Social Protection Design Issues

Targeting and Conditionality

Franziska Gassmann

Bangkok, June 2012

Outline

• Targeting• Rationale• Methods• Costs versus benefits• Evidence

• Conditionality• Definition• Rationale• Evidence

Targeting…

• …is concerned with the allocation and distribution of resources and services (Spicker, 2005)

• …will be an issue wherever a subsidy/transfer is provided to one group and financed by others (WB, 2000)

• …is a means of increasing program efficiency by increasing the benefit that the poor can get within a fixed program budget (Coady et.al. 2004)

Why targeting?

• The demand for safety nets or transfers may be indefinite, though resources are usually limited• Implies a need for targeting

• Although universal access may be the goal (e.g. health services, education), some targeting may be needed to• Channel limited public resources where financing is a

public/private mixE.g. subsidies for health insurance, fee waivers for

education• Equalize quality of services or ensure better quality

services for the poor, demand-constrained households

• Gain political support

Social transfers for everybody or only a few?

• Objective of social protection• To raise and/or protect the living standard of the

poor with public spending

• Objective of policy maker• design a policy that is effective and efficient

Effective: achieves the objective, reaches the target group

Efficient: there is no alternative policy that would have the same result at lower cost

How to identify ‘the few’?

Example: Burkina Faso

a. High poverty rate with apparently rather equal distribution among the poor

b. The poorest 10% have a consumption level of less than half of the other poor. This group is most in need of a transfer.

7

Source : Grosh et al., 2008

Targeting Methods

• Categorical/group• Geographic• Demographic (children, elderly, other)

• Individual assessments• Means test (verified, not verified)• Proxy means test• Community-based

• Self-selection• By purchase of commodity• Work requirement

• Broad targeting• Types of public spending

Means tests

• Income and/or asset-based• Eligibility is assessed individually or per household or family• Verified means-test:

• Information is verified against independent sources

• Simple means-test• No independent verification

• Examples: industrialized countries, most Eastern Europe, Kyrgyz Republic, Mozambique, India, Sri Lanka

Individual assessment

Proxy means tests

• Eligibility is assessed per household or family• Test is based on multiple indicators at household level • More easily observable than income

• Score determines whether household is eligible• Needs:

• Detailed and regular analysis of household level data (e.g. HBS)

• Poverty needs to be clearly correlated with specific household characteristics

• High capacity needs• Reduces error of inclusion• Examples: Armenia, Chile, Mexico, Colombia, Indonesia

Individual assessment

Community-based targeting

• Eligibility assessed individually or per household or family• Presumption that local authorities are better able to identify the

poor because they have more accurate information• Use of existing structures or newly established ones• With or without central directives• Little evidence on effectiveness• Possible drawbacks:

• Overburden local authorities• Capturing by local elites, reinforcing existing differences• Conflict over resources

• Example: Albania, Uzbekistan, Tajikistan, Zambia

Individual assessment

Categorical/group targeting

• Eligibility is assessed per individual, household or family characteristics that are• Easy to observe• Hard to manipulate• Correlated with poverty

• Age, gender, disability, work status, …• Lower administrative needs

• Examples: family benefits, social pensions; many pilot programs start in geographically selected target districts

Categorical/group

Categorical targeting using demographic indicators

13

Example: targeting children < 5 years

Categorical targeting based on geographical indicators

Poverty map of extreme poverty in Panama (source : poverty map 2003, MEF)

14

Self-targeting

• Universal coverage, but design discourages the non-poor to apply• Hard physical labor• Very low wage• Quality of good (e.g. broken rice)

• Often achieved by imposing a cost to participation• Work requirement• Having long queues• Attaching a stigma

• Example: Argentina, Bolivia, India

Self-selection

Summarizing…

• Frequently a combination of targeting methods – hybrid forms

• The alternative to narrow targeting • Universal access to a benefit• Broad targeting

Broad targeting

• Principle: some categories of public spending matter more to the poor than to other categories• Social services• Primary education• Health care centers

• The case for broad targeting:• Better health, education, access to clean water or

basic physical infrastructure raise well-being of the poor higher productivity, higher income

• Can be expensive, lot of leakage

Subsidies per Capita, IndonesiaSource: van de Walle, 1998

Broad targeting

Costs versus benefits

• Is targeting an effective and efficient policy instrument to reduce poverty?

