Poverty

International Poverty Centre June 2007

InequalityThe challenge of

2 International Poverty Centre

F R O M T H EE D I T O R

Dag Ehrenpreis

Poverty in Focus is a regular publication of theInternational Poverty Centre (IPC). Its purposeis to present the results of research on povertyand inequality in the developing world.

IPC is a joint project between the United NationsDevelopment Programme and Brazil to promoteSouth-South Cooperation on applied povertyresearch. It specialises in analysing poverty andinequality and offering research-based policyrecommendations on how to reduce them. IPCis directly linked to the Poverty Group of theBureau for Development Policy, UNDP and theBrazilian Government’s Institute for AppliedEconomic Research (IPEA).

IPC Director (acting)Terry McKinley

EditorDag Ehrenpreis

International Advisory Board

Desktop PublisherRoberto Astorino

Front page: Photomontage by Dr. Kurt Källblad,Malmö, Sweden, as an illustration of the diagramat the bottom of page 5: “A visualization of globalincome distribution”. It includes parts of photosby George Bosela, Wellington, Ohio, USA;Bernhard Brück, Munich, Bavaria, Germany; DavidDavis, Union, Kentucky, USA; Projeto10 (Chico,Antonio, Tiago, Eduardo), Atibaia, São Paolo,Brazil; Kurt Källblad; Birgitta Lagerström, Brasilia,Brazil; and Xiskya Valladares, Madrid, Spain.

Editor’s note: IPC is grateful for the kindpermission to use published papers and booksas a basis for articles in this issue, provided bythe authors and the following publishers: UNU-WIDER, UN-DESA, UNDP-Human DevelopmentReport Office, and Cambridge University Press.The sources of graphs and tables are thereference publications at the end of eachrespective article. Thanks also to all the authorsfor generously contributing their intellectualproducts without any monetary remuneration.

International Poverty CentreSBS – Ed. BNDES, 10º andar70076-900 Brasilia DF Brazil

povertycentre@undp-povertycentre.orgwww.undp-povertycentre.org

The views expressed in IPC publications arethose of the authors and do not necessarilyreflect the views of IPC, IPEA or UNDP.

Oscar Altimir, CEPAL, Santiago de ChileGiovanni A. Cornia, Università di FirenzeNora Lustig, UNDP/BDP Poverty GroupGita Sen, Indian Institute of Management, BangaloreAnna Tibaijuka, UN Habitat, NairobiPeter Townsend, London School of EconomicsPhilippe van Parijs, Université de Louvain

I nequality is a major challenge for poverty reduction and a crucial obstaclefor achieving the Millennium Development Goals. There are both intrinsic

and instrumental reasons why inequality matters, such as social justice and morality,putting the poor first, growth and efficiency, political legitimacy, and public policygoals (see page 13).

More rapid income poverty reduction requires both a more rapid pace of growth anda more pro-poor pattern, which implies a reduction of the inequalities that limit theprospects for poor people to share in the opportunities created by economic growth.The poverty impact of growth has been shown to be more than 10 times higher inthose countries that combined growth with falling rather than rising inequality.

However, all forms of inequality may not necessarily be harmful. While inequalitiesin opportunities are clearly both inefficient and unfair, it can be argued (see pages 20and 23) that income differences reflecting varying degrees of effort—and that provideincentives—in education, work and risk-taking entrepreneurship, are helpful for theliving-standards of the poor over time. On the other hand, such “good” inequalitiesarising in one generation may lead to “bad” inequalities for the next, which will stillbe unacceptable for all the above-mentioned reasons.

This issue of Poverty in Focus highlights inequality as one of IPC’s priority areas ofresearch. It provides some of the most important recent research results on the extentof inequality in the distribution of wealth and incomes at both the global and nationallevels, on analytical aspects of causes and patterns, and on policy conclusions andrecommendations.

UNU-WIDER first presents a summary of a new study of the global distribution ofhousehold wealth, which is seen to be even more unequal than that of incomes.

Branko Milanovic analyses the concepts of global income inequality and its componentfactors, including the role of globalisation, with a proposal for global redistribution.

José Gabriel Palma reveals and analyses a fact that has been missing in the inequalitydiscourse: while income distributions vary a lot at the top and bottom across countries,they are remarkably similar for the middle classes.

Michael Kremer shows why globalisation tends to increase inequalities in poor countries,contrary to the standard economic textbook model. Outsourcing is one major reason.

Kevin Watkins discusses the five main reasons why inequality is bad for growth,democracy, social justice and cohesion, and for the MDG prospects. Hence UNDPstresses the need for public policies to focus more on reducing inequalities.

Elizabeth A. Stanton outlines how the UNDP Human Development Index could, andshould, be adjusted to make it more sensitive to multidimensional inequalities.

Frances Stewart stresses the need to measure and monitor horizontal inequalitiesamong groups in order to promote efficiency, growth, peace and poverty reduction.

Marcelo Medeiros puts the inequality issue on its head by focusing on the rich,who should provide the resources for the necessary redistribution of income.

Francisco Ferreira underlines the policy importance of distinguishing analyticallybetween “good” and “bad” inequalities, using cholesterol as a metaphor.

Shubham Chaudhuri and Martin Ravallion apply this distinction to the recent patterns ofunequal growth in China and India that may threaten future growth and stability.

Björn Gustafsson, Li Shi and Terry Sicular summarise a forthcoming book on inequalityand public policy in China, concluding that changes are necessary in the balancebetween the processes that affect inequality.

Joana Costa, Marcelo Medeiros and Rafael Guerreiro Osório consider inequality of people’stime use, including unpaid work, across social classes and the gender divide.

This collection of articles is meant to contribute to a better understanding of theimportance of reducing inequalities in its various forms, and thus to policies andprogrammes that will more effectively reduce poverty and social injustice.

Poverty In Focus June 2007 3

The Global Distributionof Household Wealth

by James Davies1, Susanna Sandström2,Anthony Shorrocks2 and Edward Wolff 3

While the richest 10 percent of adultsin the world own 85 percent of globalhousehold wealth, the bottom halfcollectively owns barely 1 percent. Evenmore strikingly, the average person in thetop 10 percent owns nearly 3,000 times thewealth of the average person in the bottom10 percent. These are some of the resultsthat emerge from a new UNU-WIDER studyof the distribution of household wealth.

We estimate the level and distribution ofwealth across all countries in the worldusing a comprehensive concept ofhousehold wealth. In everyday conversationthe term ‘wealth’ often signifies littlemore than ‘money income’. On otheroccasions economists interpret the termbroadly and define wealth to be thevalue of all household resources, bothhuman and non-human. Our study assignswealth its long-established meaning of networth: the value of physical and financialassets less debts. In this respect, wealthrepresents the ownership of capital.Although capital is only one part ofpersonal resources, it is widely believedto have a disproportionate impact onhousehold wellbeing and economicsuccess, and more broadly on economicdevelopment and growth.

The estimates of wealth levels are basedon household balance sheets and wealthsurvey data which are available for 38countries. Fortunately, these include manyof the rich OECD countries as well as thethree most populous developing countries,China, India and Indonesia; so the datacover 56 percent of the world’s populationand 80 percent of household wealth.Careful analysis of the determinants ofwealth levels in these countries allowsimputations to be made for countrieswithout data.

The estimates of wealth distribution arebased on household asset distribution

data for 20 countries. For countrieswithout this type of direct information,the degree of wealth concentration wasestimated from income distribution data(where available), using the relationshipobserved between income and wealthinequality in countries with both kinds ofdata. The remaining countries, coveringonly a few percent of world population,were assigned the average wealthdistribution pattern for their region andincome class.

Household wealth is more concentrated,both geographically and in sizedistribution when official exchange ratesare employed rather than purchasingpower parity (PPP) valuations. Thus asomewhat different perspective emergesdepending on whether one is interestedin the power that wealth conveys in termsof local consumption options or thepower to have influence on the worldfinancial stage. Since a large share ofglobal wealth is owned by people who canreadily travel and invest internationally,it is more appropriate to use officialexchange rates when studying theglobal distribution of wealth than whenlooking at the global distribution ofincome or poverty.

Global household wealth in the year 2000amounted to $125 trillion, equivalentto roughly three times global GDP or to$20,500 per citizen of the world, by officialexchange rates. In terms of PPP dollars,the corresponding world value wasPPP$26,000 per capita, roughly the sameas the average level in Poland or Turkey.

Wealth levels vary widely across nations.Among the richest countries, meanwealth was $144,000 per person in theUSA and $181,000 in Japan. Lower downamong countries with wealth data areIndia, with per capita assets of $1,100,and Indonesia with $1,400 per capita.

Global wealth iseven more unequallydistributed than income.

The richest 2 percentof adults in the worldheld more than half ofglobal wealth.

The average personin the top decile ownsnearly 3,000 times thewealth of the averageperson in the bottom one.

The concentration ofwealth within countriesvaries significantly,but is generally high.

These are some key factsin a new UN study on globalwealth distribution, which issummarised here.

1. University of Western Ontario;2. United Nations University – World Instititute forDevelopment Economics Research (UNU-WIDER);

3. New York University.

4 International Poverty Centre

The regional pattern of asset holdingsshows wealth to be heavily concentratedin North America, Europe, and highincome Asia-Pacific countries whichtogether account for almost 90 percentof global wealth. Although NorthAmerica has only 6 percent of the worldadult population, it accounts for 34percent of household assets. Europeand high income Asia-Pacific countriesalso own disproportionate amounts ofwealth. In contrast, the overall shareof wealth owned by people in Africa,China, India, and other lower incomecountries in Asia is considerably lessthan their population share, sometimesby a factor of more than ten.

Comparing per capita wealth and percapita GDP across countries showsthat wealth is distributed even moreunequally than income. High incomecountries tend to have a higher share ofworld wealth than of world GDP becausetheir wealth to income ratios are abovethe world average. The reverse is true ofmiddle and low income nations.

Wealth to income ratios are especiallyhigh in the UK, Italy, and rich Asiannations. Lower than expected values arerecorded for eastern European countrieslike the Czech Republic and Poland, alongwith the Nordic countries and SouthAfrica. Eastern European countriesare a heterogeneous group with manydifferent features. In this region, privatewealth is on the rise, but has still not

reached very high levels. Assets likeprivate pensions and life insurance areheld by relatively few households. In theNordic countries, the social securitysystem provides generous public pensionsthat may depress wealth accumulation.South Africa is rich in resources and haswell-developed financial institutions; butthe fact that the country has a large low-income population and exhibits some ofthe characteristics of less-developedcountries, may account for the lowwealth-income ratio.

Estimation of the world distributionof wealth requires information to becombined on wealth differences betweencountries and within countries. Theconcentration of wealth within countriesvaries significantly, but is generally high.The share of the top decile ranges fromaround 40 percent in China to 70 percentand beyond in the United States and certainother countries. Typical Gini coefficientsfor wealth lie in the range 0.65-0.75, andsome are above 0.8. In contrast, the mid-range of income Ginis is 0.35 to 0.45.

Wealth inequality for the world as awhole is higher still. Expressed in termsof the adult population of the world, weestimate that net assets of $2,160 peradult in the year 2000 was sufficient toplace a household in the top half of theworld wealth distribution. At least$61,000 per adult was needed to belongto the richest 10 percent of households,while membership of the top 1 percent

required a little over $500,000 per adult.The latter figure indicates that a familyneed only be moderately wealthy inWestern terms to be among the toppercentile of world wealth-holders.

Our results show that the top wealthdecile owned 85 percent of global wealthin the year 2000. The richest 2 percent ofadults in the world held more than halfglobal wealth, and the richest 1 percent ofadults alone accounted for 40 percentof all household assets. In contrast, thebottom half of the world adult populationowned barely 1 percent of global wealth.The Gini value for global wealth isestimated to be 89 percent; the sameGini value would be obtained if $100were shared amongst 100 people in sucha way that one person receives $90 andthe remaining 99 get 10 cents each.

