indian food puzzles: growth, poverty & (mal)nutrition angus deaton & jean drèze
TRANSCRIPT
Indian food puzzles: growth, poverty & (mal)nutrition
Angus Deaton & Jean Drèze
Lots of growth
Real GDP per capita growing at 3.6 percent a year since 1980 4.6 percent a year 00−04
Real per capita aggregate consumption growing at 2.0 percent in the 1980s, 2.6 percent in the 1990s, and 4.7 percent 00−04
Poverty reduction has been less than warranted by this growth rate if equally distributed Some increase in inequality Much more important are data inconsistencies Coverage differences, and outright discrepancies Survey consumption grows less rapidly than NAS
consumption Errors on both sides Inconsistent survey instruments from year to year
Growth across the distribution
NSS growth may be too low Generally some growth at all fractiles
Percentile 10th 25th 50th 75th 90th
Rural
1983–1993/41999/00–2004.52000/01–2004.5
1.81.41.3
1.41.20.4
1.21.10.3
1.01.60.7
0.82.21.9
Urban
1983–1993/41999/00–2004.52000/01–2004.5
1.2–0.6–0.0
1.1–0.4–0.1
1.30.10.3
1.40.60.4
1.31.91.4
All India
1983–1993/41999/00–2004.52000/01–2004.5
1.71.31.1
1.40.90.3
1.20.80.3
1.11.00.9
1.11.71.5
Table 1: Growth by percentiles
NSS data
But calorie consumption is falling
YEAR PCE (real) Calories Protein Fats
Rural Urban Rural Urban Rural Urban Rural Urban
19831987–81993–41999–02000–12001–22002(2)20032004(1)
384350555657585960
142.5157.7159.9179.3185.3182.2189.0192.4194.7
230.7245.7264.9306.4311.1301.8318.2314.4315.3
2,2402,2332,1532,1482,0832,0182,0252,1062,087
2,0702,0942,0732,1552,0271,9822,0142,0202,036
63.563.260.359.156.854.855.458.056.9
58.158.657.758.455.354.254.955.556.0
27.128.331.136.034.633.634.736.435.6
37.139.341.949.646.146.147.046.747.1
Is this really correct?Data from National Nutritional Monitoring Bureau
Rural (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 MadhyaPradesh and West Bengal. The 2004-05 figures exclude Gujarat..
Fall is clearest for cereals
Year Cereals All foods
Rural Urban All India Rural Urban All India
19831987–81993–41999–02000–12001–22002(2)20032004(1)
1,6811,6481,5331,4551,4221,3911,3811,4121,419
1,3031,2961,2311,2001,1611,1301,1371,1421,165
1,5961,5691,4581,3921,3571,3301,3181,3451,357
2,2402,2332,1532,1482,0832,0182,0252,1062,087
2,0702,0942,0732,1552,0271,9822,0142,0202,036
2,2012,2022,1332,1502,0692,0092.0222,0842,075
010
020
030
040
050
0
1950 1960 1970 1980 1990 2000
Cereal availability
Cereal + pulses availability
Changes in government stocks of cerealgms
per
capi
ta p
er d
ay
From Economic Survey of India
010
020
030
040
050
0
1950 1960 1970 1980 1990 2000year
Cereal availability
Cereal + pulses availability
Changes in government stocks of cerealgms
per
capi
ta p
er d
ay
From Economic Survey of India
NSS consumption
5010
015
020
025
0
1950 1960 1970 1980 1990 2000year
RICE
WHEATOTHER CEREALS
AVAILABILITY OF CEREALS, GM PER PERSON PER DAY
Ministry of Agriculture
What about Engel?
• Engel’s law says that the share of food in the budget falls as incomes rise– Says nothing about levels of food consumption
• Calorie Engel curves show calories (including cereal calories) rising with income (or at least pce) over a range
• So we would expect calorie consumption to rise as living standards improve– India is growing rapidly, but remains poor– Average per capita calorie consumption in the bottom
decile of pce in 1983 was less than 1,400 calories, and has been around 1,500 calories for last 20 years
Why are calories falling?
