Knowledge Brief Health, Nutrition and Population Global Practice EARLY CHILDBIRTH AND UNDER-FIVE MALNUTRITION IN BANGLADESH Adenike Onagoruwa and Quentin Wodon June 2017 Child Marriage Series with Education Global Practice KEY MESSAGES:  In Bangladesh, more than three in ten children under the age of five are stunted; but for children born of mothers younger than 18, the risk of stunting is higher.  Controlling for socio-economic and other characteristics, being born of a mother younger than 18 increases the likelihood of stunting for children under-five by five percentage points in the baseline model, as compared to otherwise similar children born of older mothers. Box 1: Brief and Series Primer Nearly one-fifth of under-five deaths could be prevented with optimal feeding. Poor nutrition at a young age may How is early childbirth defined? Early childbirth is defined in also have irremediable consequences for brain this brief as a child being born of a mother younger than 18. Early development, cognitive skills, and productivity in adult life. childbearing is often related to the practice of child marriage. Malnutrition has severe consequences for children. Why a series on child marriage? Child marriage has significant negative impacts – not only for girls, but also for a range of development outcomes. Demonstrating these impacts will assist Poor nutrition weakens children’s immune systems, putting governments and others to make the case for intervening to them at a greater risk of falling sick from preventable reduce the practice. illnesses such as pneumonia and diarrhea. According to Horton et al. (2008), nearly one-fifth of under-five deaths in What are the topics discussed in the series? The series looks the world could be prevented with optimal feeding. at the impacts of child marriage on health, population, education, Research also suggests that poor nutrition at a young age employment, agency, and violence, among other outcomes. The may have irremediable consequences for brain welfare, budget, and non-monetary costs of child marriage are development, cognitive skills, and ultimately productivity in estimated. Legal/institutional aspects and options to reduce the practice are also discussed. adult life. Unfortunately, a large share of children in the developing world are malnourished. The question What is the question asked in this brief? The question is: What considered in this brief is whether early childbirth (defined is the impact at the margin of early childbirth on the probability of as a child being born of a mother younger than 18), which malnutrition for children under-five years of age? in many countries is the result of child marriage, contributes to under-five malnutrition in a significant way in How is the question answered? Econometric analysis of Bangladesh. The brief is part of a series of similar Demographic and Health Survey data is used to estimate the standardized country-specific briefs on the same topic for impact of early childbirth on under-five malnutrition. a number of countries. Page 1 HNPGP Knowledge Brief  Statistically, children from young mothers in marginal impacts in percentage terms. For example, a Bangladesh are four percentage points more likely to statistically significant coefficient of 0.05 for a mother be stunted than if the mother is between 18 and 34 younger than 18 would indicate that children of very young years of age. mothers have a likelihood of stunting five percentage points higher than otherwise similar children of older mothers. The focus in this brief is on stunting as a measure of persistent exposure to malnutrition with potentially severe With the baseline specification, table 2 indicates that, at a long-term consequences throughout a person’s life (see ten percent level, deliveries at a young age significantly box 2 on indicators used to measure malnutrition). The increase the likelihood of stunting for the children by 5.2 analysis is based on data from the 2014 Demographic and percentage points in comparison to a delivery at 18 to 34 Health Survey for Bangladesh. Estimates suggest that years of age (in Bangladesh). The difference in risk of 39.19 percent of children born of mothers younger than 18 stunting between mothers ages 18-34 and mothers above are stunted. The proportion is still high, but four points 35 is not statistically significant. lower at 35.27 percent for children born of mothers 18 to 34 years of age. For children of mothers older than 35, the Deliveries at a young age may increase the likelihood of incidence of stunting was higher at 44.30 percent. stunting for the children by 5.2 percentage points in comparison to a delivery at 18 to 34 years of age. Table 1: Incidence of Stunting by Age of the Mother Age of the mother Stunting (%) Mother younger than 18 39.19 When additional controls are added (extended model), the Mother in 18-34 age bracket 35.27 marginal effect of early childbirth changes. There is thus Mother older than 35 44.30 some evidence that in the case of Bangladesh, after Source: Authors. controlling for a wide range of other variables, early childbirth may not contribute directly (controlling for other Box 2: Measures of Malnutrition variables) to stunting, but prudence remains needed when interpreting these results given the risk of omitted variable Three main measures of malnutrition are used in applied work. A bias (see box 3). child is considered underweight if s/he has a weight more than two standard deviations below the reference median weight for Table 2: Impact of Early Childbirth on Stunting the child’s age. A child is considered wasted if s/he has a weight Age at first marriage Baseline Extended to height ratio more than two standard deviations below the model model median weight for height for the reference population. A child is Mother younger than 18 0.0518* 0.0361 considered stunted if s/he has a height more than two standard Mother in 18-34 age bracket Reference Reference deviations below the median reference height for that age. If a Mother older than 35 0.0632 0.0620 child on any of these measures is below three standard deviations of the norm, s/he is considered as severely underweight, wasted, Source: Authors. or stunted. Among the three measures, stunting and wasting tend Levels of statistical significance: *** 1%, ** 5%, * 10%. to be used the most. Stunting often results from persistent insufficient nutrient intake and infections. It may lead to delayed Box 3: Risk of Omitted