assessing targeting options for nepal’s social grants what does the evidence tell us dfid nepal

8/12/2019 Assessing Targeting Options for Nepals Social Grants What Does the Evidence Tell Us DFID Nepal

1/40


2/40


3/40

CONTENTS

PREFACE 4

EXECUTIVE SUMMARY 5

1 BACKROUND 6

2 TARGETING METHODS 7 2.1. Proxy means tes ng 7

2.2. Community-based targe ng 7

2.3. A universal approach to targe ng 8

3 HOW EFFECTIVE ARE THESE METHODOLOGIESIN REACHING THE POOR? 9 3.1. E ec veness of proxy means tes ng 9

3.2. E ec veness of community-based targe ng 14

3.3. E ec veness of universal targe ng 19

4 THE WIDER COSTS OF TARGETING 22 4.1. Social costs of targe ng 22

4.2. Moral costs of targe ng 25

4.3. Incen ve costs of targe ng 26

4.4. Poli cal costs of targe ng 27

4.5. Financial costs of targe ng 28

5 CONCLUSION 29

BIBLIOGRAPHY 33


4/40

For several years, the Government ofNepal and development partners havebeen working together to develop

a stronger evidence base to informthinking about social protec on policyand support the implementa onand scale up of social protec oninterven ons. In 2009, a Na onalSteering Commi ee on Social Protec on(NSC-SP) was formed with three mainobjec ves:

Review exis ng social protec oninterven ons and their deliverysystemsIden fy and appraise various op onsfor social protec on along withnancing and delivery mechanismsfor NepalDevelop a comprehensive,consensual and consolidatedframework of cost e ec ve socialprotec on interven ons throughextensive consulta on withpolicymakers, stakeholders withinand outside government, as well aswith development partners.

None of this can be done e ec velyunless it is based on solid evidence.

To support the Government of Nepal,the na onal development partnersSocial Protec on Task Team (SPTT)aims to build and promote a Nepal

specic evidence-base on key issuesand interven ons, including targe ngmethodologies, delivery systems andmonitoring and evalua on in order

to inform key policy and programma cdecisions.

This paper arises from DFIDs desire tosupport the achievement of this objec veand is part of DFIDs broader global

commitment to understand what works

in social protec on. DFIDs forthcomingglobal Evidence Paper on Social Protec onemphasizes the need to invest in buildingstronger evidence through robustmonitoring and evalua on.

This discussion paper is therst in a seriesof papers that aim toll the evidencegaps in Nepal by situa ng the countrys

experiences within a wider interna onalcontext. We hope that it will help policymakers take more informed choices onhow to target social protec on benets.A broader aim, however, is to contributeto the global debate on targe ng socialprotec on.While the views expressed in this paperare those of the authors, and do notnecessarily represent the views of DFID,we would like to acknowledge the helpful

comments received from DFIDs seniorsocial protec on adviser, Tim Conway.

Dr Stephen Kidd, director of policy,Development PathwaysDr Rebecca Calder, social developmentadviser, DFID NepalMs. Emily Wylde, senior economist,Development Pathways

PREFACE


5/40

This discussion paper examines andassesses the e ec veness of varioustarge ng op ons for Nepals social grants,

focusing on its old age allowances andchild grants. In par cular, it comparespoverty and universal targe ng andexamines the accuracy and costs ofdi erent op ons.

Interna onal evidence indicates thatpoverty targe ng in developing countrieshas high margins of error: it is commonfor over half of eligible beneciariesto be excluded from programmes. Anexamina on of the proxy means testmethodology within the Nepal context

reveals that it has high in-built theore calerrors, with 56 percent of intendedbeneciaries excluded when programmesaim to cover the poorest 15 percentof the popula on. Community-basedtarge ng is prone to manipula on andcapture by more powerful communitymembers and elites. In contrast, manyothers, including the World Bank (1994)argue that universal targe ng is successfulin reaching all the poor within specicdemographic categories.

However, of greater importance are the(o en hidden) costs associated withdi erent targe ng op ons. Poverty

EXECUTIVE SUMMARY

targe ng can promote increasedsocial conict, par cularly within ruralcommuni es, while universal programmes

are more likely to build social cohesion.Programmes publicly iden fyingpeople as poor can also s gma sethem, undermining their self-respect.By contrast universal programmes areregarded as en tlements for all ci zens.

There is evidence that poverty targe ngcan create disincen ves to work, whileuniversal targe ng does not.

The paper argues that the poli caleconomy of targe ng is of fundamentalsignicance. Poverty targeted

programmes tend to have less supportfrom the middle-class and tax-payers who are excluded from them and

so budgets for such programmes arerela vely small. In contrast, universalprogrammes have broader-based supportand larger budgets. As a result, universalprogrammes are likely to provide the poorwith be er and higher benets.

In conclusion and based on available

evidence the paper suggests that

Nepals current commitment to theexpansion of universal targe ng is likely tobest serve the interests of both the poorand the na on.

Interna onalevidence

indicatesthat povertytarge ng indeveloping

countries hashigh margins

of error: itis common

for over halfof eligible

beneciariesto be

excluded from

programmes


6/40

Compared to countries of similarwealth, Nepal is well advanced in termsof its social protec on commitments.Signicant na onal programmes are inplace, including:

The Senior Ci zens Allowance,providing a transfer of Rs.500 permonth to the over-70s, as well asto Dalits and Karnali residents agedover-60The Single Womens Allowance which

gives Rs.500 per month to singlewomen over-60

The Disability Grant which givesRs.1,000 per month to those with acomplete disability and Rs.300 to

those with a par al disabilityThe recent Child Protec on Grant,o ering Rs.200 per month to motherswith children underve years(with a maximum limit of two childrenper mother)

The various programmes use di erenttarge ng methods.1The Senior Ci zens

Allowance and Single Womens Allowanceuse a form of universal targe ng, wherebyeveryone above the age of eligibilityis able to access the grants. In theory,everyone with a complete disability canbenet from the Disability Grant.2TheChild Protec on Grant is targeted alsoin theory only at poor Dalits, although

anecdotal evidence indicates that, inmany places, all Dalits are being giventhe grant as poverty targe ng is provingtoo di cult. In the Karnali region,the Child Protec on Grant is targeteduniversally to all women with childrenunderve years.3

A key policy debate is currently underwayin Nepal regarding the best means ofreaching the beneciaries of the varioussocial grants. In simple terms, on oneside of the debate are those who arguethat the government should focus its

resources only on the poor. On the otherside are those who argue for con nuingwith a policy of universal targe ng asthe best means of reaching the poor. Itis important that this debate is based onsolid evidence rather than ideology.

This discussion paper will examinevarious op ons for targe ng the currentsocial grants, focusing in par cular onthe Senior Ci zens Allowance, SingleWomens Allowance and Child Protec on

Grant. It will try to determine the bestmeans of reaching poor older peopleand poor children underve within thecontext of Nepal. It will examine bothpoverty targe ng and universal targe ngand assess their e ec veness and thecosts associated with them, using bothinterna onal and Nepal specic evidence.

1BACKGROUND

A key policydebate iscurrentlyunderway inNepal regardingthe best meansof reachingbeneciaries ofsocial grants.On one sideare those whoargue that thegovernment

should focuson the poor.On the otherare thosewho arguefor universaltarge ng. Thedebate must bebased on solidevidence


7/40

This discussion paper examines threecommon targe ng methods. Two ofthe methods - proxy means tes ngand community-based targe ng - arecommonly used forms of povertytarge ng that are also prac sed in Nepal.

The third is a universal approach totarge ng.

2.1.Proxy means tes ng

The proxy means test methodology was

developed in La n America and duringthe past two decades has been usedincreasingly in developing countries acrossthe world. It uses na onal householdsurveys as its basis. Its premise is that,since household income is di cult andexpensive to measure accurately duringtarge ng, assets and other variables or proxies can be used to es matehousehold welfare instead. Analysts runregressions to calculate those proxiesthat are most correlated with welfare.

These can cover a range of areas such asdemographic characteris cs (for example,age of household members and size ofhousehold), characteris cs of the house(such as type of roof or oor), durablegoods (such as refrigerators, televisions orcars) and produc ve assets (such as landor animals). While any one proxy may be

2TARGETING METHODS

rela vely weakly correlated with welfare,correla ons improve if mul ple proxiesare used.

The proxy means test aims tond theset of proxies that best explain welfare,usually between 10 and 30 proxies intotal. Each variable is given a weightbased on its es mated impact onhousehold expenditure and a score can

be calculated for each household usingthese weights. Those households thatscore below a certain cut-o point areeligible for the programme.

Households are scored against the proxiesby means of a survey ques onnaire.Enumerators can either a empt tovisit all households in an area or canuse a public informa on campaignto encourage eligible households toregister at government o ces or centres.Enumerators then visit those householdsthat apply for the grant.

2.2.Community-basedtarge ng

Community-based targe ng has mostcommonly been used within small-scaleor pilot programmes, but there are someexamples of its use at a na onal level,


8/40

such as in Bangladesh and Indonesia.The basic argument underpinning thismethod is that communi es themselvesare best able to know and decide who ispoor and deserving of support. However,community-based targe ng can vary interms of the extent to which communi esare given discre on regarding selec on.

