12129.pdf
Media
- extracted text
-
Nutrition in India: Facts and Interpretations
Angus Deaton
Research Program in Development Studies
Center for Health and Wellbeing
Princeton University
Jean Dreze
Department of Economics
Allahabad University
April 2008
We are grateful to Reetika Khera and Thu Vu for help in the preparation of this paper, and to Dr.
G.N.V. Brahmam for detailed clarifications relating to National Nutrition Monitoring Bureau
(NNMB) data. We would also like to thank Linda Adair, Harold Aiderman, David Atkin, Anne
Case, Noel Cameron, Himanshu, K.V. Radha Krishna, Anima Rani, Abhijit Sen, Rekha Sharma,
Priyanka Sinha, Veena Shatrugna and A. Vaidyanathan for helpful discussions.
Nutrition in India: facts and interpretations
Angus Deaton, Princeton University
Jean Dreze, Allahabad University
ABSTRACT
In spite of India’s rapid economic growth, there has been a sustained decline in per capita
calorie consumption during the last twenty-five years. While the decline has been largest
among better-off households, it has taken place throughout the range of household per
capita total expenditure. For both adults and children, anthropometric indicators of
nutritional status in India are among the worst in the world. While these indicators have
shown improvement over time, the rate of progress is slow relative to what might be
expected based on international and historical experience. This paper presents the basic
facts about growth, poverty and nutrition in India, it points to a number of puzzles, and it
sketches a preliminary story that is consistent with the evidence. The reduction in calorie
consumption cannot be attributed to declining real incomes, nor to any increase in the
relative price of food. Our leading hypothesis, on which much work remains to be done,
is that, as real incomes and wages have increased, leading to some nutritional improve
ment, there has been an offsetting reduction in calorie requirements due to declining
levels of physical activity and possibly also to various improvements in the health
environment. If correct, this analysis does not imply that Indians are currently adequately
nourished; nothing could be further from the truth. Calorie intake has serious limitations
as a nutritional intake; while calories are extremely important, there are too many sources
of variation in calorie requirements for standard, invariant, calorie-norms to be usefully
applied to large sections of the population. We conclude with a plea for better, and more
regular, monitoring of nutritional status in India.
CONTENTS
1. Introduction
2. Trends in calorie consumption and nutrition indicators
2.1. Calories, food, and expenditures
Food, calories and cereal calories
Expenditure, poverty, and distribution
Calorie deficiencies and reported hunger
On calorie Engel curves
Total calories and cereal calories
Calorie Engel curves for rich and poor
Price paid per calorie
Spatial patterns ofcalorie andfat consumption
Is the decline in calories real: other evidence?
2.2. Trends in nutritional status
Anthropometric measurement
Nutrition status ofIndian children
Recent trends in child nutrition
Adult weights and heights
3. Interpretations
3.1. The calorie decline
Food prices
Coarse cereal prices
Impoverishment?
Sen s argument
Are the calorie Engel curves misleading?
Changes in food habits and non-market entitlements
Declining needs for calories?
Engel Curves reexamined
3.2. Nutritional status
Farewell to calorie-based nutrition assessment?
Multiple deficiencies ofIndian diets
Nutrition status ofprivileged Indian children
4. Concluding remarks
Nutrition in India: Facts and Interpretations
1. Introduction
The Indian economy has recently grown at historically unprecedented rates and is now one of the
fastest-growing economies in the world. Real GDP per head grew at 3.95 percent a year from
1980 to 2005, and at 5.4 percent a year from 2000 to 2005. Measured at international prices, real
per capita income in India, which was two-thirds of Kenya’s in 1950, and about the same as
Nigeria’s, is now two and a half times as large as per capita income in both countries. Real per
capita consumption has also grown rapidly, at 2.2 percent a year in the 1980s, at 2.5 percent a
year in the 1990s, and at 3.9 percent a year from 2000 to 2005. Although the household survey
data show much slower rates of per capita consumption growth than do these national accounts
estimates, even these slower growth rates are associated with a substantial decrease in poverty
since the early 1980s, Deaton and Dreze (2002), Himanshu (2007). Yet, per capita calorie intake
is declining, as is the intake of many other nutrients; indeed fats are the only major nutrient
group whose per capita consumption is unambiguously increasing. Today, more than three
quarters of the population live in households whose per capita calorie consumption is less than
2,100 in urban areas and 2,400 in rural areas - numbers that are often cited as “minimum
requirements” in India.
A related concern is that anthropometric indicators of nutrition in India, for both adults and
children, are among the worst in the world. Furthermore, the improvement of these measures of
nutrition appears to be slow relative to what might be expected in the light of international
experience and of India’s recent high rates of economic growth. Indeed, according to the
National Family Health Survey, the proportion of underweight children remained virtually
unchanged between 1998-99 and 2005-06 (from 47 to 46 percent for the age group of 0-3 years).
1
Undemutrition levels in India remain higher even than for most countries of sub-Saharan Africa,
even though those countries are currently much poorer than India, have grown much more
slowly, and have much higher levels of infant and child mortality.
In this paper, we do not attempt to provide a complete and fully documented story of poverty,
nutrition and growth in India. In fact, we doubt that such an account is currently possible.
Instead, our aim is to present the most important facts, to point to a number of unresolved
puzzles, and to present an outline of a coherent story that is consistent with the facts. As far as
the decline in per capita calorie consumption is concerned, our leading hypothesis, on which
much work remains to be done, is that while real incomes and real wages have increased (leading
to some nutritional improvement), there has been an offsetting reduction in calorie requirements,
due to declining levels of physical activity and possibly also to various improvements in the
health environment. The net effect has been a slow reduction in per capita calorie consumption.
Whatever the explanation, there is historical evidence of related episodes in other countries, for
example in Britain from 1775 to 1850, where in spite of rising real wages, there was no apparent
increase in the real consumption of food, Clark et al (1995). Per capita calorie consumption also
appears to have declined in contemporary China in the 1980s and 1990s (a period of rapid
improvement in nutrition indicators such as height and weight), see Du, Lu, Zhai and Popkin
(2002).
One of our main points is that, just as there is no tight link between incomes and calorie
consumption, there is no tight link between the numbers of calories consumed and nutritional or
health status. Although the number of calories is important, so are other factors, such as a
balanced diet containing a reasonable proportion of fruits, vegetables, and fats, not just calories
from cereals, as are factors that affect the need for and retention of calories, such as activity
2
levels, clean water, sanitation, good hygiene practices, and vaccinations. Because of changes in
these other factors, the fact that people are increasingly choosing away from a diet that is heavy
in cereals does not imply that nutritional status will automatically get worse. Nor should a
reduction in calories associated with lower activity levels be taken to mean that Indians are
currently adequately nourished; nothing could be further from the truth.
We start by documenting the decline in per capita calorie consumption (Section 2.1), as well
as the state of malnutrition (Section 2.2). We then look at possible reasons for the reduction in
calories (Section 3.1), and try to tease out how it fits into the general picture of economic growth
and malnutrition in India (Section 3.2). Section 4 concludes.
We emphasize at the outset that our analysis covers the period up to 2006, so that we do not
discuss what has happened to calorie consumption or to nutritional status in the subsequent two
years, during which there has been a marked increase in the price of food, both in India and
around the world.
2. Trends in calorie consumption and nutrition indicators
2.1. Calories, food, and expenditures
Food, calories and cereal calories
Table 1 shows estimates of per capita consumption of calories, protein, and fats using data from
the National Sample Surveys (NSS). Throughout the paper we show data from the “thick”
rounds from 1983 (38th), 1987-88 (43rd), 1993-94 (50th), 1999-2000 (55th), 2004-05 (61st)
supplemented by data from the most recent “thin” rounds collected in 2000-01 (56th), 2001-02
(57th), second half of 2002 (58th), 2003 (59th), and first half of 2004 (60th.) The per capita
consumption of calories and of protein is falling in rural India, and shows no trend in urban
3
India; this is occurring against the increase in real household per capita expenditures shown in
the first two columns of Table 2. In rural India, household per capita calorie consumption was
2,240 calories in 1983, 2,233 in 1987-88, and had fallen to 2,047 calories per head in 2004-05, a
decline of 8.6 percent from 1983; urban per capita calorie consumption was only 49 calories (2.4
percent) lower than in 1983. Over the same period, rural (urban) per capita protein consumption
fell by 12.1 percent (4.6 percent). Only per capita fat consumption, in the last two columns,
moves in the same direction as household expenditures, growing at 1.2 percent and 1.1 percent a
year in rural and urban India respectively.
Table 1: Mean per capita consumption of calories, protein, and fats.
Protein (gms)
Calories (kc)
Year
Round
Rural
Rural
Urban
Urban
Fats (gms)
Rural
Urban
37.1
27.1
58.1
63.5
39.3
28.3
58.6
63.2
41.9
31.1
57.7
60.3
49.6
36.0
58.4
59.1
46.1
34.6
55.3
56.8
46.1
33.6
54.2
54.8
47.0
34.7
54.9
55.4
46.7
36.4
55.5
58.0
46.8
35.5
55.9
56.9
47.4
35.4
55.4
55.8________
Source: Authors’ calculations from NSS data. Nutrients are calculated, following the NSS practice, by multiplying
reported quantities (purchased, grown, or received outside the market) by a set of nutrient conversion factors which
are themselves revised from time to time.
2,240
38
1983
2,233
43
1987-8
2,153
50
1993-4
2,148
55
1999- 0
2,083
56
2000- 1
2,018
57
2001- 2
2,025
58
2002(2)
2,106
59
2003
2,087
60
2004(1)
_______________________
2,021
2,047
61
2004-5
2,070
2,095
2,073
2,155
2,027
1,982
2,014
2,020
2,036
The statistics in Table 1 show calories per head, and take no account of changes in household
composition since 1983. In particular, fertility has fallen so that Indian households had a lower
proportion of children at the end of the period. The fraction of children (ages 0 to 14) in a rural
household fell from an average of 0.36 in 1983 to 0.31 in 2004-05, a decline of 14 percent, with
a larger decline, from 0.31 to 0.24, in urban India. In consequence, if we were to compute a
measure of equivalent adults, in which children counted as less than one, we would find that the
4
number of equivalents has been growing faster than the number of people, so that the decline in
calorie per equivalent would be even larger than the decline in calorie per person. Against this,
we should note that the reduction in fertility means that a shorter span of women’s lives is spent
in pregnancy or lactation, during which there is a sharp increase in recommended calorie needs.
But it is not clear that most women in India obtain these extra calories, and we suspect that the
saving from this source is small. In what follows, we mostly stick with the per capita measures
recognizing that they almost certainly provide conservative estimates of a more appropriate
measure of the fall in calories.
Table 2: Real mean per capita expenditure on all goods and on food
Year
Round
PCE at 50th Round prices Per capita food at 50th
Round prices_______
Rural
Urban
Urban
Rural
50th Round Rupees per
1000 calories_______
Rural
Urban
117.7
74.7
237.3
167.5
400.8
251.3
38
115.3
79.5
244.8
182.5
425.4
278.1
43
122.0
82.9
247.7
181.5
458.3
281.4
50
124.2
85.6
252.5
184.3
521.4
306.0
55
125.7
87.1
244.3
186.4
540.6
316.8
56
120.0
87.4
237.6
181.8
526.3
311.7
57
119.6
89.7
246.4
186.2
551.4
322.8
58
120.9
88.4
242.2
185.6
545.2
328.5
59
119.8
88.1
244.0
188.0
548.5
331.6
60
121.3
89.1
243.7
184.1
530.4
______________________________
318.3
61
____
j
’
calculations
from
NSS
data.
PCS
is
household
total
expenditure
per
capita
deflated
by
the CPIAL
Source: Authors
or CPIIW scaled to be unity in the 50th Round. Per capita food is mean per capita expenditure on food deflated by
.
.■<
x-.-r»T i t
___ i_J
the^food
components
of the
CPIAL
andi CPIIW, scaled
to be unity «««
m the 50th DrMinrl
Round. Rupees npr
per n.nlnriA.S
calories is
calculated by^dividing, for each household, per capita food expenditure by per capita calories, averaging over
households, and then deflating by the food components of the CPIAL and CPIIW.
1983
1987-8
1993-4
1999- 0
2000- 1
2001- 2
2002(2)
2003
2004(1)
2004-5
Table 2 shows that, in contrast to per capita total expenditure, and more in line with per
capita calorie consumption, there has been no real increase in per capita food expenditure,
particularly after 1987-88. The estimates of real food expenditure are constructed by dividing
food expenditure per capita by a price index of food; here the food components of the CPIAL
and CPIIW. We have also tried the survey-based food price indexes calculated in Deaton (2008);
5
these show substantially more food price inflation in the last few surveys and thus even lower per
capita food expenditure in 2004-05. There is also uncertainty regarding a possible overstatement
of food expenditures in 1999-2000 associated with the unique questionnaire design for that
survey, see Deaton and Kozel (2005a) for a summary of the literature. Even so, the real value of
per capita food expenditure in 2004-05 is essentially unchanged from its level in the late 1980s.
Any difference between the trend in real food expenditure and the trend in calories is a
consequence of a switch in the composition of foods towards foods with lower or higher calories
per kilo. That this effect has been so modest is itself something of a puzzle; we would expect
rising real incomes to generate a switch, for example, from cereals to fats and sugar, from cheap
calories to more expensive ones. The last columns of Table 2 show, in constant 1993—94 prices,
how much the average rural and urban household paid for each 1,000 calories consumed (these
figures should not be confused with calorie price indexes - these are discussed further on). If
people bought the same pattern of goods over time, this would remain constant, but it will
increase if people move away from goods that provide cheap calories
such as cereals towards
goods that provide more expensive calories—such as edible oils, milk products or meat. These
numbers show a modest increase in real price per calorie in the rural sector, consistent with the
fall in calories and flat real food expenditure, but essentially no change in the urban sector,
certainly since 1993-94. We shall return to these numbers below.
Table 3 shows that the consumption of calories from cereals has fallen even more rapidly
than have calories from all foods as shown in Table 1. From 1983 to 2004-05, rural per capita
calories from cereals fell by 295, some of which was offset by increases in calories from other
foods, so that the reduction in total calories was 193 (Table 1). Per capita cereal calories also fell
6
in urban areas, by 156 calories, essentially all of which was offset by an increase in calories from
other sources.
Table 3: Calories from cereal
Urban
Rural
Year
1,303
1983
1,681
1,296
1,648
1987-8
1,231
1,533
1993-4
1,200
1,455
1999- 0
1,161
1,422
2000- 1
1,130
1,391
2001- 2
1,137
1,381
2002(2)
1,142
1,412
2003
1,165
1,419
2004(1)
1,147
1,386
2004-5
Source: Authors’ calculations from NSS data.
All India
1,597
1,569
1,458
1,392
1,357
1,330
1,318
1,345
1,357
1,326
The decline in per capita calorie consumption has been previously noted in the Indian
literature including, without any attempt to be exhaustive, National Sample Survey Organization
(2001), Rao (2000, 2005), Meenakshi and Viswanathan (2005), Ray and Lancaster (2005),
Palmer-Jones and Sen (2001), Patnaik (2004, 2007), Radhakrishna, Rao, Ravi, and Reddy
(2004), Radhakrishna (2005), Sharma (2006) and Kumar, Mruthyunjava, and Dey (2007).
Patnaik’s two papers have been particularly influential in drawing wide attention to the decline in
this aspect of nutrition, particularly the fall in per capita calories from cereals. Sharma (2006)
shows, based on a comparison of NSS data for 1983 and 1999-2000, that the decline of food
intake is not confined to calories or proteins, but also applies to many other nutrients (fat being
the main exception, as mentioned earlier).
