The Association between Household and Community Single
Motherhood and Adolescent Pregnancy in South Africa
This study investigated the independent association of single motherhood at both
household- and community-levels with adolescent pregnancy. A sample of 14,232
female adolescents aged 10-19 years was obtained from the 2011, 2012 and 2013
South African General Household Surveys (GHS). These data were analysed using
descriptive statistics and multilevel binary logistic regression with Stata. Interaction
terms were also tested. Findings showed that living in a single motherhood household
increased the average odds of adolescent pregnancy as did high levels of single
motherhood within communities in the adjusted models. Significant interaction was
found between both household and community single motherhood variables and
education, poverty and household sex composition, after adjusting for other variables.
These results highlight the important need for supporting households headed by single
mothers. This need arises from an independent association as well as the added risk
that occurs when single motherhood occurs in the presence of school non -attendance
and poverty. Early pregnancy prevention programmes and awareness campaigns for
females growing up in homes and environments with greater levels of single
motherhood are encouraged in order to ensure the sexual and reproductive health of
young females in South Africa.
Keywords: Single motherhood, adolescent pregnancy, South Africa, multilevel
modelling, community and household level factors
Adolescent pregnancy has been identified as a pressing social and health challenge in
sub-Saharan Africa, as in many other parts of the world, due to its multiple adverse
social, health and demographic consequences. In South Africa, the 2008 National Youth
Risk Behaviour Survey indicated that one in four (24.4%) sexually active teenage
females had ever been pregnant (Reddy et al., 2010). Low educational levels, large
household size and low socio-economic status have been consistently described as
important individual- and household-level predictors (Malema, 2000; Jewkes, Morrell, &
Christofides, 2009; Macleod, 2011; Macleod & Durrheim, 2002; Macleod & Tracey, 2010;
Mkhwanazi, 2010; Panday, Makiwane, Ranchod, & Letsoalo, 2009; Timaeus and
Moultrie, 2012; Vundule, 2001) . The number of adolescents who became pregnant in
the previous year increased by 12% over 2009-13, with the adolescent pregnancy rate
rising from 33 to 38 pregnancies per 1,000 females aged 10-19 over 2011-13, posing
growing challenges (Mkwananzi, 2017).
Contextual factors accounting for the levels of adolescent pregnancy seen
nationally, as cited by previous authors, suggest addressing socio-structural factors as
an appropriate response that could empower adolescent girls and thus help them
prevent adolescent childbearing and its related risks (Macleod, 2011; Panday and
UNICEF, 2009; Jewkes, Morrell, and Christofides, 2009; Mkhwanazi, 2011). Local studies
that have investigated the effects of community factors on adolescent pregnancy have
primarily adopted qualitative approaches and have cited gender inequality as well as
poverty as important determinants (Limpopo Population and Development Directorate,
2012; Willan, 2013). There is a need to quantitatively investigate factors beyond the
individual-level that could be associated with adolescent pregnancy in South Africa,
while controlling for adolescent characteristics.
One such factor is single motherhood, both at the household- and community-
levels. Past international studies have shown a positive association between lone
parenting and the risk of adolescent pregnancy (Luo, Wang, & Gao, 2012; Omar et al.,
2010; Ugoji, 2011). For example, Francis (2008) demonstrated that the likelihood of
pregnancy among adolescent single females in Lesotho increased when they were living
in households headed by separated or divorced females. Possible explanations for the
association between single motherhood and adolescent pregnancy encompass issues
such as paternal absence, intergenerational adolescent motherhood, as well as reduced
communication and embarrassment regarding sexual matters on the part of single
female parents, leading many adolescents to seek information about sex mostly from
their peers (Domenico & Jones, 2007; Ugoji, 2011). Such factors underscore the
importance of investigating the influence of single motherhood on adolescent pregnancy
in South Africa. This study seeks to determine the independent relationship between
single motherhood and adolescent pregnancy through answering two major research
questions. Firstly, what is the association between single motherhood and adolescent
pregnancy, after controlling for demographic and socio-economic factors; and secondly,
what is the mediating effect of single motherhood on the association between
adolescent pregnancy and individual, household and community background factors?
In many sub-Saharan African countries, the proportion of female-headed families
has increased over time, but sub-regional differences exist, with West Africa having the
lowest levels, followed by East A frica and Southern A frica (Dlamini, 2010). The most
recent demographic and health surveys for Zimbabwe, Lesotho and Swaziland show
high levels of single female headedness that range from 20% to 41%, driven in large
part by high levels of premarital childbearing and HIV related deaths (Mbanefo, 2013;
Moyo & Kawewe, 2009). In South Africa, only about one third of households are
composed of nuclear families (Budlender and Lund, 2011) and the single female-headed
family is not a new phenomenon historically. Upon liberation from the Dutch and British
colonial rule, South Africa experienced apartheid, a system of institutionalised racial
segregation and of internal colonization established in 1948 (Bunting, 1964). During this
period, indigenous whites subjugated their fellow compatriots, who were unable to
purchase land except in reservations, where few employment opportunities existed. This
led to many black African males migrating for work away from their families, to urban
centres or mine settlements (Bunting, 1964; South Africa Ministry in the Office of the
President, 1994), giving rise to the widespread phenomenon of “absent fathers” at the
household-level (Timaeus and Moultrie, 2015).
