On the Always Vexing Question of Targeting: How are LAC CCTs doing? International Symposium: the Contribution of CCTs to the Creation of Rights-Based Social Protection Systems Mexico City Sept. 28-30, 2016 Margaret Grosh Practice Manager for Social Protection and Labor, World Bank
Motivations Facts Reflections 2
Why Target? To include the poor (always important) 3
Why Target? To focus expenditure on the poor (sometimes chosen) 4
Why Target? To focus expenditure on the poor (sometimes chosen) The family has five members (three children). Thepensioner s household has Single family. monthly income of GEL 20 Receives a pension (GEL The household Ranking score -has 39two 550 28) and social assistance members. (GEL 22) Thefamily household has The has three Ranking score - 47 950 monthly income of GEL members. 80 a disabled The family has Ranking 64 300 child, whoscore has a- pension The household four (GEL 28) and has social members. The(GEL household assistance 22) has monthly income of Ranking score - 155 470 1,050 GEL DATA BASE Ranking score - 665 960 5 4 3 1 2 5
The desired degree of targeting may vary by program DATABASE 5 Electricity subsidies 4 Health subsidies 3 Monetary benefits 1 2 6
Facts 7
Data on Targeting Outcomes for CCT Programs is Surprisingly Scarce World Bank ASPIRE Database classifies 64 programs as CCT. For selecting observations, we use three criteria: - Country has household survey data since 2010 which might measure participation in the programs (20 countries, of which 17 in LAC) - The survey questions reasonably clear on program (rules out Brazil, Argentina, we kept DR though a bit ambiguous) - Program still operating in similar fashion to what was observed in the survey, though scale or benefits may have evolved (rules out Guatemala) Therefore data on 17 countries for coverage and 14 for benefit incidence. 8
Coverage of the poor in CCT programs varies considerably 90.000 80.000 70.000 Share of poorest quintile and total population covered, National quintiles 60.000 50.000 40.000 30.000 20.000 10.000 0.000 World Bank ASPIRE database Poorest quintile Total Population 9
Coverage of the poor in CCT programs Is usually higher in rural areas Share of poorest quintile and total population covered, Rural quintiles Share of poorest quintile and total population covered, Urban quintiles 90 90 80 80 70 70 60 60 50 50 40 40 30 30 20 20 10 10 0 0 Quintile 1 - Rural World Bank ASPIRE database Total population Quintile 1 - Urban Total population 10
Benefit incidence is progressive in varying degrees 80 Share of benefits going to poorest and richest quintiles National quintiles 70 60 50 40 30 20 10 0 Peru Panama Mexico Philippines Chile Jamaica Colombia Ecuador Bolivia Paraguay Bangladesh Lesotho Honduras Dominican Republic Poorest Richest 11 World Bank ASPIRE database
Incidence is more progressive within urban areas than rural 80 70 Share of poorest benefits accruing to poorest and richest quintiles, Rural quintiles 80 70 Share of poorest benefits accruing to poorest and richest quintiles, Urban quintiles 60 60 50 50 40 40 30 30 20 20 10 10 0 0 Poorest - Rural Richest - Rural World Bank ASPIRE database Poorest - Urban Richest - Urban 12
Reflections 13
How good is good enough? 14
We know that targeting is NEVER perfect 15
Benchmarks: Eastern Europe s Incomes Support of Last Resort Programs Benefit Incidence is Largely Highly Progressive, but Coverage among the Poorest Decile is Low Tesliuc, Grosh, Pop, Yemtsov, 2014 16
Benefit Incidence in poorest quintile Pro-poor Benchmarks: Across all programs in ASPIRE CCTs have much more progressive incidence 45 CT CCT Social pensions Public works 40 35 30 25 20 15 10 5 0 Q1 Q2 Q3 Q4 Q5 Quintile State of Safety Nets 2015, World Bank 17
Are targeting systems dynamic enough? 18
Dynamic rosters matter because families move in and out of poverty frequently In INDONESIA, a quarter of all Indonesians are highly vulnerable to poverty as a result, more than half of the poor were not poor the year before Percentage of Population Poor or Vulnerable There are 28 million poor in Indonesia 27% 11% and a further 68 million vulnerable who live just above the poverty line Source: Susenas and World Bank calculations Note: The poor are under the national poverty line, and the vulnerable are under 1.5 times the poverty line 45 percent of this year s poor were poor the year before 2010 s poor compared to their 2009 status Newly Poor Poor Last Year while 55 percent of this year s poor were not poor the year before Source: Susenas panel 2009-10 and World Bank calculations 19
LAC shares this phenomenon of many vulnerable to falling back into poverty 20
Rosters are not very dynamic Entry is rationed by budget in many programs Applications are not always accepted continuously Re-certifications are irregular or infrequent Peru, Jamaica, Dom Republic, Honduras Irregular Every 8 years Mexico Colombia, Ecuador Ever 4-5 years Limited duration of the program Chile, Bolivia, Costa Rica Brazil Every 2 years Continuous Argentina 21
Proxy Means Tests dominate: Of countries in LAC with CCTs: Eligibility Criteria in Latin America are Focused on Chronic Poverty 12 use PMT (some combining with first step of geographical targeting); 2 use categorical targeting; 2 use verified means test; 1 does blend of geographical, PMT and categorical Should LAC move to more means testing? 22
Empirics on Hard and Easy to Verify Income Bulgaria Kyrgyz Republic Mexico Tesliuc, Grosh, Pop, Yemtsov, 2014 for Bulgaria and Kyrgyz Rep, staff calculations for Mexico and Brazil 23
Possible agendas going forward Active outreach to reduce errors of exclusion Use ID/IT to reduce errors of inclusion Of increasing importance as CCTs or their registries are used as platforms for social inclusion Increase dynamism of intake/exit Move toward more dynamic formula Would allow - Better response to disasters - Programs to serve transitory poor 24
Conclusion Targeting of CCTs may not be the thing most in need of fixing in social policy but still there are challenges and questions For some, answers exist where political and administrative resources are marshalled For others, we do not yet have very satisfactory answers So targeting will remain on the agenda