Given the high concentration of wealthin North America, Europe, and rich Asia-Pacific countries, it is not surprising todiscover that almost all of the world’srichest individuals live in these countries.The breakdown of the global wealthdistribution in the chart shows that theAmericas, Asia and Europe contributeabout one third of the members of theworld’s wealthiest decile. China occupiesmuch of the the middle third of the globalwealth distribution while India, Africa andlow-income Asian countries dominate thebottom third. For all developing regionsof the world, the share of populationexceeds the share of global wealth, whichin turn exceeds their share of people whoare among the wealthiest in the world.

Major differences are observed acrosscountries in the composition of assetholdings, a result of different influenceson household behaviour, such as marketstructure, regulation and culture. Realassets, particularly land and farm assets,are more important in less developedcountries. This reflects not only thegreater importance of agriculture, butalso immature financial institutions.

The types of financial assets that areowned also show striking differencesacross countries. A breakdown betweensavings accounts, shares and equities,and other financial assets, shows thatsavings accounts feature strongly intransition economies and in some rich

Poverty In Focus June 2007 5

Convergence and divergence between North and South

The gap between the North and the South widened continuously from 1950 to 1990, as is clear from the following chart. For Africa and LatinAmerica, the gap has continued to grow until the present; but since 1980, Asia, and China in particular, has been closing it. After 1990, thefaster growth of Asia as outweighed the relative decline of Africa and Latin America, and so, the relative position of the South as a whole hasvery marginally improved. The contrast between these two opposed tendencies is a central feature of world distribution during the last twoor three decades.

The global income distribution is represented in the following graph, in which each vertical column represents the implicit income of onedecile of one country. The tallest building in the city, so to speak, in the rear corner of the graph, represents the income of the richest decileof the population of the USA, while the scarcely visible column in the nearest corner of the graph represents the income of the poorest decileof the population of Sierra Leone. The graph illustrates extreme global inequality, in particular of the very steep ascent from the poorermajority of the world to its richer minority.

Bob Sutcliffe: A Converging or Diverging World? DESA Working Paper No. 2, UN, New York 2005.http://www.un.org/esa/desa/papers/2005/wp2_2005.pdf

Asian countries, while share-holdingsand other types of financial assets aremore evident in rich countries in theWest. Part of the explanation is poorlydeveloped financial markets in transitioncountries, while savings accounts arefavoured in Asian countries becausethere appears to be a strong preferencefor liquidity and a lack of confidence infinancial markets.

Finally, and perhaps surprisingly,household debt is relatively unimportantin poor countries. While many poorpeople in poor countries are in debt,their debts are relatively small in total.This is mainly due to the absence offinancial institutions that allowhouseholds to incur large mortgage andconsumer debts, as is increasingly thesituation in rich countries. Many people

in high-income countries havenegative net worth and—somewhatparadoxically—are among the poorestpeople in the world in terms ofhousehold wealth.

J. Davies, S. Sandström, A. Shorrocks,E. N. Wolff: The Global Distribution ofHousehold Wealth. WIDER Angle No. 2,2006. UNU-WIDER. http://www.wider.unu.edu/newsletter/newsletter.htm

Source: Author’s calculations based on data in A. Maddison: The World Economy: A Millennial Perspecitve, OECD, 2003.Note: The North = Australia, Canada, Japan, New Zealand, USA and Western Europe; The South = the rest of the worldexcept for Eastern Europe and ex--USSR; Asia includes China, but not Japan.

Data source: World Development Indicators, 2005. Online version. World Bank, Washington D. C.

6 International Poverty Centre

What is meant byglobal inequality?

How large is it?Is it increasing?Why does it matter?What can be done about it?

How much is due todifferences betweencountries and how muchto within-country gaps?

Is there a link betweenglobalisation and globalinequality?

There are three main concepts ofglobal inequality that need to be clearlydistinguished, but are often confounded:1) inequality among countries’ per capitaincomes, unweighted or 2) weighted bycountries’populations, and 3) inequalitybetween world individuals.

Concept 1 importantly deals withconvergence and divergence of countries’average incomes, but not with incomeinequality among all people in the world.Economic growth in a small country willnot have the same effect on globalinequality as growth in a poor andpopulous country. Concept 2 inequalitytakes this into account by weightingeach country by its population. Thisimplicitly assumes that each individualwithin a country receives the per capitaincome. Thus, it does not take intoaccount within-country inequalities.Concept 3 includes inequalities bothbetween and within countries.

Research using different techniquesindicates general agreement about thesize of Concept 3 inequality, but generaldisagreement about its recent directionof change. All Gini values for the 1990s liewithin a relatively narrow range between63 and 68, with the exception of the twoextremes (61 and 71), and most of theseestimates are within one standard errorof each other.

How big is a Gini of around 65? It is largerthan inequality found in any singlecountry including South Africa and Brazil,two of the most unequal countries in theworld with Ginis around 60. The Gini valuehowever does not give an intuitive feelingof how large global inequalities are.A better way to look at it is to considerhow global income is distributed acrossdifferent fractiles of the distribution.Thus, the top 5 percent of individuals inthe world receive about one-third of totalworld (PPP-valued) income, and the top

10 percent one-half. If we take thebottom 5 and 10 percent, they receiverespectively 0.2 and 0.7 percent of worldtotal income. This means that the ratiobetween the average income receivedby the richest 5 percent and the poorest5 percent of people in the world is 165to 1. The richest people earn in about48 hours as much as the poorest peopleearn in a year.

Some studies find declining values fromthe 1980s, others increasing. My ownfindings are of zigzag movements, with anincrease of 3 Gini points 1988-93 followedby a decline of 1 point by 1998, and then a1 point increase by 2002. This is explainedfirst, around 1990, by the slow growth ofrural incomes in India and China andeconomic collapse of Eastern Europe, bothof which contributed to global inequality.When both developments reversed in thenext five-year period, global inequalitydecreased. But these are zig-zags causedby specific economic events in largecountries, not a trend.

Is there a link between globalisationand global inequality? This is a verycontentious issue. Most agree on theassociation of openness with enhancedglobal growth but not on its distributionbetween rich and poor countries. Rapidgrowth in huge poor countries like Chinaand India is clearly reducing Concept 2inequality and—although to a largeextent offset by increasing nationalinequalities—probably also Concept 3inequality, at least in some periods.

Globalisation’s impact on globalinequality varies with the position ofcountries with different attributes alongthe international income distribution at agiven point in time. The effect will dependon where in the income hierarchypopulous countries are at a given point intime. If they are poor, globalisation willreduce global inequality; if they are rich,

by Branko Milanovic,the World Bank Global Income

Inequality

Poverty In Focus June 2007 7

the opposite may happen. Therelationship between globalisation andglobal inequality is not generally valid,but highly time-specific and contingenton past income history.

How much of global inequality is due todifferences in mean incomes of countriesand how much to income differenceswithin countries? Some 70 percent ofpresent global inequality is due todifferences in countries’ mean incomes.This is a sharp reversal from a situationwhich existed around the time of theIndustrial Revolution when more thanhalf of the rough estimate of globalinequality was due to income differenceswithin nations.

Yet, some people in a poor country canbe better off than some people in a richcountry. The chart plots five countries’distributions within the world one.The top curve shows that the poorest5 percent in Germany have a meanincome at the 73rd percentile of the worldincome distribution; the richest 5 percentare in the top percentile of the world.For India, the span is from the 3rd to the71st percentile in the world. These twodistributions do not overlap at all. Butif we compare Brazil and Germany, nearlya third of all Brazilians are richer than thepoorest 5 percent of the Germans, andso are over 200 million Chinese. The richBrazilians are about as rich as the richestGermans, and much richer than therichest 5 percent in India, on average.

The chart illustrates not only theconsiderable inequality due to within-country distributions, but also the practicalimplications for global transfers. Transfersfrom mean-income rich (OECD) to mean-income poor countries, when we do nota priori know who the beneficiaries are,need to take recipient countries’ incomedistributions seriously. The risk thatmoney from a German taxpayer willbenefit someone richer is higher ifGerman aid goes to Brazil than to India.

Global inequality matters, for variousreasons. From an ethical perspective,distributional justice within a nationand in the world as a whole is the samething. And pragmatically, globalisationincreases the awareness of otherpeople’s income and thus the perception

of inequalities among both the poorand the rich. Even if globalisationwere to raise everybody’s real income,it could exacerbate, rather thanmoderate, feelings of despondencyand deprivation among the poor.

Thus, globalisation is just like the processwhich led to the creation of modern nationstates out of isolated hamlets. Nationalincome distribution was similarly anabstraction for the people who didnot interact with each other, and almostignored each others’ existence and way oflife. However, once nation states came intoexistence, national inequality became anissue because people were able to observeincome differences. If the process ofglobalisation would lead toward decision-making processes at the global level, thenglobal inequality is indeed relevant.

Large income differences in the world aredue, as we have seen, mostly to the largedifferences in countries’ mean incomes.Since the early 1980s, many countriesof the world, viz. the poorest ones, havewitnessed a systematic growth failure.Thus, to reduce income differencesamong individuals, increasing thegrowth rate in poor countries is ofparamount importance. Still, therewill be a need to reduce incomedisparities through global redistribution.

Three basic progressivity rules shouldguide global income redistribution.Funds should flow from: (i) rich to poor

countries; (ii) a tax-payer who is richerthan the beneficiary of the transfer; and(iii) tax payers who are relatively richwithin their own country to relativelypoor people in the recipient country, sothat inequality decreases in both donorand recipient countries. This requiresconsideration of national incomedistributions with preference to poor andegalitarian countries, since transfers tothem are unlikely to be globally regressive.

This calls for the creation of a globalagency to be financed by taxing richpeople in rich countries and that wouldtransfer funds to poor people in poorcountries. If empowered to raise its ownfunds, it should eschew governmentsthat have often wasted foreign aid.Instead, it should deal directly withnational NGOs and individual citizens inpoor countries and distribute collectedfunds in the form of cash grants.

Vesting some modest tax-raising authorityfor the first time in history into a globalagency would be a huge political challenge.However, globalisation is rendering theeconomic gaps more obvious and thefairness of the existing global distributionmore questionable. Opponents willultimately realise that their self-interest liesin supporting some form of global actionto deal with both poverty and inequality.

Branko Milanovic: Global Income Inequality.A review. World Economics, Vol. 7, No. 1,Jan-March, 2006. See also http://www.un.org/esa/desa/papers/2006/wp26_2006.pdf

Note: World population along vertical axis by percentile from poorest to richest.Country populations along horizontal axis from poorest to richest, mean income by ventile (5 % group) in PPP dollars.Source: Calculated from World Income Distribution (WYD) data at: http://econ.worldbank.org/projects/inequality

8 International Poverty Centre

An important stylised facthas been absent from thedistributional analysis,which only reflects whathappens at the very topand at the bottom ofthe distribution.

The other half—in themiddle and upper-middleof the distribution—offersa very different picture of aremarkable homogeneity.

Reducing inequality withincountries requires transfersfrom the top to the bottomof the income distribution.

One of the oldest political economycontroversies is about why countries havesuch different income distributions. Atone end of the range of hypotheses thefocus is on anonymous market forces andoptimum equilibria that generate efficientdistributive outcomes; at the other end, thefocus is on different forms of exploitationand power relations based on specifictypes of structures of property rightsand incentives that favour some anddisadvantage others in a systematic way.

This short essay contributes to this debateby highlighting a remarkable stylised factthat has been absent from the analysisaround distributive diversity. This is thatthe picture of the distributional diversityacross the world as measured in Ginicoefficients only reflects what happenswithin half the world’s population—thoseat the very top and those at the bottom.The other half is characterised by aremarkable distributional homogeneity.

The graph below suggests four ‘layers’ ofinequality across countries. First, a moreequal layer containing Central Europeand the non-Anglophone OECD; a secondlayer contains a great variety of regions,with more than three-quarters of theworld’s population; a third one includesSub-Saharan Africa and the ‘second-tier’NICs; and a fourth, Latin America.