Consequently, many black African women were forced to raise and fend for their
offspring by themselves (Bunting, 1964; Emmett, 2003). This phenomenon has
continued to the present day, such that the country is now experiencing a fourth
generation of children raised in single female households. Gustafsson and Worku (2006)
have suggested that black African women may have opted to raise their children alone
due to husbands having higher levels of power and rights in marriage under apartheid
law. They found that exposing black African women to the white women’s marriage
market increased their likelihood of marriage by only 8%. This could imply that as single
female-headed households increased, the desire to marry among this population group
decreased. Other determinants of increasing household structure diversity have included
delays in marriage accompanied by non-marital childbearing, increases in separation
and divorce, as well as rising widowhood due to higher female life expectancy (Weeks,
In 2010, over 60% of South A frican urban families, in all race groups, were
headed by female single parents (Holborn & Eddy, 2011). In the Witwatersrand region,
the level of single female-headed households increased from 14% in 1962 to 29% in
1985 (A moateng and Setlalentoa, 2015). In 2008, approximately 40% of South African
children were living with their mothers only (Budlender and Lund, 2011). Single female-
headed households are considered a vulnerable population group and a priority for
development programmes (Kusakabe, 2000). Single female heads are arguably “triply
disadvantaged” as they face the weights of poverty, gender discrimination and the
absence of financial support (Buvinic and Gupta, 1997). Single female-headed
households experience poverty at higher levels than households headed by lone fathers
with children. They are dependent on a woman’s income, which is likely to be lower and
whose position may be more disadvantaged in the workforce. They may also be
characterized by lower labour power and control over and access to resources, all of
which decrease their economic autonomy (UN, 2015). In some settings this group of
women still confronts rejection and prejudice from society, culture and religion when
compared with their married counterparts (Essien & Bassey, 2012). Despite these
challenging consequences, single female headedness is here to stay and societies
urgently need to increase their tolerance and establish the social orientation and needs
of such households (Essien and Bassey, 2012).
Contextual overview of kinship relationships sub-nationally in South Africa
The kinship and key socio-economic characteristics of the various sub-national regions
(known as provinces in South A frica) are displayed in Table 1 below. The first column
displays levels of adolescent pregnancy across provinces. The three highest incidence
levels for 2011-2013 were seen in the North West, Northern Cape and Eastern Cape.
KwaZulu Natal, Limpopo, Mpumalanga and the Free State also display pregnancy
incidence levels of above 3% per annum.
Table 1: Levels of Adolescent Kinship Characteristics across South African
Provinces, (GHS 2011-2013)a
a Percentages of various phenomena among adolescent females in South Africa
b Adolescent pregnancy is the percentage of pregnancies among females aged 10-19 years per province
These six provinces all have relatively high levels of poverty, paternal absence
and death, single motherhood headship, and community levels of single motherhood.
The North West has the highest level of adolescent pregnancy, with poverty levels
above 60% and almost every second adolescent female here living without their fathers.
Similarly, the highest levels of poverty were seen in these same provinces
(ranging from 55.56% in the Northern Cape to 76.55% of households in Limpopo).
Adolescent females living in Limpopo also experienced the highest level of paternal
absence (60.32%), with one in five no longer having a living father and one in three
living in homes headed by their single mothers. In addition, the Eastern Cape displayed
very high levels of poverty (73.23%), paternal absence (52.23%), paternal orphanhood
(22%), single motherhood (35.32%), and with 60.44% of communities having high
levels of single motherhood.
KwaZulu Natal is noteworthy likewise for having a high incidence of adolescent
pregnancy (above 4%), poverty (65.51%), paternal absence and paternal orphanhood
(occurring among 55.31% and 24.12% of adolescent females, respectively). Further,
KwaZulu Natal had high community levels of single motherhood (46.57%) and the
highest level (36.20%) of adolescent females living in homes headed by single mothers.
The legacy of male migrant labour in South Africa, paternal mortality and low rates of
marriage can largely explain many of these patterns (Ford and Hoosegood, 2005;
Hoosegood, McGarth and Moultrie, 2009).
Data Source and Methods
The data for the study are from the 2013 updated versions of the 2011 to 2013 general
household surveys (GHS). These datasets were extracted from the Statistics South
Africa (StatsSA) database website (2015).The GHS has been conducted since 2002. It is
a nationally representative, cross-sectional survey conducted to determine levels of
development and service delivery across South Africa. Its sampling technique employs a
multi-stage design where primary sampling units (PSUs) are selected first followed by
dwelling units (DUs) in different clusters, and data stratified by geography and
population attributes based on 2001 census data (Statistics South Africa, 2013).The GHS
included samples of 25,086 households in 2011, 25,330 households in 2012 and 25,786
households in 2013. Although individuals of all age groups were included in the surveys,
our focus is on the 14,232 young females aged 10 to 19 years living in 9,286
households (5,207 from 2011, 4,912 from 2012 and 4,113 from 2013). The individuals
included in each year are from independent samples that do not allow the same clusters
to be included in each year. Therefore, the likelihood of the same individual being
present in more than one year is very low.
Adolescent Pregnancy in the past 12 Months:
A single measure for adolescent pregnancy
is used. We define adolescent pregnancy as pregnancy occurring among adolescents
(10-19 year olds). This age group was specifically used in order to include all pregnancies
below the age of 15, which are normally omitted from other studies due to lack of data.
In the survey, individuals were asked to self-report, if female, and were also allowed to
report on behalf of other females within the same household (“any female household
member who had been pregnant during the past 12 months”). This was to help capture
pregnancy in sensitive cases (with responses of Yes, No, Do not know, Not applicable
(for males) and Unspecified). Female household members who answered yes, or who
were identified by other household members from the above question and whose age
was below 20, qualified as pregnant adolescents. A dolescent pregnancy is coded as 1
and adolescent non-pregnancy as 0. Approximately 3% of the young females in our
sample were currently pregnant or had been pregnant within the preceding year,
representing 270,420 adolescent girls at the population level between the years 2011
and 2013 in South Africa.
Independent variables included variables of interest and controlling variables:
Household and Community Family Disruption:
: households with female heads who are not married or
Community levels of Single Motherhood:
municipal level of single motherhood
aggregated from household-level data as a percentage of all households within
the municipality that have single female headedness, and classified into three
equal groups of low, medium and high municipal levels.
The individual-level background variables are race, marital status, educational
level and employment status, defined as follows:
Black, White, Coloured1 , Indian/Asian;
: collapsed into two categories, those never married and those ever
married or currently cohabiting;
: highest level of educational attainment, collapsed into three
categories of not attending school, primary level and secondary level;
: employment status of respondent, employed or unemployed.
Background factors at the household-level include household density and sex
composition, defined as follows:
average number of persons per room in a household
(calculated from the number of persons and rooms per household);
: aggregated from individual-level data of all household
occupants. The five possible categories presented included only females,
predominantly females with males, evenly mixed, predominantly males with
females, and only males.
