Height of the problem
DIANE COFFEY and DEAN SPEARS
CHILDREN in India are much shorter than children in other countries. More disturbingly, they come across as much too short according to international norms for populations of healthy children. Adivasi children are even shorter, on average, than other children in India. This gives rise to two questions. First, why are adivasi children so short? And second, why are they shorter than children in other population groups in India?
We will primarily discuss existing results in the literature, but will also illustrate these sometimes technical findings with simple analysis of data on child height. Our discussion of the second question – why are Scheduled Tribe (ST) children shorter than other children in India – draws heavily on research that we have done in collaboration with Ashwini Deshpande and Jeff Hammer.1 Our research shows that while the ST-general and Scheduled Caste (SC)-general child height gaps in India are almost identical, the ST-general gap can be completely accounted for by observable differences in wealth and material resources, but the SC-general gap cannot. This is because STs tend to live in different places than other groups, and are therefore exposed to different threats and resources. SCs, in contrast, are more likely to be mixed into the same villages as higher castes.2 These results suggest that the ST height gap is no special puzzle: ST children are shorter than general children primarily because they are poorer and live in more remote places with fewer resources.
But why are ST children and general children alike so short? All Indian children, including ST children, are much too short. They are ‘too short’ both in the sense that they are shorter, on average, than guidelines for healthy growth recommend, and in the sense that being too short matters for health and economic outcomes. Children in India are also much shorter than even children in other countries of similar or lower levels of economic development.
Population height is largely determined by early-life net nutrition. Here, ‘net’ means nutrition that is consumed, net of losses due to energy expenditure, malabsorption, parasites and disease.3 The relevant early-life period in a child’s life – from conception to two years of age – is sometimes characterized as the critical ‘first 1,000 days’. Physical height is not the only part of a child growing in early life: brains and bodies, skills and cognition are all developing and shaped by health and net nutrition. Children who do not experience the health and net nutrition in the first 1,000 days that allows them to grow to their genetic height potential are also unlikely to grow to their genetic cognitive potential.
Average Child Height by Population Group
This is one reason why height is important for adult economic outcomes: taller people are, on average, paid more because taller people have greater cognitive achievement, since the same early-life health that allowed them to grow towards their height potential also allowed them to grow towards their cognitive potential.4 In India, the height cognitive achievement gradient is even steeper than in developed countries, where it was first studied by economists, suggesting that profound deficits in early-life health and net nutrition are particularly important factors shaping the distribution of human capital in India.5
Because height is so important, the puzzle of child height in India has received much attention. It is an apparent paradox, called the ‘Asian Enigma’, that children in India are shorter on average than children in sub-Saharan Africa even though children in Africa are poorer, on average. In prior research, we have shown that the India-Africa gap in average child height can be completely statistically accounted for by the fact that Indian children are exposed to particularly poor sanitation: almost every Indian child lives near many people who defecate in the open, and because population density is high, this open defecation is especially threatening for child health.6 We will draw on this research and related papers in this discussion note: sanitation is one of many factors that importantly limit the growth of adivasi children.
Here, we study child height in India using the third round of the National Family Health Survey (NFHS). In particular, we use data on 39,864 children under five for whom height-for-age was measured. Although this is the most recent Demographic and Health Survey (DHS) in India, it is almost a decade old; India has not adequately invested in even knowing just how stunted its children are. Throughout this essay, we use the DHS categorization of children into four categories: ST (or Adivasi), SC (or Dalit), OBC (Other Backward Classes), and general.7 This means that we are ignoring religion, a critically important dimension of social distance in India. Of the 6,548 ST children in the DHS with height-for-age data, a plurality (3,023) are identified as Christian, and most others (2,769) are identified as Hindu. This method has the disadvantage of grouping together middle and high caste Hindus and Muslims into the ‘general’ category; this is awkward, but is unlikely to change our findings.
The two basic facts that we attempt to explain are presented in Figure 1. First, essentially all children in India are much too short. The vertical axis is average child height-for-age, or height relative to a healthy population. Negative numbers represent children who are too short, and all of these numbers are importantly negative. Second, ST children are substantially shorter than general children. A further observation, noted by Coffey, et al.,8 is that ST children have almost the exact same average height-for-age as SC children; however, these similar levels of deprivation seem to have very different explanations.