A key ques on with community-basedtarge ng is who makes the decisions.In some cases, they can be madeby community leaders or respectedcommunity members; in other cases,a empts can be made to involve thewhole community in decision-making.Communi es usually draw up criteria fordetermining the poor and vulnerable

some mes from within a pre-determinedlist of op ons presented by theprogramme and then assess households

in the community against those criteria.4

2.3.A universal approachto targe ng

Universal targe ng is the simplest oftarge ng methodologies. Within a speciccategory of people such as the elderly

or young children everyone is eligible

for the programme. They only need to be

able to demonstrate that they belong tothat category. The majority of universalprogrammes in developing countriesare pension programmes similar to

the Senior Ci zens Allowance andpoten al beneciaries must provethat they are above a certain age to beaccepted onto the programme. Around15 such programmes can be found acrossdeveloping countries. Universal targe ngfor child grants is much rarer. While it hasbeen used for child grants in Mongoliaand Bolivia, the universal Child Protec onGrant in Karnali can be considered aninnova ve programme.


9/40

In this sec on the three methodologiesare assessed in terms of their poten ale ec veness in targe ng poor olderpeople and children. Each targe ng

methodology is examined using threecriteria:

E ec veness in reaching the poorE ec veness in excluding thebe er-oLevel of leakage to non-intendedbeneciaries

3.1.E ec veness ofproxy means tes ng

Although there are many examples ofproxy means tes ng across the world,there have been few good qualityassessments of their e ec veness inreaching the poor. Two programmesthat are well-regarded by advocates ofproxy means tes ng and have had theiraccuracy assessed are the Oportunidadesprogramme in Mexico and the PATHprogramme in Jamaica, both of whichreach around 20 percent of households.

However, both programmes havesignicant errors. In Mexico, around 70percent of the poor are excluded from

3HOW EFFECTIVE ARE THESEMETHODOLOGIES IN REACHING THE POOR?

the programme the exclusion error

while 36 percent of the non-poor are

beneciaries the inclusion error.5InJamaica, the exclusion and inclusion

errors are around 50 percent, meaningthat around half of intended beneciariesare unable to access the programme.6

There has also been an example of aproxy means test resul ng in a regressiveoutcome in Armenia with more of the

benets going to the non-poor.7

In-built into the proxy means testmethodology is an error derived from theweakness of the correla on of the proxieswith household income (or expenditure).

The regressions used in proxy means testsrarely explain more than half of householdincome.8Consequently, even the ini alweigh ngs of the mul ple proxies, whentested against the household surveysfrom which they were derived, havesignicant errors. And the smaller theprogramme, the larger the errors. Figure 1indicates the theore cal errors associatedwith the Nepal proxy means test.9At 10percent coverage, around 59 percent ofintended beneciaries would be wronglyexcluded (and the same percentage would

be wrongly included), 54 percent at 15percent coverage and 38 percent at 30percent coverage.10


10/40

Figure 1: Percentage of intended beneciaries wrongly included/excluded by the Nepalproxy means test at di erent levels of coverage.

The results found with the Nepal proxymeans test are in line with resultsfound elsewhere in the world. Forexample, when tes ng against 15

percent coverage, Kidd and Wylde(forthcoming), found errors of 53 percentin Bangladesh, 52 percent in Sri Lankaand 61 percent in Indonesia.11At lowerlevels of coverage, errors were higher:71 percent at 10 percent coverage inIndonesia, for example. Real errors arelikely to be higher since these es matesdo not include errors derived from theinaccuracies within the household survey

nor, as will be demonstrated, fromimplementa on.

The errors in the proxy means test

regressions can also be illustrated bysca er graphs. Figure 2 plots householdexpenditure derived from the householdsurvey for Nepal against the predictedexpenditure derived from the proxymeans test. These are assessed against20 percent coverage. It can be seen thatwhile many households are correctlypredicted, a large propor on are eitherwrongly excluded or included.


11/40

Figure 2: Sca er graph for Nepal indica ng households included or excluded by theproxy means test in a programme with 20 percent coverage

All households in the bo om le hand quadrant are correctly included and those in the top right handquadrant are correctly excluded. Those in the bo om right hand quadrant are incorrectly excluded and those inthe top le hand quadrant are incorrectly included.

In addi on to the errors caused by theregression methodology, there aremany other reasons explaining the largeerrors associated with proxy meanstes ng. A recent study by Kidd andWylde (forthcoming) discusses themin detail and only a summary will beprovided here. A key concern, however,is the use of the household survey asthe basis of the targe ng methodology.Household surveys include inaccuracies

and can never be regarded as morethan an approxima on of reality. Theseinaccuracies arise partly from the

inevitable problems in extrapola ngresults from a small sample surveyto the whole popula on, but equallyimportantly due to issues with the designand, cri cally, assump ons made inthe analysis and interpreta on of thedata. While surveys are useful toolsfor measuring poverty at o en fairlyaggregated levels, it is ques onablewhether they are reliable enough to beused as the founda on of a sophis cated

targe ng methodology on which thewelfare of individual households will hangin the balance. 12

In urbanMexico overhalf of those

who wereeligible for

the Progresaprogramme

neverregistered. In

fact, there wasa correla on

betweenowning a car

and beingaccepted

on to theprogramme


12/40

Indeed, the assump ons used in theanalysis of the household survey datacan also have a signicant impact onwhich individual households are includedin the programme. Instead of being fairand objec ve, as o en argued, PMTsul mately lead to quite an arbitraryselec on of households. So, not only is itan inaccurate methodology, it also scoreslow on any measure of fairness.

As well as the errors that are built intothe proxy means test methodology,interna onal evidence indicates thatfurther errors are incorporated duringimplementa on. While there tends tobe a belief that the proxy means test iseasy to implement in prac ce, in reality itsu ers from many of the same challengesas other targe ng methodologies.13

One source of error leading to theexclusion of eligible households isthat many may never be surveyed and

assessed. There is good evidence fromPakistan, Mexico and Nicaragua thateven when all households are meant tobe visited, some can be missed out.14When households are expected to visitgovernment o ces errors are usuallymuch greater, o en because householdsdo not hear about the programme or areunable to travel to the registra on centre.In urban Mexico over half of those whowere eligible for the Progresa programmenever registered.15In fact, there was a

correla on between owning a car andbeing accepted on to the programme,highligh ng the barrier to inclusion inthe programme caused by the cost oftravelling to register.16

Challenges also arise during the interviewprocess when households have to answerthe ques ons in the proxy means testsurvey. Large numbers of enumeratorsneed to be trained, but training is not

always carried out adequately.17A recentstudy in Pakistan indicated that in mosthouseholds insu cient me was set asidefor the interview, making it impossibleto gain accurate answers.18Indeed, theanswers given by beneciaries varysignicantly depending on whether a manor woman is interviewed. Furthermore,while it is commonly claimed thatproxies are easily veriable, this is o ennot the case. It is frequently not easy

par cularly in a short interview toascertain accurately age, educa onala ainment, landholdings, animals, etc.19Households are also able to manipulateproxies by lying or hiding assets. Indeed,the fear of households being able tomanipulate proxies led Castaedaand Lindert (2005) in a World Bank

publica on to recommend not makingproxies public. However, this wouldreduce the transparency of the proxymeans test, which would be inconsistentwith human rights principles.20

As a means of rec fying errors, proxymeans tests o en incorporate a processwhereby the proposed list of beneciariesis presented in community mee ngs sothat community members can eitheragree or oppose the decisions. Thereis, however, good evidence that suchmee ngs are not e ec ve. In MexicosProgresa programme only 0.1 percent ofbeneciary selec ons were disputed.21Qualita ve studies in Mexico, Peru

and Nicaragua all indicate that suchcommunity mee ngs do not func one ec vely.22Indeed, as Grosh et al.(2008:118) point out, it is highly unlikelythat such community mee ngs couldfunc on well since ques oning theinclusion of other community memberspublicly would be divisive. It wouldbe even more di cult to make suchcommunity mee ngs operate well inurban areas.

Householdsurveys includeinaccuraciesand can neverbe regarded asmore than anapproxima onof reality. It isques onablewhether they

are reliableenough to beused as thefounda on ofa sophis catedtarge ngmethodology


13/40

Further errors enter into proxy meanstes ng as a result of poverty dynamicsand delays in re-cer ca on. Proxy meanstes ng is expensive to implement and sogovernments are reluctant to re-targetfrequently. Indeed, re-cer ca on maynot be done forve years or more. Duringthis me, household circumstances willchange, with some beneciaries becomingbe er-o and some non-beneciaries

poorer. Yet, there will be no possibilityfor them to join or be removed fromthe programme. Therefore, over me,targe ng accuracy will deteriorate.

Proxy means tes ng is a complex targe ngmethodology requiring a high level oftechnical capacity. There have beenexperiences of developing simpler formsof this methodology using more intui veproxies or qualita ve consulta ons. But

The challenge of iden fying poor Dalit households in Nepal

Nepals Child Protec on Grant is supposed to use a simplied form of proxy meanstes ng when targe ng poor Dalit households. Eligible households should be iden edby fullling one of the following criteria:23

Poor Dalits without ownership or share in land and house for permanent se lement.