1 South African government race classification nomenclature for people of mixed race
Community-level background variables are place of residence and province,
which were defined as:
Place of residence:
classification according to settlement type and population
density, defined as either urban if high density and near city centre, or rural if
low density and in tribal areas or away from the city;
current province (local reference to sub-national regions) of residence.
The nine provinces are Gauteng, Eastern Cape, North West, Northern Cape,
Western Cape, Kwa-Zulu Natal, Mpumalanga, Free State, Limpopo.
The rate of adolescent pregnancy was calculated using the equation below for the whole
country and by year:
The chi-square test for trend was used to determine whether the change over time
showed a statistically significant increase or decrease in the linear trend. Bivariate
analysis used cross-tabulations of pregnancy by independent characteristics, with the
chi-squared test (p-value=0.05) used to determine statistically significant differences in
pregnancy between categories of variables.
Multilevel logistic regression models with random intercepts were used to test the
independent association between single female headedness and adolescent pregnancy.
Five models tested the heterogeneity of adolescent pregnancy in different communities
and established the association between single female headedness at household and
community levels, while controlling for socio-demographic individual-, household- and
community-level variables. Multilevel modelling is a suitable statistical technique when
individuals from the same households or geographical areas have the potential of being
included in a study sample (Goldstein, 2011). This indeed is the case for the general
household survey as households from the primary sampling units (provinces) were
sampled using systematic sampling, but every member within the household was
interviewed. However, the model was not nested at the household level due to the
numbers of adolescent females per household ranging between one and three. The two-
level model estimated the variation in the risk of adolescent pregnancy between
communities and between individuals within the same communities. Simple logistic
regression would fail to capture this accurately as members within communities are
similar, thereby violating the logistic regression assumption of independence of residuals
(Kawachi & Subramanian, 2007; Merlo, 2003; Subramanian, 2004). This would result in
an underestimation of standard errors and very small p-values, making estimates of
association appear falsely significant and results erroneous. Our model can be
represented by the following expression:
where: πij=probability of having been recently pregnant for the ith individual in the jth
community –the dependent variable,
δij are the parameter coefficients of the model, zij are the independent regressors,
and εij are the residual errors
Results from multilevel analysis incorporate a fixed and random component. The
fixed component depicts individual factors associated with adolescent pregnancy. These
are presented as odds ratios for ease of interpretation with their associated p -values. A
ratio greater than one implies that an individual in a given category would be more likely
to experience adolescent pregnancy as opposed to an individual in the base category.
The random component quantifies levels of heterogeneity between communities and
indicates the extent to which unexplained community effects are present. Also, results
of the Akaike’s information criteria statistic are included to show the suitability of added
variables in explaining the outcome. We use the linear threshold model (also known as
the latent variable approach) to calculate the variable partition coefficient (VPC) for the
multilevel logistic models.2
Interaction testing was conducted through establishing the levels of certain
variables by the two main interest variables (single female headedness and adolescent
pregnancy) and adding interaction terms in the regression analysis. The unadjusted
multilevel regression analysis of interaction terms was conducted for all possible
interaction terms, with adjusted multilevel regression performed for those interaction
terms that were significant at bivariate level. Results of interaction testing are shown in
the tables. The analyses were performed using the Stata (version 13) statistical
The adolescent females were predominantly never married, black, unemployed and
attending secondary school (Table 2). Most lived in households that consisted
predominantly of females, had low household density (up to three persons per room),
2 This assumes that the underlying binary variable is a continuous latent variable yij, with the
variance at individual level being constant (Browne, Subramanian, Jones, & Goldstein, 2005).
The individual variance is assumed to have a standard logistic distribution, with mean 0 and
variance of π2/3 =3.29.
and were neither characterized by poverty or single female headedness. Additionally,
the majority of adolescent females lived in urban areas, with relatively more in Gauteng
and Kwa-Zulu Natal provinces. A lmost half the sample (47%) came from communities
with low levels of single motherhood.
Table 2: Description of study sample by individual, household and community
Characteristics Sample n=14232 (%)
Age (median; IQR) 15; 4
10-14 year olds 45.88
15-19 year olds 54.12
Never Married 99.71
Ever Married/Cohabitting 0.29
Not Attending School 13.50
Household Sex Composition
Only Females 9.43
Predominantly Male with females 12.76
Predominantly Females with males 59.27
Equal Males and Females 18.51
Place of Residence
Western Cape 12.93
Eastern Cape 11.4
Northern Cape 2.13
Free State 4.94
North West 6.96
Municipal Levels of Single Motherhood
The levels of pregnancy across categories of independent characteristics are
shown in Table 3 with the associated chi-squared test results.
Table 3: Percentage pregnancy across characteristics in South Africa, (GHS
Pregnant n=520 P-value
Age Group 0.00
10-14 year olds
15-19 year olds
Marital Status 0.00
Never Married 3.31
Ever Married/Cohabitting 17.59
Educational Level 0.00
Not Attending School 14.57
Employment Status 0.92
Household Sex Composition 0.00
Only Females 1.91
Predominantly Male with females 4.70
Predominantly Females with males 3.49
Equal Males and Females 2.66
Household Poverty 0.00
Place of Residence 0.00
Western Cape 1.86
Eastern Cape 4.16
Northern Cape 3.94
Free State 2.59
North West 4.80
Single Motherhood 0.00
Municipal Levels of Single Motherhood 0.01
Pregnant adolescents were significantly older with a median age of 18 years as
compared to 15 years of age for non-pregnant females. Overall, 6% of 15-19 year-olds
were pregnant over 2011-13 and the level of pregnancy amongst ever married or
cohabiting adolescent (17.6%) was over five times higher than for those never married
Pregnancy levels were also higher among black adolescents (3.86%) compared
to whites ( 0.47%), as well as among those not attending school (14.57%) than for
those currently in secondary school, with the lowest levels among those attending
primary school (0.25%). There was no statistically significantly difference by
employment status. Household composition showed significant differences in pregnancy
levels across its various categories. Pregnancy levels were highest among adolescent
females living in homes constituted predominantly of males with females (4.70%) and
lowest amongst those living in homes with only females present (1.91). They were also
higher for adolescent females living in poor homes (4.62%) and in rural areas (4.46%)
than among those in homes without poverty or in urban areas.
