“ CONCEPTS OF HUMAN DEVELOPMENT AND POVERTY : A
MULTIDIMENSIONAL PERSPECTIVE “
16595
AUTOR : Anand, Sudhir, 1946-
Sen, Amartya
FOTOCOPIADO DE : Human Development Papers 1997--.New York: The
United Nations Development Programme : Oxford
University Press, 1997. p. 1-19.
CATEDRA : Economia de las Politicas Sociales
PROFESOR : Osvaldo Larraiiaga
SEMESTRE : Otofio’2006
“ USO EXCLUSIVO ALUMNOS FACEA , PARA FINES DE DOCENCIA E
INVESTIGACION “Concepts or HuMAN DEVELOPMENT AND POVERTY:
A MUuLtipiMenstonaL PERSPECTIVE
Sudhir Anand and Amartya Sen
human development has been'much discussed in the previous Human
Development Reports. This paper is concerned with a different distinc-
tion: that between (1) concentrating specifically on the living conditions of the
poor, and (2) the more traditional broader approach of looking at the condi-
tions of all the people in the society, The Human Development. Index (HDI)
belongs to the latter —- more inclusive — approach. As Human Development
Report 1997 is aimed at the conditions specifically of the poor, there is a need
for a more focused approach concentrating specifically on the lives of the
poor.
T: contrast between a commodity-focus view and the approach of
Human Development and Human Poverty
The process of development in the contemporary world can be scen in two
rather different ways. It can be characterized in terms of the progress being
made by different groups in each community, Putting weight on the fortunes of
the rich as well as the poor, the well-provided as well as the deprived. This
“conglomerative perspective” can be contrasted with an alternative, more
specialized, viewpoint in which development is judged by the way the poor
and the deprived, in particular, fare in each community. This “deprivation
Perspective” is specifically concerned with those who are forced to live
deprived lives. In this accounting of the process of development, lack of
Progress in reducing the disadvantages of the deprived cannot be “washed
away” by large advances — no matter how large — made by the better-off
people.
There are reasons for taking an interest in the process of development
1from both the perspectives. At a very basic level, the lives and successes of.
everyone should count, and it would be a mistake to make our understanding
of the process of development completely insensitive to the gains and losses of
those who happen to fare better than others. It would go against the right of
each citizen to be counted, and also clash with the comprehensive concerns of
a universalist ethic. Yet a part — a big part — of the general interest in the
progress of a nation concentrates specifically on the state of the disadvantaged.
‘The reasoning behind public concem with development relates particularly to
the advancement of those who are less privileged than others and who are
forced to live distinctly reduced lives.
This deprivational focus contrasts sharply with the conglomerative
perspective, in which even a sharp regress in the conditions of the poor can be,
quite possibly, outweighed by a suitably large surge in the fortunes of the
affluent. To see a “great advance” — on aggregative grounds — even as the
devil takes the hindmost cannot really be very convincing. The rationale of the
present paper relates to this general recognition.
The conglomerative and the deprivational perspectives are not, in fact,
substitutes of each other. We need both, for an adequate understanding of the
process of development. The plurality of our concerns and commitments
forces us take an interest in each. The specialized focus of the deprivation
Perspective is needed to supplement — not to supplant — the universalist
solicitude of the conglomerative approach.
The Human Development Reports, which have been published every
year since 1990, have been generally concemed with the removal of disadvan-
tages and the creation of opportunities to lead worthwhile lives. Right from the
beginning, these reports have taken a special interest in poverty and depriva-
tion, and have tried to provide a wide range of information — in various tables
and commentaries — on relevant features of the misery and restricted lives of
the worst off. However, the specific index of progress in the form of Human
Development Index (HDI) has taken a conglomerative perspective. For exam-
ple, a rise in the life expectancy of any group — no matter how affluent — is
reflected in a corresponding increase in the average life expectancy of that
nation, and this in tum enhances the HDI, since average life expectancy is one
of the constitutive components of HDL The particular approach of “human
development”, with its focus on human lives and the quality of living, with
which these reports have been concerned, can of course be fruitfully combined
with the deprivational perspective. While that connection has been explored in
terms of particular investigations of the conditions of impoverished people, it
has not, up to now, been reflected in any specific index of the burden of
deprivation.
2 Human Development Papers 1997Focusing on the conglomerative HDI has not, of course, been pointless,
even in understanding the nature of deprivation. The disadvantages of the
worst off have inter alia figured in the index. Given its general concentration
on the quality of life and substantive opportunities of all people, the HDI has
played a role in broadening evaluative attention from such gross measures of
economic progress as the gross national product (GNP) per head (or the
average national income) to lines of analysis that are sensitive to the removal
of deprivation of different kinds as well as expansion of opportunities in
general. Indeed, the reports have contributed to a substantial change in the
nature of public discussion and debate on the successes and failures in the
process of development. The focus has been on development in a very broad
sense, not just on the expansion of real income per head, but on the enhance-
ment of some. of the central features of the quality of life of those who lead
diminished — and often physically shorter — lives.
Nevertheless, the possibility of reflecting in a usable and uncompli-
cated index the bearing of “human development” on the “deprivational per-
spective” — concentrating specifically on people who are particularly de-
prived — has not been, so far, pursued. To undertake that task would require
the development of an index of “human poverty”, which would focus exclu-
sively on the specially deprived and impoverished. This is part of the exercise
undertaken in the 1997 Human Development Report, and calls for a Human
Poverty Index (HPI). In devising such an index, it is important that the purpose
of its contribution and the motivation behind it are properly understood.
