key: cord-0688418-yx89ofa2
authors: Alderman, Harold; Nguyen, Phuong Hong; Tran, Lan Mai; Menon, Purnima
title: Trends and geographic variability in gender inequalities in child mortality and stunting in India, 2006–2016
date: 2021-03-14
journal: Matern Child Nutr
DOI: 10.1111/mcn.13179
sha: 96d7c37fdb3f0d62ad806110a7a9aad5e5f61fe4
doc_id: 688418
cord_uid: yx89ofa2

Gender disparities in child undernutrition and mortality in India have been a topic of interest for a long time, but little is known on trends or geographic variability in recent periods. We examined the degree to which historic patterns in gender disparities in child undernutrition and mortality in India have persisted given recent progress in health and nutrition. Using two nationally representative datasets from India between 2006 and 2016, we estimated mortality rates and stunting by gender and by birth order among children under 5 years old. We then tested for differences between boys and girls within each survey round for both national and state levels using bootstrapped standard errors, controlling for cluster and sampling weights. We found striking progress in child mortality and stunting in India between 2006 and 2016 for both boys and girls. Boys were more likely to die than girls during the first year of life. Girls had a higher risk of mortality between age 1 and 5 years than boys in 2006, but the improvements in survival eliminated this gender gap in 2016. For stunting, we found no gender difference in 2006, but girls had higher height‐for‐age Z‐scores (HAZ) and lower stunting than boys in 2016. Trends in gender gaps in mortality and stunting vary substantially by birth order and between states. Our findings indicate that improvements in mortality and nutritional status among girls have started to close gender disparities. Policy efforts to close gaps must stay the course in states that have made progress and be accelerated in states where disparities are still prominent.

There has been much discussion about pervasive gender disparity in mortality in Asia in general and India specifically. This pattern was brought to the fore with a dramatically titled paper 'More than 100 Million Women are Missing' (Sen, 1990) . Few researchers doubt the basic population gender disparity in India, but there is debate as to the ages at which it occurs and the degree to which it is evolving.

For example, a study using survey data from the 1990s claims that differences in under 5 mortality are sufficient to explain the population gender disparity in India (Oster, 2009) , whereas a later study in 2010 employs World Health Organization (WHO) population data and concludes that the missing women reflect adult mortality (Anderson & Ray, 2010) . There is also some debate as to whether gender disparity in mortality has increased or decreased since the 1990s. Whereas some studies show evidence of an increase in gender disparity in mortality in India at the end of the 20th century (Das Gupta & Mari Bhat, 1997) , others see a global decrease in the number of missing women as a percentage of the population with India experiencing a smaller decline than its neighbours yet having movement in the same direction (Klasen & Wink, 2002) .

Researchers have also pointed out that gender patterns in child mortality differ substantially by age; neonatal mortality has different underlying causes as well as different outcomes than mortality in later childhood (Kim et al., 2020; Million Death Study et al., 2010) . Preventable deaths in the former largely reflect maternal nutrition, antenatal care and unsafe delivery and thus are less mediated by genderspecific patterns in childcare (Million Death Study et al., 2010) . Such childcare patterns include, for example, the fact that girls in India may be breastfed for shorter periods than boys (Jayachandran & Kuziemko, 2011) or have longer delays in seeking care when the child is ill (Malhotra & Upadhyay, 2013; Mishra et al., 2004) .

Given that malnutrition is associated with a large share of mortality in India (India State-Level Disease Burden Initiative Malnutrition, 2019), evidence on gender patterns in undernutrition is often presented along with studies of mortality (Barcellos et al., 2014; Oster, 2009 ). India is known as a country with son preferences (Hvistendahl, 2011) where boys have historically had an advantage in height and weight and lower undernutrition prevalence relative to girls (Barcellos et al., 2014) . This, again, may be a reflection of differences in childcare, either in terms of time or in terms of expenditures.

These investments vary over the lifecycle and depend on the child's age as well as the outcomes of earlier decisions if they survive. These investments can either reinforce or counteract health disparities within a household (Alderman et al., 2017) . Thus, just as patterns in neonatal mortality may differ from mortality in later childhood years, gendered patterns in nutritional outcomes need not necessarily track disparities in mortality (Alderman et al., 2017 (Chakrabarti et al., 2019) . The interventions delivered through these programmes focused on addressing child mortality and undernutrition, notably through efforts to scale-up immunization, micronutrient supplementation and deworming, food supplementation, growth monitoring and management of severely acute malnourished children (Avula et al., 2013; Vir et al., 2013) . Research has shown the impact of this improved intervention coverage on reducing undernutrition and mortality at national and subnational levels . A recent study has also examined the trends in child mortality in India between 2000 and 2017 and reported substantial variations between the states and districts in the magnitude and rate of decline in mortality but did not analyse gender inequality (India State-Level Disease Burden Initiative Child Mortality, 2020).

