35
Malnutrition: Some Measurement and Policy Issues SWP373 Wodrd Bank Staff Working Paper No. 373 rj-C- 2LI431 February 1980 Prepared by: T.N. Srinivasan Development Research Center Copyright L 1980 The World Bank 1818 H Street, N.W. Washington, D.C. 20433, U.S.A. The views and interpretations in this documeent are those of the author and should not be attributed to the World Bank, to its affiliated organizations, or to any individual acting in their behalf. Public Disclosure Authorized Public Disclosure Authorized Public Disclosure Authorized Public Disclosure Authorized Public Disclosure Authorized Public Disclosure Authorized Public Disclosure Authorized Public Disclosure Authorized

Malnutrition: Some Measurement and Policy - World Bankdocuments.worldbank.org/curated/en/800061468739194054/pdf/multi0... · is not yet fully understood and a satisfactory theoretical

Embed Size (px)

Citation preview

Malnutrition: Some Measurement and Policy Issues

SWP373Wodrd Bank Staff Working Paper No. 373

rj-C- 2LI431

February 1980

Prepared by: T.N. SrinivasanDevelopment Research Center

Copyright L 1980The World Bank1818 H Street, N.W.Washington, D.C. 20433, U.S.A.

The views and interpretations in this documeent are those of the authorand should not be attributed to the World Bank, to its affiliatedorganizations, or to any individual acting in their behalf.

Pub

lic D

iscl

osur

e A

utho

rized

Pub

lic D

iscl

osur

e A

utho

rized

Pub

lic D

iscl

osur

e A

utho

rized

Pub

lic D

iscl

osur

e A

utho

rized

Pub

lic D

iscl

osur

e A

utho

rized

Pub

lic D

iscl

osur

e A

utho

rized

Pub

lic D

iscl

osur

e A

utho

rized

Pub

lic D

iscl

osur

e A

utho

rized

The views and interpretations in this document are those of the author andshould not be attributed to the World Bank, to its affiliated organizationsor to any individual acting in their behalf.

WORLD BANK

Bank Staff Working Paper No. 373

February 1980

ALNUJTRITION: SOME MEASUREMENT AND POLICY ISSUES

The "energy" and "protein" norms published by the FAO have beenwidely misused for assessing the nutritional status of a population inspiteof warnings against such use by the FAO itself. The paper points out thatthe process governing variations in daily energy balances in human beingsis not yet fully understood and a satisfactory theoretical framework for itis yet to emerge. Nevertheless, there is some evidence that the process isstochastic-stationary and auto-regressive and there is a wide variability inenergy intakes between otherwise similarly situated healthy individuals andeven in the same individual over time. Many of the widely used proceduresfor assessing nutritional status, by classifying all individuals in a popu-lation as malnourished who have intakes below a single average norm for thepopulation as a whole, thereby ignoring intra and inter individual variancein intakes and requirements, will misclassify individuals (i.e. adequatelynourished as malnourished and vice-versa) to varying degrees. This mis-classification bias need not cancel out for the population as a whole andthere is a danger in overestimating the proportion of truly malnourished.

It is important for devising suitable policies for eliminatingmalnutrition where it exists, that the extent of it is not overestimatedand exaggerated and the malnourished get properly identified as a targetgroup. For such overestimatiion will only make the problem appear even moreintractable than it already is and lead to inaccurate targeting of policiesresulting in a wasteful use of resources. The process of probing into thetrue extent of malnutrition is likely to reveal that it is largely amonghouseholds who are not only poor but suffer other social disadvantages anddiscrimination, such as the untouchable and tribal households in India, orsimilarly disadvantaged groups in North East Brazil. Further, they oftenhave limited access to safe drinking water, sanitation and health and edu-cational facilities, since they are consigned to urban slums and ruralghettos. Attempts to tackle their nutritional problems alone in such acontext are almost futile. Integrating them more fully into the economyby providing them adequate and secure income earning opportunities, while

- 2 -

at the same time making a determined effort at reducing their formidablesocial disadvantages will go further in tackling their poverty and nutritionalproblems than nutrition interventions per se. While the interaction betweennutritional deficiency and the health status of an individual runs both ways,the policies that reduce morbidity such as provision of safe drinking water,sanitation facilities and a generally healthy environment may be the appro-priate way to break into this reaction chain, with policies for nutritionintervention following if needed.

Prepared by: Copyright @ 1980The World Bank

T.N. Srinivasan 1818 H Street, N.W.Development Research Center Washington, D.C. 20433

U.S.A.

MALNUTRITION: SOME MEASUREMENT AND POLICY ISSUES *

1. INTRODUCTION

One of the widely used indirect methods of estimating the extent

of malnutrition (i.e., the nutritional status) of a population is to estimate

it as that proportion of the population with energy intakes below some norms

called "requirements." These norms, for a reference man and for a reference

woman were published by the Food and Agricultural Organisation (FAO) and

evaluated periodically by their expert groups. See FAO (1950, 1957, 1973,

1974, 1975, 1978a, 1978b). Based on the "requirements" of a reference man

or woman, those for other age-sex groups were also made available.

The publications of FAO were very careful to point out that the

"requirements" cannot be used for examining the nutritional status of a

population. For instance, FAO (1974) states, "The figures for recommended

intakes may be compared with actual consumption figures determined by food-

consumption surveys. Such comparisons, though always useful, cannot in

themselves justify statements that undernutrition, maZnutrition or over-

nutrition is present in a community or group, as such concZusions must

always be supported by clinicaZ or bio-chemical evidence. The recommended

1/ A reference man is between 20 and 39 years of age, weighs 65 kg.,healthy and employed for eight hours in an occupation involvingmoderate activity. He spends eight hours in bed, four to six hoursin light activity and two hours in walking, active recreation, etc.

A reference woman is between 20 and 39 years of age, weighs 55 kg.,healthy and employed for eight hours of moderate activity. She spendseight hours in bed, four to six hours sitting or in very light activityand two hours in walking, active recreation, etc.

* This paper owes its inspiration to a seminar addressed in 1978 byProfessor P.V. Sukhatme at the Development Research Center of the WorldBank. Indeed many of the ideas discussed here are his and my debt tohis writings is gratefully acknowledged. I am grateful to ProfessorSukhatme, V.V. Bhatt, Robert Cassen, Dean Jamison, Philip Payne,Lance Taylor and anonymous referees for their valuable comments. Noneof them is resDonsible for anv misinterDretation of their idias.

