CONSTRUCTING THE POVERTY LINE A New Approach with Applications to Karnataka

Item

Title: CONSTRUCTING THE POVERTY LINE
A New Approach with Applications to Karnataka
extracted text: Analytical Studies, of Karnataka's Economy: 1

CONSTRUCTING THE POVERTY LINE
A New Approach with Applications to Karnataka

P. J. Nayak and G. Sumithra

MAY 1985

Analytical Studies of Karnataka's Economy: 1

CONSTRUCTING THE POVERTY LINE

A New Approach with Applications to Karnataka

P.J. NAYAKand

G.SUMITHRA*

May 1985

»

The authors work in the Karnataka Agro Industries Corporation and

the Planning Department, Government of Karnataka respectively.

The

views expressed are our own and not necessarily those of either of these
organisations.

We are grateful to the Karnataka Government Computer

Centre, and in particular to Shri Appa

Reddy and Shri R. Shankara,

for programming support in obtaining solutions to the computer aided

algorithm

proposed in this paper.

We would also like to thank Smt

Lakshmi for helpful assistance during data verification.

ABSTRACT

The paper presents a new approach to the construction of
the poverty line which examines more directly the existence
of malnutrition in households.
Jettisoning the assumption of
a monotonic relationship between per capita expenditure and
the consumption of nutrients, the paper proposes a new
definition of the poverty line.
This is generalised to the
notionof a poverty band, which assists in examining certain
relationships between nutrition and per capita expenditure.
Whereas earlier poverty line studies of the Indian economy
have defined nutritional norms solely in terms of minimal
calorie intake, this paper is sensitive also to the need for
a minimal protein intake.
The complication this causes on
account of the manner in which protein is either retained or
oxidised in the human body at low levels of nutrition, is
allowed for, and an algorithm is designed to solve for the
poverty line.

This approach is applied to 1983 NSS data for Karnataka.
The poverty line is computed at Rs.114 for rural areas and
Rs.177 for urban areas.
The proportion of the poor is 67% in
rural areas, 82% in urban areas and 73% in all Karnataka.
The possibility of estimational biases in the analytical
procedure adopted is discussed, particularly in view of the
high proportion of the urban poor.

1.Introduction and Summary
The construction of a poverty line and a suitable index
for measuring poverty are familiar issues raised in existing
economic approaches to poverty.
Recent work in the Welfare
Economics of Poverty, motivated particularly by the work of
Sen(1976), has
proposed
various indices for the measurement
of poverty, and central to this approach is the prior
specification of the poverty line.
Despite the considerable
effort put in by nutrition experts in developing subsistence
norms of individual nutrition intake and by economists in
translating these norms into expenditure equivalents, the
conceptualisation of the poverty line itself, however,
appears to have received insufficient attention and does not
rest on adequate welfare... theoretic foundations.
In the
absence of such underpinnings, the comparison of poverty
levels either over time or between regions becomes hazardous,
and the periodic generation of estimates of the size of the
poor is surely suspect.
Although existing studies of poverty line computations
in India pivot strongly round nutritional considerations, the
c o rresponden c e between individual consumption expenditures
and nutritional intake is a tenuous one, given the widespread
ignorance about the nutritional content of different food
items aswell as the pervasiveness of traditional dietary
habits.
People generally do not optimise on nutrient intake
within their budget constraints.
In this paper we recognise
the absence of such a straightforward correspondence and use
it to propose a new definition of the poverty line.
We also
investigate, within a generalised framework, how precisely
one may characterise malnourishment in a household, and use
this to identify malnourished households in Karnataka using
1983 NSS data.
The empirical work is carried out within a
more specialised framework wherein we allow for two
nutritional determinants, proteins and calories.
Such a
specialised framework is, nevertheless, more general than the
nutritional models which underpin most existing Indian
studies of the poverty line, and spawn two complications
which we are sensitive to : the problem of the dual role of
protein, and the differing ways in which individual
malnourishment
can
be
aggregated to
household
malnourishment.We then devise an algorithm to solve for the
poverty line, using 1983 NSS data for Karnataka.

The poverty line is computed at Rs. 114 for rural areas
and Rs.177 for urban areas.
The proportion of the poor is
6 7% in rural areas, 82% in urban areas and 73% in all
Karnataka.

The paper is organised as follows:
In Section 2 we
define our framework of analysis and provide a definition of
the poverty line.
This definition is generalised to the
notion of a poverty band, which assists in providing further

1

intuitive support to the usefulness of our definition of the
poverty line.
The manner in which protein is either
retained or oxidised in the human body at low levels of
nutrition is also grafted onto our framework.
In Section 3
we discuss earlier approaches to the problem and suggest that
these can be subsumed under 3 broad groups.
Section 4
discusses
the
problem
of
aggregating
individual
malnourishment into household malnourishment, and offers a
solution algorithm.
Section 5 presents applications to
Karnataka while Section 6 offers certain concluding caveats.