In particular, some possible pitfalls must be avoided. First, the Human
Poverty Index HPI must not be seen as a substitute for the Human Develop-
ment Index HDI. As has been already discussed, both perspectives — depriva-
tonal and conglomerative — have their own interests and complement each
other. Perhaps the motivational distinction between the HDI and HPI can be
explained with an analogy, drawn from the income-based evaluation of
economic success. The growth rate of GNP per head gives an account of
Progress seen in the conglomerative perspective — everyone’s income counts
in the GNP total. In contrasi, the reduction of an income-based poverty index
(such as the diminution of the proportion of the population below the poverty-
line income), while also based on income information, uses the deprivational
perspective, concentrating specifically on the incomes of the poor (in contrast
with GNP's interest in the incomes of all). Within the income-based perspec-
tive, it would make little sense to argue that since GNP is already based on
income information, any income-based poverty measure must be a substitute
for the GNP. Nor would it be sensible to suggest that the availability of the
GNP as an indicator makes it redundant to seek a measure of income poverty.
Concepts of Human Development and Poverty . 3Table 1
Income information Information on human living
Conglomerative {GNP per head Human Development Index
Deprivational —_| Income Poverty Measures Human Poverty Index
The GNP and the income poverty measures use the income information in
different perspectives, with the GNP taking a conglomerative view, while the
income-poverty measures focus specifically on the income-poor.
The relationship between HDI and HPI has to be seen in a similar way.
Both have to use the rich categories of information that are associated with
“human development”: characteristics of human lives and quality of living that
go much beyond what income information can provide. But while these
characteristics are used’ by HDI in the conglomerative perspective, the HPI
must use them in the deprivational perspective. The availability of GNP
measures does not obviate the need for an income-based poverty indicator
(using income information), and similarly, the presence of the HDI measure
does not eliminate the need for a Human Poverty Index (HPI) (using informa-
tion related to the deprivation in human lives).
The relationship between these different measures can perhaps be
usefully seen in the form a table (above) where the rows distinguish between
the conglomerative and deprivational perspectives, while the columns discrim-
inate between the use of income information and that of characteristics of
living associated with human development.
A second misunderstanding to avoid relates not to the relationship
between the HDI and the new HPI, but between income-based poverty
measures and the new HPI (that is, the contrast between the columns rather
than the rows in Table 1). Both the HPI and the income-poverty indicators
share the deprivational perspective, but while the latter sees nothing in poverty
other than low-ness of incomes, the HPI must take a much broader view, in
line with the approach of human development. It would, in fact, be useful to
see how the values and rankings of HPI relate to the results of income-based
poverty analysis.
Multidimensionality of Poverty
Poverty is, in many ways, the worst form of human deprivation. It can involve
not only the lack of necessities of material well-being, but also the denial of
Opportunities of living a tolerable life. The lives could be prematurely short-
4 Human Development Papers 1997ened, made hard, painful or hazardous, deprived of understanding and commu-
nication, and robbed of dignity, confidence and self-respect. It is ultimately in
the poverty of the lives that people can lead that poverty manifests itself.
Income-based poverty measures concentrate exclusively on deprivation
in one variable in particular, viz. income. It has the advantage of simplicity in
refraining from taking an interest on different aspects of deprivation. A
similarly rudimentary approach cannot be used in developing an index of
“human poverty”, since the lives of human beings can be blighted and
impoverished in quite different ways. Someone can, for example, enjoy good
health and live quite long, and yet suffer from being illiterate and remain cut
off from leaming as well as communication and interactions with others that
rely on literacy. Another person may be literate and quite well educated, but in
fact be particularly prone to premature mortality because of the epidemiologi-
cal characteristics of the region or country. If illiteracy were our only criterion,
the first person would be seen as deprived, but not the second, whereas with
proneness towards premature mortality as the only criterion, the second would
be seen as handicapped, but not the first. But both are, in important ways, quite
deprived, and it would be a mistake to concentrate only on one of the two types
of information and ignore the other. When we consider several other ways in
which a person can be severely deprived, we find other dimensions of
disadvantage.
This line of reasoning also throws light on the inadequacy of income-
based poverty measures. A person with an above-poverty-line personal income
may still be deprived in the sense of being illiterate, or being subject to
epidemiological vulnerabilities that can lead to premature mortality, or being
without crucial services (such as safe water or health care) which are often
provided by public provisioning (going well beyond what is purchased by
personal incomes). The need for a multidimensional view of poverty and
deprivation not only guides the search for an adequate indicator of human
poverty, it also clarifies why an income-based poverty measure cannot serve
the sarne purpose.
The HPI must, therefore, use a multidimensional procedure in assess-
ing poverty. This is entirely in line with the “human development” approach.
The Human Development Index examines different features of the quality of
life and arrives at an aggregate judgement on the extent of overall human
development. In the same way, the Human Poverty Index must examine
different features of deprivation in the quality of life, and then arrive at an
aggregate judgement on the extent of overall deprivation of the impoverished.
In getting an adequately broad view of deprivation (including its different
aspects), this multidimensionality is important and inescapable, and despite
Concepts of Human Development and Povert . 5
za ythe complications of dealing with several variables at the same time, taking a
multidimensional approach must, ultimately, be seen as an asset rather than a
liability.
This is not to deny that any reduction of a multidimensional indicator
into a numerical index, such as the HPI, must involve an exercise in weighting.
The exercise of specifying usable weights supplements the evaluative task
already involved in making a selection of the particular dimensions of depriva-
tion on which to concentrate. These evaluative decisions have to be guided not
only by practical considerations of measurability and data availability, but also
by the reading of the prevailing values as to what counts as deprivation and
how this is to be recorded, Since any choice of weights should be open to
questioning and debating in public discussions, it is crucial that the judge-
ments that are implicit in such weighting be made as clear and comprehensible
as possible, and thus be open to public scrutiny. After proposing a particular
system of weighting, we must supplement it by a thorough and transparent
investigation of its properties and implications.