• One interpretation is that poverty and hunger are increasing, especially among rural households
• “Republic of hunger”• If poor people were getting better-off, they would
consume more calories, especially more cereals• So some combination of falling incomes, rising
prices, and unemployment must be impoverishing them
• Engel curves are correct, calorie data are correct, and NSS is overstating consumption levels
Rising poverty
• A weaker argument is based on calorie adequacy
• India’s poverty lines were originally set so that at the PL, households on average obtained 2,400 calories (rural) and 2,100 calories (urban)
• So we can calculate how many people are meeting these standards over time
Calorie poverty rates
Year Round Rural Urban All India
19831987–81993–41999–02000–12001–22002(2)20032004(1)
384350555657585960
66.165.971.074.176.279.879.678.478.5
60.557.158.158.260.764.263.063.361.5
64.863.967.870.172.376.175.374.674.3
Percentages of persons below recommended daily calorie allowances
Calculations from NSS data
But, but . . . .
• The purchasing power of the original poverty lines has not changed (up to possible errors in price indexes)
• Pronab Sen (EPW) has shown that, if people around the poverty line were to pay the average price per calorie paid by people below the poverty line, they would meet the calorie norms
• So they must be reducing calories because they want to, not because they have to.
• Reports of numbers of people not getting “two square meals a day” have fallen dramatically over the last 20 years
What about malnutrition• NFHS 3 from 2005−06 • NFHS 1 (92−93) and NFHS 2 (98−99) are available but did not
consistently measure all children across India• In the 90s, children who were underweight (z-score less than −2 for
weight for age) fell from 52 to 47 percent• Latest results for most states show little improvement overall• In the 90s, stunting improved in some states, worsened in others• Latest results show improvements in most states (18 out of 21:
worse in Arunachal Pradesh and Karnataka)• Wasting has got worse (all but 4 out of 21 states)• Much worse than might have been expected given growth and
poverty estimates• Better than expected if there is widespread and increasing hunger
02
04
06
00
20
40
60
02
04
06
00
20
40
60
02
04
06
0Andhra Pradesh Arunachal Pradesh Assam Chhattisgarh Delhi
Gujarat Haryana Himachal Pradesh Jammu & Kashmir Karnataka
Kerala Madhya Pradesh Maharashtra Manipur Meghalaya
Orissa Punjab Rajasthan Uttar Pradesh Uttaranchal
West Bengal
1992-93 1998-99
2005-06
Graphs by State
PREVALENCE OF STUNTINGAges 0-3
01
02
03
04
00
10
20
30
40
01
02
03
04
00
10
20
30
40
01
02
03
04
0Andhra Pradesh Arunachal Pradesh Assam Chhattisgarh Delhi
Gujarat Haryana Himachal Pradesh Jammu & Kashmir Karnataka
Kerala Madhya Pradesh Maharashtra Manipur Meghalaya
Orissa Punjab Rajasthan Uttar Pradesh Uttaranchal
West Bengal
mean of r1 mean of r2
mean of r3
Graphs by State
PREVALENCE OF WASTINGAges 0-3
02
04
06
00
20
40
60
02
04
06
00
20
40
60
02
04
06
0Andhra Pradesh Arunachal Pradesh Assam Chhattisgarh Delhi
Gujarat Haryana Himachal Pradesh Jammu & Kashmir Karnataka
Kerala Madhya Pradesh Maharashtra Manipur Meghalaya
Orissa Punjab Rajasthan Uttar Pradesh Uttaranchal
West Bengal
1992-93 1998-99
2005-06
Graphs by State
PREVALENCE OF UNDERWEIGHTAges 0-3
149.
515
015
0.5
151
151.
5m
ean
heig
ht
10 20 30 40 50Age
Indian women
are growing taller, though little progress for thoseborn between 1965 and 1975
148
149
150
151
152
mea
n he
ight
10 20 30 40 50Age
India
Nepal
Bangladesh
and not as rapidly as women inNepal and Bangladesh, thoughthey are taller to start with
145
150
155
160
165
170
6 7 8 9 10Log of real GDP per head in year of birth
Ave
rage
hei
ght
South Asia
Africa
Latin America & Caribbean
Europe
US
China
And they remain among the smallest women in the world
Central Asia
148
150
152
154
156
mea
n w
omen
’s h
eigh
t in
198
9/99
1500 2000 2500 3000
mean per capita calories in region in 1983
black is urbanblue is rural
INDIAN NSS REGIONS
Fats and calories
• Much has been made of “nutritional transition” in countries like India
• Replacement of cereals by fats (milk, edible oil, chicken) and “animal source” foods, as well as sugar
• Concerns about consequences for health, especially diabetes and CVD
• But Indian rural poor are desperately short of fat, and for them, the nutritional transition is a good thing.