- 2 -

intakes are not an adequate yardstick for assessing health because ... each

figure represents an average requirement augmented by a factor that takes

into account inter-individual variability." (p. 2, emphasis added).2 /

Inspite of such unequivocal statements, even the FAO, in its Fourth

World Food Survey (1977), has not refrained from using average energy

requirements as the basis for estimating the proportion of malnourished in

a number of countries!

There is a further widespread use of the average energy requirement

of a population in determining an absolute poverty line through the income-

calorie intake relationship. This is apparently based on an almost naive

belief that the arbitrariness involved in defining a poverty level of

income can be escaped by linking it to an 'objective' standard such as energy

requirement. Thus the poor are defined as those who cannot afford to buy

their energy requirements. The early studies on India, by P.D. Ojha (1970),

Dandekar and Rath (1971) and a much later global study of Reutlinger and

Selowsky (1976), are of this type. There have been studies, which go beyond

estimating the proprotion of calorie-deficient in a population at a time and

estimate, the calorie-gap in terms of food grains, that needs to be met if

malnutrition is to be eliminated. Further, alternative policy (national

and international) interventions for eliminating malnutrition (and poverty)

are often compared in terms of their cost effectiveness, administrative

feasibility and their implications for international capital transfers.-/

2/ FAO also points out that the methodological basis for estimating energyrequirements is weak and much of the information on protein requirementsand energy-protein relationships come from studies on healthy young menin the USA, thereby making them of questionable applicability elsewhere.[See FAO (1978a), pp. 2-4.]

3/ See FAO (1977), Scandizzo and Graves (1978) and Scandizzo and Knudsen(1979).

- 3 -

It would appear that the analytical and empirical foundations for

the use of calorie norm in examining problems of the nutritional status or of

absolute poverty are dubious. The purposes of this paper are: firstly, to

describe a theoretical model of energy balance and its statistical properties

(Section 2); secondly, in the light of the proposed model, to examine critically

some of the widely used procedures in estimating malnutrition and their data

base (Sections 3and 4) drawing upon household survey data from India and

Sri Lanka and finally, in Section 5, to offer a few concluding observations

on the implications for policy and further research.

2. THE MEANING OF CALORIE REQUIREMENT AND MALNOURISHMENT: A CONCEPTUALFRAMEWORK

The basic concept that is at the root of the calorie requirement

calculation is that of energy balance. A human body, much like any energy

conversion device, converts the energy content of food (measured in calories)

into other forms of energy of which, a part is used up in maintaining bodily

functions (including, in particular, the maintenance of body temperature), a

part is converted into changes in body weight, a part is used up in physical

activity including work and the remaining part is 'wasted' particularly

in the energy content of bodily wastes. Of course by law of conservation of

energy, the energy input has to equal energy output. Just as there are

differences among different machines in their efficiency (i.e., the ratio of

output of usable energy to energy input), different human beings differ in

their functions as energy converters. However, the human body apparently has an

essential homeostatic or regulatory mechanism which adjusts to varying intakes

of food, without drawing on (losing weight) reserves or adding to reserves

(gaining weight), while at the same time enabling the individual to perform

- 4 -

tasks involving a given level of energy expenditure. Of course, like all

regulatory mechanisms, this one also can accommodate only a certain amount of

variability in daily intakes: too low (high) an intake sustained over a period,

can be accommodated by drawing on (adding to) reserves in terms of body weight

and/or changes in the level of activity. In more extreme cases, the individual

becomes malnourished so that many of his functions are impaired (and in severe

cases he dies) or he suffers from the illnesses arising out of obesity. Thus

we can distinguish two mechanisms operating: a regulatory mechanism which

takes care of moderate variations in intakes with no change in weight or

activity and an adaptation mechanism that adjusts to sustained levels of

too low or too high intakes through changes in bodily function and activity.

Payne and Dugdale suggest that the adaptive mechanism operates over

a much longer time frame (months and years): a prolonged over (under) feeding

in the sense of energy intakes exceeding (falling short of) maintenance energy

needs would result in an increase (decrease) of body weight, the extent of which

depends on body composition, i.e., whether an individual is metabolically

fat, average or lean. While modification of physical activity is a major

element in adaptation to changes in energy intake, it is not easy to distinguish

between cause and effect, that is between intake being low because of light

activity and activity being light because of low intake.A/

There are two alternative points of view as to how the regulatory

mechanism operates. In the Payne and Dugdale (1977) model, positive (negative)

daily energy balances, will lead over a period of time, to an increase (decrease)

4/ Payne and Dugdale (1977a) state that inspite of a large amount ofexperimental work there is as yet no adequate theoretical framework tosuggest critical experiments to test alternative hypotheses about theseprocesses. Professor Payne pointed out in correspondence that it isnot known how far adaptation can take place without an important loss ofbody function.

in body weight resulting in an increase (decrease) in Basic Metabolic Rate

(BMR) and hence a greater (lower) maintenance energy cost, thus reducing

(increasing) the energy balance, eventually turning it negative (positive)

thereby reducing (increasing) body weight (Payne and Dugdale (1977a)).

This model permits a range of genetic variability among individuals (even

of the same age, sex, body weight, etc.)through variations in the ratio of

energy stored as protein to total energy stored. In the model of Sukhatme

and Margen (1978), the relationship is a stationary stochastic process. In

this model, the daily energy balance in a healthy individual maintaining

body weight is a first order Markov process. It would appear that a fixed

non-stochastic relationship between BMR and body weight implies that the

energy requirement of individuals of the same age, sex and body weight would

be identical. But this is contrary to the empirical studies of Widdowson

(1962) that showed that on an average, the intakes of two out of forty such

individuals differ by as much as a factor of two and this difference is reduced

but not eliminated, if an allowance is made for differences in body weight.

A stochastic relationship between BMR and body weight is consistent with these

observations.

Some implications of a stochastic model of energy balance can be

derived as follows. Consider the daily energy intake Xt of a healthy

individual performing essentially the same activity while maintaining his

body weight. We can view Xt as a random variable with mean m . The

discrepancy Xt - m Et , under our assumptions that individual maintains

(i.e., there is no time trend) his body weight and activity level, has the

same distribution and behaves in the same fashion as his daily energy balance.