Expenditure, poverty, and distribution
Tables 1 and 2 show that the decline in per capita calorie consumption has been accompanied by
an increase in real average household per capita expenditure. While not all groups have shared in
7
this equally, Table 4 shows that there has been some growth of per capita expenditure at all
percentiles of the expenditure distribution. The numbers in the table are computed by calculating
the percentiles of per capita expenditure in the 38th, 50th, and 61st rounds (whose measures of
expenditure are comparable) and then calculating the annual growth rates from one period to the
next. Growth for these households was somewhat faster up to 1993-94 than in the eleven years
afterwards. Growth was also somewhat more equitable in the earlier period, with the lower
percentiles growing more rapidly than the higher percentiles (rural) or at about the same rate
(urban) while in the later period, after 1993-94, growth rates were higher in the higher
percentiles, especially in the urban sector. The differential growth rates between the top and the
bottom of the distribution and between urban and rural indicate an increase in inequality after
1993-94, something that is explored in more detail in Deaton and Dreze (2002).
Table 4: Growth of real per capita expenditure, 83 to 93-94 and from 93-94 to 04-05 (by
percentiles of per capita expenditure)
Percentile
tn
tii
75th
90
1.2
0.8
1.0
0.9
0.8
1.0
0.6
1.2
1.0
1.3
1.2
1.2
1.4
1.4
0.8
1.2
0.8
1.1
1.0
1.1
1.3
10th
25th
50
1983-1993/4
1993/4 -2004/5
1.7
0.9
1.4
0.8
Urban
1983-1993/4
1993/4-2004/5
1.1
0.4
1.1
1.7
1.7
0.9
0.9
Rural
All India
1983-1993/4
1993/4-2004/5
Source’. Authors’ calculations from NSS data
Even so, the growth in average per capita expenditures does not conceal any mixed pattern of
growth and decline, or of the rich getting richer while the poor got poorer. Note also that the
large difference in growth rates between average per capita consumption in the national accounts
and average consumption in the bottom decile (say) cannot be (entirely) attributed to increasing
8
inequality because there is a similarly large difference in growth rates across the NSS
distribution. The NSS consumption data are not consistent with the national accounts
consumption data, almost certainly due to errors in both; the NSS is almost certainly missing
some of the growth, and the NAS is almost certainly overstating it. One reason—although
certainly not the only reason—that other things are not improving at the rate one might expect in
such a rapidly growing economy is that the economy is not in fact growing quite as rapidly as the
data show.
There has, nevertheless, been a steady decline in poverty rates over the period. While there
has been much controversy over measurements associated with the NSS data from 1999-2000,
summarized in Deaton and Kozel (2005b), the 2004—05 survey was collected on the same basis
as the 1993-94 and earlier surveys, so that there is wide acceptance of the comparability of the
corresponding poverty estimates. These suggest that the rural (urban) headcount ratios declined
from 46.percent (41 percent) in 1983 to 28 percent (26 percent) in 2004-05. There is no evidence
in the NSS data of widespread impoverishment in India, although the decline in poverty is a
good deal less than might be expected given the rapid growth of per capita GDP.
Calorie deficiencies and self-reported hunger
When an increase in per capita expenditure is accompanied by a decrease in per calorie
consumption, there will be a divergence between poverty measures that are based on real
expenditures and those that are based on a calorie cutoff. Indian poverty lines were originally
justified by reference to calorie norms, and at the time that the original lines were drawn, average
per capita calorie consumption of rural households at the rural poverty line was 2,400 calories,
and average per capita calorie consumption of urban households at the urban poverty line was
9
2,100 calories. But all subsequent official poverty calculations have held the lines fixed in real
terms, so that the official poverty ratios have declined as the distribution of per capita
expenditure has moved upwards. As has been suggested by several authors, including PalmerJones and Sen (2001) and Ray and Lancaster (2005), we could take the calorie basis of the
original lines more seriously and compute the fraction of the population living in households
whose per capita calorie consumption falls beneath 2,400 calories in the rural sector and beneath
2,100 calories in the urban sector. Such calculations are shown in Table 5. Because the
distribution of per capita calories is moving to the left over time, these numbers show rising
poverty rates, from two thirds of the rural population in 1983 to four-fifths in 2004-05, and from
65 percent to more than 75 percent in India as a whole. Without understanding why per capita
calories are falling despite rising per capita expenditures, it is difficult to adjudicate between
these “calorie” poverty rates and the conventional “expenditure” poverty rates.
Table 5: Fractions of the population living in households with per capita calorie
consumption below 2,100 urban and 2,400 rural
Year
Round
Rural
66.1"
65.9
71.1
74.2
79.8
38
1983
43
1987-8
50
1993-4
55
1999-0
61
2004-5
Source'. Authors’ calculations based on NSS data.
Urban
All India
60.5
57.1
58.1
58.2
641
63.9
63.9
67.8
70.1
75.8
The fact that calorie shortfalls (based on fixed calorie norms) are not automatically
associated with self-reported hunger is shown by the evidence on the latter reported in Table 6.
Except for the 1987-88 round, the NSS consumption surveys have included a question on food
adequacy. In 1983 and 1993-94, respondents were asked whether everyone in the household got
“two square meals a day”, while in 1999-2000 and 2004-05 the question was whether everyone
in the household got “enough food every day”. Table 6 shows the fractions of people who did not
10
Table 6: Percentages of rural households reporting lack of food: India and major states
J & Kashmir
Himachal
Punjab
Uttaranchal
Haryana
Rajasthan
Uttar Pradesh
Bihar
Assam
West Bengal
Jharkhand
Orissa
Chhattisgarh
Madhya Pradesh
Gujarat
Maharashtra
Andhra Pradesh
Karnataka
Kerala
Tamil Nadu
1983
1993-94
1999-2000
2004-05
1.8
3.9
2.2
0.0
0.3
0.1
0.6
0.2
1.0
1.8
4.2
10.7
34.0
14.9
36.5
0.6
1.5
3.3
6.8
9.9
14.3
I. 7
0.2
2.4
6.3
7.2
II. 3
35.1
14.2
7.8
14.1
3.1
2.6
2.2
4.4
3.2
3.9
9.4
2.8
3.4
0.5
2.1
2.2
1.1
3.2
1.2
0.0
0.0
0.6
0.5
0.0
0.0
I.9
3.2
5.5
II. 7
0.6
5.9
2.5
2.1
0.3
1.1
0.7
0.3
2.5
0.1
13.0
14.4
17.3
17.5
17.2
_________ 2.5_____________
3.6
5.2
17.3
Total
_
___
state
did
not
exist
or
there
were
no rural households in the
Note’. Questions not asked in the 43r Round. — means
th and 50* th round is “Do all members of your household get two square meals a day?”
sample. The question in the 38
--------- ------------------- .
~
with possible responses 1 (yes, throughout the year) 2 (in some months) 3 (no). In the 55 and 61 round the words
“two square meals a day” are replaced by “enough food every day.” The table shows the fractions of people living in
households where the answer was other than 1. Note that samples are often small in the smaller states.
have adequate food (in that sense) throughout the year There are several reasons to treat these
numbers as suggestive rather than definitive*, the phrasing of the question is not identical in
different years, there may be translation issues, and the changes from 1983 to 1993-94 are
suspiciously large in several states such as Bihar and Madhya Pradesh. In so far as they are
reliable, these figures show that the fraction of rural persons going hungry has fallen from 17.3
percent in 1983 to 2.5 percent in 2004-05. In the latest survey, only West Bengal, Orissa, Assam,
and Bihar are above the national average, with Chhattisgarh and Kerala at the average. The
relatively high prevalence of self-reported hunger in Kerala is somewhat puzzling and raises
11
further questions about the interpretation of these figures. What is potentially interesting is that
the four states with high levels of hunger are all located in the rice belt of eastern India.
It is also worth noting that self-reported food adequacy is uncorrelated with calorie shortfall
(i.e. whether or not a household is below the calorie norms). The correlation across rural
households in 2004-05 is 0.02, and across NSS regions and states, it is actually negative, -0.17
for regions, and -0.09 for states. This lack of correlation might be taken as casting further doubt
on the validity of the hunger questions, or on the relevance of the calorie norms, or even both.
However, it should be noted that hunger and calorie shortfall are not the same thing, if only
because the latter, as measured here, ignores differences in calorie needs, for example those
associated with variations in activity levels
On calorie Engel curves
That calorie consumption should fall over time alongside an increase in total expenditure is in
sharp contrast to what happens when we look across households at a moment of time, where
there is a strong positive correlation between the two. This conflict comes from the fact that, over
time, there is a downward drift of the “calorie Engel curve” that plots calorie consumption
against per capita household expenditure: calorie consumption at a given level of per-capita
expenditure has steadily declined over the last 20 years. For each of the large rounds from 1983
to 2004-05, Figure 1 plots per capita calories against per capita total household expenditure
(both in log terms). The top five curves are for rural India, and are moving down over time. As is
to be expected from Table 1, the curve for 1987-88 is close to that for 1983. By 1993-94, the
downward shift has begun in earnest, but is much larger for better-off than for worse-off
households. By the end of the period, the proportional shifts are about the same at all points in
12
the distribution. The bottom of the picture shows the urban curves—at the same level of real per
capita expenditure, urban households consume fewer calories per head—and, apart from the
curve for the 38th round in 1983 which cuts across the others—there is again a steady fall in the
curves over time. Note that the rural calorie Engel curve in 2004-05 is close to the urban curve
in 1987-88, a point to which we shall return.
00 “
1983 rural.
1987-8 urban
1987-8 rural.
1993-4 rural.
o
1999-00 rural^yZ^
O
2004-05 rural^y//^
•6
J<QW
1993-4 urban
u
CL.
bJQ
1999-00 urban
2004-05 urban
™-
1983 urban
T
4.5
5
5.5
6
6.5
Logarithm of household per capita expenditure
7
Figure 1: Calorie Engel curves, rural and urban India, 1983 to 2004-05
Because the NSS has changed its questionnaires over this period, the graphs in Figure 1 are
subject to numerous qualifications, but we do not believe they are seriously misleading. They
show that in both urban and rural sectors better-off households (at least measured by per capita
expenditure) consistently consume more calories than worse-off households, at least on average.
The fall in calories in Table 1 comes about because the calorie Engel curves are shifting down
over time. Why they should do so in a country as poor and malnourished as is India will be one
of our main concerns in Section 3. For the moment, note that because the calorie Engel curves do
13
not cross (except for the 1983 urban curve), per capita consumption of calories has fallen at all
levels of per capita household expenditure. This only sharpens the puzzle; it is perhaps not too
hard to understand why people at the top of the expenditure distribution should consume less
over timfy but why those with the lowest per capita expenditures should do so is a real puzzle.
Total calories and cereal calories
The contrast between cross-section and time-series is also apparent when we look at calories
from cereals in Figure 2. These curves have lower slopes than do the curves in Figure 1 because
-
1983 rural
1987-8 rural
w
1993-4 rural
u
1999-00 rural
!
2004-05 rural
C/5 1
g
-a
1983 urban
mgs^batr^7-8 urban
’&
u
2004-05 urban
60
1999-00 urban
«. co H4.5
5
5.5
6
6.5
Logarithm of household per capita expenditure
7
Figure 2: Cereal calorie Engel curves, rural and urban India, 1983 to 2004-05
the total expenditure elasticity of cereal calories is lower than the total expenditure elasticity of
all calories; at the margin, a higher fraction of each rupee is spent on non-cereal calories than on
cereal calories. As was the case for total calories, the curves are falling through time within both
rural and urban sectors (again with a partial exception for the 1983 urban curve), and as they fall
14
they become flatter, so that the elasticity of cereal calories diminishes over time. In the lowest
curve, for the urban sector in 2004-05, the elasticity is negative except at the very bottom of the
per capita expenditure distribution. As is the case for total calories, the most recent rural curves
are similar in position and in shape to the earliest urban curves; in other words, rural India today
looks much like urban India 20 years ago as far as cereal consumption patterns are concerned.
Calorie Engel curves for rich and poor
Although all of the curves in Figures 1 and 2 are moving down over time, there are differences in
the size of the shift at different points in the per capita expenditure distribution, as well as
between cereal calories and all calories. In rural India, the total calorie curves shift down in a
roughly parallel way, so that the percentage decline in total calories is more or less the same for
the rich as for the poor. For cereal calories, the decline is much sharper among the better-off.
These comparisons hold per capita household expenditure fixed, so we are defining “better-off’
and “poor” in terms of fixed levels of real per capita expenditure. But it is also useful to arrange
rich and poor, not by expenditure, but according to their position in the distribution. Table 7
shows, for rural India, average per capita calorie and cereal calorie consumption in the bottom
decile, and for each of the four quartiles of the distribution. Because there is an increase in per
capita expenditure over time, people in each decile or quartile have rising real expenditure levels,
so that the downward movement of the Engel curves in Figures 1 and 2 is offset by movement up
the curve. And indeed, for the bottom decile, per capita calorie consumption has not fallen over
the period. For the quartiles, even the bottom quartile, the trend is either confused or declining,
so the downward shift of the Engel curves is the dominant effect, except for the bottom of the
distribution.
15
Table 7: Total and cereal calorie consumption by decile and quartile of per capita
expenditure, rural India, 1983 to 2004-05.
Second
Quartile
Third Quartile Top Quartile
1,580
1,683
1,659
1,658
1,624
2,007
2,056
2,000
1,978
1,900
2,328
2,334
2,251
2,250
2,143
3.044
2,863
2,702
2,707
2,521
Cereal
Calories
1,309
1,150
1983
1,359
1,221
1987-88
1,316
1,203
1993-94
1,289
1,197
1999-00
______________
1,259
1,189
2004-05
Source-. Authors’ calculations from NSS data.
1,589
1,598
1,504
1,591
1,690
1,738
1,715
1,591
1,509
1,430
1,974
1,894
1,690
1,566
1,471
Bottom decile Bottom
Quartile
Total calories
1983
1987-88
1993-94
1999-00
2004-05
1,359
1,488
1,490
1,496
1,485
Price paid per calorie
Richer people allocate their food expenditures differently from poorer people, switching from
cereals to fattier and sweeter foods, such as edible oils, meat, and sugars. The result is that
households that spend more in total also spend more per calorie, which is illustrated in Figure 3
for urban and rural households in 1983, 1993-94, and 2004-05. These curves plot the logarithm
of the rupees spent per calorie divided by a general index of food prices. We can therefore
imagine the changes from one period to another taking place with the price of food held constant
so that movements of the curves, like movements along the curves, come from switches from
cheaper to more expensive calories, or vice versa, not from changes in the price of food. Given
that the calories-from-cereal Engel curves in Figure 2 are moving down over time more rapidly
than the total-calories Engel curves in Figure 1, we can expect the curves in Figure 3 to move up
over time because, at any given level of total expenditure, households are reducing the relatively
16
cheap cereal calories as a share of total calories. This is exactly what happens for rural
households in Figure 3. Surprisingly, the opposite happens for urban households, in spite of the
fact that urban households are also cutting down on cereal calories (Figure 2). Indeed the urban
curves appear to be converging downwards towards the rural curves, though it should be noted
that the overall food price indexes only account for differential inflation over time, and not for
any original difference in the levels of urban and rural prices. Given that the urban curves slope
up, so that better-off households pay more per calorie, and given that living standards are rising,
these curves must fall over time to be consistent with the lack of any increase over time in the
average price paid per calorie shown in Table 2. But that mechanical result does nothing to
explain this very surprising behavior among urban consumers.
«-
1993-94
| in -
2004-05
8
8.
URBAN
1983
w
<u
1993-94
CD
E m
s
RURAL
1983
4
5
7
6
8
Logarithm of household per capita expenditure
Figure 3: Cost per 1000 calories at constant prices, urban and rural India
17
Spatial patterns of calorie andfat consumption
2004-05
1983
a.
W
L*8«f>d
c;:i tsw.swi-
sssas&z'M sw
BBB
.1
■y;
%
Figure 4: Rural per capita calorie consumption, 1983 to 2004-5
2004-05
1983
fe
M.
I
TF
W ;>■
Legend
p;T;
■ zt.
asm 4&»S6
MB >®<>
"V
%
Figure 5: Urban per capita fat consumption, 1983 to 2004-5
18
'i
Figures 4 and 5 show the regional and temporal variation in per capita calorie consumption in the
rural sector and regional and temporal variation in per capita fat consumption in the urban sector,
comparing 1983 and 2004-05. Each area of the map is a region as defined by the NSS, an area
that is a collection of districts but, in most cases, smaller than a state. The urban calorie map
shows no clear pattern over time, as is to be expected from Table 1, and the rural fat maps are
somewhat muted versions of the urban fat maps (these are not shown for reasons of space).
