how poor is poor 0

8/8/2019 How Poor is Poor 0

1/48

How poor is poor?

Towards a rights-based poverty line


2/48

ne is an independent think-and-dotank that inspires and demonstratesreal economic well-being.

We aim to improve quality o lie bypromoting innovative solutions thatchallenge mainstream thinking oneconomic, environmental and social

issues. We work in partnership andput people and the planet frst.

ne (the new economics oundation) is a registered charity ounded in 1986 by the leaders o The Other Economic Summit (TOES),which orced issues such as international debt onto the agenda o the G8 summit meetings. It has taken a lead in helping establish newcoalitions and organisations such as the Jubilee 2000 debt campaign; the Ethical Trading Initiative; the UK Social Investment Forum;and new ways to measure social and economic well-being.

It should be noted that this paper was fnalised in April 2008, beore

the World Bank/PovCalNet poverty data were updated to 2005

purchasing power parity. All o the analysis and data are thereore

based on the previous (1993 PPP) estimates.


3/48

Executive Summary 2

Introduction 4

Whats wrong with a dollar a day? 5

What IS $1-a-day? 6

Problem 1: Exclusive Focus onIncomes/Consumption 7

Problem 2: Arbitrary Poverty Lines 9

Problem 3: Over-Sensitivity to Base Years 10

Problem 4: Implicit Moral Judgments 11

Problem 5: Inconsistent Living Standards 13

Problem 6: Applying the global line to dierent countries 14

Problem 7: Updating the Poverty Line 15

Problem 8: Filling the Gaps 16

Conclusion 18

Alternative approaches

Input-Based Approaches 19

Outcome-Based Approaches 27

Towards an Alternative Approach a Rights-Based Poverty Line 30

The General Approach 30

Benefts o the Proposed Approach 33Some Preliminary Results: Poverty and the Right to Child Survival 34

Conclusion 40

Endnotes 41

Contents


4/48How poor is poor? 2

However, there are a number o important issues underlying the methodology bywhich they are produced, raising questions about the accuracy and reliability o thepicture they present o the level, composition and trends o global poverty.

P It ocuses exclusively on income (or more precisely consumption).

P It sets the poverty line at an essentially arbitrary level.

P Poverty estimates are critically dependent on the base year used or price

indices, which should make no dierence.

P The $1-a-day line itsel, used as the basis or policy, embodies an implicitmoral judgment that this level o income is morally acceptable, which is at besthighly questionable.

P It imposes inconsistent standards between countries in terms o living standards,which vary very widely between people living at $1-a-day in dierent countries.

P The conversion actors used to translate poverty lines into local currencyincomes are inappropriate, giving much greater weight to the prices o goodsand services bought by non-poor than by poor people.

P

Purchasing power parity (PPP) estimates or many countries have been basedon estimates rather than measurement (although this has improved markedlysince the paper was written).

P Updating the poverty line over time is problematic, or example ailing to reectdierences in ination as between rural and urban areas.

There are urther problems in the estimation o poverty in those years in which nosurveys have been conducted; and the method o estimating poverty since the lastsurvey are systematically skewed, so that they risk exaggerating the rate o povertyreduction in recent years.

Important as it has been to put poverty onto the global agenda, i the fgures

on which our view o poverty is based are potentially misleading, then we are inserious danger o adopting the wrong policies. More specifcally, given the natureo the methodological problems, we are in danger o being lulled into complacencythat the thrust o our global economic system is broadly consistent with reducing,and ultimately eradicating poverty.

We argue that a less arbitrary approach to defning poverty needs to be based ondefning an income level consistent with a standard o living which we considerthe minimum morally acceptable level. This can in principle be based either on aninputs approach (the cost o accessing what is needed to ulfl basic needs), or onan outcomes approach (the level o incomes associated in practice with minimumacceptable levels o indicators o physical well-being). Each o these approachescan be adopted either globally, to produce a single universal poverty line (c the

$1-a-day line), or on a country-by-country basis, to generate a poverty line oreach country. This gives us our broad categories o non-arbitrary poverty lines.

We review our alternative approaches to defning poverty, corresponding to three othese categories:

Executive summary

The World Banks estimates o global poverty, based on the$1-a-day line, have been a major intellectual and technical eat.There can be little doubt that they have also had a major political

impact in putting poverty on the international agenda.



P Kakwani and Sons International Food Poverty Line (global/input-based);

P Peter Edwards Ethical Poverty Line (global/outcome-based); and

P Morris et als Minimum Income or Healthy Living, and Reddy et als Capability-Based Approach (country-specifc/input-based)

However, each o these approaches is, in dierent ways, problematic.

We thereore propose a new approach to the defnition o poverty, which we callthe Rights-Based Poverty Line (RBPL), based on the ourth and fnal category acountry-specifc outcomes-based approach. The RBPL approach is based on theestimated statistical relation between income and indicators o well-being whichcorrespond to dierent economic and social rights (health, nutrition, education,etc). By setting a single universal threshold level o the indicator concerned, andestablishing the income at which that level is actually achieved in each country, wecan in principle defne a poverty line or each country which is at a dierent level oincome, but gives rise to an equivalent standard o living in each country.

This approach, we argue, both avoids the issues arising rom input-basedapproaches and resolves the problems inherent in any global poverty linedefned in terms o incomes, while maintaining consistency between countries.We present estimates o RBPLs or six countries (Bolivia, Eqypt, India (rural and

urban), Nicaragua, Senegal and South Arica) using the inant mortality rate as anindicator o the right to child survival, based on our alternative threshold levels.This demonstrates the wide range o incomes required to achieve equivalent livingstandards in dierent countries.

Having established a set o poverty lines, there are two ways in which poverty canbe reduced either by increasing incomes to the level at which rights are ulflled,or by reducing the income required to ulfl each right.

The picture o poverty this approach presents is much more complex than thosegenerated by other approaches, and particularly the single poverty headcount(and largely ignored poverty gap) fgures generated by approaches based on asingle global poverty line defned in dollars per day.

In the RBPL approach, we have multiple dimensions o poverty health, nutrition,education, housing, access to water and sanitation, etc. In each o thesedimensions, we have our indicators the poverty line itsel, the proportion o thepopulation below it, the poverty gap (reecting the extent to which incomes arebelow the RBPL) and the rights gap (indicating the extent to which rights are notulflled as a result o poverty).

However, this complexity is inevitable i we are to obtain a meaningul picture o themulti-aceted issue o global poverty, and we present a simple graphical approachto present poverty comparisons between countries and over time.

Through this approach, the RBPL can provide us with much more relevantinormation than the simple headline fgures o dollar-a-day approaches. Forexample, we can distinguish the extent to which poverty arises rom low incomesand rom low living standards at a given level o incomes in each country, andassess the relative importance in each case o income generation and (orexample) improving access to health services or education; we can identiy priorityareas in which access to services most needs to be improved or low-incomehouseholds; and we can avoid the misleading interpretation that poverty is allingwhere rising incomes are o-set by alling living standards relative to income.

At present, the data are not available to estimate RBPLs or all developing countries;and what data exist are ar rom ideal or the purpose. We thereore presentproposals or meeting the data requirements or a more comprehensive and reliableapplication, building on existing approaches. In view o the potential advantageso this approach it terms o presenting a uller, more nuanced and potentially moreaccurate picture o global poverty which would seem essential to eective policy-making or its reduction the relatively limited cost o such data improvementswould represent a very worthwhile investment.



This defnition o poverty has been adopted by the international community asthe basis or the frst o the Millennium Development Goals (MDG1) to reduceby hal the proportion o people living on less than a dollar a day.1 For mostpeople, this is what global poverty has come to mean.

This defnition o poverty, however, raises a number o undamental issues. Inparticular, it ocuses exclusively on one aspect o poverty, namely income, tothe exclusion o other critical aspects o deprivation, such as assets, access to

essential services, and social exclusion. Even in the context o income, it setsthe poverty line at an essentially arbitrary level. It also gives rise to a numbero serious technical problems in the measurement o poverty, comparisonsbetween countries, and analysis o changes in poverty over time.

Still more importantly, setting a poverty line as a basis or policy and themeasurement o progress represents an implicit moral judgment. By defningpoverty as the proportion o the worlds population below a particular level oincome, we are eectively saying that it is morally acceptable or people to liveat, or just above, this level o income, so long as they do not live below it. The$1-a-day line, however, is not based on any such judgment or analysis, and inmost developing countries living standards at this level o income are below

anything that could reasonably be regarded as acceptable.