We can assume that wt of the scoe individual over successive days (denoted

by t) is a first order autoregressive process:

(1) wt = t-l t

where u 's are independently normally distributed with mean zero and

variance a . This implies that:

2 22(2) Ett = 0 and Ew 2 a 2

where E is the expectation operator. The variance a2 is the intra-

individual variance in intakes. The variance of an estimate of m based

on the average of intakes of any T consecutive periods is given by:

T 2

(3) E(TT) T (l )a (lp where

1 + p 2 p(1 - pT)

1 - P T(l_ p)2

It is clear from (3) that the dampening effect on the variance of the mean as

T increases is considerably reduced, the larger the absolute value of p

Of course, if p = 0 we get the usual result that the variance of the mean2

of T observations is a

It is natural to define m = E(X ) in a healthy individual who

maintains his body weight as his 'true' energy or calorie requirement, given his

activity. In other words, the 'true' calorie requirement of an individual

is his mean calorie intake while remaining in health, at the same body weight

and performing the same task or different tasks requiring the same energy.

By the same token, we can say that an individual is malnourished, if the

mean of his actuaZ intake is below the hypothetical mean intake m that

would have enabled him to perform his assigned task while maintaining health

and activity. If the mean m for a specific individual is known, and if data

(in principle, if not necessarily in practice) relating to his actual intakes

Xt over some period of time T are available, then the question of

determining whether he is malnourished or not becomes one of testing whether

his mean intake equals m on the basis of a sample of size T .

Viewing the problem of determination of the nutritional status of

an individual from his mean intakes as a problem of testing of hypothesis

alerts us to the fact that classification errors will arise. The probability

of the two types of error (i.e., of misclassifying a weii-nourished individual

as malnourished and vice-versa) depend both on the procedure of classification

and on the inherent intra-individual variability in intakes. For instance,

if we know m and adopt the procedure of classifying an individual as

malnourished if his mean intake during the observed period T is less than

m , then assuming normality of the relevant distribution, the probability of/m -m

the individual being classified as malnourished is 0 o where ,

w(here and in the rest of the paper) is the cumulative probability function

of unit normal deviate and m is the individual's 'true' mean intake.- This

5/ We are assuming that the stochastic process characterising the deviationsof actual intakes from its mean mo is the same as that characterisingthe deviations of intakes (actual in the case of a healthy individualand hypothetical in the case of malnourished) around the requirement mHowever, in the case of malnourished since both regulatory and adaptationmechanisms may be operating, this assumption is not, strictly speaking, valid

means that this procedure will classify any individual as malnourished with

a positive probability. If in fact m0 = m , i.e., the individual has no

deficiency, there is a 0.5 probability of his being classified as malnourishedl

We can extend this model to a population of individuals following

Sukhatme (1978). Denote by X the energy intake of the ith individual onit

the t day. We can write

(4) xi= 1 + bi + wit

where p could be interpreted as the 'true' energy requirement of the referAnce

individual, bi the deviation of the 'true' requirement of individual i from

that of the reference individual and wit is a random daily fluctuation of

actual intake of individual i around his requirement p + bi . Assuming

that the same first order autoregressive process characterises the behaviour

of all healthy individuals we can write

(5) =it Pwit-l +uit

where uit's are independently (over days and across individuals) and

2normally distributed with mean zero and variance a . Let us further assume

that the distribution of individual specific bi's in the population is

normal with mean zero and variance ab2 and that b 's and wit's are

independent. While the assumption of normality is not necessarily vrlid.it

enables us to make several points sharply.

First, if we consider a population of healthy individuals, 95%

of such individuals will have intakes within the interval p ± 1.96 a

-9-

where a 2 2 + °b . Sukhatme (1970) estimates that not only that

2 2a # 0 but indeed it accounts for more than three quarters of a

Some interesting estimation problems arise in using this model.

Suppose we have data X,t on intakes for T days of a sample N individuals

from a healthy population. Then, it is easy to show that the maximum

likelihood (ignoring autocorrelation of wit) estimates of p, b1 (i = 1, ... N)

and a 2 and a 2 are:b

(6) p = X xit

t7) bi = ab (X -X) where X- Xit

ab +a 1

2 2 1 ' 2)

(8) ab =NL I .i

(9) a = N (Xit bi 1)

From (6) and (7), it would appear that as long as a 2. 0 the ML estimate

of the 'true' requirement p + bi of an individual i is not his mean intake

Xi but a weighted average of Xi and overall mean X , the weights

respectively being the shares of intra-individual and-inter-individual variances

2in total intake variance a . This has the further consequence that the

estimate of the difference in the true requirements of two healthy individuals

i and j is not the difference in their mean intakes X and X but

2

+ ^2 (Xi - X) so that - X; overstates the difference in true

requirements as long as a 2 0

- 10 -

The fact that there are intra-individual (a ) and inter-individual

(ab ) variances in intakes is important in estimating the proportion of mal-

nourished. The consequences of ignoring them altogether, as some of the

currently practised procedures come close to doing so, can be illustrated by

an example: suppose requirements, in fact, vary among individuals but intakes

do not. If the intake of each individual (which is also the average intake

for the population) is less than the average requirement of the population,

we will be classifying the entire population as malnourished while in fact

a part of it will be. If the intake exceeded the average requirement we

will be asserting that no member of the population is malnourished while in

fact some are./ Further the estimates based on this procedure (or its

variants) are highly sensitive to the level of requirement.

3. METHODOLOGICAL PROBLEMS OF A WIDELY USED PROCEDURE OF ESTIMATION OFTHE EXTENT OF MALNUTRITION

Let us waive, for the purposes of present discussion, the conceptual

problems of using calorie requirement, however defined, as a norm of comparison

with intakes to determine the nutritional status and examine some of the

procedures of estimation of the extent of malnutrition. Perhaps the most

widely used procedure in estimating the extent of malnutrition is some variant

of the one used by Ojha (1970) and Dandekar and Rath (1971) in India and later,

independently by Reutlinger and Selowsky (1976). There is a fundamental

6/ As Sukhatme (1977, p. 7) points out, "If the dividing line between theundernourished and overnourished is the average requirement, then oneaannot but conclude that the more serious problem for Brazil isovernutrition than undernutrition! One is bound to reach such an absurdconclusion if one does not allow for intra-individual variation inestimating the incidence of undernutrition."