However, as discussed in detail in Palma(2006), one has to look ‘inside’ this Gini-picture to be able to properly understandcross-country distributional diversity, inparticular the remarkable differencebetween income distribution in one halfof the world population (those at the verytop and at the bottom in each country)and in the other (those in the middle).

Figure A (page 9) shows a particularlyclose correlation between regional Ginisand the income-shares of the top 10percent; this is mainly the result of theway the Gini index is calculated. In turn,Figure B shows that the regionaldistributional structure of the bottom40 percent is the mirror image of that ofthe Ginis and decile 10. Therefore, the Giniindices are perfectly reflected both at thevery top and at the bottom of thedistribution of income.

However, when one looks at the other 50percent of the world’s population, Figure C,the ‘middle and upper-middle incomegroups’, which are located between deciles5-9, the regional distributional picturechanges completely: from huge disparityto remarkable similarity. Furthermore,as Figure D indicates, this distributive-homogeneity is even more remarkable inthe ‘upper middle’ (the 30 percent locatedbetween deciles 7-9). This contrast isclearly shown by the coefficient ofvariation; those of decile 10 and deciles1-4 are nearly four times greater than thatfor deciles 5-9 and seven times larger thanthat for deciles 7-9.

Why is Inequality soUnequal acrossCountries?

by José Gabriel Palma,University of Cambridge

LA=Latin America; AF=Sub-Saharan Africa; EA1=‘first-tier’ NICs; EA 2=‘second-tier’ NICs; SA=large South Asia andlow-income Southeast Asia; NA=North Africa; CA=Caribbean countries; OECD 1=English-speaking OECD; OECD2=non-English-speaking OECD; ex-c 1=ex-communist countries of Central Europe; ex-c 2=ex-communist countriesof the former Soviet Union.Regional figures are median values, based on a sample of 90 countries. The width of the X-axis corresponds to the actualrange of income per capita of the sample (from Ethiopia to Switzerland).Source: World Bank’s WDI data set.

Poverty In Focus June 2007 9

Thus, the half of the world populationbelonging to the ‘middle classes’ seems tobe able to appropriate about half theirrespective national incomes. The apparentparadox is that—often through verydifferent distributive mechanisms—theyseem to be able to acquire a similar‘property right’ over half their nationalincomes. They apparently do so regardlessof per capita incomes, political settlements,institutional structures for rent-seeking, thestructure of property rights and incentives,economic policies, or whether theircountries managed to get their prices,institutions or social capital ‘right’!

No such luck for the bottom 40 percent ofthe population. For them, such politicaleconomy issues can make the differencebetween getting as much as one-quarterof national income (as in the non-Anglophone OECD, or Central Europe),or as little as just over 10 percent (LatinAmerica). As far as the top income decileis concerned, the sky is (almost) the limit.

Thus, the regional distributional structuresuggested by the Gini indices only reflectsthe income disparities at the very top andat the bottom, but does not reflect theremarkable distributional homogeneityfound in middle. This phenomenon raisesserious questions regarding how usefulthe Gini index is as an indicator of overallincome inequality.

There are also major analytical implicationsderiving from this phenomenon. Inparticular, that recent political andeconomic developments (includingglobalisation) seem to have beenassociated with two very differentdistributional movements: a (better known)‘centrifugal’ one in terms of the income-shares of the top and bottom deciles, anda (lesser known) ‘centripetal’ movement interms of the income-share of deciles 5-9.

Regional distributional homogeneity inthe middle and upper-middle of thedistribution also casts doubts on manytheories trying to explain the diversityof income distributions across the world.For example, mainstream theories tend toemphasise the role of ‘human capital’ inthe determination of income distribution;according to them, the level of educationis a crucial variable (if not the most crucialvariable) in the determination of income

inequality. However, in all regions of theworld the top income decile is made upof individuals with relatively high levelsof education, while those in the bottomfour deciles have relatively low levels offormal education—in terms of eitherrelatively little schooling, or (in the moreadvanced countries), schooling of ratherdoubtful quality.

So why is it that in the mosthomogeneously-educated group (the topincome decile) one finds the greatestdistributional diversity? In turn, why isthere extraordinary similarity in the sharesof national income of the educationallyhighly heterogeneous population ofdeciles 5 to 9 (heterogeneous, forexample, in terms of the share ofthe population with secondary andespecially tertiary education)?

Obviously, more research needs to bedone on the forces shaping the nationalincome shares of these two halves ofthe world population along suchdifferent paths, particularly in suchopposite ‘centrifugal’ and ‘centripetal’directions. Remarkably, this simple

observation does not seem to havebeen emphasised before.

Finally, following this analysis, it seemsrather obvious that countries reallywishing to improve their incomedistribution should focus their efforts onhow to transfer income from the very topto the bottom of the income distribution.

One such policy, of course, would be tostrengthen workers’ property rights overhuman capital—the strengthening ofworkers’ property rights over basic skillswas one of the main mechanisms bywhich most industrialised economiestransferred income from the top tothe bottom of the distribution (andencouraged the accumulation of humancapital). In turn, the reversal of this policyhas been one of the main mechanismsby which most developed countriesmanaged to increase their incomeinequalities since the early 1980s.

Palma, J. G. (2006): Globalising Inequality:the ‘centrifugal’ and ‘centripetal’ forcesat work. DESA Working Papers No. 35.http://www.un.org/esa/desa/papers/2006/wp35_2006.pdf

10 International Poverty Centre

Under a standard textbookmodel, globalisation shouldbenefit the poor andreduce inequality.

However, recentresearch finds thattrade liberalisation isassociated more withincreased inequality inpoor countries.

Outsourcing bymultinationals increasedthe relative demandfor skilled labour,increasing the paygap to unskilled labour.

This article summarisesa model of globalisation ascross-border production.

by Michael Kremer,Harvard University Globalisation and

Inequality withinCountriesSupporters of the anti-globalisationmovement argue that globalisation hasincreased inequality between and withinnations and in particular that it hasmarginalised the poor in developingcountries and left behind the poorestcountries. Meanwhile, more moderatemainstream politicians argue that thepoor must invest in education to takeadvantage of globalisation.

However, under a standard textbookmodel, globalisation should benefit thepoor and reduce inequality, and thepoorest countries and least educatedworkers should have the greatestopportunity to benefit from globalisation.

The argument goes as follows. Supposethere are two countries, the North, with ahigh ratio of skilled to unskilled workers,and the South, with a low ratio. Supposealso high-skill and low-skill workers arecomplements. Under autarky the wageof skilled workers will be relatively lowin the skill-abundant North and relativelyhigh in the skill-scarce South. Openingtrade will equalise factor prices in thetwo countries. Hence, the wage of skilledworkers will rise in the North and fall inthe South, while the wage of unskilledworkers will fall in the North and risein the South.

Thus inequality will rise in the rich countryand fall in the poor country. The extentof, and gains from, trade will typically begreater the scarcer are skills in the South.Similar results obtain in a model withcapital and labour as the two inputs,assuming labour is equally distributedwithin each country while capital is not.

There are, however, at least two empiricalproblems with this model. First, it predictsthat bilateral trade will be greatest whenfactor endowments are most different; infact, there is little trade between

advanced countries such as the U.S.and very poor countries such as Chad.Second, evidence from specific developingcountries following trade liberalisationand from cross-country studies doesnot suggest that trade liberalisationgenerally reduces inequality in poorcountries and in fact frequently suggeststhat trade liberalisation is associated withincreased inequality.

For example, after Mexico embarked on abroad liberalisation of trade and foreigninvestment, the return to schoolingincreased; white-collar real hourly wagesincreased by 13.4 percent 1984-90, whileblue-collar wages fell by 14.0 percent. Thebiggest rise in inequality was observed infirms engaged in export industries. Risingwage inequality in Mexico is linked tocapital inflows from abroad. Outsourcingby Northern multinationals shiftedproduction towards skill-intensive goods,increasing the relative demand for skilledlabour. Multinational firms and jointventures pay higher wages, with orwithout adjustment for observablecorrelates of skill.

Increased openness is associated withreduced wage inequality in the AsianTiger economies in the 1970s and 1980sbut with increased inequality in LatinAmerica in the 1990s. Limited evidenceavailable for other countries indicatesthat liberalisation tends to be followedby increases in inequality, but causalityis doubtful. In several large countries,e.g. India, China, Russia and Indonesia,liberalisation had been only partial.In the case of China global integrationhas proceeded further in coastal thanin hinterland regions; this coincideswith increased coastal-hinterland wageinequality, while Chinese regions thatexperience a greater degree of opennessin trade also tend to have greater declinesin rural-urban income inequality.

Poverty In Focus June 2007 11

Most recent research finds either thattrade liberalisation is associated withincreased inequality in poor countriesor finds no strong association. Therelationship between openness andinequality appears positive for low-income countries and negative for highincome countries, with the turning pointoccurring somewhere in a GDP per capitarange of $6000-$13,000.

In sum, the evidence does not supportthe expected theoretical effects ofglobalisation consistently reducinginequality in poor countries. In basictrade theory, two countries tradeproduced goods with one another.However, recently globalisation has to alarge extent also involved the productionprocess. A single product can bemanufactured out of componentsmade and assembled in differentcountries, or designed in one countryand manufactured in another.

Workers in poor countries not onlyproduce labour-intensive products, butalso jointly produce with workers in richcountries. This process is modelled inthe end-reference paper. In particular,we model globalisation as cross-borderproduction; a product is designed inone country, manufactured in a secondand customer services provided by a callcentre in a third country.

The model involves production byworkers of different skill-levels and isconsistent with (i) the small scale oftrade between countries with verydifferent factor endowments and(ii) the possibility that globalisationmay increase inequality in both richand poor countries.

Our analysis assumes two countries andjust one consumption good. The richcountry has workers of two skill levels,A and B. The poor country has workersof skill levels C and D. We assume that theA-D skill levels are in alphabetical order.In a stylised two-type model this maybe a reasonable assumption. The lowestquartile of the U.S. skill distribution maywell be higher than the highest quartileof the Indian distribution; the presence ofsmall numbers of very high skill workersin India or very low skill workers in theU.S. makes little difference to our results.

The model implies that it is efficient (a) fora firm to assign a higher-skill worker tothe managerial task and a lower-skillworker to the assistant’s role, and (b) touse cross-matching, i.e. workers withdifferent skill levels working in the samefirm, than self-matching, i.e. workers withthe same skill level working together.

Before globalisation, A- and B-workerscould be cross-matched, although notnecessarily, and the same for C- andD-workers. But international matcheswere not possible: for example, B- andC-workers could not be cross-matched.By contrast, after globalisation, all cross-matches are in principle possible. Workersfrom different countries can worktogether in the same firm.

We assume that D-workers are of lowenough skill so that it is not efficientfor them to be cross-matched with anyworker in the rich country. That low-skilled workers in poor countries havedifficulty participating in cross-bordermatching is consistent with theevidence. For example, call centresin India tend to employ middle-classIndians who can speak with an Americanaccent with which U.S. customers arefamiliar. Multinationals and exportersin developing countries typically paymanufacturing wages substantiallyabove the norm for the country.

Globalisation potentially allowsefficiency gains through cross-borderproduction. If skill levels in the rich andpoor countries are sufficiently disparate,then it is inefficient for any rich countryworkers to match with any poor countryworkers. Therefore the model offers aclear explanation of why very little tradeis observed between the U.S. and Chad,for example.

Furthermore, in this model globalisationcauses inequality in the poor country(the gap between the wages of C- andD-workers) either to increase or to remainthe same. The basic reason behind theincrease in inequality in poor countriesis the additional potential matches thatglobalisation brings to each type ofworker. For C-workers, globalisation opensmore possible matches; for D-workers itdoes not. In this sense D-workers can bemarginalised by globalisation in this model

if before globalisation they matched withC-workers but are afterwards forcedto self-match.

The assumption that D-workers are so lowskilled that it is inefficient for them to beinvolved in international matching isimportant. For example, if beforeglobalisation C- and D-workers match,while afterwards A- and C-workers matchand B- and D-workers match, thenglobalisation has added new matchingpossibilities for both C- and D-workersand so no conclusion can be drawn aboutinequality changes without consideringthe exact values of A, B, C and D and thenumbers of workers of each type.

