When the mother is a child : the health impacts of having an adolescent mother : evidence from Guinea

(1)

UNIVERSITY OF AMSTERDAM

AMSTERDAM SCHOOL OF ECONOMICS

MASTER’S THESIS

2015-2016

When the Mother is a Child:

The Health Impacts of Having an Adolescent Mother - Evidence from

Guinea

Abstract: This paper aims to establish whether the age of a mother at time of birth, or at first birth has an impact on the occurrence of stunting, wasting and female genital mutilation (FGM) amongst her children. It uses data from the Demographic and Health Survey collected in Guinea in 2012 and relies on estimating results using a mother fixed effects model. The findings are that children born to adolescent mothers face a much higher likelihood of being stunted or wasted. FGM is not determined by mother’s age at birth. Although the results are not significant, they help shed some light on the ill effects of adolescent motherhood and further research could help guide policy interventions to curb this occurrence.

Author: Mayanka Vij Student Number: 11088958 Supervisor: Prof. dr. Erik Plug

(2)

Statement of Originality

This document is written by Mayanka Vij, who declares to take full responsibility for the contents of this document.

I declare that the text and the work presented in this document is original and that no sources other than those mentioned in the text and its references have been used in

creating it.

The Faculty of Economics and Business is responsible solely for the supervision of completion of the work, not for the contents.

(3)

1. Introduction

There exists a general consensus in most developed nations about the ill effects of adolescent childbearing for both, the mother as well as her offspring. The social and economic costs that accompany adolescent pregnancy are heavy and have a clear and direct sequence of consequence. Adolescent pregnancy is one of the major contributors to maternal and child mortality and adds to the cycle of poverty and ill health within a country. In low and middle-income countries, complications of pregnancy and childbirth are the leading causes of death among women aged 15-19 years.1 However, despite these overwhelming statistics, several regions in the world display strikingly high fertility rates among young girls.

Research on the health and nutritional outcomes of the children born to adolescent mothers is abundant in developed countries, yet remains scarce in developing and least-developed countries. There have not been many studies that examine the impact of being born to a mother who was just a child (between 15-19 years) when she first gave birth in a society where child marriage and subsequent adolescent pregnancies are considered acceptable social norms. In societies such as these, girls are not only disadvantaged financially, but also socially in terms of basic human rights and health violations. This brings me to the main research question driving this study; what is the impact of being born to an adolescent mother, in terms of the health and nutritional outcomes of her children?

Adolescent pregnancy has several key determinants. Sexual activity among girls is generally initiated under the context of marriage (here, girl child marriage), pregnancies arising from the low utilization of contraception (the contraction controversy), societal expectations about marriage and fertility especially for girls, lack of family planning knowledge and low education are some of them. These determinants all display the highest prevalence in Western Africa, which leads me to the country that I would like to investigate for this study; Guinea (Table 1), a country in West Africa with the adolescent fertility rate (births per 1000 women aged 15-19) at 142 - 5th highest in the world.2

1_{WHO Factsheets - Adolescent Pregnancy, September 2014.}

(5)

The research carried out on the outcomes of these adolescent mothers reveal that they are at a higher risk of health disorders, pregnancy complications, HIV/AIDS and are less equipped financially as well as psychologically to raise their child as compared to their older counterparts. However, as stated by Levine et. al. (2001), most focus is on the contextual disadvantage that stems from early parenthood but not on the disadvantages that may lead to adolescent motherhood in the first place. Through this study, I would like to address this disadvantage leading to early motherhood through the channel of social vulnerabilities present in Guinea where young girls are married at an early age and may or may not have a choice in childbearing. How do societal norms and social selection of an adolescent mother under this context enter into her child’s outcomes? In societies where long-entrenched, collective norms dictate an individual’s life decisions such as marriage and childbearing, how do the children of young mothers differ from those of older mothers? Does the age of the mother at the time of birth play a role in determining if her child would be stunted or are other mediating factors such as the presence of extended families for support enough to offset the limitation she faces in terms of less experience and resources to provide her child with adequate nutrition so as to prevent such adversities? Since most studies focus on this question from a family planning perspective, I feel it is understudied from an economic standpoint in low-income countries. To track the nutritional status of these children born into an inherently disadvantaged society along with having a mother who is barely an adult is a research area that I feel deserves notice. It can have large policy implications if this trend of ‘kids having kids’ is a perpetuating one, validating policy makers to not only implement plans that protect and economically further families like these but go one step behind and protect at-risk girls to prevent the entire cycle altogether.

Table 1 (Guinea)

Background Characteristics

% of

Women Married by Age 18 % of Women Married by Age 20 Median Age At First Marriage

1999 2005 2012 1999 2005 2012 1999 2005 2012 Education None 78.3 85.0 75.2 89.3 91.5 86.1 16.2 15.9 16.5 Primary 68.2 69.6 60.7 79.3 82.3 72.9 17.3 17.1 17.7 Secondary or Higher 37.7 41.7 28.5 53.4 56.4 38.7 20.5 20.1 24 Total 73.7 80.4 66.6 85.1 88.8 77.6 16.5 16.1 17.1

(6)

Finally, my main hypothesis is to find that the children that were born to mothers who were between the ages of 11-19 years3 would fair lower on most health indicators than the children born to older mothers. However, since the context of this study differs greatly from the context of previous studies in terms of existing family structures and lower or no social stigma surrounding teenage pregnancy hence less stress for the young mother, it is also possible that no significant difference may be observed between the children of the two groups of mothers.

