1
Early marriage and young child bearers: a study of
adolescents in Bangladesh
Thesis MSc. Development Economics
Supervised by Prof. Erik Plug, PhD
Camilla Nystrand
University of Amsterdam
August 11, 2015
Abstract
This paper uses data from Bangladesh on health conditions and utilization of health care in an instrumental variables framework. It will assess the hypothesis that females who marry at a young age bear children earlier which in turn have a negative consequence on pregnancy and birth related outcomes. The variation in the timing of the first menstruation is used as an instrument for age at marriage. The results indicate that delaying marriage by one year significantly increases the age at first child bearing by up to 0.12 years. Delaying marriage has no effect on pregnancy and birth complications, nor with utilization of health care services during the first pregnancy.
2
Contents
I
Introduction
3
II
Motivation
5
III
Data and Methods
6
A.
Data……….…...6
B.
Sample……… …...6
C.
Estimation strategy……….7
IV
Reduced form regression
14
V
Results
17
A.
Pregnancy and birth complications………...17
B.
Health care behavior during pregnancy………....22
C.
Health of the mother……….23
IV
Conclusion
24
3 I. Introduction
Every two seconds, a young female is married before the age of 18, estimating around 15 million adolescents per year. In numerous societies, marriage is an important life event for both women and men, as it signals a step towards adulthood. Early marriage is associated with a range of life changes, with one of them being the development a new family. Looking at early marriages through a human rights viewpoint becomes important if the event has a large impact on young females’ life, especially with regards to marriage age and consent. The Universal Declaration of Human Rights distinguishes an objective norm of maturity, protecting a child from forcefully being married before they are mentally, emotionally and physically ready: it allows children to be recognized in the law as being children. Hence if marriages happen forcefully before the age of 18, it may have detrimental effects on young female’s life (United Nations, 1948; Girls not Brides, 2015). Young women in developing countries often have little to no choice over the age and partner to be married as it is often determined by her parents (Jensen and Thornton, 2003). One of the highest rates of child marriage can be found in Bangladesh, especially in the rural parts of the country. Recent numbers by UNICEF (2013) estimate that over two thirds of girls marry before the age of 18, while one third are married before the age of 15.
The literature on the topic of child marriage and its consequences for child bearing and birth is still inconclusive. While some studies point towards detrimental effects resulting from early child bearing, others refer to external circumstances around marriage that may have a positive effect for both the mother and her child.
On the one hand, early marriage often causes concern because of its potential detrimental
consequences for a young woman’s’ physical and mental development, which may have an impact on her offspring. On average, girls who marry young tend to give birth earlier as they are supposed to show their worth in their new family by starting to reproduce. This directly impacts the mother and her child’s well-being during and after pregnancy (Senderowitz, 1995). Simultaneously, younger wives have less autonomy and power within the household, have limited reproductive control and suffer greater risks of maternal mortality (Jensen and Thornton, 2003).
On the other hand, there is a premium for age in the market for marriage. Ceteris paribus, younger brides may have the opportunity to marry into a wealthier household where more resources can be spent on maternal and child care, which may be able to compensate for an early child bearing age (Sekhri and Debnath, 2014). Therefore, the effect of early female marriage on pregnancy and birth related outcomes are ambiguous. At the same time, although statistics show that young brides fare worse on numerous outcomes, it is a difficult task to assess whether the outcomes are driven by the age at marriage or other related factors, such as poverty or traditional views on women’s’ role in the
society. Therefore, delaying marriage does not necessarily entail that pregnancy and birth related outcomes would improve. This paper attempts to empirically examine the pregnancy and birth related outcomes generated by early marriage of females in Bangladesh.
To estimate the influence of marriage age, the Bangladesh Health and Socio-Economic Survey from 1996 is used which contains data on household and family background characteristics, as well as female menarche, marriage and pregnancy information. Empirically, it is likely that unobserved confounding factors that have an effect on both age at marriage and pregnancy and birth related
4
outcomes will be difficult to control for in a standard ordinary least squares (OLS) regression.
Therefore, such a regression cannot be causally interpreted. For instance, men who have a preference for marrying older females may also have a preference for fewer children, and hence encourage his wife to postpone child birth. In this case, it may be the desire of the husband leading to postponed child bearing, and not the marriage age of the mother. Hence, the age of first child bearing for females cannot be fully attributed to the age of marriage. Other possible confounders may arise from women who for example marry early and pursue less education and therefore have less economics means to obtain medical care during pregnancy. However, these women may be more attractive on the marriage market and therefore marry into wealthier families who can provide for health care. Another possible scenario is that women who marry later have better economic means, but may be more attractive to a family whose traditions deny her the right to seek medical care. Such mentioned confounders may have an ambiguous effect on the outcomes of interest; hence they cannot be fully attributed to age of
marriage. To address this concern, a two-step instrumental variables (IV) methodology is used. Exogenous variation in a females’ age at her first menstrual cycle is used as an instrument for her age at first marriage. In many developing countries, there is an incentive for parents to marry off their daughters at an early stage, but girls are often withheld from marriage until they have reached pubic age (Field and Ambrus, 2008). Female menarche therefore creates a binding constraint to entering the marriage market. Because of the natural variation in the timing of menstruation, often between the ages of 11-16 in physically and mentally healthy adolescents, it creates a quasi-random variation in the age at which a young girl faces the risk of marrying (Palmert and Boepple, 2001). The estimates from this paper indicate that each additional year first menstruation is delayed significantly postpones marriage by 0.69 years. The validity of the IV is dependent on both the relevance of the instrument and whether it is uncorrelated with unobserved characteristics in the structural equation. The strategy followed in this paper has been used in previous literature and has been found to be a strong and relevant
instrument for age at first marriage. In addition, sensitivity tests are used to ensure the validity of the instrument.
This paper makes an important contribution to the existing literature on the marriage market in developing countries. While there is limited literature that have examined the direct effect of early female marriage (Field and Ambrus, 2008 on educational attainment; Dagnelie, 2010 on child
mortality; Sekhri and Debnath, 2014 on children’s human capital formation), no empirical research has been conducted on females’ well-being and health utilization during pregnancy. The results of this paper indicate that marriage age has a great impact on the age of first child bearing. I find that delaying marriage by one year universally significantly increases the age at first birth by 0.12 years. However, additional results show no effect of early child marriage on health risks during pregnancy and at birth. When the sample is restricted to females with non-educated parents, similar results are found.
The paper is structured as follows. Section II outlines the motivation for the paper. Section III introduces the data and the methodological framework. Section IV describes the reduced form estimates, while section V presents the results of the instrumental variables estimation. Section IV concludes.