Previous studies that attempted to explain sex differentials in child mortality (Arokiasamy, 2004; Gupta, 1987; Kuntla et al., 2014) or malnutrition in India and its associated factors (Corsi et al., 2015; Pillai & Ortiz-Rodriguez, 2015) used data before 2006. To our knowledge, only one recent study has explored the gender disparity in the changes of mortality using updated data in the last decade (Karlsson et al., 2019) . While the current study reaches core conclusions similar to that recent report, it also includes an exploration of the variability in these trends by state as well as evidence as to how trends in mortality compare to those in nutrition. This exploration is important because both health investments and social conditions that contribute gender differentials have not remained constant either in time or across states in India.

Specifically, this paper updates analyses on trends in mortality by gender in India, disaggregating neonatal mortality rates (NMR) from infant mortality rates (IMR) and U5MR. We compare rates in Indian • This study offers an in-depth analysis to ascertain the degree to which historic patterns in gender disparities in child undernutrition and mortality in India have persisted given recent progress in health and nutrition.

Our findings indicate that improvements in mortality and nutritional status among girls have close gender disparities.

• Staying the course to ensure that economic, social and health policies remain in place to support further improvements in health and nutrition will be important for India's girls; this is especially important in the con- 

Childhood mortality rates were estimated for Indian states, using mothers' reports on the date of birth of each of their children, their survival status and the dates of death or ages at death of deceased children. We applied the direct method as guided in the DHS reports, using the Stata Version 16.0 package 'syncmrates', which calculates age-specific mortality rates using the synthetic cohort probability method (Rutstein & Rojas, 2003) . This approach allows full use of the most recent data and is also specific for time periods. The reference period of these estimates is 5 years prior to the survey date and, thus, avoids overlap of cohorts.

IMR and U5MR are defined as the number of deaths per 1000 live births by the age of 1 and 5 years, respectively. Mortality rates for children aged 1-5 years are the differences between U5MR and IMR and are termed child mortality rates (CMR) in the analysis. NMR reflect births in the first 28 days of life, whereas postneonatal mortality (PMR) is the number of deaths that occur between 28 and 365 days per 1000 live births.

Child anthropometry was collected by trained and standardized field staff using standard methods, as described in NFHS-3 and NFHS-4 reports (IIPS, 2008 (IIPS, , 2018 . Recumbent length/height of the children was measured by Seca 417 infantometer for children below 2 years and Seca 213 stadiometer for children 2-5 years. Length/height was then converted into height-for-age Z-scores (HAZ), according to 2006 WHO child growth standards. Stunting is defined as < −2 Z-score of HAZ.

We estimated mortality rates by gender and birth order at the national level and also calculated mortality rates by gender at the state level.

For HAZ and stunting, we estimated weighted means and plotted the distributions of HAZ and stunting against child age and sex using a local polynomial smoother. Given the difference in growth patterns by age groups, we report various descriptive statistics of HAZ scores and stunting prevalence for each round of survey for children <2 and 2-5 years separately. For child mortality, we applied the 'syncmrates'

command together with the T-test option to examine the difference in mortality rates by child sex. This technique calculated bootstrapped standard errors and confidence intervals for the mortality rates and allows all Stata bootstrap options. For child HAZ and stunting, we tested for differences between boys and girls within each survey year using regression analyses with 'svy' command to control for the cluster sampling design and sampling weights used in the survey. Statistical significance was considered at three levels: p < 0.05, p < 0.01 and p < 0.001.

The mean age of children was 29.7 months in both survey rounds with 40% of children aged <2 years and 60% aged 2-5 years. The gender ratio was 108.2 boys per 100 girls in 2006 and 108.5 in 2016 ( Boys are more likely to die than girls in the first 28 days after birth ( Figure 1 Figure S1 ). However, there were appreciable differences across states; in Uttar Pradesh, Bihar and Madhya Pradesh, the decline in NMR for boys was far smaller than the decline in NMR for girls (Table S1 ). This contributed to the increase in the mortality disparity in that age bracket. Given that most deaths in the first year of life occur in the first month, the gender disparity in IMR tracks that in NMR ( Figure 1 and Table S2 ).