This paper summarises the undamental problems with the $1-a-day approach,and what they mean or what we think we know about poverty. It goes onto review other alternatives, to assess whether they provide a more viablealternative. Finally, it proposes a new approach a rights-based poverty line(RBPL), based on the level o income at which living standards consistent witheconomic and social rights are actually achieved in each country which,we argue, resolves the problems inherent in the defnition o poverty moresatisactorily than the other alternatives.

Introduction

Discussions o global poverty are conducted almost exclusively interms o the so-called $1-a-day line developed and used by theWorld Bank.



The Banks leading poverty analyst has summarised the purpose o this line, and oestimates o poverty based on it, thus:

Ultimately, the Banks purpose in producing these aggregate measures [ofpoverty, based on the $1-a-day line] is to provide a consistent assessmentof progress against absolute income poverty in the developing world.3

While the $1-a-day poverty line has become by ar the most widely used standard

o poverty, however, it is also generally acknowledged that it has importantlimitations. Many observers and analysts, particularly outside the World Bank,see these problems as seriously undermining its practical useulness and/or itsconceptual validity.

The global poverty counts produced by the World Bank are too lowdueto methodological problems with the construction of its $1-a-day povertyline.4

The Banks [$1-a-day] poverty line leads to meaninglesspovertyestimates, as some of those identied as poor have clearly greatercommand over commodities than some of those identied as non-poor.5

This crude indicator[the $1-a-day poverty line] may have been aconvenient interim measure for practical purposes, a short-term expedient,buthas not turned out to be o continuing value.6

There are good reasons to believe that global poverty counts based on thedollar-a-day international threshold are meaningless The origins of thedollar-a-day thresholdlack a solid analytical basis Global poverty countsbased on $1-a-dayhave neither normative value nor empirical relevancefor poverty analysis Clearly the current state of global poverty estimationis ar rom adequate We need to nd better ways of xing internationallycomparable poverty lines.7

The World Banks approach to estimating the extent, distribution and trendof global income poverty is neither meaningul nor reliable. The Bank usesan arbitraryinternational poverty line that is not adequately anchored inany specifcation o the real requirements o human beings. Moreover, itemploys a concept of purchasing power equivalence that is neither welldefned nor appropriate for poverty assessment. These difculties areinherent in the Banks money-metric approach and cannot be crediblyovercome without dispensing with this approach altogether. In addition,the Bank extrapolates incorrectly from limited data and thereby createsan appearance o precision that masks the high probable error o itsestimates There is reason to believe that the Banks approach may haveled it to understate the extent o global income poverty and to iner

without adequate justifcation that global income poverty has steeplydeclinedin the recent period.8

The Banks poverty estimates have an importance beyond the actual numbers theyproduce. Increasingly, what matters to policy-makers is what is measured; and, by

Whats wrong with a dollar a day?

Until 1990, there was no widely accepted international defnition opoverty. The World Bank flled this gap with the $1-a-day line in the1990 edition o its agship publication, The World Development

Report.2



highlighting the scale o global poverty, the $1-a-day estimates have succeeded inraising the profle o the issue and the resources devoted to it.

The politics of resource mobilization may demand the use of internationalpoverty lines thatsound comparable, even when they are not.9

In terms o ocusing decision-makers attention on global poverty, the existence oestimates o its extent may thus matter more than whether the numbers actuallymean anything. For evidence o this, one need look no urther than the success othe Make Poverty Historycampaign in raising the profle o the issue, and moving it

to the centre o G8s agenda at its Gleneagles Summit in 2005.

I the numbers do not accurately reect actual levels, patterns and trends o poverty,however, they may distort what decision-makers decide about whatshould bedone, howit should be done and what priorityit should be given. The reliability opoverty estimates is thereore a critical issue. As Kakwani and Son observe, policyeorts ocused on the wrong target, though simple, may be sel-deeating.10

In act, it has been argued that $1-a-day poverty estimates have been deliberatelyused as a means o deending the current model o commercial globalisation.

The main use of the USD 1 per day indicator is ideological and political.

The indicator has led World Bank researchers to claim that globalizationis working, since it seems to imply that the proportion of people livingin poverty in the world as a whole is declining at a rate that will makeMillennium Development Goal (MDG) 1 achievable.11

This sceptical view receives some support rom a recent independent review oWorld Bank research, though not in the specifc context o the $1-a-day povertyestimates. While the review, with some justifcation, highlights the Banks workin developing global estimates o poverty among the outstanding work in theBanks portolio, it also notes with respect to the Banks extremely visible work onglobalization, on aid eectiveness, and on growth and poverty, on which many oits pronouncements on poverty reduction are made:

The panel had substantial criticisms of the way that this research was usedto proselytize on behalf of Bank policy, often without taking a balanced viewof the evidence, and without expressing appropriate scepticism.12

What IS $1-a-day?At frst sight, one might think that the $1-a-day poverty line at least provides a clearand easily understood basis or assessing poverty, and or judgments about itsmoral acceptability. We know what income is; we know (roughly) how much a dollaris worth; and we know how long a day is.

Inevitably, perhaps, it isnt as simple as that. First, poverty is generally measured onthe basis o consumption rather than income. This is primarily because household

surveys (the basis or poverty estimates) are generally seen as measuring spendingmore accurately than income. The implications, however, are signifcantly dierent.Households whose incomes are in danger o alling below the poverty line canoten maintain their consumption by drawing on savings, borrowing or selling assets but generally only temporarily. Equally, the very poorest households in terms oincomes are eectively orced to fnd some means o maintaining their consumptionsimply or survival. The question o howconsumption levels are maintained inthe ace o inadequate income is a critical one; but ocusing on consumptionrather than income ignores this. Thus consumption-based fgures are likely to bemore accurate than those o income; but it is important to remember that they aredierent rom income fgures, and have dierent implications.

Second, the dollars used or the $1-a-day poverty line are not actual US (or or that

matter Canadian or Australian) dollars, but international dollars, whose value isestimated atpurchasing power parity(that is, in accordance with how much theywill buy in each country). In principle, $1-a-day doesnt mean an income o $1 perday, but an income which will buy the same amount o goods and services thatcould be bought or an income o $1 per day in the USA.



Third, the $1-a-day and $2-a-day poverty lines are not at todays prices orindeed the prices o any other year. Neither are they actually $1 or $2. They wereoriginally set at $1 and $2 at 1985 prices, but were subsequently changed to $1.08and $2.15 (at purchasing power parity) at 1993 prices.13 While this adjustment issubstantially less than ination between 1985 and 1993, this line was consideredby the Bank to be approximately equivalent to the original line. Not only are therequestions about the equivalence between the old and new lines, however, but theeects o re-basing to 1993 prices raises serious issues about the whole approach.

The implications o these three issues or the reliability o poverty estimates are

discussed later.

Problem 1: Exclusive ocus on incomes/consumptionWhile income is a key dimension o poverty, it has many other dimensions.According to the 1995 World Social Summit Programme o Action:

Absolute poverty is a condition characterized by severe deprivation of basichuman needs, including food, safe drinking water, sanitation facilities, health,shelter, education and information. It depends not only on income, but alsoon access to social services.14

Even in the material sphere, a households savings or other assets are a key

consideration in whether or not we would consider its members to be poor. Timepoverty and working conditions are other, oten neglected, aspects: to say thatsomeone is not poor because he or she keeps his or her income just above thepoverty line by working 16 hours a day 7 days a week in an unpleasant and/ordangerous occupation would seem perverse.

Lack o access to basic services, such as health, education, water and sanitation,is also an important element, as are housing quality and living environments: evenwith an income above the poverty line, someone living in a slum with no accessto these services might well be considered poor. Poverty might also be seen asarising rom chronic mental or physical illness or disability; rom powerlessness,social exclusion or unequal gender relations; rom illiteracy or lack o basiceducation; or rom physical, social or economic insecurity. None o these actors

is taken into account by the $1-a-day poverty line, except to the extent that theyimpact on measured consumption.

Most o these actors are closely inter-related with lack o income, both as causesand symptoms, in a complex vicious circle. But this is by no means always the case and, as Kakwani observes,15 it would be odd to call a disabled millionaire poor.Kakwani concludes rom this that poverty is present when basic capability ailurearises because a person has inadequate command over resources that is, lacko basic capability can be construed as poverty only when it is directly causedbyincome poverty.

It is, however, ar rom clear that appealing to causality provides a satisactory

solution to the dilemma. Suppose, or example, that our hypothetical millionairesdisability arose rom an occupational injury in a dangerous job which he was orcedto take through poverty earlier in his lie; but that he became a millionaire as a resulto large-scale damages being awarded against his employer. The lack o basiccapability is still a result o income poverty; but there is no obvious reason or this toaect whether or not we consider him poor.

