.2 % of respondents in Ethiopia, 8.4 % in Kenya, 18.5 % in Malawi, 16.3 % in Tanzania, and 13.3 % in Zimbabwe. Because these women’s marital histories are left-truncated, the estimates of the postmarital single motherhood are based only on women who were still in their first marriage at the beginning of the calendar or who got married for the first time during the observation period. To adjust for left-truncation, women contribute to our estimates only for the ages when they are under observation. For example, the risk of becoming a postmarital single mother between the ages of 30 and 35 is calculated using data only from women whose marital history calendars cover the ages of 30 to 35. Left-truncation does not influence the estimation method itself, but it has an impact on the number of observations in the given time period. We calculate the Kaplan-Meier survival function with late entry and estimate the cumulative risk of becoming a single mother after first marriage by age. Women who did not experience an episode of postmarital single motherhood by the time of the survey are censored. Last, we use Kaplan-Meier models to estimate the cumulative failure curves for entry into single motherhood either before or after first marriage. This final curve is not a simple summation of the risk of premarital and postmarital single motherhood because some women may both experience a premarital birth and later become a single mother via divorce or widowhood.

Child Mortality

To assess the correlates of child mortality, we create a sample of all children born in the six years before the survey in the DRC, Ghana, Liberia, Nigeria, Sierra Leone, and Zambia. Extending our cutoff beyond five years helps to capture children born approximately five years before the survey, whose birth dates may have been displaced. In addition, it improves the comparability of our samples from countries with marital history calendars. For these countries, we restrict our sample to all children born during the calendar (which extends up to six years before the survey). By including only these children, we avoid the problems of left-truncation and ensure that we know women’s marital status over the child’s entire lifespan. After removing a small number of children with inconsistent or missing dates of death, our final samples of children range from 3,508 in Ghana to 33,158 in Nigeria.

Mother’s Marital Status

Using the standard DHS questions available in six of our countries, only two categories of mother’s marital status can be unambiguously determined over the child’s life span. First, all children who were born after their mother’s first marriage and whose mothers are still in their first union are identified as having “continuously married” mothers (presumably married to the child’s biological father). Previous literature regarding sub-Saharan Africa suggests that if women marry within six months of a premarital birth, the marriage is generally to the child’s father (Hattori and Larsen 2007). Thus, we also code these mothers as “continuously married.” Second, mothers of children born before their first marriage are labeled as “never married.” For all other children, their mother’s union status from birth to the time of the survey is indeterminate. Instead of omitting these children (and potentially introducing selection bias into our sample), we classify their mothers according to their status at the time of the survey. These mothers may be formerly married (divorced or widowed) or remarried (thus having experienced more than one union). Because we do not know when previous unions dissolved or when new ones began, the categories of both formerly married and remarried mothers must be interpreted with caution. For example, if the child of a currently divorced woman has died, we cannot determine whether the child died before or after her divorce. Similarly, among remarried mothers, we do not know whether the child was born (or died) before or after this second union. Thus, we can assess whether there is an association between having a divorced, widowed, or remarried mother and child mortality, but we cannot draw any causal inference.

To overcome this critical limitation, we employ data from the marital history calendars available in five of our sample countries and create a time-varying covariate to capture women’s marital status in each month of the child’s life. This measure equals 0 if women have been continuously married since the time of the child’s birth and 1 if the woman was never married. Mothers who were formerly married are coded as 2, and mothers who enter into a new marriage are coded as 3. For example, if a women is married at the time of the child’s birth, each month is coded as 0 until the union ends, at which point she is coded as 2. As another example, if a woman is not married at the time of the child’s birth, she will be coded as 1 until she gets married. Thereafter, she will be coded as 3 because she is presumably not marrying the child’s father. As noted earlier, however, mothers who enter into a marriage within six months of the child’s birth are classified as being continuously married (0). In subsequent analyses, formerly married mothers are disaggregated into those who are divorced and those who are widowed. However, for formerly married mothers who enter a new union before the time of the survey, we have no information about whether her prior union ended in divorce or widowhood. Thus, we classify their status as “formerly married, status unknown” for the intervening months between union dissolution and remarriage.

Mother and Child Characteristics

Drawing on previous research, we include the main mother and child characteristics that are associated with higher rates of child mortality (Gakidou et al. 2010). The child characteristics are child’s sex, birth order, length of preceding birth interval (<2 years, 2 years, ≥3 years), and whether any siblings died previously. We also control for maternal characteristics, such as the mother’s age at time of birth (measured as <20, 20–29, and ≥30) and educational attainment (coded as no education, some primary school, and some secondary school or higher).

Potential Mechanisms

In our final analyses, we introduce a set of proxies for potential mechanisms, which could explain an association between single motherhood and child mortality. Specifically, we identify proxies for economic status, parental supervision, and critical selection effects. First, to account for potentially higher rates of poverty among unmarried women, we include an index of household wealth measured at the time of the survey. The index of wealth, which classifies women into quintiles, ranging from 1 = poorest to 5 = wealthiest, is provided in the DHS data and is based on ownership of household assets and housing characteristics. Second, we sought indicators of parental care. Unfortunately, whether the child lived with his or her mother was asked only if the child was alive at the time of the survey. Thus, it cannot be used as a predictor of child mortality.6 Instead, we focus on a critical form of maternal care: namely, breast-feeding. Data on breast-feeding practices were collected for all children born in the last five years, allowing us to create an indicator variable for whether the child was breast-fed in any given month. For children born more than five years before the survey, we include a flag indicating that their breast-feeding histories are unknown.7 Finally, to address concerns about selection into and out of single motherhood, we add indicators for whether unions are polygynous and for mother’s HIV status. Women’s HIV status was not collected in Tanzania, Ghana, and Nigeria. In all other countries, except Zimbabwe and Liberia, only one-half of the women were selected to be tested. Refusal rates were also high, ranging between 20 % and 30 % of eligible respondents. Consequently, we include a flag for whether the respondent was tested. Given these high levels of missing values for breast-feeding and HIV status and given that household wealth, polygynous unions, and HIV status are collected only at the time of the survey, these final analyses offer some exploratory insights into potential mechanisms but are not conclusive.