5 II. Motivation
This paper builds on important medical studies and country reports. Firstly, medical studies have previously examined the relationship between age of marriage and pregnancy and birth related outcomes. They have found that age at first marriage is one of the key determinants for a females’ reproductive life length and fertility level, with early marriage resulting in negative consequences for maternal health (Coale and Treadway, 1986; Rosero-Bixby, 1996; Palloni and Rafalimanana, 1999; Santhya et al. 2010). Empirical estimates also point towards the rising age at marriage for explaining declines in fertility and increased age at first child bearing (Harwood-Lejeune, 2000 in Southern and Eastern Asian countries; Timaeus and Moultrie, 2008 in South Africa; Kirdar et al. 2011 in Turkey). Adolescent pregnancy is also often associated with pregnancy and birth complications, comprising of preterm delivery and maternal anemia, which increases the risk of low birth weight, neonatal mortality and poor infant outcomes, especially significant in Bangladesh (Scholl et al. 1994; Reynolds et al. 2006). Studies also suggest that females who marry young are more likely to be younger child bearers and are more likely to terminate a pregnancy, which can result in pre- and postnatal
complications (Raj et al. 2009 in India; Godha et al. 2011 in Southern Asian countries; Nasrullah et al. 2014 in Pakistan). In sum, two channels reinforce pregnancy and birth related consequences;early marriage and adolescent pregnancy.
Secondly, the potential risks of early child bearing diminish with comprehensive prenatal care (Scholl et al. 1994). Basic data from UNICEF (2007) indicate that antenatal care coverage in
Bangladesh, the likelihood that a pregnant woman receives care during the pregnancy at least once, is less than 50 per cent. 90 per cent of women deliver births at home, and two thirds of women seek no help or receive assistance from an unqualified provider for pregnancy complications. A pregnant woman’s utilization of health care is directly affected by her education, family income and culture, but inconclusive with regards to the effect of age of the mother (Chakraborty et al. 2003 in Bangladesh; Navaneetham and Dharmalingam, 2002 in Southern India).
Thirdly, medical studies suggest that pregnancy and birth complications are directly affected by the timing of child bearing relative to menarche. The gynecological age, which is the time pasted since the onset of the first menstruation, forecasts when the body of a female is fully reproductively mature. Reproductive immaturity is a gynecological age less than three years, as it takes fully healthy and well-nourished teens between two to three years post menarche to become physically mature to bear a child. Before such time has passed, a young female is able to bear a child, but her reproductive organs have not fully developed and hence pregnancy and child birth becomes riskier (Felice et al. 1984; Stevens-Simon et al. 2002).
Fourthly, early marriage reduces educational attainment (Jensen and Thornton, 2003; Field and Ambrus, 2008). Females who marry young often drop out of school earlier to care for her new family, and hence attain less education. Young child bearers are therefore less knowledgeable about healthy practices during and after pregnancy, and have less knowledge of when to seek medical care
(Senderowitz, 1995). Simultaneously, girls living in poor households are also the most vulnerable when it comes to adolescent pregnancy, as they are the least likely to seek medical care (World Bank Group, 2014). Lack of education and less economic means due to lower education levels therefore spur the risks associated with early child bearing due to the reduced demand for health care services.
6
Fifthly, females are more likely to marry at a younger age if they come from a poor family. Previous literature indicate that poor families on the one hand face higher opportunity costs for maintaining the well-being of their children, both related to nutrition, health and educational attainment. On the other hand, in marriage markets where a dowry system is prevalent, less has to be paid to the new husband’s family the younger the female is at marriage (Bates et al. 2004; Raj et al. 2009). Since females show their worth in the new family by producing children, a smaller dowry due to a younger age may put even more pressure on the wife to bear children soon after the marriage. In addition, a smaller dowry provides less financial means for the new family to cover for medical expenses (Jensen and Thornton, 2003).
Lastly, the younger the female is at first marriage, the less likely she is to have autonomy and freedom to make her own decisions in the household, and the less she will be able to utilize health care services (Senderowitz, 1995; Field and Ambrus, 2008). Education has a direct and an indirect effect on empowerment through knowledge and higher income. The later the marriage, the higher the education level and the higher the power to be able to decide for yourself whether to utilize health care services.
III. Data and methods
A. Data
The quantitative data used in this paper is collected from the 1996 Matlab Health and Socio-Economic Survey, MHSS, a region in rural Bangladesh. The survey consists of household- and individual-level information on 4,364 households which were clustered in 2,687 baris (residential compounds), amounting to approximately one third of the sample of all baris in the surveillance area randomly chosen. A maximum of two households in each bari were randomly selected for interview. Baris are a cluster of extended families that function as the basic family unit for economic, social and security activity. The head of the household, the spouse, parents of the head, every individual over the age of 55 and a random sample of members between the ages of 15 and 49 were selected for interview.
The data contains comprehensive information needed for the analysis, such as household characteristics and family background, education and health conditions, as well as complete ever-married woman information with marriage, menarche, and pregnancy information. The main outcome variables are age at first child birth, pregnancy and birth complications, demand for health care services and adult health. The endogenous variable of interest is age at first marriage, which is instrumented by age at menarche. The control variables used are age and whether the female is Hindu. The analysis is used universally on the whole sample, as well as on a smaller sample of females whose parents have no education.
B. Sample
The sample used in this paper is limited to 2,097 ever-married women between the ages 25 and 44 for whom it is possible to find full information on marriage and menarche history. Every woman in the sample has given birth at least once. The lower limit cut-off was chosen to minimize the possibility of
7
observational outliers of women who marry late for the first time, while the upper limit cut-off was chosen to reduce the risk of selective mortality. If earlier marriage or earlier menarche is correlated with higher rates of mortality amongst female adults, the estimates could be biased. In addition, the sample could be biased by the likeliness that some females have suffered pregnancy related
complications which has resulted in maternal mortality. However, as it is not possible to control for or gain information on these females, it has to be neglected in the further discussion. By excluding these females, the results will be biased downwards. This is due to the fact that the outcome variables include complications during pregnancy and child birth, which may have resulted in mortality due to child bearing for the excluded women in the analysis. In addition, the mortality could have been due to serious health conditions, which may also have had a negative consequence on child bearing. Being able to account for these women may therefore have had an impact on the results, hence the downward bias by excluding them.
Summary statistics for the sample selected for analysis is presented in Table I. The full sample is presented in the first (1) column, while the following columns are divided according to marriage age and intervals for age of menarche. The patterns of low marriage age in Bangladesh are found in the data, with the average marriage age being 16.5 years. Almost half of the sample comes from a non-educated family, and while 13 percent of females give birth for the first time within three years after menarche, the statistics indicate no significant difference in age of first birth between women married before or after the age of 15. Females have low utilization rates of antenatal care, and the likeliness of giving birth at home is almost unitary. However, pregnancy and birth complications are rare, and the likeliness of miscarriage in the full sample is five percent.
As reported in columns two (2) and three (3), differences in marriage age areconsistent with the trends in Bangladesh, where 45 percent of the sample were married at age 15 or younger. Females who marry young are more likely to come from a family where her parents have no education, and are less likely to utilize antenatal care. However, the average age of first birth is similar across the marriage intervals, as is the low rates of pregnancy and birth complications.