Similarly, without making strongerassumptions, we cannot pin down whateffect globalisation has on inequality inthe rich country other than to say thatthe wages of the A- and B-workerscannot both go down; all othercombinations are possible.

Finally, observe that our model makesno clear prediction on trends in globalinequality, even in the cases whereinequality increases within both richand poor countries. Precise resultsdepend on the relative numbers of A-,B-, C-, and D-workers as well as on therelative skill levels. However, if peoplemeasure their status relative to othersin their own society, then they willperceive inequality as increasing. Thisanalysis corresponds to the view ofmany anti-globalisation protestors thatglobalisation benefits elites in both richand poor countries.

Of course, while joint production may bea good model of some types of trade, theproduct trade model may be appropriatein other cases. Low-end shoe factories inpoor countries may hire relatively low-skill workers. Still, the presence of someindustries in which foreign investorstypically hire medium-skill workers whoare high-skill relative to others in theircountry may help explain why there isnot a clear equalising effect of trade inpoor countries.

Michael Kremer and Eric Maskin:Globalisation and Inequality.http://www.economics.harvard.edu/faculty/kremer/papers/GlobalizationInequality_Oct06.pdf

12 International Poverty Centre

Inequalities based onwealth, region, genderand ethnicity are bad forgrowth, democracy, socialjustice and cohesion,and for the MDGs.

The disparities hamperingprogress towards the MDGsare systemic—transmittedacross generations; andthey reflect publicpolicy choices.

This should haveimportant implicationsfor MDG and povertyreduction strategies.

There are five mainreasons—intrinsic andinstrumental—whyinequality matters forhuman development.

Inequality is a fundamental issuefor human development. Extremeinequalities in opportunity and lifechance have a direct bearing on whatpeople can be and what they can do—that is, on human capabilities. Childrenfacing a higher risk of death becausethey are born into a poor household orbecause they are girls clearly have lessopportunity to realize their potential.Inherited disadvantage in opportunityis wrong for intrinsic reasons: it violatesbasic precepts of social justice. There arealso strong instrumental reasons for aconcern with inequality. Deep disparitiesbased on wealth, region, gender andethnicity are bad for growth, bad fordemocracy and bad for social cohesion.

They are also bad for the MillenniumDevelopment Goals (MDGs). The MDGsdo not directly address inequality. Inthis sense they are distribution neutral.Progress is measured by aggregatingand averaging change at a nationallevel. In theory, the MDGs could be meteven if, say, households with lowincomes were falling behind on theincome poverty and health targets,or if the rate of reduction in childdeaths among boys was sufficientto compensate for a slower rate ofreduction among girls.

The distributional blind spot of theMDGs is a weakness on two counts. First,the MDGs themselves are rooted in ideasabout global justice and human rights.They are universal entitlements, notoptional or discretionary allowances. Itfollows that progress should be for all,regardless of economic status, gender,parents’ wealth or location in a country.

Yet the MDGs do not remindgovernments that success in advancingtowards the MDGs should be measuredfor all of society and not just in the

aggregate. The disparities hamperingprogress towards the MDGs are systemic.They reflect complex hierarchies ofadvantage and disadvantage that aretransmitted across generations—andthey reflect public policy choices.

Secondly, poor people are beingleft behind across many of the MDGs;progress among the poorest 20 percentof the population is far below thenational average in a large group ofcountries. Apart from being unjust, this issuboptimal from the perspective of MDGattainment. People who are poor accountfor a far larger share of deprivation thanpeople who are not. It follows thataccelerated progress among poor peopleis one of the most effective routes tofaster national progress. Current patternsare slowing the overall advance becausethe smallest gains are being registeredamong the households that account forthe biggest part of the problem.

These considerations have importantimplications for the design of MDGstrategies. The evidence shows that a“trickle down” approach to reducingdisparities and maintaining overallprogress will not work. The MDGs setquantifiable targets that lend themselvesto policy responses rooted in technicaland financial terms. Ultimately, however,the real barriers to progress are socialand political. They are rooted in unequalaccess to resources and distribution ofpower within and among countries.Unless these inequalities are corrected,the first principles of the MillenniumDeclaration—commitment to socialjustice, equity and human rights—fromwhich the MDGs are derived will not betranslated into progress in humandevelopment at the required rate.From a human development perspectivethere are a range of mutually reinforcingintrinsic and instrumental reasons why

by Kevin Watkins,United Nations

Development ProgrammeInequality andHuman Development

Poverty In Focus June 2007 13

inequality matters. These can be broadlysummarized under five headings:

Social justice and morality. The view thatthere are limits to tolerable deprivationis fundamental to most societies andvalue systems. All major religionsexpress concerns with equity andplace obligations on their adherentsto address extreme deprivation as amoral duty. Public ideas reflect widernormative concerns. Surveys showstrong opinions in many countries thatthe gap between rich and poor is toolarge, thus indicating an underlyingperception of social justice.

Putting the poor first. More weightshould be given to improvementsin the well-being of the poor anddisadvantaged than to the rich andprivileged. National income is not agood measure of welfare, because itignores the distribution of income.Beyond income, many of the samearguments apply. For example, anadditional unit of public spendingdirected towards reducing child deathsor extending access to primary schoolwould be preferable on social groundsto a similar amount spent on transfers toservices for high-income groups.

Growth and efficiency. Extreme inequalityis not just bad for poverty reduction—it isalso bad for growth. Long-run efficiencyand greater equity can be complementary.Poor people remain poor partly becausethey cannot borrow against future

earnings to invest in production, theeducation of their children and assets toreduce their vulnerability. Insecure landrights and limited access to justice cancreate further barriers to investment andpro-poor growth.

Political legitimacy. Extreme inequalitiesalso weaken political legitimacy andcorrode institutions. Inequalities inincome and human capabilities oftenreflect inequalities in political power.Poor people (especially women), ruralpopulations and indigenouscommunities are disadvantaged partlybecause they have a weak political voice,and vice versa. Where politicalinstitutions are seen as perpetuatingunjust inequalities or advancing theinterests of elites, democracy andstability can be undermined.

Public policy goals. Most societies seereducing poverty and removing unjustinequalities as important goals for publicpolicy. Extreme disparities undermine thepursuit of these goals, and limit the rateat which growth can be converted intopoverty reduction. Similarly, extremedisparities in health and educationreduce the scope of disadvantagedgroups to take advantage ofopportunities for improving welfare.

The appropriate response is to ensurethat inequality and the measures toovercome disparities in life chancesfigure more prominently in the designof poverty reduction strategies.

Inequalities in income reflect thedistribution of assets and opportunityand the operation of markets. But theyare influenced by government taxationand spending. In many countries fiscaltransfers are already narrowing extremeinequalities. In Chile, for example, theynarrow the gap between the incomeratios of the richest and the poorest20 percent of the population from 20:1to 10:1. From a human developmentperspective the fiscal transfers with thehighest returns are investments thatbuild capabilities and provide protectionduring periods of acute vulnerability

An obvious requirement for meaningfulfiscal transfers to alleviate poverty is thewillingness—and capacity—of the stateto mobilize revenue. In much of LatinAmerica aversion to taxation restricts thiscondition. Mexico raises only 13 percentof GDP in revenue—less than Senegaldoes. India’s capacity to redistribute thebenefits of higher growth through thefiscal system is similarly constrained bya tax to revenue ratio of only 10 percent.After two decades of growth that ratiohas not increased.

Fiscal transfer is one mechanism forraising the income of the poor abovethe level dictated by current growthand distribution patterns. More broadly,pro-poor growth requires a publicinvestment focus on the markets inwhich poor people operate. In manycountries the challenge is to shift thepolicy focus to the smallholder producersand to the more marginal areas thataccount for the bulk of poverty.

Control over assets is critical. It issometimes argued that there is apotential trade-off in agriculturebetween greater equity through landreform and greater growth. Here too thetrade-offs are more apparent than real.Redistributive reforms in agriculturehave proven results in reducing poverty,leading to major advances in manycountries. In West Bengal, India, agriculturaloutput and incomes rose followingtenancy reform and recognition ofthe land rights of the poor.

Inequality and Human Development.Chapter 2 of UNDP Human DevelopmentReport 2005. http://hdr.undp.org/reports/global/2005/pdf/HDR05_chapter_2.pdfSource: UNDP Human Development Report 2005, page 57.

14 International Poverty Centre

Both per-capita income andHDI are based on nationalaverages, thus ignoringdistributional inequalities.

The correlation acrosscountries between thethree components ofHDI are very weak.

Inequalities in eachcomponent may havecorrosive effectson social well-beingand social cohesion.

An inequality-adjustedHDI could incorporateseveral new elementsthat make HDI moresensitive to inequality.

A first test of such anIHDI shows significantdifferences from the HDIcountry rankings.

Adjusting the HumanDevelopment Indexfor Inequality

by Elizabeth A. Stanton,Tufts University

This histogram shows all rank changes from HDI to IHDI; five countries kept the same rank, 22 had worse ranksby IHDI, and the remaining 19 had better ranks. The average absolute change in rank was 3.3, which is quite largefor only 46 countries.

Inequalities in health, education,and income—key components of humandevelopment—matter deeply to socialwell-being. Yet the best-known measuresof well-being either ignore distributionalinequalities altogether or at best accountfor only some of their effects. Per capitaincome, the most common measure, is asimple average. Its main alternative, theUNDP Human Development Index (HDI),is likewise based on national averages,albeit for a wider set of welfare indicators.

The practice of identifying averages withnational well-being ignores potentialtrade-offs between increasing averagesand decreasing differences in distribution.For example, as the rich get richer,average income may increase, butincome inequality simultaneously mayincrease so sharply that the incomes ofthe poor decline, arguably resulting in adecrease in well-being. More generally,measures based solely on nationalaverages record unambiguous changesin well-being in circumstances madeambiguous by changes in inequality.

HDI is derived from three componentindices: health, as proxied by average lifeexpectancy; education, as proxied by aweighted average of literacy and schoolenrollment rates; and income per capita,

using the natural logarithm in order toaccount for the diminishing marginalutility of income. The rationale for HDI isthat average health and education arenot simple functions of average incomeper capita. There are two reasons for this.

First, health and education have asubstantial public goods component;they are not private goods, distributedentirely according to income. Publiclyprovided goods and services may beunequally distributed as well, becauseaccess to them is politically driven andaffected by discrimination on the basisof race, ethnicity, religion, or gender.Because inequalities in the distributionof health and education have negativeeffects on human well-being, and are notsimply a function of income inequality,they too should enter into measures ofsocial welfare.

Second, if privately-provided goods andservices for health and education purposesdiminish on the margin as income rises (i.e.when the relationship between individualincome and individula health/education isconcave in mathematical terms), thencountries with the same average incomebut different income distributions will havedifferent levels of average health andeducation; the country with greater incomeinequality will have lower average healthand education levels, and hence a lowerHDI. This is called the “aggregation effect”of income inequality on health andeducation. It implies that a redistribution ofincome would change the average levelof health and education.

Thus, HDI’s inclusion of averagehealth and education goes some waytoward capturing the effects of incomeinequality on well-being. But it fails toaccount for other well-being effects ofincome inequality, as well as for relevanteffects of inequalities in the distributionof health and education outcomes.For example, inequalities in all threecomponents of HDI may have corrosiveeffects on social well-being through theirassociation with decreasing social

Poverty In Focus June 2007 15

Source: Author’s calculations using data from the reference study.

cohesion, increasing violence, orincreasing environmental degradation.Moreover, there is evidence that many, ifnot all, people put some intrinsic valueon limiting inequality as an end in itself.