Calorie consumption is highest in the northwest of India, and lowest in the south, and the
successive maps show the calorie collapse with the high consumption zone retreating to the far
north by 2004-5. The fat maps for the urban sector show a different pattern with fat consumption
higher in the west and lower in the east, with a wave of fat washing in from the west. For
proteins, which we do not show, the pattern is very similar to the pattern for calories, again
echoing Table 1.
Is the decline in calories real: other evidence?
It is puzzling that a country as poor and malnourished as India should react to growing prosperity
without increasing real food consumption and by actually cutting back on its calorie
consumption, so it is important to check that the facts are indeed correct. Table 8 shows data on
calorie consumption from an independent source, the National Nutritional Monitoring Bureau
(NNMB) which periodically collects data on nutrition in the rural areas of a number of states.
Comparison with these data is also useful because the NNMB measures calories by direct recall
over a short reference period rather than by measuring quantities of foods over a longer period
and then imputing calories using tables for the calorie contents of foods. Table 8 shows values of
calorie and protein consumption that are lower than those in Table 1, most of which is explained
19
by the fact that the NNMB covers mostly southern states where calorie consumption is relatively
low; indeed, the averages from the NSS from the comparable states are quite close, see the
footnote to the table. More importantly, we see the same declining trends as in Table 1.
Table 8: Calorie and Protein consumption from National Nutritional Monitoring Bureau
Rural per capita consumption (nine states)
1975-79
1988-90
1996-97
2000-01
2004-05
Energy
2,340
2,283
2,108
1,954
1,907
Protein
62.9
58.4
53.7
50.7
48.8
Note'. Andhra Pradesh, Gujarat, Karnataka, Kerala, Madhya Pradesh, Maharashtra, Orissa, Tamil Nadu, and West
Bengal. 1988-90 and 1996-97 estimates exclude Madhya Pradesh and West Bengal. The 2004-05 figures exclude
Gujarat. NSS estimates for the comparable states in nearby years are as follows: calories, 2,131 (1983), 2,139
(1987-88), 2,076 (1993-94), 2,020 (1999-2000), 1,960 (2004-05); protein, 57.5 (1983), 57.0 (1987-88), 54.7
(1993-94), 52.7 (1999-2000), and 50.8 (2004-05).
Source'. Various NNMB reports and authors’ calculations from NSS data.
Another cross-check, at least for cereal intake, comes from aggregate availability figures
compiled from production data with the addition of net imports and the subtraction of addition to
stocks—in practice mostly government stocks. Figure 6 shows the data from the 2006-7
Economic Survey, for cereals as well as for cereals plus pulses. The figure shows that there has
been little net change in aggregate cereal (or cereal plus pulses) availability in India since
independence and that there has been a clear downward trend since the early 1990s, up to
fluctuations that are largely attributable to the build up and release of government stocks. Figure
6 also shows the estimates of aggregate cereal consumption from the large rounds of the NSS
and, as has typically been found in the literature, at least since Minhas (1988), there is reasonably
close agreement between the NSS estimates and the data on availability, something that is in
sharp contrast to the discrepancies between the NSS and National Accounts data on total
consumer expenditure. These aggregate data are therefore consistent with the per capita calorie
20
decline that we see in both NSS and NNMB data. There can therefore be very little doubt that the
decline is real. And far from offsetting the decline in cereals, consumption of pulses has also
declined.
Cereal + pulses availability
o
O -
in
\
A
„
X-y
(A
%
o
4
O -
Cereal availability
NSS consumption
TD
(D
Q.
CLCXJ
8
0)
a2w
E
Changes in government stocks of cereal
O)
o -
1950
1960
1970
1990
1980
year
2000
From Economic Survey of India, 2006-07
Figure 6: Availability of cereal, 1950 to 2004
Figure 7 shows the disaggregation of cereal availability into three main groups: rice, wheat,
and other cereals. The last group is mainly the “coarse” cereal group consisting of maize, barley,
jowar (sorghum), ragi (finger millet) and bajra (pearl millet). The most notable feature here is
the steady long-term decline of the coarse cereals whose consumption has fallen by a half over
the last half century; although there have been temporary increases, the fall in coarse cereals is
long established. For many years, the place of coarse cereals was taken by an increase in
consumption of wheat. However, the long-run increase in wheat consumption appeared to come
to an end in the late 1990s. Rice consumption began to trend downward somewhat earlier, from
about the early 1990s so that, in the last few years, there has been no increase in wheat
consumption to offset the decline in the other two categories.
21
°
J
ID “
AVAILABILITY OF CEREALS, GM PER PERSON PER DAY
CXI
RICE
o
o “
CXI
o
LO -
OTHER CEREALS
WHEAT
o
O -
o
ID I ___
1950
1960
1970
1990
1980
year
2000
Ministry of Agriculture
Figure 7: Availability of rice, wheat, and other cereals
2.2. Trends in nutritional status
In the preceding section we have scrutinized the recent decline in per capita calorie
consumption—one of the many variables that influence nutritional status. We now turn to the
evidence on nutritional status itself, particularly the anthropometric evidence. Our main focus is
on child nutrition.
Anthropometric Measurement
For children, anthropometric indicators are typically based on age, height and weight. Three
standard indicators are “height for age”, “weight for height”, and “weight for age”. Low heightfor-age is often referred to as “stunting”, low weight-for-height as “wasting”, and low weightfor-age as “underweight”.
22
Stunting is a cumulative indicator of nutritional deprivation from birth (or rather, conception)
onwards. It is relatively independent of immediate circumstances, since height does not change
much in the short term. Wasting, by contrast, is usually taken to be an indicator of short-term
nutritional status. From the measurement point of view, one advantage of wasting is that it does
not require information on the age of the child, which is often hard to ascertain precisely.
“Weight-for-age” can be seen as a more comprehensive indicator, which captures stunting as
well as wasting: both stunted and wasted children are likely to fall in the “underweight”
category. Thus, if a single “summary” indicator is to be used, weight-for-age would claim special
attention.
The rationale of this approach is not that there is anything intrinsically “wrong” with being
short or lean. While Amitabh Bachchan’s height probably served him well, Sunil Gavaskar and
Sachin Tendulkar are not doing too badly either. In many cases, being short or lean is not a
serious impairment. However, there is evidence that pronounced stunting or wasting in childhood
is associated with serious deprivations, such as ill health, diminished learning abilities, or even
higher mortality. More precisely, there is a great deal of variation in the genetic potential of
individuals to be small or large, but nutritional deprivation in early life, resulting in a failure to
attain one’s genetic potential, is likely to cause lasting harm, see for example the review article
by Victora et al (2008) which focuses on low and middle-income countries. While genetics are
important at the individual level, they are much less so—and arguably completely unimportant
at the population level, so that populations with a high fraction of people who are stunted or
underweight are populations where there is evidence of nutritional deprivation.
23
Children’s anthropometric status is usually evaluated on the basis of international standards,
which reflect the anthropometric features of children in a well-nourished reference population.1
The distribution of heights and weights in the reference population is used to set “cut-offs”
below which a child would be considered stunted, wasted or underweight. A standard cut-off is
“median minus two standard deviations”, based on the reference population so that, for example,
a child of a given age who is shorter than this cut-off would be considered stunted.
This procedure is based on the assumption that the anthropometric achievements of children
in the focus population (here, India) would be much the same as in the reference population, if
these children were well nourished. In other words, anthropometric standards are similar in both
populations. Since this procedure tends to be applied all over the world, it amounts to saying that
there are “universal” anthropometric standards for children. We shall return to this assumption,
but accept it for now.
Nutrition Status ofIndian Children
There are two major sources of anthropometric data for Indian children: the National Nutrition
Monitoring Bureau (NNMB) at the National Institute of Nutrition, Hyderabad, and the National
Family Health Survey (NFHS). Considering the former first, the NNMB surveys go back to the
1970s, but are available for nine states only: Andhra Pradesh, Gujarat, Karnataka, Kerala,
Madhya Pradesh, Orissa, Maharashtra, Tamil Nadu and West Bengal (hereafter the “NNMB
1 Until 2006, the World Health Organization (WHO) recommended the US National Center for Health
Statistics (NCHS) standard, and this was used inter alia in the first and second rounds of the National
Family Health Survey. In April 2006, the WHO released new standards “based on children around the
world (Brazil, Ghana, India, Norway, Oman, and the United States) who are raised in healthy
environments, whose mothers do not smoke, and who are fed with recommended feeding practices
(International Institute for Population Sciences, 2007, p. 268). These new standards were used in the third
National Family Health Survey.
24
states”).2 In each state, the NNMB surveys cover 20 households per village in 120 villages.
From 1991-2 onwards, the NNMB sample villages are sub-samples of National Sample Survey
(NSS) samples. In the “repeat surveys” (1975-79, 1988-90 and 1996-97), the 120 sample
villages consist of 90 villages from the previous survey and 30 “new” villages. The repeat
surveys are meant to be fully comparable and provide perhaps the safest basis for assessing
nutrition trends from NNMB data, at least if height selective out-migration is not important. In
Table 9, we present indicators of child nutrition based on the repeat surveys, extended using the
most recent NNMB surveys.
One difficulty with this exercise is that some states are “skipped” in some surveys:
specifically, West Bengal in 1975-79 and 1988-90, and Madhya Pradesh in 1996-7. Further, it is
not possible to construct a consistent series of aggregate figures for the remaining seven states
from the NNMB reports, due to missing data. However, weight-for-age data are available state
wise for each survey year from the NNMB reports, and close scrutiny of these data shows that
omission of one of these two states makes little difference to the aggregates. Thus, the aggregate
figures presented in the NNMB reports (for nine states, with Madhya Pradesh missing in one
year and West Bengal missing in two years) are good enough for the purpose of assessing broad
nutrition trends. These are the figures shown in Table 9 for child nutrition, and later in Table 12
for adult nutrition.
The child nutrition indicators presented in Table 9 point to a steady retreat of severe
undernourishment during the last thirty years or so. For instance, between 1975-9 and 2004-5,
2 The rationale for the selection of these nine particular states (a tenth state, Uttar Pradesh, was added later
on) is not clear from the NNMB reports. According to a personal communication from NNMB, they just
happen to be the states that responded to initial requests for cooperation from the Indian Council of
Medical Research. For a useful appraisal of nutrition monitoring arrangements in India, see
Ramachandran (2006).
25
Table 9: Child Nutrition Indicators, 1975-9 to 2004-5
1975-79
Proportion (%) of undernourished children3
1988-90
1996-97
2000-1
2004-5
Percentage
decline (19759 to 2004-5)a
Weight-for-age
Below 2 SD
Below 3 SD
Height-for-age
77
37
69
27
62
23
60
21
55
18
29
51
Below 2 SD
Below 3 SD
Weight-for-height
79
53
65
37
58
29
49
26
52
25
34
53
Below 2 SD
Below 3 SD
Prevalence of
nutritional deficiency
signs (%)
18
2.9
20
2.4
19
2.5
23
3.1
15
2.4
17
17
0.4
1.3
1.8
5.7
0.1
0.6
0.7
5.7
0.1
0.1
0.7
2.1
0.0
0.2
0.8
1.4
0.0
0.0
0.6
0.8
100
100
67
86
Oedema
Marasmus
Bitot spots
Angular stomatitis
a Based on NCHS standards (see Table 11).
b Difference between the 1975-79 and 2004-5 figures, as a ratio of the former.
Source-. National Nutrition Monitoring Bureau (1991, 1999, 2002, 2006). All figures pertain to children aged 1-5
years. These figures apply to the nine “NNMB states”: Andhra Pradesh, Gujarat, Kamataka, Kerala, Madhya
Pradesh, Maharashtra, Orissa, Tamil Nadu, West Bengal. Data for 1975-79 and 1988-90 exclude West Bengal, data
for 1996-7 exclude Madhya Pradesh. See text for further discussion.
there have been reductions of around 50 per cent in the prevalence of severe undemutrition
(measurement below 3 standard deviations of the median of the reference distribution), whether
one looks at weight-for-age (37 percent of the child population to 18 percent) or height-for-age
(53 percent to 25 percent).3 The retreat of severe undernourishment can also be seen from the
3 The undemutrition figures are very similar for boys and girls. There is, in other words, little evidence of
major gender differences in anthropometric achievements, according to NNMB data. The same
conclusion follows from NFHS data. For instance, according to NFHS-3, boys and girls had exactly the
same “mean z-scores” in 2005-6, not only in terms of weight-for-age but also in terms of weight-forheight and height-for-age (International Institute for Population Sciences, 2007, p. 270). This pattern is
26
sharp decline over the same period in the prevalence of clinical signs of nutritional deficiency,
such as marasmus and oedema. These findings are consistent with the decline of self-reported
hunger presented earlier in Table 6.
Having said this, the overall levels of child undemutrition in India (including not only severe
but also “moderate” undernourishment) are still very high, both in absolute terms as well as
relative to other countries. Even today, close to half of all Indian children are underweight, and
about half suffer from anemia. These are appalling figures, which place India among the most
“undernourished” countries in the world. According to the 2007 World Development Indicators,
only two countries have higher proportions of underweight children (based on the same
standards): Bangladesh and Nepal. While Pakistan and Sri Lanka have somewhat lower levels of
child undemutrition, the whole South Asian region stands apart from the rest of the world in this
respect. In particular, child undemutrition is much higher in South Asia (48.5 percent
underweight in 1999) than in Sub-Saharan Africa (29.6 percent underweight in 2005), although
the most undernourished countries in both regions fare much the same, as Table 10 indicates.
National Family Health Survey (NFHS) data corroborate these basic patterns, with some
qualifications. Three rounds of the NFHS are available so far. They were conducted in 1992-3,
1998-9 and 2005-6. We shall refer to them as “NFHS-1”, “NFHS-2” and “NFHS-3”
respectively. As far as levels of undemutrition are concerned, NNMB and NFHS data are broadly
consistent. For instance, both NNMB and NFHS place the proportion of underweight children at
around half in the latest year for which data are available (2004-5 and 2005-6, respectively).
perhaps a little surprising, considering that there is much evidence of various forms of discrimination
against female children in India, reflected inter alia in higher mortality rates for girls than for boys. For
further discussion of this issue, see e.g. Harriss (1990) and Tarozzi and Mahajan (2007).
27
However, there are significant differences between these two sources as far as trends in child
nutrition are concerned.
Table 10: Countries with the Highest Levels of Child Undernutrition, 1996-2005
Country
Proportion (%) of children with
____ low ‘‘weight for age”____
48.3
Nepal
47.5
Bangladesh
46.7
India
45.8
Timor-Leste
45.6
Yemen
45.1
Burundi
41.9
Madagascar
40.7
Sudan
40.4
Lao (People’s Dem Rep)
40.1
Niger
39.6
Eritrea
39.3
Afghanistan______________________________________________________
Source -.World Development Indicators, 2007. Figures apply to the most recent year for which data are available
within the reference period. There is a significant margin of error for individual countries.
Recent Trends in Child Nutrition
Until recently, trends in anthropometric indicators presented a reasonably clear picture. Whether
one looked at, say, the heights or weights of children, or the “Body Mass Index” (BMI) of adults,
the dominant pattern was one of sustained improvement. We have already noted some of these
trends in the preceding section, as they emerge from NNMB data (Table 9). The first two rounds
of the National Family Health Survey (NFHS-1 and NFHS-2) appeared to be broadly consistent
with the NNMB data in this respect. For instance, according to NFHS data, the proportion of
underweight children declined from 52 per cent in 1992-3 to 47 per cent in 1998-9 (International
Institute for Population Sciences, 2000a, p. 267). NNMB data suggest a similar rate of decline—
about one percentage point per year in the 1990s (Table 9). As discussed below, BMI data for
Indian adults also show a sustained improvement over time, albeit from levels of height and
weight that were (and still are) among the lowest in the world.