Context Dependence and Deprivation in Poor Countries
The identification of principal deprivations tends to vary with social and
economic conditions of the community in question. Premature mortality is an
extensive problem in impoverished countries in a way it may not be in highly
developed countries where most people live quite long. Also, illiteracy may be
a frequent form of significant deprivation in many countries with educational
backwardness, but not in others. The choice of the indicators to be used in the
Human Poverty Index cannot but be sensitive to the context of the evaluation,
and in particular to the characteristics of the countries for which this index is
primarily intended. For example, an index that concentrates on illiteracy and
premature mortality may be able to discriminate between, say, Sri Lanka and
Pakistan much more easily than it can between, say, Germany and France.
To some extent this problem is also present for the Human Develop-
ment Index itself, and the case for having more specialized indicators in the
conglomerative perspective may well be worth considering. But the difficulty
is particularly immediate and powerful in characterizing poverty and in
examining the condition of the deprived. Issues of poverty in the developing
countries crucially involve such matters as hunger, illiteracy, epidemics, lack
of health service or of safe water, but these deprivations may not be at all
common in the more developed countries, where hunger is rare, literacy is
close to being universal, most epidemics are well controlled, health services
are typically widespread, and safe water is easy to access. Not surprisingly,
studies of poverty in the more affluent countries have tended to concentrate on
6 Human Development Papers 1997other variables, such as social exclusion, or inability to take part in the life of
the community. These too can be quite forceful sources of deprivation of
human lives and very hard to eliminate, but they involve different types of
deprivation. There is no real possibility of getting an index of human poverty
that would be equally relevant in the different types of countries.
Given the peculiar importance of poverty in poor countries, the HPI to
be presented here is aimed at that context, The variables chosen reflect that,
The nature of poverty in rich countries deserves a separate study — and a more
specialized index, focusing on those deprivations that are particularly relevant
for these countries. This motivational point has to be borne in mind in
interpreting and examining the HPI, in the light of its context,
Deprivation in Survival and in Education
Of the three basic components of the Human Development Index, two deal
respectively with survival aiid education. The aspect of survival in the quality
of life is captured in the Human Development Index through the use of life
expectancy at birth. This draws on the age-specific mortality rates that cur-
rently obtain. A weighted average of survival years is calculated, using this
information. Different people do, of course, die at different ages, but the
pattern of ages at death can be calculated by using the prevailing age-specific
mortality rates, and the life expectancy gives the statistical expectation of an
average person's length of life at the prevailing mortality rates,
The average value of life expectancy uses a firmly conglomerative
approach, and is an average over the entire community. The deprivational
perspective applied to survival makes us concentrate on vulnerability to death
at a remarkably carly age. By using the same age-specific mortality rates, it is
possible to calculate the proportion of people who can be expected to die
before a specified age (such as 40 years), at the prevailing rates. That
proportion is a measure of vulnerability to having a substantially shortened life
and reflects the incidence of serious deprivation in terms of length of life.
Turning now to education, it is reasonable to argue that illiteracy is
indeed a strikingly sharp educational deprivation. The proportion of illiteracy
expresses the incidence of this kind of deprivation in the population. The
illiteracy rate can, thus, be used for the educational component in deriving a
human poverty measure HPI. The consequent narrowing from the broad
coverage of the educational component in the Human Development Index to
the specific concentration on illiteracy in the Human Poverty Index should be
an appropriate move.
‘There is an issue of data availability and reliability which also deserves
attention here. The narrowing of the education focus from a general educa-
Concepts of Human Development and Poverty . 7
"pmtional index in the HDI to just illiteracy in the HPI will probably be an
improvement in terms of data access and quality, since, comparatively speak-
ing, the literacy data tend to be better than the data on enrolment at educational
establishments at different levels.’
So far as survival is concerned, the calculation of life expectancy and
that of non-survival to a specified age (such as 40 years) have to draw on the
same informational base, viz. the prevailing age-specific mortality rates. The
latter needs a little less, since it is independent of mortality rates beyond the
specified age (40 years). However, doubts about the quality of mortality
Statistics that are available apply to both, and this should be taken into account
in making use of the results of the HPI calculation (just as it is relevant for
using the HDI results as well).
Deprivation in Economic Provisions
The third component ‘of the Human Development Index HDI is based on
income, and concentrates on per-capita national income or GNP. It is used at
an aggregative level for each country, and reflects not only the average
personal incomes of individuals but also the provision of public services (such
as public health care) paid out of the aggregate national income. The income
component of HDI is, thus, an amalgam of private and public facilities,
attempting to reflect overall economic provisioning. In adapting this aspect of
living standard in the measurement of poverty, the focus has to be disaggrega-
tive and oriented towards individual life experiences. In line with the general
idea behind the use of the income variable in HDI (as an indicator of overall
economic provisioning), the HPI must pay attention to personal affluence as
well as public services.
One of the problems in assessing the prevalence of income poverty is
that the use of the same cut-off poverty line in different countries can be very
misleading. Adam Smith noted in the Wealth of Nations, as early as 1776, that,
depending on the prevailing consumption patterns in the respective societies,
the minimum income needed to achieve the same elementary freedom from
deprivation would vary from one community to another. In Adam Smith's
example, a person who cannot afford to buy a linen shirt or leather shoes in
England may well be ashamed to appear in public, but that inability may not be
seen as a crucial deprivation in a society where.linen shirts or leather shoes are
not widely worn.