Rather, it would seem a closer approximation to what we understand by poverty tosay that we do not think the millionaire is poor because his income is so far aboveanything that could be considered income poverty as to make his income irrelevant.I the poverty-line were $1 per day, and his income were $1.01 per day, intuitivelywe might very well still consider him as poor as someone living on $0.99 per daywithout a disability either because his disability gives rise to additional costs (e.g.

or a wheelchair, or increased transportation costs, or to pay or services others areable to perorm or themselves), or because his quality o lie is reduced or anygiven level o income.



The issue o non-fnancial dimensions o poverty suggests two possibleapproaches. One is to develop a composite indicatoro poverty, whichamalgamates income (or consumption) with other aspects o deprivation. This is theapproach taken by UNDPs human poverty indicator. While useul as a supplementto income poverty data, however, this approach also has important limitations (Box1), leading one analyst to conclude:

Important non-income aspects of poverty, such as deprivations in health,educational attainment and enjoyment of citizenship rights, cannot bemeaningfully combined with consumption measures to dene a comprehensive

poverty indicator of relevance.16

Box 1: Composite indicators the human poverty indicator

While there is a need or measures o income poverty, an important shortcoming o the $1-a-day measure is that itneglects other important dimensions o poverty. This leads some analysts to argue that any indicator that identifes

an individual as poor has to be multidimensional.17

In 1997, UNDP sought to broaden the defnition o poverty beyond income poverty by developing a compositeindicator, the human poverty index (HPI).18 This combines three components, each given an equal weight:

1. Premature death (the percentage o people expected to die beore the age o 40).

2. The adult illiteracy rate.

3. Standard o living (combining the percentage o the population without access to health services and withoutaccess to sae water, and the percentage o children under fve suering rom malnutrition).

The HPI thus broadens the defnition o poverty beyond income in act, it will be noticed that it does not include

an income component. It combines a number o important indicators o the eects o the most extreme poverty,and provides a valuable additional indicator o national perormance, together with income poverty indicators andthe human development indicator (HDI). In eect, it provides a deprivational counterpart o the conglomerativeHDI that is, it indicates the proportion o the population who suer severe absolute shortalls o humandevelopment within a particular country rather than its overall level in much the same way that income povertyindicators provide a deprivational counterpart to GDP per capita.19 At the same time, its potential to supplementincome poverty indicators is demonstrated by the act that, while the individual components o the index arestrongly correlated with income at the country level, the correlation between income poverty and human poverty[in dierent countries] is weak, and there is a large spread in levels o human poverty among countries with similarlevels o income poverty.20

As a poverty indicator in its own right, however, the HPI has a number o limitations.

P As the UNDP itsel recognises, there is some inescapable arbitrariness in the selection o indicators, and in theirweighting.

P It does not distinguish between the incidence o poverty and its depth or example, to take the extreme case,between a country where a dierent 30 per cent o the population qualifes as poor under each o the threecriteria, so that 90 per cent are moderately poor, and a country where the same 30 per cent qualifes under allthree criteria, so that 30 per cent suer rom acute poverty.

P It can only provide an aggregate picture o the extent o poverty in a particular country (or subnationalpopulation), and does not provide a means o identiying which households are poor.

Because it does not identiy households or individuals as poor, except by each o the individual criteria, it does not

allow a link to be made with incomes.

In consequence, while the HPI provides an important complementto income poverty data, it does not represent aneective substitute.



The problems o combining income and non-income dimensions o poverty intoa single composite indicator suggests a second alternative: to consider otheraspects o deprivation, such as disability, chronic illness, poor living or workingenvironments, lack o access to basic services, insecurity etc., as shiting thepoverty line or the individual or household concerned relative to those who do notsuer these deprivations, according to the impacts on well-being and/or additionalcosts associated with them. This principle, applied at the country level, underlies therights-based approach we present later in this paper.

Whether or not they are linked with lack o income, the broader aspects o

deprivation discussed in this subsection are clearly just as important to those theyaect, and we certainly must not lose sight o them. Nonetheless, we would arguethat we still need a distinct concept o income poverty as such. Within the broadercontext o deprivation, income is clearly a critically important aspect o poverty inits own right, and a major source (though by no means the only source) o mosto the other aspects o deprivation among those who are in income poverty. It alsorepresents a critical link with economic policies and perormance, which representthe major cause o the extent, severity, distribution and evolution o global poverty.

Whether we use the word poverty as shorthand or income povertyor or a broaderdefnition o deprivation is a matter o terminology. Whatever we call it, the analysiso income poverty is a critically important aspect o any consideration o poverty in a

broader sense. This, rather than the broader deprivation defnition, is the one we usein this paper. We recognise, however, that income poverty lines should in principlebe adjusted according to other aspects o deprivation; and this is implicit (at least atthe country level) in the approach we develop later in this paper.

Problem 2: Arbitrary poverty linesEven as a defnition o income poverty, the $1-a-day line is essentially arbitrary. Itwas established as a measure o poverty by the World Bank in the 1990 editiono its agship publication, The World Development Report, ocusing on poverty.21It was selected as the median o the ten lowest poverty lines in a sample o 33developed and developing countries (including only ten low-income countries).

As the chie architect o the Banks poverty work readily (and repeatedly)

acknowledges, this was deliberately a conservative choice.22 In act, while the$1-a-day standard is applied to the developing world as a whole, using the mediano the lowest ten poverty lines means that

Only ve countries have poverty lines which are below this norm. The restof the countries all have poverty lines above the $1-a-day in 1985 PPPs.The $1-a-day poverty line therefore may be more appropriate for measuringextreme absolute poverty in the poorest countries. For the range of incomesof countries normally included in measuring global poverty trends, the$2-a-day poverty line is arguably a more appropriate indicator of extremepoverty on a global basis.23

In any case, given how the analysis was conducted, as Kakwani and Son observe,

The Banks claim that $1-a-day is representative of poverty lines among low-income countries has a very weak foundation.24

Specifcally, many o the national poverty lines used were not ofcial defnitionsbut estimates rom independent researchers; they were based on widely dierentmethodologies; and, where more than one line was used (e.g. where separate lineswere used or rural and urban areas), the lowest was used rather than the weightedaverage. Together with the selection o the ten countries with the lowest povertylines, rather all the developing countries in the sample or even the ten poorestcountries, this results in an artifcially low fgure.

Kakwani and Son also point out:

Since econometric analysis [based on these gures] failed to yield areasonable international poverty line, the Bank then decided to determine thepoverty line by eyeballing the scatter plot of [the econometric] equation.25



Thus the process by which the $1-a-day line was selected was essentially arbitrary.The primary motivation in choosing this fgure appears not to have been that itprovided a minimally adequate living standard, but rather a combination o thesimplicity o the headline fgure (leading to a line which could be described as awhole number o dollars) and a deliberate choice to use the most conservativedefnition (leading to the choice o $1-a-day rather than $2-a-day as thepredominant line).

Problem 3: Over-sensitivity to base yearsThe process described above generated the original poverty line o $1 per day,

in purchasing power parity (PPP) terms, at 1985 prices. As noted above, however,the line has since been updated, based on new estimates o PPP exchange ratesestimated or 1993.

Estimating PPP exchange rates requires a comparison o prices between countries;and this requires weights to be given to dierent goods and services, in line withtheir importance in consumption. However, consumption patterns change over time,so that the weights, and thus estimated PPP exchange rates, become progressivelyurther out o line with reality. They are thereore updated periodically on the basis orevised weights.

However, the eects o this process demonstrate still more strongly the arbitrary

nature o $1-a-day line, leading poverty analysts to criticise the updating asgiving rise to egregious errors,26 and play[ing] havoc with the poverty estimates,changing them in ways that have little or nothing to do with the actual experienceo the poor.27 Others have demonstrated that the results are highly sensitive tothe arbitrary choice o PPP base year, which is completely irrelevant to anyonesstandard o living,28 and that poverty estimates or the same country and year canchange dramatically purely as a result o shiting the base year.29

This raises serious questions about the whole approach, irrespective o which baseyear is used.