In 2006, girls who survived their first year had a higher risk of subsequent mortality before their fifth birthday than did boys ( Figure 1 ). This was mainly driven by wide disparities in CMR in Uttar

Pradesh as well as Rajasthan (Table S3 ). This pattern is also observed in the NFHS-1 and NFHS-2 (CMR for boys vs. girls was 29.4 vs. 42.1 in 1992 and 24.8 vs. 36.6 in 1999) ( Figure S1 ). By 2016, these disparities had largely disappeared in Rajasthan and Uttar Pradesh as well as the nation overall. U5MR reduced substantially in the last two decades ( Figure S1 ), and the gender differences in U5MR ( Figure 1 and In keeping with common risk factors of mortality and nutrition, patterns in nutrition largely follow those in mortality. As illustrated in Younger boys are shorter than girls, but these gaps close as the children approach their second birthdays, after which the curves eventually cross. Finally, the figure confirms that children in the 2016 sample were taller and less likely to be stunted than the corresponding age and gender counterpart in the 2006 sample.

There was a small but significant gender gap in average HAZ for children under 2 in 2016, with boys having lower HAZ than girls (Table 3 ). The differences in stunting in 2016, however, was appreciable, and, as with average HAZ, the gender gap is slightly to the detriment of boys. In partial contrast, there is no gender difference in HAZ at the national level for children 2-5 years and only a small difference in stunting, with girls at a small disadvantage. Over the entire under 5 population in 2016, girls had higher HAZ and lower stunting than boys (Table 3) .

Stunting prevalence increases and mean HAZ declines with higher birth order. These patterns are monotonic in all ages and survey years, although the gender difference is not quite monotonic; the point estimate for the gap closes for the relatively few children who are the fourth born or of higher birth order compared with those who are third.

Just as there are differences in overall HAZ and stunting across states, there are differences in gender disparities (Tables S5-S8 1990s (Tarozzi & Mahajan, 2007) , but by the time of the third NFHS, the disparity in nutrition status had largely been eliminated (Corsi et al., 2015) . Although a previous study found small improvements in mortality, the probability of mortality for girls was higher than for boys in the north and lower in the south at the end of the last century (Tarozzi & Mahajan, 2007 Our results on mortality differentials contrast slightly with a study using UNICEF data, which reported that in 2015, male U5MR exceeded that of females in 193 out of 195 countries (Iqbal et al., 2018) , India being an exception. Although more recent UNICEF estimates still show a slightly lower U5MR for boys in India (UNICEF, 2019), the decline in the male to female ratio from 0.94 in 2015 to 0.97 in 2019 is similar to the trend derived from NFHS data indicated in Figure 1 showing a crossover to slightly higher male mortality.

Results for child stature also differ from those in other studies (Jayachandran & Pande, 2017; Oster, 2009) . What may account for this difference? One possible reason is that some studies compare z-scores in India with those in Africa despite the fact the z-scores are already standardized relative to the same international norms. As there is a pervasive pattern of boys being shorter than girls in Africa (Svedberg, 1990; Wamani et al., 2007) , such a comparison is not based on a neutral benchmark. Another reason for the difference with previous studies is the simple fact that the impressive improvements in stature in India between 2006 and 2016 have been slightly faster for girls. Although the current study does see a small disparity in stunting to the disadvantage of girls of higher birth order relative to boys, this difference is far smaller than the difference in stunting between either boys or girls of earlier birth order and their gender counterparts of latter birth order.

This study is not designed to address why the trends are not parallel, but we can reference reasons that plausibly contribute to the reduction of both gender differences in anthropometry and in mortality. A previous study offers a model postulating that whether gender discrimination is due to market returns to investments, greater propensity of boys to contribute to their parents' household or simply a greater indirect satisfaction of parents from their son's welfare than their daughter's, gender differences in investments should decline as resources increase (Alderman & Gertler, 1997) . That study also postulates that households will be more responsive to prices in the case of girls, including changes in the availability of services (Alderman & Gertler, 1997) programmes, many states also rolled out a range of cash incentive programmes targeted towards the education and support of girls (Sekher, 2012) .