The rest of this paper is structured as follows; Section 2 reviews existing literature on the subject, Section 3 provides details on the data used and the outcome, control and endogenous variables studied. Section 4 talks about methodology employed, Section 5 lists the empirical results, Section 6 provides a discussion on these results and Section 7 concludes this paper.

2. Literature Review

Frongillo et. al (1997) aimed to establish worldwide trends in stunting and wasting of children, using a conceptual model that built on three types of causes of the health disorders; namely Basic causes (political structure, economic structure, formal/informal institutions, resources), Underlying causes (maternal health, food security, health services/environment) and Immediate causes (diet and health). This model is relevant for my study because I would like to exploit the extent of the impact of an underlying cause (maternal health) on adverse health outcomes while controlling for some of the other causes as mentioned above. One would assume that a young girl having a child between the ages of 15-19 years would undergo severe health and psychological complications. These would pass on to her child, either genetically or through the channel of upbringing.

Pogarsky et. al. (2006), use a ‘Life Course Perspective’ as a framework stating that, deviations (here, adolescent pregnancy) from major life course trajectories lead to risks for the individuals and their families. It works through different channels - disrupting mother’s formation of social and human capital, low educational attainment due to early parenthood and hence later financial disadvantage, disorders in family formation and stress from the structural disadvantage of being a young mother. Hence, a mother’s age at the time of birth should have an impact on the health of her child either directly through her incapability’s or indirectly through her worse health.

3_{Even though adolescent pregnancies are classified from 15 years onwards, the minimum age at birth reported in}

this sample was 11 years. Hence, the group of adolescent mothers also includes girls who had their children between the ages of 11-15 years.

(7)

Broadly speaking, there have been two generations of studies that have looked at the relationship between the well being of the children born to young mothers. The first generation of studies did not account for the potential selectivity bias of young girls into adolescent motherhood and hence found largely overstated negative effects of early childbearing for both the mother and her child. The second generation of studies surrounding this topic takes into consideration the potential bias of selecting young mothers who are likely to be less economically equipped than older mothers and hence their children fare worse because of the existing inequality and financial disadvantage. This disadvantage is the cause for the selection bias, which needs to be controlled for by finding suitable comparison groups of children and their mothers to be able to credibly deduce any findings or associations between the mothers’ age and her child’s outcomes.

In their study, Abdullah et. al (2007), seek to determine whether there exists a difference in the health of children born to mothers in the age groups of 15-19 years, 20-24 years and 25-29 years. Their main comparison groups were the children of the mothers from the first and last two groups and the children from the middle group served as an interim group to determine which group they were more similar to. Using a logistic regression analysis, they find that in Bangladesh, a country with similar health and economic characteristics as Guinea, the children born to adolescent mothers were more likely to be stunted, underweight and hospitalized for longer durations compared to the children of older mothers who were more likely to be immunized and have mothers with higher literacy rates. Since all mothers who brought their children to the public hospital came from similar financial backgrounds, the significant difference between the children was the age of their mothers, controlling for the potential disadvantages that could have arisen due to reasons that were not financial.

Another study by Marteleto and Dondero (2013) looks at the implications of maternal age at first birth for the education of her children in Brazil. Realizing the numerous confounding factors that could influence the educational outcomes of these children, they utilize an instrumental variable approach, using the teen mother’s miscarriage as an instrument to measure the impact of mother’s age on child outcomes. The instrument allows the creation of two groups of children; the first where the mothers gave birth as adolescents and the second who gave birth as adults but would have had children earlier but were unable to because of the miscarriage. This identification strategy worked as it created two comparison groups of mothers with similar background characteristics and the only reason in the different ages at birth was the naturally occurring miscarriage. Through their 2SLS method, they find that children born to teen mothers exhibit a less favorable socio-economic profile; they come from families with low incomes, their mothers have lower levels of education and the children themselves display a higher age-grade disparity than the children of mothers who didn’t have a

(8)

child when she was an adolescent. These children are also found to be at a higher risk of grade repetition.

Using such biological fertility shocks as instrumental variables to address the endogeneity of motherhood timing has proved to be an effective method. A study by Miller (2008) exploits these shocks and one of the instruments utilized by her is one that I make use of as well to set-up a sub-sample to reduce selection bias of the young mothers. She constructs an indicator variable of contraception use at the time of pregnancy which allows her to differentiate between two kinds of mothers, one where the mother indicates yes for contraception when she got pregnant and the second who indicates no for the same. The first indicator acts as a proxy for an accidental pregnancy, as mothers were using contraceptives and did not intend to have a child then. She addresses the potential pitfalls of using self-reported variables as key indicators in her study by cross-referencing them against data from a family planning survey and finds matching observations. Miller through her study estimates that a year of motherhood delay leads to a 0.927 increase in test scores of her first-born child.

Two important points to remember in the context of Guinea are the role of household structure and the socio-cultural context of the area under study. In households with multiple adults who can act as caretakers of the child, it is possible that effects of a young mother might not be apparent since the child is mainly brought up by another adult such as a grandparent. This links to second point of taking into account the social and cultural values that prevail in the society being studied. In societies where early childbearing is normative and an intentional life course strategy, children might face lower risks of adverse outcomes than in areas where it is not due to low psychological stress.

(9)

3. Data and Outcome Variables

For this study, I used the Demographic and Health Survey (DHS) data collected from Guinea in 2012. The DHS are nationally representative surveys of women of childbearing age (i.e., 15-49 years) carried out in developing countries and are well suited for conducting analyses on maternal and child health.