- 11 -

unity of approach in these studies in spite of their difference in scope,

the data base and the treatment of income-calorie intake relationship. As

such we discuss only the approach of Reutlinger and Selowsky (R-S). They

postulate a linear relation between the calorie intake, i , (of presumably

an individual) and Y , the logarithm of his income, so that i = a + bY .

Given the parameters of this relationship, one can solve for the income Yc

needed to ensure a given calorie consumption c by the antilogarithm of

Yc = (c - a)/b . If we now bring in the data on distribution of individuals

in the population according to their income, denoted by the cumulative dis-

tribution function F( ) , then the proportion of individuals consuming c

calories or below is clearly F(Y ) . By setting c equal to the average

calorie requirement, they get their estimates of the proportion of malnourished.

They vary c around the requirement and find that the estimates are quite

sensitive to the choice of c . It should be observed, and this is a crucial

point, regardless of how the average c is computed, the basis of classification

is the comparison of average calorie intake (estimated or actual) of individuals

in an income class with a single number c , thus in effect ignoring intra and

inter individual variations discussed above. All individuals within an income

class are classified as malnourished, if the average calorie intake of the

class is below c . R-S (1977) in their response to a comment of Payne (1977)

argue that even though any given individual may be misclassified by this

procedure, for the population as a whole the estimates will be unbiased if

either the deviation of each individual's intake from the mean intake of his

income class is the same as the deviation of his requirement from the mean

- 12 -

requirement (assumed the same for all groups) or if the number of individuals

misclassified as malnourished in the income classes that are classified as

malnourished as a whoZe cZass equals the number of individuals who are mis-

classified as adequately nourished in those income classes which are classified

as adequately nourished as a whoZe cZass. They further claim that even if

these conditions are not satisfied, the bias is unlikely to exceed 10%.

Several comments are in order. Firstly, neither condition is

verifiable with their grouped data. Secondly, the first condition (or its

variant that the ratio of an individual's intake to his requirement is the

same as that of the respective means for his income class) simply ensures

that if the income class is getting less calories on an average than its

requirement then every member of the class is suffering from the same

deficiency. A case can be made for such a condition being satisfied, within

a household in the unlikely event that intra-household allocations are so

equitable that each member is allocated calories from the household supply

in proportion to his or her requirement. But it is highly unlikely that

the variations of intakes and requirements of a member of an income class from

the respective class averages satisfy this condition. Thirdly, the second

condition is highly unlikely to be satisfied for the following reasons:

at the upper reaches of the income scale requirements are likely to be fully

met so that there is very little malnutrition and at the lower end the

average requirements themselves are likely to be below the average for the

population as a whole to the extent lower (per capita) income households have

a larger proportion of children and older dependents. In addition there

is likely to be less wastage of food in poorer households. Finally, even in

- 13 -

the event of the R-S method yielding an unbiased estimate of the extent of

malnutrition it will almost surely err on its incidence in the population

since it locates all of it in the group who have incomes below the cut-off

point. Clearly from a policy point of view errors in this way of identifying

the target groups is of some consequence.

It is possible to assess the extent of some of the biases discussed

above on the basis of household survey data. Pravin Visaria (1979) examined the

consequences of applying the R-S procedure with household survey data from

Sri Lanka for 1969-70. The data included information on the age-sex

composition of each household, its total calorie intake from food consumption

and its total consumption expenditure. These were tabulated for Sri Lanka

as a whole as well as for Rural, Urban, Estate (i.e., tea plantations) sector

separately. The calorie norms used were 2,220 Kcal. per capita (FAQ) and

2,750 Kcal. per adult equivalent. When using the latter norm, the number

of individuals of each age-sex group in a household was converted to the

number of equivalent weights by using the weights published by the Indian

National Sample Survey. In determining the level of household expenditure

at which caloric norm was met, he used a regression relationship between

caloric intake per equivalent adult of a household and its per capita

consumption expenditure. Thus the poverty line was defined as that per capita

expenditure at which estimated adult equivalent caloric intake was 2,750 Kcal.

From our point of view the following results based on Visaria (1979) are of

interest.

(1) Age-sex composition does matter: that is, whether one uses a

per capita or per adult equivalent caloric norm makes a difference as shown

below:

- 14 -

Table 1

PERCENTAGE OF POPULATION (SRI LANKA)

\ Intake < 2220 > 2220

Intake PerEquivalent Aul

< 2750 47.1 4.5 51.6

> 2750 4.8 43.6 48.4

51.9 48.1 99.9

Thus, even though the proportion of Sri Lanka's population classified by

either criterion as malnourished is almost the same, households accounting

for 9.3% of the population gets misclassified.

(2) The comparison of the results from classifying the members of

aZZ households with per capita expenditure below the poverty line as mal-

nourished (i.e., by the R-S procedure) with classifying each household by

comparing its adult equivalent calorie intake with the norm of 2750 Kcal.

is as follows:

- 15 -

Table 2A

PERCENTAGE OF POPULATION: SRI LANKA

By Household Adequately\ Specific Malnourished Nourished Totalo mparison

By R-S \Comparison \

Estate 24.2 6.1 30.3

Rural 34.8 8.6 43.4Malnourished Urban 42.9 11.8 54.7

All 32.8 8.4 41.2

Estate 14.5 55.2 69.7

Adequately Rural 17.2 39.4 56.6

Nourished Urban 14.7 30.6 45.3

All 18.8 40.0 58.8

Estate 38.7 61.3 100.0

Total Rural 52.0 48.0 100.0Urban 57.6 42.4 100.0

All 51.6 48.4 100.0

There is a maximum downword bias in using the R-S procedure of the

order of 10.4%. However the percentage of population living in misclassified

households varies from as high as 20.6 in the estate sector to an even higher

27.2 for all island.

Indian data from National Sample Survey Organisation (1976) reveal

a similar picture. The calorie norm used was 2700 Kcals. per consumer unit

(adult equivalent). While in Rural India the R-S procedure has a downward

bias as in every sector of Sri Lanka, in Urban India it has an upward bias.