This comparison may appear unfair: 95% of ST children in our height sample live in rural places, while only 65% of general caste children do, and only 75% of all children in the sample. Because rural children are importantly shorter than urban children (by 0.41 height-for-age z-points), is there still an ST gap when we compare ST children with general caste children from rural areas? Yes, there is: rural ST children are still 0.38 z-points shorter than rural general caste children. Although rural residence accounts for 19% of the ST-general height gap, there is still 81% of the gap left to explain. Rural homes are only a small part of the gap.
More than a billion people worldwide defecate in the open without using a toilet or latrine. India, with some of the world’s worst stunting, also has one of the very highest rates of open defecation: more than half of the Indian population does not use any toilet or latrine, and most people worldwide who defecate in the open live in India. Worse still, high population density in India means that children are especially likely to be exposed to neighbours’ germs: the same amount of open defecation is more harmful to early-life health where population density is greater.9
Researchers have long recognized that disease is an important part of early life ‘net nutrition’, and therefore disease control has historically been an important part of improvements in height.10 In a recent study of the historical increase in European heights, Hatton11 found that improvements in height occurred when disease control (measured as infant mortality) improved. Evidence in the medical and epidemiological literature has documented that germs in faeces can stunt children’s growth. This is in part due to diarrhoea, parasite infections, energy spent fighting disease, and possibly in part due to enteropathy,12 which is a change in the lining of the intestines13 that may make it harder for the body to use nutrients. New observational evidence is consistent with the idea that enteropathy may lead to stunting.14 Econometric papers focusing on cause and effect have also shown a causal link from sanitation to child height.15
Therefore, Spears16 asked the quantitative, accounting question: are differences in child height associated with differences in exposure to open defecation big enough such that, given the differences in exposure to open defecation, sanitation could explain some international differences in child height? In particular, could it explain the Asian enigma? Using different statistical methods, one finds similarly sized effects of sanitation on child height. These effects are big enough that sanitation could account for the entire India-Africa height gap. This result suggests that if Indian children faced similar exposure (or lack of it) to open defecation as African children, Indian children would be about as tall as African children. That restatement makes clear that we are not claiming that open defecation is the only important threat to child height in India: ‘as tall as African children’ is still much too short for good health and human capital outcomes!17
What are the effects of sanitation on child height among adivasis? Faecal germs are impersonal, and there is every reason to expect the effect of open defecation on height to be comparable to what it is on other Indians. A greater fraction of the average adivasi child’s neighbours defecate in the open than other groups: 35 %age points more of the average ST child’s neighbours defecate in the open than the average general child’s neighbours; this disadvantage falls to only 20 percentage points if we only look at rural children. On the other hand, if ST children tend to live in low population density places – which the DHS data do not let one assess – this might decrease the harmfulness of a given difference in open defecation.18
Figure 2 plots the average height of ST and general children at each level of exposure to open defecation. The horizontal axis is the fraction of households surveyed in the DHS living near the child who report defecating in the open; for rural children, this can be thought of as the fraction of the child’s village that defecates in the open. Rural children are plotted separately so that we can see how much of the difference is due to the fact that STs are more likely to live in rural places.
Average Child Height by Local Open Defecation
Two conclusions are visible in the figure. First, the lines slope down for both groups: this means that among adivasi children and among general caste children, those exposed to more open defecation are shorter, on average. Exposure to open defecation is one of the reasons ST children are so short, relative to healthy norms. Second, the vertical distance between the lines suggests that, even at the same level of exposure to open defecation, and even looking only within rural India, ST children are still shorter than general children.19 Sanitation alone cannot account for the ST-general height gap.
Of course, this figure by itself is not enough to prove that open defecation has a causal effect on child height: places with more open defecation almost certainly have more of other health hazards too, on average. Nevertheless, we know that open defecation is indeed bad for child height, from the totality of a literature that uses multiple identification strategies – such as randomization, fixed effects, discontinuities, and instrumental variables – and that also impacts related outcomes such as haemoglobin levels,20 infant mortality,21 and cognitive achievement.22
To understand why two population groups differ in some outcome, economists use decomposition methods. These techniques decompose the average difference in outcome between two population groups into the part that can be explained by other observable differences between the two groups and the part that is still left unexplained after taking those differences into account.