C. Estimation strategy
To estimate the causaleffect of early marriage on birth and pregnancy related outcomes, age of menarche will be used as an IV for the age of marriage. Numerous medical studies discuss the distributions of age of menarche across populations and argue that in normal healthy adolescents, the transition to a reproductive age occurs in any population in a wide distribution of ages (Scott and Johnston, 1985; Palmert and Boepple, 2001). If first marriage is bounded by tradition by the age of menarche, as in Bangladesh, it creates an exogenous disparity in marriage timing of young girls menarche (Begum, 2003). However, only if age of menarche is exogenously determined is the
instrument valid, which will be discussed further on. Although prearrangements of marriage occur, it is rare and in the sample it amounts to less than ten observations.
8 TABLE I
SUMMARY STATISTICS
FIRST MARRIED MENARCHE
15 OR OLDER ALL YOUNGER THAN 15 11-13 14 15-17 (1) (2) (3) (4) (5) (6) Age 33.388 34.492 32.497 33.903 33.332 33.026 Height (cm) 150.133 150.152 150.117 150.155 150.174 150.074 Family Background Likeliness of parents .473 .435 .504 .508 .461 .456 having an education Marriage outcomes: Age at marriage 16.513 13.544 18.908 15.306 16.649 17.354 Value of dowry 3,421 2,545 4,127 3,286 3,330 3,621
Pregnancy health care outcomes:
Age at first birth 17.442 17.355 17.512 17.338 17.428 17.541 Likelihood prenatal care .287 .189 .368 .223 .323 .305
9
Average visits to antenatal .728 .476 .933 .605 .837 .718
clinic Likelihood anti tetanus shot .921 .916 .925 .905 .925 .931
Likelihood giving birth at home .972 .984 .962 .965 .971 .979
Pregnancy related complications At least one complication .009 .0 .017 .007 .005 .016
Birth related complications Likelihood of miscarriage .044 .040 .048 .049 .047 .038
At least one complication .004 .001 .006 .003 .003 .005
Adult health outcomes Diabetes .073 .075 .070 .081 .063 .076
Arthritis .414 .460 .377 .385 .422 .430
Urinary infection .178 .180 .177 .199 .152 .188
Respiratory disease .031 .040 .024 .044 .028 .024
Gastritis .487 .494 .482 .475 .507 .477
Other health condition .094 .103 .087 .102 .091 .092
Overall health (self-rated 1-3) 1.531 1.549 1.517 1.571 1.512 1.519 Observations: 2097 941 1156 606 746 745
NOTE. –Data are taken from the MHSS. Likeliness of parents having an education is an indicator of the likeliness that any or both of the female’s parents have attended school. Likelihood of prenatal care is an indicators of prenatal care during the first pregnancy; average visits to antenatal care is the average number of visits to an antenatal clinic during the first pregnancy; likelihood anti tetanus shot is an indicator of a prenatal anti-tetanus shot during the first pregnancy; at least one pregnancy complication indicates the likeliness of experiencing at least one of the following during the first birth: excessive bleeding, enema and anaemia; at least one birth complication indicates at least one of the following during the first birth: excessive bleeding, convulsions, reversed position (legs first); likelihood of miscarriage and birth at home are indicators of the likeliness of the event happening during the first pregnancy.
10
Dowry payments are also a part of the marriage market in Bangladesh. Although illegal, it is a common practice amongst Hindu families (Bates et al. 2004). In the sample, 47 percent of women report some form of dowry being paid to her husband’s family, and while dowry payments often increase with age, it indicates an incentive for parents to marry off their daughters at a younger age. At the same time, preserving a daughter’s virginity post menarche could be of great difficulty for young girls’ parents, as girls traditionally dress differently when menarche is reached. This creates a further incentive for parents to marry off daughters early, as protecting them from intercourse becomes more difficult when traditional costumes signals both menarche and maturity (Begum, 2003).
The idea is that girls in many parts of the world are withheld from marriage until the onset of puberty, which creates a binding constraint on exposure to early marriage opportunities. The age of menarche therefore acts as a physical constraint to marriage, assuming it is independent of birth and pregnancy outcomes, and can therefore be used as an exogenous event when assessing the impact of child marriage. Although menarche itself impacts the timing of child bearing, due to the gynaecological age, pregnancy occur post marriage, given menarche history. In the sample, a negligible amount of females become pregnant before marriage, hence marriage acts as a constraint to pregnancy.
In the sample statistics, these patterns are mostly reflected. Columns four (4) to six (6) in Table I show the average age of first marriage corresponding to different menarche intervals. 59 percent of the females marry within the first two years after the menarche.
The IV-approach adopted in this paper makes use of a two-stage least squares model of the following form:
𝛭𝛭𝑖𝑖 = 𝛾𝛾0+ 𝛾𝛾1𝛧𝛧𝑖𝑖+ 𝛾𝛾2𝑋𝑋𝑖𝑖+ 𝑣𝑣𝑖𝑖 (1)
and
𝑌𝑌𝑖𝑖 = 𝛿𝛿0+ 𝛿𝛿1𝛭𝛭𝑖𝑖+ 𝛿𝛿2𝑋𝑋𝑖𝑖 + 𝜐𝜐𝑖𝑖 (2)
where 𝛭𝛭𝑖𝑖 is the individual i’s age at first marriage, and 𝛧𝛧𝑖𝑖 is i’s age at menarche, which is the instrument used to identify the first-stage equation (1). 𝑌𝑌𝑖𝑖 is the outcome of interest and 𝑋𝑋𝑖𝑖is a full set of controls including: religion and age.
The analysis also makes use of a reduced form equation which estimates the relationship between the IV and the outcomes of interest. This provides an estimate for the direct effect of the IV, in addition to an idea regarding the direction of the effect for the structural equation (2). The reduced form equation takes the following form:
11
Data is clustered at the bari level where robust standard errors are used to correct for this, and single year of birth fixed effects and five year age interval dummies are used in the regressions.
Because age of marriage is likely to be an endogenous regressor, age of menarche as an instrument requires a strong correlation with age of first marriage. In the first-stage regression presented in Table II, with and without the full set of controls, there is a significantly strong correlation between age of menarche and age at first
marriage, which is consistent with earlier findings (Field and Ambrus, 2008 and Sekhri and Debnath, 2014). In the first (1) column, without controls, each year that puberty is delayed increases the age at first marriage by 0.69 years. In column two (2), which is the baseline estimation with a full set of controls, the estimation is very similar to column one and statistically significant, with an F-statistic of 161. Columns three (3) and four (4) allow for nonlinearity by including age of menarche as dummy variables to estimate its effect on marriage age. The unit of analysis in the regressions is the sample of ever-married women between the ages 25-44 collected by MHSS. Throughout the paper, the specifications in columns two (2) and four (4) will be used in the regressions.
Whether age at menarche is a valid instrument is dependent on whether it is correlated with age at marriage (relevance) and uncorrelated with the error term 𝜐𝜐𝑖𝑖 in the structural equation (2). The first requirement was tested and presented in Table II2.