.4.5
.6.7
.8
4 4.5 5 5.5 6 6.5
1983 1987-81993-4
RURAL FOOD SHARES
Fra
ctio
n of
the
bud
get
spen
t on
foo
d
Logarithm of household total per capita expenditure
“Thin” rounds 1994-98
1999-00
“Thin” rounds 2000-4
.3.4
.5.6
.7
4.5 5 5.5 6 6.5 7
1983
1987-8
1993-4
URBAN FOOD SHARES
Fra
ctio
n of
the
bud
get
spen
t on
foo
d
Logarithm of household total per capita expenditure
“Thin” rounds 1994-981999-00
“Thin” rounds 2000-4
7.2
7.4
7.6
7.8
8
4 4.5 5 5.5 6 6.5
R38
R43
R50R55
R56-60
RURAL INDIA: Nonparametric Calorie Engel Curves
Logarithm of household per capita expenditure
Log
per
cap
ita h
ou
seh
old
calo
ries
7.2
7.4
7.6
7.8
8
4 4.5 5 5.5 6
R38
R43
R50
R55
R56-60
URBAN INDIA: Nonparametric Calorie Engel Curves
Logarithm of household per capita expenditure
Log
per
cap
ita h
ou
seh
old
calo
ries
7.4
7.6
7.8
8
4 4.5 5 5.5 6
R50
R55
R50 with 365 days for low frequency items
Logarithm of household per capita expenditure
Log
per
cap
ita h
ou
seh
old
calo
ries
RURAL INDIA: Nonparametric Calorie Engel Curves
77.
27.
47.
6
4 4.5 5 5.5 6
38 43 50
55
56-60
Rural calories from cereals
logarithm of household total expenditure per capita
loga
rithm
of
per
capi
ta c
alor
ies
from
cer
eals
6.8
77.
27.
47.
6
4 5 6 7
RURAL
URBAN
Calories from cereals, rural and urban together
2.5
33.
54
4.5
4 5 6 7
Per capita fat consumption, ruraland urban, all rounds
logarithm of household total expenditure per capita
loga
rithm
of
per
capi
ta f
ats
The puzzle remains
Engel curves for nutrients have positive slopes Even for calories from cereals (rural households)
Yet calorie consumption (& certainly cereal consumption) is falling Even among the poorest rural households
What can be shifting the curve? Not prices
Perhaps the curves are misleading? Cereal consumption really falls with income
Activity patterns Urban population is more sedentary
But fall is within rural population too Rural population may need less energy
Changing occupational mix More mechanization Less fetching of water and firewood Better public health: clean water & immunization
But not clear that this helps Less work often yields relatively few calories Occupational mix (out of agriculture) is real, but too small
relative to effects in regression equations’ Especially if the slopes of the Engel curves are correct Bigger people need more calories Falling for children too, according to NNMB
Biased Engel curves
NNMB data show that better off households consume less cereal and fewer calories from cereals
NCAER data appear to show the same thing But both of these have weak income measures
Comparison of APL and BPL households in Rajasthan
Literature on bias in Engel curves based on “indirect” measurement of nutrients (like NSS) Measurement error in quantities induces positively
correlated measurement error in nutrient counts and in total expenditure, biasing slope towards unity
Some evidence on the bias
Aggregation up to states, regions, districts The importance of varying tastes
Instrumental variable estimates None perfect and at best suggestive Almost always other interpretations
Aggregation and sample splitting reduces the slope But measurement error biases
Perhaps evidence that the slope is too high But not convincing
Other reasons for bias
If NSS measures food OK, but progressively understates non-food, true Engel curves would be flatter But can’t explain why it is falling over time
The rich feed servants and guests, the poor get meals that are not recorded, so steepening the Engel curve Unlikely to be a large effect Engel curve should flatten over time, but not
shift down at the bottom
But maybe Engel curves are OK?
For the rural poor in agriculture, calories may not generate utility directly, but fuel for work
Those who are healthier and stronger eat more, especially cereal calories, to generate earnings
Causality is from calories to income Among non-manual workers, cereals are
unresponsive to income, or even inferior As real wages rise, people are less willing to do hard
physical labor, which moves calorie consumption down as incomes and consumption rises
Which reconciles the Engel curve evidence