Models

Analyses were conducted in Stata (version 11.0). Random-effects discrete-time hazard models are used to account for clustering of births within mothers and right-censoring (Sear et al. 2002; Singer and Willlet 2003). Our basic model is as follows:
https://static-content.springer.com/image/art%3A10.1007%2Fs13524-013-0220-6/MediaObjects/13524_2013_220_Equa_HTML.gif
in which htij indicates whether child i of mother j died in month t. Children who were alive at the time of the survey but had not reached their fifth birthday were treated as censored. MarStat is a time-constant variable for the six standard DHS countries and a time-varying variable for the five DHS countries with marital history calendars. In the latter countries, we lagged this variable by one month to ensure that marital transitions preceded child mortality events. Child’s age is grouped into one of five age categories (<1 month, 1–11 months, 1 year, 2 years, and 3–4 years) to best fit the shape of the child mortality curve. α represents the log odds of child mortality in each age category. Although u accounts for mother-level random effects, we further adjust for clustering of deaths within households by including a variable for whether an older sibling died (Das Gupta 1997). Weights are not used because the analysis is conducted at the child level rather than at the woman level, although dummy variables for each region and urban residence are included. All models include child characteristics and maternal characteristics, and our final models add proxies for potential causal mechanisms. We assess the significance of individual predictors using a Wald statistic and the goodness-of-fit with a likelihood ratio test.

Results

Pathways Into Single Motherhood Over the Life Course

Figure 1 shows the cumulative probability of having a premarital birth in the six countries with standard DHS questions. About 10 % of women in Nigeria have a premarital birth, compared with nearly 35 % in Liberia. In Sierra Leone, premarital fertility rises early but slowly, reflecting both early marriage and childbearing. Yet, the total percentage of women in Sierra Leone who will have a premarital birth (24 %) is only slightly less than that in Zambia (26 %), where premarital single motherhood does not begin until after the age of 15 but continues to rise until well past the age of 20 as a result of the later age of marriage. Both Ghana (17 %) and the DRC (15 %) exhibit similar patterns of premarital single motherhood. Figure 2 plots all three single-motherhood curves for Ethiopia, Kenya, Malawi, Tanzania, and Zimbabwe, respectively. In these countries, rates of premarital single motherhood rise fastest between the ages of 15 and 20; but in countries with later ages of first marriage, such as Kenya, Tanzania, and Zimbabwe, the cumulative risk continues to rise well into women’s early 20s. As a result, about 30 % of women in Kenya and 17 % of women in Tanzania and Zimbabwe have a premarital birth. These numbers contrast with 5 % in Ethiopia and 12 % in Malawi.
https://static-content.springer.com/image/art%3A10.1007%2Fs13524-013-0220-6/MediaObjects/13524_2013_220_Fig1_HTML.gif
Fig. 1

Cumulative risk of becoming a premarital single mother, by age

https://static-content.springer.com/image/art%3A10.1007%2Fs13524-013-0220-6/MediaObjects/13524_2013_220_Fig2_HTML.gif
Fig. 2

Cumulative risk of becoming a single mother, by age

Figure 2 shows even greater variation in the cumulative risk of postmarital single motherhood. Although only about 25 % of women in Ethiopia will experience an episode of postmarital single motherhood, approximately one-half of women in both Malawi and Zimbabwe are expected to become single mothers via divorce or widowhood. Moreover, with the notable exception of Kenya, women are far more likely to become single mothers following divorce or death of a spouse than by having a premarital birth. In all countries, the total likelihood of ever being a single mother by the age of 45 via either a premarital birth or a union dissolution is quite substantial: 29.9 % in Ethiopia, 59.5 % in Kenya, 61.0 % in Malawi, 52.1 % in Tanzania, and 68.8 % in Zimbabwe.

The Effects of Single Motherhood on Child Mortality

The descriptive statistics presented in Table 1 reflect the diversity in the level of economic development, population health, and women’s status across these countries. The percentage of children in our samples who have died was lowest in Ghana (6.1 %) and highest in the DRC (10.7 %).8 Among the standard DHS countries, there is considerable variation in the proportion of mothers with young children who were still in their first union, ranging from more than 80 % in Nigeria to slightly more than one-half in Liberia. Similarly, although only 2.7 % of children in Nigeria were born before first marriage, 14.5 % of children in Liberia had never-married mothers. At the time of the survey, mothers were substantially more likely to be divorced than widowed in all countries except Nigeria. Last, a large proportion of mothers were remarried at the time of the survey. From the perspective of the child, however, these may not be new unions given that many of these young children are likely to be the biological offspring of the current spouses.
Table 1

Descriptive characteristics of mothers and children

 

Standard DHS

 

Congo (DRC)

Ghana

Liberia

Nigeria

Sierra Leone

Zambia

n

10,443

3,508

6,644

33,158

6,711

7,425

Child Died Before Age 5 (%)

10.7

6.1

7.3

9.8

10.6

8.3

Mother’s Marital Statusa (%)

      

 Continuously marriedb

73.7

69.1

54.9

82.3

67.9

67.6

 Never married

3.7

5.7

14.5

2.7

8.5

8.9

 Divorced

6.4

4.3

7.2

1.5

3.3

7.2

 Widowed

1.2

1.1

1.3

1.2

1.7

2.4

 Remarried/newly married

15.0

19.8

22.1

12.2

18.6

13.9

Child’s Agea (mean months)

30.0

31.5

31.6

30.6

29.4

30.6

Resides in Urban Area (%)

40.0

33.8

35.0

26.7

33.5

32.6

Male (%)

49.9

51.0

51.0

51.1

50.4

49.4

Birth Order (%)

      

 First

20.1

23.1

22.1

18.8

22.1

20.2

 Second

17.8

20.8

17.8

17.4

20.4

18.8

 Third

14.9

16.9

15.8

15.3

16.6

16.4

 Fourth

12.7

13.0

13.0

13.0

13.4

13.3

 Fifth or higher

34.6

26.3

31.4

35.5

27.4

31.3

Preceding Birth Intervalc (%)

      

 < 2 years

26.8

14.2

19.3

24.3

19.2

15.9

 2 years

37.7

28.8

30.7

38.5

32.4

39.3

 ≥ 3 years

35.5

57.1

50.1

37.2

48.3

44.8

Older Sibling Died (%)

5.8

2.8

3.8

5.8

4.8

4.0

Mother’s Age at Birth (%)

      

 < 20

16.2

12.1

17.3

15.6

18.0

17.9

 20–29

51.2

50.2

48.7

51.9

51.5

54.3

 ≥ 30

32.6

37.7

34.0

32.5

30.4

27.9

Mother’s Education (%)

      

 None

24.7

37.8

48.8

50.2

74.0

13.4

 Primary

42.5

23.9

36.0

23.1

13.0

62.4

 Secondary or more

32.9

38.3

15.2

26.7

13.0

24.2

Wealth Index (quintiles) (%)

      

 Poorest

22.6

31.9

24.0

26.5

22.0

21.5

 Poorer

20.5

21.8

24.0

23.9

19.3

22.0

 Middle

19.6

17.0

21.4

19.6

20.8

22.6

 Richer

20.6

17.0

19.6

16.7

20.4

21.1

 Richest

16.7

12.4

11.1

13.4

17.6

12.8

Breast-fed (first month)