This variation of “necessary” commodities between one society and
another makes the use of the same poverty line in different countries (with
diverse levels of general opulence and disparate consumption patterns) ex-
tremely deceptive. Depending on the prevailing patterns of observable con-
8 Human Development Papers 1997sumption (of clothing, accommodation, means of communication and interac-
tion such as radios, televisions, telephones, etc.), many provisions are taken to
be essential for social participation in one community without being treated to
be so in another. As a result, the minimum income needed to escape social
estrangement can be quite different in different communities. Given the social
pressure generated to fulfil these felt “needs”, they may compete — for
relatively poor people — even with the provision of resources for food,
nutrition and health care, which can be significantly squeezed — often for the
children — as a result of this socially induced influence (and a corresponding
sense of cultural compulsion). This can, in fact, help to explain the observed
presence of hunger and undernourishment even in societies such as the United
States in which the general income levels are very high, but where inequalities
generate a heavy burden of “necessity” in the direction of socially obligated
consumption, often to the detriment of nutritional and health expenditure.
‘Thus, the assessment of poverty on the basis only of the level of income, with
the same cut-off point in different societies, can record little or no poverty in
generally affluent societies, even when the relatively poor in these societies
may lack social participation and — more starkly — may suffer even from
hunger and undernourishment (because of the competing pressure of “social
necessities”).*
An altemative is to use different poverty lines in different countries.
But it is not easy to decide what the appropriate variations would be and how
the respective poverty lines could be estimated. The officially published
national “poverty lines” cannot, obviously, serve this Purpose, since those
variations reflect other influences, particularly the impact of politics and
Pressures. The general need for a variable cut-off line of poverty is easy to
appreciate, though it may be hard to find an adequate procedure for determin-
ing the appropriate values of a variable poverty line in different communities.
A more practical possibility in dealing with the component of eco-
nomic deprivation in the HPI is to choose to be less ambitious and to
concentrate on material deprivation in the form of hunger and undernourish-
ment in particular, rather than on income in general. Since a very high
Proportion of personal income is expended on acquiring food and nourish-
ment, this is not a tremendously severe departure for poor countries. We can
either use information on the intake of food and nutrition, or go by estimates of
Prevailing undernourishment. The former relates more closely to personal
incomes, but the latter incorporates also the influence of other variables that
affect nourishment, such as metabolic rates, climatic conditions, activity
patterns and epidemiological circumstances. Since our ultimate concern is
with the nature of the lives that people can lead, there is a case for going
Concepts of Human Development and Poverty : 9straight to the prevalence of undernourishment, rather than to the intake of
calories and other nutrients.’ This is what has been done in the Human Poverty
Index, concentrating specifically on the undernutrition of children, which is
easier to diagnose (by such criteria as weight for age) and on which useable
data are easier to get.
So far as specifically public provisions are concerned, access to health
service and to safe water have been chosen as the indicators to be used. The
absence of access to health service or the unavailability of safe water can
indeed be seen as a significant deprivation. Combining these two access
variables with the prevalence of undernutrition, we get a fairly broad picture of
the deprivation of economic provisioning — private and public ~~ to supple-
ment the information on survival and literacy.
These, then, are the basic informational ingredients of the HPI. It must
be emphasized that there is some inescapable arbitrariness in any such choice.
The choice was made on the basis of balancing considerations of relevance, on
the one hand, and the availability and quality of data on the other. There are
inevitable compromises here. Even though some variables could not be used
because of the quatity of data seemed particularly bad (information on access
to sanitation was a variable that was considered seriously, but could not
ultimately be used given the limitations of the available data), it would be idle
to pretend that even the variables that have been included have high-quality
data for every country. There has been an attempt, in these selections, to strike
a balance between the demands of relevance and the need for tolerably useable
data, and these choices would certainly remain open to criticism and public
scrutiny.
Weighting and Aggregation
There are two distinct problems of weighting in moving from this diverse and
multidimensional informational base to a combined index of human depriva-
tion in the form of HPI. While two of the three components to be dealt with
have clear numerical values (viz. proportion of population expected to die by
the age of 40 and the proportion of illiteracy), the description of economic
deprivation includes disparate elements in the form of undernourishment, lack
of access to health care and lack of access to safe water.
These sub-components are, however, themselves numerically specific,
and we can identify the percentage of population without access health care
(h), the percentage without safe water (w), and the percentage of children who
are undernourished (n). In the absence of any clearly agreed way of discrimi-
nating between the respective importance of these three sub-components, the
economic provisioning component has been chosen to be simply the mean of
10 ‘Human Development Papers 1997these three deprivations, that is the unweighted average of these three percent-
age deprivations: ¢ = (1/3)[h+w+n]. It should be pointed out, to make the
scrutiny of the implicit assumptions easier, that this procedure amounts to
assuming that the three deprivations act additively and have the same relative
values vis-a-vis each other throughout.
There is now the further problem of moving from this three-
dimensional indication of human poverty to a numerical index of human
poverty HPI. The three components are reflected respectively by:
(1) survival deprivation (s), given by the expected incidence of mortality by
age 40 (that is, the proportion that would be expected to die before becoming
40 years old at the current age-specific mortality rates);
(2) deprivation of education and knowledge (k), given by the percentage of
people who are illiterate;
(3) economic deprivation (e), given by the mean of the three sub-components
already discussed,
While these three components of human poverty are all important, it is
not unreasonable to assume, given their dissimilarity, that the relative impact
of the deprivation of each would increase as the level of deprivation becomes
sharper. For example, as we consider higher and higher percentages of people
who may perish before the age of 40, this deprivation will become more and
more intense per unit, compared with other deprivations. This is, of course,
equivalent to assuming that as the incidence of early death diminishes (and,
correspondingly, as there is more plentiful survival prospects), the relative
value of this consideration will diminish.