2.The Construction of a

Poverty Line

Although there are differing conceptions of what
constitutes a condition of poverty, and although in some
societies poverty may be more appropriately viewed as a
relative phenomenon — as an aspect ofa
skewedincome
distribution— nevertheless for a developing economy the most
useful approach to poverty does appear to lie in
characterising it as an absolute condition arising from a
lack of desirable minimal subsistence.
The usefulness of
viewing poverty as an essentially physical phenomenon arises
as much from the need to focus policy attention where
physical deprivation and consequent human agony is most
severe, as from noting the stark consequences that emerge
from the lack of subsistence nutrition to the physical growth
of the human body.
Economic approaches to the construction
of the poverty line in India have therefore necessarily begun
with minimal nutritional norms for subsistence as suggested
by nutritional experts (such as those generated by the Indian
Council of Medical Research (1968).)
It must be recognised,
of course, that subsistence is not an unambiguous concept and
refers to a condition where a state of reasonable health can
be maintained. Subjective interpretations of these terms are
inevitable and nutritional studies have differred on adequate
norms of minimal subsistence nutrition.
Nevertheless,
"Economists can do no better than respect their judgements,
imperfect though admittedly they are. What they should guard
against, however, is the mistake of employing a particular
norm in every situation, regardless of its specific context.
It is obvious that minimum nutritional requirements may vary
from country to country depending on the ethnic and climatic
factors" (Osmani (1982)).
In this paper we begin, therefore, by adopting a minimum
subsistence nutrition basket which has emerged from
nutritional studies.
We regard an individual as being
malnourished if his actual
nutritional
intake shows a
deficiency in any one nutrient when compared to the
subsistence nutrition basket.
We then construct the poverty
line such that (in an intuitive sense) the number of
malnourished people equals the number of people whose
individual consumption expenditure does not exceed the
poverty line.
We now formalise this definition.

2

Assume an economy with in individuals indexed by i
(i = 1 ,...., m ), k food items indexed by h (h = 1
,k) and 1
nutrients indexed by j (j = l,...,l).
Let individual i have a
personal.consumption expenditure of e1 and consume a food
vector f1 = (f j ,.... f 1 ).
Let N be an 1 x k matrix
which
we call the f ood-*nut r 11 i on transformation matrix, the (j,h)
component o f which indicates the extent of nutrient j
available in a unit o.f food item h.
It is then clear that
the matrix productN.f1 represents the 1.vector of nutrients
actually consumed by individual i through his food vector f1.
Let M1 be the set of nutrition vectors in Euclidean 1-space
representing malnourishment for individual i.
Clearly, the
i th individual is malnourished if N.f1 is an element of the
set M1.

Observe at this stage that the Malnourishment Set will
not be identical for all individuals, and we assume later
that variations will occur on account of the age and sex of
individuals.
We do not for the moment, however., offer a
precise specification of the Malnourishment Sets M1. Instead,
let S = {'i : N.f 1 6M
be the set of malnourished individuals
and let S(e) = {i:e1<e}refer to the set of individuals whose
personal consumption expenditure does not exceed some non.
negative number e.
Let the notation ISI and
IS(e)|
refer to
the number of elements in the sets S and S(e) respectively.
We now propose a definition of the poverty line.
DEFINITION:

The poverty line P is defined by
P = mini mum {e: ]S (e )| >(S|}

REMARK:
The motivation
behind
this
definition
is
straightforward: the number of people whose consumption
expenditure does not exceed the poverty line should equal the
number of malnourished people.
However, equality may never
exist as a number of individuals may have a per capita
consumption expenditure equal to the poverty line.
Thus P is
the minimum expenditure level at which the number of people
with a consumption expenditure nd higher than P is no less
than the number of malnourished people.
Observe that not everyone with a consumption expenditure
not exceeding P is necessarily malnourished, just as some
individuals with a consumption expenditure exceeding P may be
malnourished. This in itself cannot constitute a critique of
our definition of the poverty line, for such a scatter of
malnourished and non.malnourished individuals on the
"wrong"
side of the poverty line is necessarily a feature of any
n u t r i t i o n-. b a s e d definition of the poverty line, unless one
presupposes a monotonically non->decreasi ng relationship
between consumption expenditure and the quantity of..every
nutrient consumed (which we henceforth, refer to as the
monotonicity property).
Indeed, if we define:

3

P
= m a x^e1 : i G S]
.
and Pm-jn = max^e^iGS and e1

< e1

for all

i^Sj

then all individuals with a consumption expenditure fl-o-fe
exceed i n-g P , n
are
malnourished and no individual with a
consumption expenditure exceeding Pmax is malnourished.
We
referto the closed interval [Pm-jn>Pn)axl as the poverty band.
The following proposition then nolds:

LEMMA:
The poverty line lies within the poverty band.
Further, if the monotonicity property holds within the
economy and if the minimal nutrition basket needed to avoid
malnourishment is uniform for all individuals, then the
poverty band is identical to the poverty line.