Persons with less than two ropani of land in the hills and one ka ha in the Terai.

Subsistence farming families who produce food that is su cient for less than threemonths of the year

But these criteria are by no means failsafe. For instance no correla on has been donewith poverty to assess their accuracy nor the number of households that qualify.Furthermore, they are unlikely to work well in urban areas where few households willcul vate their own food. And, while they may seem simple criteria, they will not be easyto measure. A par cular challenge will be assessing the extent to which householdsproduce food that is su cient for fewer than three months. It will depend on answersprovided by households and it will not be possible to verify the informa on.

Errors are likely to be signicant with this targe ng. Indeed, there is evidence from visitsto certain communi es that all Dalits are being regarded as eligible. This may turn out tobe a common occurrence across Nepal.

using fewer proxies is likely to result ina weaker correla on with poverty thanwith more sophis cated proxies derivedfrom the household survey. Evidenceappears to bear this out. The KenyanOrphans and Vulnerable Childrens

(OVC) Cash Transfer programme hasused a simplied form of proxy meanstes ng and the results as demonstratedin Figure 3 indicate that it is only a

li le be er than random selec on.24

In fact, despite a high coverage of 51percent of households with OVCs, 43percent of eligible households were notincluded in the programme. Indeed,an eligible household had only a 13percent greater chance of being includedin the programme than a non-eligiblehousehold. Nepals Child Protec onGrant also uses a simplied proxy meanstest, which is discussed in Box 1.

Box 1


14/40

Figure 3: The propor on of all beneciaries in each income quin le for the Kenya OVCCash Transfer programme ( Quin le 1 is the poorest)

So proxy means tests are inherentlyinaccurate for a range of reasons andresult in rela vely high levels of leakagewhile many eligible poor people o enthe majority are excluded. They are

also very di cult to implement andrequire good technical capacity. Indeed,the World Bank in its social protec onhandbook25 advises that: Proxy meanstests are most appropriately usedwhere a country has reasonably highadministra ve capacity. It is unlikely that

Nepal could be regarded as fullling thisrequirement.

3.2.E ec veness ofcommunity-based targe ng

Interna onal evidence on community-based targe ng indicates that it su ersfrom signicant errors. There are veryfew na onal programmes that haveused community targe ng: one exampleis the Food for Educa on programmein Bangladesh, which used SchoolManagement Commi ees to select

beneciaries. As Figure 4 indicates, thetarge ng errors have been signicant,with results li le be er than randomselec on. Indeed, a high propor on ofeligible children more than half were

excluded from the programme. In anotherlarge-scale programme using communitytarge ng in Indonesia, the World Bank(2006:192) found that, although theprogramme was targeted at the poorest30 percent of the popula on, 55 percentwere excluded from the programme and

the leakage rate was 50 percent.


15/40

Figure 4: Propor on of beneciaries and non-beneciaries of the Bangladesh Foodfor Educa on Programme in each income decile (decile 1 is poorest): par cipa ngvillages only26

Small pilot programmes that haveused community-based targe ng alsodemonstrate poor targe ng outcomes.In general, beneciary households tendto be clustered around the middle ofthe wealth spectrum. Results improvewhen signicantly more resources areinvested in administra on, but are s llproblema c.27In addi on it is highly likelythat community-based targe ng willperform less well at na onal level thanat pilot programme level because pilot

programmes usually employ unrealis callyhigh levels of administra ve support.28

Although community-based targe nghas not been used for cash transfers inNepal, it has been used to target other

programmes to poor households.Despite the small scale of the pilots and the likely signicant levelof support from outside agencies performance has not been good.

Figure 5 shows the results from astudy by Hobley and Paudyal (2008)in which they examined community-based targe ng in two communi esin Nepal. They found signicantexclusion of the poor and leakage tothe non-poor. Around 47 percent of

eligible households were excluded.Indeed, a subsequent study in Nepalby Jha et al. (2009) asserts thatcommunity-based targe ng leads tosignicant errors of exclusion of thepoor and inclusion of the non-poor.


16/40

Figure 5: Propor on of households in wealth categories receiving and not receivingbenets as a result of community-based targe ng (in communi es of Pakhad andOkhreni in Nepal)29

One key explana on for the poorperformance of community-basedtarge ng is linked to the reality ofcommunity life. For community-basedtarge ng to be successful, it depends onthe existence of benevolent communi es,in which all people are concerned aboutthe welfare of the poorest and wantto channel resources towards them. It

is unlikely that many such benevolentcommuni es exist. An alterna ve vision ofcommunity life is described by Abrahamand Pla eau (2000):

Rather than idyllic village democracies

whose members interact in a freeatmosphere of trus ul coopera on basedon well-accepted social norms, theyappear as repressive socie es where

mutual control is constantly exercised,suspicions are con nuously entertainedabout others inten ons, inter-personalconicts are pervasive, and a rigid rank-based hierarchical structure governspeoples life.

In communi es that are divided along,for example, the lines of caste, ethnicity,

poli cs, religion and language, andcharacterised by high levels of socialexclusion, it is di cult to believe thatdecisions on beneciaries will alwaysbe made in the best interests of thepoor. Self-interest will clearly featureand, as with many other experiences ofcommunity-based targe ng, it is likelythat people will want to capture resourcesfor themselves, their friends and rela ves,


17/40

as indicated by the experiences above. Ofall targe ng methods, community-basedtarge ng is the most likely to be capturedby local elites and poli cians. In Malawi,village leaders have inuenced theprocess to ensure that their friends andrela ves are included in the programme.30Indeed, both the World Bank31and Jhaet al. (2009) indicate that this is likely to

happen in the Nepal context.

In most communi es, it is likely that themost vulnerable and poorest maynd itdi cult to speak up and represent theirown interests in community decision-making processes. Indeed, the reasonthat they are the poorest could well bebecause they have already been excluded.

There is good evidence that this is thecase in Nepal.32To expect that they willbe included by communi es when making

decisions on the sharing of resourcesmay be unrealis c. As the World Banknotes, in Nepal exis ng power rela onsand pa erns of social exclusion maywell be replicated in community-basedtarge ng.33Chinsinga (2005:298) foundevidence of communi es in Malawiac velynding jus ca ons to excludethe poor. And, as with proxy meanstes ng, in community mee ngs peoplend it di cult to speak out publiclyagainst the interests of others for

example, by sugges ng they not be

included in a programme as this maydamage their social rela ons.34Yet, this isof cri cal importance if community-basedtarge ng is to be successful.

A further challenge of community-basedtarge ng is that, in many communi es,the vast majority of people are poor.When communi es are asked to choosea small propor on of households, this

can be very di cult as the di erencesbetween households may be minimal(and, indeed, may change quite regularlywith seasonal and other varia ons). Ine ect, communi es are asked to selectthe poor from the poor, which is boundto lead to inaccuracies. As Sec on 4.1indicates, it is common for membersof rural communi es to view everyone

as poor, making selec on par cularlyproblema c.

Community-based targe ng assumes thatcommuni es can be easily iden ed. Yet,this is not so straigh orward. Decisionsneed to be made on the size of thecommunity. In Nepal, it would make asignicant di erence whether decisionswere made at ward or VDC level. At VDClevel, communi es are much larger andpeople are less likely to be well-informed

about the situa on of others in thecommunity. In this case, community-based targe ng begins to lose one of itspurported advantages. On the other hand,undertaking targe ng on a small scaledemands greater e ort and costs.

Decisions made by communi es can bevery subjec ve. And, as the World Banknotes, decisions can change dependingon who makes the choice.35In Honduras,for example, Bergeron et al (1998) askeddi erent randomly selected groups to

rank households and each came upwith very di erent lists. In Indonesia, itwas found that the inclusion of womendepended, to a large extent, on whetherthe selec on of beneciaries was doneduring the day or night.36In Malawi,di erent communi es arrived at almostiden cally opposite decisions whendeciding whether to include specic typesof household in the programme.37

In manycommuni es

the vastmajority ofpeople are

poor, so its

very di cult forcommuni es to

iden fy a smallpropor on of

householdsbecause thedi erences

between themmay be minimal

and vary

according tothe seasonsfor instance.

In e ect,communi esare asked to

select the poorfrom the poor,which is bound

to lead toinaccuracies


18/40

When community-based targe ngis undertaken on a na onal scale,its accuracy further decreases as aresult of the heterogeneity betweencommuni es. If each community in aregion is given axed and equal quota,

errors will increase due to di erencesbetween communi es in terms ofwealth. So, if 20 percent of householdsare targeted in a wealthy community, itis unlikely that anyone selected will bein the poorest 20 percent of householdsna onwide. Similarly, in very poor partsof the country, some communi es willcontain only households that fall belowthe na onal poverty line, yet they will beforced to select a small propor on (i.e.20 percent) to receive a transfer when,in reality, all should be en tled.