- 16 -

Table 2B

PERCENTAGE OF POPULATION: INDIA

By Household Adequately\ Specific Malnourished Nourished Total

romparison

By R-S\Comparison

Rural 27.9 7.2 35.1Malnourished

Urban 51.5 14.5 66.0

Adequately Rural 19.5 45.4 64.9

Nourished Urban 11.2 22.8 34.0

Rural 47.4 52.6 100.0Total

Urban 62.7 37.3 100.0

The percentage of Indian population living in misclassified

households is 26.7 and 25.7 in rural and urban areas, respectively.

- 17 -

(3) Perhaps the most interesting result of Visaria's analysis

throws into considerable doubt the use of caloric norm as a criterion for

determining nutritional status, thereby supporting the conceptual criticism

of the first section of this paper. He classifies households according

to the extent of departure of their adult equivalent caloric intake from

the norm and the decile of per capita consumption expenditure to which the

household belongs. We reproduce here the results for the whole of Sri

Lanka only.

(a3nNILN03)

LOT I 1'2t I L'bT I s'St ? R9t t Z'4.1 I 9'£t I L'tt I 6'L I 9'f I n' I . T0 001 I 6'9 1 I1'll I f'2T I f'ST T '7t I I'ft I 8'11 I 2'W I 6'! 1 b7 I O OZ+ 01 T'OT+

I Sgll I ?bSt I 2t91 I 6002 1 SRe1 I 9iLT I h.SST I ILOI I t75 I OS I ItI-------- I------- -J-------- I-A .... To------- I-------- I-------- I-------- I-------- I--X--~I

.t T q Z'9 t !8 I f'9 I £'O1 t O'6 I 9'6 I L'b I 1'9 I O' I L I 0-01+ 03 T-S+O000 r g'q9 i p'oo I o0'o r I Zb It'2 I 0'btl I L'thT I S'6 Z L'9 I 2 + I9L99 I 16s I !F6 I 6q9 I O£21 T 0501 1 12! I LL21 I Ie8 I 915 I SOl I 01

I-------- I--------IT-------- I--------tI-------- I ------- S "I-------- I....... -- ^--I-----IOIL 0 0' 1 L9 I 9'L I 5'9 I go 0'6 I R'6 I P'OT I 6' I E's I £l I

^00T I P9 I 2'9 I 2'6 I 0'6 t 6b'!1 0"!t I OsS I 6'StI 9 'g I £'2 I 0 S+ (I TVO+t£98 I o05 1 IOL I 26L I SLL 5 T 9201 I 9211 I 6621 I 69£1 I ISI I 661 I 6

I-B-------I------- I.....I*"*----s" .....xi-------I-------- I--------I-------- I----^---**-- .- - I--j9-8 I S'h I 9'5 1 2'S I 9'9 t f'g I B'Ot I 0'01 I S'6 I 9'S I V'? IO-COT I 0'5 I t'L I f'9 I Z'h t f'lT I 9'S' I S'S1 I 8'SI I 9'6 I 2*t I 6 7- 03 1 O-9hS9 I 62" I S09 I Ibs I £9L t 996 1 LhET I L2fl I Lb£1 I gig I i9s I L

I9---- 4 -- I-------- I-------- I-------- t-------- I-------- I-------- I-------- I .... ------ IS'8 I b'7 I I'S I I's I 6'0t i 9'L I 6'0t I 9'0T I 6'tl I 6'"' I 6't Igi 60 P I I£ I I'S I I St I `21Y 9'g I O'fl I I q55 I sT I £'ql I lot 1 6'6 - 0 -S

VL£O I SZ9 I OfS I 17jS I O0t t 206 I 19S1 I h2bl I £191 I 20I. I ' 1 9I........ I-------I-'------- I-.-------E ............................... I--------I.b---t----*-dI------o-MI----- -- 1-

0'1 I l'o I S'99 I OIL I L'O % 0'01 I 9'hl I S' 1 I £'0Z I 2i'4 I Loft IO OGT T f'2 I 0' I L'hr I hIL t 9'9 I 9'01 I V'7I I 0'91 L'22 I 2'?t I 6 61 03 OT-22zzt t O&V I tg99 I S19 I 6L21 f9t11 I £291 I Zl? I ?S!2 I 106£ * hOIZ I S

I-------l~-------- I-------- I--------,?-------- I-------- I.......&I ------- I--........--------'O OT I 1V2 1 az2 I 6'£ I Of I t S's I 0'i 1 2'OI I1 b'bI I z*r ' e2 I 't,0-001 I S't I 212 1 l I O'£ I L' t d'b I S'9 I l'olI I 't7 I 9'9Z I 9 \ Z2 6 6Z- 03 Oz -10t LOF I 02 I P&2 I ZOb7 I 09f t 5£9 I ILQ I £h91t I 1561 I 2PS£ I 10Jf I b

I-------I--------I--------I--------t--------I--~I------- &I .......--...... I -------- I------'--IZS-L I h'~O I l't I O't1 I £'Z t. 0'2 t I 2' I V ,ty I 6'9 I l'Zl I f'of I

RnPnOT ? '7 I '14 I t'lt I O'£ I S'Z I Z'£ I f'9g I 0'i1 I 9'91 I £'0S I 6 6£ - 01 OC-{£! I 9f r 91T i SOl I LLZ T Zf r 662 I S85 'I 66tt I 6fL1 I 909h I 9

9-Z I is I t' r £'0 1 I Z' t 2e 1 Z't I L' I 9' I L'Z I 9'-St I 6' 6'- 03 O'/-0, 0,1 I £' I S' I 9', 1 9' I L62 I L'h I 6'2 I 9'1t I 6'tl I O Ip9 I9b2f I DI I St I 12 I 6t T98 I SSt I £6 I 9L£ I L9fE I tLLt IZ

;i--------I-------- I-------- I--------........ I'--------IZ--------I----- ---I----b----I-------- I-ao sI I is I 11 I Z' I O, t O, I 2Z' I SI I bi I 6' I t'9 I ;am1 S

0 00 1 T I rs I t 1 M 1 Ill I.l t 2' I 9 6 I £bI b'£ 7 I £'9g I Z'9L IDlst I 92 I ht I L6 I h ! £ I 92 I B9 I iS I Til I 'ta2 I