PROOF :
As
there
is
no malnourished
individual
with
a
consumption
expenditure exceeding
Pmax>
it
follows
that
| S (P m a x )| > |S| . Hence P< Pmax- Further, as every individual
with a consumption expenditure less than
pmin
s
malnourished, it follows that for any e < P[n-jn we have |S(e)|
< IS) , implying that Pml-n < P.
Hence the poverty line is
contained within the povery band.
To show that the poverty line is identical to the
poverty band (ie. that Proax = Pmin) when
the monotonicity
property holds and all individuals have a uniform minimal
nutrition requirement, assume
otherwise.
Let m = (mp...m-|)
be the minimal nutrition basket indicating the minimal
quantity of each nutrient required to avoid malnourishment.
By hypothesis
this is identical for all individuals.
Let
the malnourished individual with consumption expenditure Pmax
be deficient in his net consumption of some nutrient j.
If
his consumption of this nutrient is n't,
we have n't < m j.
Further, if the non malnourished individual with consumption
expenditure< Pm£^^ has a net consumption
n't,
of nutrient j,
we have n j > m j. Finally, as the monotonicity property holds
and as
PJma-x >
Pmin» we have nt > n j. These inequalities
imply a contradiction. Hence P m
, Y = P = P111
m<
n
111 Cl A
I 11
REMARK: The assumptions that the minimal nutrition basket
needed to avoid malnourishment is uniform across individuals
and that monotonicity is satisfied are stringent assumptions
and will not be imposed later in this paper when we examine
NSS data from Karnataka.
In the absence of these 2
assumptions, however, the poverty band is in general non-.
degenerate (i.e. does not necessarily collapse to a point).
As there is no ambiguity about the nutritional status of
those either above or below the poverty band, and as it is
therefore desirable that the poverty line lie within this
band (which is assured by the LEMMA above), this reinforces
intuitive support for the definition adopted in this paper of
the poverty line.

4

We now introduce a simplification to our framework by
assuming that our nutritional baskets consist of just two
nutrients which we call energy (measured by the consumption
of calories) and proteins.
To be precise, energy is a
nutritional characteristic rather than a nutrient,
and the
label "nutrient" is used figuratively.
Observe that there
are other nutrients (particularly vitamins) which are
essential in minimal amounts for subsistence.
Most studies
on poverty line computations,
however,
have typically
considered just one nutrient (energy through calories) and
our inclusion of proteins provides greater generality.
Whether the exclusion of other nutrients is a limitation of
our exercise will be considered later.
With just 2 nutrifents entering into the nutrition basket,
the food-nutrition transformation matrix N is clearly a 2 x k
matrix.
At this stage, however, it is necessary to
discriminate between the gross and net intake of energy and
proteins, in view of the special way Tn which the human body
absorbs these nutrients.
Although the principal source of
energy in the human body is fats and carbohydrates, proteins
can also generate energy, as all these organic matters are
converted into heat energy (measured through calories).
The
role of protein
as a source of energy through oxidisation
competes, however, with its more fundamental role of being
the principal building block of blood cells and tissue, which
enables the human body to grow.
It is crucial, therefore, to
comprehend the manner in which these competing uses of
protein are resolved.

In practice fats and carbohydrates are much more readily
oxidised than are proteins.
As long as these readier sources
of energy are available in sufficient amounts to provide
adequate energy, protein's prime role of facilitating the
building of blood cells and tissue is fully retained.
If,
however, fats and carbohydrates fail to provide adequate
energy, protein is available for oxidisation for filling the
gap, resulting in a di in unition of protein's central role.
It
is thus crucial
(particularly at subsistence levels of
nutrition) to be conscious of the extent to which protein is
being oxidised to provide energy.
Thus, although the food
nutrition transformation
matrix
N
refers
to the
gross
consumption of energy and proteins (the nutritional content
of food items), the body's actual retention must be measured
by their net consumption, and these may well be different in
malnourished people.