For the second requirement, the exclusion restriction requires that the relationship between age of menarche and pregnancy and birth related outcomes is fully arbitrated through age at first marriage, so that there are no other channels of correlation than through marriage age that maturation has an influence on the outcome variables. However, the second requirement cannot be tested and has to be assumed according to knowledge and previous literature. According to numerous biological research studies, there are several external factors which may influence the timing of menarche and has been tested in laboratory experiments. These comprise of strenuous physical activity or stress, geographical components, exposure to endocrine-disrupting chemicals and toxics, sex composition of peer group and sudden changes in diet during childhood or in utero (malnutrition).
Through medical examinations, it has been indicated that hard physical labour during childhood affects menarche (Bronson, 1987). The onset of puberty may be affected by the economic situation of her family, as less well-off families may have to put their children to strenuous work. More physically demanding work at an early age is thought to delay age of menarche. From Table III, it is clear that education, which is used as an indicator of the economic status of the family, is insignificantly
1A strong instrument requires an F-statistic of at least 10, which this result satisfies.
2
The first requirement for the IV is valid for the sample in Bangladesh, but external validity of the following IV regressions has to be carefully interpreted. With data from Indonesia on a similar sample of females, the first stage regression showed insignificant results not different from zero. Although outside of the discussion in this paper on the reasoning behind this, the results may not be extrapolated and evaluated for other countries or settings.
12 TABLE II
FIRST-STAGE REGRESSION: AGEOF FIRST MARRIAGE
(1) (2) (3) (4) (5) (6) Age of menarche .693 .692 .817 .626 (.033)** (.076)** (.112)** (.098)** Menarche 11 -4.303 -4.288 (.122)** (.852)** Menarche 12 -2.877 -2.878 (.296)** (.501)** Menarche 13 -2.215 -2.213 (.093)** (.318)** Menarche 14 -1.047 -1.048 (.410)* (.296)** Menarche 15 -.420 -.424 (.206) (.309) Age -.053 -.053 -.133 -.096 (.040) (.040) (.030)** (.039)* Hindu .172 .143 -.008 -.357 (.336) (.356) (.419) (.569) Universe Yes Yes Yes Yes Parents Parents
without with education education F-Statistic 83.6 16.0 82.7 15.8 53.3 40.2 Observations 2097 2097 2097 2097 1063 1034
NOTE. –Data are taken from the 1996 MHSS. OLS regression; dependent variable is age at first marriage. Standard errors are in parentheses and account for individual roster weights and sample clustering (baris). Regressions also include 5-year age dummies and are robust to single-year age fixed effects.
* Significant at the 5 percent level. ** Significant at the 1 percent level.
correlated with age of menarche. The possible effect of hard labour can therefore be neglected in the sample of analysis.
Geographical components, such as weather conditions and altitude, are difficult to test on the sample of analysis since there is no data on the geographical location where the woman grew up. Literature suggests that the higher the altitude and the colder the weather, the more puberty is delayed (Nazian and Piacsek, 1976). Because the women are sampled from a uniformly rural district in Bangladesh, the variation in climate and altitude is small (Field and Ambrus, 2008). However, if women have migrated to the area post marriage, there is little to no way to control for its influence without migration data.
Different endocrine-disrupting chemicals have an ambiguous effect on the onset of puberty, with different forms of toxics having different effects (Der et al. 1974, McLachlan and Dixon, 1977). These are often due to environmental pollution which is strongly
13
the regression in Table III, no significant effects are found.
Sex composition of the peer group has an ambiguous direction (Vandenbergh 1969, Mucignat-Caretta, Caretta, and Cavaggioni, 1995) and is difficult to assess in the analyzed sample. However, it is likely that the composition is distributed evenly across individuals and therefore has no significant influence.
Early childhood malnutrition that could have an effect on timing of menarche would show itself through its effect on growth stunting. Commonly linked with nutrition and stunting is adult height, and the data can be assessed and its effect on age of menarche can be found by regressing height on timing of menstruation. When menstruation is broken down in age intervals, the coefficient estimate is close to zero and insignificant3. In Table III, the regression reveals no significant association between age of menarche and height. Similarly, in Table I, row two and columns four (4) to six (6), the data reveal no significant association with the onset of menstruation and height.
Furthermore, a recent medical study conducted in Bangladesh found no significant
differences in health status between girls who were menstruating and
non-menstruating below the age of 16 (Chowdhury et al. 2000). Table I reinforces these findings as no significant association is found between age of menarche and adult health indicators.
The exogeniety of age of menarche as an instrument is tested in Table III where the variable has been regressed on variables that may have an impact on age of menarche. The results are found in Table III.
TABLE III
SENSITIVITY TEST OF THE INSTRUMENT
(1) (2) (3) (4) Education of -.041 -.060 parents (.748) (.078) Hindu .110 .116 (.113) (.120) Height .002 .003 (.006) (.007)
Universe Yes Yes Yes Yes Observations 2097 2097 2097 2097
NOTE. –Data are taken from the 1996 MHSS. OLS regression; dependent variable is age of menarche. Standard errors are in parentheses and account for individual roster weights and sample clustering (baris). Regressions also include 5-year age dummies and are robust to single-year age fixed effects.
14
* Significant at the 5 percent level. ** Significant at the 1 percent level.
The results from Table III indicate that the independent variables are both statistically insignificant in isolation as well as jointly. This sensitivity test of the instrument further validates its usage and its exogeniety, and the exclusion restriction of the validity holds for the variables that are arguably exogenous.
In addition, biological studies suggest that genetic factors are the strongest forecast of adolescent development because of its natural randomness. It therefore plays an important role in estimating the age at which a girl reaches puberty. From a study on dizygotic and monozygotic twins, the correlation in the age of menarche was much more profound for monozygotic twins. The more similar the genetics, the higher the correlation between ages of menarche. The study on twins therefore suggest a high degree of influence from genetics and hence the role for environmental factors on maturation becomes minor (Campbell and Udry, 1995; Kaprio et al. 1995).
The evidence therefore suggest that age at menarche is uncorrelated with health background and well-being, and differences across individuals when it comes to
health status and nutritional background are unlikely to confound the further analysis. The second requirement for a valid instrument can therefore be assumed, according to
previous literature on the topic. In addition, any poor childhood nutritional status that may affect age of menarche would have a downward bias on the IV estimates, as it is likely to be correlated with poorer birth and pregnancy related outcomes. To further control for potential bias, the sample is reduced to females reaching puberty between the ages of 11-16 years. This is to control for potential outliers which are linked to chronic medical conditions and above normal physical activity and stress (Palmert and Boepple, 2001).