Researchers have explored the origins and persistence of son preferences in terms of cultural features including religion and patrilocal inheritance patterns across regions of India (Jain, 2014) .

Although the current study does not attempt to parse out the relative contribution of culture, we note that culture seldom changes as rapidly as did relative mortality rates in the period studied. In contrast, improved access to health services has been dynamic. Although a portion of gender difference in anthropometry reflects anticipation of future child investments, the main contribution to nutritional differences accrues in children of higher birth order (Jayachandran & Pande, 2017) . This is not necessarily direct discrimination of later born children but rather a reflection that birth order and the number of siblings are correlated with household resources (Spears et al., 2019) .

This, again, implies that economic progress-as well as any other trends that result in reduced fertility-will also reduce gender imbalances.

How much of the reduction in gender disparity is due to trends in fertility? We explored this by assuming that the share of children born in different birth orders in 2006 was the same as observed in 2016.

That is, we assume that a larger share of children was first or second born than actually was observed in the earlier year. We also assumed that the mortality rates and stunting prevalence per gender and birth order remained as they were in 2006. Shifting to this scenario, however, makes very little difference. We estimated only a drop in overall mortality from 74.1 to 72.5 and stunting from 47.2% to 46.2%. This is about 6% of the total decline in mortality and 2% of the total decline in stunting. Although this estimate does not take into account that birth order is related to overall family size, which is not included in this exercise, it illustrates that the direct contribution of birth order changes over the decade to the favourable mortality and nutrition trends is modest. This ties to an additional possible reason that improvements in mortality and nutritional status have been faster for girls than boys in recent years (Anukriti et al., 2020) , and the availability of sex selection technology has benefited survival rates of girls by 2006 additional to any improvements in health services that benefit both genders. To the degree that this occurs, it contributes to the observed decline. But because mortality rates are defined as a share of live births, the reported mortality that is analysed here is not a biased estimate even if it differs from a counterfactual of what might have occurred in the absence of ultrasound technology (Anukriti et al., 2020; Hu & Schlosser, 2015) . These authors do not deny the prevalence of son preference but claim that the preference is manifested differently with modern technology. Nor do the authors downplay other negative consequences of sex selection (Amaral & Bhalotra, 2017) . Similarly, we do not view these improvements in the survival and nutrition of girls as the final step in reducing disparities. Many remain; the government needs to continue to address these, both in the supply of services and in changing incentives for education and later marriage (Sekher, 2012) . born for every 100 girls following the birth of a male first child and 118 if the first child was female. In 2016, these ratios were 108 and 110, respectively. Although these numbers raise questions well outside the current study, they provide support to the view that the changes in gender disparities in mortality and nutrition between survey rounds are not primarily a reflection of access to prenatal determination of the child's gender.

A limitation of this study stems from the fact that anthropometry can only be collected on survivors. If there are gender-specific trends in mortality between periods studied, this will have an effect on trends in anthropometry (Alderman et al., 2011; Harttgen et al., 2019) .

It is unlikely, however, that the different rates of improvements in survival appreciably mask or obscure the patterns in anthropometry discussed here. In particular, improved survival has a tendency to slightly increase stunting in the overall population; thus, the greater improvements in CMR and U5MR among girls would somewhat reduce any bias relative to the earlier period. Because the bias on HAZ from not observing heights of children who died prior to the survey is upwards, the decrease in mortality implies a small downward impact on observed changes in stature. In the absence of any bias in both periods, the measured improvements in HAZ for girls would have been slightly larger relative to boys. Similar expectations can be made regarding any small bias in stunting, which in this case reduces observed stunting due to loss of stunted children from the sample.

Likewise, the more rapid decline in IMR for girls attenuates rather than exaggerates the relative improvement in measured stunting for girls under 2 over the period studied. Mortality may also be slightly underestimated because many children have not lived through the risk period covered in the mortality statistic. That is, if a child is only 36 months at the time of the survey, we cannot tell if s/he will survive to 60 months. However, this should not have a strong effect on a comparison of relative disparity over time (Karlsson et al., 2019) .

India remains a challenging country to be a girl. At the same time, our research demonstrates that the basic biological disadvantages of sheer survival or of being undernourished can change and quite dramatically. The reasons behind the reduced sex differentials need more exploration, but our analyses are a step in identifying states where further investigation would be useful. It is more likely than not that a complex set of positive forces-improvements in health services, improvements in household conditions and changes in societal perceptions-came together to improve survival chances and physical growth for millions of girls in India. Staying the course to ensure that economic, social and health policies remain in place to support further improvements in health and nutrition will be important for India's girls;

this is especially important in the context of massive disruptions like the COVID-19 pandemic. Similar efforts to understand and close disparities in education can build upon such progress.