This discussion above immediately introduces a caveat
into the specification of the .set M . of millnourishment
vectors for individual i. If n1 = (n j1 ,, n-j 1 ) represents
the minimal nutrition vector needed to avoid malnourishment,
then it may appear reasonable at first blush to define M
as
(h
=(nj,....,ni):
< n 1 , f o r some j}.However, this will be
inadmissible as the vector n1 represents net consumption,
5

while M1
as defined in this paper is a set of _£££s.$
consumption vectors.
The existence of these biochemicaT
relationships
w h i c h induces the need to discriminate between
gross and net consumption of nutrients was first recognised
by
Osmani (1982), though
in his
work he continues
to
define the ma 1nourishment set in
the above inadmissible
manner.
That this distinction between gross and net consumption
of nutrients necessarily alters the boundary of the
Malnourishment Set is apparent from FIGURE 1, where all
nutrients are measured in gross terms.
We assume that all
individuals fall into one of several age-sex groups in the
population, each of
which is character!sed by
a different
minimal protein and energy norm.
If nefa
and np^1 are the

minimal net consumption of energy and protein required to
avoid malnourishment for an individual who belongs to some
age-sex group G , then the Malnourishment Set for that
individual is the shaded part in the diagram.
If the gross
consumption of energy is less than ne ,
then gross protein
consumed (measured in grams) is initially converted
into
energy (measured in
kilo calories) at the rate of 1:4 upto
the point where the aggregate availability of energy
reaches
n^ .
Thus the
line AB has a negative slope of 4.
The Household Malnourishment Set is then given by the
aggregation of
individual Malnourishment Sets and we discuss
the method of aggregation later.
FIGURE 2 illustrates such a
Household Malnourishment Set.
Its boundary is piecewise
linear with 4 linear segments, indicating that the household
consists of individuals who fall into 3 different minimal
nutrient groups.

To determine whether the household is malnourished, it
suffices to ascertain whether the actual household intake of
gross protein and gross energy (P,E) is within the Household
Malnourishment Set.
We assume that all individuals in a
household are malnourished if and only if the household is
malnourished.
Although this abstracts from certain important
issues
of intra-household feeding practices, the assumption
is inevitable
if individual consumption details are
unavailable.
We discuss later whether this is likely to bias
our results.
The method outlined above
enables us to identify all
malnourished individuals and to thereby compute the poverty
line in accordance with the definition given earlier.
We
propose later an algorithm for determining which households
are malnourished; meanwhile, in the next Section, we compare
our
methodology
with
other
approaches
towards
the
construction of the poverty line.

6

3.

Earlier Approaches

Earlier approaches to the construction of the poverty
line fall conveniently into 3 conceptual categories.
The
first, and most widely used approach in
the context of
Indian data, is to postulate a minimum individual calorie
norm for avoiding ma 1nourishment, identify the expenditure
class (from grouped NSS data) within which such a calorie
norm falls, and estimate the poverty line as falling within
such an expenditure class (e.g. mid-point of the expenditure
class).
Although there have been variations in the
estimational procedures used (which have become increasingly
sophisticated over the years), they conform to a uniform
logical prototype.
The studies of Sukhatme (1965), Dandekar
and Rath (1969),
Bardhan (1974),
Minhas (1971) and the
Planning Commission (1979) follow such a logic, as do all
official documents in India which have attempted poverty line
computations.
These efforts must now be seen as based on a
tentative conceptualisation, assuming as they do that:
(a) calorie consumption i s a sufficient indicator o f
malnutrition ;
(b) no distinction is made between the gross and n e t
consumption
of
calor i e s ,
despite
the
biochemic a 1
relationship that exi sts between protein and calor i e
consumption which ought not to be ignored at low levels o f
nutrition; and
(c) there is
a monotonic relationship between average
calorie consumption and average per capita expenditure within
each expenditure class.

All

the

3 assumptions are relaxed in our present paper.

A second approach, associated particularly with
Pannikar (1972) and Rajaraman (1974) is to postulate a
minimum calorie norm as before and, given a vector of food
commodity
prices, to estimate a least-cost diet which can be
consumed at these prices.
Given data on the proportion of
total consumption expenditure spent on food, this provides
the rationale
for an alternative concept of the poverty line
Although such
an approach dispenses
with the monotonicity
assumption (though it retains the
other 2 assumptions of the
first approach discussed above), this conceptualisation must
be seen as
being strongly normative.
An individual may not
necessarily
purchase
such a
least-cost
bundle
of food
commodities, either because of ignorance or on account of
ingrained food habits, and the conceptualisation is .largely
insensitive to this.
To argue that people should change
their food habits in order to consume a higher quantity of
nutrients
and thereby escape being labelled poor is not very
helpful in an estimation of poverty today. Although Pannikar
and Rajaraman's exercises
do recognise this through their
introduction of 'palatability constraints' in their least*
cost linear programming exercises, these constraints are