      

 Not breast-fed

2.0

2.0

1.3

2.0

0.8

2.5

 Breast-fed

75.1

78.0

75.6

75.3

68.1

79.5

 Status unknown

22.9

20.0

23.1

22.8

31.1

18.0

Months of Breast-feeding (mean)

15.1

15.6

13.7

14.3

13.7

15.6

HIV Status (at survey)

      

 Negative

47.8

N/A

91.0

N/A

48.3

68.8

 Positive

0.8

N/A

1.7

N/A

0.8

11.1

 Unknown

51.4

N/A

7.3

N/A

50.9

20.1

Polygynous Union (at survey)

18.0

18.7

13.3

32.3

29.5

11.7

Child Lives With Motherd (%)

95.9

95.1

89.4

95.6

89.8

95.0

The timing of childbirth relative to union-status changes can be determined more precisely in our five countries with marital history calendars. The large majority of children have mothers who were continuously married from the time of the child’s birth through the last observation.9 Relatively few mothers remained never married by the last month of observation. Mothers were more likely to be divorced than widowed, although rates of new marriages were relatively low given that children in all our samples were observed for only about 30 months, on average.

Women’s education varied considerably across countries. Only 4.1 % of women in Zimbabwe had never been to school, compared with 77.1 % of women in Ethiopia. Among children who were still alive at the time of the survey, more than 89 % lived with their mothers. However, coresidence with mothers was significantly lower for women who had ever been single mothers (ranging between 78 % and 91 %) than for women who were continuously married (ranging between 92 % and 99 %) in all countries. Breast-feeding is also very common. Less than 3 % of children were not breast-fed in the first month. Among respondents who were tested for HIV, less than 2 % of mothers were infected in all countries except Kenya (8.2 %), Malawi (10.9 %), Zambia (13.9 %), and Zimbabwe (20.4 %). Last, although in some countries (Ethiopia, Malawi, Zambia, and Zimbabwe) only about 10 % of mothers were in polygynous unions, roughly 30 % of mothers had co-wives in Sierra Leone and Nigeria.

Table 2 provides estimates for the association between mother’s marital status at the time of the survey and child mortality for countries with standard DHS questions. Compared with children whose mothers were married and still in their first unions, children whose mothers were not married at the time of their birth were significantly more likely to die in Liberia, Nigeria, and Sierra Leone, and marginally more likely to die (p = .07) in the DRC. Having a mother who was formerly married (i.e., divorced or widowed) at the time of the survey is significantly positively associated with child mortality in all countries except Ghana. Finally, in five of the six countries, children whose mothers remarried were significantly more likely to have died, with odds ratios ranging from 1.30 in Sierra Leone to 1.72 in Ghana. This finding is surprising considering that a substantial fraction of these children are potentially the biological children of the current spouse. However, it is also plausible that many of these children died before their mothers remarried.
Table 2

Mother’s marital status (at time of survey) and child mortality (random-effects, discrete-time logistic regression)

 

Congo (DRC)

Ghana

Liberia

Nigeria

Sierra Leone

Zambia

O.R.

Coef.

SE

Sig.

O.R.

Coef.

SE

Sig.

O.R.

Coef.

SE

Sig.

O.R.

Coef.

SE

Sig.

O.R.

Coef.

SE

Sig.

O.R.

Coef.

SE

Sig.

Mother’s Marital Status

 Continuously married (ref.)

1.00

0.00

––

 

1.00

0.00

––

 

1.00

0.00

––

 

1.00

0.00

––

 

1.00

0.00

––

 

1.00

0.00

––

 

 Never married

1.32

0.28

0.16

1.42

0.35

0.31

 

1.48

0.39

0.15

**

1.36

0.31

0.11

**

1.39

0.33

0.15

*

1.19

0.17

0.16

 

 Formerly married

1.37

0.32

0.11

**

1.43

0.36

0.29

 

1.51

0.41

0.16

**

1.36

0.31

0.09

***

1.72

0.54

0.15

***

1.30

0.26

0.13

*

 Remarried

1.32

0.27

0.08

***

1.72

0.54

0.17

***

1.48

0.39

0.12

***

1.36

0.31

0.05

***

1.30

0.26

0.10

**

0.98

–0.02

0.13

 

Male (ref. = female)

1.13

0.12

0.06

1.13

0.12

0.14

 

1.15

0.14

0.09

 

1.16

0.15

0.04

***

0.99

–0.01

0.08

 

1.19

0.17

0.08

*

Birth Order

                        

 First

1.43

0.36

0.13

**

1.08

0.08

0.31

 

1.07

0.07

0.21

 

1.14

0.13

0.08

 

1.04

0.04

0.17

 

1.72

0.54

0.19

**

 Second

0.79

–0.24

0.11

*

0.79

–0.24

0.27

 

0.73

–0.31

0.17

0.75

–0.29

0.07

***

0.50

–0.69

0.14

***

1.14

0.13

0.16

 

 Third

0.90

–0.11

0.11

 

0.92

–0.08

0.25

 

0.84

–0.18

0.16

 

0.78

–0.25

0.06

***

0.66

–0.41

0.13

**

1.16

0.15

0.16

 

 Fourth

0.81

–0.21

0.11

0.86

–0.15

0.25

 

0.75

–0.29

0.17

0.79

–0.24

0.06

***

0.69

–0.37

0.14

**

1.25

0.22

0.15

 

 Fifth or higher (ref.)

1.00

0.00

––

 

1.00

0.00

––

 

1.00

0.00

––

 

1.00

0.00

––

 

1.00

0.00

––

 

1.00

0.00

––

 

Preceeding Birth Interval

 < 2 years

1.98

0.69

0.09

***

1.49

0.40

0.22

2.16

0.77

0.13

***

2.12

0.75

0.05

***

2.64

0.97

0.12

***

2.01

0.70

0.12

***

 2 years

1.31

0.27

0.09

**

1.23

0.21

0.19

 

1.31

0.27

0.13

*

1.39

0.33

0.05

***

1.92

0.65

0.11

***

1.02

0.02

0.11

 

 ≥ 3 years (ref.)

1.00

0.00

––

 

1.00

0.00

––

 

1.00

0.00

––

 

1.00

0.00

––

 

1.00

0.00

––

 

1.00

0.00

––

 

Older Sibling Died

3.25

1.18

0.08

***

2.94

1.08

0.24

***

3.97

1.38

0.14

***

2.94

1.08

0.06

***

2.97

1.09

0.14

***

3.46

1.24

0.12

***

Mother’s Age at Birth

 < 20 (ref.)