Variable Bias motor development and poor cognitive skills that can affect school performance as well as productivity and earnings later in life. In many countries, although not in Bangladesh, early childbirth Wasting tends to result more from acute food shortage or disease appears to be positively correlated with the risk of stunting after and may lead to death. For the purpose of this brief, given a controlling for other factors that may also contribute to stunting. separate brief of under-five mortality, stunting is the best measure This could indicate a causal effect. However, other variables to focus on. correlated with both early childbirth and stunting not included in the analysis could be at the source of the correlation between early childbirth and stunting. Because of the risk of omitted Controlling for other factors, early childbirth still variable bias, the results cannot be considered as fully conclusive increases the likelihood of stunting substantially. regarding a causal impact of early childbirth on the risk of stunting. The difference in the likelihood of stunting between children A number of results from the regression analysis not shown of young and older mothers does not necessarily imply a in table 2 are worth mentioning. The impact of wealth on causal effect of the age at delivery, but it does suggest that the likelihood of stunting is statistically significant, with early childbirth may contribute to stunting. To check children from wealthier households tending to be less whether controlling for other factors early childbirth is stunted. For example, compared to children in the poorest indeed associated at the margin with higher under-five 20% of households, those from the fourth and fifth quintiles malnutrition, regression analysis is used (see the annex for have a likelihood of stunting ranging from 8-14 percentage details on the methodology). points lower. A number of other effects are also statistically significant, as discussed in the more detailed study on Table 2 provides key results with baseline and extended which the brief is based. As shown in table 3, the marginal models. The interpretation of the coefficients is in terms of impact of a mother having a secondary education or better Page 2 on the likelihood that her child will be stunted is not corresponds (roughly) to the product of the marginal effect statistically significant, which is somewhat surprising. of early deliveries on stunting (0.052 in table 2) times the share of children born of mothers younger than 18. Given Note though more generally that the inclusion of education the rate of stunting nationally, two in 100 stunted children as a control points to the possibility of indirect effects of could be considered as stunted due to the direct effect of early childbirth on stunting. Because early childbirth may early childbirth on the likelihood of stunting. This may have an impact on other variables used as controls in the appear low in comparison to the total number of children regression, its overall effect on stunting, including indirect who are stunted, but still represents a large number of effects through these other variables, may be larger than children. the direct effect documented in table 2. For example, for some girls having a baby at a young age, early childbirth Because only a small share of deliveries are by mothers could have reduced education attainment, which could lead younger than 18, only two in 100 stunted children can be to a higher risk of stunting (although not in Bangladesh said to be stunted directly due to early childbirth. according to the regression results). In addition, early deliveries, by increasing the number of household Conclusion members may also contribute to lower standards of living. In Bangladesh, as mentioned earlier, the regression results Early childbirth contributes to the risk of malnutrition for suggest that the level of welfare as measured through children, directly and indirectly. This brief has provided wealth quintiles have an effect on stunting. Still, in terms of estimates of the direct impact of early childbirth on stunting magnitude, those indirect effects are likely to be small in in Bangladesh using the latest DHS survey. Four in ten comparison to the direct effects in table 2. children under the age of five are stunted. For children born of mothers younger than 18, the risk of stunting is higher Early childbirth may affect under-five malnutrition through by four percentage points versus mothers ages 18-34. lower education attainment for mothers or lower socio- Controlling for socio-economic and other characteristics, economic status, but these effects are likely to be smaller. being born of a mother younger than 18 appears to increase the likelihood of stunting by five percentage points Table 3: Impact of the Mother’s Education on Stunting versus children born of older mothers in the baseline Age at first marriage Baseline Extended model. Given the share of children born of mothers model model younger than 18, nationally for every 100 stunted children, No education or below primary Reference Reference at least two could be stunted directly because of early Primary education -0.0267 -0.0176 childbirth. This may appear low, but still represents a large Secondary education or higher -0.0532* -0.0429 number of children. Source: Authors. Levels of statistical significance: *** 1%, ** 5%, * 10%. References Overall, the results suggest that as a first order Fall, C. H., et al., 2015, Association between maternal age at approximation, one may rely on the estimated direct effects childbirth and child and adult outcomes in the offspring: a of early childbirth on stunting when simulating the impact prospective study in five low-income and middle-income countries of the elimination of early childbirth on malnutrition in the (COHORTS collaboration), Lancet Glob Health 3(7):e366-77. country. Doing so does likely entail underestimation of the overall effects, but probably not by a wide margin. Finlay. J. E., E. Özaltin, and D. Canning, 2011, The association of maternal age with infant mortality, child anthropometric failure, Simulations suggest that about two in 100 stunted diarrhoea and anaemia for first births: evidence from 55 low- and children are stunted directly due to early childbirth. middle-income countries, BMJ 1:e000226. Horton, S., H. Alderman, and J. Rivera. 