Decomposition of ST-General Height Gap
A classic application is the U.S. wage gap between whites and blacks: blacks are paid less, on average, than whites in the U.S. labour market. How much of this difference in wages can be explained by the fact that blacks have, on average less education? Often, the part of the difference that is left over and cannot be explained by the differences in inputs is interpreted as an effect of discrimination, although in fact it could be an effect of any factor that is not accounted for in the decomposition. So, if blacks are still paid less even at the same level of education, skills, and experience, an economist might interpret this as evidence of discrimination.
We apply similar decomposition techniques to the height gap between ST and general children. This section follows Coffey, et al. (2014), in which we perform a similar but more detailed decomposition of these height differences. We will proceed step by step, first asking what fraction of the general-ST gap can be explained by the mere fact that STs are more likely to live in rural places; then adding the difference in exposure to open defecation, and finally adding a measure of relative wealth and poverty. For this analysis we will differentiate between richer and poorer children simply by using the division of the population into asset wealth quintiles that is included with DHS data. In Coffey et al. we use a more detailed accounting for wealth and poverty differences and find similar results.
Figure 3 graphs the decomposition results. Each bar after the first is the remaining ‘unexplained’ height gap after the listed variables are accounted for. As we have already argued, rural or urban location can account for very little of the height gap. While differences in sanitation can account for more of the height gap, there is still an important smaller gap left to be explained. The five wealth categories, however, can completely account for the ST-general height gap, even without taking rural location and sanitation into account.23 Therefore, in this statistical accounting sense, the fact that ST children are shorter, on average, than general children can be completely accounted for by the fact that they are poorer. Presumably, this is in part because they live in more remote locations with fewer resources, but this decomposition cannot assess this directly.
The decomposition result above suggests that ST children are shorter than general children in large part because they are poorer. Of course, this does not estimate any sort of impact of any sort of policy, and tells us little about what sort of improvements in the material environment of ST children might make them richer or healthier most effectively.
But the ST-general height gap is only 22% of the more important gap between ST children and the healthy reference population. Therefore, the most important question may not be what can be done to make ST children as tall as general children, but what can be done to make ST children grow to healthy heights. The graph above, along with a large and growing body of evidence, suggests that reducing the amount of open defecation to which ST children are exposed will help them grow taller. Improving sanitation is a particularly appropriate policy recommendation because sanitation is a public good and open defecation has negative externalities – meaning bad effects on other people. In public economics, such externalities are a classic situation requiring public action to achieve a good outcome.
Improving sanitation is an important step towards helping Indian children grow taller, but it will not be easy. Open defecation in India has remained stubbornly resistant to policy initiatives. New evidence suggests that the fraction of people in India who defecate in the open has declined so slowly that it has not kept pace with population growth. Spears shows that most people in India live in a district where their exposure to open defecation density increased between the 2001 and 2011 census rounds.24 Many people in India do not believe that open defecation is harmful, and some even prefer to defecate in the open. There is no clear, ready-made solution to the problem of sanitation behaviour change in rural north India. Widespread child stunting demands that we get busy experimenting with many different ways to address open defecation.
1. Diane Coffey, Ashwini Deshpande, Jeffrey Hammer and Dean Spears, Unpublished research on differences in child height across population groups in India, available on request. 2014.
2. Among SCs, social rank can account for height gap that remains after controlling for economic factors. Additionally, we find that after controlling for material resources, SC children are no shorter than general caste children when they live in villages where they are not outranked by higher caste people.
3. Angus Deaton, ‘Height, Health and Development’, Proceedings of the National Academy of Sciences 104(33), 2007, pp. 13232-237.
4. 4. Anne Case and Christina Paxson, ‘Stature and Status: Height, Ability, and Labor Market Outcomes’, Journal of Political Economy 116(3), 2008, pp. 499-532.
5. Dean Spears, ‘Height and Cognitive Achievement Among Indian Children’, Economics and Human Biology, 10(2), 2012, pp. 210-219.
6. Dean Spears, How Much International Variation in Child Height Can Sanitation Explain? Policy Research Working Paper 6351, World Bank, 2013.