I----^---I--------- I........I-------- ........ I....... MI-------- - --.... I-1--------I.... &-"-IZ-- - (ileaaIad)10001 1 S'1 I 2'2 I 0'! I 1' 1 9I9 I S I bT I L I ZI 2 t 9 I sjuaanbaX yAO£!I I £02 011OZ-

mob I q-jjvjuj -3I usa TUD

aTn3TPu9dxg P31Fdvo Tad jo aTTz)aa I u9am2aq;

V)t I I?S :OL-6961 ' S!NE0i' I £ a0i tXINI 7 I bOWI NaaPa1 aIo£a aIaOVINTO &I a V tIC N IU0' ! IS'2 I 2a'a ao 'Ia i 0a N Ioi¾vioa £oo xoi I 6i-isiai

VE aTqel

Table 3A(Continued)

Differencebetween COUNT :I

Caloric Intake ROW PCT I ROW

& Requirements COL PCT I TOTAL(percent) … 1 ! 2 I 3 . 4 S …I. 1 7 I e I 9 S to I- ........---- I---------I........ I ---- _----I----------- ---.....I------- ..-.-.... I-------- I----- ;--I -------- I

12 1 0 1 162 1 631 I 1025 I 1249 1 1630 1 1367 I 999 I 1342 I 919 I AM+20.1 to + 30.0 I 0 I 1,7 I 6,8 I 11,0 I 13,4 I 17,5 £ 14,7 I 10,7 I 14,4 I 9,9 SIlUv-

I 0 I 1,1 I 4,7 I 7,8 I 10,0 1 14,0 i 11,4 I 9,6 1 12,8 I 9,6 I 7.6a l------rw--9- .... _W--w---I--I---w---I-------- I---wt------ WJ-I------ -I-T------- I------- I.... "

13 I 33 7 65 1 159 1 514 I 606 1 781 i 104? I 1150 I 1027 I 772 I b!53A+30.1 to +40.0 I ,s I 1,1 I 2.L I 8,4 I 9,8 I 12,7 1 17,0 I 18,7 I 16,7 I 12,5 I 10

I ,2 I ,5 1 1,2 1 3,9 I 4,9 1 6 ,7 * 8,8 I 11,0 1 9,8 I 8,0 I 5.0* I'9 ...---I.......I..-.....I…-------…-I------- . '…-------"I….....I... …I.... ..-------- I

14 I 0 1 33 1 77 I 189 1 253 1 482 i 591 I 726 1 646 I 1126 I 4127 H+40.1 to +50.0 I 0 I .8 I 1,9 I 4,6 I 6,l I 11,7 1 14,3 I 17,b 1 15,7 I 27,3 I 100.0

I 0 1 ,2 I ,6 I 1,4 I 2,0 I 4,1 i 4,9 I 7,0 I 6,2 I 11,7 I 3.4... …….. I-.--------------I........I--------I--------I--------I--------I--.----. I

15 I 52 1 0 I 123 1 216 I 430 I 788 i 904 1 1824 1 2035 I 3033 I 9405+50.1 or more I ,b6 0 I 1,3 I 2,3 I 4,6 I 8,4 I 9,6 I 19,4 I 21,0 I 32,2 1 100.0

I ,3 I 0 I ,9 I 1,6 I 3,4 I 6,8 i 7,6 I 17,5 I 19,4 I 31,6 I 7.6* I-_----- .I…I---I-- .------I …----.I…-- -- .- _-I--____--I.__.. -1_

COLUMN 15311 14350 13549 13225 12479 11633 11948 10413 10499 9606 123012TOTAL 12 4 11,7 11,0 io,8 10,1 9,5 9,7 8,5 8,S 7,8 100,0

100.0 100.0 100.0 100.0 100.0 100.0 100.0 100.0 100.0 100.0 100.0

Note: In each cell, the first figure estimated shows the number of persons(in hundreds); the second (third) figure shows the number of personsin the cell as a percentage of the row (column) total.

- 20 -

The Indian data reveal the following picture:

Table 3B

DISTRIBUTION OF POPULATION BY PER CAPITA EXPENDITUREAND DIFFERENCE BETWEEN CALORIE INTAKE

AND REQUIREMENTS: INDIA

Difference Between Share of PopulationEnergy Requirement _ .and Average Intake Bottom Middle Topper Consumer Unit Third Third Third All

Rural 39.3 6.4 1.4 16.4< -30%Urban 47.2 13.5 2.8 20.6

Rural 56.8 75.0 41.0 57.9-30% < S 30%

Urban 51.4 79.3 68.4 66.7

Rural 3.8 18.6 57.6 25.7> 30%Urban 1.5 7.3 28.9 12.7

- 21 -

It is seen that in Sri Lanka as high as 34.0% of the population in the

top quintiles have calorie intakes exceeding requirements by more than 40%!

Similarly, 12.9% of the population in the bottom quintile have calorie

intakes falling short of requirements by more than 40%. In India, in the

top third of the population as high as 57.6% in rural areas and 28.9% in

urban areas have intakes exceeding requirements by more than 30%. In the

bottom third, 39.3% in rural and 47.2% in urban areas have intakes falling

short of requirements by more than 30%. If the energy requirements mean

anything at all, departures from these norms to this extent and in this large

a proportion of households should mean visible ill-health for these people.

There is no independent evidence for such a state of affairs. This brings

out sharply the inadequacy of using average calorie requirement as determinants

of nutritional status. Besides it may also reflect some inherent problems

in the use of survey data on consumption to which we now turn.

4. THE DATA BASE OF MALNUTRITION STUDIES

A few brief remarks on the data base of many of the malnutrition

studies are in order. The data base of most studies is a household estimates

of average annual consumption (per capita or per household), often at a national

and sometimes at the regional levels within a country. Given that income dis-

tribution is likely to be highly skewed, the absolute number of high income

households in the sample will be far fewer than lower income ones even with a

large sample size, so that the estimate of any characteristic is likely to be

subject to greater sompZing error for the upper income groups. Besides even

the inherent variation of the characteristic itself may be higher for upper

income groups though this is likely to be less of a problem for necessities,

like food. Some of the better designed surveys, such as the National Sample

- 22 -

Survey in India canvass independent sub-samples of households during different

sub-rounds spread throughout the year so that even if the data relating to

households covered in any sub-round are affected by seasonality (if the reference

period was part or whole of the sub-round), the estimate averaged over-all house-

holds taken in all the sub-rounds is free of seasonality. However, any analysis

that is based on individual household data will be affected by seasonality.