28
However, the most recent data, from the third round of the National Family Health Survey,
(NFHS-3), tell a different story. According to NFHS-3, the proportion of underweight children in
2005-6 (using the same standards as in NFHS-1 and NFHS-2) was 46 per cent—virtually the
same as in 1998-9. This apparent “stalling” of nutritional improvement came as a rude shock
when these numbers were published in late 2006, at a time of widespread euphoria about India s
rapidly growing economy. If correct, this statistic raises serious questions about what is currently
holding up the progress of child nutrition in India.
The NFHS-3 data on weight-for-age, however, are not quite conclusive in this respect, for
several reasons. First, the evidence on weight-for-age needs to be read together with other
indicators of child nutrition, such as height-for-age and weight-for-height. A fuller picture of
NFHS-3 data (and the corresponding NFHS-2 estimates) is presented in Table 11, which shows
that while the proportion of underweight children remained virtually unchanged between 199899 and 2005-6, there was a significant decline in stunting (from 51 to 45 per cent), and a small
increase in wasting (from 20 to 23 per cent). The stagnation of underweight indicators can be
thought of as an averaging of the opposite movements of stunting and wasting but it is far from
clear why wasting should increase while stunting goes down. Perhaps the decline of stunting in
this period reflects continued improvements in maternal health, themselves related to nutritional
improvements that occurred twenty or thirty years ago (when today’s mothers were children).
The increase in wasting, for its part, is quite puzzling, and it is difficult to come to any firm
conclusion without a better understanding of these opposite trends.
Second, there are also contrasting indications from NFHS and NNMB data. Indeed, the latest
NNMB survey, conducted in 2004-5, suggests a continuation of the earlier decline of
undemutrition as reflected in weight-for-age data, in contrast with the “stalling” pattern in the
29
Table 11: Trends in Child Nutrition: NFHS Data
Proportion (%) of children under the age of three years who are
undernourished
TVew WHO Standards
NCHS Standards
1992-3
1998-9
2005-6
1998-9
2005-6
Below 2 SD
Below 3 SD
Height-for-age
52
20
47.0
18.0
45.9
n/a
42.7
17.6
40.4
15.8
Below 2 SD
Below 3 SD
Weight-for-height
n/a
n/a
45.5
23.0
38.4
n/a
51.0
27.7
44.9
22.0
Below 2 SD
Below 3 SD
n/a
n/a
15.5
2.8
19.1
n/a
19.7
6.7
22.9
7.9
Weight-for-age
Source'. International Institute for Population Sciences (2000), pp. 266-7, and International Institute for
Population Sciences (2007), p. 274. The 2005-6 figures based on NCHS standards are taken from the
“National Fact Sheet” (International Institute for Population Sciences, 2007b).
Notesdl) The National Center for Health Statistics (NCHS) standards, recommended until 2006 by the
World Health Organization (WHO), were used in NFHS-1 and NFHS-2, and are also used by the National
Nutrition Monitoring Bureau (see Table 9). The new WHO Child Growth Standard, released in 2006 (see
WHO, 2008), are used in the NFHS-3 report (UPS, 2007a), although the provisional NFHS-3 “Fact
Sheets” (UPS, 2007b, 2007c) present figures based on the NCHS standards that appear to be comparable
with the NFHS-1 and NFHS-2 figures; these are shown here in the middle column. 2) The NFHS-2
figures are based on the last two children under three years of age of the sample women (ever-married
women in the age group of 15-49 years).For purposes of comparability, the NFHS-3 figures presented in
the last column focus on the same group.
NFHS data (Table 9). Further, this decline is associated with a sharp decline in wasting, and
some increase in stunting—almost the “reverse” of the NFHS trends on these indicators.
Third, there is a possible issue of comparability between NFHS-2 and NFHS-3 data. This is
not immediately obvious from anthropometric data, but other indicators give reason for doubt in
4 The NFHS trends mentioned earlier were based on all-India figures, but population-weighted figures for
the nine NNMB states indicate that the same broad trends (including a marginal decline in underweight
between 1998-99 and 2005-06, a larger decline in stunting, and an increase in wasting) apply to the
NNMB states as well. The changes are of a similar order of magnitude in both cases.
30
this respect. For instance, the NFHS surveys suggest that there has been a decline of school
attendance between 1998-99 and 2005-06, even though there is much independent evidence of
major increases in school attendance during that period, from the National Sample Survey
(Sankar 2007) and other sources (e.g. Samson et al, 2008). There are also inconsistencies
between NFHS-2 and NFHS-3 in the measurement of adult height, which we discuss in the next
subsection. In short, the “crisis” signals emerging from NFHS-3 data on weight-for-age are yet to
be corroborated, and further research (and evidence) is required to make sense of thee
contrasting indications on recent nutrition trends.
Having said this, even if we ignore the most recent crisis signals and take a somewhat longer
view (for which there is evidence of sustained improvement, as mentioned earlier), the progress
of nutrition indicators in India seems to be undistinguished, given the country’s high rates of
economic growth from the early 1990s onwards. To illustrate the point, one recent study of
international data concludes that the rate of decline of child undemutrition (based on weight-for
age) tends to be around half of the rate of growth of per-capita GDP (Haddad et al, 2003). If this
rough benchmark can be applied to India, which grew at 4.2 percent a year from 1990 to 2005,
we would have expected the proportion of underweight children to have declined by 2.1 percent
a year, or by about 27 percent since 1990. But the actual decline was only about 20 per cent,
according to NNMB data for nine states (Table 9), and much less (barely 10 per cent) according
to NFHS data. Over the longer period, from 1980 to 2005, which includes the earlier decade of
somewhat slower growth, the predicted decline is 38 percent, and the actual improvement was 29
percent. As we have already noted, the growth rates of per capita consumption in Table 4 are a
good deal lower than the measured growth rates of per capita GDP, almost certainly because of
measurement errors on both sides, and it is possible that the growth rate of per capita GDP is
31
somewhat exaggerated, and the nutritional improvements are close to what might be expected
based on the international benchmark. For example, a growth rate of 3 percent a year from 1990
to 2005, which is more than twice the growth rates recorded in Table 4, would be consistent with
the 20 percent improvement in Table 9. Of course, there is little comfort in saying that there is
little evidence of high economic growth among the malnourished. In any case, these are sobering
calculations. If India is indeed growing as fast as is claimed, there are important requirements of
better nutrition that are not taking place as they should, and this is in spite of sustained recent
improvements in other important determinants of child nutrition, such as maternal education and
the availability of safe water.
The contrast between India and China is also of some interest in this context. There is
evidence of a steady growth in the heights of Chinese children in recent decades, not only during
the period of fast economic growth that followed the “economic reforms” of the late 1970s, but
also before that. For instance one recent study reports that, in a representative sample of Chinese
children aged 2-5 years, the average increase in height between 1992 and 2002 was 3 cm in rural
areas (for both boys and girls), and even higher in urban areas (3.6 cm and 3.8 cm for boys and
girls, respectively); see Yang et al (2005).5 And, according to an earlier study, “the average
heights of Chinese children between the ages of 7 and 14 years increased by approximately 8.04
cm between 1951-8 and 1979” (Harris 2000 based on Piazza 1984).
NNMB data suggest much slower growth rates and heights of Indian children. For instance, in
the “NNMB states” the increase in children’s heights between 1975-9 and 2004-5 was a little
5 This article is in Chinese; the statement in the text is based on the abstract in English. According to this
abstract, the findings are based on a national, representative sample: "‘Subjects of 71,971 households from
31 provinces were selected by stratified multi-stage cluster random sampling method .
32
below 2 cm per decade at age 3, and barely 1 cm per decade at age 5.6 The NNMB data also
suggest that the growth rates of heights and weights were particularly slow in the later part of
this period, with, for instance, very little growth in the heights of children at age 5 between 19967 and 2004-5.
Here again, however, there are significant differences between NNMB and NFHS data. For
instance, a comparison between NFHS-1 and NFHS-3 suggests that children’s height at age 3
increased by about 2.5 cm per decade between 1992-3 and 2005-6, which looks much closer to
the Chinese rates of increase. Further work is required to reconcile these different sources, and to
assess the comparative progress of nutrition indicators in India and China.
Adult Weights and Heights
Recent nutrition trends can be further scrutinized from available data on adult weights and
heights. A useful starting point is the “Body Mass Index” (BMI), defined as the ratio of weight
(in kilos) to the square of height (in meters). Table 12 presents the proportion of men and women
Table 12: Nutrition Status of Indian Adults, 1975-9 to 2004-5 (Body Mass Index)
Proportion (%) of adults with Body Mass Index below 18.5
2004-05
2000-01
1996-97
1988-90
1975-79
Men
Women
56
52
49
49
46
48
37
39
33
36
% decline
(1975-9 to
2004-5)
41
31
Sources’. National Nutrition Monitoring Bureau (1999, 2002, 2006). These figures apply to the nine “NNMB states”:
Andhra Pradesh, Gujarat, Karnataka, Kerala, Madhya Pradesh, Maharashtra, Orissa, Tamil Nadu, West Bengal. Data
for 1975-79 and 1988-90 exclude West Bengal; data for 1996-7 exclude Madhya Pradesh. See text for further
discussion.
6 This is subject to the earlier qualifications about missing data for specific states in some years. However,
as with other anthropometric data for children, it is unlikely that the basic trends would be altered if a
consistent series for the same states were to be constructed.
33
Table 13: International BMI Data (Women Aged 15-49 Years)
Mean BMI
Proportion (%) ofwomen
with BMI <18.5
South Asia
India
Bangladesh
Nepal
20.5
20.2
20.6
35.6
34.3
24.4
Sub-Saharan Africa
Eritrea
Ethiopia
Burkina Faso
Chad
Madagascar
Niger
Senegal
Nigeria
Zambia
Congo 2005
Guinea
Mauritania
Kenya
Uganda
Benin
Tanzania
Rwanda
Ghana
Malawi
Zimbabwe
Mozambique
Gabon
Lesotho
20.0
20.2
20.9
20.8
20.8
21.4
22.3
22.3
21.6
22.9
21.8
24.3
22.7
22.2
22.4
22.3
21.8
23.1
22.0
23.1
22.1
23.5
25.1
37.3
26.5
20.8
20.3
19.2
19.2
18.2
15.2
15.0
13.2
13.2
13.0
12.3
12.1
10.7
10.4
9.8
9.3
9.2
9.2
8.6
6.6
5.7
Population-weighted
average for sub-Saharan
Africa (23 countries)
21.9
15.8
Source’. “Demographic and Health Surveys” (DHS) data available at www.measuredhs.com. The reference years
vary between 2000-1 and 2005-6. India’s National Family Health Surveys (NFHS) are part of the DHS series.
with BMI below 18.5 (a standard cut-off conventionally associated with “chronic energy
deficiency”) in the nine NNMB states. The proportion of individuals with low BMI, like that of
underweight children, declined steadily during the last 30 years or so. In spite of this, Indian
adults today (like Indian children) have some of the highest levels of undemutrition in the world,
34
with 36 per cent of adult women suffering from low BMI (rising to well over 40 per cent in
several states).7 Among 23 countries of sub-Saharan Africa for which comparable data are
available from the Demographic and Health Surveys, only one (Eritrea) is doing worse than
India in this respect (Table 13). In fact, the proportion of adult women with low BMI is above 20
per cent in only four of these 23 countries (Burkina Faso, Chad, Eritrea and Ethiopia), and the
population-weighed average for all these countries together is 16 per cent, much less than half of
the Indian figure.
Data on the height of Indian adults also shed further light on nutrition and nutritional trends
in the past. Because people’s heights do not change after they are fully grown—at least until age
50 after which there is some shrinkage—and because adult height is set in childhood and in
adolescence, we can look at the history of nutrition and the disease environment by comparing
the adult heights of contemporary Indians who were bom at different dates in the past. This can
be done using NFHS-2 which collected data on the heights of adult women, and NFHS-3 which
collected data on both men and women. Figure 8 shows plots of average height by year of birth
for women from both surveys, and for men and women from NFHS-3. In order to fit both men
and women on the same graph, we have subtracted 10 centimeters from average men’s heights.
The scale for women is on the left, and the scale for men is on the right; note that apart from the
10 centimeter shift, the scales are the same.
Recall that NFHS-2 was collected in 1998-99 and NFHS-3 in 2005-06, so that adult height
is attained only for those bom prior to around 1978 in NFHS-2 and around 1985 in NFHS-3, and
that the apparent declines of heights after those dates in the figure come from the fact that the
7 International Institute for Population Sciences (2007), page 304. This is consistent with the NNMBbased figures presented in Table 12 for 9 states. The international figures are available at
http://www.measuredhs.coni/aboutsurveys .
35
relevant individuals are not fully grown. That adult height is attained so late in India—in the
early or mid 20s, compared with age 18 in the West—is itself a mark of poor nutrition. And
indeed, Indian women, along with Nepalese and Bangladeshi women—for whom data are also
available from the DHS surveys—are today among the shortest in the world, see Deaton (2007).
CD
CD
CD J
IO “
Men in NFHS3
M-
CD
ID “
U)T’(D
U)
'(1)
-C
C
05
c
CTJ
0)
Women in NFHS3
E
E
Sc
^CN
o
E
o
§
Women in NFHS2
O
CD
O
ID -
1950
1960
I
—i—
1970
1980
1990
Year of birth
Figure 8: Average heights of Indian men and women by age
Figure 8 shows also that later-bom men and women are taller, an indication that net nutrition
has been improving, which is consistent with the NNMB data on child nutrition discussed above.
The figure also shows that the rate of growth of men’s heights, at 0.056 cm a year, is more than
three times the rate of growth of women’s heights which is 0.018 cm a year in NFHS-3 and
0.012 cm a year in NFHS-2. Since we do not see any such differential improvement in stunting
or wasting in infants, the relatively favorable growth of boys must reflect discriminatory factors
that operate in later childhood. We do not know what these factors might b
36
■though differential
access to healthcare is certainly a possibility—and the finding echoes the similar (largely
unresolved) puzzles about the decline in female-male ratios in the Indian population in the 20
century, see Dreze and Sen (2002, Chapter 7) and Mari Bhat (2002).
To set adult heights in India in context, it is of interest to compare their growth with other
countries, as well as to look at differences between Indian states. Data from the China Health and
Nutrition Survey between 1989 and 2004 show that Chinese adults, both men and women, have
been gaining height at around one centimeter per decade. Not only have Chinese men grown
taller at twice the rate of Indian men, but there is no evidence in China of the differential
disfavoring of women that we see in India. In this respect, other neighboring countries also
appear to be doing better than India, although not as well as China. Women in Bangladesh have
been growing at about 0.24 cm. a decade, and women in Nepal at 0.59 cm per decade; these data
come from Demographic and Health Surveys which do not include data on mean’s heights, at
least as yet. That the Chinese growth rates are not impossible in India is clearly shown by the
data for Kerala, where men and women are growing taller at 1.29 and 1.16 cm per decade
respectively, even faster than in China, Deaton (2008, Table 1). Indeed, decadal rates of height
increase of around one centimeter have previously been reported for a range of European
countries between 1870 and 1970, Floud, Wachter, and Gregory (1990, Figure 1.7). For those
bom between 1950 and 1980, Scandinavians grew taller by a little less than one centimeter per
decade, while southern European countries such as Spain, Portugal, and Italy, where people are
shorter, were catching up with growth rates closer to 2 cm. a decade, Bozzoli et al (2008).
Taking all this together, the growth rates of heights in China and Kerala are in line with historical
experience, while India as a whole is making much slower progress, especially but not
exclusively for women.
37
The graphs in Figure 8 show that the data on women’s heights are inconsistent between
NFHS2 and NFHS3, with women of the same birth cohort apparently taller in the later survey.