With the exception of gender disparities,the inequality measures sporadicallyreported in the UNDP HumanDevelopmentReport (HDR) have been restrictedexclusively to income distribution, andnone have been incorporated into theHDI itself. The first HDR stated that allthree average measures of humandevelopment “conceal wide disparitiesin the overall population,” but thatcompared to income inequality, the“inequality possible in respect to lifeexpectancy and literacy is much morelimited: a person can be literate onlyonce, and human life is finite.” Althoughhealth and education inequalities arequantitatively more limited than incomeinequality, the replacement of binaryvariables, like literacy and schoolenrollment, with continuous variables,like years of schooling, allows for thedetection of more inequality.

A ranking of countries by HDI differs fromranking by per capita income, because(i) average life expectancy and educationregister the effects of income distribution,and (ii) health and education levels aredetermined in part by non-incomefactors, in particular public provisionof social services. While these effectsmake HDI a better measure of socialwelfare than per capita income, threeadditional ways in which inequality canaffect well-being are missing from HDI:

The aggregation effects of inequalityin health and education on individualwelfare (as opposed to theaggregation effects of inequality inincome on health and education);Shifts in the way individual well-being is related to personal income,health and education, due to theeffects of inequality of other welfare-relevant factors that are not includedin the HDI; andInequality’s intrinsic negative effectson social well-being.

For these reasons, it is both desirableand feasible to reformulate the HDIto push the boundaries of well-beingmeasurement beyond national averages.An Inequality-adjusted HumanDevelopment Index (IHDI) canincorporate several new elements thatmake HDI more sensitive to inequality.

First, in IHDI, social welfare functions withrespect to health, education, and incomeare modeled with diminishing marginalreturns to individual welfare to capturethe aggregation effects of inequality.Second, IHDI can vary the weights ofindividuals in calculating social well-beingso as to reflect the degree of emphasison equality. Third, Gini coefficients areused to adjust the resulting compositeindices—in a way that maintains balancebetween the components—to take intoaccount further instrumental andintrinsic costs of inequality beyond theaggregation effect.

GDP per capita and HDI are commonlyused as measures of social welfare toindicate which countries’ policies havebeen the most effective in providing thebest quality of life. When social welfare ismeasured without reference to inequality,these rankings incorporate conceptualflaws. HDI thus ranks some countries, like

South Korea, too unfavourably, andothers, like Brazil, too favourably.

Because distributional inequalities areomitted from HDI, progress in improvingsocial welfare may be overlooked—as maycertain kinds of deterioration of socialwelfare. IHDI can both provide a betterranking of countries at any given timeand better illuminate changes in socialwelfare over time. While the data necessaryto calculate IHDI are not yet available forthe full set of countries covered in theHDRs, the IHDI proposal provides aroadmap to a more robust measure ofsocial welfare for use in internationaland inter-temporal comparisons.

Stanton, Elizabeth A.: Accounting forInequality: A Proposed Revision of theHuman Development Index. Working PaperNo. 119, November, 2006. PERI: Amherst,Mass. http://www.peri.umass.edu/fileadmin/pdf/ working_ papers/working_papers_101-150/WP119.pdf

16 International Poverty Centre

Horizontal inequalities areimportant because theyaffect well-being and canlead to violent conflict.

Group inequality providespowerful grievances thatleaders can use to mobilisepeople to political protest.

It should be measuredand monitored.When acute, policiesare needed to countersuch inequalities.

This may help achieveseveral objectives,including efficiency,growth, peace andpoverty reduction.

Why HorizontalInequalities Matter

by Frances Stewart,Oxford University

Horizontal inequalities (HIs)among culturally defined groupsencompass economic, social, politicaland cultural dimensions. They areimportant because they affect well-being and can lead to violent conflict.Yet almost all economic analysis andmeasurement of inequality, e.g. the Ginicoefficient, concerns vertical inequalityamong individuals, while groupinequality tends to be ignored.

People can be grouped in many ways,and most people are members of manygroups including national, racial,ethnic, religious, gender and agecategorisations. These may emergefrom self-identification, legislation,and/or as a result of categorisation byothers. There are also many groups ofwhich membership is short-lived—suchas social clubs, or producer networks,and so on.

Some group affiliations are clearlymore important than others in terms ofpeople’s identity, their permanence andthe impact on access to resources.While group boundaries are to someextent arbitrary, socially constructedand fluid, nonetheless some groupaffiliations are salient, affectingpeoples sense of self, their welfare andtheir political affiliations. The groupforms part of a person’s identity, andrelative impoverishment of the groupincreases perceptions of members thatthey are likely to be permanentlytrapped in poverty.

In addition to being important in itself,reducing HIs may help achieve otherobjectives. One is efficiency:discrimination is likely to lower efficiency,since a number of talented people will beheld back while resources and highpositions go to less talented people inthe favoured group. Also, povertyreduction may require tackling the

position of the group. For example,programmes to advance credit to poorproducers, or to promote universaleducation, may not be achievable solong as group inequality remains, e.g inattitudes towards women’s rights andgirls’ education.

Group inequality provides powerfulgrievances that leaders can use tomobilise people to political protest, bycalling on cultural markers (a commonhistory or language or religion) andpointing to group exploitation. This isespecially likely where there is political aswell as economic inequality, so that theleaders are excluded from political power,for example Côte d’Ivoire, Rwanda,Northern Ireland, Chiapas, and the Sudan.

HIs are an important source of grievanceand potentially of instability, independentlyof the extent of vertical inequality. WhetherHIs lead to political violence or notdepends on a number of factors, e.g.the nature of the HIs, which can becategorised into four areas: politicalparticipation, economic resources, socialservices, and cultural recognition.

While these categories are applicableto every society, the elements that arerelevant in a particular case depend onthe nature of the society, its politicalsystem, economy, and social structure.For example land is clearly ofparamount importance in the ruraleconomies of Africa but not in modernurban societies, while employmentseems to be important in mostcountries. In natural resource richeconomies, the control over suchresources, either directly or via the state,is an important source of group rivalry.Access to housing is of criticalimportance in more developedeconomies, such as Northern Ireland,but less so where people mostlyconstruct their own homes, e.g. in Africa.

Poverty In Focus June 2007 17

Tackling horizontal inequalities in Malaysia

Malaysia faced large economic HIs between the Chinese and Malay (Bumiputera)populations. Following anti-Chinese riots in 1969, the government introduced reformsto reduce HIs and thus secure national unity. Policies included expanding the Malayshare of capital ownership through low-interest credit allocations; settling Malayson 95 percent of new lands; and quotas in access to education. Inequalities were thusreduced in incomes, capital ownership and education. Conflict was avoided, even afterthe 1997 financial crisis when other countries in the region faced anti-Chinese violence.Moreover, there was high growth and rapid poverty reduction. Political HIs favoured themajority Malays, facilitating the comprehensive and effective policy reforms.

Strong causal connections betweendifferent HIs cause and perpetuate HIs.For example, HIs in political power oftenlead to similar social and economicinequalities through biased access togovernment jobs and services. There arealso mutual linkages between economicand social elements, e.g. education andeconomic opportunities, thus reinforcingcycles of privilege and deprivation.

HIs may be spatially distributed, whichcan lead to separatist claims whereresource rich provinces seek autonomy,resenting the redistribution of localresources to other parts of the country(for example, Biafra in Nigeria, or Aceh inIndonesia). Yet, sometimes the poorerregions feel exploited by the richer areas(for example, in Bangladesh and Eritrea).

Different types of conflict emerge wherepeople from competing groups live inthe same geographic area. In such cases,the deprived may seek political andeconomic rights or control overgovernment institutions. Similarly, theremay be attacks on particular groups andpressure for ethnic cleansing withoutdirect government involvement.

The violent conflict in Cote d’Ivoire,the long war in Guatemala and thegenocide in Rwanda are examples wheredeep political and socio-economic HIswere an important cause of the groupviolence. However, in Malaysia andNorthern Ireland effective policies wereintroduced to reduce HIs, in each casecontributing to peace.

While the existence of HIs, especiallywhen consistent in political andeconomic dimensions, is likely to lead topolitical mobilisation on group lines,

whether there is conflict or not dependson other factors, including cultural,demographic, economic and politicalconditions (especially the nature ofthe state).

The potential strength of any violentmovement depends on the relative sizeof the population in different groups. Atone extreme where there are many smallgroups they find it difficult to mobilisecollectively, while if the population ishomogeneous violence may be lesslikely. In between there are manypossibilities. Where an underprivilegedgroup is a small minority, any violentprotest is likely to be readily suppressedor may not occur (as in the northern partof Kenya or among the Roma people inEurope). Where there is small privilegedgroup and a large underprivileged one,the majority can be mobilised against theprivileged—the Jews and the Chineseand the Lebanese, for example, havebeen subject to such attacks periodicallyover many centuries.

Another important factor is the cohesionof the group; strong cohesion and hencemobilisation potential may be theoutcome of cultural unity, or of politicalleadership emphasising group unity.However, cohesion is inherently easier insome contexts. For example, where thereis cultural fragmentation (e.g. manylanguages being spoken), mobilisationbehind some overarching identity may beless likely, and geographic dispersion mayalso make cohesion less likely. Thus, inPeru, an overarching affiliation could be an‘indigeneous’ identity which would bringtogether about half the population—a clearly powerful political entity. Yetdifferences within the indigenouspopulation as to language, history,

location may mean that such anoverarching identity is not strongly feltand hence political action on the basis ofindigeneity may be unlikely.

The nature of the economy is anothertype of conditioning factor which helpsdetermine the outbreak (or not) of violentconflict. Impoverished societies seem tobe more prone to violence accordingto econometric investigations. This mayreflect lack of viable occupations amongpoor populations, who therefore find waran attractive proposition; or lack of stateresources which means that the stateoffers little to its citizens which mightlead them to respect their civicobligations; or that weak states that areunable to repress violence effectively.Low growth economies seem to be moreviolence prone than high growth. This isprobably because with higher growth allgroups benefit, and so inequalities seemto matter less.

Another factor is the presence of naturalresources which has been shown to makeconflict more likely. Natural resourcesoften lead to quite acute HIs—oftengeographic—between those with accessto the resources and those without.

General investigation of country casesshows that HIs are important both asa source of conflict and because theyreduce welfare. They should bemeasured and monitored and, wherethey are acute, policies introduced tocounter them. But caution is needed.Countering HIs can itself provokeconflict from the groups whoseprivileges are being reduced—as seemsto have been one factor behind theconflict in Sri Lanka. Moreover thepolitical conditions for the sort ofcomprehensive and systematic policiesneeded are rare. But leaving HIs to festeris dangerous, in Northern societies aswell as in Southern. Monitoring andtackling HIs should be part of thedevelopment agenda, along with otherdevelopment objectives includinggrowth and poverty reduction.

Stewart, F.: Horizontal Inequality: a NeglectedDimension of Development. Helsinki:WIDER Annual Development Lecture 2002.

Stewart, F., G. Brown and A. Langer (2007):Policies towards HIs, CRISE Working Paper 42,www.crise.ox.ac.uk

18 International Poverty Centre

Growth alone will not besufficient to eradicateincome poverty within areasonable time frame oftwo or three decades.

Eradicating poverty willrequire reducing inequalitythrough direct redistribution.

It is not a matter of choosingbetween labour and transferstrategies, but of recognisingthem as complementary.

We know a lot about the poorbut little about the rich, whoshould provide the resourcesfor redistribution.

Learning more about the richis important for improvingthe lives of the poor.

Now, the Richby Marcelo Medeiros,International Poverty Centre

Economic growth is a very tempingstrategy to combat poverty. After allgrowth can be good for everyone. Atleast in theory, with growth both thepoor and the rich can win, thus avoidingany serious distributive conflicts. Andeven if growth comes accompanied byincreases in inequality—as seems to bethe most common case with acceleratinggrowth—it can still improve the life ofeveryone. However, what is a possibilityhas become almost a mantra: growth isgood for the poor.