The primary sample here was constructed using the main data source from the DHS program that was classified on the basis of survey questions that covered maternal health and fertility. The original sample contained 22,900 observations, i.e., data on 22,900 children that were part of the households that were surveyed. This sample was created using detailed health, nutritional and medical information of the last 5 children of each mother. Keeping this in mind, to accurately track the impact of teenage motherhood from the first child of each mother, households with no children or with more than 5 children were excluded from the sample. Only observations with accurate data on the height and weight of children were included, and those that reported an answer to the question on Female Genital Mutilation4 (henceforth, FGM). This led to a loss of 13,107 observations and a final sample of 9,793 children5.

No systematic differences were observed in the sample on the basis of gender of the child or the wealth index6 of a household; however, some differences were observed on the basis of region of residence (Appendix - 1).

The outcome variables analyzed pertain to the health of the children being studied. They are Stunting, Wasting and the prevalence of FGM (among female children).

Stunted growth and Wasting in children have several long term and short-term consequences. Their immediate effects appear in the form of increased health costs for an unwell child, risk of morbidity and reduced cognitive development. In the long run, these health disorders manifest themselves in various ways such as decreasing earning and working capacity and poor reproductive health. Both disorders are representative of the health and nutritional history of the population since they are born out of a culmination of different factors, some of which are experienced by all members of the community. Stunting occurs when the child’s height-for-age ratio is below -2 standard deviations of the reference median. Similarly, wasting occurs when the child’s weight-for-height ratio is below

4_{WHO defines}_{female genital mutilation as procedures that intentionally alter or cause injury to the female genital}

organs for non-medical reasons.

5_{Complete and accurate data on the height and weight of children was available for 3,058 children. For the FGM}

outcome variable, the question was posed only to the mothers with female children and out of 11,085 observations, 8,166 reported an answer.

6_{“The wealth index is a within-country measure of the wealth of the household relative to other households in that}

survey based on its ownership of household assets. This measure has been standardized by Measure DHS across most of the DHS and is widely used as a measure of relative wealth within a country.” (DHS Recode Manual 6)

(10)

-2 standard deviations of the reference median. In the dataset the reference median is set according to WHO Child Growth Standards median and the height-for-age and weight-for-height ratios are calculated. The standard deviations of these ratios from the reference median are then used for this study to construct dummy variables for whether or not a child falls below the -2 SD threshold, making him/her stunted or wasted.

FGM finds its roots in religious, hygienic and traditional justifications7, however it remains banned by law in Guinea (1965, 2000). The reason the age of the mother can have an impact on if her child underwent FGM is because younger mothers are easily pressured and are in a powerless position to prevent the practice on their daughters if the decision is being made by the elders of the family who are more likely to believe in such outdated practices. Teen mothers are also more likely to drop out of school and have lower education and awareness levels than older mothers and hence might even be unaware of the harmful effects of FGM or may just be unable to form their own opinion on the practice and hence follow the established norm where FGM is accepted.

The FGM variable is again a dummy variable constructed on the basis of the answers to the questions asked during the survey. A girl child will fall under the category of ‘Underwent FGM’ if her mother answered yes to that question and will not if her mother answered no.

There are two independent variables in this study. The first one is whether a mother was ever a teenager when she had a child. This includes all mothers who had children when they were teenagers, irrespective of whether or not those specific children are being analyzed. The second one is whether a mother was an adolescent at the time of birth of the specific child being analyzed. This differentiation between the two variables allows us to examine whether the effect of young motherhood is only on the child in question and fades away as the mother grows older, or whether the effect carries over onto subsequent children as well. Two questions from the survey make it possible to set up the dummies for these two variables. 8

Adolescent pregnancy is expected to have a “carry-over” effect on all her children because many young girls who become pregnant might have to drop out of school in a society where termination options are not easily accessible and may also be considered religiously unethical. In such cases, she is left with fewer skills and human capital to help her find a job later on. This leads to a financial and socio-economic disadvantage for herself and her children in terms of the facilities she can access, nutrition she can afford etc. Such a carry over effect can be checked for using the birth order of a child (Appendix 2).

7_{UNFPA groups the reasons why FGM is performed under 5 categories, namely, Psychosexual, Cultural &}

Sociological, Hygiene & Aesthetic, Religious and, Socio-Economic factors.

8_{Age of the respondent (mother) at first birth;}

(11)

Control variables here are important since the health outcomes of a child, specially height and weight based measures are easily influenced by external factors. Since on average, boys tend to be taller than girls, controlling for the gender of the child when looking at Stunting and Wasting allows us to control for this natural variation. However, some studies have found that boys under the age of five are also more likely to be stunted than girls (Wamani et al., 2007), hence making the gender and age of the child important to control for. Controlling for the age of the child also allows us to control for the natural variation in height and weight that occur, as children grow older. Where a child is brought up can also have an impact on the outcome variables being analyzed. Rural areas display greater following of traditions and old practices such as FGM vis-à-vis urban areas where such practices would be considered barbaric (In this sample 72% of the girls on whom FGM was performed live in rural areas versus the 27% who live in urban areas). Hence controlling for region of residence lets us hold constant the difference in attitudes of people and only examine the effect of the mothers’ age.