Community-based targe ng requireswell-trained facilitators if its chancesof success are to be enhanced. Thismay be possible to achieve on a pilotscale level when signicant resourcescan be invested in facilita on. But theWorld Bank is scep cal that adequatetraining of consistent quality couldbe undertaken on a na onal scale inNepal.38In Pakistan, it was found thattraining surveyors for the much simpler

task of undertaking a proxy means testin a large pilot programme was very

challenging, resul ng in surveyors havingwidely variable abili es. 39In Malawi,despite training, 48 percent of thosefacilita ng the community targe ngprocess did not even fully know thetarge ng criteria.40

Community-based targe ng su ers fromsimilar challenges as proxy-means tes ngwith regard to poverty dynamics andre-cer ca on. If re-cer ca on is notundertaken with su cient frequency,targe ng accuracy deteriorates. However,frequent re-cer ca on is likely to be tooexpensive to undertake and, poten ally,overly disrup ve for communi es.

In sum, therefore, evidence indicatesthat community-based targe ng is

unlikely to deliver accurate targe ng.As with the proxy means test, it leads tosignicant leakage with a large propor onof resources going to non-intendedbeneciaries. It is also unlikely to bepossible in urban areas since cohesivecommuni es are probably even lessiden able compared to rural areas.Hobley (2010) who has undertaken

research on community-based targe ng inNepal has concluded that it is important

to move away fromcommunity based

approaches as they risk projec ng pastinjus ces into the future.

By includingthe be er-o , targe ngof the poor ismade moreaccurate.Indeed, thebe er-omay self-exclude fromprogrammesbecause theydo not feelthat it is worththe e ort toapply for arela vely smallbenet


19/40

3.3.E ec veness ofuniversal targe ng

As indicated by the World Bank(1994:240), universal targe ng performs

very well in terms of ensuring theinclusion of poor people in programmes.41So, in contexts when the inclusion ofpoor people in programmes is priori sed,many countries have adopted universaltarge ng in their non-contributorypension programmes.42On the otherhand, of course, non-poor are alsoincluded in the programmes and thenumber can depend on the level ofpoverty in a country.

Yet it is important to recognise that this is

not an error or leakage.43By including thebe er-o , targe ng of the poor is mademore accurate. And there are also goodpoli cal economy reasons for includingthe non-poor, which are explained inSec on 4.4.

Indeed, the be er-o may self excludefrom programmes because they do notfeel that it is worth the e ort to apply fora rela vely small benet. This happensin Nepal, as is evidenced by the lowtake-up of the Senior Ci zens Allowancein Kathmandu where the propor on ofpeople living below the poverty line is afrac on of the na onal average. In NewZealands universal pension, the benetis recovered from the be er-o bycon nuing to tax people in old age.44

Targe ng Dalits is no solu on

In Nepal, both the Senior Ci zens Allowance and Child Protec on Grant are targeted atexcluded groups, specically the Dalit popula on.

The Senior Ci zens Allowance is given to all Dalits aged 60-69 years. (The allowance isuniversal for all over 70s). In contrast the Child Protec on Grant is given only to poorDalit children aged 0-4 years (with the excep on of Karnali).

Although there is a correla on between being a Dalit and being poor, it is far fromperfect. Three signicant challenges arise if excluded groups such as Dalits are targeted.

Firstly, many poor households in Nepal are not Dalit indeed, 83 percent of poor and

vulnerable households (using twice the na onal poverty line) are non-Dalit andthey would be excluded by programmes only targeted at Dalits.

Secondly, Dalit targe ng poten ally further s gma ses Dalits, by highligh ng them asa dis nct group in need of special assistance.

And, thirdly, it could create resentment among the majority of the poor some

of whom will be poorer than their Dalit neighbours thereby threatening social

cohesion.

Box 2


20/40

In countries where poverty levels arehigh and many people are in danger of

falling into poverty, it may make li lesense to use poverty targe ng. In Nepal,when a poverty line twice that of thena onal poverty line is used as a meansof measuring both poor householdsand those vulnerable to poverty,households with children underveyears have a poverty rate of 81 percentwhile households with people over-60have a poverty rate of 69 percent.45It isimportant to provide benets to those indanger of falling into poverty as a meansof building their resilience. Indeed, it

should be borne in mind that the povertyrate can disguise the real number of poorpeople. The poverty rate is a very sta cmeasure and, over a three-year period,a much higher number of people willspend some me in poverty. For example,in Pakistan, while the poverty rate iscalculated at 17 percent, the World Bank(2009c:77) has es mated that 67 percentspend some me in poverty over threeyears. Any reasonable measure of povertywould regard those who spend some me

in poverty as poor.

There are countries that have tried to usepoverty targe ng to reach a similar highpropor on of households, but many eligiblepoor people s ll miss out. In rural Mexico,for example, the Progresa programmecovered 78 percent of people, yet 16 percentof intended beneciaries were omi ed.46And, in South Africa, the non-contributorypension reaches 68 percent of thepopula on, but 14 percent of eligible peoplemiss out.47Although both programmes haveinvested a signicant amount of me, moneyand e ort to exclude a small propor on of

the popula on, this s ll leads to many poorpeople being unable to access them.

Therefore, if the policy priority is to includeall poor people, the chances of success willbe enhanced if everyone is targeted. Thisis of par cular importance to consider incontexts such as Nepal where there are highlevels of poverty and vulnerability and largenumbers of people are in danger of fallinginto poverty. As indicated above, the be er-o may exclude themselves anyway.

Lower the age of Nepals Senior Ci zens Allowance to close poverty gap

Although everyone aged over 70 is eligible for Nepals Senior Ci zens Allowance, the age

of eligibility for Dalits and people living in Karnali is 60 years. Currently, around 682,000people receive the Senior Ci zens Allowance48and 314,000 single women receivea benet from 60 years of age. A total of 996,000 over 60s currently receive old agebenets. The total number of over-60s is around 1,756,000.49

The programme as currently implemented is es mated to have reduced the povertyrate among households with members aged 70 and older by 20 percent and reduced the

poverty gap for this group by 35 percent, compared to a scenario without the transfers.Among households with members aged 60 and older, these reduc ons are es mated tohave been 12 percent and 23 percent respec vely.

Lowering the age of en tlement of the Senior Ci zens Allowance and incorpora ngthe Single Womens Allowance within it would mean that even more poor people

could be reached. For example, an allowance of Rs.500 per month targeted at everyoneover-60 would reach a further 18 percent of the popula on na onally (in addi on tothe 22 percent reached currently) and reduce the poverty gap in these households withmembers aged 60 and older by a further 17 percent, resul ng in an overall reduc on inthe poverty gap of 36 percent.

Box 3


21/40

If there are concerns about the costs ofuniversal targe ng, one op on may be to

gradually expand programmes. There aretwo means of doing this:

The programme could be targetedat a more restricted age range. Forexample, Nepals Senior Ci zens

Allowance was ini ally targeted ateveryone aged over 75 years beforebeing reduced to age 70. Box 3indicates the addi onal impact theprogramme would have if the age ofeligibility were to be lowered to 60years.

The other op on is to target universalgrants in specic areas with higherpoverty levels before rolling them outto other poor districts and eventuallyna onwide. For instance the ChildProtec on Grant is currently targetedat all mothers in the Karnali zone.Nepal could progressively expandcoverage of the Child Protec on

Grant to further poor districts un luniversal coverage across the country

is achieved.

Universal coverage of the Child Protec onGrant for all underves would costapproximately 0.8 percent of GDP,reach 48 percent of households (and 66percent of poor households) and reducethe poverty gap in these householdsby a further 13 percent resul ng in anoverall impact from the child grant of17 percent. Providing such a grant to allchildren underve, rather than limi ngit to only two children per household,

would dispropor onately benet poorhouseholds who tend to have largernumbers of children as well as those fromexcluded groups, but add li le to thecost of the programme while simplifyingadministra on.50Box 4 indicates theimpacts that would be achieved fromcombining a grant for all childrenunderve and an old age allowance foreveryone aged 60 and over.

Cut poverty in Nepal by combining a universal old age allowance with a child grant

If Nepal were to implement both an old age allowance for everyone over 60 years ofage and a grant for all children below ve, the impacts on poverty would be signicant.The programme would reach around 75 percent of the popula on and 88 percent of thepoor. While the current old age allowances and child protec on grant reduce the na onalpoverty gap by 8 percent, the expansion to all over-60s and all children under ve wouldbring a further reduc on in the poverty gap of 16 percent. The overall reduc on from bothcurrent and expanded programmes would be 24 percent.

The cost of implemen ng both of these programmes would be around 1.5 percent of GDP(approximately equal to what Lesotho spends on a pension for the over-70s alone).51

Universal targe ng as prac sedwithin Nepal with the Senior Ci zensand Single Womens Allowance is

an e ec ve means of maximizing theinclusion of the poor. Extendingthe Child Protec on Grant to allchildren underve would go a longway towards ensuring that all pooryoung children receive the grant. This

cannot be achieved by poverty targe ngwhich would, in contrast, necessarilylead to many poor children missing out.Furthermore, leakage through universaltarge ng would probably be minimal, inpar cular when compared with povertytarge ng which, as explained in sec ons3.1 and 3.2, is likely to have a leakage rateof around 50 percent.52

Universaltarge ng is an

e ec ve meansof maximizingthe inclusionof the poor.