If it is good for the poor, it is welcome.But good does not mean best and thefact is that in many countries realisticgrowth rates will not be sufficient toeradicate income poverty within areasonable time frame of two to threedecades. If middle-income countries likeSouth Africa, Peru and the Philippines donot reduce inequality, they may doubletheir GDP and still have more than 10percent of their population living belowa $2/day income threshold; low-incomecountries like Kenya would have suchpoverty ratios above 20 percent andIndia would be above the 30 percentlevel. The calculation here is simple: if theeconomy grows without changes ininequality, the headcount ratio for theextreme poverty threshold of a $1/daywill be the share of the population underthe $2/day poverty line after theduplication of the domestic product.

Where increasing the average level ofincome is not sufficient, it is necessaryalso to better distribute the existingresources. For many developingcountries income poverty is not aproblem of generalized scarcity ofresources but mainly a matter of skeweddistribution of national income. Muchof the poverty in the world could bereduced if there were less inequality inthese countries. This is particularly true in

highly unequal middle income countriesin Africa and Latin America, such as SouthAfrica, Botswana, Colombia, Brazil, Chileor Mexico, in which 5 percent of theincome of the richer decile of thepopulation is sufficient to double oreven quadruplicate the income of thepoorest decile, according to the HDR 2005data on poverty and inequality.

It is not just a matter of having lessinequality. Destroying the wealth of therich reduces inequality but it is hard tosee how this would help the poor. Thetype of inequality reduction thesecountries need is the one that resultsin increases of the income of the poor.There are countless ways to do it andof course there is no “one-size-fits-all”solution to this problem, but basicallysuch increase can be obtained by raisingthe labour earnings of the poor andraising social transfers.

Urban and rural land reforms, betteraccess to credit and markets, labourmarket regulation, production subsidiesand the remodelling of control agenciesand bureaucracies can and should beused to increase labour earnings. But notall the poor would be benefited by thesemeasures. In some cases they havelimited scope and affect only certainoccupational groups or regions. In othercases they have to be complemented byso many structural changes in theproduction and distribution ofcommodities that in practice isolatedpolicies are of little use.

Moreover, changes in the demand forlabour will hardly be sufficient toeradicate poverty if they are notaccompanied by changes in the supply,which involves increasing the skills ofthe poor workers and allowing them tofreely migrate to areas and countrieswhere they can find better jobs.

Poverty In Focus June 2007 19

Improving the labour earnings of thepoor requires not only a good effort, butalso some time. Not all alternatives toreducing inequality by increasing thelabour income of the poor will bringsubstantial results in the relatively shortperiod of twenty or so years. Forexample, it takes more than ten years toeducate a cohort of children and evencountries that may accomplish a largeexpansion in their educational systemswill have to wait many more years untilthe new generations of educatedworkers become a majority in the labourmarket. Parallel to that, these countrieswill also have to change the productionand occupational structures to reducesegmentation and absorb new wavesof educated workers. This may take along time as well.

The poor cannot wait that long. If the ideaof eradicating poverty in a reasonabletime frame of two to three decades is tobe taken seriously, direct redistributionwill be inevitable. Improving the labourearnings of the poor is crucial but forthe present generations it will hardly beenough to drastically reduce poverty;as a consequence, income has to bedirectly redistributed to the poor viasocial transfers. It is not a matterof choosing between labour andtransfers but of recognizing theyare complementary strategies.

Redistribution means to take incomefrom one group and give it to another.In the last twenty years there was animpressive progress in terms ofknowledge about the group that will bebenefited by this redistribution, the poor.Ethnographers, statisticians, sociologists,economists and a multitude of otherprofessionals have visited, mappedand described the poor, analysing theirhabits and even testing their behavioursexperimentally. However, little is knownabout the group that preferably shouldprovide the resources for redistribution,i.e. the rich.

Indeed, the rich receive so little attentionfrom researchers that even operationaldefinitions of the group need to bebetter developed. Take, for instance, theever increasing sophistication in thedebate about poverty lines. Drawingsuch lines allows the stratification of the

population and is one of the fist steps inany statistical study of poverty. There ismuch in the literature about the methodsdo draw absolute, relative, objective,subjective, uni- and multidimensionalpoverty lines but not so many attemptsto draw richness or affluence lineshaving redistribution in mind.

Many years ago the first studies of thepoor had to overcome huge obstacles.But thanks to the ingenuity of a largenumber of scholars, today there areseveral tools that allow us to identifyand aggregate the poor, analyze thedeterminants of poverty and estimatethe possible impacts of public policies onpoverty. The study of the rich can benefitfrom these previous efforts. With fewadaptations, the tools used to studypoverty can be easily applied on theresearch about the rich.

Data, however, is still a problem. In thedeveloping countries the existing surveyquestionnaires were not designed tocorrectly collect data on the wealth andincome of the richest groups of thepopulation, although many of thesesurveys have very long and detailedsections on the income and assets of thepoor. This limitation, however, does notprevent research in these countries.

Even underestimating incomes andassets, the existing household surveyscan bring us useful information aboutthe rich. For instance, they allow us toseek answers to questions that areextremely important for theories ofjustice and, therefore, for egalitarianpublic policies: How fluid is thetransition between the non-rich andthe rich strata? Does regular educationopen to everyone the opportunity tobecome rich? How strong is the roleof dynasties for generational mobility?What is the composition of the incomeof the rich? What are their consumptionand saving patterns?

Most of these questions can be exploredusing data that is already available.Although not perfect, this data should beused because, as in any other field ofresearch, a growing number of studiesmay justify improvements in datacollection and lead to the developmentof better theories to understand social

inequality. And, of course, surveys are notthe only source of data to study the rich.There are studies that lay hold of taxinformation, some governmentadministrative records and data fromcompanies, but most of them refer tohighly developed countries. Similarstudies in developing countries arestill in their infancy.

Different from the “positive-sum game”promised by growth, redistribution haswinners and losers and consequently isassociated with conflicts of interest.Because of the interests it affects, it canhardly happen without imposition by thestate. Therefore, redistribution has also apolitical dimension, in addition to theeconomic one. The rich have power anduse it to gain or maintain economicadvantages. However, what seems anobvious fact actually requires moredetailed analysis so the mechanisms thatlink political power to wealth can be fullyunderstood. The approach used by alarge number of studies about socialnetworks and social capital of the poorseem to be a promising path tounderstand where the political powerof the rich comes from, how it relates totheir wealth and how it can blockredistribution initiatives.

It is true that there is some discomfortamong economists about approachingthe rich—surprisingly, since the studyof elites has been consolidated as a fieldof political science for almost a century.But the fact is that to better understandinequality and what can be done toreduce it, the rich must be increasingly puton the research agenda of developmentstudies. There is a need for informationabout who they are, what makes themrich, how they use their wealth, whathappens with their consumption andinvestment when they are taxed, whatare the relations they have with thestate, how much political power theyhave and use, how they benefit frompublic policies and so on. By knowingmore about the rich we will be betterprepared to improve the life of the poor.

M. Medeiros: Poverty, inequality andredistribution: A methodology to definethe rich. IPC Working Paper No. 18, 2006.http://www.undp-povertycentre.org/newsletters/WorkingPaper18.pdf

20 International Poverty Centre

Like cholesterol,inequality can beeither good or bad.

“Good” inequality rewardseffort and leads to betterperformance, while“bad” inequality wasteshuman potential.

Credible decompositions ofinequality into “opportunity”and “effort” componentsare increasingly feasible.

Clarity about differentkinds of inequality mightcreate some politicalcommon ground betweenthe extreme modes ofpublic opinion.

Outside Latin America, where mostpeople view it as an unambiguously badthing, there is no universal consensusthat all inequality is unjust. This is true inacademic and policy circles, as well as inthe broader popular opinion. The WorldValues Survey once asked people in 69countries whether incomes should bemade more equal or, instead, whetherlarger differences were needed asincentives for individual effort. Answerscould take any one of ten values,between those two “extreme” views.The distribution of responses was bipolar:40 percent of respondents were evenlydivided between the two extremes withno pattern in between.

The phrasing of the survey questionitself was revealing. A positive view ofinequality is usually associated withrewards and incentives for theapplication of effort. On the other hand,negative views tend to be stronger ifthe inequalities in question involveopportunities. Many people might beless offended by inequality they seeas caused by differences in work habitsand personal responsibility, than byinequality attributed to differences ininitial life chances, due to factors suchas gender, race or family wealth.

The concept of inequality of opportunityhas a long and distinguished history.Franklin D. Roosevelt once stated that“We know that equality of individualability has never existed and never will,but we do insist that equality ofopportunity still must be sought.”Economists like John Roemer havesuggested that equal opportunitieswould be attained in a situation whereindividual advantage is independent fromcircumstances—exogenous attributesover which individuals have no control,such as family background, place of birth,etc. In the words of Vito Peragine:

“according to the opportunity egalitarianethics, economic inequalities dueto factors beyond the individualresponsibility are inequitable andto be compensated by society,whereas inequalities due to personalresponsibility are equitable and notto be compensated.”

Such a decomposition of observedinequalities into a component dueto pre-determined circumstances andanother due to individual effort mightbe helpful in a number of ways. Clarityabout the different kinds of inequalitymight create some political commonground between the extreme modes ofpublic opinion. It might also help shednew light on the muddled debate aboutthe relationship between inequality andgrowth. Perhaps one reason why theempirical evidence on the causal effectof past income inequality on subsequenteconomic growth is so ambiguous is thatthere really are two offsetting effects:while “good” inequality rewards effortand leads to better performance, “bad”inequality wastes human potential bydenying certain groups the opportunityto invest in themselves. Like cholesterol,there may be more than one kindof inequality.

The distinction may even helppolicymakers choose between moreand less efficient forms of redistribution:policies that reduce inequality ofopportunity affect the premia todifferent predetermined circumstances,and so should give rise to fewerincentive problems. Policies that attackoutcome inequalities by taxing the fruitsof effort, rather than by leveling theplaying field, are likely to be costlier.On the other hand, there may be adanger in overstating the conceptualdifferences. Today’s outcomes clearlyaffect tomorrow’s circumstances. Large

Inequality asCholesterol

by Francisco H. G. Ferreira,the World Bank

Poverty In Focus June 2007 21

permissible inequalities arising fromdifferences in efforts in one generationmay lead to morally unacceptableinequalities in opportunity for the next.

Be that as it may, a practical problemwith this concept is that inequality ofopportunity could not be measureduntil recently. Given the perils andpitfalls of household surveys, it seemedchallenging enough just to get themeasurement of income inequality right.How was one to capture the distributionof “opportunities”, which by definition arepotential, rather than actual, in nature?

Roemer’s distinction betweencircumstances and efforts asdeterminants of advantage has helpedconsiderably. Recent studies haveattempted to quantify the share ofinequality that might be attributable toeconomically exogenous circumstances.Most of these studies share the samebasic approach: they select a numberof observable circumstance variableslike gender, race, place of birth and familybackground; divide the population intogroups that have identical circumstances;and calculate the share of total inequalitywhich is accounted for by inequalitybetween the groups. This share is ameasure of inequality of opportunity.

Two challenges come immediately tomind: (i) researchers do not observe allincome determinants—some inequalityis always due to unobserved efforts andcircumstances, and there is no way todistinguish between those; (ii) theseunobserved variables may be correlatedwith their observed counterparts, makingit difficult to identify the true individualeffects of each observed circumstance.A study of male earnings inequality inurban Brazil uses a simulation approachto help address the second problem;it measures the opportunity share ofinequality corresponding to the effectof five observed circumstances, namelyrace, place of birth, father’s occupation,mother’s education and father’seducation. Between 10 and 37 percentof observed earnings inequality could beattributed to differences in opportunities,driven solely by these five characteristics.

The same study recognizes two channelsthough which circumstances can affect

outcomes. There can be a direct impact,as when two equally productive workersreceive different pay, because of race orgender differences. But there can also bean indirect impact, as when a personfrom a disadvantaged background exertsless effort, often as a result of therational expectation of futurediscrimination. On average, the studyfound that 40 percent of the overalleffect of circumstances on earningsinequality operates indirectly—throughdifferences in the distribution of effortsconditional on circumstances.