A simple, symmetric and regular way of reflecting this requirement is
through the formula:
H = [(1/3)s" + (1/3)k° + (1/3)e"]",
if we choose a value of a more than 1." This is in line with a general procedure
we had recommended for constructing indices in earlier Human Development
Reports (see, for example, Anand and Sen 1995 on the construction of
gender-sensitive development index, GDI), and this general approach has
indeed been used in Human Development Reports already (see, for example,
UNDP 1995). The procedure is discussed in. the accompanying Technical
Note, which also examines the properties of this type of weighting and their
implications.
Concepts of Human Development and Poverty . itTechnical note 1. Properties of the human poverty index
‘This techncd not states, establishes and dizcuses some important properties
ofthe human porn index Iecnded a a ad o understanding the inde,
these properties are derived with respect to 4 more genera definition of the
hunan poverty index Pf) than tht actly wed in this Report. This allows
the pons tht the weights onthe tree poverty sbindices may fe, 0
that P(a)is «weighted mean of order a of P, , and P
‘Ths eting > Obe the weight on P20) ford 1, 2,3, we define the
severlited mean Pla) as
w ry-(stkeateent)
‘The weighted mean reduces to the ordinary mean of order auwhen w= 1 for
very, Wiha = tt, = 1, we have sims
ele Be Chl
‘The mean of order 1 (a = 1) the simple weighted or unweighted arth:
metic mean ofP, Py and P,. Thus
roe Eater
sLer en) where, = 1 os
Can the human poverty index P(a) be interpreted as headcount oc inci>
dence of poverty? While P,P, and Pare the headcount orincidence of porerty
in cach of three separate dimensions, P(a) cannot be generally thought of #5
the headcount ratio with respect toa povery line (hyperplane) drawn in the
prodhct space of the three variables. Instead, P(a) isan average albeit of order
4, of the thee subindices P,P, and P,. If the incidence of poverty happened
tobe thesameinevery dimension, then P(a) would clestlybe equal to this eom-
‘mon aumber, since
[=e +u,Pla)* + w, Play
= Pla) +{ shes wyPs 6 wsPe r
‘This observation allows ust interpretP(a) asthe degree of overall poverty that
is equivalent to having « headcount eatio of P(a)% in every dimension.
‘The first property of Pa) that we establish is central to understanding is
samean of P,P, andP,, Thispropertyis that P(a) alway esberween the all-
ext and largest values of P, for = 1,2, 3.
‘Paoposrriok
sin .8, 8] Pa ma 2,8,2
PROOF. By definition of Pla), we have
waa
0) Pat = tw, tw, ate ew,
12
But foreach = 1,2,3,
min (R.%.R} sR < male. 885)
ects >
[ref
Using the rght-hand-side inequality for each Pin equation 3 gives
sf em oat
ay st FF a
roe fray
Similarly,
Pay o[min fr rai
fois {2.%,0)f tar s[maxle. 2.7}
Sree > 0, follows ht
win 7,8} sRensmale,.2.8) 0
‘The generalized mean Pla) is constructed for valves of 2
limiting value when @ and P,.
proposition 7 we show that the lrg i, the larger Pt) wil be. For expos
tional reasons, iis convenient to demonstrate st this stage that ov tenls
Infinite limiting value of P(g) is mex (P,, Py, Py)
Prorosmox 2, Ave,
Proprius}
PROOE. Let P, be the largest—or in the case of tes, one ofthe lagest—P,
fori = 1,2, 3. Thus
cax{t. 2.24)
‘Then from proposition 1, for any & > 0, we have
w ads, aunt, %.8)
Now
Pe era art area
tothe since P, is one of ,,Pyor?,,
Human Development Papers 1997‘Therefore, since @ > 0,
Letting a => =,
sothat lim PladzP,.
Dat or equines have
ati fod
ene
ig 007, «ne .7,.8)
homo:
the ine
“The next property of Pla) that we demonstrate that the indes
geneous of degree {inthe subindices P,P, and Py. In other words,
dence of poverty in each dimension is haved (mukipled by A > C), the value
ofthe aggregate index Pa) will be halved {changed to hmipled by Pla)
PRoposmnon 3. P(a) ishomogencous of degree Lin (Py, PP)
Poor, Let > 0 be sclar number, and et P(a) be the value ofthe human
poverty index corresponding ta (P,.
Then
‘The value of the human poverty index corresponding to (IP, LP, AP, Jig then
venby
| ee ea [seen
ara || pay
=a. Oo
‘The next propesty of P(a) that we derive i that P(Q) fe manetonic increas:
ing ineach P, ford 1,2, 3.
ProFOsttion 4, For each =1,2,3,
an)
®
>a
Proor. From the definition of the generalized mean Pa) we have
(ue, $10, +14) Pea =e Rays tay
Difernsiating parially with respect to P,
ama)
ay, te, ty) Pg) PE) yep
tay tes) PO ca
Concepts of Human Development and Poverty
Therefore
a), ‘
1 a, “eta, re, Far
D0 becmusew,>0,
Inthe unit weights case We, = 1, for# = 1,2, 3) this reduces to
ane) af 2 J
3) Fa
Moreover, for c= 1, so that P(t) is simply the weighted or unweighted arith
metic mean of P, we have
yw,
eR
a)
ey
For an aggregate poverty index Pla) composed of distinct povery
subindices P,P, and Pit seems early desirable that Pla) should be increns
ingin each F, Also desirable is tha P(a) should increase at an inereasing rte
hin other words, that Pla) should be conver with respect to P,. Thi
lent to saying that Pla) decreases with reductions in P, and tir
ihngrate. Thenon propio exebshe tat ouragerepior anton Ma),
fora > 1, does saishy tis property.
Prorostmow 5. For each? = 1, 2,3,
ame
0.