As mentioned earlier, family structure plays an important role in whether a mother’s age has an impact on her children due to the availability of other adults to raise the child. Mednick & Baker (1980), find that children of adolescent mothers who lived with their grandmothers had a better mean health score than the children who lived with biological mothers who weren’t adolescents. This was because these mothers lacked the parental support from grandparents but also lacked the expertise that came with maturity and age of mothers older than 30 years. Hence, by observing changes in the estimates after controlling for the family structure or caretaker of the child, the impact of mother’s age and familial support can be separated. Similarly, the wealth of a household must be controlled for to see how estimates change when the wealth index (acting as a proxy for access to better health services) is included to separate the impact of mother’s age and financial background on the health of her child.

From the descriptive statistics displayed in Table 2, it can be seen that children of younger mothers display slightly higher sample means on outcome variables that those of older mothers, however, the difference is not a lot. Similarly, mother characteristics also do not differ much between teenage and adult mothers. The majority of the population sampled would be classified under rural regions, however, given the ratio of rural-urban population in Guinea9; this is partly resolved by the weights set to take into account the sampling design. Teenage mothers on average display more years of completed education than older mothers, though the overall years of completed education of the sample is extremely low, with 83% of the sample reporting 0 years of completed education.

(12)

The age of the mothers in this sample at first birth ranges from 11-42 years with 73% of mothers reporting their first child by the age of 19 years. The average of mothers at the time of birth of specific children is 25.12 years with 21% of the sample being 19 years or less.

(13)

Variable

Teenage Mothers (≤19 years) Adult Mothers (>19 years)

Variable Description

Observations Mean SD Observations Mean SD

Child Characteristics/ Outcome Variables

Stunting 554 -1.108 1.594 2539 -0.929 1.737 SD of Height-for-Age Ratios

Wasting 554 -0.576 1.144 2539 -0.470 1.228 SD of Weight-for-Height Ratios

Female Genital Mutilation 1576 0.521 0.499 6590 0.497 0.500 1=Yes 0=No Mother’s Characteristics Education in Years 1871 1.701 3.636 7922 1.090 2.995 Min=0 years Max=19 years FGM 1871 0.987 0.112 7922 0.994 0.145 1=Yes 0=No Family Characteristics Place of Residence 1871 1.704 0.456 7922 1.724 0.440 1=Urban 2=Rural

Wealth Index 1871 2.781 1.393 7922 2.796 1.358 1-5 in Increasing Order of Wealth

(14)

4. Methodology

In the subsequent section, I explain the identification strategies used in this paper and their potential limitations.

4.1 Specification

The regression model used to examine the impact of mothers’ age on child outcomes rests on two right hand side variables; hence the two different regression equations.

Y

ij

= α

1

+ ß

1

X

ij

+ ρ

1

W

ij

+ ε

1ij (1)

Y

ij

= α

2

+ ß

2

X

i

+ ρ

2

W

ij

+ ε

2ij

(2)

In both of the above specifications, Yij represents the outcome variables (Stunting, Wasting,

FGM) of interest for child i born to mother j. Wij represent the control variables (simple and

additional).

In the first equation, Xij is the variable for whether a mother was between 11-19 years at the

time of birth of that specific child and in the second equation; Xi is the variable for whether the

mother ever had a child as an adolescent.

From this,

ß

1 and

ß

2are the parameters of interest as they capture the impact of the mothers’ age on the outcomes variables.

ß

1would capture the direct effect of adolescent motherhood on health of her child whereas

ß

2would capture the carryover effects of young motherhood onto subsequent children, if any. For each outcome, the regressions are run twice, once with a basic set of control variables and the second time with an extended set of control variables and the changes in the estimates are compared.

As mentioned earlier, to credibly deduce any results, the selectivity bias of teenage mothers needs to be accounted for. To do so, I employ two complementary strategies such that the possible selection bias is diminished.

The first strategy is to use an entity fixed effects regression model, with mothers as the entity. These fixed effects control for the unobservable (cultural beliefs, genetic features etc.) time invariant features. For instance, it may be that women who are more devoted mothers choose to have children sooner or women who are more ambitious choose to delay (Miller, 2008). They

(15)

will understate the negative impact of teenage motherhood in the first case and overstate it in the second.

The regression specification for this model would be drawn from equation 1 and Ui would

capture mother fixed effects.

Y

ij

= α

1

+ ß

1

X

ij

+ ρ

1

W

ij

+U

i

+ ε

1ij (3)

The second strategy is to analyze a sample where motherhood is to some extent an unexpected event. If the arrival of the child is random and independent from the age of the mother, then a causal estimate of the impact of mothers’ age on health outcomes of her children can be obtained. Mothers in this sample would be similar in most aspects (checked with descriptive statistics in Table 4) except for when they had their children and the estimates of the impact of her age on the health of her child can be deduced.

This identification strategy is based on the biological shock instrument variable utilized by Miller (2008) in her study. The setup of this sub-sample is possible due to a variable in the DHS questionnaire where the mothers indicate whether they wanted that specific child ‘Then’, ‘Later’ or ‘Wanted No More Children’. From this, children that were not wanted at the time that they were conceived or not wanted at all, can be grouped under “accidental/unwanted pregnancies”, making the pregnancy more or less exogenously determined. These observations are then cross-referenced with two other variables from the questionnaire. One is where the mothers indicate if they were using any form of contraception at the time of conception of that specific child. The second is where the mothers indicate the reason for using said contraception.10 Observations where mother’s reply ‘yes’ for using contraception for ‘Limiting’ or ‘Spacing’ are included in the final sub-sample.