Provincial differentials of pregnancy ranged from a low of 1.86% for adolescent
females living in the Western Cape to a high of 4.80% for those living in the N orth
West. Adolescent females living in homes headed by their single mothers experienced
significantly higher pregnancy levels (4.71%) than those without single mother heads
(2.78%). Furthermore, pregnancy levels were higher among adolescent females from
communities with high levels of single motherhood (4.07%) than for those living in
communities with low single motherhood.
Multilevel Modelling Results
The random effects and model characteristics are presented first in Table 4 to give
details of the community effects on adolescent pregnancy through multilevel modelling.
Regression analysis tested variability in adolescent pregnancy of 14232 adolescent
females across 66 communities of South A frica. The base model shows results of the
empty model, from which we are able to establish the levels of variability in adolescent
pregnancy across communities. The odds of pregnancy for an adolescent female from
an ‘average’ community (µ0j =0) is 0.04. Therefore, an adolescent female picked at
random from an average community is less likely to be pregnant, and this risk decreases
as we control for further factors to 0.0003 in Model 5. The between community variance
in the log odds of adolescent pregnancy show that adolescent females across
communities differ significantly by 10% with respect to pregnancy and as variables are
controlled for, the level of difference in the risk of pregnancy decreased.
Table 4: Random Effects for Multilevel Logistic Regression of Adolescent
Pregnancy in South Africa, (GHS 2011-2013)
a Statistical Significance: * p<.1; * *p<.05; ** *p<.01
The addition of background factors decreased the variance in pregnancy risk by
5% in all the models. Intra-cluster correlation reveals a 3% significant similarity
between two adolescent females living in the same community. Controlling for more
factors in subsequent models decreases the level of similarity between adolescent
females from the same community to 1% in the fully adjusted model. Model 5 is the
best fit for adolescent pregnancy compared to the other models built, as indicated by
the goodness-of-fit statistics (higher Akaike information criterion (AIC) and log likelihood
The fixed effects results of the unadjusted models for each variable and
hierarchically adjusted multilevel logistic regression models are shown in Table 5. Model
1 represents the bivariate association between the various characteristics and
adolescent pregnancy. All tested variables increase the likelihood of adolescent
pregnancy, except for race where a negative yet significant unadjusted relationship
exists. Model 2 shows the adjusted relationship of demographic and socioeconomic
factors at various levels with the outcome and reveals a similar pattern of association to
that seen in bivariate regression for all variables except employment status, where the
association reverses and becomes negative.
The unadjusted models depict a significant positive relationship between
household- and community-levels of single motherhood with adolescent pregnancy in
South Africa between 2011 and 2013. The fully adjusted model shows that the presence
of a single mother heading a home at the household-level, and high levels of single
motherhood at the community-level, significantly increase the average odds of
adolescent pregnancy by 26% and 22%, respectively. A comparison of Model 2 with
Models 3, 4 and 5 indicates the mediating effects of single motherhood. Models 3 and 4
show that the addition of single motherhood at both the household- and community-
levels increases the magnitude of the association between household sex composition
Random Effects Base Model
Model 2Model 3Model 4Model 5
Constant 0.04*** 0.0003*** 0.0004*** 0,0004*** 0,0003***
Local Municipality Variance 0.10*** 0.05** 0.05** 0,05** 0,05***
Proportional Change in Variance 0.05 0.05 0.05 0.05
Intra Cluster Correlation 0.03** 0.02** 0.02** 0,01** 0,01**
AIC 4446.06 3544.20 3545.17 3544.52 3543.51
Units: Local municipalities 66 66 66 66 66
Units: Individual Teenage Females 14232 14175 14175 14175 14175
Log Likelihood -2221.03 -1747.60 -1746.08 -1747.26 -1745.75
and adolescent pregnancy. Finally, Model 5 (the full model) shows that the magnitude in
the association of marital status, educational status and household poverty decreases,
magnitude of the association with household sex composition and province increases
and that of race, employment status, household density and place of residence largely
remain the same.
Table 5: Multile vel Logistic regression of Adolescent Pregnancy in South A frica, (GHS 2011-2013)
a Statistical Significance: * p<.1; * *p<.05; ** *p<.01
Model 1 Model 2 Model 3 Model 4 Model 5
Single Female Headedness (No†)
Yes 1.34*** 1,20* 1,26**
Municipal Levels of Single Female Headedness (Low†)
Medium 1.15 1.05 1.00
High 1.39*** 1,25** 1,22**
Marital status: (Never Married†)
Ever Married/Cohabiting 5.36*** 2.58** 2,53** 2.55** 2.52**
Coloured 0.72** 0.86 0,88 0,87 0,89
Indian/Asian 0.14** 0.15*** 0,16*** 0,15*** 0,15***
White 0.06*** 0.09*** 0,09*** 0,09*** 0,09***
Educational Level: (Primary†)
Secondary 13.80*** 14.56*** 14,46*** 14,56*** 14,47***
Not Attending School 101.03*** 112.29*** 111,32*** 112,30*** 111,30***
Employment Status: (Unemployed†)
Employed 1.03 0.32*** 0,28*** 0,32*** 0,32***
Household Sex Composition (Only Females†)
Predominantly Male with females 2.92*** 2.62*** 2,91*** 2.62*** 2,90***
Predominantly Females with males 2.16*** 2.04*** 2,18*** 2.03*** 2.18***
Equal Males and Females 1.61*** 1.59* 1,73** 1,59* 1,73**
Household Poverty (No†)
Yes 2.32*** 1.60*** 1.56*** 1.60*** 1.56***
Household Density 1.25*** 1.12*** 1.11** 1,12*** 1,11***
Place of Residence (Urban†)
Rural 1.64*** 1.33** 1.35** 1.32** 1.35***
Province: (Western Cape†)
Eastern Cape 2.23*** 1.88** 1,85** 1,95** 1,92**
Northern Cape 2.12*** 2.39*** 2.38*** 2.51*** 2.50***
Free State 1.66** 1.63 1.62 1.73 1.72
KwaZulu-Natal 1.95*** 1.51 1,47 1,56 1,52
North West 2.29*** 2.06** 2.07** 2,19** 2.19**
Gauteng 1.36 1.38 1.38 1.36 1.35
Mpumalanga 1.77** 1.59 1,58 1,64 1,63
Limpopo 1.88*** 1.64 1,61 1,71 1,69
Constant 0.0004*** 0,0003*** 0,0004*** 0,0003***
The results indicate a possible interaction between household- and community-
levels of single motherhood, education, sex composition and poverty. The cross-
tabulation of household-level factors (single motherhood and sex composition) showed
that 73% of female-only households were headed by single mothers with significant
differences of sex composition by single motherhood. In addition, 76% of households
headed by single mothers had males (either predominantly so or not) and 71%
experienced poverty as opposed to 53% that were headed by others. These differences
were statistically significant. A dditionally, cross-tabulation of community-levels of single
motherhood and poverty showed significantly higher levels of poverty in communities
with greater levels of single mothers.