The poverty estimates that the Banks method generates when applied withone PPP base year are inconsistent with those it generates when applied

with another. This sufces to discredit the method even when it is alwaysapplied with the same base year The discrepancies between the estimatesyielded by two applications of the Banks method [using 1985 and 1993 asPPP base years] are so large as to render this method unacceptable.30

The extraordinary scale o the problem, as demonstrated by Pogge and Reddy,makes it clear why they and others consider the Banks poverty estimates to bemeaningless. The eect o the rebasing was to increase the estimated poverty ratein Sub-Saharan Arica by a quarter in 1993, while reducing that in Latin America bya third. Thus the ratio between the two nearly doubled, rom 1.66 to 3.24, purely asa result o the change in base year.31 Global trends are also aected: whereas theglobal incidence o poverty ell by 4.23 per cent between 1987 and 1993 based on

the 1985 PPP, it ell by only 0.57 per cent based on 1993 PPP.32

The problem can be ar worse at the country level. The result o updating the baseyear rom 1985 to 1993 was to raise the poverty line or Nigeria in 1993 by 42 percent, while lowering that or Mauritania by 61 per cent, changing the ratio betweenthe two (in the same year) by a actor o 3.7. The eect on the estimated incidenceo poverty is still more dramatic. Using the 1985 base-line, the Bank estimated thepoverty rate in the two countries to be almost identical in 1999, at 31.1 per centand 31.4 per cent respectively. Using the 1993 base-line the ollowing year, it oundthat Nigerias poverty rate was nearly 20 times that in Mauritania, at 70.2 per centas compared with just 3.8 per cent.33 For the application o the same method tothe same data about the same peoples income to give rise to such vastly dierentresults, purely as a result o a changing the base year, amply demonstrates theshortcomings o this approach.



In view o these problems, Angus Deaton proposes setting an initial internationalpoverty line, defning its equivalent in a particular year in each country (allowingsome adjustment or local contexts), and simply increasing it progressively over timeby local price ination.34 Srinivasan suggests a similar approach, but starting romthe existing $1-a-day line.35 I combined with the development o appropriate priceindices based on the consumption patterns o poor households both to establishthe initial poverty lines and or subsequent adjustment, this would avoid the need toupdate the base year, and the serious distortions which arise rom it although theresults would remain critically dependent on which base year was used to set theinitial line.

In practice, however, the switch rom 1985 to 1993 as a base year also signifcantlylowered the levelo the $1-a-day poverty line in real terms.36 Keeping the $1-a-dayline equivalent to $1 per day at 1985 PPP would imply increasing it in line withUS dollar ination between 1985 and 1993, to a fgure o about $1.30 per day at1993 prices. Instead, the Bank recalculated the originalpoverty lines o the originalsample o 33 countries at 1993 PPP exchange rates, and again took the median othe lowest ten. This process gave rise to a poverty line o just $1.08 per day at 1993prices. The median o more recent poverty lines in an alternative sample o 19 low-income countries indicates a fgure in the order o $1.50.37

According to Pogge and Reddy:

The redenition has lowered national poverty lines in 77 countries,containing 82% of the population of the 92 countries [for which PPPestimates are available] and raised national poverty lines in only 15countries.38

The net result was to reduce the recorded number o poor households in 1993 by58 million more than two-thirds as much as the estimated reduction in povertybetween 1987 and 2001 based on 1993 PPP.39

This is part o a more general problem, as shiting to a later base year can beexpected to lower the poverty line progressively over time. According to Pogge andReddy:

[Base year] adjustments are like [sic] to introduce a systematic bias towardpainting too rosy a picture of global poverty trends. Adjustment of thebase year reduces reported poverty headcounts insofar as internationalconsumption is shifting toward commodities (such as services) that are verymuch cheaper in poor than in rich countries and away from commodities(such as food) that are not much cheaper in poor than in rich countries.40

Applying a poverty line o $1.30$1.50 to the World Banks PovCal data41 wouldimply that an additional 1015 per cent o the population o the developing world(some 400700 million people) have been in poverty over the last 25 years than isindicated by the Banks $1.08 line. Moreover, while the absolute rate o reduction

in the incidence o poverty has been similar, the higher absolute level o povertymeans that this translates into a slower proportionalrate o reduction, at around 2per cent per year compared with nearly 3 per cent.

Problem 4: Implicit moral judgmentsAs noted above, the level o the $1-a-day line is essentially arbitrary. This makesits use as a basis or policy particularly problematic because o the moral judgmentembodied in poverty lines. As the Banks leading poverty analyst observes:

Poverty lines are inherently subjective judgments people make about whatconstitutes a socially acceptable minimum standard of living in a particularsociety.42

By adopting a particular poverty line or policy purposes, we are thus eectivelysaying two things:

1. It is morally acceptable or people to live atthat level o income, so long as theydo not live belowit.



2. Further increases in income beyond this level no longer matter rom a povertyreduction perspective.

Because this judgment is generally unstated, however, the issue is rarely consideredin moral terms.

There are a number o important actors to take into account in judging whether adollar a day is enough or someone to live on in a developing country. The frst trap isthe assumption that it is easier to live on $1-a-day in a developing country, becausemany prices are much lower. Because the $1-a-day line uses PPP exchange rates,

these dierences are (in principle) already accounted or in the way the fgures arecalculated. Allowing or exchange rate changes and ination, the $1-a-day line is, inprinciple, equivalent to living on about 0.60 per person per day in the UK in 2007.This means a amily o our living on an income o about 75 a month.

Second, we can all too easily assume that it is possible or people living on suchlow incomes to survive by scavenging in urban areas, or by enjoying the ruits o theland hunting wild animals or collecting wild ruits, or example; or that they growtheir own ood; or that they receive ood aid; or that they receive gits or support romamilies or neighbours, beg or even steal.

Again, however, this is a allacy. Since poverty is generally measured on the basis o

consumption, not income, anything that is consumed is included, and valued at itsmarket price, whether it is bought, ound, scavenged, hunted, collected, received as agit or charity, or stolen.

Even i one doesnt all into either o these traps, one encounters a third problem inmaking a moral judgment about the $1-a-day poverty line that an income o 0.60per person per day in the UK is so ar below anyones actual experience as to beunimaginable. In act, it is in the order o one-hundredth o the average income in theUK. Looked at another way, it was equivalent to around 37 people living on a singleminimum wage, with no recourse to welare benefts in early 2006.43 (In act, as wediscuss later, the equivalent incomes in the UK are actually considerably worse thanthese fgures suggest.)

As a result, most people would not be in a position to judge whether the $1-a-dayline is consistent with their own moral values about what constitutes an acceptablestandard o living, even i they were inclined to do so. (It also seems likely that mostpeople never ask themselves the question.)

An alternative way o approaching this question is to consider what it actuallymeans in practice to live at the $1-a-day line, in terms o aspects o lie on whichpeople could be expected to have a defnite moral opinion. This is a question whichremained unaddressed until surprisingly recently, partly because o serious datalimitations.

However, recent World Bank research does provide a frst opportunity to make such a

judgment. World Bank economist Adam Wagtsa used the Banks Poverty and Healthdataset to estimate child mortality rates and under-nutrition at this level o income indierent countries.44 His results suggest that between one in six and one in twelveo all children in households at (not below) the $1-a-day poverty line in a typicaldeveloping country die beore their fth birthdays (compared with an averageo around one in 160 in developed countries); and between one-third and halo those ortunate enough to survive are stunted through chronic malnutrition.45Conditions are better than this in about a quarter o developing countries but theyare also even worse in another quarter.

Moral judgments are inevitably subjective in nature. At a time o unprecedentedprosperity and conspicuous consumption among the better-o, however, it wouldseem surprising in the extreme i most people, asked directly, considered that it was

morally justifable or people to live at a level o income which resulted in the deathso one-sixth or one-twelth o all children beore the age o fve. It certainly doesnot seem consistent with the Right to Child Survival established by the 1989 UNConvention on the Rights o the Child Health,46 or the right to the highest attainablestandard o health, a ounding principle o the World Health Organisation in 194647



enshrined in the International Covenant on Economic and Social Rights o 1966.48This suggests, at the very least, that there is a undamental inconsistency betweeninternational standards in terms o poverty and o economic and social rights.

Problem 5: Inconsistent living standardsI we are to make moral judgments about poverty lines based on what it means inpractice to live at a particular level o income, then there is a serious problem withthe very concept o a global poverty line fxed in monetary terms, whatever notionalexchange rate is used to apply it to individual countries, because the same level oincome may give rise to very dierent living standards in dierent contexts.

However, or Martin Ravallion, the World Banks leading poverty analyst and one othe chie architects o the $1-a-day line, this is the very oundation o the defnitionand measurement o poverty.