The difference between the two lines is only 0.16 cm, but this is quite large relative to the annual
growth rate and, because it happens for all dates of birth, it is clearly not attributable to random
sampling error in the estimation of mean heights. We noted earlier that other questions have
arisen about the NFHS-3 so there was perhaps some sampling difference between the two
surveys. With a taller sample of women in NFHS-3, we would expect to see taller children—so
that the reduction in stunting between the two surveys may be overstated—as well as thinner
children—the children of taller women tend to have lower weight for height—which would
mean that the increase in wasting over the two surveys may be overstated. We can assess this
using NFHS-2 by estimating the effects of mother’s height on the probability of her child being
wasted or stunted. But when we do so, the effects are tiny and make no difference to the trends
discussed earlier; a sample of women that is 1.6 millimeters taller will have a fraction of children
stunted that is 0.26 of a percent lower, and a fraction wasted that is 0.03 of a percent higher.
3. Interpretations
3.1. The calorie decline
Food prices
Returning to the puzzle of falling calorie intake, the first possibility to consider is that the decline
was driven by an increase in food or calorie prices relative to the prices of other things. In spite
of the fact that people buy food, and not calories, we can think of price indexes of food as price
indexes of calories. To see this, suppose the price of food i is pi per kilo, and its calorie content
is k. per kilo, then the price per calorie through good i is
38
/ k., so that provided the calorie
Ratio of food to general price indexes
CPIAL
CPIIW
ID CT)
O) -
i
i
i
i
l
1985
1990
1995
2000
2005
Figure 9: Ratio of food to non-food price, urban and rural
contents do not change over time—an issue to which we will return—calorie prices are just food
prices in different units and any index of food prices can be reinterpreted as an index of calorie
prices, at least for foods that have non-zero calorie content. In consequence, calories become
relatively more expensive or cheaper according to whether food prices rise faster or slower than
non-food prices. Figure 9 shows the monthly ratios of food to general price indexes for the rural
(CPIAL) and urban (CPIIW) sectors. All four indexes, CPIAL food, CPIAL general, CPIIW
food, and CPIIW general have been scaled to be 100 in 1993-94, and are shown as ratios of food
to general within sectors. In rural India, food (and therefore calorie) prices moved along with
general prices from 1983 until about 2000, and then fell by a little less than five percent relative
to general prices. In urban India, there was a slow secular increase in the relative price of food,
by less than five percent, from 1983 until the late 1990s, followed by a more pronounced decline,
39
by more than ten percent, until the end of the period. In both sectors, the relative price of food
was lower in 2004-05 than at the start of the period in 1983. The decline in calorie consumption
cannot therefore be attributed to any increase in the relative price of food. In Deaton (2008) one
of us has argued that the food component of the CPIAL, as well as the CPIAL itself, is
understated after 2000 through the use of outdated weights in the official indexes. However,
there is no question about the relative decline in food prices after 2000, or about the long term
decline between 1983 and 2004-05.
Coarse cereal prices
Although food prices may not have risen in general, it is conceivable that some component of
food prices has risen, and in particular that the decline in coarse cereals was driven by an
increase in their relative price, and that this, by itself, might go at least some way to explain the
decline in cereal and in total calorie consumption. This explanation is somewhat implausible in
theory, because we would expect an increase in the price of one cereal group to lead to a
substitution into another cereal—in this case wheat—not to a reduction in overall calories, whose
general price is falling in relative terms. It is also false on the numbers, see Table 14. Rural food
Table 14: Chained Laspeyres price indexes, round by round, food prices and coarse cereal
prices
URBAN
RURAL
38-43
43-50
50-55
55-61
Food
Coarse Cereals Food
1.353
1.780
1.804
1.118
1.139
1.598
2.075
0.940
1.399
1.796
1.785
1.168
5.238
3.550
4.857
38-61
Source'. Authors’ calculations from the unit values in the NSS surveys.
40
Coarse Cereals
1.150
1.762
2.253
1.000
4.565
prices rose by 4.8 times from 1983 to 2004-05, and urban food prices by 5.2 times. The
corresponding figures for coarse cereals are 3.6 times and 4.6 times, so that coarse cereals (and
calories from coarse cereals) have become systematically cheaper than food (and calories from
food) as a whole. Only in the period from 1993-94 to 1999-2000 did the price of coarse cereals
rise more rapidly than other food prices.
Impoverishment?
If it is not relative prices that are shifting the calorie Engel curves, what can be causing the
calorie decline? One account, most prominently argued by Patnaik (2004, 2007), is that hunger
and deprivation are increasing, especially in agriculture, and that people are purchasing fewer
calories, particularly cereal calories, because they cannot afford to do otherwise. Some
combination of falling incomes, rising prices, and rising agricultural unemployment has
impoverished a substantial fraction of the rural population.
However, there is little direct evidence of sustained, widespread impoverishment of the rural
population in the 1980s and 1990s. Further, the impoverishment argument jars with the fact that
the decline of nutrient intakes has been proportionately larger much larger in the higher per-
capita expenditure groups, see Figures 1 and 2. There is strong evidence that these groups have
enjoyed sustained increases in living standards in the 1980s and 1990s. Clearly, something other
than “impoverishment” (e.g. a reduction in nutrient requirements associated with reduced
activity levels or better health) needs to be invoked to explain the decline of nutrient intakes
among these groups. And it is quite possible that this hidden factor, whatever it was, also played
a role in the stagnation of nutrient intakes among lower-income groups.
41
The impoverishment view also requires us to challenge some aspects of the data presented so
far. In particular, it implies that the rise in real per capita expenditure in the NSS must somehow
be incorrect, possibly because expenditures are overstated, price inflation understated, or some
combination of the two. Or perhaps there has been a substantial widening of the income
distribution, with the majority of people getting worse off, so that growth at the mean is driven
only by progress at the top. But with a few minor exceptions to do with weighting, the inflation
rate in the CPIAL is supported by the unit values observed in the NSS surveys. Nor is there any
evidence in Table 4 that the poor are getting worse off, even if they have done less well than the
average. And while the NSS expenditure figures have been robustly challenged, the general
complaint—and one that we endorse, at least to some extent—is that, in light of the national
accounts, they appear to underestimate (rather than overestimate) the growth of per-capita
expenditure over time.
Sen’s argument
In an interesting paper (concerned mainly with the method of poverty measurement), Pronab Sen
(2005) takes up the question of why it is that people around the current poverty line are
purchasing many fewer calories than the 2,400 (rural) and 2,100 (urban) recommended
allowances that played a part in the original derivation of the lines. Since the purchasing power
of the poverty line has been held constant, by construction, the recommended allowances are, in
principle, no less (or more) affordable than when they were established. As Sen points out,
however, it is possible that the food budget has been squeezed out because “the cost of meeting
the minimum non-food requirements has increased” (p. 4612). To illustrate, suppose that poor
households have access to schooling facilities that were not available earlier. Schooling expenses
42
(including, say, adequate clothes for children) would become a new item on the household
budget, and food expenditure may be curtailed to make room for it.
In response to this argument, Sen notes that households around the poverty line could still
achieve the recommended calorie intakes within their current food budget by spending their
actual food budgets, not as they do, but on the foods that yield the same number of calories per
rupee as are actually purchased by people below the poverty line on average. In other words,
people at the line could meet the calorie norms within their current actual food budgets if they
had the same food purchasing patterns as the typical poor person. This is a useful point to note, if
food purchases below the poverty line are not of much lower quality than food purchases at the
line. But it does not establish that there is no “squeeze’ in the food budget.
What would be more useful to know is whether households around the poverty line can or
cannot use their current expenditures on food to purchase their baseline food basket, which meets
the calorie requirements, by definition. The answer to this is likely to be “:no”. Indeed, since the
poverty line has constant purchasing power in real terms, and since the relative price of food
(and thus of calories) did not change much from 1983 until 2004-5, purchasing that baseline
food basket with its recommended calorie content would require these households to spend the
same share of total per-capita expenditure on food as in the initial scenario. In fact, the food
share has fallen at all levels of per-capita expenditure, including the poverty line. Thus, it is not
possible for households around the poverty line to purchase their initial food basket within their
currentfood budget
In short, there is no simple “revealed preference argument” that refutes the hypothesis of a
squeeze in the food budget due to rising essential expenditure on non-food items. It is worth
noting, however, that in the event of such a squeeze, one would expect poor households to switch
43
towards cheaper calories. But the opposite is actually happening poor rural households have
made a modest switch towards more expensive calories, away from cereals and other sources of
cheap energy, see Figure 3. This does cast some doubt on the hypothesis of a squeeze in the food
budget. It is also consistent with the possibility that people need fewer calories today than they
used to, an alternative hypothesis we examine below.
Are the calorie Engel curves misleading?
Another possibility is that the Engel curves in Figures 1 and 2 are biased, and that the true Engel
curves are actually downward sloping, which could possibly reconcile the cross-section and
time-series evidence without any need to account for downward shifts in the curve. This is only a
real possibility for the cereal calorie curves; for total calories, the estimated Engel curves are
steep—so that there is a high slope to explain away —and it is not plausible that richer people
consume fewer total calories than poorer people do. Some support for the idea that cereal
calories fall as people get better-off comes from NNMB data, which show per capita
consumption of cereals declining with household income (e.g. NNMB 1999, Table 6.9).
Somewhat similar results are reported by Subramanian (2003) using an NCAER survey of 2,000
rural households. He estimates cereal Engel curves that slope up among the poorest households,
but are flat throughout most of the range of per capita income, similar in shape to some of the
recent urban curves in Figure 2. Income is notoriously difficult to measure in rural India,
especially for self-employed cultivators, and measurement error can easily cause the Engel curve
to be spuriously flattened. However, measurement error will only turn a negative slope into a
positive one if richer people systematically report lower incomes than poorer people, which is
quite unlikely.
44
One possible source of bias in the NSS Engel curves arises from the fact that, in the NSS
surveys, both total expenditures and total calories are estimated by “adding up” expenditures and
quantities applying to various goods (in the NNMB surveys, by contrast, there are independent
questions on income and food intake). Thus, measurement errors in reported quantities of goods
will find their way both into total expenditures and totals (or subtotals) of calories and this
common measurement error biases the estimated slope of the calorie Engel curve towards unity,
and could conceivably convert a true negative slope into an apparent positive one, see Bouis and
Haddad (1985) and Subramanian and Deaton (1988). It is difficult to deal with this problem in a
convincing way. The standard treatment is to look for some variable that is correlated with total
expenditure, but which is measured in a way that is not contaminated by the measurement errors
that are common to calories and to expenditures. We have done some experiments along these
lines, for example using the ownership levels of the various durable goods that are collected in
the NSS surveys. Better-off people are more likely to own a bicycle or an electric fan, so we can
follow Filmer and Prtichett (2001) and compute an index of durable ownership that can be
thought of as a proxy for living standards.
These experiments yield suggestive but ultimately inconclusive results. The simple count of
the number of durable goods owned is (a) strongly positively correlated with per capita total
household expenditure, and (b) negatively correlated with per capita consumption of cereal
calories, which is consistent with the view that per capita cereal calories are negatively correlated
with income, and that the Engel curves in Figure 2 are biased by the common measurement
errors. One problem with this story is that it is possible that ownership of at least some of the
durable goods is directly linked to the consumption of cereals, not just indirectly as a proxy for
living standards. Consider, for instance, the fact that household ownership of a bicycle or motor
45
cycle is a positive predictor of per capita expenditure and negative predictors of cereal
consumption. This could simply reflect that people who own bicycles do not use as many
calories in walking to work or to school, rather than the fact that ownership of a bicycle indicates
higher income and, through higher income, lower cereal consumption. Indeed, the increased
ownership of (human) energy-saving durable goods over the last decade or two is likely
implicated in the calorie decline, at least to some extent. Another problem is that these results are
not replicated if we run the regressions state by state, or even if we pool states but allow for state
dummies. Yet if the common measurement error account is correct, it should hold within states
as well as across them.
Changes in food habits and non-market entitlements
A simple explanation for the shift from cereals, albeit a not very informative one, is that there
has been a change in “consumer tastes”, and some econometric studies have taken this route, at
least implicitly, by including unexplained time trends in the analysis. While invoking changes in
tastes is not helpful in itself, the last few decades have seen important changes in food habits in
rural India that are not easily explained by standard factors such as changes in income and prices.
Consider for example the sharp decline in the consumption of “coarse cereals”. In so far as
coarse cereals are “inferior goods” (with a negative income elasticity of demand), this decline
can be interpreted as an income effect. But there is obviously more to it, since the consumption
of coarse cereals is also falling at any given level of per-capita expenditure.
There are various possible reasons for this decline: growing emulation of the consumption
patterns of affluent groups, exposure to new food items, influence of advertisement, and changes
in food habits (specifically, substitution towards wheat and rice) induced by the Public
46
Distribution System. It is possible that some of these changes have nutritional consequences
about which people are not fully informed, for example, for the consumption of important
micronutrients. To illustrate, while “coarse cereals” are generally considered quite healthy and
nutritious by nutritionists (some of them even argue for calling them “nutritious grain” rather
than “coarse cereals”), they may or may not be perceived as such by the consumers. Conversely,
the nutrition value—or lack of it—of “junk foods” (some of which are rapidly gaining popularity
in India, and not just in urban areas) is not always adequately understood. Thus, it is conceivable
that recent changes in food habits (in particular, the decline in coarse cereal consumption) have
had nutritional consequences that are not folly appreciated or taken into account by consumers.
A related issue concerns the decline of certain types of food entitlements, not necessarily well
captured in NSS-type data. For instance, buttermilk used to be widely available, often free of
cost, in many Indian villages, particularly in the north-western region. In those days, large
quantities of buttermilk were available as a by-product of ghee, butter and other local milk
products. In many villages there was no market for buttermilk—it was just consumed at home or
given away to neighbours, friends and visitors. With the growing commercialization of milk,
these local processing activities have declined, and so has the availability of buttermilk. It is
unlikely that this steep decline in the consumption of buttermilk is adequately captured in
standard NSS data. Similarly, as the role of home-grown consumption has lessened, the effective
price of food is likely to have risen in a way that is not captured in the usual indexes and which
may help explain some of the decline in cereal consumption among the poor; the size of this
effect is hard to document with available data.
Some studies (e.g. Das Gupta, 1985, Jodha, 1986) have also noted the decline of a range of
traditional food entitlements in various parts of rural India, such as the growing scarcity of wild
47
spinach (saag) due to recent changes in cropping patterns, the reduced availability of fish in rice
fields due to pesticide use, and deprivation of forest products or common property resources due
to environmental degradation. All this could have an adverse impact on food intake and
nutritional status. However, it does not help to explain the decline in measured calorie
consumption (if these items were not covered by household survey data in the first place), or for
that matter in cereal consumption, especially among the better off sections of the population.
Declining needs for calories?
The declining consumption of calories needs to be examined relative to changes in the various
factors that might affect the need for calories. For example, as mentioned earlier the decline of
the total fertility rate—from 6.6 in 1960 to 5.0 in 1980 and 2.8 in 2005—is a source of reduction
in requirements (since pregnant and lactating women need more calories), though if those
requirements are not met, this may not have had much effect on actual calories consumed. For
the same reason, but working in the opposite direction, households today have fewer children
relative to adults, so that the decline in per capita calories has most likely understated the decline
in calories per adult equivalent.
One possible source of falling calorie requirements is an improved epidemiological
environment, with less exposure to disease and infections. For example, there has been a major
improvement in access to safe drinking water during the last twenty-five years or so. the
proportion of households with access to piped water has risen sharply across the country, see
Figure 10. While piped water is not necessarily safe, much of the increase came from the
construction of hand-pumps and tube-wells, whose water is usually safe, and certainly safer than
water obtained from rivers, tanks, or open wells. Better water reduces the prevalence of diseases,
48
g
especially diarrhoeal disease, and removes a potentially major source of calorie wastage. Other
improvements in the disease environment may also reduce calorific needs. For instance, child
vaccination rates have risen, and child health improves with mother’s education, which has also
risen rapidly in recent years. To the extent that these and other improvements in the health
environment reduce the susceptibility of children (and adults) to disease and infection, calorie
requirements would be reduced.