Extendingthe Child

Protec onGrant to allunderves

would goa long wayto ensuring

that all pooryoung children

receive thegrant

Box 4


22/40

Any analysis of targe ng that onlyassesses its accuracy is limited. It isalso important to examine the costs oftarge ng methods. Some of these costsarenancial, such as administra ve costsor the overall cost of the programme.However, there is a range of other typesof costs that are associated with targe ng,including moral, social, incen ve andpoli cal. All need to be included in anassessment of targe ng methods.

Amartya Sen (1995:11) observed that theuse of the term targe ng assumes thatpeople are passive agents. However, hepoints out that people have agency andcan respond in various ways to targe ngmethodologies, o en undermining them.

This sec on will examine how targe ngmethods can change human behaviour,looking specically at the broadercosts. It will conclude by examining theimplica ons of targe ng for thenancialcosts of programmes.

4.1.Social costs of targe ng

Poverty targe ng can have signicantsocial costs. There is solid interna onalevidence that poverty targe ng can causeconicts in communi es when somepeople are chosen and others are not. The

4THE WIDER COSTS OF TARGETING

problem is exacerbated by the fact thatmany poor people who are eligible for theprogramme are excluded. Social conicthas been observed with both proxy meanstes ng and community-based targe ng.

The fact that many communi es acrossthe developing world regard themselvesas egalitarian poses a challenge for thepoverty targe ng method.53When povertytarge ng is imposed on communi es it iscommon to hear members exclaim: We

are all poor here. Such a refrain has been

noted in African, Mexican, Nicaraguanand Indonesian communi es.54In Malawi,for example, a community argued:We are one group of people therefore

targe ng some and leaving out others isnot right.55Indeed, Ellis (2008) points outthat it also reects the economic reality ofmany poor communi es: the di erencein income between the poorest and manyothers in the community is minimal.O en, as noted earlier, poverty targe nga empts to single out some poor people

from a large number of poor.

In Mexico, Nicaragua and Indonesia, proxymeans tes ng has led to varying degreesof social conict in communi es. Studiesin Mexico and Nicaragua demonstratehow feelings of despair, frustra on,envy, resentment and jealousy fromnon-beneciaries many of whom are

Proxy meanstes ng hasled to varyingdegreesof socialconict incommuni es.Studies inMexico andNicaraguademonstrate

how feelingsof despair,frustra on,envy,resentmentand jealousyfrom non-beneciaries many of

whom are poor are common


23/40

poor are common.56

Indeed, in Mexico,Nicaragua and Indonesia, divisions causedby the proxy means test have led non-beneciaries to withdraw their labourfrom voluntary community ac vi es.

There is evidence of direct conict: insome communi es in Mexico, whenbeneciaries were cleaning the streetsthe non-beneciaries threw rubbish; inothers, fences mended by beneciarieswere subsequently knocked down by non-beneciaries.57In a community visited byHannigan (2010), the ini al distribu on ofthe Indonesia PHK programme provokedstone throwing and the burning downof a building. And, in Lebanon, theintroduc on of proxy means tes ng led toriots in some refugee camps.58

The poten al for social conict maywell be greater with community-basedtarge ng than with proxy means-tes ng. With the proxy means test, if itis not well-implemented, it is possibleto blame outsiders.59This is not the

case with community-based targe ngsince decisions are made by communitymembers themselves. For example, 36percent of people in the Mchinji schemein Malawi thought that the choicesmade by community representa veswere unfair, with many deserving

people (as much as 47 percent in somecommuni es) missing out. Yet this was inthe context of a small pilot programmewhere signicant expense was investedin the targe ng process.60In Malawicommunity members even expressedfears that community-based targe ngwould lead to resentment, social tensionsand retribu on against the beneciaries.61And, as with proxy means-tes ng, non-beneciaries may withdraw their labourfrom community ac vi es.62

To avoid crea ng social conict,communi es can ac vely underminethe community-based targe ng processby refusing to use poverty-targe ng.For example, in South African workprogrammes, instead of iden fyingthe poor, communi es preferred tochoose lots out of a hat so that everyonehad an equal chance of selec on and,presumably, no one could be blamed.63In Malawi, it has been reported thatcommuni es have taken the cash benets

in a poverty-targeted programmeand distributed it equally among allbeneciaries.64The World Bank in Nepalindicates that communi es may refuse topoverty target by classifying everyone aspoor so that benets are distributed toeverybody.65

In contrastto poverty

targe ng, thereis no evidencethat providing

benets toeveryone

in a speciccategory such

as all old oryoung people

leads to

social conict.Given that such

groups areo en regardedas deserving,

people tendto accept that

they shouldreceive benets


24/40

In Nepal, there is also evidence ofpoverty targe ng causing socialproblems in communi es. Hobleyand Paudyal (2008) report stories ofdivision and social conict as a resultof community-based targe ng.66Such experiences are consistentwith a empts by the World FoodProgramme to introduce community-based poverty targe ng into theirfood distribu on, which has generatedopposi on in communi es.67

In contrast to poverty targe ng, there isno evidence that providing benets toeveryone in a specic category suchas all old or young people leads to

social conict. Given that such groupsare o en regarded as deserving, andthat at some me in their lives people

will be part of these groups themselves,people tend to support these benets.In fact, it is common with community-based targe ng for communi esthemselves to select categoriesthat they regard as deserving, in

par cular the elderly and people withdisabili es.68For example:

In the Mchinji cash transferprogramme in Malawi, the meanage of heads of beneciaryhouseholds is 62 years69

In Rwanda, 73 percent ofhouseholds chosen for its DirectSupport programme have an olderor disabled person70In the Kalomo programme inZambia, around 65 percent of theheads of beneciary households areover 60 years of age71

Good social rela ons are a key assetof the poor. There needs to be seriousreec on on whether poverty targe ng

given its inaccuracies is a worthwhile

price to pay for undermining poorpeoples social capital. As Adato and

Roopnaraine (2004:79) conclude, whenreec ng on the impact of poverty-targe ng in Nicaragua:

It is necessary to take these social

costs into account in evalua ng futuretarge ng op ons.

As Sen (1995:13) and Grosh et al.(2008:104) point out, a further socialcost of poverty targe ng is the poten als gma sa on of beneciaries. Sen(1995:13) argues:

Any system of subsidy that requires

people to be iden ed as poor and thatis seen as a special benefac on for those

who cannot fend for themselves wouldtend to have some e ects on their self-respect as well as on the respect accordedthem by others.

Many poor people do not want to beiden ed as poor, yet poverty targe ngmechanisms can ac vely use a processof s gma sa on to help iden fy them.For example, pu ng lists of beneciariesup on walls or reading them out incommunity mee ngs can be a source

of embarrassment and s gma sa on.In Malawi, some beneciaries foundthis process to be very painful.72In theNepal context, Jha et al. (2009) havenoted how community-based targe ngcan lead to greater s gma. The proxymeans test prepared by the World Bankin Nepal proposes publishing the list ofbeneciaries in na onal newspapers,which may be regarded as a poten allys gma sing measure.73

There isevidenceof poorhouseholdsdeliberatelyprovidingmisleading

informa onabout theirlevel of povertyso as to avoidbeing iden edas poor


25/40

Indeed, there is evidence of poorhouseholds deliberately providingmisleading informa on on theirlevel of poverty so as to avoid beingiden ed as poor.74This can lead toeligible households not being includedin programmes. In Nepal, Hobley andPaudyal (2008) found that:

There is .. clear evidence that people

are manipula ng the well-being rankingas they do not want to be labeled as pooror extreme poor as this will a ect thesocial status of the household and themarriage chances for their daughters.

One of the key advantages of a universalprogramme is that it can be presentedas an en tlement and the right of

everyone. As a result, it is unlikely to be

s gma sing. Again, policy-makers needto consider whether it is worthwhileimposing s gma on poor people when,as Rawls (1971) suggests, self respect isperhaps the most important primary

good.75

4.2.Moral costs of targe ng

As Sen (1995:12) points out, targe ng

is a process that rewards chea ng andpenalises honesty. If someone cansuccessfully lie about their well-beingto qualify for a programme, they willbe rewarded by receiving the benet;in contrast, those who are honest indeclaring that they are not poor will notreceive the benet.

While all targe ng methods can bemanipulated by poten al beneciaries,it is more likely to happen when the

informa on required for targe ng isdi cult to verify. With poverty targe ng,obtaining evidence of income is verydi cult, in par cular when the majorityof people work in the informal sector orsubsistence agriculture. As indicated inSec on 3.1, many of the proxies usedin proxy means-tes ng can be di cultto verify. Indeed, in Perus Juntos

programme which uses proxy means-

tes ng Huber et al. (2009:45) notedthat:

A certain quan ty of beneciariesentered into the programme becausethey lied, while others who respondedhonestly to the ques ons damaged theirown interests because they were notincorporated.