1.00

0.00

––

 

1.00

0.00

––

 

1.00

0.00

––

 

1.00

0.00

––

 

1.00

0.00

––

 

1.00

0.00

––

 

 20–29

0.86

–0.16

0.10

 

0.76

–0.27

0.26

 

0.73

–0.31

0.15

*

0.73

–0.32

0.06

***

0.61

–0.49

0.12

***

0.92

–0.08

0.13

 

 ≥ 30

0.85

–0.17

0.13

 

0.95

–0.05

0.31

 

0.64

–0.44

0.19

*

0.76

–0.27

0.07

***

0.54

–0.61

0.15

***

1.05

0.05

0.18

 

Mother’s Education

 None

1.44

0.36

0.10

***

0.98

–0.02

0.21

 

1.01

0.01

0.15

 

1.10

0.10

0.06

 

1.26

0.23

0.14

 

1.36

0.31

0.15

*

 Primary

1.20

0.18

0.09

*

1.21

0.19

0.19

 

0.99

–0.02

0.15

 

1.06

0.06

0.06

 

1.26

0.23

0.16

 

1.21

0.19

0.11

 Secondary or more (ref.)

1.00

0.00

––

 

1.00

0.00

––

 

1.00

0.00

––

 

1.00

0.00

––

 

1.00

0.00

––

 

1.00

0.00

––

 

Urban (ref. = rural)

0.77

–0.26

0.79

***

1.19

0.17

0.17

 

1.28

0.25

0.12

*

0.76

–0.28

0.05

***

1.22

0.20

0.10

*

1.21

0.19

0.10

*

Child’s Age

 < 1 month (ref.)

1.00

0.00

––

 

1.00

0.00

––

 

1.00

0.00

––

 

1.00

0.00

––

 

1.00

0.00

––

 

1.00

0.00

––

 

 1–11 months

0.12

–2.12

0.07

***

0.06

–2.77

0.16

***

0.11

–2.20

0.10

***

0.08

–2.55

0.04

***

0.12

–2.08

0.09

***

0.08

–2.52

0.09

***

 1 year

0.05

–2.94

0.09

***

0.02

–3.80

0.24

***

0.03

–3.66

0.17

***

0.07

–2.68

0.04

***

0.05

–3.07

0.12

***

0.04

–3.13

0.12

***

 2 years

0.01

–4.29

0.18

***

0.01

–4.40

0.35

***

0.01

–4.31

0.25

***

0.02

–3.78

0.08

***

0.01

–4.31

0.22

***

0.01

–5.08

0.31

***

 3–4 years

0.01

–4.69

0.20

***

0.00

–5.36

0.51

***

0.00

–5.33

0.39

***

0.01

–4.63

0.11

***

0.01

–5.05

0.30

***

0.00

–5.30

0.32

***

Constant

 

–3.73

   

–3.99

   

–4.02

   

–3.53

   

–3.61

   

–4.17

  

Random Part

 Between-mother variance

0.006

0.007

0.082

0.132

0.360

0.018

 Intramother correlation (rho)

.000

.000

.002

.005

.040

.001

Log-Likelihood

–6,272.8

–1,241.7

–2,834.8

–18,761.4

–3,957.2

–3,515.26

Person-Months (n)

313,419

110,494

209,731

1,013,825

197,137

227,666

Number of Children

10,443

3,508

6,644

33,158

6,711

7,425

Number of Mothers

5,732

2,296

4,239

18,720

4,263

4,348

Note: All regressions control for region.

p ≤ .10; *p ≤ .05; **p ≤ .01; ***p ≤ .001

Table 3 uses an improved measure of mother’s marital status based on the marital history calendars. Compared with children whose mothers were married since their birth, children of never-married mothers faced a significantly higher rate of mortality in Malawi, Tanzania, and Zimbabwe. In Zimbabwe, the odds ratio is higher than 2. In comparison, children born to formerly married women were more likely to die in all countries except Zimbabwe. Having a formerly married mother is associated with between a 58 % increase in the odds of dying in Tanzania and a 75 % increase in Kenya. Finally, in both Ethiopia and Malawi, children whose mothers were in a new union are at an elevated risk of dying relative to children whose mothers were continuously married. The effect of having a mother in a new union is especially strong in Ethiopia, although this coefficient should be interpreted with caution given the relatively small percentage of children whose mothers formed a new union during the period of observation. Moreover, although all models control for the child’s age (exposure time), children whose mothers formed a new union were, on average, older.
Table 3

Mother’s marital status (time-vaying) and child mortality (random-effects, discrete-time logistic regression)

 

Ethiopia

Kenya

Malawi

Tanzania

Zimbabwe

O.R.

Coef.

SE

Sig.

O.R.

Coef.

SE

Sig.

O.R.

Coef.

SE

Sig.

O.R.

Coef.

SE

Sig.

O.R.

Coef.

SE

Sig.

Mother’s Marital Status (time-varying)

                    

 Continuously married (ref.)

1.00

0.00

––

 

1.00

0.00

––

 

1.00

0.00

––

 

1.00

0.00

––

 

1.00

0.00

––

 

 Never married

1.51

0.41

0.39

 

1.17

0.16

0.18

 

1.63

0.49

0.16

**

1.45

0.37

0.15

*

2.16

0.77

0.21

***

 Formerly married

1.68

0.52

0.18

**

1.75

0.56

0.19

**

1.72

0.54

0.11

***

1.58

0.46

0.14

***

1.34

0.29

0.22

 

 Newly married

6.42

1.86

0.42

***

1.65

0.50

0.52

 

1.75

0.56

0.29

*

1.48

0.39

0.39

 

2.08

0.73

0.72

 

Male (ref. = female)

1.25

0.22

0.07

***

1.25

0.22

0.09

*

1.30

0.26

0.06

***

1.12

0.11

0.07

 

1.17

0.16

0.10

 

Birth Order

                    

 First

1.45

0.37

0.15

*

1.06

0.06

0.21

 

1.52

0.42

0.15

**

1.05

0.05

0.17

 

0.58

–0.54

0.25

*

 Second

0.67

–0.40

0.13

**

0.94

–0.06

0.17

 

1.09

0.09

0.12

 

0.87

–0.14

0.13

 

0.63

–0.46

0.21

*

 Third

0.91

–0.09

0.12

 

1.01

0.01

0.16

 

1.03

0.03

0.12

 

0.85

–0.16

0.14

 

0.73

–0.31

0.20

 

 Fourth

0.92

–0.08

0.12

 

0.92

–0.08

0.17

 

1.12

0.11

0.12

 

0.76

–0.28

0.14

*

0.75

–0.29

0.21

 

 Fifth or higher (ref.)