7

entered in the most rudimentary fashion, ana their exercises
are undoubtedly normative.
No such
prescriptiveness
exists
in our present paper.
A third approach, initiated by Osmani (1982), pioneers
the notion of the distinction between the gross and net
consumption of nutrients (wh'ich we have adopted in this
paper) but does not dispense with the monotonicity assumption
which is used to generate a
convergent algorithm for
estimating the poverty line (monotonicity being crucial for
ensuring convergence).
Our present paper dispenses with this
assumption.

One of the first attempts at computing the poverty line
for India, which falls into none of the above 3 categories,
emerged from the recommendations in 1962 of a Study Group set
up
by the Planning Commission which suggested that a per
capita annual consumption of Rs.240 at 1960-61 prices
(excluding expenditure on health and education which should
be provided free by
the State) be treated as the nationally
desirable minimum level of consumption expenditure.
However,
there is no rigorous derivation of such a poverty line from
nutritional norms and
the
formal logic employed by this
Study Group in making their computations continues to be a
little unclear.

4.

A Solution Algorithm

Determining whether a household is malnourished or not
necessitates (in FIGURE 2) an identification of which side of
the boundary
of the Malnourishment Set a particular
(E,C)
household
consumption is located.
In this Section we
construct
a solution algorithm for this purpose which
precludes the need for information about the entire boundary
of the Household Mainourishment Set.
Observe first, however,
that we have still not given a precise definition of the
Household Mainourishment Set, beyond stating that it
is an
aggregated derivative of the Mainourishment Sets of the
individuals contained in the household.
There can be several
ways of deriving such a
Household Ma 1nourishment Set.
Consider the following two methods:
(a) The Household M a 1 n o u r i s h rn e n t Set is the vector sum
of
elements of the individual Ma 1nourishment Sets
n

e.
i.

M,
1

n

= {(E x i

y i ): ( x i , y 1

)

6

M

,

for

all

i}

<.=■)

L-1

where.we assume that the household consists of n individuals,
and M1 is the Malnourishment Set of individual i.
(b) The complement of the Household Malnourishment Set is the
vector sum of elements in the complement of individual
Mainouri shment Sets
n

i.e.M2

= -£ { ( £

n

xi,£yi

8

):

(x1

, y1) G

£ M1 .for all'ij

d
<9

where ,6
refers to the complement of a set in the non.
negative quadrant of the cartesian plane.
It is clear that the first definition M, offers a very
liberal coverage under malnourishment:
if a particular
household consumption can be derived as the vector sum of
individual malnourishment consumption (however unlikely in
practice these individual arrays of consumption be) then that
household consumption is malnourished.
On the other hand,
the second definition Mg is more conservative in admitting
malnourishment : if any household consumption can be derived
as a vector sum of individual non«malnourished consumption
(however improbable that such n o n > m a 1 n o u r i s h e d consumption
will exist in practice), then that household consumption will
not be regarded as malnourished.

»
9

9
■i
9
9
■*

FIGURE 3 refers to a household with 2 individuals, the
boundaries of whose individual Malnourishment Sets are given
by lj and Ig.
The
2
kinds of Household Malnourishment Sets
M J . and Mg are also shown. We now seek a notion
of
household malnourishment which
is
more liberal than Mg yet
less so than Mj.
We
use polar
coordinates (r,0) to define
such a Household Malnourishment Set M.

d

>

9
9
9

DEFINITION:
The Household Malnourishment Set M is defined by
the use of polar coordinates as:

9

n

M = {(£r1,

9
9

9
9
9

9
9

9
9
9

9

9
9
9

9
9
9
9

0) :

(r1, 0)

9
9

9

M1 > for all

i}.

REMARK:
We recommend this
definition in view of its
intuitive simplicity.
A point on the boundary of such a
Household Malnutrition Set is then arrived at by moving up
the line connecting the origin to that point in steps
representing the intersection
of each individual
Malnourishment Set with this line.
FIGURE
3 reveals that
the boundary of M lies
between the boundaries of M1 and Mg.
The definition also motivates our solution algorithm which,
in FIGURE
2, seeks to compare the position of the point A
(the intersection of the line joining
(P,E)
to
the origin
with the
boundary of the Household
Malnourishment Set) with
the
point
(P,E).
Again,
the
point
A in our algorithm is
arrived at by moving
up the line
connecting the origin to
(P,E)
in steps representing the intersection of each
individual Malnourishment Set with this line.
The point A is
reached when all individuals in the household have been
considered. A and the point
(P,E)
are
then
compared by
noting their
projections
onto the
protein axis.
(Equally,
of course, we may also project onto the energy axis).

