This document summarizes research on Egypt's Takaful and Karama social safety net program, which uses a proxy means test (PMT) to target benefits. The summary is:
1) PMT targeting in Egypt resulted in 45% of benefits going to the poorest 20% of households, though exclusion and inclusion errors were still significant issues.
2) Factors like self-selection, geographic targeting, and household verification helped improve targeting accuracy, while lack of urban data and unclear criteria hurt accuracy.
3) Beneficiaries generally viewed targeting as fair, while non-beneficiaries had more mixed views and perceived more exclusion errors, highlighting challenges communicating PMT processes.
2024: The FAR, Federal Acquisition Regulations - Part 24
Targeting social safety nets in Egypt using proxy means tests and evidence from Takaful and Karama program
1. “Targeting Social Safety Nets Using Proxy
Means Tests: Evidence from Egypt’s Takaful
and Karama Program”
Sikandra Kurdi, Clemens Breisinger, Hagar ElDidi, Hoda El
Enbaby, Dan O. Gilligan, and Naureen Karachiwalla
2. Why PMTs?
• Proxy Means Test:
• acceptance to a program is based on applying a formula for predicting
poverty based on household characteristics that are easy to verify
(usually asset ownership, education levels, and housing
characteristics)
• formula is developed earlier using detailed household survey from a
sample of households
• Latin American CCT programs starting in 1990s using PMT
targeting were widely used as models
• Almost all large CCT programs in 2000s use PMT as a tool for
targeting
3. Evaluations of PMT Targeting Outcomes
• PMT targeting is imperfect, but performs better than the alternatives on
average
• Simulation based papers show that high inclusion and exclusion errors are
inescapable, at least 30% of non-poor are included even assuming perfect
implementation and reducing exclusion rates by focusing program on the
poorest can increases inclusion error above 50%
• Brown, Ravallion, and van de Walle (2015)
• Kidd and Wylde (2011)
• Cross-county evaluations shows wide variety in targeting outcomes:
countries with better implementation capacity, higher rates of inequality,
and more detailed data on household characteristics do better
• Devereaux et al. (2015)
• Coady, Grosh, and Hoddinott (2004)
4. Disadvantages of PMT Targeting
• High administrative costs compared to geographic or category
based targeting
• Data management system
• Household verification
• Hard to explain or justify to the public compared with simpler
targeting schemes
• Kidd and Wylde (2011) qualitative studies in Latin America show that
poor households perceive PMT targeting as highly random
• Alatas et al. (2012) experimental comparisons in Indonesia finds
community based targeting is seen as more legitimate even if it
objectively performs worse than PMT and hybrid model is preferred
5. Egypt’s Takaful and Karama Program
• Egypt’s Takaful and Karama part of wave of CCT programs in
Africa
• Takaful is much larger sub-program focused on families,
monthly cash transfer of 325 EGP plus 60-100 EGP per child
• Context of macroeconomics reform and reductions in poorly
targeted universal subsidies
7. PMT Formula and Threshold
• Formula based on 85 variables including housing
characteristics, employment, ages, and education of household
head and members, and 17 household assets
• Threshold selected based on targeting the lowest 40% of the
population (relatively inclusive)
• Size of program can only accommodate 9% of eligible
households (those with children under 18)
8. Targeting Outcomes in Egypt
Poorest 20% 20-40% 40-60% 60-80% Richest 20% Total
Share of
Households That
Applied to Takaful 0.50 0.42 0.33 0.30 0.17 0.35
Acceptance Rate
of Applicants 0.41 0.23 0.22 0.18 0.13 0.27
Share of
Households
Receiving Takaful
20% 10% 7% 6% 2% 9%
Share of Takaful
Beneficiaries in
this Quintile
45% 22% 16% 12% 5% 100%
9. Factors that Increase Targeting Accuracy
• Self-selection and geographic roll-out (poorer households more
likely to apply)
• Based on acceptance rate alone 55% of households would be in lowest
40% while actually 67% in lowest 40%
• Household level verification
• Inclusion error rates were much higher for households that applied
early in the program before verification was implemented
10. Factors that Decrease Accuracy
• Exclusion factors
• Among households in the first quintile, 17 percent of households would
not have been eligible for Takaful due to these exclusion factors (some
households applied and were rejected while others may have decided
not to apply knowing that they would not qualify)
• Urban households less likely to be accepted conditional on
applying than rural households
• PMT formula is slightly different for urban and rural areas, but does not
appear to be as accurate for urban households
11. Perceptions of PMT Targeting Outcomes
• Ultra-poor beneficiaries were the most likely to perceive the
targeting process as fair or very fair, while nonbeneficiaries
generally, and particularly nonbeneficiaries near the threshold,
tended to see less fairness in the selection process
• Concerns about inclusion errors especially common in areas
where lots of men migrate for work
• Male focus groups saw more exclusion error and female focus
groups saw more inclusion error
• Higher exclusion error perceived in urban communities
12. Perceptions of the Targeting Process
• General acceptance of need for targeting mechanism and checks
• Widespread awareness of the exclusion factors but no awareness of what
else goes into the decision
• Confusion caused by cards being stopped without explanation or not
getting any response to the application
“The local MoSS employees do not go out to see the people’s living
conditions. There is personal preference and laziness involved.”
• Concerns about favoritism amplified by lack of clarity about selection
procedure
“We didn’t know the acceptance criteria until they filtered people out and did
the checks and their transfers stopped. Nothing was clear, and everyone
applied anyway. The papers were clear, but not the criteria.”
13. Lessons to Take Away
• PMT targeting is not a magic bullet, there will still be targeting
errors
• More data on applicants and household verification are costly
but are improve targeting
• Clear communication is challenging with PMTs but essential
especially in contexts where overall trust in government is low