1.00

0.00

––

 

1.00

0.00

––

 

1.00

0.00

––

 

1.00

0.00

––

 

1.00

0.00

––

 

Preceding Birth Interval

                    

 < 2 years

2.51

0.92

0.10

***

1.68

0.52

0.13

***

1.99

0.69

0.09

***

1.55

0.44

0.11

***

2.23

0.80

0.16

***

 2 years

1.31

0.27

0.10

**

0.87

–0.14

0.13

 

1.13

0.12

0.09

 

1.05

0.05

0.10

 

0.89

–0.12

0.15

 

 ≥ 3 years (ref.)

1.00

0.00

––

 

1.00

0.00

––

 

1.00

0.00

––

 

1.00

0.00

––

 

1.00

0.00

––

 

Older Sibling Died

3.56

1.27

0.14

***

5.37

1.68

0.13

***

3.46

1.24

0.10

***

3.10

1.13

0.12

***

4.81

1.57

0.16

***

Mother’s Age at Birth

                    

 < 20 (ref.)

1.00

0.00

––

 

1.00

0.00

––

 

1.00

0.00

––

 

1.00

0.00

––

 

1.00

0.00

––

 

 20–29

0.67

–0.40

0.11

***

0.95

–0.05

0.15

 

0.81

–0.21

0.09

*

0.83

–0.19

0.12

0.83

–0.19

0.16

 

 ≥ 30

0.74

–0.30

0.14

*

1.14

0.13

0.20

 

0.90

–0.11

0.14

 

0.84

–0.17

0.16

 

0.64

–0.45

0.23

Mother’s Education

                    

 None

1.34

0.29

0.20

 

1.82

0.60

0.18

***

1.49

0.40

0.14

**

1.68

0.52

0.18

**

0.98

–0.02

0.27

 

 Primary

1.26

0.23

0.21

 

1.46

0.38

0.14

 

1.38

0.32

0.13

*

1.48

0.39

0.18

*

0.95

–0.05

0.12

 

 Secondary or more (ref.)

1.00

0.00

––

 

1.00

0.00

––

 

1.00

0.00

––

 

1.00

0.00

––

 

1.00

0.00

––

 

 Urban (ref. = rural)

0.92

–0.08

0.14

 

0.98

–0.02

0.13

 

0.68

–0.38

0.12

**

0.99

–0.01

0.11

 

1.02

0.02

0.20

 

Child’s Age

                    

 < 1 month (ref.)

1.00

0.00

––

 

1.00

0.00

––

 

1.00

0.00

––

 

1.00

0.00

––

 

1.00

0.00

––

 

 1–11 months

0.07

–2.71

0.08

***

0.09

–2.39

0.10

***

0.12

–2.12

0.07

***

0.10

–2.33

0.08

***

0.09

–2.38

0.11

***

 1 year

0.03

–3.68

0.11

***

0.03

–3.41

0.14

***

0.04

–3.13

0.10

***

0.05

–2.95

0.10

***

0.02

–3.92

0.20

***

 2 years

0.02

–4.19

0.15

***

0.01

–4.35

0.24

***

0.02

–4.11

0.16

***

0.02

–4.11

0.18

***

0.01

–4.39

0.27

***

 3–4 years

0.00

–5.41

0.26

***

0.00

–5.89

0.51

***

0.00

–5.39

0.27

***

0.00

–6.42

0.50

***

0.00

–5.30

0.39

***

Constant

 

–3.91

   

–3.98

   

–4.21

   

–3.82

   

–2.88

  

Random Part

                    

 Between-mother variance

0.100

0.004

0.006

0.005

0.004

 Intramother correlation (rho)

.003

.000

.000

.000

.000

Log-Likelihood

–4,967.9

–2,903.6

–6,389.7

–4,794.7

–2,274.7

Person-Months (n)

347,165

184,572

359,167

311,946

191,594

Number of Children

11,210

6,409

12,268

10,040

6,073

Number of Mothers

6,882

4,104

7,553

6,002

4,385

Note: All regressions control for region.

p ≤ .10; *p ≤ .05; **p ≤ .01; ***p ≤ .001

Tables 2 and 3 suggest that other characteristics of mothers and children have the expected effects on child mortality: males, children born to adolescent mothers, children whose older sibling died, and those with shorter birth intervals are, in most countries, less likely to survive. Interestingly, although the strong positive effect that mother’s education has on child survival is well known (Bicego and Ties 1993; Gakidou et al. 2010), we find that the difference in mortality rates between women with no education and those with secondary schooling is significant in only 5 of the 11 countries (odds ratios of 1.36–1.82). Last, the intramother correlation (rho) is close to 0 in all our models, suggesting very little within-mother correlation.

Model 1 of Table 4, which disaggregates formerly married mothers into divorcées and widows at the time of the survey, shows that divorce is positively associated with child mortality in all six countries, although this effect is not significant in Ghana and is only marginally significant in Zambia. Children whose mothers were divorced experienced a 36 % increase in the odds of dying in the DRC, a 52 % increase in Liberia, a 57 % increase in Nigeria, and almost a twofold increase in Sierra Leone, compared with children whose parents were continuously married. In contrast, although the coefficient is usually positive, the association between currently widowed women and recent child mortality is never significant. The addition of proxies for mechanisms in Model 2 has little effect on the coefficients of mother’s marital status. Nonetheless, several indicators have a direct effect. As expected, greater wealth and breast-feeding are negatively associated with children’s mortality, whereas having an HIV-positive mother or a mother in a polygynous union is positively correlated with mortality.
Table 4

Potential mediators of the effects of single motherhood (at time of survey) on child mortality (random-effects, discrete-time logistic regression)

 

Congo (DRC)

Ghana

Liberia

Nigeria

Sierra Leone

Zambia

O.R.

Coef.

SE

Sig.

O.R.

Coef.

SE

Sig.

O.R.

Coef.

SE

Sig.

O.R.

Coef.

SE

Sig.

O.R.

Coef.

SE

Sig.

O.R.

Coef.

SE

Sig.

Model 1

                        

 Mother’s marital status

  Continuously married (ref.)

1.00

0.00

––

 

1.00

0.00

––

 

1.00

0.00

––

 

1.00

0.00

––

 

1.00

0.00

––

 

1.00

0.00

––

 

  Never married

1.32

0.28

0.16

1.45

0.37

0.31

 

1.48

0.39

0.15

**

1.36

0.31

0.11

**

1.39

0.33

0.15

*

1.19

0.17

0.16

 

  Divorced

1.36

0.31

0.12

**

1.65

0.50

0.31

 

1.52

0.42

0.17

*

1.57

0.45

0.12

***

1.99

0.69

0.17

***

1.28

0.25

0.15

  Widowed

1.44

0.36

0.25

 

0.80

–0.22

0.73

 

1.48

0.39

0.39

 

1.12

0.11

0.15

 

1.17

0.16

0.30

 

1.31

0.27

0.24

 

  Remarried

1.32

0.27

0.08

***

1.72

0.54

0.17

***

1.48

0.39

0.12

***

1.36

0.31

0.05

***

1.30

0.26

0.10

**

0.98

–0.02

0.13

 

Model 2

                        

 Mother’s marital status

  Continuously married (ref.)