We assume that all individuals fall into one of 10 age*
sex groups in the population, each
of which is characterised
by a different minimal net requirement of proteins and, energy
for subsistence.
For any such group G, let (nD , ne '
represent the minimal net nutrition required by an individual

9
9

G

9

contained in this group to avoid malnourishment.
FIGURE 1
then indicates that the line with slope Z = E/P from the
origin intersects the
boundary of the Malnourishment Set of
that individual at some point (in b, m e J), where:

f

m G =
P

<
I

> neG

npG.

i f Z

(npG)

ncG+4noG>
7+4

1 f z

(%G) < neG
H

This enables us to construct the following
aided algorithm, as a sequence of instructions:
1. Consider the first

individual

2.

Identify the individual's
and compute m p b.

3.

Put Mp = nipG

* 4.

5.

computer

in the household.

age-sex group G

If Mp > P, then we conclude the household is not
malnourished, and move to the last instruction 5"elow.

If Mp < P, and all household individuals have
been
considered, we conclude the household is malnourished
and move to the last instruction below

6.If Mn < P and not all household individuals have been
considered, then select the next individual in the household

Identify the individual's age.sex group G and compute
mp .
p
8. Increment Mp by m p
and let thi s be the new value of
7.

V
9.

Go to the instruction marked *

10.

STOP

The algorithm enables
us to identify
all
the
malnourished households and to thereby compute the poverty
line and the proportion of the poor.
Clearly, the
algorithm is fully decisive and can be
separately appTi ed on
urban and rural data.
We do so in the next Secti on.

5. Applications to Karnataka
We now use the ideas developed above to estimate, using
1983 NSS data, the poverty line for that year in Karnataka,
as well as the proportion of people below the poverty line.
3 8 7k
eg) 'fKa, //£s w 'fKz

10

household
consumer
expenditure
and
employment
and
unemployment
(the earlier' surveys being in the 27th Round
1972.73 and the 32nd Round 1 9 7 7. 7 8 ).
The State sample for
Karnataka covers all 19 districts and has a two.stage
stratified design : the first stage units being the villages
and urban blocks, and the second stage being the households
in them.
A household is taken to be "a group of persons
normally eating together and taking food from a common
kitchen".

The survey covered 336 villages and 240 blocks,
involving 3320 rural households and 2390 urban households.
202 rural households and 236 urban households were deleted,
largely because data pertaining to them was either incomplete
or inconsistent, but also because individuals in some of
these households ate partially or completely outside the
household, and details of such external food consumed were
unavailable. Thus a total of 3118 rural households involving
17,242 individuals, and 2154 urban households involving
12,295 individuals, were processed for this study.

From the exhaustive data schedules available, data on
the consumption of 107 food items were culled out and grouped
into 26 aggregated food items.
The consumption of each
household of these 26 food items includes data on quantities
purchased as well as quantities consumed from home.grown
stock .
The food«nutrition transformation matrix N defined in
Section 2
is thus a (2 x 26)
matrix whose coefficients
emerge from the data contained in TABLE 1.
Further, TABLE 2
contains the definitions of the 10 age.sex groups as well as
the minimal net consumption necessary of protein and energy
for avoiding ma 1 nourishment.
The data in TABLE 1 is drawn
from the Indian Council of Medical Research (1971), while the
minimal protein and energy norms contained in TABLE 2 are
from the Indian Council of Medical Research (1968).
The results emerging for Karnataka in 1983 are presented
below:
RURAL KARNATAKA:
No.of individuals covered
17,242
No.of individuals malnourished
11,5-08
Rural poverty line
Rs. 113.85
% of people below the poverty line
66.7
URBAN KARNATAKA
No.of individuals covered
12,295
No.of individuals malnourished
10,112
Urban poverty line
Rs. 176.65
% of people below the poverty line
82.2
ALL KARNATAKA
No.of individuals covered
29 , 537
No.of individuals malnourished
21,620
% of people below the poverty line
73.2

'/
«
INFORMATION y

ft

fc

'i (j

/ j

n85

6.

