Hunger in India: Facts and challenges — 2  

  Vol II : issue 6

  Amartya Sen
  Peter Svedberg
  M.S. Swaminathan
  Swadesh Deepak
Jayanta Mahapatra
  A.K. Shiva Kumar

  Only in Print

Subscribe to The Little Magazine
Order the print edition of this issue
Browse our bookstore
Browse back issues

   Mail this page link
   Enter recipient's e-mail:

Peter Svedberg

Trends in Anthropometric Status of Indian Children

The three anthropometric indicators reported by the WHO and in the DHS are weight for age (W/A), height for age (H/A) and weight for height (W/H). In all the surveys, the shares of children who have a height and/or weight two and three standard deviations (sd) below the median in the reference population (from the US) are reported. In the following, the cut-off points to be applied are two standard deviations below (-2sd) the median for these three anthropometric indicators. We will also report results based on the share below three standard deviations (-3sd) of the H/A norm. The latter indicator is applied to capture "extreme" stunting. (That a child at the age of five is -3sd below the norm means that she/he is about 85-90 cm tall, while a "normal" child of this age is 110 cm, signifying a 20-25 cm shortfall.)


4The broad picture at face value

There are seven anthropometric surveys of young children in India reported in the latest WHO (2001a) Global Database on Child Growth and Malnutrition that are sufficiently large, detailed and comparable to merit consideration in the present context. The first such survey was undertaken during the years 1974-79 and the latest in 1998-99. The surveys cover between 3,000 and 46,000 children. All the seven surveys were originally carried out by the Indian National Nutrition Monitoring Bureau (NNMB). The original data have subsequently been re-tabulated by the WHO so as to increase comparability and ensure the use of uniform norms and ways of defining anthropometric "shortcomings".

The shares of the Indian children that fall below norms for the four anthropometric measures used here are depicted in Figure 1 for the seven surveys. At face value, it seems that the share of children with a W/H (-2sd) has remained more or less unaltered at slightly below 20 per cent since the 1970s, although it declined somewhat over the latter part of the 1990s. The shares of children with height and weight for age below the norms have declined more or less consistently since the late 1970s (the trends are highly statistically significant). By the indicator aimed at capturing extreme stunting (H/A -3sd), the decline is dramatic. In the 1970s, half the Indian children were below this norm; in the late 1990s, less than a quarter. The declines in modest stunting (H/A -2sd) and in underweight (W/A -2sd) are also quantitatively large, from about 70 to less than 50 per cent. A worrying message brought out by Figure 1, however, is that the shares of children below the three height- and weight-for-age norms have increased somewhat during the late 1990s.


The picture revealed by Figure 1 has to be qualified since the seven surveys are not strictly comparable. First, most results reported in Figure 1 are based on averages for surveys covering eight to 18 Indian states; only two are claimed to be nationally representative. Second, some surveys cover children in rural areas only, while others include both rural and urban children. Third, the surveys cover different age cohorts; most surveys include children between 0 and 5 years, while others include 0-3 or 0-4 year olds. These differences should be controlled as far as possible before one can say more definitely by how much the prevalence of anthropometric failure among children in India has declined. But some tentative controls undertaken have not revealed anything that fundamentally changes the initial conclusion that there have been notable declines in the share of children who are stunted and underweight since the second half of the 1970s up to the late 1990s (Svedberg 2001b). The share of children who are wasted has declined only during the 1990s, but this period is too short to judge whether this is the beginning of a new trend or not.

Inter-State Differentials

As is well known, there are large differences between and within the 25 Indian states in dimensions such as per-capita income, educational attainment, child mortality, longevity and the female/male population ratio (Drèze and Sen 1995; Sen 2001). There are also large differences in the anthropometric status of children across the Indian states.

4Inter-state variation in 1992-93

The survey that provides data for the largest number (20) of states is the one from 1992-93. This survey reveals marked differences in all the four anthropometric indicators across the 14 states for which sufficiently many observations are available (Figure 2). The highest state prevalence rate, as a ratio to the lowest, ranges from 2.2 (modest stunting) to 4.4 (extreme stunting) for the various indicators. The number on top of the "bars" at the lower and upper end of the inter-state distribution indicate that the observation for these particular states differs in a statistically significant way from the 20-state average at the 1, 5 or 10 per cent level.