PRoor,
2 Pla) 9 [aPiad
a | OF
eof
from equation 5
a
fk
[7]
928] /
Bw,
ie, Fay $y) Pra
J 4
alt
=ta-v4
Pave
arta
substituting for 220. fromm equation 5
whos a
(uy +10, +0, )PI0I*
13Hee
a, AA gay sey itar ut]
BPE (w, 4, #w, ayer Cee a
>o
because o> 1 and
(w, +0, +u))Plal*—w,P* = Ju, P2>0. 0
‘The next propery we consides isthe effect onthe azareate index Pa) of
increasing the weight w, on 1 pacticular poverty subindes P, We expec that
increasing the weigh onthe largest subindee, max (P, P,P) wil inrease
(a), while increasing the weight on th snlicstsubindse, min (Py Py P,)y
will reduce P(o), But wha: would be the effect of increasing the weigh on 4
‘middle? The answer depends onthe relationship between P, and Pa).
Provosmion 6, For any,
aPa)
PO sons 7, 3 Pa)
Be, BOE Pe
Proor From the defition of (a) we have
(u, 0 40, )PCa) = uF + PE +0, Pe
Differentiating both sides panially with respect ow,
Bie)
uy, $40, +, aPCaye! 2), playe = pe
(uy, +, aPtaye' Play =
Therefore
Peay.
Sb avg-b uo arta! PLD
(uy yt wy yartaye
Hence, sinee @> 0,
apa)
2D) so as po ,
Te, BO Pe wR
that is,
2 2Ro) O
Fora = Iwehave
mo.
2, mPa,
Souk PI
‘The nem property we consider isthe effect on P(a) af xsing the pararne-
ter vahieo forgiven values af the rubindicesP, for = 1, 2,3 It shows th
the value of the aggregate index will be higher when a higher-order mean ia
formed of P,P, and P,, In particular, men of order a> 1 will reeul in
Plot) thai greater than P(), the simple arkmetic mean of P,P end Py
14
Pnorosmont 7, For given Py P, and P, that ae not equal, fa > 7> 0, ther,
Pla) > Phy
PROOF. Let «> ¥> 0. By definition of Pla) and Pr, we have
and
Pet ae — Py,
Raising both sides of the second equation to the power (a/7) (> 1 because
a>y> 0)
(owt (orton tasting tt :
eat"
Now fi) = 27s a sccly convex function, since
Fors (a/y) xr
wd
ewatarvifary ifm
30 bso
Hence, by Jensen's inequality applied to atric convex umetions ((), since Py
P, and P, are not equal, wehave the strict inequality
{ py pr,
<4 — 109) 2 pn, 97),
Using he seen fain fs) = ge
(ool athe arate Marea
ae,
roi 1, we have the corallay that
we 40 ,Py + wy?
the simple weighted arithmetic mean of P,P, and P,.
Human Development Papers 1997‘We next investigate the “decomposability” of the human poverty index
simon aronp sithin county Suppor the popstion of county dived
ino m mtu exclusive and exeustve groups, The groupe may be dened
invterms of seat (rbun ral eon (ys province det org
dle female. Let be the si of poping for = 1,2...
ara er be hese of the toed popitin ofthe county Then
wis
LetP,, P, and P, be the values ofthe thrce poverty subindices Py, and Py
for groupy, where = 1,2. 7m. Finally, let P(@) denoce the mean f order
tof P,P, and , for groups. By defnition, we have
WP 4 WPS ty
ro ES
‘Wha isthe rclaonship bervcen Pa) sd the PO) for 1,2.
Sic decomposbiity ofthe index Pa) woud equ that (a) be» pops
Non-wcghted verge of th a, the population weighs being, But sit
decompoisblity does nat generally obtain,
“Thecelutionshipberwcen the values of ive obindes fo diferent groups
(Gor example, Py, for = 1.2,.. andthe overal vale ofthe uhindex or
cxample Pi Ssightfooward enough. As the indices are simple headcourts
of power have
Epon,
£4
But when the c-averages of Py P, and Py ate formed for each to give Pa),
the population-weighied average ofthe Pas exceeds Pa)
PROPOSITION &. For 2 1
¥ rtarr ra
Proof. Foreschj = 1, 2
wm, wehave
Concepts of Human Development and Poverty
Applying Minkowsk's inequality (Hardy, Litlewood and Paya 1952, p.30) to
GnnDPya nly bnedP,, fori = 1,2, 4m yds
tw 5 2
Home
Tele
Et rman o
‘The wae i een vibe si age ues ee
STEER LEER Aoachogcs ng
‘A simple exarnple with aon proportionality ofthe group povery subindices
shows why decomposbily(qualtyin propestion 8) docs not chtsinforck> 1.
Suppore the population is divided into two rurualy exclosive and exhaustive
of equa size rn =n fn ="), vith values ofpoverysubindices
(yy Pyy Py) = 025,0.5,079),
and (Phy Py Py) = (0.75, 05, 025).
Hence
(P,, Py Py) = (05,05, 0.5),
sod obvious Poa) = 05,
Now for group t
Pa) » ((4)0.25) + (110.5) + (1,075
03, by proposition 7 since a> f,
and for group 2
Poy = (10.75) + (Y,NO.% + (4, 10.257)
> 05, by proposition 7 since & > I
‘Therefore
Cia cay + CIP gay > (03) + (4903)
705
= Pla)
“Taking the group arithmetic means of each poverty subindex tends to
reduce or leave unchanged the relative disparity among the thece poverty
subindices. As eslt of his feature te @-averge ofthe arithmetic meens of
‘rou sobindices is smaller than she arithmetic mean of averages of group
subindices.