When trying to make a global comparison of absolute poverty in terms ofconsumption, there is (in my view) a compelling case for using the samereal consumption level as the poverty line in all countries.49

This, or Ravallion (and the World Bank), is what comparability means in the contexto poverty.

For our global poverty counts, we have but one overriding concern thattwo people with the same standard of living, measured by command overcommodities, be treated the same way no matter where they live.50

As Adam Wagstas results on health standards at the $1-a-day line demonstrate,however, there is a considerable dierence between what it means to live withthe same real consumption level in dierent local contexts.51 This may partly bea result on the use o inappropriate conversion actors (as discussed in the nextsub-section), and also with the arbitrariness associated with the choice o base year as demonstrated by the Nigeria and Mauritania example highlighted above. Thescale o the variation in our own results, however, presented later in this paper, aswell as Adam Wagstas, suggest a larger problem than this.

This is not surprising, or at least three reasons.

1. Consumption needs vary substantially between countries. For example, theamount o energy and clothing required is much greater in cold climates than inhot climates; the cost o household maintenance may be much greater in areasprone to oods and storms than elsewhere, etc.

2. Levels o social provision vary widely between dierent contexts. While thereis generally a signifcant association between household income on the onehand and access to (and quality o) health services and education on the other,dierent levels o provision and user costs mean that the level o access andquality associated with a particular level o income may be very dierent in

dierent places.

3. In the case o health indicators, health risks vary widely between countriesand areas as a result o geographical and social actors, or example climaticconditions (particularly or diseases such as malaria and dengue ever), andpoverty and limited access to health services (which increase the incidence ocommunicable diseases). This will result in a considerable variation, not onlyin the health outcomes associated with a given level o income, but also in theconsumption required to maintain health at a given level (or example, throughthe need or treatment and or preventive measures such as mosquito nets).

As a result, the living standards associated with any specifed level o income willvary widely between countries and contexts. This means that applying a single

poverty line defned in monetary terms to dierent countries is incompatiblewith defning poverty on the basis o living standards which is preciselywhat we are bound to do i our concern about poverty is based on the moralunacceptability o the consequences o living on extremely low incomes. A singleglobal poverty line, wherever it is set, will inevitablymean either that some people



whose living standards are above the threshold o acceptability are classifed aspoor, or that some whose living standards are unacceptably low will be classifed asnot being poor.

Problem 6: Applying the global line to dierent countriesA urther challenge to the principle o a $1-per-day approach arises rom the wayin which purchasing power parity (PPP) exchange rates are estimated. As notedabove, these were used both in the original estimation o the $1-a-day poverty line,and in translating it into supposedly consistent poverty lines in each country.

As Ravallion emphasises:

If one is comparing living standards in terms of household consumptionper capita then comparisons of absolute poverty across regions, sectors ordates can be misleading unless the poverty line has constant purchasingpower (based on a cost-of living index appropriate to the poor).521

However, because the PPP approach was designed or comparing nationalaccounts fgures (e.g. or total national income, investment or consumption) ratherthan or estimating poverty, PPP exchange rates are based on overallconsumptionpatterns in the economy as a whole, rather than those o poor households. Sincenon-poor households account or the great majority o consumption, even where

the poor represent a large majority o the population, PPP estimates reect mainlythe consumption patterns o the better-o, which are very dierent rom those o thepoor. As a result:

The use of such PPPs is quite inappropriate for poverty assessment andseverely distorts the resulting poverty estimates.53

Worse, this introduces a systematic distortion into the analysis. A much higherproportion o the spending o poor households is on ood, and especially stapleoods, than or better-o households sometimes as much as 7080 per cent.Conversely, non-poor households tend to spend a much larger proportion o theirincome on services. However, the high local labour content o most servicesmeans that they tend to be much cheaper in poorer countries, where wages levels

are lower, whereas the price dierences or oodstus between richer and poorercountries are generally much smaller. As a result:

Existing data about the prices of foodstuffs and, more specically, of staplebreads and cereals, show that these items (a large part of the consumptionrequirements of the poor) cost far more in poor countries than general-consumption PPPs suggest. The same is true for many basic necessitiesother than food.54This means that converting the $1-a-day poverty lineat the overall PPP exchange rate results in purchasing power well below$1in poorer countries in terms o the goods and services actually consumedby poor households, leading to serious under-estimation o the number opeople defned as poor by this standard. Pogge and Reddy estimate that

using an index or bread and cereal prices in place o overall PPP estimateswould increase poverty lines in poor countries by an average o 3040 percent, suggesting a fgure o around $1.401.50 at 1993 PPP broadly in linewith the line resulting rom adjusting the original (1985 PPP) $1-a-day line inline with ination.

Price indices based on the consumption patterns o poor households do notgenerally exist in developing countries. A project is currently being undertakenby the World Bank, however, in collaboration with the International ComparisonProgramme, with the aim o constructing such indices. This approach has thepotential to bring signifcant improvements in the quality o poverty estimates, byreducing the role o irrelevant goods and services in the PPP conversion actors although, as Reddy observes:

This initiative fails to address other equally important shortcomings ofcurrent approaches to poverty estimation. Among these are the computationof prices from irrelevant countries in the construction of the PPPs, the lackof a meaningfully dened international poverty line, and the unreasonable



dependence of poverty estimates upon the base year of the internationalpoverty line.55

There is also a problem o circularity: it is impossible to know what commodities thepoor consume without frst identiying who the poor are56 and this dependson how the poverty line is defned. This makes using the consumption patterns othe poor as a means o defning the local poverty line problematic. More generally,the problem o standardising across the very dierent consumption bundles o poorhouseholds in, or example, Vietnam and Mexico, would remain intractable.

A second problem is that even the PPP exchange rates currently used suer romimportant data inadequacies.

The general-consumption PPPs currently in use are based on limited andhighly questionable evidence As a result a massive element of guessworkand gap-lling underlies current poverty estimates.57

Direct estimates o PPP exchange rates, based on actual price data, do not existor all countries (about 60 or the 1985 dataset, and 110 or 1993). Figures or othercountries are estimated indirectly on the basis o the estimated relationship o theserates with other variables such as GDP per capita. While a statistically signifcantrelationship exists, this provides, at best, very approximate fgures or other countries.

The level o uncertainty attached to PPP estimates can be gauged by comparingthose used by the World Bank with the corresponding fgures rom the Penn WorldTables (PWT6.1) produced by the International Comparison Project, the other majorsource o PPP exchange rates. The dierences are very substantial, and sufcient toaect poverty estimates considerably according to which source is used.

Close to 50 per cent of the World Bank estimates diverge by at least 20 percent on either side from the PWT6.1 PPP estimates in 1993. In more than15 per cent of the countries the World Bank estimates are higher than thePWT6.1 gures by 40 per cent or more.58

The problem o missing PPP data is particularly acute in terms o global poverty

estimates because some o the largest countries are among those with limited orunreliable data.

No price surveys for constructing PPP prices are available for importantcountries such as China. For others in particular, India such surveys tookplace only in the distant past. 59

India last participated in a benchmark survey or PPP in 1985,60 and accordingto Reddy and Pogge the current estimates o PPPs or China that are consideredplausible dier by a actor o two. 61

Since India and China together account or 44 per cent o the population o the

developing world, and more than hal o the World Banks estimates o total povertyat the $1-a-day level (down rom two-thirds in 1981), any inaccuracy in the PPPs orthese countries could have a very considerable eect on the Banks global povertyfgures. The potential eect on the time trend o poverty is also considerable,particularly in the case o China.62 According to the Banks estimates, the incidenceo ($1-a-day) poverty in China ell by more than fve-sixths rom 63.8 per cent in1981 to just 9.9 per cent in 2004, while poverty in the rest o the world ell by onlyone-third, rom 31.3 per cent to 20.7 per cent. While the number o people estimatedto be below the $1-a-day line in China ell by 80 per cent between 1981 and 2004,the estimated number in the rest o the world actually increased marginally, rom837 million to 841 million.63

Problem 7: Updating the poverty line

A urther problem arises rom the need to adjust the poverty line, set at 1993 prices,or use in earlier and later years. This is done by converting the $1.08 poverty lineinto local currency in 1993, using the PPP exchange rate, and adjusting it in linewith a national price index (e.g. the consumer price index). However, this use ogeneral prices indices generates similar problems to those caused by the use o



general PPP exchange rates, because again such price indices are based on totalconsumption, and thus primarily reect the consumption patterns o the better-o.