To further test for interactions, we ran multilevel logistic regression models that
included interaction terms for household and community single motherhood with all
control variables. After controlling for other variables, statistical significance remained
between the interaction terms with education, household poverty and household
composition categories. Results for these six significant interaction terms in the
unadjusted and adjusted models are seen below in Table 6 and 7.
Table 6 shows the results of interaction testing for household single motherhood
with education, household poverty and household composition. The highest risk of
adolescent pregnancy was among adolescent females who were not attending school, at
secondary level education and then at primary education level. In all cases, adolescent
females living with single mothers had a higher risk (in the adjusted model adolescent
females not attending school and with single mothers had a 116 times higher likelihood
of pregnancy compared to adolescent females that were attending primary school and
not from a single mothered home). The magnitude and strength of association were
higher among adolescent females who lived in poverty and with their single mothers,
with poverty exerting a greater influence than single motherhood. Finally, the
interaction term of single motherhood and household composition, again, showed a
higher risk of adolescent pregnancy among young females living with single mothers as
well as with males (be it predominantly so or not).
Table 7 displays interaction testing results for community levels of single
motherhood with education, household poverty and composition separately. Each of
these interactions remained significant in the unadjusted and adjusted multilevel
models. Specifically, the interaction of education with community levels of single
motherhood revealed that adolescent females that were not attending school had the
highest risk of adolescent pregnancy, but this was modified according to the level of
single motherhood within their communities. The association with adolescent pregnancy
was 14% to 21% greater for adolescent females that were not attending school from
communities with medium and high levels of single motherhood, compared to those not
attending school and living in communities with low levels of single motherhood. A
similar pattern of modification of association by community levels of single motherhood
occurred for adolescent females attending secondary and primary schooling.
Table 6: Testing of Single Motherhood Interaction Terms in South Africa, (GHS
a Statistical Significance: * p<.1; * *p<.05; ** *p<.01
Likewise, single motherhood at community level modified the association
between poverty and adolescent pregnancy as the association was greater in poverty
yet increased with the level of single motherhood. In the adjusted model, adolescent
females living in poverty and in communities having medium or high levels of single
motherhood had 6% to 8% higher likelihood of adolescent pregnancy than did their
counterparts, also living in poverty, but in communities with low levels of single
motherhood. All comparisons are with adolescent females not living in poverty and from
communities with low single motherhood levels. Although similar patterns of
Fixed Effects of Interaction Terms n=14232 n=14175
Single Motherhood and Education (No # Primary†)
Yes # Not Attending School 113.32*** 116.29***
No# Not Attending School 105.01*** 105.36***
Yes # Secondary 19.07*** 16.95***
No# Secondary 12.77*** 13.44***
Yes # Primary 1.23 1.07
Single Motherhood and Household Poverty (No # No†)
Yes # Yes 2.64*** 1.65***
No # Yes 2.20*** 1.47***
Yes # No 1.10 0.85
Single Motherhood and Household Composition (No # Only Females †)
Yes # Predominantly Male with Females 3.42*** 2.39**
Yes # Predominantly Female with Males 2.37** 1.69
No # Predominantly Male with Females 2.15** 1.53
Yes # Equal Males and Females 1.95* 1.31
No # Predominantly Female with Males 1.47 1.38
No # Equal Males and Females 1.13 1.12
Yes # Only Females 0.73 0.65
modification seemed to occur for adolescent females not living in poverty, these results
were not statistically significant.
Table 7: Testing of Community Single Motherhood Interaction Terms in South
Africa, (GHS 2011-2013)
a Statistical Significance: * p<.1; **p<.05; ***p<.01
Finally, interaction consistently occurred in the unadjusted and adjusted models
for community levels of single motherhood and household composition. Levels of single
motherhood modified the relationship between sex composition and adolescent
pregnancy, with greater association shown for higher community levels of single
motherhood. However, adolescent females living in homes consisting predominantly of
Fixed Effects of Interaction Terms n=14232 n=14175
Community Single Motherhood and Education (Low # Primary†)
High # Not Attending School 223.53*** 165.22***
Medium # Not Attemding School 187.15*** 151.88***
Low # Not Attending School 119.58*** 130.72***
High # Secondary 29.78*** 21.38***
Medium # Secondary 21.06*** 20.62***
Low # Secondary 20.06*** 16.87***
High # Primary 2.53 2.05
Medium # Primary 1.39 1.19
Community Single Motherhood and Poverty (Low # No†)
High # Yes 2.58*** 1.67**
Medium # Yes 2.46*** 1.63**
Low # Yes 2.34*** 1.53**
High # No 1.39 1.18
Medium # No 0.98 0.93
Community Single Motherhood and Household composition (Low # Only Females†)
Low # Predominantly Male with Females 4.10*** 4.93***
High # Predominantly Male with Females 2.99** 2.50**
High # Predominantly Female with Males 2.96*** 2.49**
High # Equal Males and Females 2.69** 2.42**
Medium # Predominantly Male with Females 2.49** 2.39**
Medium # Predominantly Female with Males 2.48** 2.29**
Medium # Equal Males and Females 1.93 1.89**
Low # Predominantly Female with Males 1.92* 2.12*
Low # Equal Males and Females 1.22 1.65
High # Only Females 1.18 1.32
Medium # Only Females 1.14 1.12
males with females also present, and in communities with low levels of single
motherhood, had almost 5 times (4.93) higher likelihood of becoming pregnant. This
finding needs further examination. Furthermore, the association between household
composition and adolescent pregnancy was attenuated as the community level of single
motherhood decreased. Greater likelihood of adolescent pregnancy was seen for all
participants living with males in their household as compared to their counterparts only
living with females.