Actions for Nutrition in India (POSHAN), led by International Food Policy Research Institute.

The authors declare no conflicts of interest.

HA conceived the idea for the manuscript, supported data interpretation and wrote significant sections of the manuscript. PHN conceived the idea for the manuscript, conducted the statistical analysis and wrote significant sections of the manuscript. LMT conducted the statistical analysis and reviewed and edited the manuscript. PM supported data interpretation and reviewed and edited the manuscript. All authors read and approved the final submitted manuscript.

The data that support the findings of this study are available in the tables/figures and supporting information of this article. Additional data are available upon request.

Phuong Hong Nguyen https://orcid.org/0000-0003-3418-1674

Family resources and gender differences in human capital investments: The demand for children's medical care in Pakistan

Evidence of impact on growth and development of interventions during early and middle childhood

Tall claims: Mortality selection and the height of children in India

Progress in reducing child mortality and stunting in India: An application of the Lives Saved Tool

Population sex ratios and violence against women: The long-run effects of sex selection in India

Missing women: Age and disease. The Review of Economic Studies

On the quantity and quality of girls: Fertility, parental investments, and mortality. Policy Research Working Paper, 9390

Regional patterns of sex bias and excess female child mortality in India

The operational evidence base for delivering direct nutrition interventions in India: A desk review

Child gender and parental investments in India: Are boys and girls treated differently?

India's Integrated Child Development Services programme

No female disadvantage in anthropometric status among children in India: Analysis of the 1992-1993 and 2005-2006 Indian National Family Health Surveys

Fertility decline and increased manifestation of sex bias in India

Ministry of Health & Family Welfare. Five years (2009-2014) of National Health Mission. Achievement & new initiatives

Selective discrimination against female children in rural Punjab

Selective mortality and the anthropometric status of children in low-and middle-income countries

Prenatal sex determination and girls' wellbeing: Evidence from India

Unnatural selection: Choosing boys over girls, and the consequences of a world full of men. United States: Public Affairs

India report. NFHS-3 (National Family Health Survey-3), International Institute for Population Studies

Subnational mapping of under-5 and neonatal mortality trends in India: The Global Burden of Disease Study

The burden of child and maternal malnutrition and trends in its indicators in the states of India: The Global Burden of Disease Study

Girls' hidden penalty: Analysis of gender inequality in child mortality with data from 195 countries

Where there is a will fertility behavior and sex bias in large families

Why do mothers breastfeed girls less than boys? Evidence and implications for child health in India

Why are Indian children so short? The role of birth order and son preference

Trends in selective abortions of girls in India: Analysis of nationally representative birth histories from 1990 to 2005 and census data from

Socioeconomic and gender inequalities in neonatal, postneonatal and child mortality in India: A repeated cross-sectional study

Precision-weighted estimates of neonatal, post-neonatal and child mortality for 640 districts in India, National Family Health Survey

A turning point in gender bias in mortality? An update on the number of missing women. Population and Development Review

Explaining sex differentials in child mortality in India: Trends and determinants

Why are there delays in seeking treatment for childhood diarrhoea in India?

Causes of neonatal and child mortality in India: A nationally representative mortality survey

Sex differentials in childhood feeding, health care, and nutritional status in India. Population and Development Review

Proximate sources of population sex imbalance in India

Child malnutrition and gender preference in India: The role of culture

Guide to DHS statistics. Calvert, MD: Demographic and Health Surveys, ORC Macro

Ladlis and Lakshmis: Financial incentive schemes for the girl child

More than 100 million women are missing. The New York Review of Books

Birth order, fertility, and child height in India and Africa. IZA Discussion Papers, No. 12289

Undernutrition in Sub-Saharan Africa: Is there a gender bias?

Child nutrition in India in the nineties. Economic Development and Cultural Change

Worldwide timing of growth faltering: Revisiting implications for interventions

Understanding the landscape of national policies and strategic plans to tackle undernutrition in India: A review

INFORMATION Additional supporting information may be found online in the Supporting Information section at the end of this article

Trends and geographic variability in gender inequalities in child mortality and stunting in India