4.2 Limitations

As is always the case with Ordinary Least Squares models, estimates are prone to omitted variable bias (OVB). Here, even with additional control variables, the ß parameters can suffer from OVB since unobservable factors such as cultural and normative differences and unreported factors such as differences in family planning policies in different areas are not captured and can bias the results. For instance, most family planning policies (PSI, Guinea - Family Planning 2020) that have been rolled out in Guinea have targeted areas such as the

10_{The different responses to this question are Limiting, Spacing, Prevention of STDs, Other. Only the first two are}

(16)

capital city of Conakry and Forest (a part of rural) Guinea. Since these areas are part of the sample being analyzed, and a large part of the primary sample resides in areas covered by such policies, it is possible that the estimates may be downward biased due to the availability of large-scale family planning services that enable young pregnant girls to have safe terminations or low-cost vaccinations and immunizations for their children.

Similarly, the assumptions on which the two complementary strategies are based, that aim to control for selection bias among young mothers may not hold true in practice. For instance, using the mother fixed effects model relies on the assumption that some mother characteristics remain unchanged over all her children. It is true however, only to a certain extent. Using mother’s education as an example, this assumption can hold true if a young mother would drop out of school once she has had her child. Then, her education level would remain unchanged. However if she decides to re-enroll once her child has reached a certain age, the fixed effects assumption is violated. Unobservable factors such as attitude towards FGM can change with age; as a mother grows older, she may come to realize the redundancy of the practice, hence once more violating the assumption.

The second strategy also has limitations such that even if children arrive unexpectedly or exogenously, adolescent mothers would be very different from adult mothers. However, such large differences are not observed in the descriptive statistics (Table 4). The small size of the sub-sample can also affect the significance of the results, as it can be no longer representative. This however is unavoidable given the design of the dataset and hence the results must be interpreted with caution.

(17)

5. Empirical Results

From the above mentioned identification strategies, eight different regressions are run, differing on the sample being analyzed, adolescent motherhood and the set of control variables utilized. The results from these regressions are reported in Table 3. In section 1, the results are divided on the basis of whether the mother was ever an adolescent when she had a child or if she was an adolescent at the time she had the specific child being analyzed.

An ordinary least square regression shows that a child is 0.6% more likely to be stunted if his/her mother had a child while she was a teenager, regardless of whether it was him/her or an older sibling. Similarly, these children are 2.5% more likely to be wasted and the girls are 1.8% more likely to undergo FGM. On controlling for mothers education level and the wealth index of a household, the estimates on stunting and FGM decrease to almost 0 but wasting remains around the same, approximately 2.2% more likely to be wasted.

When the outcomes of these children are analyzed on the age of their mothers when they themselves were born, all coefficients except stunting decrease in magnitude and even change direction. Children whose mothers were adolescents when they were born are 3.5% more likely to be stunted, but 1.6% less likely to be wasted than children whose mothers were older than 19 years. The FGM coefficient is close to zero and insignificant, implying that a mothers’ age at birth has no impact on if her daughter undergoes FGM or not. Even with extended controls, this change of direction in the wasting variable remains and the estimate increases slightly. The estimate of stunting increases. The OLS regressions on the full sample hence provide ambiguous results in no definite direction.

Section 2 of Table 3 reports the estimates on health outcomes of the children whose births were determined exogenously. This splicing of the sample helps diminish the selection bias into young motherhood that could arise while just running OLS on the entire sample.

Descriptive statistics of this sample (Table 4) show almost the same results as the statistics for the primary sample, where the children of adolescent mothers display slightly higher sample means on the health outcomes as the children of older mothers. However, the only significant difference is in the education level of the mothers. Here, the younger mothers have a higher number of completed school years than older mothers, but only because the maximum number of years completed by the two groups is 13 years and 17 years respectively. This could imply that younger mothers are more likely to drop out of school or pause their schooling when they have an unplanned child as compared to older mothers who have probably already completed

(18)

their schooling (giving form to the carry-over effect of a teenage pregnancy). The average age at birth in this sample is 26 years, and 25% of this sample falls below the age of 19 years at time of birth. The average age at first birth is 17 years and 65% of the sample had had their first child before turning 19.

The same specification (equation 1) is used for this sub-sample where the age of the mothers at the time of the exogenously determined pregnancy is examined. Using equation 2 (adolescent mother ever) would bring back the selection bias that the sub-sample was created to avoid. The results show that all coefficients are insignificant. All the coefficients except for the one of Stunting increase slightly, but still remain negligible. Wasting retains a negative coefficient in both samples however; the impact is around -3% likelihood of being wasted. Like in the full sample, OLS shows ambiguous results in the sub-sample as well.

Columns 7-9 of section 1, report the estimates of using the Mother-Fixed Effects model. In the full sample, the impact of having a child as an adolescent does not significantly enter into the child’s health outcomes even while controlling for the unobservable variables that manifest themselves through mothers such as her own health conditions, genes etc. What is different about the results of this specification from the ones of the previous two is that wasting is no longer inversely related to mother’s age. Children that are born to adolescent mothers approximately 7.1% more likely to be wasted than the children born to mothers over the age of 19 years.

When employing mother fixed effects in the sub-sample, the results change drastically. The likelihood of being stunted is 29% higher for children that are born to adolescent mothers. Similarly, children are 21% more likely to be wasted and 8.9% less likely to undergo FGM if their mothers were adolescents at the time of their births. These results are not significant possibly due to the small size of the sub-sample.