The modern day experience of adolescent females in South African society is a distant
cry from the desired ideal. High levels of household poverty and sub-optimal family
structures, as well as violence at the societal level, expose many young females to dire
consequences (Delius & Glaser, 2002; Emmett, 2003). These are more likely to include
difficult relationship dynamics that, in turn, may lead to a higher incidence of coerced
sex, unintended pregnancy, sexually transmitted diseases and HIV (Dunkle et al., 2003;
Jewkes et al, 2010; Wood, Maforah, & Jewkes, 1998). The adolescent pregnancy rate in
South A frica increased from 33 to 38 pregnancies per 1000 adolescent females over
2011-2013, despite numerous prevention campaigns and legislation that has legalized
termination of pregnancy as well as access to contraceptives from the age of 12
(Department of Health, 2012; Macleod & Tracey, 2010; South A frica Department of
Health, 1996). It is imperative to raise awareness and to improve all aspects of access
to family planning services in order to help adolescents avoid unwanted pregnancies.
Our results show that after controlling for background demographic and socio-
economic factors, a positive significant relationship with adolescent pregnancy remains
for single motherhood both at the household- and community-levels. These results
confirm earlier findings from Southern Africa and elsewhere that adolescent females
living with a divorced or separated female have a higher risk of non-marital
childbearing. For example, Brahmbhatt et al. (2014) found that being raised by a single
parent increased the odds of pregnancy 18 times among teenage females in
Johannesburg, South Africa.
Previous literature has posited a number of pathways through which these
findings may occur. Firstly, the family enables socialisation of children and adolescents
including the transfer of values and norms such as those concerning parental views
concerning sex and contraception (Moore & Chase-Lansdale, 2001;) Muindi, 2007).
Single parented family structure may elevate predisposition to premarital birth due to
early sexual socialisation (Lee 2001; Powers, 2005). These results may also be due to
the risk posed through intergenerational adolescent motherhood. Studies from many
other parts of the world indicate that daughters of adolescent mothers are significantly
more likely than daughters of older women to have an adolescent childbirth (Meade,
Kershaw, and Ickovics, 2008)Santos and Rosário, 2011). .
Another form of the socialisation theory states the importance of parental
presence in the development of children’s personality (McLanahan and Bumpass, 1988).
The 2011 South Africa census indicated that 95% of female adolescents living in single
parented households were not living with their fathers. Paternal absence is the most
common type of parental absence globally, but despite the social expectation of
mothers as primary care givers (McLanahan and Bumpass, 1988), fathers can be just as
important as mothers in the development and well-being of children (Luo et al., 2012).
For example, an earlier study in Cape Town, South Africa, indicated that not living with
one’s biological father almost tripled (RR=2.62) the risk of adolescent pregnancy
(Vundule et al., 2001). Research has found paternal absence to increase the risk of
identity crisis, low self-esteem, the dependence of daughters on men, early sexual
debut and adolescent pregnancy (Franklin, 1988; Luo et al., 2012; McLanahan &
Bumpass, 1988). Paternal presence assists in cognitive development, nullifies the need
for compensatory masculine attention and ensures strict relationship rules for daughters
with the opposite sex. Validation of the girl child from the male parent has a lasting
impression on her personal value and raises confidence levels. Paternal absence also
affects the reassurance of identity.
It is important to understand the systematic creation of the phenomenon of
children growing up without their fathers because of South Africa's past. Programmes
need to be set up for such children by non-governmental organisations (NGOs) or other
stakeholders to ensure father figures are present in such children's lives and, if absent,
children should be paired up with suitable "substitute" father figures. Screening is
necessary to mitigate against the possibility of sexual or physical abuse and other
problems. The "substitute" father’s main function would be to act as a form of support,
love, validation and encouragement for the child. This could be facilitated with the aid of
relatives. Programmes could also incorporate organisation of activities and meetings
with these "substitute" father figures and their "adopted" children periodically, beyond
encouraging private supportive relationships. Measures to reduce adolescent pregnancy
among females growing up with single mothers could include the provision of some
forms of special assistance for single mothers. For example, to help them navigate the
journey of cautiously finding love again, in the presence of children, this could include
special needs that some single mothers may not be fully aware of, such as counselling
on introducing children to potential mates only once they are sure of the direction of a
This study has further highlighted the mediating effect of single motherhood on
the association between adolescent pregnancy and individual, household, and
community factors. Findings indicate the strong interaction between both household and
community levels of single motherhood with education, poverty and sex composition.
This includes the additive effects of single motherhood and nonattendance of schooling
on the risk of adolescent pregnancy, although the cross-sectional nature of the data
precludes determination of the direction of these effects.3. The higher risk of pregnancy
for adolescent females completing secondary education as compared to primary school
attendees is a function of the correlation of educational level with age in the data. As
shown in the results, 15 to 19 year-olds had 23 times higher levels of pregnancy than
the younger girls who would still be attending primary school. A ge and educational level
variables were significantly correlated (0.86).
The risk of adolescent pregnancy also appears to be highest in the presence of
single motherhood and poverty. Cross-tabulation of poverty and single motherhood
found that 75% of single mother-headed households were poor and that communities
with medium or high levels of poverty comprised 66% of single-motherhood headed
households, which were significant results according to the chi-squared test. This
predisposition of single female-headed households to poverty could be rooted in three
intrinsic characteristics of such households, as suggested by Buvinić and Gupta (1997).
Firstly, these households may experience greater economic burdens than do their male-
headed counterparts, because they often have more dependents. Secondly, the female
headship of the household introduces economic vulnerability because women generally
have lower wages, fewer resources, lower assets and worse employment opportunities.