IV. Reduced form regression
Table IV reports estimates for age of menarche as a continuous variable in Panel A, as well as each separate age of menarche in Panel B and its effect (𝛼𝛼1) for the sample. In both panels, it is prevalent that each year of additional delay of menarche significantly increases the age at first birth. The continuous measure shows that when timing of menarche is increased by one year, age at first birth increases by almost 0.09 years. The association can be analysed from previous empirical research, where age of
menarche can been seen as a binding constraint to entry into the marriage market. While Table IV indicates some significant effects for other pregnancy and birth
related outcomes, the estimates are small and close to zero. The results already provide an idea of the impact of age at marriage on pregnancy and birth related outcomes, and the reduced form estimates reveal little hope for finding any results in the IV estimates. Intuitive reasoning as to why no correlation is found will be further discussed in the next section.
16 TABLE IV
REDUCED FORM REGRESSION: EFFECT OF AGEOF MENARCHE ON PREGNANCY AND BIRTH RELATED OUTCOMES
Age at Pregnancy Antenatal Number pre- Anti-tetanus Birth Giving birth Health First Birth Miscarriage Complications care natal visits shot Complications at home condition
(1) (2) (3) (4) (5) (6) (7) (8) (9) Panel A Age of menarche .086 -.003 .001 .003 -.036 .008 -.000 .004 .011 continuous (.007)** (.001)* (.001) (.010) (.030) (.004) (.000) (.002)* (.010) Age .004 -.001 -.001 -.036 -.081 .002 -.001 .003 .017 (.002)* (.001) (.001)* (.004)** (.010)** (.001) (.001) (.003) (.007)* Hindu .113 -.002 -.004 .110 .215 .016 -.003 -.026 -.121 (.057)* (.007) (.004) (.042)** (.110)* (.006)* (.001)** (.005) (.045)** Panel B Age of menarche 11 -.192 .084 -.001 -.199 -.011 -.306 .002 .006 .234 (.096)* (.090) (.005) (.219) (.275) (.180) (.001) (.014) (.047)** Age of menarche 12 -.369 .005 .008 -.009 .361 -.026 -.000 -.046 -.061 (.153)* (.015) (.017) (.060) (.273) (.029) (.001) (.034) (.064) Age of menarche 13 -.280 .006 -.007 -.017 .074 -.021 .004 -.002 -.042 (.104)** (.009) (.005) (.037) (.102) (.018) (.004) (.014) (.043) Age of menarche 14 -.247 .002 -.003 .034 .152 .008 .001 .005 .013 (.095)** (.009) (.006) (.035) (.092) (.014) (.002) (.013) (.039)
17 Age of menarche 15 -.107 -.005 .008 -.021 .093 -.006 .003 .002 .010 (.077) (.009) (.009) (.039) (.101) (.017) (.003) (.015) (.043) Age -.004 -.001 -.001 -.036 -.081 .002 -.001 .003 .017 (.009) (.002) (.001) (.004)** (.011)** (.002) (.001) (.002) (.007)* Hindu .131 -.002 -.005 .108 .208 .014 -.003 -.026 -.121 (.085) (.009) (.005) (.042)** (.110) (.013) (.002) (.019) (.045)** Observations 2097 2097 2097 2097 2097 2097 2097 2097 2097
NOTE. –Data are taken from the 1996 MHSS. OLS regression; dependent variable is pregnancy and birth related outcomes. The dependent variable in col. 1 is the age of first birth of the mother; the dependent variable in col. 2 is the likeliness of miscarriage during the first pregnancy; the dependent variable in col. 3 is whether the female
experienced any pregnancy complications, which are defined as excessive bleeding, convulsions, enema or anaemia; the dependent variable in col. 4 is the likeliness of antenatal care during the first pregnancy; the dependent variable in col. 5 is an indicator of the number of prenatal visits to a health clinic during the first pregnancy; the dependent variable in col. 6 is the likeliness of the mother receiving a prenatal anti-tetanus shot during the first pregnancy; col. 7 is the likeliness that the female experiences any birth related complications, defined as excessive bleeding, convulsions or revered leg position; the dependent variable in col. 8 is the likeliness that the female gave birth at home to her first born child. Column 9 is the likeliness that the female experiences any health related problems in adulthood, defined as: diabetes, arthritis, gastritis, respiratory disease, urinary infection or other illness In Panel A, age of menarche is a continuous variable defined as the onset of menarche between the ages 11-17. In Panel B, age of menarche is divided into single years. Standard errors are in parentheses and account for individual roster weights and sample clustering (baris). Regressions also include 5-year age dummies and are robust to single-year age fixed effects.
* Significant at the 5 percent level. ** Significant at the 1 percent level.
18 V. Main results
A. Pregnancy and birth complications
To measure the impact of early marriage on pregnancy and birth related outcomes, I measure the impact of delayed marriage on eight different pregnancy and birth related outcomes amongst women universally and those who were raised by parents with no education. This amounts to comparing the outcomes between females who grew up in a poor family (no education) to the whole sample. The sample is restricted to females who have given birth at least once, and the effects are measured on each female’s first birth. By such construction, it provides more precise results as confounding unknown factors affecting successive births cannot be accounted for.
Results from the IV estimates for pregnancy and birth related outcomes are reported in Table V together with the corresponding OLS estimates. Each regression in Table V controls for age of the female and whether she is Hindu. Panel A is regressions for the full sample, while Panel B is restricted to females whose parents have no education. Each variable is estimated with two IV alternatives. The first uses the continuous measure for age at menarche, while the second uses a combination of ages of menarche divided into single years4.
Columns 1-3 report the age at first birth for all women (Panel A) and women from a poor background (Panel B). While all the estimates are highly statistically
significant, the coefficients for the IV estimates are much larger than the OLS
estimates. Using either of the two IV alternatives, a one year delay in age of marriage significantly delays age at first child birth by 0.10-0.12 years. These estimates imply that universal marriage postponement could result in later ages at first child birth, ceteris paribus. When the sample is restricted to non-educated parents, the estimates are similar but lack significance except for in column 2. The reason as to why the IV estimates are almost eight times as large as the OLS estimate is likely to stem from omitted variable bias. Since age at first marriage was assumed to be an endogenous variable correlated with unobservable characteristics that could not be controlled for in an OLS regression, the IV estimates shows that these characteristics were likely to have a downward bias on the OLS results, but reduced in the IV regression.
Columns 4-6 indicate the likeliness of suffering a miscarriage from the first
pregnancy. The only significant result is from the IV estimate, where the coefficient is negative but close to zero. Similarly, outcome variables related to problems during pregnancy and at birth, such as the likeliness of suffering from pregnancy
complications (Columns 7-9) and birth complications (Columns 16-18), all show
4
Age of menarche is 11, 12, 13, 14 and 15.
6
The variables for complications were constructed as the likeliness of suffering from at least one of the following complications during: (Pregnancy): excessive bleeding, convulsions or revered leg position, and (Birth): excessive bleeding, convulsions, enema or anaemia.