1.00

0.00

––

 

1.00

0.00

––

 

1.00

0.00

––

 

1.00

0.00

––

 

1.00

0.00

––

 

1.00

0.00

––

 

  Never married

1.21

0.19

0.16

 

1.75

0.56

0.32

1.57

0.45

0.15

**

1.30

0.26

0.11

*

1.54

0.43

0.17

*

1.13

0.12

0.16

 

  Divorced

1.40

0.34

0.12

**

1.46

0.38

0.32

 

1.60

0.47

0.18

**

1.63

0.49

0.12

***

2.29

0.83

0.23

***

1.11

0.10

0.15

 

  Widowed

1.35

0.30

0.26

 

0.69

–0.37

0.74

 

1.27

0.24

0.41

 

1.05

0.05

0.16

 

1.27

0.24

0.37

 

0.79

–0.23

0.25

 

  Remarried

1.27

0.24

0.08

**

1.51

0.41

0.17

*

1.39

0.33

0.12

**

1.26

0.23

0.05

***

1.19

0.17

0.12

 

0.86

–0.15

0.13

 

 Wealth index (quintiles)

  Poorest (ref.)

1.00

0.00

––

 

1.00

0.00

––

 

1.00

0.00

––

 

1.00

0.00

––

 

1.00

0.00

––

 

1.00

0.00

––

 

  Poorer

0.94

–0.07

0.09

 

0.78

–0.25

0.23

 

1.05

0.05

0.14

 

0.94

–0.06

0.05

 

0.64

–0.44

0.15

**

1.11

0.10

0.13

 

  Middle

0.80

–0.22

0.09

*

1.17

0.16

0.26

 

0.99

–0.01

0.16

 

0.85

–0.16

0.06

**

0.76

–0.27

0.15

0.99

–0.01

0.14

 

  Richer

0.85

–0.16

0.11

 

1.00

0.00

0.29

 

1.00

0.00

0.17

 

0.73

–0.31

0.07

***

0.70

–0.36

0.17

*

1.04

0.04

0.18

 

  Richest

0.53

–0.63

0.17

***

0.80

–0.22

0.38

 

1.06

0.06

0.22

 

0.63

–0.47

0.10

***

0.65

–0.43

0.22

*

0.91

–0.09

0.23

 

 Breast-fed

1.00

   

1.00

   

1.00

   

1.00

   

1.00

   

1.00

   

  Not breast-fed (ref.)

1.00

0.00

––

 

1.00

0.00

––

 

1.00

0.00

––

 

1.00

0.00

––

 

1.00

0.00

––

 

1.00

0.00

––

 

  Breast-fed

0.08

–2.50

0.10

***

0.03

–3.38

0.23

***

0.10

–2.31

0.17

***

0.08

–2.54

0.06

***

0.19

–1.64

0.17

***

0.05

–3.03

0.13

***

  Unknown

0.38

–0.96

0.10

***

0.25

–1.37

0.20

***

0.56

–0.58

0.16

***

0.44

–0.82

0.06

***

1.40

0.34

0.18

0.21

–1.58

0.13

***

 Mother’s HIV status

1.00

   

1.00

   

1.00

   

1.00

   

1.00

   

1.00

   

  Negative (ref.)

1.00

0.00

––

 

1.00

   

1.00

0.00

––

 

1.00

   

1.00

0.00

––

 

1.00

0.00

––

 

  Positive

1.07

0.06

0.34

 

1.00

   

1.51

0.41

0.28

 

1.00

   

1.75

0.56

0.44

 

2.18

0.78

0.11

***

  Unknown

1.00

0.00

0.06

 

1.00

   

1.09

0.09

0.19

 

1.00

   

0.92

–0.08

0.10

 

1.06

0.06

0.11

 

 Polygynous union

1.22

0.20

0.08

*

1.10

0.10

0.19

 

1.36

0.31

0.14

*

1.16

0.15

0.04

***

1.49

0.40

0.11

***

1.17

0.16

0.13

 

Log-Likelihood

–5,898.02

–1,114.76

–2,669.42

–17,618.36

–3,744.33

–3,214.14

Person-Months (n)

313,419

110,194

209,731

1,013,825

197,137

227,666

Note: All regressions control for all mother and child characteristics shown in Table 2.

p ≤ .10; *p ≤ .05; **p ≤ .01; ***p ≤ .001

Turning to our data with marital history calendars, we separate formerly married mothers who were divorced or widowed and remained unmarried through the time of the survey from those who remarried by the time of the survey (Model 1, Table 5). These fine distinctions produce interesting results. Similar to our findings from the other countries, children whose mothers divorced tend to fare poorly. The effects of divorce are quite large, more than doubling the odds of dying in Kenya. Widowhood is significantly correlated with child mortality only in Malawi. However, in all countries except Kenya, children of formerly married mothers who eventually married had the highest rate of mortality. Children whose mothers remarried by the time of the survey may potentially have experienced their parent’s union rupture at younger ages. However, subsequent analyses reveal little difference in length of exposure post-union rupture. Thus, it appears to suggest a strong selection into new unions among formerly married women whose children have died.
Table 5

Potential mediators of the effects of single motherhood (time-varying) on child mortality (random-effects, discrete-time logistic regression)

 

Ethiopia

Kenya

Malawi

Tanzania

Zimbabwe

O.R.

Coef.

SE

Sig.

O.R.

Coef.

SE

Sig.

O.R.

Coef.

SE

Sig.

O.R.

Coef.

SE

Sig.

O.R.

Coef.

SE

Sig.

Model 1

                    

 Mother’s marital status

                    

  Continuously married (ref.)

1.00

0.00

––

 

1.00

0.00

––

 

1.00

0.00

––

 

1.00

0.00

––

 

1.00

0.00

––

 

  Never married

1.51

0.41

0.39

 

1.19

0.17

0.18

 

1.63

0.49

0.16

**

1.46

0.38

0.15

*

2.16

0.77

0.21

***

  Divorced

1.68

0.52

0.23

*

2.18

0.78

0.23

***

1.42

0.35

0.17

*

1.68

0.52

0.18

**

1.00

0.00

0.34

 

 Widowed

0.96

–0.04

0.51

 

1.13

0.12

0.37

 

1.86

0.62

0.27

*

0.90

–0.10

0.39

 

1.31

0.27

0.37

 

  Formerly married, status unknown

2.66

0.98

0.35

**

1.65

0.50

0.59

 

2.14

0.76

0.17

***

1.82

0.60

0.23

**

2.46

0.90

0.39

*

  Newly married

6.42

1.86

0.42

***

1.67

0.51

0.52

 

1.77

0.57

0.29

*

1.48

0.39

0.39

 

2.08

0.73

0.72

 

Model 2

                    

 Mother’s marital status

                    

  Continuously married (ref.)