Proportion of households
with access to piped drinking water, 1981
Proportion of households
with access to piped drinking water, 2001
Legend
Legend
03-26
CO9-25
(7^26-40
26-40
<i -so
41 - 60
» 61 - 80
MI61-80
HH81 -100
□J Missin9
mi si-icn
!...... J M,si,n9
Figure 10: Access to piped water, 1981 and 2001
Another possible source of reduction in calorie requirements is the reduction of activity
levels (especially in rural areas), see for example Rao (2000, 2005). Aside from reducing
8 Calorie requirements increase sharply during episodes of diarrhoea. Note, however, that adults or
children often reduce calorie intake during episodes of diarrhoea, contrary to expert recommendations
(Scrimshaw et al, 1983). In that case, reduced exposure to diarrhoea may be irrelevant as an explanation
for the calorie decline, though it will enhance the nutritional effects of any given calorie intake.
49
exposure to disease, improved access to piped water is likely to reduce the energy requirements
associated with fetching and carrying water, a strenuous task (typically assigned to women and
children in rural India). Similarly, the extension of road coverage and transport facilities has
enabled more people to use motorised transport, and to save some of the energy spent earlier on
long walks (another strenuous activity, especially when it involves carrying heavy loads). The
general mechanisation of domestic activities and agricultural work would also contribute to
reduced calorie requirements. For instance, aside from fetching water, rural women used to
spend much energy on grinding flour at home (using heavy stone mills known as chakkis), but
today this is typically done outside the home with energised devices such as electric mills.
The NSS survey data show a large increase in the ownership of durable goods over the last
two decades, and many of these goods are likely to reduce the expenditure of human energy. In
rural and urban India in 2004-05, 56 and 50 percent of households reported owning a bicycle,
compared with 17 and 19 percent in 1987-88. In 2004-05, 31 percent of urban households
owned a motorcycle, and 6 percent owned a car. Although the rural figures are only a third as
high, there has been an almost tenfold increase in ownership rates over 17 years. The most recent
figures are confirmed by NFHS-3, which estimated that in 2005—06, 12 percent of rural
households and 32 percent of urban households owned a means of motorized transportation.
Television watching, a sedentary leisure activity, has also increased: 30 percent of rural and 74
percent of urban households now own a television, and these estimates are the same in both
surveys. In 1987-88, less than one percent of Indian households possessed a television.
The importance of accounting for activity levels in any calorie-based assessment of
nutritional status arises from the fact that calorie requirements increase quite sharply with the
level of activity. According to the Indian Council of Medical Research (ICMR), the calorie
50
requirements of an “average Indian man” (weighing 60 kgs) are almost 60 per cent higher if he is
engaged in “heavy activity” rather than “sedentary activity”—3,800 calories and 2,425 calories,
respectively. Clearly, looking at calorie intake independently of activity could be quite
misleading when the former varies so much with the latter.
Table 15: Activity levels and calorie requirements (adult men)
Proportion of adult
men engaged in
“moderate activity”
10%
20%
40%
60%
Proportion of adult men engaged in “heavy activity”
10%
2608
2653
2743
2833
2745
2790
2880
2970
20%40%
60%
3020
3065
3155
3245
3295
3340
3430
n/a
Source: Calculated from Indian Council of Medical Research (1990), page 29. Each cell indicates average
calorie requirements when the proportion of adult men engaged in “heavy activity” is as indicated in the
column title, the proportion engaged in “moderate activity” is as indicated in the row title, and other adult
men are engaged in “sedentary activity”. The estimated calorie requirements for each group are: heavy
activity 3800 calories; moderate activity 2875 calories; sedentary activity 2425 calories. All this applies to
the “average Indian man”, weighing 60 kgs.
For further illustration, Table 15 shows how the average calorie requirements of adult men
vary with activity patterns, according to ICMR data. In this table, the column title indicates the
proportion of the population engaged in “heavy” activity, the row title indicates the proportion
engaged in the “moderate” activity, and it is understood that the rest of the population is engaged
in “sedentary” activity. The cell entries give the corresponding average requirements, e.g. 2,790
calories per day when the proportions of adult men engaged in moderate and heavy activity are
20 per cent each. As the table indicates, relatively moderate reductions in activity levels would
lead to reductions of calorie requirements of the same order as those that have been actually
observed during the last twenty-five years or so (see Table 1). For instance, a reduction in the
proportion of adult men engaged in “heavy activity” from 40 to 20 per cent, with no change in
51
\ £-1 1-7
the proportion engaged in “moderate activity”, would lead to a decline in average calorie
requirements of close to 10 per cent, similar to the actual decline in average calorie consumption
in rural areas between 1983 and 2004. It is also reasonable to expect that the percentage activity
reductions would be larger among the better-off households, especially if a substantial part of the
calorie reduction reflects a substitution of mechanical for human power. This would help to
explain why the reduction in calorie intake is itself larger at higher levels of per-capita
expenditure.
This line of explanation also helps to interpret the rural-urban contrasts. As Figures 1 and 2
show, there have been marked reductions of cereal and calorie intake (at all levels of per capita
total expenditure) in both sectors, but the decline has been larger in rural areas, and the rural-
urban gap has narrowed over time, especially in the lower part of the per capita expenditure
distribution. The rural-urban contrast, with lower intake levels in urban areas, has been
traditionally attributed to differences in activity patterns, including the more sedentary lifestyle
of a large fraction of the urban population, the heavy energy demands of agricultural and
domestic work in rural areas, and the lack of transportation in the rural sector. The fact that
calorie intake is going down in both sectors, but with larger reductions in rural areas, fits the
activity story. The continuity across the two sectors in Figures 1 and 2, albeit with a lag of more
than 20 years, is consistent with the notion that the rural sector is becoming more like the urban
sector in its activity patterns over time.9
9 It is also worth noting that, within the rural sector, households that describe themselves as self-employed
in agriculture or as agricultural labourers use more calories than do non-agricultural rural households at
the same level of per capita expenditure—between six and fourteen percent less. But the shift out of
agriculture within the rural sector has been modest (from 74 percent of the rural population in 1983 to 64
percent in 2004-05), and can only account for a small part of the calorie decline.
52
This interpretation of the calorie decline, in terms of reduced requirements associated with
lower activity levels, is difficult to substantiate further in the absence of reliable data on activity
patterns. Time-use studies would help to shed further light on this account. Meanwhile, the
activity story does have some plausibility, and this also applies to the more general notion that
calorie requirements have declined, due not only to reduced activity levels but also to better
health.
Engel curves reexamined
The focus on activity can also help to understand some features of the NSS-based Engel curves
for calories and cereal calories in Figures 1 and 2. The basic point to recognize is that food
consumption has a “dual” role. On the one hand, food is like any other commodity, which is
valued (a source of “utility”) for various reasons such as the pleasure of eating, the pursuit of
better health, and the role of food in social occasions. On the other hand, calories are required to
meet one’s immediate energy requirements, themselves contingent on activity patterns. In
particular, energy expenditure is required to work and earn.
io
From this perspective, the distinguishing feature of cereals is that they provide a cheap (often
the cheapest) source of calories. At low levels of income, energy requirements tend to be met
mainly from cereals, because other foods are unaffordable. As income rises, people typically
switch to other, more expensive sources of calories that are also valued from the point of view of
taste, healthiness, status, and so on. The calorie requirements, for their part, depend partly on the
amount of work people do.
10 In terms of standard consumer theory, this suggests that attention needs to be paid not only to die usual “budget
constraint”, but also to the “energy constraint”: calorie consumption has to be no less than what is required to earn
the amount of income associated with the budget constraint.
53
To see the implications of cereals being valued mainly as a source of cheap calories, and in
particular as a cheap “fuel” for work, suppose that this is their only purpose. Further, consider a
population of rural workers who vary in terms of the amount of work they do, for reasons that
might include differences in health, in physical fitness, in work opportunities, or in individual
preferences. For instance, some may be able to find work, others not; some may be able to put in
long hours of hard work while others have limited strength or suffer from illness; and some may
simply chose to work hard and earn more while others prefer a more leisurely life. To keep
things simple, assume (to start with) that these workers face the same wage rate, and that calorie
requirements depend exclusively on work.
In this setting, which we think of as applying to agricultural workers, we would find a
positive relationship (across households) between total expenditure and calorie intake, because
both are driven by a third factor, which is the amount of work people do and the effort they put
into it. At these low levels of income, this relationship would be much the same for calories and
cereal calories, since calorie requirements are met mainly from cereals. At somewhat higher
income levels, the Engel curve for cereal calories would flatten out (and possibly even have a
negative slope) as other sources of calories substitute for cereals.
What happens over time is different. As the wage rises, and with less work being required to
earn a given amount of income, there will be a downward shift in the calorie Engel curve. This
may be accentuated by technological change of the sort we have discussed in the preceding
section. In the process, calorie intake may decline as incomes rise, in contrast to calorie intake
increasing with income in the cross-section at a given point of time. To look at it another way,
and emphasizing once again that we are talking about the lower portion of the Engel curve,
variations in income over time are driven (in this model) by wage variations, while those
54
observed at a given point of time are driven by variations in work and effort. The corresponding
segments of the Engel curves may have opposite slopes, positive in the cross-section but
negative over time.
This is a simplified model that applies, at best, to people with low wages whose income
depends mainly on physical labour. To bring it closer to reality, we also have to bring in the
people at the top of the Engel curve, among whom variations in income at a moment in time are
more likely to be due to variations in wages (or salaries) than to variations in work and effort.
Among these better off workers, it is reasonable to expect a negative relationship between cereal
intake and income, partly because people are less likely to be involved in strenuous activities as
they become better off, and partly because they would substitute away from cereals towards
other foods. Calorie intake may rise or decline with income: even if activity levels are lower at
higher levels of income, there are likely to be positive income effects that cause people to
consume more. When wages rise over time, the earlier reasoning would still apply to the extent
that the Engel curve is driven by variations in effort and work opportunities across people at any
given wage. In addition, at higher levels of income, there may be substantial scope for reducing
activity levels over time by adopting labour-saving technology such as agricultural machinery,
courtyard hand-pumps, domestic appliances, and so on, some of which are expensive and more
affordable by the better off.
This reasoning is also relevant in interpreting contrasts in calorie consumption between
different places, e.g., Indian states. Within states, the Engel curves would have the shapes
discussed earlier, with a relatively steep positive slope at the bottom, and some flattening (and
possibly even a negative slope, in the case of cereal calories) at the top. But across states, things
would look different, because the interpersonal differences in ability and opportunity to work
55
would largely average out. Thus, across states, there would be a negative correlation between
per-capita expenditure and cereal calories, driven by differences in wage rates. The correlation
between per-capita expenditure and total calories, for its part, may be positive or negative,
depending on the strength of different income effects. In fact, the correlation between state
averages of per-capita cereal calories and per-capita expenditure (in logs) is —0.44, see Figure 11,
<p
(/)
■§
o r-
• < •••
“ • 1
2o
• Me •
e *
,• -
•K’.
• •
’•••
Q.
CO
•
•n s
o
CD
CL
CD
_O
•
<sc
<D
E co _
(D. CD
W
w
CD .
CD Lt
5
5.5
6
6.5
State mean log of per capita total expenditure
Figure 11: State averages of cereal calories and PCE, large rounds
while for per-capita total calories, the correlation is 0.53. This pattern for cereal calories, which
appears to conflict with the upward-sloping Engel curves emerging from household data in
Figure 2, is quite plausible in this activity-focused approach. It is, of course, also possible that
this negative correlation is coincidental, in the sense that it reflects regional variations in calorie
requirements (related, for instance, to the climatological or epidemiological enviromnent) or
food habits that have little to do with the factors discussed in this section. Nevertheless, it is
56
interesting that, here again, the observed features of NSS-based Engel curves are consistent with
the recognition of activity levels as an important determinant of calorie requirements.
This completes our brief tour of alternative explanations for the recent decline of calorie
intake in India. If only by default, the hypothesis of declining calorie requirements commands
serious attention as a possible clue to this puzzle. However, it would be premature to endorse it
in the absence of adequate evidence on activity levels and other determinants of calorie
requirements.
Because of the uncertainty as to what is going on, it is also difficult to assess the welfare
implications of the decline in calorie intake. We have already noted that, had real wages and real
incomes increased faster at the bottom of the distribution, it is likely that calorie consumption
would have increased, at least for some sections of the population, and possibly even on average.
On the other hand, to the extent that the calorie decline is driven by lower activity levels or a
better health environment, it may be associated with improvements in some aspects of well-being
(such as freedom from hard labour) that escape assessment in the standard poverty estimates and
well-being indicators.11
3.2. Nutritional status
Farewell to Calorie-based Nutrition Assessment?
One important lesson from the preceding discussion is that average calorie intake per se is a poor
indicator of the nutritional status of the population (on this see also Vaidyanathan 2002). Calorie
requirements seem to be highly context-specific, depending on activity levels, the
11 In so far as the decline in calorie requirements comes from increased ownership of labour-saving durable goods,
such as bicycles, motor-cycles, or even televisions, the associated expenditures will be captured in the surveys. But
the surveys record purchases, not the use-flow associated with durable stocks. In consequence, the Indian boom in
durable ownership was recorded in the expenditure figures too early, which biases downward the rates of
expenditure growth and poverty reduction.
57
epidemiological environment, the composition of the population, and other factors. Thus, simple
comparisons of nutrition levels (say, between different regions or periods of time) based on
average calorie intake can be very misleading. Indeed, average calorie intake in India is lower
today than it was twenty years ago, yet the nutritional status of the population has improved.
Similarly, looking across the districts and regions of India, the correlation between average
calorie intake and, say, the nutrition status of children is negative, as Figure 13 illustrates: child
Per capita calorie consumption 2004-5
Proportion of underweight children, 2004-5
I
Legend
Ufytttid
WA2S0
□»-»
HBi*'
Figure 12: Underweight children and per capita calorie consumption
nutrition is worse in regions with higher levels of per-capita calorie intake.
12
Even if there were a
simple relationship between individual nutritional status and calorie consumption, there would be
12 Figure 12 juxtaposes region-level NSS data on calorie intake in 2004-5 with district-level child
nutrition data in 2002-4 from the second round of the Reproductive and Child Health (RCH) survey, also
58
serious perils in making comparisons of nutrition based on NSS-type data on per capita calorie
intake at the household level, with no idea of its distribution across people, let alone whether the
distribution over individuals matches individual calorie needs.
As discussed below, this does not detract from the importance of calorie intake in achieving
better nutrition. It does mean, however, that approaches to nutrition assessment based on calorie
intake data and “fixed” calorie norms (or on simple comparisons of average calorie intake) are
both incomplete and misleading. There is a strong case not only for paying attention to other
aspects of food intake (such as the consumption of essential vitamins and minerals) but also for
shifting the focus away from food intake towards more “direct” indicators of the nutrition status
of the population, such as the growth curves of children or other evidence of nutrition-related
impairments.
Energy adequacy depends on net intake, which is calorie intake less the calories absorbed by
a range of demands depending inter alia on the epidemiological environment as well as activity
levels, and is influenced by variables such as age, birthweight, mother s education, breastfeeding
practices, the composition of the diet, to cite a few. Adequate nutrition also requires a host of
other inputs that are not adequately summarized by total calories, including a range of
micronutrients and, at certain stages of life, especially childhood, a varied diet. Looking at
calorie intake, or for that matter other specific nutrition inputs, independently of other relevant
variables, is an exercise of limited value for assessing what matters, which is nutritional
outcomes. These can be usefully measured by the various nutrition indicators such as
anthropometric measurements, at least in so far as they capture the combined effects of various
inputs (food and non-food) net of other demands.
known as District Level Household Survey (DLHS). On the DLHS-RCH surveys, see Ramachandran
(2006).
59
Having said this, there are also possible difficulties with the use of anthropometric indicators
based on currently available data. As noted earlier, for instance, different anthropometric
indicators (e.g. weight-for-age and weight-for-height) do not always move in the same direction,
and sometimes the same indicator moves in different directions according to different sources
(e.g. NFHS and NNMB). There are also unresolved issues about the interpretation of current
anthropometric standards, as discussed below.