400350

150 200

300250500

700

1100

1700

0

200

400

600

800

1000

1200

1400

1600

1800

1st 2nd 3rd 4th 5th 6th 7th 8th 9th 10th

US$/

year

Source : Devereux

Initial welfare distribution

150 200 250 300 350 400500

700

1700

1100

70

70

70

70

7070

7070

7070

0

200

400

600

800

1000

1200

1400

1600

1800

1st 2nd 3rd 4th 5th 6th 7th 8th 9th 10th

US$/

yea

rCase 1 – Universal Transfer

Source : Devereux

150 200 250 300 350

400

1700

1100

700

500

15

65115165215

0

200

400

600

800

1000

1200

1400

1600

1800

1st 2nd 3rd 4th 5th 6th 7th 8th 9th 10th

US$/

yea

rSource : Devereux

Case 2 – Perfect targeting

From a purely theoretical perspective…

• …targeting the poorest is the most effective and efficient option in reducing the poverty gap

• …targeting achieves the biggest reduction of the poverty gap taking into account budget constraints

Is this plausible?

• Targeting comes at a cost• Human rights• Stigmatization of the poor• Political economy

Benefits of targeting versus its costs

Targeting comes at a cost

Administrative Private Social Political Disincentive Leakage

Means test High High High High Low

Proxy MT High Low Low Low Low Low

Community Low Low High Low Low

Categorical Low Low Low Low Low High

Self-targeting Low Low Low Low Low High

Appropriate when…

Means test High share of verifiable income, large benefits, high admin capacity

Proxy MT High admin capacity, good data, large benefits or multiple programs

Community Low admin capacity, strong community structure, low benefit

Categorical Low admin capacity, strong correlation with poverty

Self-targeting Crisis situation, chronic poverty, low admin capacity

For example: Administrative costs

Source: Grosh, et al. 2008

Source: Grosh, et al. 2008

Easy vs. hard-to-verify income0

2040

6080

100

Sha

re (

%)

Figure 1a: Share of Easy-to and Hard-to-verify incomes: Kyrgyzstan

Easy-to-Verify Hard-to-verify

Targeting: Trade-off between costs and efficiency

• Type I errors: eligible families are excluded Exclusion error (reduced horizontal efficiency)

• Type II errors: non-eligible families are included Inclusion error (reduced vertical efficiency)

• The aim to reduce Type II errors tends to increase Type I errors

• Higher targeting costs are acceptable if they lead to sufficiently better targeting (Coady et.al. 2004)

Errors of Inclusion

Source: Grosh et.al. 2008

Potential reasons for targeting errors

• Targeting methods not appropriate given the country context and nature of poverty

• Insufficient financing limits inclusion of all the poor

• Poor do not take up benefit• Lack of information• High transaction costs• High opportunity costs• Stigma

Evidence of targeting effectiveness

Targeting Methods – Comparison

• Study by Coady, Grosh & Hoddinott (2004)• 122 programs from 48 countries• 49 cash transfers, 18 near cash, 12 food transfers• 18 food-subsidy and 5 non-food subsidy programs• 20 public work programs

Targeting Performance by Targeting

Method Source: Coady et.al., 2004

Main findings

• Targeting can work…• Mean = 1.25, top-10 between 2 and 4

• …but not always• One quarter of programs is regressive

• No clearly preferred method• Within method variation of 80%

• Weak ranking• Means, geographic, work requirement• Proxy means, community based, to children• To elderly, community bidding, commodities

• Context, program design and implementation matter

Conclusion

• Targeting is never perfect!• Trade-off between costs and benefits

• How much imperfection is tolerable?