Finally forgiven value of (21), we discuss che degree of subsivatabil:
15iy berwees the poverty subindices P,P, and P, in the aggregate measure PCa)
The ehstcty af substitution bervezn, Say, P, and P, along an iso-P(a) curve
(helaing P, constant) is defined asthe percemage change i (P/P,) fra uit
percentage change in the stope ofthe tangent along thir curve (projected onto
PeP, space athe given value ofP), For the index P(a) the elaiciy of subs
stitution isconstant along each level set of P(a) and the same fo different level
sets. By proposition 3, Pa) is homogeneous of degree | in Py, P,P), and
therefore its leve sets are homotheti.
Propesinox 9. The elasticity of substitution & berseen any twa sbindices of
a, thatis between any two ofP,Pyand®,, constant and equal to 1A0-1)
Poor. Consider the elasticity of substuson berween P, and P, holding P,
‘constant, Te slope of the tangent along an iso-P(a) curve in P,P, space ie
ven by
ara) Jara
oR fa,
By definition, the elasticity of substation @ berseen P, and Py is
Alogi, /P,)
Blogs
From equation 5 in proposition 4 we have
ama Jane (PY
aR fw, LP
Therefore
Pole
AL (ey atu
and
Hence the elasiciy of substation
Alog(h,/P,)
Dee
16
‘Thus, if = 1, thee is infinite, or perfect, subsitutailiy between Py &
Py, And st "+ , there is no aubetutabiy between P, and PAs
increas fron 1, the elasticity of substitution decreases tmonotonicll ec
wt00.
Ifwe choose a= 1 (the ese of perfect substhabilty), the aggregate iné
Pa) isthe simple aritsetic meanof the three subindicesP, P,P, Aster
Lo infinity, che substitutability becomes zero, and the agereyate index tends
the maximum of the three subindces, max (Py, Py P,) In general, the el
ticity of substitation berween any s40 ofthe subindies, holding the other ce
stant, so = 1a
With a = 1 and infinite subsituabiig the impact on Pla) from a
increase (or decreas) of any subindex i the same, irespective ofthe vel
deprivation ia the eliferent dimensions. This contradicts the usual assum
tion that as the extent of deprivation in any dimension Increases (given ¢
cthers}, the weight on farther additions to deprivation in that dimensi
Should alo inesease, For this we need & > 1, The vale of azo influenc
ccomespondingy, he elative weigh to be placed on deprivation in the dif
‘ent dimensions. Consider, for example, P, = 60% and P, = 30% (with «
, = 435%) In this case, for any the relative impact of @ unit increase in
‘compared with a unt increase in P, which i given in general by (P,/P,)*
equals 2°, With ct = 1 the relative impact i given by 1, As was remark
cealies, 160 tends to infinity, P, becomesthe only determinant of Pa), so
its impact is infrtely larger than that of unit increase InP, which has,
this ease, no impact at all.
“The relative impact ineveases 2 is rised Fow 1. With = 3, the relat
impact is 4, giving the dimension of doubly greater deprivation (P,) mu
greater weight. The elatve impact iss ver fast with the raising ofa, esiscle
from the formula Fora = 5, she elatve impact of a unit increase in P is
much a8 16 times that of a eit increat in P
or calculating the human pavery index, & = 3 has been chosen. TI
ives an elasticity of substitution of "and places greater weight on the
dimensions in which deprivation ts lnget. It docs not, however, heve ¢
extremism of zero substitutability (given by c tending to infinity), nor t
very high values af relive impact that are generated as is essed (ince:
ing the relative impact, in the eate discussed shove, from 4 10 16 48. ge
from } 405}. Theve is an inescapable arbitrariness in the choice of &. T
Fight way co deal with tissue isto explain clearly what is being assume
15 has been aitempied here; so that public ertcism of this assumption
iit should be mentioned thatthe vat
corresponds exactly fo the weighting used to ealeulate the geneb
telated development index (GDM.
Human Development Papers 1997" Notes
1, See Dréze and Sen (1995).
2. On this question of interdependence, see Sen (1992).
3. On these and related issues, see the important study of Peter Svedberg (1997). He
also provides extensive comparison of the levels of nutritional deprivation respectively in
sub-Saharan Africa and in South Asia.
4. In the Human Development Report 1997, the HPI is constructed with an assumed
value of «= 3.
References
Anand, Sudhir (1977): “Aspects of Poverty in Malaysia", Review of Income and Wealth,
Series 23, No. 1, March, 1-16.
(1983): Inequality and Poverty in Malaysia: Measurement and Decomposition,
New York: Oxford University Press.
(1993): “Inequality Between and Within Nations”, mimeographed, Center for
Population and Development Studies, Harvard University, Cambridge, MA.
Anand, Sudhir, and Ravallion, Martin (1993). “Human Development in Poor Countries: On
the Role of Private Incomes and Public Services”, Journal of Economic Perspectives,
2
and Amartya K. Sen (1993): “Human Development Index: Methodology and
Measurement”, Human Development Report Office Occasional Paper 12. New York:
United Nations Development Programme.
(1995): “Gender Inequality in Human Development: Theories and Measure-
Human Development Report Office Occasional Paper 19, New York: United
Nations Development Programme.
Atkinson, Anthony B. (1970): “On the Measurement of Inequality”, Journal of Economic
Theory, Vol. 2, No, 3, September, 244-263.
(1973): “How Progressive Should Income-Tax Be?” in (ed.) M, Parkin, Essays
on. Modern Economics, Longman. Reprinted in (ed.) E.S. Phelps, Economic Justice,
Penguin Education, Harmondsworth, Middlesex, 386-408.
(1987): “On the Measurement of Poverty”, Econometrica, Vol. 55, No. 4, July,
749-164.
and Frangois Bourguignon (1982): “The Comparison of Multi-Dimensional
Distributions of Economic Status”, Review of Economie Studies, Vol. 49, 183-201.