Multiple Deficiencies ofIndian Diets
Before we proceed, two clarifications may help. First, nothing we have said justifies a
downplaying of calorie deficiencies among poor households. Even if the recent decline of calorie
intake is due to falling requirements, and has little to do with impoverishment, the fact remains
that poor households have serious energy deficits. Had poverty decline been faster, it is quite
likely that calorie intake would have risen among underprivileged households (and perhaps even
on average). Reducing calorie deficits among poor households remains very important for the
removal of nutritional deprivation in India, and this, in turn, would be hard to achieve without
substantial income growth at the bottom of the distribution.
Second, by a similar reasoning, the findings we have presented do not detract from the
crucial importance of improved diets in India, including not only reduced calorie deficits but also
higher—possibly much higher—intakes of various nutrients for large sections of the population.
While economic research on nutrition issues in India has tended to focus mainly on calorie
intake, nutritionists have been deeply concerned for a long time about a range of nutritional
deficiencies, including those of essential minerals and vitamins, animal protein, and fats. As Tara
Gopaldas (2006) notes, for instance, ^all members of low-income (and even middle-income)
60
families are likely to be deficient in vitamins and minerals” (p. 3671). In fact, the deficiencies
are large, in relation to current ICMR recommendations. For instance, in the age group of 4-6
years, the ratio of average intake to “recommended daily allowance” is only 16 per cent for
Vitamin A, 35 per cent for iron and 45 per cent for calcium (NNMB data presented in Gopaldas,
2006). While there is some debate as to how these deficiencies are best addressed, including
lively controversies about the merits of various forms of “micronutrient supplementation”, what
is not in doubt is that large sections of the Indian population have diets that are missing much
that they need.
Similar concerns apply to the intake of fats (Dorin, 1999). In rich countries, and among
affluent sections of the population in poor countries (including India), fat has acquired a bad
name as a harbinger of diabetes, cardio-vascular diseases and other illnesses. Yet, fat intake
among large sections of the Indian population is almost certainly too low, possibly much too
low. This applies in particular to children in poor households. Fat is important for children not
only because fat is a calorie-dense food and helps them to achieve adequate calorie intake despite
small stomachs, and generally facilitates the absorption of various nutrients, but also because fat
is useful in its own right, e.g. for the development of the brain (Uauy et al, 2001). Current expert
recommendations regarding optimal fat intake as a proportion of calorie intake for young
children appear to be in the range of 30 to 45 per cent (World Health Organization, 2005, p. 16).
In India, however, fat accounts for barely 15 per cent of average calorie intake, according to
National Sample Survey data, with much lower figures among poor households. According to
the National Nutrition Monitoring Bureau, fat intake among Indian children is only about 30 per
cent of the “recommended daily allowance” (National Nutrition Monitoring Bureau, 2006, p.
14).
61
Overcoming these massive deficiencies would require a substantial shift from cereal-based
diets to more diversified diets. Indeed, it is virtually impossible to meet these diverse
requirements from a cereal-dominated diet. Even calorie requirements are difficult to meet from
cereals alone, especially for children, due to absorption problems. Sometimes, cereal
consumption even has an adverse effect on the consumption of essential nutrients. For instance,
cereals, legumes, peas and nuts contain phosphorus compounds known as phytates which inhibit
the absorption of iron (certain compounds found in tea have similar effects). This inhibition of
iron absorption is one reason why iron-deficiency anaemia is so widespread in India, Baynes and
Bothwell (1990).
In this connection, there is also evidence from international data, suggesting that the
diversification of diets is associated with better child nutrition (Arimond and Ruel, 2004). While
the shift from cereal-based diets to more diversified diets is already happening to some extent in
India (Kumar et al, 2007), Indian diets remain very frugal. To illustrate, according to NFHS-2
data for 1998-99, only 55 per cent of adult women in India consume milk or curd at least once a
week, only 33 per cent eat fruits at least once a week, and 28 per cent get an egg (International
Institute for Population Sciences, 2000, p. 242). Seven years later, the NFHS-3 survey yielded
the same figure for “milk or curd” (55 per cent), and only slightly higher figures for fruits and
eggS—40 and 32 per cent respectively (International Institute for Population Sciences, 2007a, p.
299). It is hard to see how mass undernourishment could be eliminated in India without a major
improvement in the quantity, quality and diversity of food intake among poorer sections of the
population.
62
Nutrition Status ofPrivileged Indian Children
So far we have proceeded on the assumption that international anthropometric standards are
“applicable” to India. As the NFHS-3 report puts it:
“Evaluation of nutritional status is based on the rationale that in a well-nourished population, there is a
statistically predictable distribution of children of a given age with respect to height and weight... The use
of a reference population is based on the empirical finding that well-nourished children in all population
groups for which data exist follow very similar growth patterns before puberty.” (International Institute
for Population Sciences, 2007a, page 268).
The puzzlingly high levels of child undemutrition in India (and South Asia) point to the possible
need to reexamine this assumption. Consider for instance Kerala. According to NFHS-3, 23 per
cent of children below the age of 3 years in Kerala are underweight, 16 per cent are wasted and
25 per cent are stunted, International Institute for Population Sciences (2007a, p. 273). However,
other indicators of child health and well-being suggest that children in Kerala are doing quite
well. For instance, the infant mortality rate is only 13 per 1,000—about the same as in, say,
Kuwait, Costa Rica or Malaysia. Generally, child development indicators (other than
anthropometric measurements) are much better in Kerala than in countries with similar
proportions of underweight children. For instance, the proportion of underweight children in
Kerala is not very different from the average for sub-Saharan Africa, but the infant mortality rate
in sub-Saharan Africa (around 100 per 1,000) is nearly eight times as high as in Kerala. It is, of
course, possible that children in Kerala would be doing even better in some respects if they were
not “held up” by low weights and heights. But it is not clear, from available data, what these
impairments actually are.
A related puzzle is that child undemutrition is also quite high (according to the
anthropometric approach) among well-off households. Here, two different bodies of evidence
63
need to be considered. On the one hand, some studies (e.g. Agarwal et al 1987, 1991) suggest
that the anthropometric achievements of children in affluent Indian families are much the same
as those of well-nourished children elsewhere, and indeed these Indian children are included in
the international reference population that forms the basis of the most recent WHO “child growth
standards” (World Health Organization, 2008). On the other hand, the NFHS data tell a different
story, whereby a substantial proportion of Indian children are undernourished even among welloff households. To illustrate, Table 16 shows standard child nutrition indicators for different
economic classes, defined in terms of the “wealth index” presented in the NFHS-3 report.
13 A
As
this table makes clear, child undernourishment declines sharply with increasing values of the
index, as one might expect. However, even among the top 20 per cent of households, substantial
proportions of children are undernourished: 20 per cent of children are underweight, 25 per cent
are stunted, and 13 per cent are wasted (again, much the same figures as the sub-Saharan African
averages).
Table 16: Child Nutrition and “Wealth”
“Wealth index”
Proportion (%) of undernourished children3
Weight-for-height
Height-for-age
Weight-for-age
25 (-1.2)
60 (-2.3)
57 (-2.2)
Lowest
22 (-1.1)
54 (-2.1)
49 (-2.0)
Second
19 (-1.0)
49 (-1.9)
41 (-1.8)
Third
17 (-0.9)
41 (-1.6)
34 (-1.5)
Fourth
13 (-0.7)
25 (-1.1)
20 (-1.1)
Highest
19.8 (-1.0)
48.0 (-1.9)
42.5 (-1.8)
All categories
a “Mean z-scores” in brackets (expressed in terms of standard deviations).
Source: International Institute for Population Sciences (2007a), page 271. All figures pertain to children below the
age of five years (note that the corresponding figures in Table 11 apply to children below the age of three years).
13 This index is constructed through “principal components analysis” from a list of 33 household assets
and bousing characteristics (such as ownership of various durables, access to electricity, and type of
cooking fuel). For further discussion, see International Institute for Population Sciences (2007a), pp. 4345.
64
Also presented in this table are the “mean z-scores” of children in different economic classes.
If children from wealthy households had the same “growth curves” as children from the
international reference population, their mean z-scores would be zero. In fact, the mean z-scores
are negative throughout the wealth scale.
It is perhaps not surprising that wealth alone provides inadequate protection against stunting
or wasting. Following on this, Tarozzi (2007) used NFHS-2 data to investigate the nutrition
status of highly privileged Indian children, where privilege is defined not only in terms of wealth
but also involves urban residence, piped water, a flush toilet, and both parents having a high
school degree. A little over 400 households in the NFHS-2 sample fall in this category (say
“privileged households” for short). For this group of privileged households, an interesting
pattern emerges: the mean z-score for “height-for-age” is still negative (-0.88), but the mean z-
score for “weight-for-height” is very close to zero (0.105), Tarozzi (2007, Table 4). In other
words, privileged Indian children are somewhat stunted, but they are not wasted.
The absence of any wasting among privileged Indian children is consistent with the plausible
notion that these children have an adequate food intake. The question remains as to why they are
stunted, in spite of the absence of any significant food deprivation, and presumably also of other
basic deprivations (e.g. lack of sanitation), given the privileged environment in which these
children live. Three possible (and not mutually exclusive) hypotheses are as follows.
•
“Social determinants " hypothesis'. Stunting among privileged children reflects social
factors such as a poor epidemiological environment, inadequate social support, and
inappropriate social norms relating (say) to breastfeeding or child feeding.
65
“Geneticpotential’’ hypothesis: Indian children do not have the same genetic potential as
children in the international reference population—they are “naturally” shorter, even
when they are well nourished.
•
"Gradual catch-up ” hypothesis: Indian children have the same genetic potential as
children in the reference population, but it takes time for the heights of privileged
children to catch up with the genetic potential, given the history of undemutrition.
The social determinants hypothesis is consistent with the claim that the international
anthropometric standards are applicable in India. It implies that privileged children in India are
often undernourished, though not due to lack of food. For instance, Bhandari et al (2002) found
that, in a sample of well-off families in South Delhi, only 61 per cent of infants were breastfed
within 12 hours of birth; only 25 per cent were exclusively breastfed up to the age of four
months; almost two-thirds stopped breastfeeding before the age of one year; and only 2 per cent
were fed non-vegetarian foods within a year (even though 58 per cent of the sample families
were non-vegetarian). These observations indicate that, even in privileged families, breastfeeding
and weaning practices are often at variance with current recommendations, such as breastfeeding
within an hour of birth and exclusive breastfeeding for six months (Gupta, 2006, Government of
India, 2006). It is quite possible that the same children would have benefited from more effective
breastfeeding and weaning in a more conducive social environment, e.g. a society where there is
better awareness of these matters and where breastfeeding counseling services (or, say, facilities
for breastfeeding at the work place) are available. Having said this, if the social determinants
hypothesis is correct, one would expect it to be reflected in wasting as well as in stunting, and
not just in stunting. Yet, as we have seen, there is little wasting among privileged Indian
children.
66
The genetic potential hypothesis, although certainly not disproved, is becoming less accepted
in the scientific literature, if only because there is a long history of differences in population
heights that were presumed to be genetic, and that vanished in the face of improved nutrition.
The “gradual catch-up” hypothesis has some plausibility, if only because (1) nutritional
status is highly influenced by birthweight, and (2) birthweight, in turn, is highly correlated with
mother’s weight and height. This observation, which is at the root of the notion of
intergenerational perpetuation of undemutrition, suggests that it would indeed take time for well-
fed children to overcome the burden of undemutrition in the past. The catch-up hypothesis is also
consistent with recent evidence about the secular increase in heights in Europe (Cole, 2003).
This hypothesis raises the further question whether well-nourished children suffer from any
impairments during the catch-up period, due to being relatively short. The “small but healthy”
debate suggests an affirmative answer to this question (Dasgupta and Ray, 1990). A number of
developed country studies also suggest that failure to reach one’s genetic potential as a child is
correlated with a range of negative outcomes throughout the life course. This applies whether
this failure is due to low birthweight, or to poor nutrition after birth, and this suggests that
stunted children do suffer from significant impairments even if they are well fed. The possibility
remains that the impairments experienced by these short but well-fed children (e.g. those
belonging to well-off families in Kerala) are less serious than those of children who suffer from
inadequate nutrition as well as low birthweights (e.g. stunted children in Bihar).
The gradual catch-up hypothesis, if correct, would not invalidate the use of international
reference standards to assess the nutritional status of Indian children. Indeed, it is appropriate for
reference standards to be based on children who are not only well fed, but also have adequate
birth-weights and healthy mothers (so that they are free from any burden of past undemutrition).
67
However, this hypothesis might modify our interpretation of the data, depending on the nature of
the impairments associated with stunting during the period when population height levels are
rising towards the international norm. It would also make it a little easier to understand why
anthropometric indicators appear to improve rather slowly at times (without necessarily implying
that the consequences of slow improvement are any less serious), even when there is not only
substantial poverty reduction but also major improvements in other determinants of child
nutrition, such as maternal education and access to safe water. The gradual catch-up hypothesis
may also be of some help in explaining why South Asians are so short: this is what one would
expect if the region has a particularly long history of mass undernourishment, and if it takes time
(perhaps many generations) to overcome this historical burden.
Having said this, the gradual catch-up hypothesis is of no help in explaining the recent
“stalling” of weight improvement among Indian children, if such a stalling has actually taken
place (as we saw, there is room for doubt on this). Indeed, if children’s weights were improving
steadily in the eighties and nineties, it is not clear why they should have stopped improving
between NFHS-2 and NFHS-3. This is, as things stand, an unresolved puzzle.
Concluding Remarks
In this paper, we have examined recent evidence on nutrition in India, and discussed possible
interpretations of the facts. The main findings are as follows.
One, there is strong evidence (not only from the National Sample Survey but also from
NNMB surveys) of a sustained decline in per-capita calorie consumption during the last twentyfive years or so. According to NSS data, average calorie consumption in rural areas was about 10
per cent lower in 2004-5 than in 1983. The proportionate decline was larger among better-off
68
sections of the population, and close to zero for the bottom quartile of the per-capita expenditure
scale. In urban areas, there was little change in average calorie consumption over this period.
Two, the decline of per-capita consumption is not limited to calories. It also applies to
proteins and many other nutrients, the major exception being fat consumption, which has
increased steadily (in both rural and urban areas) during this period.
Three, it would be difficult to attribute the decline in calorie consumption to declining per-
capita incomes, or to changes in relative prices. Indeed, the evidence points to rising per-capita
incomes (especially—but not exclusively—among the better-off), with little change in the price of
food - or calories - relative to other commodities. Thus, the main point is that calorie
consumption is lower today at a given level of per-capita household expenditure, and this applies
across the expenditure scale, at low levels of per capita expenditure as well as high. In other
words, the decline in calorie consumption in rural areas is associated with a steady downward
drift of calorie Engel curves, the plots of per capita calorie consumption against per capita total
expenditure.
Four, one possible explanation for this drift is that calorie requirements have declined, due to
better health as well as to lower activity levels. There are fragments of evidence to support this
hypothesis, such as major expansions in the availability of safe water, vaccination rates, transport
facilities, and the ownership of various effort-saving durables. Given that calorie requirements
rise sharply with activity levels, fairly moderate reductions in activity levels (reductions that
appear to be well within the realm of plausibility) would go a long way to explain the 10 per cent
reduction in average calorie intake since 1983. However, this hypothesis remains somewhat
speculative, in the absence of direct evidence on activity levels and the associated calorie
requirements.
69
Five, we have argued that average calorie intake has serious limitations as a nutrition
indicator. While calorie deficiency is an extremely important aspect of nutritional deprivation,
there are too many sources of variation in calorie requirements for standard, time-invariant
“calorie norms” to be usefully applied to large sections of the population. Further, close attention
needs to be paid to other aspects of food deprivation, such as the intake of vitamins and minerals,
fat consumption, the diversity of the diet, and breastfeeding practices. This point is relatively
well understood among nutritionists, but is yet to be adequately taken on board in economic
analyses of nutrition issues in India.
Six, because there is considerable uncertainty about the causes of the calorie decline, it is
difficult to assess its welfare implications. It is likely true that faster wage growth among the
poor would have moderated or reversed the calorie decline. But some of the calorie decline may
come from a better health environment or a reduction in the burden of hard labor, each of which
is a positive development in its own right.