7. Aficionados of the DHS: we are using question s118 in the birth recode. We use the birth recode throughout the paper, except when we merge in local open defecation computed from the household recode: the fraction of households reporting open defecation in a child’s PSU. Primarily because it is consistent with the language used in the DHS data, on which we rely, we will often follow the DHS terminology of ST and SC. This terminology emphasizes the origins of these categories in the Indian state. We find it awkward to refer to ‘non-SC, non-ST, and non-OBC’ children as ‘general’ children, but this appears to be a convention, and we know of no better alternative, so we shall do so.
8. Diane Coffey, Ashwini Deshpande, Jeffrey Hammer and Dean Spears, op. cit., fn. 1.
9. Payal Hathi, Sabrine Haque, Lovey Pant, Diane Coffey and Dean Spears, Do Toilets Spill Over? Population Density and the Effect of Sanitation on Early-Life Health. Working paper, RICE, 2014.
10. Carlos Bozzoli, Angus Deaton and Climent Quintana-Domeque, ‘Adult Height and Childhood Disease’, Demography 46(4), 2009, pp. 647-669.
11. Timothy Hatton, ‘How Have Europeans Grown so Tall?’ Oxford Economic Papers, 2013.
12. Jean H. Humphrey, ‘Child Undernutrition, Tropical Enteropathy, Toilets, and Hand-Washing’, Lancet 374, 2009, pp. 1032-1035.
13. Poonum S. Korpe and William A. Petri, Jr., ‘Environmental Enteropathy: Critical Implications of a Poorly Understood Condition’, Trends in Molecular Medicine 18(6), 2012, pp. 328-336.
14. Audrie Lin, Benjamin F. Arnold, Sadia Afreen, Rie Goto, Tarique Mohammad Nurul Huda, Rashidul Haque, Rubhana Raqib, Leanne Unicomb, Tahmeed Ahmed, John M. Colford Jr. and Stephen P. Luby, ‘Household Environmental Conditions are Associated with Enteropathy and Impaired Growth in Rural Bangladesh’, American Journal of Tropical Medicine and Hygiene, 2013; Margaret Kosek, et al. and the MAL-ED Network, ‘Fecal Markers of Intestinal Inflammation and Permeability Associated with the Subsequent Acquisition of Linear Growth Deficits in Infants’, American Journal of Tropical Medicine and Hygiene 88(2), 2013, pp. 390-396.
15. Dean Spears, Effects of Rural Sanitation on Infant Mortality and Human Capital: Evidence From a Local Governance Incentive in India. Working paper, Princeton, 2012.
16. Dean Spears, 2013, op. cit., fn. 6.
17. Indeed, we have other research joint with Reetika Khera exploring social rank within Indian rural joint households to show that women’s social status is another important constraint on child height in India (Diane Coffey, Reetika Khera and Dean Spears, ‘Women’s Status and Children’s Height in India: Evidence From Joint Rural Households’. Paper presented at Economic Demography Workshop, PAA, 2013).
18. Payal Hathi, et. al., 2014, op. cit., fn. 9.
19. Strikingly, sanitation does seem able to account for the difference in this figure between general children who live in rural and urban areas. A regression confirms this, general rural children are 0.354 height-for-age points shorter than general urban children, but this disappears (the point estimate becomes + 0.040, in the other direction) when the fraction of households in the child’s PSU who defecate in the open is linearly controlled for. Of course, this is merely suggestive.
20. Diane Coffey, Open Defecation and Hemoglobin Deficiency in Young Children. Paper presented at PAA, 2014.
21. Michael Geruso and Dean Spears, Sanitation and Health Externalities: Resolving the Muslim Mortality Paradox. Paper presented at PAA, 2014.
22. Dean Spears and Sneha Lamba, Effects of Early-Life Exposure to Rural Sanitation on Childhood Cognitive Skills: Evidence from India’s Total Sanitation Campaign. Paper presented at PAA, 2013.
23. See Coffey, et. al., (2014), op. cit., fn. 1, 8. For more details, and a more transparent accounting for asset wealth than this use of the DHS quintiles.
24. Dean Spears, Increasing Average Exposure to Open Defecation in India, 2001-2011. Working paper, RICE, 2014.