Finally, there exists an unfavourable combined effect of household characteristics and
female headship, which leads to such females working fewer hours in lower-skilled
employment to ensure their availability to perform household chores and domestic care-
giving duties (Buvinic and Gupta, 1997). It has been shown for South Africa that “when
young children grow up in residential areas where poverty is entrenched, they are at
risk of experiencing an early pregnancy” (Panday et al., 2009). A participant in another
study suggested that: “Insufficient spaces in the family where parents are sharing a
room or a shack with their children tend to expose children to sexual activities” (Makola,
2011). This problem may be cyclic in nature as adolescents in such homes may choose
to escape the hardship through transactional sex, premarital childbearing, leaving home
or cohabitating with the father of their child (McLanahan and Bumpass, 1988).
Consequently, single female-headed households need to be supported, socially and
financially, in order to help mitigate poverty as well as unwanted pregnancy among the
home’s female dependents. Initiatives led by NGOs or community leaders could also
assist such households in becoming better accepted and integrated in society.
The interaction of single motherhood with sex composition is that which is least
understood. Female adolescents from households having females only were shown to
3 Also, some cases may have dropped out due to pregnancy in the past year.
have the least likelihood of pregnancy, which was highest for households with men and
increased with the proportion of men. The relationship of these males to the adolescent
females in question is unclear and needs further investigation, including by well-
executed, qualitative, research. In addition, this study used cross-sectional data, and
cannot establish causation, with pregnancy in the preceding year used as the outcome
to ensure a short time lapse between the outcome and tested variables. Future studies
investigating similar phenomena would benefit greatly from using a longitudinal and
mixed methods approach to establish the correct sequence of events and understand
the links of these issues in the local setting of South Africa.
This study has shown the complex association and mediating effects of single
motherhood on adolescent pregnancy in South Africa. The apartheid regime
systematically disrupted the fundamental structure of families, particularly those of black
Africans, by effectively manufacturing the phenomenon of single female headedness
and normalizing paternal absence nationally. Women gradually became accustomed to
raising and fending for offspring alone. The responsibility of raising a child is already
immense for two loving parents even with the assistance of their families and relatives.
Much greater strain can only be expected when a single mother, heading a household,
has to shoulder this responsibility by herself. Early pregnancy prevention programmes
and awareness campaigns for females growing up in homes and environments with
greater levels of single motherhood need to be strengthened as a matter of urgency to
assist in ensuring the sexual and reproductive health of young females in South Africa.
Brahmbhatt, H., Kagesten, A., Emerson, M., Decker, M. R., Olumide, A., Ojengbede, O.,
Lou, C., Sonenstein, F., Blum, R., & Delaney-Moretlwe, S. (2014). Prevalence and
Determinants of Adolescent Pregnancy in Urban Disadvantaged Settings A cross
Journal of Adolescent Health,
Browne, W., Subramanian, S., Jones, K., & Goldstein, H. (2005). Variance Partitioning in
Multilevel Logistic Models that Exhibit Overdispersion.
Journal of the Royal
Statistical Society, 168
Budlender D. and Lund F. (2011). South A frica: A Legacy of Family Disruption.
Development and Change
Bunting, B. (1964).
The Rise of the South African Reich
(Vol. 2). London Penguin Books
Buvinić, M., & Gupta, G. (1997). Female-headed Households and Female-maintained
Families: are They Worth Targeting to Reduce Poverty in Developing Countries?
Economic development and cultural change, 45
Carlson, B., & Davis, L. (1977). Prevention of Domestic Violence: An Ecological Analysis.
Social Service Review, 58
Centre for Social Development Humanitarian Affairs, UNICEF, United Nations Population
Fund, of, U. N. D. F. f. W. S. D., United Nations, O. f. P. S., World Food
Programme, . . . World Health Organization. (2010).
The World's Women: Trends
. New York: United Nations.
Delius, P., & Glaser, C. (2002). Sexual Socialisation in South Africa: A Historical
African Studies, 61
Department of Health. (2012). National Contraception Clinical Guidelines. Pretoria.
Dlamini, N. S. (2010).
Measurement and characteristics of single mothers in South
Africa: analysis using the 2002 general household survey.
Domenico, D., & Jones, K. H. (2007). Adolescent Pregnancy in America: Causes and
Journal for Vocational Special Needs Education, 30
Dunkle, K., Jewkes, R., Brown, H., McIntyre, J., Gray, G., & Harlow, S. (2003). Gender -
based violence and HIV infection among pregnant women in Soweto.
Health Group, Men on relationships with and abuse of women, Medical Research
Council Technical Report, Medical Research Council, Tygerberg
Edin, K., & Tach, L. (2012). Becoming a Parent: The Social Contexts of Fertility during
Young Adulthood. In A. Booth & W. Manning (Eds.),
Early Adulthood in a Family
(pp. 185-207). New York Springer.
Emmett, T. (2003). Social Disorganisation, Social Capital and Violence Prevention in
South Africa: Original Contribution.
African Safety Promotion, 1
(2), p. 4-18.
Essien, A ., & Bassey, A . (2012). The social and religious challenges of single mothers in
American Journal of Social Issues and Humanities, 2
Ford, K., & Hoosegood, V. (2005). AIDS mortality and the mobility of children in
KwaZulu Natal, South Africa.
Francis, G. (2008).
The Effect of Household Characteristics on Adolescent Childbearing
(Master of A rts), University of the Witwatersrand, Johannesburg.
Franklin, D. (1988). Race, class, and adolescent pregnancy: An ecological analysis.
American Journal of Orthopsychiatry, 58
Goldstein, H. (2011).
Multilevel Statistical Models
(Vol. 922). Hoboken, New Jersey: John
Wiley & Sons.
Gottlieb, S. L., Low, N., Newman, L. M., Bolan, G., Kamb, M., & Broutet, N. (2014).
Toward global prevention of sexually transmitted infections (STIs): The need for
(14), 1527-1535. doi:
Gustafsson, S., & Worku, S. (2006). Marriage markets and single motherhood in So uth
Africa: Tinbergen Institute.
Holborn, L., & Eddy, G. (2011). First steps to healing the South African family. In S. A. I.
o. R. Relations (Ed.). Johannesburg: South African Institute of Race Relations.