Arrow, Kenneth J. (1965): Aspects of the Theory of Risk-Bearing, Yj Jahnsson Lectures.
Helsinki: Yrj6 Jahnssonin Said,
Basu, Kaushik (1987): “Achievements, Capabilities, and the Concept of Well-Being", Social
Choice.and Welfare, Vol. 4, 69-76.
Blackorby, C. and D, Donaldson (1978): “Measures of Relative Equality and their Meaning in
Concepts of Human Development and Poverty 17Terms of Social Welfare", Journal of Economic Theory, Vol. 18
(1984): “Ethically Significant Ordinal Indexes of Relative Inequality”, Advances
in Econometrics, Vol. 3.
Desai, Meghnad J. (1991): “Human Development: Concepts and Measurement”, Ewropean
Economic. Review, Vol. 35, 350-357,
Diamond, Peter A. and Michael Rothschild (eds). (1989): Uncertainty in Economies: Read-
ings and Exercises, Revised Bdition, New York: Academic Press.
Foster, James B. (1984): “On Economic Poverty: A Survey of Aggregate Measures”, Advances
in.Econometrics, Vol. 3, 215-251,
(1985): “Inequality Measurement”, in (ed.) H.P. Young, Fair Allocation. Provi-
dence, RI: American Mathematical Society.
+ Joel Greer and Erik Thorbecke (1984): “A Class of Decomposable Poverty
Measures”, Econometrica, Vol. 52, No. 3, May, 761-765.
Graaff, J. de v. (1977). “Equity and Efficiency as Components of General Welfare”, South
African Journal of Economics, Vol. 45.
Hammond, Peter J. (1975): “A Note on Extreme Inequality Aversion”, Journal of Economic
Theory, Vol. 11, 465-467.
Hammond, P.J. (1978). “Economic Welfare with Rank Order Price Weighting”, Review of
Economic Studies, 45.
Hicks, JR. (1940). “The Valuation of the Social Income”, Economica, Vol. 7.
Hardy, G.H., JE. Littlewood and G. Pélya (1952): Inequalities, Second Edition. Cambridge:
Cambridge University Press.
Kolm, Serge Ch. (1969): “The Optimal Production of Social Justice”, in (eds.) J. Margolis and
H. Guitton, Public Economics. London: Macmillan.
Martineti, Enrica Chiappero (1994): “A New Approach to Evaluation of Well-Being and
Poverty by Fuzzy Set Theory”, Giornale deggli Economisti e Annali di Economia,
Luglio - Settembre, 367-388.
Nussbaum, Martha C. (1988): “Nature, Function, and Capability: Aristotle on Political
Distribution", Oxford Studies in Ancient Philosophy (supplementary volume).
Orshansky, Molly (1965): “Counting the Poor: Another Look at the Poverty Profile”, Social
Security Bulletin, Vol. 28, 3-29.
‘Osmani, Siddig R. (198:
Pratt, John W. (1964): “Ri
122-136,
Ravallion, Martin (1994): Poverty Comparisons. Chur, Switzerland: Harwood Academie
Publishers,
Roberts, K.W.S. (1980b). “Price Independent Welfare Prescriptions”, Journal of Public
Economics, Vol. 13,
Rothschild, Michael and Joseph B. Stiglitz (1970): “Increasing Risk: I. A Definition”, Journal
of Economic Theory, Vol. 2, No. 3, September, 225-243,
Sen, Amartya K. (1973): On Economic Inequality. Oxford: Clarendon Press.
‘conomic Inequality and Group Welfare. Oxford: Clarendon Press.
k Aversion in the Small and in the Large”, Econometrica, Vol. 32,
18 Human Development Papers 1997(1976a): “Poverty: An Ordinal Approach to Measurement”, Econometrica, Vol.
46, 219-2315 reprinted in Sen (1982).
(1976b). “Real National Income”, Review of Economic Studies, 43; reprinted in
Sen (1982)
(1979), “The Welfare Basis of Real Income Comparisons”, Journal of Economic
Literature, Vol. 17.
(1982). Choice, Welfare and Measurement. Oxford: Blackwell, and Cambridge,
MA: MIT Press,
(1983): “Poor, Relatively Speaking”, Oxford Economic Papers, Vol. 35.
(1985). Commodities and Capabilities (Amsterdam: North Holland).
(1992): Inequality Reexamined. Oxford: Clarendon Press; and Cambridge, MA:
Harvard University Press.
(1993); “Life Expectancy and Inequality: Some Conceptual Issues”, in (eds.)
P.K. Bardhan, M. Datta-Chaudhuri and T.N. Kristnan, Development and Change.
Bombay: Oxford University Press.
(1997): On Economic Inequality, with a new Annexe by James Foster and
‘Amartya Sen. Oxford: Clarendon Press,
Stern, Nicholas H. (197): “Welfare Weights and the Elasticity of the Marginal Valuation of
Income”, in (eds.) M. Artis and R. Nobay, Current Economic Problems. Oxford:
Basil Blackwell
Svedberg, Peter (1997): Poverty and Undernutrition in Sub-Saharan Africa, mimeographed,
WIDER, Helsinki, 1997, 10 be published by Clarendon Press, Oxford,
Streeten, Paul, with Shahid J. Burki, Mahbub ul Hag, Norman Hicks, and Frances Stewart
(1981): First Things First: Meeting Basic Human Needs in the Developing Coun-
tries, New York: Oxford University Press.
United Nations Development Programme (UNDP) (1990): Human Development Report 1990.
New York: Oxford University Press,
United Nations Development Programme (UNDP) (1995); Human Development Report 1995.
New York: Oxford University Press.
(1996): Human Development Report 1996, New York: Oxford University Press.
Concepts of Human Development and Poverty »