a

e
a
a
a

a
a
a
a

a

a
a

Conclusions and Caveats

This paper has suggested a critique and a reformulation
of the notion of the poverty line and we would like to argue
that its main contribution lies
in the specific
investigative logic proposed.
Such a conceptualisation is
also fully operationalisable if data on individual household
consumption is available, as our application to 1983 NSS data
from Karnataka reveals. The possibility of estimational bias
in our results does require scrutiny, however, particularly
in view of the very high proportion of the
urban poor that
our results provide.
The Government of Karnataka's Draft
Seventh Five Year Plan (which has also analysed 1983 NSS data)
has computed the proportion of the poor as 61% in rural
areas, 58% in urban areas and 60% in all Karnataka.
The
methodology used by that document conforms essentially to the
first category of exercises discussed in Section 3 above, and
relies on the Planning Commission's (1979) poverty line
threshold of 2400 calories in rural areas and 2100 calories
in urban areas.
This is linked to monetary consumption
expenditure and for 1979-80 the poverty line was computed at
current prices as Rs.76 per capita per month in rural areas
and Rs.88 in urban areas.
Deflating these estimates by the
Consumer Price Index for Agricultural Labourers (in rural
areas) and for Industrial Workers (in urban areas) yields
the 1983 Poverty Line in the Draft Seventh Plan document of
Rs.108 (rural) and Rs.130 (urban).

a

a
a
a
a
a
a
a
a
a
a
a

a
a
»
a

a
a

a

a
a

A major reason why poverty levels as assessed in our
present paper are higher than in the Draft Seventh Plan
document of the Government of Karnataka must lie in our
inclusion of protein adequacy as an indicator of nourishment;
consequently, therefore, a major conclusion of our study must
be that protein deficiency is more acute in urban rather than
rural areas.
Indeed, the wide divergence in the estimates of
the urban poor in Karnataka as revealed in these 2 studies
lends further justification to the incorporation of protein
inadequacy as an indicator of malnourishment.

This necessarily also raises the issue of whether the
exclusion of other crucial nutrients (particularly
vitamins)
leads to an underestimation of poverty.
Clearly, a view on
this must depend on an assessment of the categories of
malnourishment which induce poverty through their
impact on
the capacity to earn.
Thus, for instance, a plausible case
can be argued that acute deficiency in vitamin A consumption
can induce poverty through blindness, as can the prevalence
of goitre
which can lead to mental retardation in children.
Our calculations of the size of the poor are in this respect
probably biased downwards.
Observe, however, that the
general framework we set up in Section 2 does permit us in
principle to take such a comprehensive view of malnourishment
in its impact on poverty.
12

The aggregation of individual Mainourishment Sets into
the Household Malnourishment Set may also, of course, lead to
esti (national bias.
We $ r g u e d earlier, in the context of
FIGURE 3, that
the set M^ probably overestimates the extent
of malnourishment, whereas the set M2 underestimates it.
We
believe intuitively that the set M which we have constructed
reduces such estimational bias.
Given the information
structure contained in NSS household schedules, which yield
no information about the consumption patterns of individuals
in
the household, the boundary of the Household
Malnourishment Set is inevitably blurred.
Despite this,
however, in obtaining information from household schedules
rather than grouped NSS data, we have utilised a richer
information structure than have most earlier studies of the
poverty line in
the Indian economy, and thereby reduced
estimational bias.

Aggregation from individual to household malnourishment
can, however, introduce a different sort of bias.
Given the
considerable evidence that exists of sequential feeding
within poor families ("male breadwinners eat first, male
children eat next
adult females eat last whatever
remains"), a malnourished household (as identified
by our
solution algorithm)
does
not
necessarily
imply
that a 11
individuals in the household are malnourished.
In this sense
our results do probably overestimate poverty, though the
magnitude of this bias is difficult to quantify in the
absence of data on intra«household food consumption.
As against this, an underestimation of rural poverty is
also implicit in our minimal nutritional norms being
dependent on just age and sex, and not on the worker's
occupation.
For adult males our norms refer to moderate
physical activity at work, whereas more
strenuous manual
activity would necessitate a higher nutritional threshold for
avoiding malnourishment.
Poverty amongst agricultural
labourers may, therefore, be higher than our approach allows
for.

REFERENCES
BARDHAN, P.K.
(1974) “On the Incidence of Poverty in
Rural
India in the Sixties" in Srinivasan and Bardhan (ed.)Poverty and
Income Distribution in India, Calcutta.
DANDEKAR, V. M. AND RATH, Nl (~1 96 9 ): Poverty in India, Indian
School of Political Economy, Pune.
INDIAN COUNCIL OF MEDICAL RESEARCH (1 968): Recommended Dai ly
Allowances of Nutrients and Balanced Diets, Hyderabad.
INDIAN COUNCIL OF MLblC'AL RESEARCH (1971 ): Nutritive Value of
Indian Foods, Hyderabad.
MINHAS, B. S. (”1971): "Rural Poverty and the Minimum Level of
Living",Indian Economic Review, Vo1.6
■LHS-M A N I, S. R ^ ■ ■ ■( 1-9 8-2-) ;-—E-e-e-n o m i c I n e q u a 4-i^y—-a-n-d—G:F-ew-p W e-1 f-a-F e , Oxford