• Targeting effectiveness depends on country context and implementation

• Are programs for the poor poor programs?• Political economy should not be underestimated

‘People underestimate how difficult it is to target the poor’ (van de Walle, 1998)

Conditionality

Conditionality – a first glance

• Definition of conditionality: • conditionality asks for some form of behavioral

compliance in exchange for the money in order to continue receiving support

beneficiaries have to meet certain conditions that are spelled out by the program

• Frequent conditions• Education : enrolment, attendance of 85%• Health: regular check-ups for children 0-5,

vaccinations, assisted birth, preventive health care measures

One term with multiple meanings

There are many CCT programs worldwide

But they differ substantially…• Exemptions

• Majority of programs, no official exemptions, at times indirect exemptions

• In Kenya & Tanzania those without access are exempt

• In Jamaica, children with disabilities are exempt

• Coverage• Non-compliance affects the

entire transfer in 40% of CCTs vs. 13% of CCTs where only a supplementary grant is affected

• Objective• 22% of all 41 CCTs reviewed

have an exclusive health or education focus

• Expected performance• In 67% of all CCTs

conditionality is output-related

• Passing grade in Cambodia; single repetition in Turkey

• Monitoring• No monitoring in Ecuador• Light monitoring in Pakistan,

Bangladesh & Cambodia• More intensive monitoring in

Brazil, Chili, Mexico & Kenya

• Enforcement• Punitive in Mexico & Jamaica• Developmental in Brazil &

Chili

Why opt for conditionality?

Argument 1 : Private efficiency

• Households under-invest in the education and health of their children from a private point of view• Parents are not always well informed about the (future)

value of education• Parents (or at least one of them) do not always behave

altruistically towards their children• Parents have self-control problem and are willing to

sacrifice future for present consumption• Parents of low-income households tend to be more risk-

averse and shy away from investments in education Conditionality could guide beneficiaries, encourage them to

invest and function as a self-control device

Argument 2: Social efficiency

• Households under-invest in the education and health of their children from a social point of view• They do not factor in positive side effects from

educating children or from preventive health care measures

• They do not consider future social costs of under-investment such as increased crime, lost labor contribution and less informed political participation

Conditionality can encourage households to increase their investments

Argument 3 : Political economy

• Social cash transfers often have difficulties in securing long-term political support due to worries about• Beneficiaries not knowing how to spend the money and

misusing this flexibility • Beneficiaries becoming dependent and not being

interested in improving their situation• Not having any control over the impact of a rather cost-

intensive intervention Conditionality can make a social cash transfer program more

attractive by controlling the use of funds, giving beneficiaries an extra nudge and guaranteeing a certain impact.

Redistribution may be better acceptable if beneficiaries ‘behave’ accordingly (condition).

Argument 4 : Empowerment

• Beneficiary households of social transfers and/or individual members are often in an inferior position• Government support is often presented as charity as

opposed to an entitlement• The use of the transfer is often decided by the

household head without any further consultationConditionality can give beneficiaries / individual

household members a negotiation device and allow them to exercise their rights

Argument 5 : Equity

• Targeting is bound to be imperfect and costly.• Households who do not qualify for the transfers

apply and benefit nevertheless (inclusion error)• Trying to reduce inclusion errors can have a

negative impact on exclusion errors and end up being costly

Conditionality can make a program unattractive to those who don’t quality and therefore serve as a screening device

Evidence ?