Earnings and incomes are not the onlyadvantages a person might enjoy.Cognitive achievement arising fromeducation is another example of animportant outcome that is affected bothby individual effort and by initialcircumstances. An ongoing study ofstudent achievement in standardizedmath tests for five Latin Americancountries finds that a sizable proportionof variation in educational achievementcan be associated with unequalopportunities. For the four countries withcomplete data (Argentina, Brazil, Chileand Mexico), 43-56 percent of totalvariation is due to inequality betweengroups defined by the gender of thechild; the location of the school; mother’seducation and father’s occupation. Whenindividual attributes are consideredseparately, the family backgroundvariables (father’s occupation and

mother’s education) play the larger role.Consistent with what we know aboutgender and schooling in Latin America,the gender of the student is of almostno relevance. Spatial differences are stillimportant, particularly in Mexico.

A cross-country comparison isalso informative. There is no directcorrespondence between the overalllevels of inequality in test scores and theiropportunity shares. Mexico, for instance,has approximately half of Brazil’s inequalityin test scores, but a greater opportunityshare. There are also differences in thesalience of specific circumstance variables:whereas mother’s education is the mostimportant in Mexico, father’s occupationdominates in Argentina.

While the findings for education arepreliminary, taken together with theearlier results on earnings, they dosuggest that credible decompositions ofinequality into “opportunity” and “effort”components are increasingly feasible.Just as distinguishing between good andbad cholesterol is important in choosingthe best policy for your heart, thisdistinction may become relevant forchoosing what inequalities to combat,and how best to do so.

F. Bourguignon, F. Ferreira and M. Menéndez(2003): Inequality of Outcomes andInequality of Opportunities in Brazil, WorldBank Policy Research Working Paper 3174.http://econ.worldbank.org/docsearch

Note: Each bar corresponds to the between-group share of total inequality for each partition, using a measureof test-scores in the PISA 2000 mathematics examination. NB: Peru lacks school location data.

22 International Poverty Centre

Both China or India havebeen influenced by andhave generated both “good”and “bad” inequalities.

Inequality is rising, butIndia has not yet hadthe same trend risein inequality as China.

Poverty in both countriesis now less responsiveto economic growthand is becomingmore responsiveto rising inequality.

If the problem is notaddressed, the highgrowth rates may not bemaintained, affecting tradeand growth elsewhere.

Addressing it may involvesome short-term costs togrowth, although redressingthe bad inequalities wouldactually be good for growth.

Aggregate economic growth israrely balanced across regions or sectorsof a developing economy, and neitherChina nor India is an exception. The post-reform pattern of growth has not beenparticularly pro-poor in either country.It has been uneven geographically,sectorally and across households. Becausethe more rapid growth of both countrieshas been so uneven in these dimensions,it has sometimes brought disappointingoutcomes in terms of progress againstincome poverty and other dimensionsof well-being.

In China, growth in the primary sector(primarily agriculture) did more to reducepoverty and inequality than growth ineither the secondary or tertiary sectors.In India, with higher initial inequality inaccess to land than China, agriculturalgrowth was less important than tertiarysector growth. In both countries, there hasbeen a marked geographic unevennessin the growth process, with numerouslagging regions, including some of thosethat started off among the poorest.

Income inequality is rising, althoughIndia has not yet experienced the sametrend increase in inequality that Chinahas seen. The Gini index of incomeinequality for China rose from 28percent in 1981 to 41 percent in 2003,though not continuously, and more insome periods and provinces. In the caseof India, one finds that the Gini indexrose in the 1990s, although the increasewas less pronounced than in China.However, it is too early to say if Indiais undergoing a trend increase ininequality similar to what China hasexperienced. As can be seen from thefigure, rising inequality in India is seento be a recent phenomenon. Indeed,there is no statistically significant trendincrease in consumption inequality inIndia up to the early 1990s.

Perceptions “on the ground” thatinequality is rising markedly in Indiado not appear to sit easily with theimpression given by the figure. Popularopinion can be mistaken, but nor are thedata perfect. The survey-based numbersmay well understate the relative gains tothe rich, and that is consistent with theevidence from tax returns. The visiblechanges in consumption patterns andlifestyles that the rich have achieved maywell not be reflected properly in thesurvey-based inequality measures. Also,and possibly more importantly, theperception of sharply rising inequalityin India may well reflect rising absoluteinequality, as reflected in the absolutegaps between the rich and the poor, asdistinct from the proportionate gaps.There is evidence that many people viewinequality in absolute terms rather thanrelative terms

Poverty in both countries is notbecoming any more responsive toaggregate economic growth and isbecoming more responsive to risinginequality. India’s poor did not start thereform period with the same advantagesas China’s poor, in terms of access to landand education.

Persistent inequalities in human resourcedevelopment and access to essentialinfrastructure within both countries, butprobably more so in India, are clearlyimpeding the prospects for poor peopleto share in the aggregate economic gainsspurred by reforms. The geographicdimensions of their inequalities and theassociated disparities in fiscal resourcesand governmental capabilities loom largeas policy concerns for both countries.

In the future, it will be harder for eithercountry to maintain its past rate ofprogress against poverty withoutaddressing the problem of high and

by Shubham Chaudhuri andMartin Ravallion, the World Bank Good and Bad

Inequalities in Chinaand India

Poverty In Focus June 2007 23

rising inequality. However, it is notparticularly useful to talk about“inequality” as a homogeneous entity inthis context. Policy needs to focus on thespecific dimensions of inequality thatcreate or preserve unequal opportunitiesfor participating in the gains from futureeconomic growth.

We make a crucial distinction betweengood and bad inequalities—drivers anddimensions of uneven growth that aregood or bad in terms of what they implyfor how the living standards of poorpeople evolve over time. We argue thatthe post-reform development paths ofboth India and China have beeninfluenced by and have generated bothtypes of inequalities.

Good inequalities are those that reflectand reinforce market-based incentivesthat are needed to foster innovation,entrepreneurship and growth. Scatteredevidence suggests that the rise ininequality with the introduction of marketreforms in both India and China is at leastin part a reflection of newly-unleashedmarket-based incentives at work, incontrast with the earlier period ofartificially low levels of inequality broughtabout by regulatory distortions andinterventions that suppressed incentivesfor individual effort and innovation.

There are two key dimensions of badinequalities. The first relates to locationin the presence of externalities,impediments to mobility and heavydependence of local states on localresources. These features can generategeographic poverty traps whereby livingin a well-endowed area entails that apoor household can eventually escapepoverty, while an otherwise identicalhousehold living in a poor area seesstagnation or decline. This is onepossible reason why initially poorerprovinces have often seen lowersubsequent growth.

The second dimension of undoubtedimportance relates to inequalities inhuman resource development—oftenlinked to credit market failures on thedemand side but also reflectinggovernmental failures in service delivery.The rising returns to schooling andincreasing dispersion of wages represent

good inequalities because they reflectfreer labor markets with increasedincentives for work and skill-acquisition.But naturally, those with relatively littleschooling and few assets, or little accessto credit are less able to respond to theseincentives and are less well positioned totake advantage of the new opportunitiesunleashed by market-oriented reforms.And thus, inequalities in human capitalare ‘bad inequalities’ in that they haveretarded poverty reduction throughgrowth in both countries.

While both countries need to beconcerned about the bad inequalities, wesuspect that it is China where the near-term risk that rising inequality willjeopardize growth and poverty reductionis greater. Arguably, the Chinese authoritieshave been able to compensate for risinginequality by achieving high growth rates;by this view, it is the rising inequality thatfuels growth in China, through thepolitical economy of maintaining socialstability. However, the emerging badinequalities in China will make it harder topromote the growth that will be neededto compensate for those inequalities.Maintaining sufficient growth will requireeven greater efficacy of the policy leversused to promote growth.

Whether or not the problem of risinginequality is successfully addressed, thereare likely to be implications for the rest ofthe world. If the problem is not addressed,then there is a risk that the high growthrates will not be maintained, with spillover

effects for trade and growth elsewhere.If it is addressed, and depending onexactly how this is done, there may besome short-term costs to growth,although redressing the bad inequalitieswould actually be good for growth. Theremay also be consequences for the patternof trade, such as through a change in thesectoral composition of growth; forexample, in both countries there appearsto be potential for cash crop expansion,which would attenuate one importantsource of concern about rising inequality,and it can be expected that a large shareof this expansion in domestic cash-cropoutput would be exported.

The new initiatives underway in bothcountries are probably steps in theright direction, although continuousevaluative research will be needed onthe efficacy of these approaches relativeto alternatives. There are important butpoorly resolved issues concerning theappropriate balance between typesof interventions. But an even harderchallenge remains, namely to improvegovernance—capacity, accountabilityand responsiveness—notably (but notonly) at the local level. If this challenge isleft unmet, the ultimate efficacy of any ofthese initiatives will be in doubt.

Shubham Chaudhuri and Martin Ravallion:“Partially Awakened Giants: Uneven Growthin China and India” in Dancing with Giants:China, India, and the Global Economy, editedby L. Alan Winters and Shahid Yusuf, WorldBank, 2007. See also World Bank PolicyResearch Working Paper 4069, Nov. 2006.http://econ.worldbank.org/docsearch

24 International Poverty Centre

China’s recent economicgrowth performance isoutstanding, and averageincomes have risen fast.

However, the speedof increased incomeinequality, and its currentlevel is also striking.

Poverty declinedbetween1995 and 2002,reversing past trends,while inequality in thewealth distribution rose.

Public policy needsto change the balancebeween the processesthat reduce inequalityand those that increase it.

Inequality and PublicPolicy in China:Issues and Trends

by Björn Gustafsson1, Li Shi 2

and Terry Sicular3

1. University of Göteborg;2. Chinese Academy of Social Sciences;

3. University of Western Ontario.

In recent decades China’s economyhas grown more rapidly and in a moresustained fashion than that of any othercountry. As a result, most people in Chinahave higher incomes, consume more andbetter goods, and live in better housingthan ever before. Between 1988 and 2002,household income per capita on averagenearly tripled in real terms. Also, lifeexpectancy has increased, and educationlevels continue to rise.

With growth, however, has comea significant widening of incomedifferences among households andindividuals. Income inequality has risenfrom a relatively low level in the early1980s to a level that is now consideredhigh by international standards.Although increased inequality often goeshand in hand with economic growth anddevelopment, in China the speed withwhich inequality has increased, and thelevel to which it has risen, is striking.

Inequality is not necessarily a problem.Most would agree that past policies inChina had excessively compressedpersonal income differentials, so someincrease was expected. Inequality reflectingdifferences in effort, experience, skills,investments, and risk can be justifiablefrom economic and social standpoints.

Concerns arise, however, when incomesdiffer excessively in ways that reduceefficiency or violate accepted views offairness and justice. In such situationsinequality can erode social cohesion,generate social and political instability,and hinder economic growth. Concernsalso arise if segments of the populationare left behind, with insufficient resourcesto meet basic needs or entitlements.

For most of the reform period, China’seconomic policies have placed highestpriority on growth and tolerated

widening inequality. This approach wasencapsulated in Deng Xiaoping’s well-known statement that some people shouldbe allowed to “get rich first.” China’s leadershave nevertheless been mindful of equity;the government has adopted a range ofprogrammes for rural and regional povertyalleviation and for social insurance andwelfare, mainly in urban areas, aimed ataddressing particular distributional issues.These measures reflect concerns about theimpact of income gaps on stability, as wellas efforts by the Communist leadership tomaintain political legitimacy.

Interest in China’s income distributionextends beyond its borders. China is themost populous country in the world, andso changes in the level and distributionof its income hold implications for globalinequality and poverty. When trying tounderstand changes in inequality andpoverty on a global scale, then,knowledge of trends in China is essential.

China is also of general interest becauseits experience may provide evidence andinsights relevant for understanding therelationship between inequality anddevelopment. The conventional wisdomsuggests that rising inequality is acommon phenomenon in early stagesof development, but this relationshiphas been critically examined by economicresearchers and is subject to ongoingdebate. The experiences of formerlyplanned economies in Eastern Europeand Central Asia suggest that transitionis also a major force at play. Some wouldargue that trade liberalisation is also acontributing factor.