2008. “Copenhagen The last step in the analysis consists in assessing the Consensus 2008 Challenge Paper: Hunger and Malnutrition.” potential impact of eliminating early childbirth on stunting. http://www.copenhagenconsensus.com. This is done by predicting (i.e. simulating) the likelihood that children who were born of mothers younger than 18 National Institute of Population Research and Training (NIPORT), would have been stunted if they had been born of older Mitra and Associates, and ICF International 2016. Bangladesh mothers. In other words, we are considering the direct Demographic and Health Survey 2014. Dhaka, Bangladesh, and effects of the age of the mother on stunting, shifting in the Rockville, Maryland, USA: NIPORT, Mitra and Associates, and data deliveries by young mothers to deliveries at a later ICF International age, and observing the difference that this makes for stunting rates nationally. The simulations suggest that without early deliveries, the share of children stunted could decrease by about 0.75 percentage points. This Page 3 HNPGP Knowledge Brief  Annex: Methodological Note regressions, the dependent variable is whether a child is stunted or not. There is an existing literature on the relationships between early childbirth, child marriage, and the risks of under-five malnutrition Different specifications are estimated to assess the robustness of (e.g., Fall et al., 2015, and Finlay et al. 2011). This literature the results to the choice of models. Overall, the results are fairly suggests that children born of young mothers are at higher risk of robust to different specifications. For the baseline model, the malnutrition. How much higher in the case of Niger is the question independent variables are the following: (1) the age of the mother to be answered. To answer this question, this brief focuses on at the time of delivery by categories; (2) the child’s age and stunting as the measure that tends to best capture persistent gender; (3) whether the child had siblings born at the same time exposure to malnutrition with likely long term negative (multiple birth); (4) the birth order of the child and the child’s birth consequences for children. weight by categories; (5) the length of time between the child’s birth and a previous birth for the mother; (6) whether the child has There is no doubt that children born to very young mothers are received recommended immunizations; (7) whether the delivery more likely to be malnourished than other children, but this might took place in a health facility and was attended by skilled not specifically be due to the fact that the mothers are young. Girls personnel; (8) the mother’s height and education level, as well as who have children before the age of 18 tend to be poorer. This whether she works and the type of work involved; (9) the father’s implies that their children are at higher risk of inadequate occupation and his level of education; (10) the location of the child nutritional intake. Early pregnancies are more common in areas by region and by urban-rural category; (11) whether the where access to healthcare is limited to prevent or treat household has access to an improved water source and improved malnutrition. Young girls also often suffer from a lack of agency sanitation; (12) whether the household has more than two and decision-making power in the household, which may reduce children under-five; and finally (13) the wealth quintile of the their ability to seek care for their children when needed. Girls who household. give birth early are likely to have dropped out of school due to pregnancy or marriage, which may also affects the nutritional In the extended model, additional controls are added: (14) status of their children due to lack of education or knowledge on whether the household practices polygyny; (15) whether the how to best take care of young children. distance to health facility is a major problem for the household; (16) the age gap between the spouses; (17) indicators of These risk factors correlated with early childbirth do not decision-making power for the mother; (19) tolerance towards necessarily imply that early pregnancies by themselves contribute wife beating; and (20) whether the mother is able to get in a direct way to child malnutrition. Controlling for other factors, permission to access healthcare. it could be that early childbirth does not lead to a higher risk of malnutrition for children. But it could also be that there is a direct causal link between early childbirth and child malnutrition, for example if some young mothers giving birth are not yet be ready This brief was produced by a World Bank team as part of the Economic physiologically to give birth, which could in turn affect the health Impacts of Child Marriage study. The synthesis report under the study was of their children. This brief estimates the direct impact of early produced jointly with the International Center for Research on Women. childbirth on under-five malnutrition. The study benefitted from support from the Bill & Melinda Gates Foundation, the Children’s Investment Fund Foundation, and the Global Partnership for Education. Comments from colleagues and peer In addition, the brief provides an assessment of the extent to reviewers are gratefully acknowledged. The opinions expressed in this which under-five malnutrition would be reduced if early brief are those of the authors only and need not reflect the views of the pregnancies/deliveries were eliminated. In order to measure the World Bank, its Executive Directors, of the countries they represent. potential impact of early childbirth at the margin on stunting for children under the age of five, regression analysis is used. In the more detailed paper on which this brief is based, both tobit regressions (to measure the degree of stunting among children who are stunted) and probit regressions (to measure the likelihood of stunting) are provided. In this brief, due to space constraints and because of the interest in the share of stunting that could be attributed to early childbirth, the focus is on reporting results from probit regressions. In those The Health, Nutrition and Population Knowledge Briefs of the World Bank are a quick reference on the essentials of specific HNP-related topics summarizing new findings and information. These may highlight an issue and key interventions proven to be effective in improving health, or disseminate new findings and lessons learned from the regions. For more information on this topic, go to: www.worldbank.org/health. Page 4