Evidence on behavioral changes

• Few (long-term) studies on the impact of conditionality at household level

• In the past, the good performance of a conditional cash transfer program was erroneously cited as evidence of the value added of conditionality

• The impact varies greatly from a magnifier effect of conditionality of 16 times (Mexico) to no impact at all in Malawi

Evidence on the political economy

• Very little evidence on the political effects

• A study by Taylor-Gooby (2001) confirms the rising popularity of activation / conditionality in the Western World

• Conditionality was seen as an important tool by politicians in Brazil to guarantee impact, avoid dependency; better monitoring reduced however mayor’s chances of reelection

• Mixed results for Zambia: popularity of conditionality as such but unpopularity of consequences of conditionality

Evidence on empowerment

• Very little but anecdotal evidence

• Adato et al (2002) show that the trainings women had to attend increased their confidence & awareness

• Molyneux (2006) & Bradshaw (2008) hint at disempowering effects with conditionality confining women to their female duties & suggestion that they were bad mothers before

• Zambia

• Individuals with a partner unanimously confirmed that conditionality had helped them to negotiate household expenditure with the partner

Evidence on equity

• Limited evidence• Alvarez et al (2008) find that

conditionality improved the targeting effectiveness in Mexico

• In Zambia, a monthly transfer of 10 US$ wouldn’t have discouraged the better off from participating

• In rural areas in Zambia, the better off even had a greater likelihood of participating than the poorer segments

Self-assessed Yes No

Very poor 10% 90%

Poor 13% 87%

Middle 11% 89%

Upper middle 0% 100%

Rich 0% 100%

Conditionality may lead to inefficiency

• Impact inefficiency• Conditionality in one area can negatively effect

behavior in another area• Households can perceive conditionality as lack of trust

and get demotivated• Design inefficiency

• Conditionality does not address actual/real problem• Implementation inefficiency

• Direct/indirect administrative costs• Capacity constraints• Promotion of corruption

How to decide?Source: Fiszbein & Shady, 2009

1. The rationale behind conditionality is to optimize household behavior, empower beneficiaries, increase the political attractiveness of social cash transfer programs and improve the targeting effectiveness

2. The evidence base on which conditionality rests, in particular for the empowerment, political economy and equity effects is rather thin

3. Conditionally also has the potential to introduce a number of impact, design and implementation inefficiencies that lower the cost-effectiveness of a social cash transfer program.

4. Prior to deciding for or against conditionality, it is important to analyze the potential of conditionality to effect any positive changes, the costs and inefficiencies associated with it as well as the capacity of a country to successfully manage the process

Key messages: Conditionality

Recommended readingTargeting:• Grosh, M. et.al. (2008), For Protection and Promotion. The Design and

Implementation of Social Safety Nets, The World Bank, Washington DC.: chapter 4 (Targeting) http://siteresources.worldbank.org/SAFETYNETSANDTRANSFERS/Resources/281945-1223052223982/For_Protection_and_Promotion-Ch4.pdf

• SAMSON, M., VAN NIEKERK, I. & MAC QUENE, K. 2006. Designing and Implementing Social Transfer Programmes, Cape Town, Economic Policy Research Institute.

Conditionality• Fiszbein, A. and N. Shady (2009), Conditional Cash Transfers. Reducing present and

future poverty, The World Bank, Washington DC.• de Janvry, A. and E. Sadoulet (2005). Conditional cash transfer programs for child

human capital development: Lessons derived from experience in Mexico and Brazil. GRADE 25th anniversary Conference, "Investigación, Politicas y Desarrollo". Lima, Peru. Available from: http://www.cgdev.org/doc/event%20docs/MADS/SAIS-CCT.pdf

• Standing, G. (2007). Conditional Cash Transfers: Why Targeting and Conditionalities Could Fail. One Pager No. 47. Brasilia, Brazil, International Poverty Centre. United Nations Development Programme. Available from: http://www.undp-povertycentre.org/pub/IPCOnePager47.pdf

• Schuering, E. (2010), Conditions, conditionality, conditionalities, responsibilities – finding common ground, MGSoG Working Paper, 2010WP014, Maastricht Graduate School of Governance, Maastricht. http://www.merit.unu.edu/publications/mgsog_wppdf/2010/wp2010-014.pdf