Advocates of community-basedtarge ng claim that it is more di cultfor beneciaries to lie because othercommunity members know abouttheir real situa on. Yet, in the Mchinjiprogramme in Malawi, nine percentof all households were found to havecreated ghost members, presumably

to increase their chances of beingselected for the programme.76One ofthe main reasons why Mauri us movedfrom a poverty targeted to a universal

non-contributory pension in 1958 wasbecause of complaints that those whohonestly declared their income lostthe pension while those who falsiedinforma on were accepted on to thescheme.77

Policy-makers need to decide whetherit is worthwhile introducing a targe ngsystem that rewards the dishonesty oflarge numbers of people and poten allyerodes the moral fabric of a society.

Targe ngis a process

that rewardschea ng.

Those who liesuccessfullyabout their

well-being toqualify for aprogrammereceive the

benet, whilethose who

are honest indeclaring that

they are notpoor do notreceive the

benet


26/40

4.3.Incen vecosts of targe ng

Poverty targe ng can inuence peopleseconomic behaviour. Both Sen (1995) andthe World Bank (1994) point out that theprospect of losing a benet by earning toomuch could deter people from engagingin economic ac vi es. The level of thisdisincen ve may be di cult to measurebut Sen (1995:13) argues that:

It would be natural to expect that there

would be some signicant distor ng shi sif the qualica on for the subsidy is basedon a variable (such as income) that isfreely adjustable through changing ones

economic behavior [sic].

There is evidence of poverty targe ng

crea ng perverse economic incen vesin developing countries. For example,the South African non-contributorypension is poverty-targeted. Peopleare reported withdrawing savings fromcontributory pension schemes so thatthey can appear poor and qualify for thenon-contributory pension.78Sen (1995:18)provides examples of families in faminerelief situa ons keeping one child hungryso that they can qualify for nutri onalsupport. Morris et al. (2004) found that

a condi onal cash transfer programmein Brazil the Bolsa Alimentacao had asurprisingly nega ve impact on childrensweight and height.79They surmise thatit may be because parents believed thatchildren needed to be underweight toqualify for the benet.

One common belief with proxymeans tes ng is that disincen ves forbeneciaries to engage in labour arereduced with proxy means tes ng.80

Yet, there is no reason why proxymeans tes ng should not createperverse incen ves if implemented in atransparent manner. If beneciaries areaware that possessing certain proxies

such as animals or a good qualityroof will count against them, they

may well not invest in such assets. Infact, one likely reason for the absenceof evidence on perverse incen vesin proxy means tes ng is that re-cer ca on is rarely undertaken. As aresult, beneciaries may well not haveexperienced people being removedfrom the programme as a result ofbecoming wealthier.

As Sen (1995:18) notes, targe ngprogrammes based purely ondemographic characteris cs thatcannot be manipulated by beneciaries such as old age, young age, gender

and disability are not likely to create

perverse incen ves. Indeed, both theWorld Bank (1994:240) and Mkandawire(2005:15) argue that one key advantageof universal programmes is that theydo not introduce disincen ves to work.Furthermore, universal pensions donot discourage people from savingin contributory pension schemes.

This could have signicant economicbenets for countries that want toincrease the number of contributorsto such schemes, thereby expandingna onal savings and providing increasedfunds for investment in business.

If beneciariesare aware thatpossessingcertain proxies such as

animals or agood qualityroof will

count againstthem, they maywell not investin such assets


27/40

4.4.Poli calcosts of targe ng

Mkandawire (2005:13) points out thatexperience in developed and middle-income countries indicates that universaltarge ng is one of the best means ofbuilding middle class poli cal supportfor the nancing of social protec onprogrammes. Poverty-targetedprogrammes exclude the middle-classand, as a result, are less likely to receivetheir support. As Fiszbein and Schady(2009:59f) of the World Bank note,standard economic theory wouldargue that:

Transfer schemes narrowly targeted

to the poor would tend to have limitedsupport because a small share of the

popula on benet, whereas the costs aredispersed across all tax-payers.

This concurs with Sens earlier

assessment (1995:14) that:

The beneciaries of thoroughlytargeted poverty-allevia on programsare o en quite weak poli cally andmay lack the clout to sustain theprograms.

The comparison between the levelof government spending investedin universal and poverty-targetedprogrammes tends to bear this out.As Figure 6 indicates, universalpension programmes across the worldhave signicantly higher budgetscompared with narrowly targetedprogrammes.81Indeed, the various oldage programmes in Nepal have muchhigher poli cal andnancial support

than the poverty-targeted pensions ofIndia and Bangladesh.

Figure 6: Comparison of budgets allocated to universal and poverty-targeted pensionprogrammes


28/40

Therefore, as Pritche (2005) argues,poverty targe ng can deliverworse benets for the poor thanuniversal programmes. Budgets and

benets may be smaller when comparedwith universal programmes.82Indeed,Gelbach and Pritche (2002) havedeveloped a model in which they suggestthat poverty targe ng will lead to lowerbenets for the poor.83Moene andWallerstein (2001:22) point out:

The op mal policy for the very poor is notnecessarily a policy that targets benetsas narrowly as possible, once the impactof targe ng on poli cal support is takeninto account.84

One addi onal benet of universalprogrammes is that they are more likelyto strengthen the social contract betweenci zens and the state. As an en tlement,they are given to everyone as a right. Incontrast, poverty targeted programmesgive more opportuni es to unscrupulous

poli cians to direct benets tocertain groups rather than to others.In a country like Nepal, where it isimportant to build and strengthen thestate, providing benets to all ci zenswithin a certain category such as the

elderly, people with disabili es andyoung children could, over me, helpbuild the na on and increase peoplesawareness of themselves as ci zens.

4.5.Financial costsof targe ng

As indicated in the previous sec on,poverty-targeted programmes costless than universal programmes.However, this should not be confusedwithscal sustainability. Universalprogrammes, despite higher levels ofspending, may well be morescallysustainable due to their greaterpoli cal support.

Poverty-targe ng can lead to worse outcomes for the poor

In the 1990 World Development Report, the World Bank highlighted the danger thatpoverty targe ng could undermine poli cal support for programmes, leading to worseoutcomes for the poor: 85

In prac ce, the success of public interven ons involves more than cost-e ec veness.The demands made by di erent sec ons of the popula on, and their ability to exertpressure on the authori es, are o en more inuen al than the governments economiccalcula ons. Fine targe ng based on a single-minded concern for cost-e ec veness canreduce public interest in the vigorous implementa on of government programmes to help

the poor. For example, in the late 1970s Sri Lanka replaced a universal food subsidy witha less costly, targeted food stamp program. In me, the benets delivered by the newprogram declined. The middle classes no longer gained from the scheme, and althoughthe new program was more cost-e ec ve, it lost crucial poli cal support. Similarly, a foodsubsidy directed to poor consumers in Colombia was so ghtly targeted that it lacked ane ec ve poli cal cons tuency, and it was dropped at a change of administra on. Theanalysis of public policy has to be alive to these considera ons of poli cal economy.

In terms of administra ve costs,poverty targe ng is more expensivethan universal programmes, which arevery simple, not requiring anything

like as sophis cated and expensive anadministra ve structure. Indeed, there isnoxed rule on the administra ve costs

of poverty-targe ng; the lower the errorsdemanded, the more that needs to beinvested in poverty targe ng. In MexicosProgresa programme, a desire to reduce

administra ve costs in urban areas leddirectly to a signicant deteriora on intarge ng outcomes.86

Box 5


29/40


30/40

1

No one in receipt of another form of income from the state such as a salary orcivil service pension is eligible to receive a social transfer. Therefore, in theory,

Nepals grants enable universal coverage of benets rather than the use of strict

universal targe ng.2The grant for the par ally disabled is heavily targeted with a maximum of two

beneciaries per VDC. There are only 7,000 beneciaries across Nepal.3While all mothers in the Karnali region are targeted, each mother is limited to amaximum of two children who are able to receive the grant.4There are forms of community-based targe ng in which community membersprimarily serve as the extended arm of the social protec on bureaucracy andadminister a ques onnaire. This is not considered here as it is primarily a meansof saving costs in administering some form of proxy targe ng.5Veras et al. (2007:2).6World Bank (2009a).7Coady et al. (2004).

8 R-squared results for proxy means test regressions tend to oscillate between0.4 and 0.6. For example, The World Bank (2009b) found an r-squared result forBosnia of 0.488 and gave r-squared results for other countries of: Armenia, 0.2;

Bosnia, 0.49; Georgia, 0.62; Russia, 0.58; Sri Lanka 0.59. Ahmed and Bouis (2002)calculated an r-squared for their work in Egypt of 0.43 and stated that in La n

America r-squared values were in the range 0.3-0.4. In Indonesia, Alatas et al.(2009) gave an r-squared value of 0.48.9This proxy means test is based on the PMT formula used for the Girls

Scholarship Programme, modied slightly to reect the availability of variable inthe NLSS 2003/4 dataset10These exclusion errors are based on extending benets to the lowest-rankedhouseholds (according to the proxy score, or in other words the predictedvalues) in the popula on un l coverage matches the targeted popula on . Thismeans that inclusion and exclusion errors are necessarily the same (since for

each individual wrongly excluded there is another that is wrongly included).