19 TABLE V
OLS AND IV: THE EFFECTOF MARRIAGE AGE ON PREGNANCY AND BIRTH RELATED OUTCOMES
Age at Pregnancy
First Birth Miscarriage Complications
OLS IV1 IV2 OLS IV1 IV2 OLS IV1 IV2
(1) (2) (3) (4) (5) (6) (7) (8) (9) Panel A Age of marriage .016 .124 .106 .001 -.005 -.005 .002 .002 .002 (.006)* (.012)** (.308)** (.001) (.002)* (.004) (.000) (.002) (.003) Age -.005 -.116 -.138 -.001 -.032 -.020 -.001 -.003 .003 (.002)* (.197) (.259) (.001) (.015) (.031) (.001) (.005) (.004) Hindu .134 .107 .112 -.003 -.002 -.001 .005 -.005 -.005 (.056)* (.068) (.089) (.007) (.006) (.009) (.004) (.004) (.005) Observations 2097 2097 2097 2097 2097 2097 2097 2097 2097 Panel B Age of marriage .011 .120 .090 .001 -.003 -.001 .001 .002 .001 (.018) (.051)* (.054) (.001) (.004) (.003) (.001) (.003) (.003) Age -.012 .436 .374 -.001 -.038 -.034 -.001 -.004 -.003 (.009) (.249) (.227) (.001) (.023) (.023) (.001)** (.007) (.006) Hindu .190 .185 .190 .002 .003 .003 -.006 -.006 -.006 (.149) (.153) (.150) (.011) (.011) (.011) (.004) (.004) (.004) Observations 1063 1063 1063 1063 1063 1063 1063 1063 1063
20
Number of Anti-Tetanus Birth
Prenatal Visits Shot Complications
OLS IV1 IV2 OLS IV1 IV2 OLS IV1 IV2
(10) (11) (12) (13) (14) (15) (16) (17) (18) Panel A Age of marriage .036 -.052 -.044 .000 .012 .015 .001 -.001 -.001 (.011)** (.044) (.046) (.001) (.005)* (.008) (.000)** (.000) (.001) Age -.077 -.840 -.830 .001 .190 .192 -.001 -.000 -.000 (.011)** (.385)* (.382)* (.001) (.047)** (.097)* (.001) (.001) (.002) Hindu .203 .225 .223 .017 .014 .013 -.003 -.003 -.003 (.108) (.114)* (.113)* (.006)* (.008) (.013) (.001)** (.001)** (.002) Observations 2097 2097 2097 2097 2097 2097 2097 2097 2097 Panel B Age of marriage .024 .011 .023 .002 .005 .005 -.001 -.000 .000 (.011)* (.043) (.042) (.001) (.003) (.003) (.002) (.001) (.000) Age -.072 .036 .060 -.002 .011 .011 -.000 -.000 .000 (.014)** (.096) (.097) (.001) (.009) (.002) (.002) (.001) (.000) Hindu .110 .110 .109 -.003 -.003 -.003 -.003 -.002 -.002 (.148) (.148) (.148) (.016) (.016) (.016) (.002) (.002) (.002) Observations 1063 1063 1063 1063 1063 1063 1063 1063 1063
21
Antenatal Giving Birth General health
Visit at Home of mother
OLS IV1 IV2 OLS IV1 IV2 OLS IV1 IV2
(19) (20) (21) (22) (23) (24) (25) (26) (27) Panel A Age of marriage .013 .005 .007 -.002 .007 .006 .157 .016 .017 (.003)** (.014) (.014) (.001) (.002)* (.006) (.225) (.015) (.014) Age -.035 -.258 -.254 .003 .016 .016 -.073 .222 .223 (.004)** (.113)* (.113)* (.003) (.011) (.011) (.042) (.165) (.165) Hindu .107 .109 .108 -.026 -.028 -.027 .280 -.124 -.124 (.042)** (.042)** (.042)** (.005)** (.005)** (.020) (.356) (.045)** (.045)** Observations 2097 2097 2097 2097 2097 2097 2097 2097 2097 Panel B Age of marriage .006 .014 .017 .000 .004 .003 -.325 -.037 -.033 (.004) (.016) (.016) (.002) (.005) (.005) (.338) (.025) (.021) Age -.034 -.036 .044 .003 .009 .007 -.115 .084 .092 (.005)** (.040) (.042) (.003) (.014) (.013) (.045)** (.201) (.206) Hindu .063 .062 .062 -.060 -.060 -.060 .447 -.074 -.076 (.055) (.055) (.055) (.032) (.032) (.032) (.605) (.071) (.070) Observations 1063 1063 1063 1063 1063 1063 1063 1063 1063
NOTE. –Data are taken from the 1996 MHSS. OLS and IV regressions; dependent variable is pregnancy and birth related outcomes. IV estimate is age of menarche. The dependent variable in cols. 1 to 3 is the age of first birth of the mother; the dependent variable in cols. 4 to 6 is the likeliness of miscarriage during the first pregnancy; the dependent variable in cols. 7 to 9 is whether the female experienced any pregnancy complications, which are defined as: excessive bleeding, convulsions, enema or anaemia; the dependent variable in cols. 10 to 12 is the number of visits to a prenatal clinic during the first pregnancy; the dependent variable in cols. 13 to 15 is an indicator of a
22
prenatal anti-tetanus shot received during the first pregnancy; the dependent variable in cols. 16 to 18 is the likeliness that the female experiences any birth related complications, defined as: excessive bleeding, convulsions or revered leg position; the dependent variable in cols. 19 to 21 an indicator of antenatal care during the first pregnancy; the dependent variable in cols. 22 to 24 is the likeliness that the female gave birth at home to her first born child; the dependent variables in cols. 25 to 27 is the likeliness that the female experiences any health related problems in adulthood, defined as: diabetes, arthritis, gastritis, respiratory disease, urinary infection or other illness. Panel A includes the whole sample, while Panel B is restricted to females with parents who did not obtain any education. Standard errors are in parentheses and account for individual roster weights and sample clustering (baris). Regressions also include 5-year age dummies and are robust to single-year age fixed effects.
* Significant at the 5 percent level. ** Significant at the 1 percent level.
1 Instrument is a continuous measure of age of menarche
23
results that is close to zero5. An interpretation of this pattern is that only a small sample of the females in the sample gives birth within three years of their first menstruation. Giving birth post the three year limit empirically suggests that the female body is fully physically developed to bear a child without problems that are related to her maternal development. The large amount of females bearing their first child post three years of menarche, 87 percent of the sample, is likely to be driving the results; hence the estimates show little to no correlation to age at marriage.
B. Demand for health care behaviour during pregnancy
Other outcome variables analysed in this paper are related to the health care utilization of females during their first pregnancy. Past research has shown that increased schooling, decision-making power and economic means may positively influence the likeliness of females appropriating health care during and after pregnancy. However, Bangladesh suffers from low ages at first marriage, which negatively influences both schooling attainment, empowerment and income.