The validity of concerns over inequalityin China, and the sorts of policiesappropriate to addressing thoseconcerns, depends on the leveland characteristics of that inequality.Empirical knowledge about the level and

Poverty In Focus June 2007 25

characteristics of inequality in China hasimproved greatly over time, but significantgaps remain. A lot is known aboutinequality among provinces and regions,but at the household and individuallevels most studies examine inequalitywithin urban areas or within rural areas.Studies on inequality nationwide arerelatively scarce. Research on key facetsof inequality, such as the distribution ofwealth or the level of poverty, is spotty.

Gaps also exist in knowledge aboutparticular groups, such as migrants,women, minorities, and the elderly. Moststudies rely on data more than a decadeold. In a rapidly changing environmentlike China’s, effective policy requires moreup-to-date information. Data remain anunderlying constraint, limiting the scopeand timeliness of available studies.

A major finding is that the level ofincome inequality in China remainedrelatively stable between 1995 and 2002.The forces generating more inequalityhave been offset by emerging equalisingprocesses including: rising off-farm wageemployment in rural areas, convergencein household per capita income amongprovinces in eastern China, and widelyshared macroeconomic growth. Yet,disequalising forces continue. The largeurban-rural income gap increased andbecame an even more dominant sourceof overall inequality. Education hasemerged as an important factorunderlying inequality.

Poverty, reversing past trends, declinedbetween 1995 and 2002. This encouragingtrend in poverty was accompanied byrising inequality in the distribution ofwealth. The level of wealth inequality,however, was not overly high byinternational standards.

Including rural-to-urban migrants in thecalculations changes measured levels ofinequality and poverty in China, but theimpact is relatively modest; inequalitydeclines and poverty increases. Althoughtheir income falls between that of theurban and rural groups, migrants havethe highest rate of poverty. Migrantpoverty is higher than that of registeredurban households. It is also higher thanrural poverty, mainly because the urbanpoverty line is higher than the rural one.

China’s path of economic reform andtransition since the late 1970s has beenmarked by a wide range of significantreform measures, most aimed atpromoting growth and development.A major theme of the reforms since thelate 1980s has been to expand the roleof markets. In the early reform periodmarkets were encouraged to emergealongside planning in what was knownas the “dual-track” system, and the reformpolicies focused on markets for goods.Planning continued to replace marketsfor production factors—land, labour, andcapital. The 1990s saw the continueddismantling of the system ofadministrative planning and pricing andthe improved functioning of markets forgoods and services. By the mid-1990splanned allocation had been eliminatedfor almost all consumer goods. Inaddition, at this time the governmentbegan to take steps to develop marketsfor production factors.

The urban labour market reforms havegiven enterprises more freedom to layoff and hire workers. This, together withthe loosening of restrictions on rural-to-urban migration, has allowed labourmarkets to function more fully. Barriers tosuch migration were put in place duringthe planning period and embodied inthe household registration system.

During the 1990s factor marketdevelopment was also furthered byownership reforms. Small and medium-sized state-owned enterprises wereprivatised, converted into shareholdingoperations, or leased; for larger state-owned enterprises the governmentinitiated a programme of governancereform and corporatisation. Suchmeasures have made enterprises in bothrural and urban areas more marketoriented, with implications for theiremployment and other behavior. Theexpansion of shareholding and privateenterprise has provided an impetus forthe expansion of financial and capitalmarkets. The urban housing reforms areanother aspect of factor market reformsin the 1990s. In a remarkably short timethey have created, from scratch, marketsfor private real estate in urban areas.

In a market environment, earningsdepend on households’ endowments of

factors such as labour, entrepreneurialability, education or skills, experience,and physical or financial assets, aswell as on the prices or returns ofthese endowments and other factors.The expansion of markets during the1990s thus led to greater differentiationof incomes on the basis of householdendowments and characteristics.Also, with the expansion of marketsfor financial instruments and real estate,the importance of income from assetsincreased, although not necessarilyequally among households.

The process of factor price equalisation,however, is as yet incomplete. Despitecontinued market reform in the 1990sand early 2000s, and while markets formost goods and many services are nowwell established, factor markets remainfragmented and incomplete. Thus, forexample, worker mobility among urbanresidents in China is higher than before,but still much lower than in marketeconomies. Unemployment and otherforms of non-employment haveemerged, reflecting frictions and lack ofmarket clearing in urban labour markets.This has affected the urban incomedistribution; poverty is no longer a solelyrural phenomenon.

Our findings warn against making simplepredictions from one period to the next.We have found processes workingtoward more inequality as well as thoseworking toward less. Public policy caninfluence both types of processes andalter the relative balance between them.Noteworthy here are policies affectinglabour mobility, especially between ruraland urban areas, education, publicfinance, and the provision of socialinsurance and social welfare programmes.In view of the rising importance of assetownership, steps to strengthen propertyrights and factor markets will also be ofcrucial importance.

Finally, policies promoting broadmacroeconomic growth remain importantso as to promote further progress towardthe stage of development where thebenefits of growth spread from leadingsectors to the broader population.

B. Gustafsson, L. Shi, T. Sicular (Eds.):Inequality and Public Policy in China.Cambridge University Press, 2007.

26 International Poverty Centre26 International Poverty Centre

I N T I M E

Studying individual timeuse can tell us much aboutthe distribution of well-being among populationgroups, which is useful forpublic policy and povertyreduction strategies.

Women work more thanmen, but their labour ismostly unpaid; thus they donot achieve more economicpower as a result.

However, most inequalityin time use is located within,not between gender groups.

For poor people, escapingincome deprivationoften means increasingtime deprivation.

The way people use their time has direct implications for their well-being. Time isparticularly important because it is a scarce resource and virtually all human activitiesare time consuming; allocating time to one activity requires restricting time for otheractivities. The study of the distribution of individual time allocations can tell us muchabout the distribution of well-being in a population.

Take, for instance, overwork, which has a clear impact on a person’s life. Excessive workprevents people from living a reasonable family and social life and can even affectseriously the health of an individual when it does not allow for adequate rest. Work,here, does not mean only paid work but also unpaid domestic work. It is easy toimagine that excessive domestic work can negatively affect the schooling of childrenor the labour market participation of women.

One can always argue that there is no such thing as reduction of well-being due toexcessive work or any other activities since time is allocated according to choices andrational individuals will always choose what is best for them. Nevertheless, for themoral judgments involved in the issue, what matters is not which choices are madebut the restrictions affecting these choices. In order to understand these restrictionsthe distributional analysis is crucial.

The figure illustrates that time allocation is not only a matter of individual choice. Itshows a generalized Lorenz curv—that is, a Lorenz curve in which the values weremultiplied by the mean of the distribution—of time spent working, be it paid or unpaidwork, by urban adult women and men in Bolivia, using data for 2001. As the populationsare ordered according to the amount of total time spent at work, these curves shouldbe interpreted as the cumulative workload along the population of each gender.

The first thing this figure shows is that the curves for men and women are very distinct.This means that there is a clear differentiation by gender in the patterns of allocationof time to work. It becomes hard to sustain that free choice prevails here since in a freeallocation—free from social roles and other constraints—there would be no reason forsuch a clear group differentiation.

Second and perhaps more relevant is that the curve of women always lies above thatof men. This dominance of the curves means that the workload of women is higherthan that of men, regardless of the point of the distribution we assess it. Such asustained difference gives no room to doubt: there is a clear gender division of labourin the Bolivian society leading women to work more than men. Some aspects of thisdivision of labour have implications to other spheres of life. As shown in the bar chart(page 27), in many societies, particularly in the developing world, men systematicallyparticipate more in the labour market but less in unpaid domestic activities. This leadsus to the conclusion that women work more than men but the fruits of this work donot necessarily translate in more economic power.

Using gender as an example of the relevance of studying inequalities in time allocationwas not a random choice. Over the last twenty years feminist economists have correctlyinsisted on the need of incorporating time use information in the system of nationalaccounts. Several non-market activities such as those related to the reproduction of thelabour force—child care, the preparation of food, among others—are valuable for the

Time Use, Inequalityand Well-beingby Joana Costa, Marcelo Medeiros and Rafael Guerreiro Osório,International Poverty Centre

Source: IPC Working Paper No. 34, page 6.

Poverty In Focus June 2007 27Poverty In Focus June 2007 27

Source: UNDP (2006). Human DevelopmentReport 2006, page 379. New York: UNDP.(Selected countries).

society but are not accounted as such. Knowing the time spent on them is a first step toevaluate this hidden contribution of women to economy and society. National accountsare based on market prices and there are obvious difficulties to give prices to non-marketlabour but despite the obstacles the fact is that the incorporation of time use in theaccounts gives visibility to this important but often neglected part of the economy.

National accounts, however, focus on aggregates. Equally important is to fullyunderstand the distribution among individuals and social groups underlying theseaggregates. The concern with time brought about by the desire of giving womenrecognition for their non-market contributions can incorporate the concern with thereasons why some people work so much and others so little. As in the case of incomedistribution analysis, there are several reasons to look at time allocation inequalitiesand studying them can contribute to explaining inequalities in other spheres of life.

In fact, gender is not the single determinant of inequalities in time allocation. Asdepicted by the arc of the generalized Lorenz curves in the figure (page 26), much ofthe total inequality is located not between, but within gender groups. Indeed, in theBolivian example, a decomposition of the total workload inequality shows that only asmall fraction of it can be related to the differences between men and women. Most ofit is inequality among males and among females because both are very heterogeneousgroups in terms of factors such as family composition, class position, and so on. As anexample, richer women can buy domestic services on the market, and therefore are lessprone than poorer women to suffer from overwork due to a double-shift of marketand non-market activities.

Besides gender, time use analysis can also contribute to a deeper understanding ofpoverty and other issues. For pragmatic reasons, a typical income poverty study equateshigher income with more well-being. This seems to be the best feasible approach withlimited data, but it should be noted that the gains in well-being from additional incomemay vary depending on how this income is earned. Earning income has costs and time ispart of these costs. For the individuals more work means more welfare coming from thecommodities their wages can buy but, on the other hand, less welfare coming from othersources, such as family activities. There is a trade-off between time deprivation andincome deprivation; in other words, between time poverty and income poverty.

A simulation exercise using Brazilian data for 2004 shows how important this trade offcan be. Using a simple decile-based threshold to define as poor those who belong to thetwo bottom deciles of the per capita household income distribution, we depart from anincome poverty incidence (headcount ratio) of 20 percent of the entire population. Thisvalue ignores the amount of time spent in paid and unpaid work—that is, total work.However, reducing the amount of hours of unpaid work to limit total work to 60 hoursper week increases the incidence to 21 percent. Going the other way around, reducingpaid work hours to limit the total workload at 60 hours per week increases the incidenceratio even more, to 25 percent. Such simulations are always imperfect and depend onassumptions but, in rough terms, this means that about one fifth of the poorest peoplein Brazil are escaping income deprivation by falling into time deprivation.

A better understanding of the dynamics of time allocation can be very useful for publicpolicies. It can be used to enhance strategies for reducing inequalities and achievingthe MDGs. For instance, publicly run or subsidised crèches and pre-schools maycontribute to releasing female labour force to the market, thus reducing poverty andincreasing economic autonomy of women; they may also reduce overwork amongboth men and women. Similar benefits may be accomplished in poor countries, wherewomen and children spend huge amounts of time on fetching water and firewood,by extending public water supply, rural electrification and fuel subsidies and/orimproved fuel-efficient stoves.

Marcelo Medeiros, Rafael Guerreiro Osório, Joana Costa: Gender Inequalities in AllocatingTime to Paid and Unpaid Work: Evidence from Bolivia. IPC Working Paper No. 34. April 2007.http://www.undp-povertycentre.org/pub/IPCWorkingPaper34.pdf

International Poverty CentreSBS – Ed. BNDES, 10º andar70076-900 Brasilia DFBrazil

povertycentre@undp-povertycentre.orgwww.undp-povertycentre.orgTelephone +55 61 2105 5000 Ju

ne

200

7