On the other hand, if the reference period is the whole previous year, the

seasonal effect disappears but recall biases enter. In countries such as

India and Pakistan, a large proportion of rural households grow their own

food, exchange food in kind, etc. and for some of these food items there is

hardly any market. Even though the investigators are instructed to record

consumption out of home-grown stock as well as out of gifts and exchanges in

kind, it is widely held that not all of this get recorded. Since many of the

poorer peasant households engage in these activities to a greater extent than

others, it is likely that a downward bias exists in their recorded food (and

hence calorie) intakes. At the other end, the food served by richer peasants

to their farm workers and guests often get recorded as their consumption so

that fantastically high calorie intakes (as high as 5000 Kcals. per consumer

unit per day) are observed.

There is another source of bias--even if the food item consumed by

a rich and poor household is called by the same name, the quality and price

per unit are often different. If the calorie intake is inferred from food

expenditures, the bias is obvious. Even if consumption data in physical units

are recorded, the quality differences affect calorie (and more often the

- 23 -

protein) content-for instance, the rich consume more processed and

possibly less nutritious foods than the poor. To the extent the various

biases tend to result in making the calorie intake-income relationship

steeper than it really is, removing such biases would make the estimate

of the extent of malnutrition obtained by applying the R-S procedure even

more sensitive to the variations in the average calorie requirement than

the already high sensitivity recorded by them.

5. SOME POLICY IMPLICATIONS AND ISSUES FOR FURTHER RESEARCH

To summarize the discussion so far: The "energy" and "protein"

norms published by the FAO have been widely misused for assessing the

nutritional status of a population inspite of warnings against such use

by the FAO itself. It was argued that the process governing variations in

daily energy balances in human beings is not yet fully understood and a

satisfactory theoretical framework for it is yet to emerge. Nevertheless,

there is some evidence that the process is stochastic-stationary and auto-

regressive and there is a wide variability in energy intakes between otherwise

similarly situated healthy individuals and even in the same individual over

time. Many of the widely used procedures for assessing nutritional status,

by classifying all individuals in a population as malnourished who have

intakes below a single average norm for the population as a whole, thereby

ignoring intra and inter individual variance in intakes and requirements,

will misclassify individuals (i.e. adequately nourished as malnourished and

vice-versa) to varying degrees. This misclassification bias need not cancel

out for the population as a whole and there is a danger in overestimating the

proportion of truly malnourished.

_ 24 -

It is important for devising suitable policies for eliminating

malnutrition where it exists, that the extent of it is not overestimated

and exaggerated and the malnourished get properly identified as a target

group. For such overestimation will only make the problem appear even more

intractable than it already is. And inaccurate targeting will result in

a waste of scarce resources. The process of probing into the true extent

of malnutrition is likely to reveal that it is largely among households

who are not only poor but suffer other social disadvantages and dis-

crimination, such as the untouchable and tribal households in India, or

similarly disadvantaged groups in North East Brazil. Further, they often

have limited access to safe drinking water, sanitation and health and

educational facilities, since they are consigned to urban slums and rural

ghettos. Attempts to tackle their nutritional problems alone in such a

context are almost futile. Many such programs have failed in the past.

Given their extreme poverty and their tenuous link with the rest of the

economy through inadequate and uncertain employment opportunities, it is

not surprising that any shortfall in aggregate food supply finds them priced

out of the market. Integrating them more fully into the economy by providing

them adequate and secure income earning opportunities, while at the same

time making a determined effort at reducing their formidable social dis-

advantages will go further in tackling their poverty and nutritional problems

than nutrition interventions per se. Given the known inequalities in the

intra-family allocation of food in poor families, calorie availability has

to be greater than each family's requirement so as to ensure that every member

gets his due share. An implication of this argument is that an individual

- 25 -

can consume more than his requirements indefinitely without adjusting his

bodily functions and activity. The above discussion on the regulatory

mechanism governing energy balance shows this to be fallacious. Similarly

by constuning more than one's requirement in good years, one could not avoid

malnutrition in bad years, though maintaining health and nutritional status

would undoubtedly improve one's capacity to sustain mild deprivation.

According to Payne and Dugdale (1977b), in subsistence agricultural

societies the pattern of eating is often of feasts in good times and fasts

in bad times, so that the ability to store excess energy efficiently and

release it when needed has survival values. As such individuals' with such

genetic capability have an advantage. But it is bizarre to go from this to

argue that in modern times, fluctuations in food availability should be

evened out by inducing people to eat excessively and put on body weight

during good years and lose it during bad years! Even the least efficient

developing country should be able to design and manage a better buffer stock

arrangementl

While the interaction between nutritional deficiency and the health

status of an individual runs both ways, the policies that reduce morbidity

such as provision of safe drinking water, sanitation facilities and a gen-

erally healthy environment may be the appropriate way to break into this

reaction chain, with policies for nutrition intervention per se following

if needed. And since severe malnutrition is predominantly a symptom of

extreme poverty, the importance of policies that augment the income earning

opportunities (and their growth over time) of the poor and socially dis-

advantaged can hardly be overemphasized.

- 26 -

References

Dandekar, V. M. and N. Rath (1971): "Poverty in India," Economic and PoZiticaZWeekly, Bombay, Vol. 6, Nos. 1 and 2 (January 2 and 9). Laterissued as a pamphlet with the same title by Saneeksha Trust, Bombay.

Food and Agricultural Organisation (1950): "Report of the Committee on;Calorie Requirements," FAO NutritionaZ Studies, No. 5.

------ (1957): "Report of the Second Committee on Calorie Requirements,"FAO NutritionaZ Studies, No. 15.

------ (1973): Energy and Protein Requirements, Report of a Joint FAO/WHOAd Hoc Expert Committee, Rome, FAO.

------ (1974): Handbook on Human NutritionaZ Requirements, Rome, FAO.

------ (1975): "Energy and Protein Requirements: Recommendations by aJoint FAO/WHO Informal Gathering of Experts," Food and Nutrition,1 (2).

------ (1977): The Fourth WorZd Food Survey, Rome, FAO.