Columns 10-12 show how age at marriage affects the rate at which a woman seeks prenatal care during her first pregnancy6 and Columns 19-21 look at the likeliness of an antenatal visit7. Delaying marriage by one year has a negative but insignificant effect on the number of antenatal visits, even when the sample is reduced to females with non-educated parents. The effect is close to zero and insignificant for the latter estimate. In a study on an almost identical sample of females from the MHSS survey in Bangladesh, a strong positive correlation was found, which is in line with empirical suggestions of correlation (Field and Ambrus, 2008). However, the study doesnot limit the outcome variable to first born children, but to the average over all
pregnancies. One interpretation of the divergence in the result is that marriage age itself has no direct effect on utilization of prenatal care, but rather through other factors.
From Table I, the mean utilization rate of antenatal care for the whole sample is approximately 30 percent. For reasons unexplored in this paper such as income and education, a large amount of females might be constrained in seeking medical care. Hence, regardless of marriage and child bearing age, women universally demand health care but are for unknown reasons constrained to appropriate it. The positive correlation found by Field and Ambrus (2008) may stem from the almost 30 percent of women utilizing health care during the first pregnancy, who continue to and even increase utilization because of their higher educational attainment, income and
6
The variable was constructed for the first pregnancy, where a survey question asked: ”How many antenatal visits did you have during your first pregnancy?”
8
The variable was constructed for the first pregnancy, where a survey question asked: ”When you were pregnant with your first child, were you given an injection in the arm to prevent the baby from getting tetanus, that is, convulsions after birth?”
7 The variable was constructed for the first pregnancy, where a survey question asked:
24
empowerment. With higher educational attainment, women are more likely to understand the risks associated with pregnancy and gains from seeking health care. More education, higher income and thereby also empowerment within the family may result in increased spending for health care in general and more so on prenatal care, which is what the literature on the topic suggests. In sum, the close to zero estimates may result from the universal demand of health care during the first pregnancy, regardless of age, but may simultaneously be constrained for specified reasons. The minority of women appropriating health care may continue and increase the
utilization over the following pregnancies, which explain the results found by Field and Ambrus (2008).
Columns 13-15 indicate the likeliness of a pregnant woman receiving an anti- tetanus shot while being pregnant8. A one year delay in the age at first marriage is associated with a significant increase of 1.2 percent in the likelihood of receiving an anti-tetanus shot. The estimate suggest that females who postpone marriage may, through for instance increased education or income, utilize some forms of health care slightly more than females who marry at a younger age.
Columns 22-24 look at the likeliness of giving birth at home. The estimate is significantly positive but close to zero. By looking at the sample statistics from Table I, almost all women in the sample give birth at home. Although, like previous
discussion on channels of influence for health care utilization, women desire to give birth at hospitals or health care centres, they may for unknown reasons be restricted to do so. Whether it is education, income or distance to health facility that drives the correlation down towards zero, age of marriage is not a determinant for the likeliness of giving birth at home.
Hence, the estimates related to health care behaviour during the first pregnancy do not seem to be largely affected by age at marriage, and the impact of marriage age on these outcomes is small and negligible. Following previous research, it seems likely that the effect is only prevalent on average over many pregnancies due to education, income or other unexplored factors.
C. Health of the mother
The health condition of the mother may be affected by the age when she first married. Previous research points towards more health problems associated with early
marriage, where the effect is indirect through age of first child bearing. Columns 25-27 show insignificant results and indicate that there is no correlation between age of first marriage and adverse health outcomes. The columns represent the outcome variable of adult health, which comprise of several different illness conditions. It shows the likeliness of having suffered from at least one of the health problems, comprising of diabetes, gastritis, respiratory disease, urinary infection or other illness. In line with previously discussed findings, it is likely that insignificant results are
25
found due to the timing of the first menstruation in relation to the age of first child bearing, as a majority of the sampled females give birth post the critical time when the female body is not fully matured to bear a child. Hence, in the sample of women living in a rural part of Bangladesh, marriage age does not have an impact on adult health outcomes.
VI. Conclusion
This paper aimed to identify a relationship between marriage age and pregnancy and birth related outcomes. I intended to look at both health related problems as well as utilization of health care services during pregnancy and birth for two reasons. Firstly, health related consequences during pregnancy and birth are larger amongst
adolescents, whom also utilize less health care. Secondly, it fills an important gap in the existing literature on the topic of marriage markets. By using individual-level data from a rural part of Bangladesh from 1996, an instrumental variables framework was set up to provide for an interpretation of the results that deals with the endogeniety issue related to marriage age. Age of menarche is used as an instrument for age at marriage, as it strongly correlates with age at marriage and is assumed from previous research to be unrelated to confounding factors. The biological delay caused by the distributions of menarche in entering the marriage market significantly postpones the age of marriage. The findings provide empirical evidence that early marriage of girls has an effect on the age at which these girls bear their first child. A one year delay in marriage increases age at first birth by up to 0.12 years. The estimates for pregnancy and birth related consequences, as well as utilization of health care services, are close to zero and a majority are insignificant. The intuitive reasoning behind the
insignificant results of pregnancy and birth complications are likely to stem from the higher gynaecological age at first child bearing, where a large majority of the females in the sample bear children when their bodies are already fully maternally
development. In addition, the findings on health care utilization may be due to the analysis of first births’ where the demand for health care services is universal, regardless of marriage age.
In sum, the hypothesis that early age at first marriage has a negative effect on pregnancy and birth complications is invalidated by the results found in this analysis. Age at first marriage has an effect on age at first child bearing, but small and often insignificant results are found for other outcome variables. However, to be fully able to draw extensive conclusions from the research, further evidence from other
countries facing large numbers of child marriages is essential to understand its consequences during child bearing.
26 References
Bates, L. M., Islam, F., Islam, K. and Schuler, S. R. (2004), “Legal Registration of Marriage in Bangladesh: An Intervention to Strengthen Women’s Economic and Social Position and Protect Them against Domestic Violence?” Manuscript, Acad. Educ. Development, Washington, DC.
Begum, L. (2003), “Meaning Given to Adolescents’ Reproductive Health in Bangladesh.” J. Population, 9: 81-103.
Bharadwaj, R. (2009), ”Impact of Changes in Marriage Law: Implications for Fertility and School Enrolment.” Availablie at SSRN: http://ssrn.com/abstract=1091789 Bronson, F. H. (1987), ”Puberty in Female Rats: Relative Effect of Exercise and Food Restriction.” American J. Physiology-Regulatory Integrative and Comparative
Physiology 252, no. 1, pt. 2: R140-R144.
Campbell, B. C. and Udry, J. R. (1995), ”Stress and Age at Menarche of Mothers and Daughters.” J. Biosocial Sci. 27: 127-34.
Chakraborty, N., Islam, A., Chowdhury, R. I., Bari, W. and Akhter, H. (2003), “Determinants of the use of maternal health services in rural Bangladesh”, Health Promotion International, Oxford Journals, 18(4): 327-37.
Chowdhury, S., et al. (2000), ”Nutritional Status and Age at Menarche in a Rural Areas of Bangladesh.” Ann. Human Biology, 27(3): 249-56.