(19)

All Results

Stunting Wasting FGM Stunting Wasting FGM Stunting Wasting FGM

(1) (2) (3) (4) (5) (6) (7) (8) (9) OLS OLS FE Full Sample Section 1 Adolescent Mother at Time of Birth 0.035 (1.44) -0.016 (1.11) -0.003 (0.23) 0.0402 (1.63) -0.0181 (1.23) -0.0017 (0.13) 0.0183 (0.26) 0.0709 (1.52) 0.0176 (0.74) Adolescent Mother Ever 0.006 (0.31) 0.025 (2.08)* 0.018 (1.55) 0.0009 (0.05) 0.022 (1.85) 0.007 (0.66) N 3,058 3,058 8,166 3,058 3,058 8,166 3,058 3,058 8,166 Sub Sample Section 2 Adolescent Mother at Time of Birth -0.0082 (0.19) -0.0307 (0.99) 0.0053 (0.13) 0.0103 (0.22) -0.0346 (1.09) 0.0082 (0.22) 0.2912 (1.21) 0.2133 (1.13) -0.0891 (1.37) N 519 519 475 519 519 475 519 519 475

Extended Controls Yes Yes Yes

Mother Fixed Effects Yes Yes Yes

Table 3: Results

(20)

Variable

Teenage Mothers (≤19 years) Adult Mothers (>19 years)

Variable Description

Observations Mean SD Observations Mean SD

Child Characteristics/ Outcome Variables

Stunting 128 -0.836 1.425 397 -0.616 1.774 SD of Height-for-Age Ratios

Wasting 128 -0.558 1.152 397 -0.492 1.315 SD of Weight-for-Height Ratios

Female Genital Mutilation 116 0.120 0.327 359 0.116 0.321 1=Yes 0=No Mother’s Characteristics Education in Years 268 4.145 4.015 776 1.921 3.770 Min=0 years Max=13 years/17years FGM 268 0.970 0.170 776 0.985 0.118 1=Yes 0=No Family Characteristics Place of Residence 268 1.537 0.499 776 1.630 0.483 1=Urban 2=Rural

Wealth Index 268 3.272 1.375 776 3.141 1.379 1-5 in Increasing Order of Wealth

(21)

6. Discussion

The OLS results of this paper show conflicting evidence on the impact of mother’s age at birth on her children’s likelihood of being stunted or wasted. However, after employing a mother fixed effects model, the results take on a definite direction in line with the hypothesis, showing that stunting and wasting are very likely to occur in the children whose mothers were adolescents at the time of their births. In the sample where these pregnancies are not determined and planned by the mothers and time invariant characteristics of the mother are controlled for, the impact of her age at birth is magnified. This implies that the earlier OLS and FE results were confounded by unobservable mother qualities such as her desire to have children and take care of them and perhaps the most relevant confounder being the normative practice of child brides and consequent family formation. In a society where it is acceptable and a common practice for young girls to be married and start their families, they are less likely to go through the psychological stress and social stigma as would happen in a society where this was not regular. This would, in a way, make this mother more equipped to raise her children and they would display lower risk of any health disorders compared to the children of older mothers. In developed nations young mothers face a structural disadvantage, financially and socially, in Guinea this disadvantage does not exist. The social and cultural set-up in Guinea works as a mediating factor between the negative consequences of adolescent motherhood and the health of her children. When such norms are controlled for, the true negative impact of young motherhood is observed (Table 4, Section 2, columns 7&8).

For a country like Guinea where the rate of adolescent pregnancy is very high, this could have important implications for maternal and child health policies. Family planning schemes and a higher contraception provision rate could not only improve maternal health but also potentially improve child health. Organizations aiming at eradicating child-marriage could use such implications and provide policy advises in conjunction with reproductive and early child development practices.

Female Genital Mutilation of a girl child in Guinea is also not influenced by how old her mother was at the time of her first birth or at the time the girl was born. This result goes to show that this practice is influenced by external factors such as religion, mother’s education level and region of residence. A great level of intergenerational mobility is observed among mothers who underwent FGM and their daughters (Appendix 3).

(22)

7. Conclusion

The objective of this paper was to examine whether a mother’s age at first birth or a mother’s age at the time of birth of her child has an impact on whether her children are Stunted, Wasted or undergo Female Genital Mutilation. It was hypothesized that mother’s age would have an impact on her children’s outcomes because younger mothers are assumed to be psychologically, physiologically and financially less equipped to raise their children compared to their older counterparts. The results show an ambiguous impact of mother’s age on the practice of FGM and stunting and wasting of her children when using OLS. However, the results take a definite direction when a sub-sample with exogenously occurring pregnancies is analyzed using a mother fixed effects model. Interpreting the results with caution, as they are not significant, it is observed that children of young mothers have a much higher chance of being stunted or wasted compared to the children of their older counterparts. The practice of FGM displays a very high intergenerational trend between a mother who underwent the procedure and her daughters.

Even though the practice of child marriage is banned, child brides make up the main population where adolescent pregnancies originate. Adolescent pregnancy is typically defined using a very Western construct, ignoring the prevailing cultures in a specific country. The consequence of this can be severe, just like in the case of Guinea where young pregnant girls are considered a status symbol for their older husbands and economically constrained families. In such contexts where young girls conceiving and birthing children is not regarded as a violation but instead a tradition to be upheld, the impact of having an adolescent mother appears greatly diminished, disguising its true extent. This channel of mediation is an interesting topic for future research and could be undertaken to analyze better the impact of young motherhood on her children.