This can be demonstrated by the eects o recent major increases in cereal prices,largely as a result (directly or indirectly) o increased biouel production. In March2008, the world prices or major cereals (rice, maize, wheat and sorghum) werearound double their average levels in 2006.64 For the poorest households, suchstaples may already account or more than hal o their expenditure, so that doublingthe price represents a very considerable increase in their living costs. But or thepopulation as a whole, staples represent a much smaller part o total spending; so,

since the ination rate is calculated on the basis o overall consumption, a doublingo cereal prices might add only 5 per cent to the price index. Adjusting the povertyline by the overall ination rate would then increase it rom (say) $1.00 to $1.05,although a poor households income would need to rise to $1.50 to maintain thesame level o consumption.

A second problem is that national price indices do not necessarily apply equallyin all areas and particularly as between rural and urban areas, where prices maychange at signifcantly dierent rates. (They may also dier markedly betweenregions, e.g. due to dierences in dietary patterns and ood availability.) Thus:

Indias ofcial poverty estimates, which use price indices for updating, show

diverging trends in urban and rural poverty between 198788 and 199394.When these estimates were recalculated using commodity weights and unitvalues based on household surveys, no great difference was found in therate of decline of urban and rural areas between those years.65

In other words, the appearance o a relative increase in rural poverty arose purelybecause an overall price index was used rather than separate ones or rural andurban areas. The global fgures are also distorted by not taking account o rural-urban dierences in living costs. In presenting the World Banks latest estimates,Chen and Ravallion estimate the impact o this eect:

More than 100 million people are added to the global count of the[$1-a-day] poor when we allow for the higher cost of living in urban areas,

and about half of the 100 million come from South Asia and one third from[Sub-Saharan Africa].66

While Chen and Ravallions fnding that this makes much less dierence to thetrends over time is true in aggregate, this masks signifcant eects at the regionallevel and in all probability still stronger eects at the national level. Thus one-thirdo the apparent reduction in poverty in Sub-Saharan Arica between 1993 and 2004appears to be a result o the ailure to take account o rural-urban price dierences.While the Banks headline fgures suggest that the proportion o people in Aricabelow the $1-a-day line ell rom 45.5 per cent to 42.6 per cent, correcting or rural-urban price dierences indicates both a higher level and a smaller reduction, rom49.2 per cent to 47.2 per cent.67

Problem 8: Filling the gapsPPP exchange rates are not the only area in which directly estimated data aremissing, and other means have to be ound to fll the gaps or where the methodso doing so raise important questions about the reliability o the Banks estimates olevels o, and changes in, poverty. There are also major gaps in data on incomes,because the household surveys on which poverty estimates are based aregenerally conducted only at long and irregular intervals. In order to reach global (oreven regional) estimates o changes in poverty over time, it is thereore necessaryto estimate poverty between, beore and ater these observations.

Estimating poverty rates between surveys is relatively unproblematic this is doneon the basis o the survey beore and the one ater, assuming that actual incomes

changed in line with the pattern o overall economic growth between the two. Thisis not necessarily accurate, because actual changes in poverty will also reectchanges in income distribution, which may vary over time; but it should generallyprovide a reasonable approximation.



Much more serious problems arise, however, rom estimating poverty rates beorethe frst household survey and ater the last. This is done on the basis o theoverall growth rate o consumption in the national accounts statistics, keeping thedistribution o income constant. This is problematic or two reasons.

First, and most obviously, it ignores changes in the distribution o income. Sincemost countries (developed as well as developing) have seen a systematic increasein inequality since 1980, in marked contrast with the previous 20 years,68 this islikely to result in a signifcant over-estimation o the rate o reduction in poverty.

The second problem is that there is a very substantial dierence betweenconsumption as measured by national accounts statistics and as measuredby household surveys. More seriously, the gap between the two is wideningconsistently over time, as the national accounts measure is growing substantiallyaster than the household survey measure.69,70 Thus simply using the growth raterom the national accounts statistics without adjusting it to compensate or thisdierence, means that increases in household incomes and thus the reduction inpoverty will be systematically over-estimated.

The Banks estimates eectively assume that the discrepancy between the twoconsumption measures arises primarily because o under-reporting by better-o households in household surveys, so that poverty data are not substantially

aected.71 The divergence, however, almost certainly also arises at least partlybecause the defnition o consumption in the national accounts statistics is dierentrom that used in household surveys.

Specifcally, national accounts statistics on consumption include expenditure bynon-proft organisations and imputed rents or accommodation (the rental valueo owner-occupied properties) which are not generally included in the householdsurvey/poverty defnition; and they do not eectively capture subsistenceproduction and inormal sector activity. The latter is particularly important becausethese are oten the most important sources o income or the rural and urbanpoor respectively. Equally, expenditure by non-proft organisations may well besubstantially greater (particularly relative to total consumption) in low-income

countries, where the incidence o poverty is greatest, than it is in middle-incomecountries.

Thus the fgures or the earlier and later parts o the period covered by the Banksglobal poverty estimates (19812004) are based on systematically over-optimisticestimates both o the growth rate o consumption and o changes in its distribution.Even in the absence o the other problems highlighted above, this would impart asignifcant optimistic bias both to the estimated rate o poverty reduction and (to amore limited extent) to the current level o poverty or any given poverty line.

This does not substantially aect the results or India or China, both o which havesurvey data or 1980 or 1981 and or 2004 (that is, or the beginning and the end othe period covered by the World Banks latest global poverty estimates) although,

as discussed previously, there are substantial problems with the estimates or thesecountries as a result o the estimation o their PPP exchange rates. (There are alsosome questions about the reliability o the survey data or China in the 1980s.72)

Elsewhere, however, the issue is a serious one, especially in the 1980s, when thenumber o surveys conducted was particularly limited. There are almost no actualdata or Sub-Saharan Arica or Eastern Europe and Central Asia beore the late1980s, and ew or the Middle East and North Arica.73

The scale o the problem can be assessed by considering the proportion o totalestimated poverty accounted or in each year by countries where lack o surveydata means that the orward or backward projection method must have been used.

In 1981, 97 per cent o total poverty outside China and India (at the $1-a-day level)appears to have been estimated in this way. Only rom 1987 is even a quarter o thetotal based on actual survey data or estimates based on surveys beore and aterthe year in question. The problem in the most recent years is limited to some extentby the increased requency o household surveys; but in 2004, the latest available



estimate, 90 per cent o $1-a-day poverty outside India and China was estimated byorward projection.

Two other points are noteworthy in this context. First, it is inevitable (barring a veryconsiderable increase in the requency o household expenditure surveys) thatthe Banks estimates o poverty outside India and China or the most recent yearswill always be based very largely on orward projections. The optimistic bias o thismethod thereore suggests that they will consistently tend to present an artifciallyavourable picture o recent developments, exaggerating any acceleration in povertyreduction or concealing any deceleration or reversal.

Second, the absence o reliable estimates or the 1980s is particularly unhelpulbecause o the considerable uncertainty surrounding poverty during this period,when the debt crisis frst struck, and the frst and most painul phase o structuraladjustment took place. This issue is particularly acute in Sub-Saharan Arica,where nearly 90 per cent o estimated poverty prior to 1987 (and 60 per cent in19871990) was based on backward projections rom later household surveys. Thissuggests that even the Banks estimate o the increase in poverty during this period(rom 42.3 per cent in 1981 to 47.2 per cent in 1987) is likely to understate the trueimpact.

Conclusion

At frst sight, the $1-a-day poverty line might appear to provide a clear and easilyunderstood way o defning and measuring poverty consistently on the globallevel. In practice, however, it is ar rom clear, and prone to misinterpretation by theuninitiated. It is inconsistent between countries and over time, highly sensitive torelatively minor changes in methodology which should make no dierence, andprone to serious methodological and analytical problems; and there are a numbero serious gaps in the data which can at best be flled only very approximately.

While some o these problems merely give rise to major uncertainties regardingthe true picture, some lead systematically to an under-estimation o the extent opoverty or over-estimation o the rate at which it is alling, and others to distortionsin the pattern o poverty between poorer and less poor countries. The scale o theseproblems is such that this approach cannot be considered to provide a reliable

basis or estimating the extent, severity, distribution or composition o poverty, or otrends over time.

All o these problems arise (more or less) equally, irrespective o the actualincome level at which the poverty line is drawn. In addition, the $1-a-day line itselis essentially arbitrary, and clearly does not provide a standard o living whichcould reasonably be regarded as morally acceptable in an era o unprecedentedprosperity or the better-o. More generally, because o the wide dierences in theliving standards associated with a given level o income between countries, thistype o approach would appear to be incapable o generating a poverty line whichgenuinely reects a morally-based defnition o poverty, in any meaningul sense, ina consistent way across countries.