Coale, A. and Treadway, R. (1986), ”A summary of the changing distribution of overall fertility, marital fertility, and the proportion married in the provinces of Europé,” in The Decline of Fertility in Europe, Princeton University Press, pp. 31-181.
Dagnelie, O. (2010), ”Too Young to Marry: Evidence from Senegal on Child Mortality*” Available at:
http://www.csae.ox.ac.uk/conferences/2010-EdiA/papers/154-Dagnelie.pdf
Der, R., Fahim, Z., Hilderbrand, D. and Fahim, M. (1974), ”Combined Effect of Lead and Low Protein Diet on Growth, Sexual Development and Metabolism in Female Rats.” Res. Communications Chemical Pathology and Pharmacy, 9: 723-38.
Duflo, E. (2011), ”Women’s Empowerment and Economic Development.” National
27
Felice, M. E., James, M., Shragg, P. and Hollingsworth, D. R. (1984), ”Observations related to chronologic and gynecologic age in pregnant adolescents.” Yale Journal of
Biology and Medicine, 57(5): 777-85.
Field, E. and Ambrus, A. (2008), ”Early Marriage and Female Schooling in Bangladesh.” Journal of Political Economy, 116(5): 881-930
Godha, D., Hotchkiss, D. and Gage, A. (2011), ”The Influence of Child Marriage on Fertility, Fertility-Control, and Maternal Health Care Utilization: A Multi-country Study from South Asia.” MEASURE Evaluation PRH. Working Paper 11-124. Girls not brides, (2015). URL: http://www.girlsnotbrides.org/child-marriage-law/ Harwoon-Lejeune A. (2000), ”Rising Age at Marriage and Fertility in Southern and Eastern Africa.” European Journal of Population, 17: 261-80
Jayachandran, S. and A. Lleras-Muney (2009), “Life Expectancy and Human Capital Investments: Evidence from Maternal Mortality Declines,” Quarterly Journal of
Economics 124(1): 349-98.
Jensen, R. and Thornton, R. (2003), ”Early Female Marriage in the Developing World.” Gender and Development, 11(2): 9-19
Kaprio, J. Arja, R. Torsten, W. Viken, R. and Rose, R. J. (1995), ”Common Genetic Influences on BMI and Age at Menarche.” Human Biology, 67: 739-53.
Kirdar, M. G., Tayfur, M. D. and Koc, I. (2011), ”The Effects of Compulsory
Schooling Laws on Teenage Marriage and Births in Turkey.” Institute for the Study of
Labor (IZA). Discussion Paper No. 5887.
Malhorta, A. (1997), ”Gender and the Timing of Marriage: Rural-urban differences in Java.” Journal of Marriage and the Family, 59(2): 434-50.
McLachlan, J. A. and Dixon, R. L. (1977), ”Toxicologic Comparison of Experimental and Clinical Exposure to Diethylstilbestrol during Gestation.” In Advances in Sex
Steroid Hormone Research, vol 3. Baltimore University Park Press.
Mucignat-Caretta, C. Caretta, A. and Cavaggioni, A. (1995), ”Acceleration of Puberty Onset in Female Mice by Male Urinary Proteins.” J. Physiology, 486: 517-22.
Nasrullah, M., Muazzam, S., Bhutta Z. A. and Raj, A. (2014), ”Girl Child Marriage and its Effect on Fertility in Pakistan: findings from Pakistan Demographic and Health Survey, 2006-2007.” Maternal and Child Health Journal. 18(3): 534-43.
28
Navaneetham, K. and Dharmalingam, A. (2002), “Utilization of maternal health care services in Southern India”, J. Social Science and Medicine, 55(10): 1849-69.
Nazian, S. J. and Piacsek, B. E. (1976), ”Vaginal Opening and Early Estrous Cycles in Rats Raised at a Low Ambient Temperature.” J. Experimental Zoology. 198: 13-16. Palloni, A. and Rafalimanana, H. (1999), ”The Effects of Infant Mortality on Fertility Revisited: New Evidence from Latin America.” Demography, 36(1): 41-58.
Palmert, M. R. and Boepple, P. A. (2001), ”Variation in the Timing of Puberty: Clinical Spectrum and Genetic Investigation.” Clinical Endocrinology and
Metabolism, 86: 2364-68.
Rahman, O. et al. (1999), ”The Matlab Health and Socioeconomic Survey: Overview and User’s Guide”, DRU-2018/1, RAND; Santa Monica CA.
Raj, A., Niranjan, S., Balaiah, D., and Silverman, J. G. (2009), ”Prevalence of child marriage and its effect on fertility and fertility-control outcomes of young women in India: a cross-sectional, observational study.” Lancet 2009; 373: 1883-89.
Reynolds, H. W., Wong, E. L. and Tucker, H. (2006), ”Adolescent’s Use of Maternal and Child Health Services in Developing Countries.” International Family Planning
Perspectives 32(1): 6-16.
Rosero-Bixby, L. (1996), ”Nuptiality trends and fertility transition in Latin America”, in Guzmán, J. M.; Singh, S.; Rodríguez, G., Pentelides, E.A. (Editors), The Fertility Transition in Latin America, Oxford University Press, pp. 135-50.
Santhya, KG., Ram, U., Acharya, R., Jejeebhoy, SJ, Ram, F. and Singh, A. (2010), “Associations between early marriage and young women’s martial and reproductive health outcomes: evidence from India.” Int. Perspect Sex Reprod. Health, 36(3): 132-39.
Scott, E. C., and Johnston, F. E. (1985), ”Science, Nutrition, Fat, and Policy: Tests of the Critical-Fat Hypothesis.” Current Anthropology, 26(4): 463-73.
Sekhri, S. and Debnath, S. (2014), ”Intergenerational Consequences of Early Age Marriage of Girls: Effect on Children’s Human Capital” The Journal of Development
Studies, 50(12): 1670-86.
Senderowitz, J. (1995), ”Adolescent health: reassessing the passage to adulthood”, No. 272 World Bank Discussion Papers, Washington, D.C.
29
Scholl, T. O., Hediger, M. L. and Belsky, D. H. (1994), ”Prenatal care and Maternal Health During Adolescent Pregnancy: A Review and Meta-analysis.” Journal of
Adolescent Health, 15: 444-56.
Stevens-Simon, C., Beach, R. K. and McGregor, J. A. (2002), ”Does Incomplete Growth and Development Predispose Teenagers to Preterm Delivery? A Template for Research” Journal of Perinatology, 22(4): 315-23.
United Nations, (1948), “Universal Declaration of Human Rights”. Office of the High Commissioner for Human Rights.
UNICEF. (2013), ”Ending Child Marriage: Progress and Prospects”, Data and Analytics Section, New York.
”UNFPA, State of the World Population 2005”, United Nations Population Fund. Vandenbergh, J. G. (1969), ”Male Odor Accelerates Female Sexual Maturation in Mice.” Endocrinology, 84: 658-60.