The fact that for three of the anthropometric indicators, only two to four states have higher failure rates than the 20-state averages may seem puzzling. This follows, however, from the fact that the average is weighted by the number of observations in the various states, and that these numbers are by far the highest in Uttar Pradesh and Bihar, the two states with the highest incidence of anthropometric failure. This also explains why only one state (Bihar) has a prevalence of (extreme) stunting that is statistically significantly higher than the (weighted) average.

The surveys suggest a widening urban-rural gap in anthropometric failure rates in all four indicators. The rural disadvantage is the largest for the indicator of extreme stunting (H/A -3sd), and has been intensified the most between the two surveys

Unfortunately, the fourteen states represented in Figure 2 are not "randomly" distributed over the Indian continent. There is an under-representation of large states in the southeast (Andhra Pradesh, Tamil Nadu and West Bengal), while north-western India is well represented (although Madhya Pradesh is not covered). The hypothesis that there could be a north-west and south-east divide (see Sen 2001) can hence not be satisfactorily answered by the data from the 1992-93 survey.

4Inter-state variation over time

Three of the seven surveys permit the study of inter-state variation in the anthropometric status of children in India over time (the surveys from 1974-79, 1988-1990 and 1996-97). These surveys are comparable in the sense that they provide separate data for the same eight states (mainly southern), based on rural samples of 0-5 year olds. This allows us to investigate

i. to what extent there are time trends in these particular states,

ii. the extent to which inter-state differences have been stable over time and

iii. whether the inter-state variation has increased or not.

When it comes to the W/H (-2sd) indicator, there is little consistency in any of these dimensions. In a few states, there has been an increase, in others a decrease, and in still others a reversal. No single state has had the lowest or the highest score throughout the periods covered. The inter-state spread was high in the first and third period and less so in the middle period (1988-90).

The indicators for height and weight for age show more consistent developments over time. With one exception (Madhya Pradesh), there has been a steady decline in the share of children below the norms in all the states covered. Most of the states have not changed their position in the "ranking" notably (as indicated by crossing lines in Figure 3). By the H/A (-2sd) and W/A (-2sd) indicators, there have been notable increases in the inter-state differences in the anthropometric status of Indian children since the 1970s. In those days, over 70 per cent of the children observed had a weight and height for age below the standard norms (-2sd), and there were practically no differences across the eight states covered. Also, the indicator of extreme stunting (H/A -3sd) shows between 45 and 60 per cent of the children to be below norm in the 1970s (Figure 3).


A few cautious words on the limitations of the above analysis of inter-state differences in anthropometric failure rates are in order. The time analysis permitted by the data is restricted to eight of the 25 Indian states, and these eight states do not include any one from northern India. A more detailed comparison of states along the south-east and north-west divide a comparison that would be interesting in the perspective of notable differences along this "border" in other welfare indicators (Drèze and Sen 1995; Sen 2001) is hence not possible to conduct.

Urban vs Rural areas

Most of the seven surveys scrutinised above are restricted to rural areas. Only two surveys permit comparisons of anthropometric failure rates in urban and rural areas. These two surveys, from 1992-93 and 1998-99, are based on nationally representative samples and differ only in that the earlier one covers children aged 0-4 years, while the latter covers 0-3 year olds. The surveys suggest a widening urban-rural gap in anthropometric failure rates in all four indicators. The rural disadvantage is the largest for the indicator of extreme stunting (H/A -3sd), and has been intensified the most between the two surveys. In 1998-99, the incidence of extreme stunting in urban areas was only 61 per cent of that in rural areas, down from 71 per cent in the early 1990s.

p. 1 p. 2 p. 3 Sources

Peter Svedberg is Professor of Development Economics at the Institute for International Economic Studies, Stockholm University. He is the author of Poverty and Undernutrition: Theory, Measurement and Policy (Clarendon Press, Oxford, 2000, to be published by OUP India soon), a definitive study of undernutrition in Sub-Saharan Africa and South Asia. He lives in Stockholm