In short, the current approach to poverty estimates does not provide anadequate basis either or the estimation o poverty as it is (or appears to be)currently defned, and still less or a morally based defnition. An alternativeapproach is urgently needed.



This suggests the need or a dierent approach to defning and measuring poverty,which:

P takes eective account o non-income aspects o poverty (while also ideallybeing measured in terms o income);

P corresponds with equivalent living standards in dierent countries and contexts;

P sets a threshold or poverty at a standard o living which might reasonably beconsidered as morally acceptable; and

P avoids the technical problems associated with purchasing power parityexchange rates and price indices.

In considering alternatives, we start rom the concept o poverty as a moral issue that a household should be defned as (income) poor i its income is insufcientto achieve what is considered to be a minimum morally-acceptable standard oliving. There are various ways o translating this type o approach in practice, which

can be summarised in terms o two key distinctions.

1. One can distinguish between input defnitions o poverty, based on estimatingthe level of income requiredby a household to achieve a consumption patternthat provides an adequate standard o living; and outcome defnitions, based onthe living standards actually achievedby households at a given level o income.

2. One can either defne a single universal poverty line, defned in money termsat the global level, which is then converted into local currency units in eachcountry; or a set o consistent national poverty lines, defned according tospecifc local contexts, in such a way that they correspond with equivalent livingstandards in each country.

This typology is illustrated in Table 1.

Input-based approaches

As Ravallion notes, going back to Rowntrees work in the UK in the late nineteenthcentury:

The most common approach in dening an absolute poverty line is toestimate the cost of a bundle of goods deemed to assure that basicconsumption needs are met in the specic domain of the povertycomparison. The difculty lies in identifying what constitutes basic needs.74

The frst problem is that o defning what minimum consumption needs are inany universal sense, or example by reerence to basic human needs. This is, inprinciple, easible or absolute physical needs, such as nutrition, protection romthe elements, health care and physical security. The question here is, primarily,where the threshold should be drawn although this is by no means always asstraightorward as it might frst appear, as discussed later in the context o nutrition.

Alternative approaches

In the light o the serious problems discussed earlier, it is difcultto disagree with the dismissive statements about the $1-a-day linealready quoted, or that an alternative is needed which both provides

a truer picture o the actual situation and better reects the moralcontent implicit in the concept o poverty.



Human needs go ar beyond this most basic physical level, however, and simplydefning a set o goods which a household requires merely or its continued

physical existence excludes equally undamental needs, such as social interactionand mental stimulation. Even people unable to ulfl their physical needs may wellvalue the ulflment o such non-physical needs as highly as what might, to anoutside analyst, be considered necessities.

Moreover, even i we could defne both needs and appropriate thresholds in ameaningul way, the bundle o goods and services which would be required to fullthese needs would dier very considerably between and oten within countries,or example between hot and cold climates, between rural and urban areas,or according to the local availability o dierent basic oods. It would also dierbetween households within a particular locality according to their demographiccomposition. For example, the number and ages o children have a major eect onthe types and quantities o oods (and other products) a household requires. (SeeBox 2 later.)

As a result, even within a particular national context:

Unavoidably, determining which goods and services are to be included in apoverty bundle, and in what amounts, is an arbitrary decision.81

One might even question the concepto such absolute needs, and considerthat it is only coherent to consider what it means to be poor in a particular socio-cultural context. This view has a longer pedigree even than Rowntrees subsistenceapproach.

In dening necessities, Adam Smith argued that one must include, not onlynecessary commodities but also those which the custom of the countryrenders indecent for creditable people even of the lowest order to bewithout.82

A comprehensive input-based approach: Morris et als Minimum Income orHealthy Lie (MIHL)Perhaps the most rigorous approach to developing a country-specifc input-based poverty line is the work by Jerry Morris et al. at the London School oHygiene and Tropical Medicine. Taking as their starting point attainable levels ohealth as a human right and prime goal o society,83 Morris et al. frst defne ademographic group, and assess its material needs or a healthy lie, covering dietand nutrition, housing, physical activity, medical care and social integration, basedon epidemiological evidence (supplemented by pragmatic judgments). They thentranslate this into a poverty line on the basis o estimated cautious pragmatic,representative minimal costs per week.84

Table 1: Typology o morally-based approaches to the defnition o poverty

Input-based Outcome-based

Universal povertyline

International FoodPoverty Line (IFPL)

NanakKakwaniand HyunSon75

nutrition Ethical PovertyLine (EPL)

Peter Edward76 health(narrowdefnition)

Country-specifcpoverty line

Minimum Incomeor Healthy Living(MIHL)

JerryMorriset al77,78

health(broaddefnition)

Rights-BasedPoverty Line(RBPL)

David Woodwardand SaamahAbdallah79

multiplecriteria(childsurvival asillustration)Global Capability-

Based Approach(GCBA)

SanjayReddyet al80

multiplecriteria(nutrition asillustration)



They have applied this approach to two groups: single, healthy men, 18 to 30years, living away rom their amily and on their own in the UK,85 and people aged65 and above, living independently (i.e. in non-assisted housing), retired romemployment and without signifcant defned disability in England.86 For 1830-year-olds, they ound costs varying between 106.47 and 163.86 per week (average131.86) in dierent regions at 1999 prices; and or people over 65, 131 or asingle person, and 208 or a couple in April 2007.

It will be noted that these poverty lines are ar above any considered in thecontext o developing countries. In act, they are in the order o $2550 per person

per day. Moreover, even setting aside the issue o appropriate price indices, thiscomparison actually understates the dierence signifcantly, because the HMILfgures exclude subsidies such as ree health services, ree or subsidised publictransport (or people over 65 in many areas) and ree television licences (or peopleover 75).

The scale o this dierence largely reects the much higher standards o livingwhich are regarded as the minimum acceptable in developed than in developingcountries. Thus among the costs included or those over 65, under the heading opsychosocial relations/social inclusion/active minds, are:

a telephone, occasional gifts to grandchildren and others, modest

recreational and entertainment costs, membership fees, a television set (andlicence for those under 75), a daily newspaper, an annual UK holiday and alittle money for hobbies.

While such social needs are (largely) taken or granted in the developed world,their consideration in the analysis o poverty in developing countries is virtuallyunheard-o.

Morris et als approach also highlights the high level o specifcity which is requiredor a comprehensive needs-based approach. The goods and services required by1830-year-olds are very dierent rom those o people over 65. Moreover, the lattergroup excludes 40 per cent o older people who have disabilities, and thereoreare likely to have extra personal costs that require urther ad hoc study. Within

the 1830 group, there are wide regional variations (largely as a result o housingcosts), the highest-cost region indicating fnancial needs 50 per cent above thelowest or an equivalent living standard.

The results also raise questions about the use o per capita consumption as a basisor applying a poverty line: it will be noted that the per capita cost or a couple over65 (104) is substantially less than or a single person (131) although single-person poor households are likely to be much less common in most developing-country settings than in the UK.

The MIHL approach is probably the most comprehensive attempt yet to evaluatea country-specifc poverty line based explicitly on consumption needs although

some arbitrariness inevitably remains in the assumptions about both theconsumption basket and the prices. No equivalent appears to have been attemptedin a developing-country setting, no doubt partly reecting the very considerabledata requirements, and the extent o the analysis required to estimate poverty linesspecifc at a sufcient level o regional and demographic disaggregation.

This may also reect in part the implicit abandonment o the right to health in thedeveloping world in the ace o poverty on a scale which makes it unthinkablein any meaningul sense or the majority o the worlds population. At best,considerations o health in this context all ar short o the defnition establishedby the constitution o the World Health Organisation in 1946: not merely theabsence o disease, but a complete state o physical, mental and social well-being. In relation to developing countries, the defnition rarely extends beyond

the absence o [physical] disease, and is oten still more narrowly drawn in termso mortality.



Food-based input approaches: general considerationsWith the notable exception o Warren et als MIHL approach, as described above,input-based approaches to poverty are in practice almost invariably based onnutritional needs. This largely reects the serious constraints o data and analyticalcapacity in developing countries, and:

the absence of any consensus on what non-food capabilities are ofconcern, on the characteristics of the commodities which promote them, onthe transformation function that relates these characteristics to capabilities,and on the levels of each capability that ought to be deemed minimally

adequate.87

Nutritional needs, in turn, are generally simplifed to the ulflment o calorierequirements, which can readily be assessed on the basis o ood purchases, usingstandard calorifc values.