(23)

References

A. B. Soura, M. Winner (2014). Trends in Family Planning and Age at First Marriage among Women in Guinea: Further Analysis of the 2012 Demographic and Health Survey. DHS Further Analysis Reports, No. 94.

A. L. Cherry, M. E. Dillon (2014). Biological Determinants and Influences Affecting Adolescent Pregnancy. International Handbook of Adolescent Pregnancy, 39-53.

A. M. Dorélien (2015). Effects of Birth Month on Child Health and Survival in SubSaharan Africa, Biodemography Social Biology

A. R. Miller (2008). Motherhood Delay and the Human Capital of the Next Generation. American Economic Review, 99(2):154-58.

A. Raj (2010). When the Mother is a Child: the Impact of Child Marriage on the Health and Human Rights of Girls. Arch Dis Child.

B. R. Mednick & R. L. Baker (1980). Consequences of family structure and maternal state for the child and mother's development. Progress Report, National Institute of Child Health and Human Development.

C. H.D. Fall, H.S. Sachdev, C. Osmond, M.C. Restrepo-Mendez, C. Victora, R. Martorell, A. D. Stein, S. Sinha, N. Tandon, L. Adair, I. Bas, S. Norris, L.M. Richter, and the COHORTS investigators, (2015). Association Between Maternal Age At Childbirth And Child And Adult Outcomes In The Off Spring: A Prospective Study In Five Low-Income And Middle-Income Countries. Lancet Glob Health; 3: e366–7.

E. A. Frongillo, Jr., M. de Onis, K. M. P. Hanson (1997). Socioeconomic and Demographic Factors Are Associated with Worldwide Patterns of Stunting and Wasting of Children. American Society for Nutritional Sciences.

E. L. Lipman, K. Georgiades, M. Boyle (2011). Young Adult Outcomes of Children Born to Teen Mothers: Effects of Being Born During Their Teen or Later Years. Journal of the American Academy of Child and Adolescent Psychiatry, Vol 50, No. 3.

F. Frustenberg, J. Brooks-Gunn, S. Philip Morgan (1987). Adolescent Mothers and Their Children in Later Life. Family Planning Perspectives, Vol 19, 142-151.

G. J. Duncan, K. Lee, M. Rueda, A. Kalil, K. Ziol-Guest (2015). Maternal Age and Child Achievement. (Under Review)

G. Pogarsky, T. Thornberry, A. Lizotte (2006). Developmental Outcomes for Children of Young Mothers. Journal of Marriage and Family, Vol 68, 332-344.

(24)

H. Wamani, A. Åstrøm, Nordrehaug and Peterson, Stefan and Tumwine, J. K. Tumwine, and T. Tylleskär, (2007). Boys Are More Stunted Than Girls In Sub-Saharan Africa: A Meta-Analysis Of 16 Demographic And Health Surveys. BMC Pediatrics, Vol 7, 1-10.

J. A. Levine, H. Pollack, M. E. Comfort (2001). Academic and Behavioural Outcomes Among the Children of Young Mothers. Journal of Marriage and Family, Vol 63, 355-369.

J. Brooks-Gunn, G. J. Duncan (1997). The Effects of Poverty on Children. The Future of Children, Vol. 7, No. 2, Children and Poverty, pp. 55-71.

J. E. Finlay, E. Ozaltin & D. Canning, (2011). The Association Of Maternal Age With Infant Mortality, Child Anthropometric Failure, Diarrhoea And Anaemia For First Births: Evidence From 55 Low- And Middle-Income Countries. BMJ Open 2011;1: e000226. doi:10.1136/ bmjopen-2011-000226 K. Abdullah, M. A. Malek, A. S. G. Faruque, M. A. Salam & T. Ahmed (2007). Health And Nutritional

Status Of Children Of Adolescent Mothers: Experience From A Diarrhoeal Disease Hospital In Bangladesh. Acta Pædiatrica Vol 96, pp. 396–400.

L. J. Marteleto, M. Dondero (2013). Maternal Age at First Birth and Adolescent Education in Brazil. Demographic Research, Vol 28, 793-820.

M. I. Varela-Silva, H. Azcorr, F. Dickinson, B. Bogin, & A. R. Frisancho. (2009). Influence Of Maternal Stature, Pregnancy Age, And Infant Birth Weight On Growth During Childhood In Yucatan, Mexico: A Test Of The Intergenerational Effects Hypothesis. Am. J. Hum. Biol., 21: 657–

663

M. R. Rosenzweig, K. I. Wolpin (1995). Sisters, Siblings, and Mothers: The Effect of Teen-age Childbearing on Birth Outcomes in a Dynamic Family Context. Econometrica, Vol 63, 303-326. M. Witwer (1993). Health of Infants Born to Teenage Mothers Affected More by Family Background

than by Mother's Age. Family Planning Perspectives, Vol. 25, No. 4, pp. 191-192.

S. A. Richard, R. E. Black,W. Checkley (2012). Revisiting the Relationship of Weight and Height in Early Childhood. American Society for Nutrition. Adv. Nutr. 3: 250–254.

S. Jaffee, A. Caspi, T. E. Moffit, J. Bleksy, P. Silva (2001). Why are Children Born to Teen Mothers at Risk for Adverse Outcomes in Young Adulthood? Results from a 20-year Longitudinal Study. Development and Psychopathy, 13, 377-397.