Hoosegood, V., McGarth, N. & Moultrie, T. (2009). Dispensing with marriage: Marital
and partnership trends in rural KwaZulu Natal, South Africa 2000 -2006.
20 (13): 279 -312
Jewkes, R., Morrell, R., & Christofides, N. (2009). Empowering teenagers to prevent
pregnancy: lessons from South Africa.
Culture, health & sexuality, 11
688. doi: 10.1080/13691050902846452
Jewkes, R. K., Dunkle, K., Nduna, M., Jama, P. N., & Puren, A. (2010). Associations
between childhood adversity and depression, substance abuse and HIV and HSV2
incident infections in rural South African youth.
Child abuse & neglect, 34
Kawachi, I., & Subramanian, S. (2007). Neighbourhood influences on health.
Epidemiology and Community Health, 61
(1), 3-4. doi: 10.1136/jech.2005.045203
Lee, M. (2001). Family and Adolescent Childbearing.
Journal of Adolescent Health,
Limpopo Population and Development Directorate. (2012). Factors associated with
teenage pregnancy in Limpopo province. In Department of Population and
Factors Associated with Teenage Pregnancy in South Africa
Polokwane: Department of Population and Development,.
Luo, J., Wang, L. G., & Gao, W. B. (2012). The influence of the absence of fathers and
the timing of separation on anxiety and selfesteem of adolescents: a
Child: care, health and development, 38
Macleod, C. (2011).
'Adolescence', Pregnancy and Abortion-Constructing a Threat of
. New York: Routledge, Taylor & Francis Group.
Macleod, C., & Durrheim, K. (2002). Racializing Teenage Pregnancy:'Culture' and
'Tradition' in the South A frican Scientific Literature.
Ethnic and Racial Studies,
Macleod, C., & Tracey, T. (2010). A decade later: follow-up review of South A frican
research on the consequences of and contributory factors in teen-aged
South African Journal of Psychology, 40
Makola, M. (2011).
Teenage Pregnancy: Views of Parents/Caregivers, Teenagers and
Teachers at Two High Schools in Soweto, Gauteng.
(Masters of Arts), University
of the Witwatersrand, Johannesburg, Johannesburg.
Mbanefo, C. (2013).
Levels and Correlates of Single Motherhood in Southern Africa.
(Master of A rts), University of the Witwatersrand, Johannesburg.
McLanahan, S., & Bumpass, L. (1988). Intergenerational Consequences of Family
American Journal of Sociology
Meade, C., Kershaw, T., & Ickovics, J. (2008). The intergenerational cycle of teenage
motherhood: an ecological approach.
Health Psychology, 27
Merlo, J. (2003). Multilevel Analytical A pproaches in Social Epidemiology: Measures of
Health Variation Compared with Traditional Measures of Association.
Epidemiology and Community Health, 57
Mkhwanazi, N. (2010). Understanding Teenage Pregnancy in a Post-Apartheid South
Culture, health & sexuality, 12
(4), 347-358. doi:
Moore, M., & Chase-Lansdale, P. (2001). Sexual Intercourse and Pregnancy among
African A merican Girls in High-Poverty Neighborhoods: The Role of Family and
Perceived Community Environment.
Journal of Marriage and Family, 63
1157. doi: 10.2307/3599820
Moyo, O., & Kawewe, S. (2009). Lone Motherhood in Zimbabwe: The Socioeconomic
Conditions of Lone Parents and their Children.
Social work in public health, 24
Muindi, K. (2007).
Adolescent Sexual Behaviour in Navrongo: Does Family Count?
(Master of Science), University of the Witwatersrand, Johannesburg.
Omar, K., Hasim, S., Muhammad, N., Jaffar, A., Hashim, S., & Siraj, H. (2010).
Adolescent pregnancy outcomes and risk factors in Malaysia.
Journal of Gynecology & Obstetrics, 111
Panday, S., Makiwane, M., Ranchod, C., & Letsoalo, T. (2009).
Teenage Pregnancy in
South Africa: With a Specific Focus on School-going Learners
Department of Basic Education.
Powers, D. (2005). Effects of family structure on the risk of first premarital birth in the
presence of correlated unmeasured family effects.
Social Science Research,
Reddy, S., James, S., Sewpaul, R., Koopman, F., Funani, N., Sifunda, S., . . . Omardien,
R. (2010). Umthente Uhlaba Usamila-The 2nd South African National Youth Risk
Behaviour Survey 2008
National Youth Risk Behaviour Survey
. Cape Town: South
African Medical Research Council.
Santos, M., & Rosário, F. (2011). A score for assessing the risk of first-time adolescent
Family Practice, 28
Choice on Termination of Pregnancy Act, 92 of 1996 (1996).
South Africa Ministry in the Office of the President. (1994).
White Paper on
Reconstruction and Development
. Cape Town: Government gazette 16085,
Statistics South Africa. (2013). General Household Survey, 2012
Pretoria: Statistics South Africa.
Subramanian, S. (2004). The Relevance of Multilevel Statistical Methods for Identifying
Causal Neighborhood Effects.
Social Science & Medicine, 58
Timaeus, I. & Moultrie T. (2015). Teenage Childbearing and educational Attainment in
Studies in Family Planning,
Ugoji, F. (2011). Parental marital status and peer influence as corelates of teenage
pregnancy among female teens in south-South Nigeria.
Gender and Behaviour,
UN (United Nations), 2015.
The World's Women 2015: Trends and Statistics
. New York:
United Nations, Department of Economic and Social Affairs, Statistics Division.
Vundule, C., Maforah, F., Jewkes, R., & Jordaan, E. (2001). Risk factors for teenage
pregnancy among sexually active black adolescents in Cape Town.
Medical Journal, 91
Weeks, J. (2011).
Population: An Introduction to Concepts and Issues
Belmont, California: Wadsworth, Cengage Learning.
Willan, S. (2013). A Review of Teenage Pregnancy in South A frica–Experiences of
Schooling, and Knowledge and Access to Sexual & Reproductive Health Services.
Cape Town: Partners in Sexual Health.
Wood, K., Maforah, F., & Jewkes, R. (1998). “He forced me to love him”: putting
violence on adolescent sexual health agendas.
Social Science & Medicine, 47