The general approach is to assess minimum nutritional needs (generally basedon average calorie requirements); to estimate the cost o this, on the basis o theaverage cost o ood per calorie o a reerence group o the population; and toadd an allowance or non-ood expenditure, based on actual expenditure patternsamong poorer households. Two such approaches are discussed below.

In practice, however, this process is less straightorward than is oten assumed.First, reducing nutritional needs to calorie requirements is a serious over-simplifcation, as calories are only one aspect o nutritional requirements. From ahealth perspective, an adequate nutritional intake also depends on other attributeso ood, such as its contents o protein, vitamins, minerals and other micronutrients.For households with young children, especially while weaning, energy density isalso very important.88

It also neglects other aspects o ood which may also be valued by households asmuch as their nutritional content, even i they are below minimum calorie intakes, orexample taste, variety89 and time required or preparation.

Such dierences among oods in terms o non-calorie attributes (nutritional and

non-nutritional) give rise to a considerable variation in average costs per caloriebetween income levels, even below the level at which adequate nutritional intakesare reached. In the case o Bangladesh, or example, Kakwani and Son fnd that thecost per 1,000 kilo calories to the richest 20 per cent o the population is 2 timesthat o the poorest 20 per cent.90 This makes it critically important which incomegroup is used as the reerence group to estimate per calorie costs.

Secondly, even calorie requirements vary considerably between individuals, notablyaccording to age, gender, pregnancy and lactation, levels o physical activityand health status. This is o particular importance because these actors varysystematically though in dierent directions with household income levels(see Box 2).

While demographic considerations are likely to reduce the calorie requirements opoor households relative to the population as a whole, health and activity levels willhave the opposite eect, so that the overall eect will be determined largely by thebalance between the two. While the scale o the eects suggests that the net eectcould be substantial, it may vary considerably between dierent contexts.

It is clear, however, that these eects will tend to distort fndings on the compositiono poverty. Specifcally, it is likely to result in methods based on national averagesor nutritional requirements under-estimating poverty in rural and mining areasand in areas with a high incidence o parasitic and other diseases; and to over-estimate it among amilies with above-average numbers o children, emale-headedhouseholds and among older people.



There is also a more undamental challenge to calorie-based approaches, in termso the hierarchy o human needs.92 Calorie consumption, together with drinking,is the most basic human need which requires fnancial resources (others, such asbreathing and sleep, being costless). The imperative or survival thereore meansthat some level o calorie intake (though not necessarily the notional minimumrequirement) will be met, i necessary, at almost any cost in terms o other needs.

I, however, we are interested in a broader defnition o income poverty that is,i we think that people also have a moral entitlement to the ulflment o materialneeds other than an adequate calorie intake then this does not provide theappropriate criterion. Rather than setting the poverty line at the point where the rsthuman priority is met in ull, we should be seeking to defne the poverty line asthe point at which the lasto the needs which we consider to be an entitlement isulflled. On this basis, by defnition, a poverty line based on calorie intake inevitablyrepresents a minimalist approach to defning a moral poverty line.

Food-based approaches two illustrations

Two recent proposals or input-based poverty lines using the approach outlined aboveare noteworthy. Kakwani and Sons International Food Poverty Line (IFPL) approachestimates a ood poverty line or an anchor country (Bangladesh), which is then usedas a basis or estimating poverty lines in other low-income countries.93 The median othese lines ($1.22 per day at 1993 PPP) is then used as a global poverty line.

Box 2: Individual variations in calorie requirements and relationship with

household income

Calorie requirements vary substantially between individuals in ways that are signifcantly linked to income levels.91

P Age: children require ewer calories than adults, rising rom 950 at age 612 months to 2210 (emale) and 2895(male) in adulthood. To the extent that the proportion o children is higher in poorer households, this will tend toreduce calorie requirements at lower incomes. Conversely, requirements all ater age 60, by 30 per cent or menand by 17 per cent or women. The eect here is ambiguous: since lie expectancy is shorter at lower incomelevels, one might expect ewer over-60s in poorer households; but the more limited earning potential o theelderly puts them at higher risk o poverty.

P Gender: the calorie requirement or an active adult man is some 30 per cent higher than that or an activeadult woman, with smaller dierences or children rom the age o 10. To the extent that households withoutan adult male breadwinner are more likely to be among the poor, this may on average tend to reduce calorierequirements among the poor slightly relative to the non-poor.

P Pregnancy and lactation: womens calorie requirements are increased by 9 per cent during pregnancy, and by23 per cent during lactation. To the extent that poorer households on average have higher ertility rates, this willtend to increase calorie requirements among the poor, partly o-setting the eects o their higher proportion o

children. Pregnancy and lactation also substantially increase needs or protein and micronutrients, and henceper-calorie costs or an adequate diet.

P Activity: people who expend more energy require a commensurately higher energy intake; and the work opoorer households is typically much more physically demanding than or the better o (especially in agriculture,but also, or example, in mining, construction and other hard physical labour). A larger proportion o themembers o poorer households, including women and children, are also likely to be economically active, otenin such occupations, than at higher income levels. Similarly, economically inactive women in poor householdsalso typically have to expend much more energy on very physically demanding activities such as poundingbasic oods, carrying water over long distances, growing ood crops or household consumption, etc. This islikely to increase calorie requirements considerably among poor households.

P Health: ill-health, and especially chronic or repeated ill-health, requires substantially higher calorie intakes toachieve a given nutritional outcome (as measured by anthropometric indicators such as height-or-age, weight-or-age and weight-or-height). O particular importance are some parasitic diseases, which may substantiallyreduce the nutritional value to the body o a given intake o calories and other nutrients. Since poorerhouseholds typically suer considerably higher burdens o disease (not least parasitic diseases) than the better-o, this can be expected to increase their calorie requirements substantially.



While the resulting poverty line is less arbitrary than the $1-a-day line, it has threeproblems (beyond those inherent in calorie-based approaches).94 First, Kakwaniand Son base the ood poverty line on the cost per thousand kilocalories o thepoorest 20 per cent o the population, rather than o households which haveadequate calorie intakes. This results in an artifcially low ood poverty line; and,since the non-ood component is based on this fgure, it too will be artifcially low.

Secondly, while the theoretical part o their paper rests on using PPP exchange ratesbased on ood prices, they appear to use general PPP exchange rates in their actualestimates. As noted above, this will result in poverty lines being under-estimated in

countries poorer than the anchor country, and over-estimated in less poor countries.Even using ood PPPs, the problem remains that these will be dominated by theconsumption patterns o non-poor households and thereore may not accuratelyreect dierences in ood costs to the poor. This will distort the conversion o theood poverty line between countries.

Thirdly, there is the general problem o using a global line: those countries wherethe estimated poverty line is above the median level will have a poverty line belowthe level appropriate to that country on nutritional grounds, while those where theline is below the median will have a line higher than would be appropriate.

These problems are dealt with better by Reddy et als proposed Global Capability-

Based Approach to poverty. Reddy et al. avoid the last two problems by proposingthe use o country-specifc lines, estimated on a consistent basis or all countries.They also use ood expenditure at the level o income where (average) calorierequirements are met, thus avoiding the frst problem. This results in a much widerrange o poverty lines than ound by Kakwani and Son: while the ratio between thehighest and lowest o the 19 national poverty lines estimated by Kakwani and Sonis 1.55, the corresponding fgure or the three countries considered by Reddy et al.is more than double this, at 3.4.

I, however, our primary interest in poverty arises rom the living standards itresults in, the real test o a poverty line based on nutrition is to compare its resultsagainst nutritional outcome indicators and this raises serious doubts about bothapproaches.

In Figures 1 and 2, we compare the incidence o poverty implied by Kakwani andSons 19 estimated national poverty lines (as applied to the World Banks PovCalNetdatabase95) in each o the years in which income data are available, with WorldBank fgures or the incidence o malnutrition in each country or the nearestavailable year, in terms o height-or-age and weight-or-age.96

Two eatures stand out rom these graphs. First, there is no systematic relationshipbetween the proportion o households defned by Kakwani and Sons methodas poor, and the proportion o children who suer rom malnutrition. In act, thecountries with the highest rates o malnutrition have ood poverty rates in themiddle o the distribution, while those with the highest rates o ood poverty have

average (height-or-age) or below-average (weight-or-age) malnutrition indicators.This suggests that the method does not eectively reect dierences in purchasingpower or nutritional standards between countries.

Second, the incidence o malnutrition is higher than the estimated incidence opoverty (indicated by points ab

how poor is poor 0

Documents