------ (1978a): Report of the First Joint FAO/WHO Expert Consultation onEnergy Intake and Requirements, Danish Funds-In-Trus, TF/INT 297(DEN), Rome, FAO.

------ (1978b): "Requirements for Protein and Energy: An Examination ofCurrent Recommendations," Report by a Group of Consultants, Rome,FAO (mimeo).

McCarthy, F. D., W. M. Rand and L. Taylor (1978): "How Wide is the CalorieGap," Mimeographed, Department of Nutrition and Food Science,and International Nutrition Policy and Planning Program, Massachu-setts Institute of Technology, Cambridge.

National Sample Survey Organization (1976): CaZorie and Protein VaZues ofFood Items Consumed Per Diem Per Consumer Unit, Report 238, re-lating to Twentysixth Round (July 1971-June 1972), Vol. I (RuralAreas) and Vol. II (Urban Areas), Government of India, Departmentof Statistics, Ministry of Planning.

Ojha, P. D. (1970): "A Configuration of Indian Poverty," Reserve Bank ofIndia BuZletin, January.

Payne, P. R. (1977): Review of Malnutrition and Poverty, World Bank,Ocasional Paper, No. 23, in Food PoZicy, Vol. 2, No. 2, May 7.

------ and A. E. Dugdale (1977a): "A Model for the Prediction of EnergyBalance and Body Weight," Annats of Human Biology, Vol. 4, No. 6.

------ and ------ (1977b): "Pattern of Lean and Fat Disposition in Adults,"Nature, Vol. 266, No. 5600.

- 27 -

Reutlinger, S. and M. Selowsky (1976): Malnutrition and Poverty, World Bank,Occasional Paper, No. 23.

------ (1977): Letter to the Editor of Food PoZicy, Vol. 2, November.

Scandizzo, P. and J. Graves (1978): "The Alleviation of Malnutrition: Impactand Cost Effectiveness of Official Programs," Working Paper No. 19 ofthe Economic Policy Division of Agriculture and Rural DevelopmentDepartment, The World Bank.

------ and 0. Knudsen (1979): "Nutrition and Food Needs of Developing Countries,"Staff Working Paper No. 328, The World Bank, May.

------ and ------ (1978): "The Evaluation of Benefits of Basic Needs Policies,"Working Paper No. 18 of the Economic Policy Division of Agricultureand Rural Development Department, The World Bank.

Sukhatme, P. V. (1977): "Malnutrition and Poverty," Ninth Lal BhaduriShastri Memorial Lecture, Indian AgricuZturaZ Research Institute,New Delhi.

------ (1978): "Assessment of Adequacy of Diets at Different Income Levels,"Economic and PoZitical Weekly, Bombay, Special Number, August.

------ (1979): "Nutrition in India in Current Five Year Plans," ThirteenthConvocation Address, Indian Statistical Institute, Calcutta, March.

------ and S. Margen (1978): "Models for Protein Deficiency," The AmericanJournaZ of ClinicaZ Nutrition, 31, July.

Visaria, P. (1979): "The Incidence of "Absolute" Poverty in Sri Lanka,1969-70," Joint ESCAP-IBRD Project on the Evalutaion of AsianData on Income Distribution, Working Paper No. 6, June (mimeo).

Widdowson, E.M. (1962): "Individual Variation," Proceedings of the NutritionSociety of London.

RECENT PAPERS IN THIS SERIES

No. TITLE OF PAPER AUTHOR

343 The Capital Goods Sector in LDCs: J. Datta MitraA Case for State Intervention

344 International Technology Transfer: F. StewartIssues and Policy Options

345 Family Planning Programs: An Evaluation of R. CucaExperience

346 Prospects for Traditional and Non-Conventional D. HughartEnergy Sources in Developing Countries

347 National Urbanization Policies in Developing B. RenaudCountries

348 Private Direct ForeiRn Investment in K. BillerbeckDeveloped Countries Y. Yasugi

349 Adjustment Policies and Problems in Developed M. WolfCountries

350 Energy Options and Policy Issues in Developing D. Fallen-BaileyCountries T. Byer

351 Growth and Equity in Semi-Industrialized Countries J. Bergsman

352 Capital Flows and Developing Country Debt D. Keesing

353 Trade Policy for Developing Countries D. Keesing

354 Development Problems of Mineral-Exporting G. NankaniCountries

355 The Global Framework R. Cheetham, S. GuptaA. Schwartz

356 The Distribution on Income in Brazil G. Pfefferman, R. Webb

357 Estimating Shadow Prices for Colombia in an W. SchohlInput-Output Table Framework

358 Inter-Country Comparison of "Real" (PPP) Incomes: P. IsenmanRevised Estimates and Unresolved Questions

359 Price Distortions in Agriculture and Their Effects: E. LutzAn International Comparison M. Bale

360 Costs and Benefits of Agricultural Research: G. SchuhThe State of the Arts H. Tollini (consultants)

- 2-

No. TITLE OF PAPER AUTHOR

361 Investment in International Agricultural Research: G. ScobieSome Economic Dimensions (consultant)

362 Identification and Appraisal of Rural Roads H. BeenhakkerProjects A. Chammari

363 Small Enterprises in African Development: J. Page (consultant)A Survey

364 Income, Consumption and Poverty in Thailand, 0. Meesook1962/63 to 1975/76

365 A Survey and Critique of World Bank Supported R. MarrisResearch on International Comparisons of Real (consultant)Product

366 Paradigms in the Study of Urban Labor Markets D. Mazumdarin LDCs: A Reassessment in the Light of anEmpirical Survey in Bombay City

367 Incentives for Resource Allocation: S. AcharyaA Case Study of Sudan

368 Why the Emperor's New Clothes Are Not Made in D. MorawetzColombia (consultant)

369 Economic and Social Analysis of Projects andof Price Policy: The Morocco Fourth Agricultural K. CleaverCredit Project

370 The Tokyo Round and the Developing Countries B. Balassa

371 Bus Ownership and Efficiency in Urban Areas A. Walters

372 The Tokyo Round: Results and Implications ' ' (consultant)for Developing Countries

HG3881.5 .W57 W67 no.373 c.3Srinivasan, T. N., 1933-Malnutrition, some measurement|

and policy issues /

DATE NAME AND EXTENSION ROOMjuw nlw ' - ' ' NUMBE