13

OSMANI,S.R. (1982): Economic Inequality and Group Welfare, Oxford
University
Press.
PANNIKAR, P.G.K. ( 1 9 72 ): "Economics of Nutrition", Economi c and
Pol. itical Week 1y, Annual Number.
PLANNING COMM IS S ION ( 1 9 7 9 ): Report of the Task Force on
Projections of Minimum Need's and Effective Consumption
Demand? New Delhi.
RAJARAMAN, I (1974): "Constructing the Poverty Line: Rural
Punjab, 1960*61 ",Discuss ion Paper 4 3, Research Program in
Economic Develop ment, Princeton Univers i t y
(M i m e o)
SEN, A.K.
(1976):
"Poverty:
An Ordinal Approach to
Measurement",
Econometri ca, Vo1.44
SUKHATME, V.V. (1965): Feeding India's Growing Millions, Asia
Publishing House, Bombay

14

TABLE 1
COEFFICIENTS FOR THE FOOD-NUTRITION TRANSFORMATION MATRIX

Unit of
Food Item Measu
rement

Rice
Kgs
2*. Wheat
Kgs
3. Jowa r
Kgs
4. Ba j ra
Kgs
5. Maize
Kgs
6. Ba r 1 e.y
Kgs
7. Small
Millets Kgs.
8. Ra gi
Kgs.
9. Gram
Kgs.
10. Pulses
Kgs .
11 . Milk
Li tres
12. Ghee
Kgs.
13. Edible
Oi 1
Kgs
14. Meat
Kgs
15. Poultry Nos
16. Eggs
Nos
17. Fresh
Fish
Kgs
18. Dry Fish Kgs.
19. Vege
tables
Kgs.
20
Bananas Nos.
21
Guavas
Nos.
22
Oranges/
Mosambi s Nos .
23. Mangos
Kgs .
24. Grapes
Kgs .
25. Ground
nut
Kgs
26.Sugar
Kgs

Edible
Portion
where less
than full
(%)

Calorie/Unit
(k.cal .)

Protein/Unit
(gms.)

3178,0
2988.0
3074.0
2642.6
2976.0
2900.0

68.0
118.0
104.0
97.4
111.0
115.0

2822.0
2988.0
2648.0
2322.0
656.0
9000.0

62.0
73.0
205 .5
2 8 2.0
34.8
0.0

9000.0
1020.0
54.0
35.9

0.0
157.3
259.0
4.0

66

23 7 .6
76 6.0

148.5 .
681.0

95
71

953.0
79.0
48.9

39.0
0.9
0.1

79
74

33.0
53.0
690.0

0.7
0.4
5.0

73

3400.3
3976.0

184.7
1.0

84

85
70

15

r

r
r

e
r
e

r
9
9

TABLE 2
AGE-SEX GROUPS AND SUBSISTENCE NORMS OF PROTEIN
PUR PloKSi-i
ANU ENERGY CONSUMPTION
a » Ji A * M k \ ii k > i li l ■ M ^ > > > I
a « n « > i» a a x » > j » W n * » a « a
Group
Age-Range Sex
Minimal Protein
Minimal Energy
Consumpti on
Consumpti on
(k . ca1s . )
(fc'gs.)
- — — ft-W — - -- -- -- -- -

i

1
2
3
4
5
6
7
8
9
10

>19
>19
16*18
13-18
13*15
10*12
7*9
4*6
1-3
<1

M
F
M
F
M
M,F
M,F
M, F
M.F
M,F

1650
1350
1800
1500
1650
1230
990
660
540
450

•* sii».ajtsM»*asiJtk 1 JX lllixi 1 a * * W > a i* * k a W • k M

•
9

9
9

9
9
9
9
9
9
9

9

9

16

84,000
66,000
90,000
66,000
75,000
63,000
54,000
45,000
36,000
27,000
M it m ii 'a * m a i * a a 1 j i S 1 J X S J J S X > > 1 j X ■a k 9 k W .•» 'a.

FIGURE 1
MALHOURISHMENT SET FOR INDIVIDUAL IN GROUP G

protein (gmd.)

FIGURE ?

TIE AGGREGATION OF INDIVIDUAL MALNOURISHIE NT
SETS INTO A HOUSEHOLD MALNOURISHMENT SET

energy
(k.cal)

protein (gms)

Media: 10975.pdf

Position: 2821 (3 views)