Another approach commonly used by analysts is to set the cut-o based on actualconsump on (instead of the predicted value). However this method is highlyproblema c as it produces another (en rely ar cial) error whereby coverage ofthe programme will be less than the target. For a full cri que of this alterna vemethodology, see Kidd and Wylde (forthcoming).11The results found by Kidd and Wylde (forthcoming) are consistent with theerrors found by other analysts in a range of countries.12See Kidd and Wylde (forthcoming) for a more in-depth discussion on theweaknesses of household surveys and their use in proxy means tes ng.13Kidd and Wylde (forthcoming) provide a comprehensive analysis of the errorsthat are introduced into the proxy means test during implementaon.14Adato et al. (2000), Adato and Roopnaraine (2004:16), Huber et al. (2009:46f)and GHK (2009:107 and 155).15Coady and Parker (2005).16Coady and Parker (2005:26).17See, for example, GHK (2009).18GHK (2009).19Cf. Narayan and Yoshida (2005).20See Sepulvedra (2009) for a discussion of human rights principles in cashtransfer programmes.21Skouas et al. (1999:iv).22See: Adato (2000), Adato et al. (2000), Adato and Roopnaraine (2004:17, 21),

and Huber et al. (2009:46f).23Source: MLD (2010).24Ward et al. (2010).


31/40

25

Grosh et al. (2008).26Source: Galasso and Ravallion (2001). Decile one is the poorest.27See: Watkins (2008), Miller et al. (2008), and Seaman et al. (2008).28Cf. Watkins (2008).29Hobley and Paudyal (2008).30Chinsinga (2005) and Miller (2008:27, 38f).31See: Isik-Dikmelik (2009).32See: Department for Interna onal Development and the World Bank (2006).33See: Isik-Dikmelik (2009).34Cf. Miller et al. (2008:28f).35See: Isik-Dikmelik (2009).36Alatas et al. (2009:36).37Miller et al. (2008:28).38See: Isik-Dikmelik (2009).39GHK (2009).

40Miller et al. (2008:25).41A recent World Bank paper on targe ng in Nepal appears to misunderstand

this important point (Isik-Dikmelik 2009). It stated that: the targe ng of the .

elderly allowance program is likely not very e ec ve, an interpreta on thatcontrasts signicantly with the views of pension experts within the World Bank.42Developing countries with universally targeted non-contributory pensions in

addi on to Nepal include: Bolivia, Namibia, Lesotho, Botswana, Mauri us,Seychelles, Maldives, Brunei, Samoa, Kiriba , Tuvalu, East Timor and Kosovo.In addi on, in Mexico there are universal pensions in Mexico City and smallrural towns and the rural pension in Brazil is almost universal. The pension inSwaziland is, in theory, poverty targeted but, in prac ce, universal. Universaltarge ng of cash transfer programmes is also common in developed countries.43A recent paper by the World Bank on targe ng in Nepal makes the error of

regarding the inclusion on non-poor in universal programmes as leakage. See

Isik-Dikmelik (2009).44St. John and Willmore (2001:1292).45These es mates are calculated using a poverty line set at twice the na onalpoverty line, using per-capita consump on es mates as per the 2003/4 NLSS.46Skouas et al. (1999).47Samson et al. (2007).48Thegures of 682,000 beneciaries for the Senior Ci zens Allowance and314,000 Single Womens Allowance is based on the number of beneciaries

in 2008 increased by 3 percent a year, the approximate rate of growth of theover-60 popula on. Figures were provided by Shizu Upadhaya as part of anassignmentnanced by UNICEF49The current over-60 popula on is based on data from the Central Bureau ofSta s cs popula on projec ons.50Although opponents of child grants o en argue that they will encouragefamilies to have more children, there is almost no evidence of this interna onally

in developing countries, when grants are small in size. Recent studies in Mexico,

Nicaragua and South Africa have shown no impact at all (Stecklov et al. 2006;Department of Social Development 2006; Makiwane and Udjo 2006). SeeFiszbein and Schady (2009) for an overview of some of the evidence. It is unlikelythat families in Nepal would be encouraged to have more children for a monthlygrant of only Rs.200.51Thegure of 1.5 percent assumes that 10 percent of the popula on do nottake up the grant mainly as a result of self-exclusion by the be er-o anddoes not include administra ve costs.52Leakage in this context means the propor on of grants that go to non-intendedbeneciaries. In the context of a universal child grant for 0-4 year olds, leakagewould refer to children above the age of four receiving the programme. In termsof a poverty targeted child grant, leakage would refer both to children above theage of four and those from families who are richer than the poverty line used fortarge ng the grant.


32/40

53Cf. Mgemezulu (2008:73).54See Kidd (1999), Adato (2000), Adato et al. (2000), Adato and Roopneraine

(2000), Mgemezulu (2008:73f), Huber et al. (2009:49), Ellis (2008) and Hannigan(2010).55Mgemezulu (2008:73).56Cf. Adato (2000), Adato et al. (2000), Adato and Roopnaraine (2004) andHannigan (2010).57Adato (2000).58Source: Georgia Rowe, personal communica on.59Adato (2000).60Miller (2008:38).61Mgemezulu (2008:73 ).62Cf. Mgemezulu (2008:76). The refusal by non-beneciaries to engage incommunity ac vi es has been observed in Ethiopia by one of the authors of thispaper.63Adato and Haddad (2002).64Personal communica on: Paul Msoma and Blessings Chinsinga.65See: Isik-Dikmelik (2009).66Jha et al. (2009) state that community-based targe ng can be divisive in Nepal.67Personal communica on: Siemon Hollema, WFP.68Cf. Alatas et al. (2009).69Miller (2008:33).70Personal communica on: Louis Marie Asselin. A similar situa on of

communi es choosing older people has been observed in the Kalomoprogramme in Zambia71(MCDSS/GTZ 2007:19).72Miller (2008:29).73See Proxy Means Tes ng (PMT) Guidelines, Version November 2009.74Cf. Adato et al. (2000).75Cited in Sen (1995:13).76Miller et al. (2008:27).77See Willmore (2006).78Samson et al. (2007). A similar process happens in Australia (Sass 2004).79See also Fiszbein and Schady (2009:150f).80For example, see Grosh et al (2008:100).81South Africa is an example of one poverty targeted pension programme with a

signicant budget a ached. However, the South African pension is an anomalyas it is received by almost all of the black popula on who dominate electoraldecisions, which, in e ect, proves the point. The focus here, therefore, is onnarrowly targeted programmes that reach a minority of the popula on.82

Cf. Subbarao et al. (1997).83Cited in Fiszbein and Schady (2009:60).84Cited in Mkandawire (2005:13).85See World Bank (1990:92).86Coady and Parker (2005).87Holzmann et al. (2005:95).


33/40

Abraham, A., and J-P. Pla eau(2000) Par cipatory

development in the presence of endogenous communityimperfec ons. Dra paper, Department of Economicsand CRED (Centre de Recherche en Economie duDveloppement), University of Namur, Belgium.

Adato, M.(2000) El impacto de Progresa sobre las relacionessociales en la comunidad. Unpublished paper by IFPRI.

Adato, M., D. Coady and M. Ruel (2000) Evaluacin deoperaciones de progresa desde la perspec va de losbeneciarios, las promotoras, directores de escuela ypersonal de salud. Unpublished paper by Interna onal

Food Policy Research Ins tute (IFPRI).

Adato, M., and L. Haddad(2002) Targe ng poverty throughcommunity-based public works programmes: experiencefrom South Africa. In Journal of Development Studies38(February): 1-36.

Adato, M., and T. Roopnaraine(2004)A social analysis of theRed de Proteccion Social (RPS) in Nicaragua.Final report,Interna onal Food Policy Research Ins tute (IFPRI).

Alatas, V., A. Banerjee, R. Hanna, B. A. Olken and J. Tobias(2009) How to target the poor: evidence from a eldexperiment in Indonesia. Unpublished manuscript.

Bergeron, G., S. S. Morris and J. M. Banegas(1998) Howreliable are group informant ra ngs? : a test of foodsecurity ra ngs in Honduras. InWorld Development,26(10): 1893-1902.

Castaeda, T., and K. Lindert(2005)Designing and implemen nghousehold targe ng systems: lessons from La n Americanand The United States.Social Protec on Unit, HumanDevelopment Network, The World Bank.

BIBLIOGRAPHY


34/40

Chinsinga, B.(2005) The clash of voices: community-basedtarge ng of safety-net interven on in Malawi. InSocialPolicy and Administra on,39(3): 284-301.

Coady, D., M. Grosh and J. Hoddino (2004)Targe ng oftransfers in developing countries: review of lessons andexperience.Interna onal Food Policy Research Ins tute(IFPRI).

Coady, D., and S. Parker(2005)Program par cipa on undermeans-tes ng and self-selec on targe ng methods.FCND discussion paper 191, Food Consump on andNutri on Division, Interna onal Food Policy ResearchIns tute. Washington D.C.

Ellis, F.(2008)We are all poor here: economic dierence,social divisiveness and targe ng cash transfers in sub-Saharan Africa. Paper prepared for the conference,Social protec on for the poorest in Africa: learning fromexperience, Uganda 8-10 September 2008.

Department for Interna onal Development (UK) and the WorldBank(2006)Unequal ci zens: gender, caste and ethnicexclusion in Nepal.