1.00

0.00

––

 

1.00

0.00

––

 

1.00

0.00

––

 

1.00

0.00

––

 

1.00

0.00

––

 

  Never married

1.32

0.28

0.41

 

1.22

0.20

0.19

 

1.62

0.48

0.16

**

1.54

0.43

0.15

**

1.95

0.67

0.21

**

  Divorced

1.99

0.69

0.24

**

2.23

0.80

0.24

***

1.57

0.45

0.17

**

1.86

0.62

0.18

***

0.93

–0.07

0.35

 

  Widowed

0.93

–0.07

0.52

 

0.99

–0.01

0.38

 

1.86

0.62

0.28

*

0.90

–0.11

0.39

 

0.93

–0.07

0.37

 

  Formerly married, status unknown

2.29

0.83

0.37

*

1.58

0.46

0.61

 

1.82

0.60

0.18

***

1.58

0.46

0.23

*

2.03

0.71

0.39

  Newly married

6.11

1.81

0.44

***

1.55

0.44

0.52

 

1.57

0.45

0.29

 

1.19

0.17

0.39

 

1.88

0.63

0.72

 

 Wealth index (quintiles)

                    

  Poorest (ref.)

1.00

0.00

––

 

1.00

0.00

––

 

1.00

0.00

––

 

1.00

0.00

––

 

1.00

0.00

––

 

  Poorer

1.09

0.09

0.11

 

0.84

–0.18

0.15

 

1.08

0.08

0.09

 

1.21

0.19

0.11

0.92

–0.08

0.15

 

  Middle

1.25

0.22

0.11

0.94

–0.06

0.15

 

1.05

0.05

0.09

 

1.15

0.14

0.11

 

0.99

–0.01

0.16

 

  Richer

1.13

0.12

0.12

 

0.86

–0.15

0.17

 

0.97

–0.03

0.10

 

0.96

–0.04

0.12

 

0.79

–0.23

0.21

 

  Richest

0.66

–0.42

0.16

*

0.95

–0.05

0.21

 

0.80

–0.22

0.14

 

0.91

–0.09

0.17

 

1.01

0.01

0.29

 

 Breast-fed

                    

  Not breast-fed (ref.)

1.00

0.00

––

 

1.00

0.00

––

 

1.00

0.00

––

 

1.00

0.00

––

 

1.00

0.00

––

 

  Breast-fed

0.05

–2.97

0.11

***

0.10

–2.26

0.14

***

0.07

–2.70

0.09

***

0.07

–2.60

0.11

***

0.07

–2.66

0.17

***

  Unknown

0.34

–1.07

0.10

***

0.80

–0.22

0.15

 

0.25

–1.39

0.10

***

0.28

–1.26

0.11

***

0.28

–1.28

0.17

***

 Mother’s HIV status

                    

  Negative (ref.)

1.00

0.00

––

 

1.00

0.00

––

 

1.00

0.00

––

 

1.00

   

1.00

0.00

––

 

  Positive

1.49

0.40

0.41

 

1.77

0.57

0.21

**

2.12

0.75

0.15

***

1.00

   

2.56

0.94

0.12

***

  Unknown

0.98

–0.02

0.07

 

1.11

0.10

0.10

 

1.15

0.14

0.08

1.00

   

1.03

0.03

0.17

 

 Polygynous union

1.14

0.13

0.11

 

1.06

0.06

0.14

 

1.11

0.10

0.09

 

0.91

–0.09

0.10

 

1.07

0.07

0.16

 

Log-Likelihood

–4,513.27

–2,685.90

–5,972.24

–4,515.45

–2,107.83

Person-Months (n)

347,165

184,572

359,167

311,946

191,594

Note: All regressions control for all mother and child characteristics shown in Table 3.

p ≤ .10; *p ≤ .05; **p ≤ .01; ***p ≤ .001

In Model 2 of Table 5, we include our proxy indicators for economic status, parental care, and potential selection effects. Consistent with the results from Table 4, the addition of these proxies has little effect on the coefficients associated with mother’s marital status. Similarly, children who are breast-fed are strikingly less likely to die, and children with HIV-positive mothers face a greater risk of dying, especially in Malawi and Zimbabwe, which have high levels of HIV. In contrast with Table 4, however, here there is no statistically significant association between mother’s polygynous union and child mortality.

Discussion and Limitations

A strikingly large percentage of women in sub-Saharan Africa become single mothers. Our estimates range from roughly 30 % in Ethiopia, where nonmarital childbearing is rare, to nearly 70 % in Zimbabwe. Moreover, although high levels of premarital childbearing in sub-Saharan Africa have attracted some attention, we find that women are more likely to become single mothers through union dissolution. Roughly one-half of women in Malawi and Zimbabwe will become single mothers as the result of either divorce or widowhood. Despite these high rates of single motherhood, the implications of nonmarital childbearing and child rearing for child well-being have received little attention.

Our analyses indicate that having a single mother significantly increases the risk of a child dying before age 5. However, whether premarital or postmarital single motherhood is most detrimental varies across countries. In six countries, children of never-married women were more likely to die; and in nine countries, children of formerly married women experienced an elevated mortality rate compared with children whose parents were continuously married. Children whose mothers were divorced were especially vulnerable. This finding is noteworthy given that children of divorce are rarely studied in sub-Saharan Africa. One of our most unexpected results is that widowhood was rarely significantly associated with child mortality. Given that death tends to cluster in households and that infectious diseases could affect everyone in a family, it is surprising that widowhood had so little effect. One possible explanation is that widows may receive more support than divorced women from paternal kin. They may also inherit some, if not all, of their former spouses’ assets. Another explanation is the relatively small sample of widows with young children. Indeed, in several countries, the coefficient on widowhood remains insignificant, although it is large and positive. In addition, because widows are themselves more likely to die, their children would be excluded from our sample (Hallett et al. 2010). Finally, we note that although Table 4 shows a strong association between child mortality and remarriage at the time of the survey, Table 5 suggests that many of these children probably died before their mothers remarried.

Explaining why children of unmarried women face a greater risk of dying is difficult with limited data, and our explorations yield few conclusive results. The addition of proxies for economic status, maternal care, and selection effects slightly alters the coefficients associated with never-married and formerly married women, but not in a consistent direction across all countries. Only the coefficients associated with remarriage are moderately, but consistently, lower in Model 2 than in Model 1 (Tables 4 and 5). These coefficients fall most noticeably after controls for breast-feeding are added, suggesting that reductions in maternal care may partly explain why new unions are harmful to young children.

