Mexico’s Oportunidades: Self-Selection in Targeted Social
Programs
César MartinelliProfessor of Economics, ITAM, Mexico City and
Wilson Center/Comexi Public Policy Scholarand
Susan W. ParkerProfessor of Economics, CIDE, Mexico City
Introduction
Conditional cash transfers (CCT) have become a major poverty alleviation strategy across Latin America Mexico, Honduras, Nicaragua, Jamaica, Colombia,
Brazil, Argentina, Ecuador
Key innovation: condition grants to investment in children’s human capital Current and future poverty reduction goals, e.g. by
improving HK of children today, expectation that future poverty will be reduced
Introduction, continues
How are beneficiaries selected, in context where income not observable? As in other social programs in
developing countries, based on information provided by applicant
Obvious incentives to misreport
Introduction, continues
According to economic theory, applicants will underreport given material incentives
Lying is consider immoral by many, perhaps most moral thinkers … need to look at the evidence
Stigma (social embarrassment) and self-deception may lead to overreporting
Objectives of paper
What is prevalence of misreporting in social programs?
Who misreports? What impact on who receives
benefits?
Objectives, continues
Previous studies limited by lack of information on misreporting
Take advantage of context where program verified economic conditions, allows us to compare reported with actual conditions
Description of Oportunidades
Large scale poverty alleviation program Begun in 1997 in rural areas, expanded to
urban areas in 2001. By 2004, 5 million beneficiary families (1/4 of all households)
Cash transfers conditional to investment in human capital Regular school attendance of children Regular clinic visits of family members Transfers given directly to mother Average monthly benefits: US$30, overall increase of
30% over monthly income
Description of Oportunidades cont.
Health and nutrition component: Basic health care package Fixed income linked to clinic attendance Nutritional supplements given to pregnant
women and children aged 0 to 2 (or up to age 5 if malnutrition perceived)
Monthly Benefits
Description of Oport. cont.
Targeting differs between rural and urban Oportunidades
Rural targeting: Geographic: poor communities identified using
community poverty index Household: ALL HH in eligible communities
applied socio-economic questionnaire, regression analysis (proxy means test) carried out to determine who was poor/eligible on basis of characteristics
Description of Oport. cont.
Urban areas: interviewing all households deemed too costly, Self-selection introduced
Urban targeting: Geographic: poor urban areas identified Program modules set up in poor areas during 2 months,
where household go to apply for benefits Households applied socio-economic questionnaire in the
module, based on answers, households declared initially eligible
Initially eligible HH receive follow up visit to verify info reported in module. Final classification made
Issues in Self-Selection
1. Who is aware of the program?2. Who applies to the program?3. How is eligibility determined?... at these three stages it is possible
to introduce mistakes (give program to who is “not poor” or deny it to who is “poor”)
We focus on stage 3 in this talk—stages 1 and 2: work in progress
Eligibility: Poverty Regression
Problem to determine who is “poor”: unreliability of income or consumption data
Methodology: poverty regression… search for observable household characteristics that are correlated with income and estimate a “poverty score”
Oportunidades poverty regression(poor: score ≥ 0.69)
Family size/number of rooms Family size Female head Number of children below 11 Head has no schooling 1 to 5 years of schooling Age head (years) No toilet Toilet w/o running water Unpaved floor No gas boiler No refrigerator No washing machine Neither car nor truck Rural area Region Constant
0.139 0.176 -0.02 0.255 0.380 0.201 0.005 0.415 0.220 0.475 0.761 0.507 0.127 0.159 0.653 From -0.657 to 0 -1.579
Selection of urban beneficiaries
Applicants to Oportunidades were asked to answer questionnaire about household characteristics
Self-report was used to determined preliminary eligibility using poverty regression
Household visits to verify answers to questionnaire and assess definitive eligibility status
Data
ENCASURB:1. Inclusion questionnaire, carried out
at the module2. Additional questionnaire after
initial eligibility determines, carried out at the module
3. Verification questionnaire, carried out at household
Characteristics of applicants
Definitions
For each good g, Ag: applicants who assert having it
Dg: applicants who deny having it
Hg: found having the good
Ng: found not having the good
Then:
A model of misreporting
wg: belief of applicant about weight of good g in poverty score
δ: applicant’s assessment of probability of verification visit
p: belief about penalty for getting caught underreporting
Ba: (monthly) benefits from program for applicant’s household
Ya: (monthly) household expenditure (as proxy for income)
Xa: education, age, gender, etc of applicant
A model of misreporting, continues
A model of misreporting, continues
Applicant underreport good g if
Applicant overreport good g if
Assumptions: Risk aversion coeff. = 1 Random terms are Gumbel distributed
(i.e. use logit regressions)
( ) ln( / )w p B Y Xg g a a g a ga 1 1 1
w B Y Xg a a g a galn( / )1 2 2
Mg. effects of program benefits
Beliefs vs. true relative weights
Mg. effect of years of education
Male rather than female reporter
Underreporting vs. final status
Conclusions
Underreporting is widespread Overreporting occurs for goods whose absence is
associated with poverty—so have weight in poverty regression!
Both under and overreporting are sensitive to program benefits: evidence of a cost of lying and a stigma consideration (even in front of strangers!)
Need to revisit poverty regression methodology More generally: room for cross-fertilization
between social policy design and behavioral economics
Reference
Deception and Misreporting in a Social Program (by Cesar Martinelli and Susan Parker) available at http://ciep.itam.mx/~martinel/