Department of Social Development(2006) Incen ve structuresof social assistance grants in South Africa. Pretoria:Department of Social Development.

Fiszbein A., and N. Schady(2009)Condi onal cash transfers:reducing present and future poverty.World Bank,Washington DC, USA.

Galasso, E., and M. Ravaillion(2001)Decentralized targe ng ofan an -poverty program.Development Research Group,World Bank, Washington DC, USA.

Gelbach, J., and L. Pritche (2002)Is more for the poor lessfor the poor? In BEP topics in economic analysis andpolicy, 2(1).

GHK Consul ng Ltd.(2009) Pakistan process evalua on ofBenazir Income Support Programme (BISP) scorecard-based poverty targe ng under the test phase: report onndings. Unpublished report.

Grosh, M., C. del Ninno, E. Tesliuc and A. Ouerghi(2008)Forprotec on and promo on: the design and implementa onof e ec ve safety nets.Washington, DC: The Interna onalBank for Reconstruc on and Development/The WorldBank.


35/40

Hannigan, L.(2010) Social assistance in Indonesia -understanding the gaps. Dra version of paper presentedat the annual Indonesia update conference, Canberra,

Sept 24-25 2010.

Hobley, M. and B. R. Paudyal(2008) Study of income-genera ng ac vi es in Kapilvastu and Ramechhapdistricts, Nepal. Unpublished manuscript.

Hobley, M. (2010) Background paper for SSPP design team.Unpublished manuscript.

Holzmann, R., R. Hinz, et al.(2005)Old-age income support inthe twenty-rst century: an interna onal perspec ve onpension systems and reform.World Bank, Washington,

DC, USA.

Huber, L., P. Zarate, A. Durand, O. Madalengoi a, and J. Morel(2009)Programme Juntos: certezas y malentendidos entorno a las transferencias condicionadas estudio de

caso de seis distritos rurales del Peru.UNFPA, Ins tuto deEstudios Peruanos and UNICEF. Lima: Peru.

Isik-Dikmelik, A.(2009)Targe ng systems for safety nets inNepal: a review.South Asia Development Department,Social Protec on Unit, World Bank, Washington DC.

Jha, C., S. Prasai, M. Hobley and L. Benne (2009)Ci zenmobilisa on in Nepal; building on Nepals tradionof social mobilisa on to make local governance moreinclusive and accountable.Report funded by World Bankand Swiss Development Coopera on.

Kidd, S. W. (1999)Love and hate among the people withoutthings: the social and economic rela ons of the Enxetpeople of Paraguay.Doctoral disserta on. University ofSt. Andrews.

Kidd, S. W., and E. Whitehouse(2009)Pensions and old agepoverty in closing the coverage gap: the role of social

pensions and other re rement income transfers.WorldBank; Washington DC.

Kidd, S. W., and E. Wylde(forthcoming) Just like winning thelo ery: an assessment of proxy means test targe ngmethodology. Dra report for AusAID, Australia.

Makiwane, M., and E. Udjo(2006) Is the child support grantassociated with an increase in teenage fer lity in SouthAfrica?Evidence from na onal surveys and administra vedata. Pretoria: Human Sciences Research Council, 2006.


36/40

MCDSS/GTZ(2007)Final evalua on report: Kalomo social cashtransfer scheme.Ministry of Social Development andCommunity Services, Zambia.

Mgemezulu, O. P.(2008)The social impact of community basedtarge ng mechanisms for safety nets. A qualita ve studyof the targeted agricultural input subsidy programme inMalawi.Masters disserta on. University of KwaZulu-Natal,Durban.

Miller, C., M. Tsoka and C. Reichert(2008) Impact evalua on

report: external evalua on of the Mchinji social cashtransfer pilot. Unpublished manuscript, Boston University

School of Public Health.

Miller, C., M. Tsoka and C. Reichert(2008) Targe ng report:external evalua on of the Mchinji social cash transferpilot. Unpublished manuscript, Boston University School

of Public Health.

Mkandawire, T.(2005)Targe ng and universalism in povertyreduc on.Social policy and development programmepaper, Number 23, December United Na ons ResearchIns tute for Social Development.

MLD(2010)Child protec on grant opera onal procedure 2066.Ministry of Local Development, Nepal.

Moene, K.O. and M. Wallerstein(2001) Targe ng and thepoli cal support for welfare spending. In Economics ofGovernance,2 (1):324.

Morris, S., P. Olinto, R. Flores, E. A. F. Nilson and A. C. Figueiro(2004) Condi onal cash transfers are associated with asmall reduc on in the weight gain of preschool children inNortheast Brazil. InJournal of Nutri on134: 2336-41.

Narayan, A., and N. Yoshida(2005) Proxy means test for

targe ng welfare benets in Sri Lanka. Working paper33258, South Asia Region, World Bank, Washington D.C.

Pritche , L. (2005)A lecture on the poli cal economy of targetedsafety nets.Social protec on discussion paper series no.0501, Social Protec on Advisory Service, The World Bank,Washington DC.

Rawls, J.(1971)A theory of jus ce. Cambridge, Mass.: HarvardUniversity Press.


37/40

Samson, M., K. Mac Quene, I. van Niekerk, S. Kaniki, K.Kallmann, M. Williams(2007)Review of targe ngmechanisms, means tests and values for South Africas

social grants.Economic Policy Research Ins tute, CapeTown, South Africa.

Seaman, J., C. Pe y and P. Kambewa(2008) The impact onhousehold income and welfare of the pilot social cashtransfer and agricultural input subsidy programmes inMlomba TA, Machinga district, Malawi. Unpublished

manuscript, Malawi Vulnerability Assessment Commi ee.

Sen, A.(1995) The poli cal economy of targe ng. In van de

Wall, D., and Nead, K. (eds.)Public spending and the poor:theory and evidence.World Bank. Bal more: The JohnHopkins University Press.

Sepulvedra, M.(2009)Promo on and protec on of all humanrights, civil, poli cal, economic, social and cultural rights,including the right to development.Report to the UNGeneral Assembly, prepared by the independent expert onhuman rights and extreme poverty. Human Rights Council,eleventh session, agenda item 3, March 27th 2009

Skouas, E., B. Davis and S. de la Vega(1999)Targe ng the poor

in Mexico: an evalua on of the selec on of householdsinto Progresa.Interna onal Food Policy Research Ins tute,Washington, D.C.

St. John, S. and L. Willmore (2001) Two legs are be er thanthree: New Zealand as a model for old age pensions. In

World Development29(3):1291-1305.

Stecklov, G., P. Winters, J . Todd and F. Regalia(2006)Demographic externali es from poverty programsin developing countries: experimental evidence fromLa n America. Working paper 2006-1, Department ofEconomics, American University, Washington DC.

Subbarao, K., C. Graham, K. Ezemenari, J. Braithwaite, A.Bonnerjee, S. Carvalho and A. Thompson(1997)Safetynet programs and poverty reduc on: lessons from cross-country experience.The World Bank, Washington DC.

Veras, F., R. Peres and R. Guerreiro(2007) Evalua ng theimpact of Brazils Bolsa Famlia: cash transfer programmes

in compara ve perspec ve. IPC evalua on note, 1:Interna onal Poverty Centre, Brasilia.


38/40

Ward, P., A. Hurrell, A. Visram, N. Riemenschneider, L.Pellerano, C. OBrien, I. MacAuslan, and J. Willis(2009).Cash transfer programme for orphans and vulnerablechildren(CT-OVC), Kenya opera onal and impactevalua on, 20072009, nal report, prepared by OxfordPolicy Management.

Watkins, B.(2008)Alterna ve methods for targe ng socialassistance to highly vulnerable groups.Independentmonitoring and evalua on study. Prepared for Zambiasocial assistance technical working group.

Willmore, L.(2006) Universal age pensions in developingcountries: the example of Mauri us. In Interna onalSocial Security Review59:4 (October): 67-89.

World Bank(1990)World development report: poverty.A WorldBank Publica on, Oxford University Press.

World Bank(1994)Aver ng the old age crisis: policies to protectthe old and promote growth,New York: Oxford UniversityPress.

World Bank(2005)Egypt toward a more e ec ve social policy:subsides and social safety net.Social and EconomicDevelopment Group - Middle East and North AfricaRegion, The World Bank, Washington. D.C.

World Bank(2006)Making the new Indonesia work for the poor.The World Bank, Washington DC.

World Bank(2009a) Implementa on comple on and resultsreport on a loan in the amount of US$40 million to theGovernment of Jamaica for a social safety net project.Report No: ICR00001226, September 17, 2009. La nAmerica and the Caribbean Regional O ce, The World

Bank, Washington. D.C.

World Bank(2009b)Social transfers in Bosnia and Herzegovina:moving towards a more sustainable and be er targetedsafety net.The World Bank, Washington. D.C.

World Bank(2009c) Project appraisal document on a proposedcredit to the Islamic Republic of Pakistan for a social safetynet technical assistance project. Report No: 47288-PK,

May 8, 2009. World Bank, Washington, D.C.


39/40


40/40

assessing targeting options for nepal’s social grants what does the evidence tell us dfid nepal

Documents