Research from North America has typically found that controlling for changes in economic status accounts for roughly one-half of the negative effects of divorce on children’s well-being (Amato 2000). We find that including indicators of current household wealth did not alter the relationship between single motherhood and child mortality in sub-Saharan Africa. However, because our measures of current wealth do not capture changes in women’s economic status over time, it remains plausible that women’s wealth would decrease substantially following either divorce or widowhood and that this decrease could result in greater child mortality. Data from longitudinal studies are needed. Ideally, these longitudinal economic measures would capture not only changes in household wealth but also in the intrahousehold allocation of resources. Although single mothers may be relatively poorer, they may be better able to direct households’ resources toward their children compared with women in new marriages (Bledsoe 1995). This differential treatment of children from former unions may partially explain their higher mortality levels in Ethiopia and Malawi.

In addition to having fewer economic resources, single mothers are less likely to coreside with their children, thus limiting the amount of maternal care their children receive. In our samples, children of unmarried mothers were more likely than children whose parents were married to live elsewhere (up to 22 % versus less than 8 %). Studies that assess not only where children live but also the time mothers spend working away from the household are needed to understand how single mothers manage the care of their children. In addition, future studies should include indicators of support from extended kin, who may or may not coreside with the child. Although single mothers are sometimes referred to as “lone parents” in the Western literature, this phrase poorly describes unmarried mothers in Africa because they often receive extensive help from both maternal and paternal relatives (Gage 1997; Madhavan and Townsend 2007; Sear et al. 2002; Townsend et al. 2002). Some researchers have noted that in matrilineal ethnic groups, single mothers are likely to receive more support from kin with raising their children (Takyi and Gyimah 2007). Additional analyses (not shown), however, found that children of single mothers belonging to matrilineal ethnic groups in Malawi (Yao and Lomwe) or in Ghana (Akan) did not fare any better than children of single mothers from patrilineal groups.

Finally, our analyses provide little evidence that selection effects drive the relationship between single motherhood and child mortality. Although HIV-positive women are more likely to be divorced or widowed and single mothers are more likely to enter into new unions that are polygynous, these differences do not account for the effects of being formerly married or newly married on child survival. In sum, although our understanding of the causal mechanisms linking single motherhood to poor child health remains speculative, we suggest that single mothers in Africa, as elsewhere, face both increased economic hardships and greater time constraints. In severely impoverished settings like those found in many parts of sub-Saharan Africa, these additional burdens can have dire consequences for their children. Perhaps most importantly, the support single mothers traditionally received from both the maternal and paternal kin of their children may be eroding as rapid urbanization, increased internal migration, and a sustained HIV/AIDS epidemic strain these kinship networks.

Research and Policy Implications

Despite considerable data limitations, our results across a wide range of countries in eastern, western, central, and southern Africa indicate that the implications of both premarital and postmarital single motherhood warrant additional research. The consistently negative effects associated with single motherhood contrast with the mixed, and often positive, results found in the previous literature using measures of female-headed households (Adhikari and Podhisita 2010; Hargreaves et al. 2004; Kennedy and Peters 1992; Onyango et al. 1994). Overall, we find a very low correspondence between women who are household heads and those who are single mothers at the time of the survey. In all countries, more than one-third of single mothers are not household heads; and in 7 of the 11 countries, more than two-thirds of female household heads are not single mothers. In short, we contend that female headship is a poor proxy measure of single motherhood and should not be used in studies examining the effects of family structure on children’s well-being. Using measures of women’s marital status would require gathering additional data—specifically, the beginning and ending dates of all unions—but these data could prove quite valuable.

To meet the fourth U.N. Millennium Development Goal (http://www.un.org/millenniumgoals/) of reducing child mortality by two-thirds by 2015 and to continue to reduce child mortality beyond 2015 requires not only sustained commitment to well-known determinants of child mortality but also the identification of new factors. Women’s education is often considered to be one of the most important determinants of child health (Gakidou et al. 2010). Our results suggest that a mother’s marital status may be as important as her educational attainment. We find that the effects of single motherhood—particularly, postmarital single motherhood—on child mortality are quite large: of roughly the same order of magnitude as the difference between having a mother with no education and having a mother who completed secondary school. Yet, few programs are designed to reach single mothers and their children. Such programs could go beyond providing economic assistance to encourage unwed adolescent mothers to stay in school, offer livelihood training programs for formerly married mothers, subsidize the costs of daycare or crèche services, and ensure that child health services (vaccinations and pediatric care) are available during weekends or evenings. Designing programs to assist single mothers could play an important role in protecting this vulnerable population and ultimately reducing child mortality in sub-Saharan Africa.

Footnotes
1

We refer to all union dissolutions that are not a consequence of a spousal death as “divorces,” although this term does not necessarily indicate a formal or legal divorce.

 
2

The countries covered are the Democratic Republic of the Congo (DRC), Ethiopia, Ghana, Kenya, Liberia, Malawi, Nigeria, Sierra Leone, Tanzania, Zambia, and Zimbabwe.

 
3

In the Ethiopian survey, the Ethiopian calendar is used, which corresponds to a start date of September 12, 1999, in the Gregorian calendar.

 
4

For missing dates, we use values imputed by DHS. The DHS imputation procedures first identify an initial logical range for the missing date based on other relevant dates (such as the age of first marriage, age of first sexual intercourse, and children’s birth dates) and then randomly assign a date within this range.

 
5

Women who delivered their first child or got married before age 10 are dropped from the analysis (N = 37 in Nigeria; 19 in Ethiopia; 17 in Zambia; 11 in Zimbabwe; 5 in Tanzania; 1 in Malawi, Ghana, and Sierra Leone; and 0 in Kenya, DRC, and Liberia).

 
6

Additional analyses (not shown) also included a dummy variable for whether fathers resided in the household at the time of the survey. This variable was not significant in any country after controlling for mother’s marital status.

 
7

The percentage of children with missing information on their breast-feeding status ranges from only 12 % in Kenya to 31 % in Sierra Leone, which partly reflects differential levels of birth date displacement across surveys.

 
8

The percentages presented in Table 1 are lower than true under-5 mortality rates (5q0) because they do not take into account right-censoring of children below age 5.

 
9

Unlike our estimates using standard DHS questions, continuously married mothers in countries with marital history calendars are not restricted to women in their first marriage.

 

Acknowledgements

Funding to support this work was generously provided by the Nike Foundation. We wish to thank Judith Bruce at the Population Council and Kathleen Beegle at the World Bank for their valuable comments and insights.

Copyright information

© Population Association of America 2013