Seven, the limitations of intake-focused nutritional assessments reinforce the case for
supplementing intake data with outcome-focused indicators, such as anthropometric
measurements. However, anthropometric data have some limitations too. For one thing, there are
unresolved puzzles about anthropometric indicators in India, such as the high prevalence of
stunting among privileged children (which is seemingly at variance with the premises of
accepted “growth standards”). For another, there are inconsistencies between different sources
of anthropometric data (specifically, the National Family Health Survey and the National
Nutrition Monitoring Bureau), as well as puzzling contrasts between nutrition trends based on
different anthropometric indicators, such as height-for-age and weight-for-height. While broad,
long-term trends are reasonably clear, there is some confusion about recent changes.
70
Eight, some of the available anthropometric evidence, particularly that from the National
Family Health Survey, raises troubling questions about the speed of nutritional improvement in
India. According to the most recent NFHS data, there has been virtually no change in the
proportion of underweight children between 1998-9 and 2005-6 (close to 50 per cent in both
years), and no change in the proportion of adult women with low BMI (also close to 50 per cent).
These undemutrition figures are among the highest in the world, making it all the more troubling
that so little progress should have been made during a period of rapid economic growth. Data
from the NFHS, also suggest that the adult height of Indians—an indicator of nutrition and
disease in childhood and adolescence—has improved more slowly than has been the case in
other countries, even other countries with similar deficits. Indian women have done particularly
poorly. These height data indicate slow progress in nutritional status over a long period—from
the late 1950s Complicating things further, and as discussed in the text, the overall anthro
pometric evidence on recent nutrition trends for children is far from clear: NNMB data shed a
different light on these trends, and even the picture emerging from NFHS data is actually quite
mixed, with, for instance, a substantial decline in stunting rates (among children) between
NFHS-2 and NFHS-3. Clarification of this is an important area of further research and an urgent
priority for data improvement.
Among other messages emerging from this enquiry is the urgent need for better nutrition
monitoring arrangements in India. There are serious gaps in India’s nutrition statistics, and even
the most basic nutrition trends are far from clear. The NNMB surveys are not particularly
informative, given their small sample sizes, limited geographical coverage, obscure
methodology, and uncertain quality. The NFHS surveys, for their part, are conducted at wide
intervals (six to seven years), and the reports focus mainly on state-level indicators. Effective
71
action in this field requires regular and reliable large-scale surveys that would make it possible to
monitor the nutrition situation at the district level at intervals of, say, two to three years at most.
Revamping the District Level Health Surveys (DLHS), and conducting NFHS surveys at more
regular intervals, are two possible steps in this direction. Another useful option would be to
“link” these surveys with NS SO surveys from time to time, making it possible to combine
detailed nutrition data with household expenditure data.
This examination of recent evidence on nutrition in India is far from conclusive. Indeed, it
makes clear that the nutrition situation in India is full of “puzzles”. We hope that the puzzles, if
not resolved, are at least clearer.
References
Alderman, Harold, 2005, “Linkages between poverty reduction strategies and child nutrition: an
Asian perspective”, Economic and Political Weekly, 40 (46), 4837—42.
Agarwal, D.K., Upadhyay, S.K., Tripathi, A.M., Agarwal, K.N., 1987, “Nutritional status,
physical work capacity and mental function in school children”, Report 6, Nutrition Foundation
of India, New Delhi.
Agarwal, K.N., Agarwal, D.K., Benakappa, D.G., Gupta, S.M., Khanduja, P.C., Khatua, S.P.,
Ramachandran, K., Udani, P.M. and Gopalan, C., 1991, “Growth performance of affluent Indian
children (under-fives): growth standard for Indian children”, Report 11, Nutrition Foundation of
India, New Delhi.
Baynes, R.D., and Bothwell, T.H., 1990, “Iron deficiency”, Annual Review ofNutrition, 10, 133—
48.
Bhandari, N„ Bahl, R„ Taneja, S„ de Onis, M., and Bhan, M.K., 2002, “Growth performance of
affluent Indian children is similar to that in developed countries”, Bulletin of the World Health
Organization, 80 (3), 189-95.
Bouis, Howard and Laurence J. Haddad, 1985, “Are estimates of calorie-income elasticities too
high? A recalibration of the plausible range,” Journal ofDevelopment Economics, 39, 333-64.
Bozzoli, Carlos, Angus Deaton, and Climent Quintana-Domeque, 2007, "Child mortality,
income, and adult height," Princeton University, Research Program in Development Studies,
March.
72
Clark, Gregory, Michael Huberman, and Peter Lindert, 1995, “A British food puzzle, 17701850,” The Economic History Review, 48(2), 215-37.
Cole, Tim J., 2003, “The secular trend in human physical growth: a biological view”, Economics
and Human Biology, 1, 161-8.
Das Gupta, Monica, 1985, “Informal security systems and population retention in rural India”,
NCAER Working Paper No.2, National Council of Applied Economic Research, New Delhi.
Dasgupta, Partha and Debraj Ray, 1990, “Adapting to undernourishment: the biological evidence
and its implications”, in Dreze, J.P., and Sen, A.K. (eds.), 1990, The Political Economy of
Hunger Vol. I (Oxford: Clarendon).
Deaton, Angus and Jean Dreze, 2002, “Poverty and inequality in India: a reexamination,”
Economic and Political Weekly, September 7th, 3729—48
Deaton, Angus, 2008, “Price trends in India and their implications for measuring
poverty ^Economic and Political Weekly, February 9th, 2008, 43—9.
Deaton, Angus, and Valerie Kozel, 2005a, “Data and dogma: the great Indian poverty debate,”
World Bank Research Observer, 20(2), 177-99.
Deaton, Angus, and Valerie Kozel, eds., 2005b, The great Indian poverty debate, New Delhi.
Macmillan.
Dorin, Bruno, 1999, “Food policy and nutritional security: the unequal access to lipids in India,”
Economic and Political Weekly, 34(26), June, 1709-17.
Dreze, Jean, 2006, “Universalization with quality: ICDS in a rights perspective”, Economic and
Political Weekly, 26 August.
Dreze, Jean, Reetika Khera, and S. Narayanan, 2008, “Early childhood in India: facing the
facts”, Indian Journal ofHuman Development, forthcoming.
Dreze, Jean and Amartya Sen, 1989, Hunger and Public Action Oxford. Clarendon.
Dreze, Jean and Amartya Sen, 2002, India: Development andparticipation f New Delhi. Oxford
University Press.
Du, S., B. Lu, F. Zhai, and B. M. Popkin, 2002, “A new stage of the nutrition transition in
China”, Public Health Nutrition, 5(1 A), 169-74..
Flood, Roderick, Kenneth Wachter, and Annabel Gregory, 1990, Height, health, and history,
Cambridge. Cambridge University Press.
73
Filmer, Deon, and Lant H. Prtichett, 2001, “Estimating wealth effects without expenditure
data—or tears: an application to educational enrollments in states of India,” Demography, 38(1),
115-32.
Gopaldas, Tara, 2006, “Hidden hunger: the problem and possible interventions”, Economic and
Political Weekly, 41(34), 26 August, 3671-4.
Gupta, Arun, 2006, “Infant and young child feeding: an optimal approach”, Economic and
Political Weekly, 41(34), 26 August, 3666—71.
Haddad, Lawrence, Harold Aiderman, Simon Appleton, Lina Song, and Yisehac Yohannes,
2003, “Reducing child malnutrition: how far does income growth take us?”, World Bank
Economic Review, 17(1), 107-31.
Harris, Bernard (2000), “Height and nutrition”, in Kiple, K.F. and Ornelas, K.C. (eds),
Cambridge World History ofFood, Vol 2, Cambridge: Cambridge University Press, 1427-38.
Himanshu, 2007, “Recent trends in poverty and inequality: some preliminary results,” Economic
and Political Weekly, 10 February, 497-508.
International Institute for Population Sciences, 2000, National Family Health Survey 1998-99
(NFHS-2): India, Mumbai, UPS.
International Institute for Population Sciences, 2007a, National Family Health Survey 2005-06
(NFHS-3): India, Mumbai, UPS.
International Institute for Population Sciences, 2007b, 2005-2006 National Family Health Survey
(NFHS-3): National Fact Sheet, Mumbai, UPS. (Also available at www.nfhsindia.org.)
International Institute for Population Sciences, 2007c, State-wise Fact Sheets based on 20052006 National Family Health Survey (NFHS-3), Mumbai, UPS. (Also available at
www.nfhsindia.org.)
Jodha, N.S., 1986, “Common property resources and rural poor in dry regions of India”,
Economic and Political Weekly, 21(27), 5 July, 1169—81.
Kumar, Praduman, Mruthyunjaya, and Madan M. Dey, 2007, “Long-term changes in Indian food
basket and nutrition,” Economic and Political Weekly, 42(35), September 1 , 3567—72.
Mari Bhat, P. N., 2002, “On the trail of ‘missing’ Indian females I: search for clues,” Economic
and Political Weekly, 37(51), December 21, 5105-5118.
Meenakshi, J. V. and Brinda Viswanathan, 2005, “Calorie deprivation in rural India between
1983 and 1999/2000,” in Angus Deaton and Valerie Kozel, eds.. The great Indian poverty
debate. New Delhi. Macmillan, 570-83.
74
Minhas, B. S., 1988, “Validation of large scale sample survey data: the case of NSS estimates of
household consumption expenditure,” Sankhya, The Indian Journal of Statistics, 50, Series B,
Part 3, Suppl., 1-63.
National Nutrition Monitoring Bureau, 1991, Report ofRepeat Surveys (1988-90), Hyderabad,
National Institute of Nutrition.
National Nutrition Monitoring Bureau, 1996a, Nutritional Status ofRural Population: Report of
NNMB Surveys, Hyderabad, National Institute of Nutrition.
National Nutrition Monitoring Bureau, 1996b, Report ofNNMB Surveys: Additional
Tabulations, Hyderabad, National Institute of Nutrition.
National Nutrition Monitoring Bureau, 1997, 25 Years ofNational Nutrition Monitoring Bureau,
Hyderabad, National Institute of Nutrition.
National Nutrition Monitoring Bureau, 1999, Report ofSecond Repeat Survey - Rural (1996-7),
Hyderabad, National Nutrition of Nutrition.
National Nutrition Monitoring Bureau, 2000, Report on Food and Nutrient Intake ofIndividuals
(Rural), Hyderabad, National Institute of Nutrition.
National Nutrition Monitoring Bureau, 2002, Diet and Nutritional Status ofRural Population,
Hyderabad, National Institute of Nutrition.
National Nutrition Monitoring Bureau, 2005, Assessment ofDiet and Nutritional Status of
Individuals and Prevalence ofHypertension in Adults & Anemia Among Men and NPNL Women
in Rural Communities, Hyderabad, National Institute of Nutrition.
National Nutrition Monitoring Bureau, 2006, Diet and Nutritional Status ofPopulation and
Prevalence ofHypertension among Adults in Rural Areas, Hyderabad: National Institute of
Nutrition.
National Nutrition Monitoring Bureau. 1990?, Report ofNNMB-NSSO Linked Survey,
Hyderabad, National Institute of Nutrition.
National Nutrition Monitoring Bureau, 2006, Assessment ofDiet and Nutritional Status of
Individuals and Prevalence ofHypertension in Adults & Anaemia among Men and NPNL
Women in Rural Communities, Hyderabad, National Institute of Nutrition.
National Sample Survey Organization, 2001, Nutritional intake in India 1999-2000, Ministry of
Statistics and Program Implementation, Government of India, August.
Palmer-Jones, Richard, and Kunal Sen, 2001, “On Indian poverty puzzles and statistics of
poverty,” Economic and Political Weekly, 36(3), January 20, 211—7.
75
Patnaik, Utsa, 2004, “The republic of hunger,” Social Scientist, 32(9-10), Sep-Oct, 9-35.
Patnaik Utsa, 2007, “Neoliberalism and rural poverty in India,” Economic and Political Weekly,
42(30), July 28, 3132-50.
Piazza, Alan, 1986, Food Consumption and Nutritional Status in the People’s Republic of China,
Boulder, Colo. Westview.
Radhakrishna, R., 2005, “Food and nutrition security of the poor: emerging perspectives and
policy issues,” Economic and Political Weekly, 40(18), April 30, 1817-21.
Radhakrishna, R., K. Hanumantha Rao, C. Ravi, and B. Sambi Reddy. 2004, “Chronic poverty
and malnutrition in 1990s,” Economic and Political Weekly, 39(28), July 10, 3121-30.
Ramachandran, K., 2006, “Nutrition monitoring and surveillance”, available at
http://nutritionfoundationofmdia.res.in.
Rao, C. H. Hannmantha, 2000, “Declining demand for foodgrains in rural India: causes and
implications,” Economic and Political Weekly, 35(4), January 22, 201—6.
Rao, C. H. Hanumantha, 2005, Agriculture, food security, poverty and environment: essays on
post-reform India, New Delhi. Oxford University Press.
Rao, Shobha, 2001, “Nutritional status of the Indian population”, Journal ofBiosciences, 26(4),
Suppl., 481-9.
Ray, Ranjan and Geoffrey Lancaster, 2005, “On setting the poverty line based on estimated
nutrient prices: condition of socially disadvantaged groups during the reform period,” Economic
and Political Weekly, 40(1), January 1,46-56.
Sankar, Deepa, 2007, “What is the progress in elementary education in India during the last two
decades? an analysis using NSS education rounds”, mimeo, World Bank.
Sen, Pronab, 2005, “Of calories and things: reflections on nutritional norms, poverty lines and
consumption behaviour in India,” Economic and Political Weekly, 40(43), October 22, 4611-8.
Scrimshaw, Nevin S., Oscar Brunser, Gerald Keusch, Ayesha Molla, Imran Ozalp, and Benjamin
Tonin, 1983, “Diarrhea and nutrient requirements,” Chapter 16 in Lincoln C. Chen and Nevin S.
Scrimshaw, eds., Diarrhea and Malnutrition: Interactions, mechanisms and interventions,
Geneva. United Nations University Press, pp. 269-86.
Sharma, Rekha, 2006, “Quantifying Malnutrition in India”, PhD thesis, Delhi School of
Economics, Delhi.
Subramanian, A., 2003, “Consumption of food and calories: some implications for nutrition and
income,” New Delhi, National Council for Applied Economic Research, processed.
76
<
Tarozzi, Alessandro, 2005, “On the nutritional status of Indian children”, mimeo, Department of
Economics, Duke University.
Tarozzi, Alessandro, 2007, “Growth reference charts and the status of Indian children,” Duke
University, Department of Economics, processed. (August.) Available at
http://www.econ.duke.edu/~taroz/
Tarozzi, Alessandro, and Aprajit Mahajan, 2007, “Child Nutrition in India in the Nineties”,
Economic Development and Cultural Change, 55(3), 441-86.
Uauy, R., Hoffinan, D.R., Peirano, P., Birch, and Birch, E.E., 2001, “Essential fatty acids in
visual and brain development”, Lipids, 36(9), 2962S-72S..
United Nations Development Program, 2006, Human Development Report 2006, New York:
UNDP.
Venkaiah, K., Damayanti, K., Nayak, M.U., and Vijayaraghavan, K., 2002, “Diet and nutritional
status of rural adolescents in India”, European Journal of Clinical Nutrition, 56 (11), 1119-25.
Victora, Cesar G, Linda Adair, Caroline Fall, Pedro C. Hallal, Reynaldo Martorell, Linda Richter
and Harshpal Singh Sachdev, for the Maternal and Child Undemutrition Study Group, 2008,
“Maternal and child undemutrition: consequences for adult health and human capital,” The
Lancet, 371, 340-57.
World Health Organization, 2005, Guiding Principles for Feeding Non-Breasfed Children 6-24
Months ofAge, Geneva, World Health Organization.
World Health Organization, 2008, “The WHO Child Growth Standards”, available at
http://www.who.int/childgrowth/en/.
Yang, X.G., et al^ 2005, “Study on weight and height of the Chinese people and the differences
between 1992 and 2002”, Chinese Journal ofEpidemiology.26(7), 489-93.
77
Position: 1036 (5 views)