V. J. Hotz, S. W. McElroy, S. G. Sanders (1999). Teenage Childbearing And Its Life Cycle Consequences: Exploiting A Natural Experiment. Nber Working Paper Series, Working Paper 7397. W. Baldwin, V. S. Clan (1980). The Children of Teenage Parents. Family Planning Perspectives, Vol

(25)

Appendix

1. To check for systematic differences in stunting, wasting, FGM in the sample, the

observations are split on the basis of their occurrence in children by their region of residence (urban or rural), their gender and the wealth quintile that their household falls under.

In Table 5, a clear split is visible for the observations under stunting, wasting and FGM by region. 72.4% of the observations are from rural areas and the rest from urban areas. This could be a potential limitation of the sample as this could exacerbate the impact of a young mother as children from urban areas are more likely to have access to better health, sanitation and

nutritional facilities than those from rural areas. It can be observed by the data that only 16% of children from urban areas are stunted whereas, in rural areas this percentage is double. The same is not true for Wasting in the children from the two areas, with the percentages being 8% and 10% for urban and rural respectively. Similarly for FGM, there are almost an equal number of girl who underwent the procedure in both areas. This makes the control variable for Region of Residence crucial in the regression equations, especially when analyzing the Stunting outcome variable.

The split in the observations by gender is almost an equal one, with 53% of the sample being male and 47% being female. Within this split as well, the percentage of children that are stunted or wasted is roughly equal for both males (28%) and females (24%). However, since males are said to be at a higher risk for stunting than females, the control variable for gender is included to help reduce any bias that may arise due to this.

The three outcome variables have their observations roughly equally split among the first four wealth quintiles. For stunting, as could be expected, the richest households display the lowest percentage of stunted and wasted children (12%) compared to the poorest (30%) and poorer (33%) households. For wasting, the prevalence is low but the highest numbers of children come from the poorest households as well. For FGM, the poorest households have a prevalence rate of 56% and for the rest the rate is around 48%. Hence, controlling for the wealth index of a household is also important so as to not let the poor households upward bias any estimates.

(26)

Table 5: Checking for Systematic Differences

Region of Residence Gender of Child Wealth Quintiles of the Household

Variable Urban YES | NO Rural YES | NO Male YES | NO Female YES | NO Poorest (1) YES | NO Poorer (2) YES | NO Middle (3) YES | NO Rich (4) YES | NO Richest (5) YES | NO Total YES | NO Stunting 842 2216 1626 1432 743 656 597 690 372 3058 136 706 673 1543 457 1169 352 1080 216 527 223 433 176 421 153 537 41 331 809 2249 Wasting 842 2216 1626 1432 743 656 597 690 372 3058 68 706 242 1974 176 1450 134 1298 95 648 65 591 52 545 71 619 27 345 310 2748 FGM 2268 5898 1927 1702 1709 1790 1038 8166 1128 1140 2973 2925 1082 845 790 912 854 855 934 856 441 597 4101 4065

(27)

2. If there exists a carry-over effect of adolescent pregnancy, we could expect to see it in

the outcomes of children separated by their birth order. More recent children (birth order 4 & 5) would show more or less similar likelihood of being stunted, wasted or FGM as the older children who were born when their mother was a teenager. An OLS regression of the health outcomes on if a mother ever had a child as an adolescent is run with birth order of the child included as dummy variables; we observe no trend for stunting. However, regardless of birth order, children of mothers who were ever adolescents at the time of birth display approximately the same likelihoods of being wasted or undergoing FGM.

Table 6: Carry Over Effects of Adolescent Childbearing

Variable Stunting Wasting FGM

Adolescent Mother Ever 0.0006 (0.03) 0.0229 (1.86) 0.0062 (0.54) Gender of Child -0.0230 (1.31) -0.0084 (0.70) Region of Residence 0.1023 (4.02)** 0.0087 (0.50) -0.0984 (6.09)**

Current Age of Child 0.0731

(9.74)** -0.0204 (4.48)** 0.0640 (38.81)** Mothers Education -0.0062 (2.37)* 0.0003 (0.17) -0.0057 (3.18)** Wealth Index -0.0162 (1.88) -0.0134 (2.21)* -0.0463 (8.75)** Birth Order 1 0.3720 (7.63)** Birth Order 2 0.1035 (0.30) 0.0670 (1.41) 0.4070 (8.58)** Birth Order 3 -0.1753 (1.83) 0.1083 (2.10)* 0.3989 (8.67)** Birth Order 4 0.0189 (0.25) 0.1368 (2.69)** 0.3451 (7.82)** Birth Order 5 0.0453 (0.64) 0.1652 (3.32)** 0.3356 (7.98)** R2 0.32 0.12 0.66 N 3,058 3,058 8,166 * p<0.05; ** p<0.01

(28)

3. From running OLS regression, it is observed in Table 7 that a daughter is 22.6% more

likely to undergo FGM if her mother has undergone the same well. Other statistically significant factors include the current age of the child (as FGM is usually performed before the girl arrives at puberty) and mother’s education level (more educated mothers are less likely to have FGM performed on their daughters). Mother’s age at birth has no significant impact.

Variable Daughter Undergoes FGM

Mother Underwent FGM 0.2260 (3.82)** Region 0.012 (2.18)* Mother’s Education -0.0071 (3.98)**

Current Age of Child 0.0673

(61.91)**

Mother’s Age at Birth -0.0008

(1.09) Required by Religion -0.0078 (2.02)* Intercept -0.1793 (2.83)** R2 0.35 N 8,166

Table 7: Intergenerational Mobility of FGM