Targeting Efficiency: How well can we identify the poor

1
Targeting Efficiency How well can we identify
the poor?

• IFMRCMF Seminar
• May 5, 2008
• Jyoti Prasad Mukhopadhyay
• Abhijit Banerjee, Esther Duflo, Raghabendra
  Chattopadhyay
• and Jeremy Shapiro

2
Motivation

• Nearly all poverty alleviation programs target a
  particular sub-population
• Thus, accurate targeting is crucial to program
  success
• Evidence that targeting is often sub-optimal
• National Sample Survey Organization finds that
  18 of the wealthiest 20 of rural population
  (ranked by monthly per capita expenditure) held
  Below Poverty Line (BPL) rationing cards.
• Which targeting methods work? Which dont?

3
This Study

• Evaluates the targeting efficiency of . . .
• various government anti-poverty programs
• Targeting the Ultra Poor, operated by Bandhan (a
  Kolkata MFI)
• Participatory Rural Appraisals (PRAs)

4
Preview of Results

• Government targeting
• Does not identify the poorest of the poor
• In our sample, eligible households appear no
  worse off than ineligible households
• Bandhans targeting
• Identifies a group which is disadvantaged in some
  respects
• Own less land and fewer assets lack credit
  access
• Per capita expenditure does not appear lower
• PRA
• Generates wealth ranking of participants which
  accords with various measures of poverty

5
Overview of Bandhans Program

• Name of the program Targeting Hard-core Poor
  (THP)
• Background Benefits of microfinance do not
  accrue to the poorest of the poor (Morduch 1999,
  Rabbani et al 2006). Microfinance Institutions
  (MFIs) are in general reluctant to lend to the
  poor.
• Reasons
• ultra-poor households tend to use loan for
  meeting consumption needs
• Productive investment of loan is unlikely
• Ultra-poor households are extremely vulnerable to
  shocks and hence more prone to default
• Murdoch (1999) aptly commented, poorer
  households should be served by other
  interventions than credit.
• Objective of the program To provide income
  generating assets livestock, inventory etc. as
  grant to help ultra poor households secure a
  regular source of income. Also imparting training
  and other assistance required for starting a
  small scale enterprise so as to graduate them to
  potential microfinance clients.

6
Overview of Bandhans Program

• Area of intervention Murshidabad district, West
  Bengal
• Why Murshidabad district?
• This is one of the poorest districts of West
  Bengal.
• district level statistics
• HDI Ranking (2004) -15 (out of 17 districts)
• Targeted no. of beneficiaries 300
• To date, the identification process has occurred
• in 60 villages, with an average of 15 households
  
• identified as Ultra Poor in each village and
• 300 beneficiaries have received assets so far.

7
Key Phases of the Program

• Identification of Ultra-poor households(Potential
  beneficiaries)
• Half of the identified beneficiaries are randomly
  chosen for asset distribution
• Enterprise Selection
• Training
• Asset distribution
• Weekly Follow-up and monitoring

8
The Identification Process

• Identifying the poorest villages and hamlets of
  the district
• Conducting PRAs in the identified hamlets (Social
  mapping)
• Identifying ultra-poor households through
  wealth-ranking during PRA
• First verification of the identified ultra-poor
  households household survey
• Second verification by the THP program
  coordinator
• Final selection

9
What is Wealth Ranking?

• Wealth ranking is done after the completion of
  social mapping in PRA. It is an effective way of
  classifying households into different categories
  based on households occupation, assets, land
  holdings and general economic well being. Each
  household is given a rank in a scale of 1-6
  corresponding to each category where lower ranks
  corresponds to relatively better-off households.
  Lively discussion among villagers generate the
  most precise definition of (relative) poverty and
  facilitate accurate wealth ranking.

10
The Second and Final Verification

• The second and final verification is done by THP
  program coordinator. During verification, the
  program coordinator observes economic condition,
  educational attainment and nutritional status of
  the women and children of the households. Some
  mandatory requirements to be eligible for
  obtaining grants are as follows
• The household must have at least one active woman
  capable of undertaking some enterprise
• the household must not be associated with any MFI
  (in keeping with the aim of targeting those who
  lack credit access) or receive sufficient support
  through a government aid program
• primary source of income should be informal labor
  or begging
• land holdings below 20 decimals
• no ownership of productive assets other than land
  
• no able bodied male in the household and having
  school-aged children working rather than
  attending school

11
Data Collection

• At Bandhans request, we interviewed households
  identified as Ultra Poor by Bandhan and other
  poor households
• Conducted economic census (similar to that used
  for government targeting) in 5 villages where
  Bandhan operates Ultra Poor program
• Identified poorer population from census
• From this population, interviewed 170 random
  households not identified by Bandhan as Ultra
  Poor
• Also interviewed 92 Ultra Poor households

12
The Dataset Summary Statistics
13
Targeting Efficiency of Government Programs

• Targeting for many government programs is based
  on BPL census
• Concern that census incorrectly classifies
  households (Jalan and Murgai, 2007)
• Speculation that lists of BPL households are
  manipulated to include non-poor households
  (Mukherjee, 2005)
• To assess these concerns, we compare recipient
  and non-recipient households
• For various programs BPL and Antodaya rationing,
  Indira housing and employment generating schemes
• Compare according to expenditure measures, land
  holdings, whether members eat two meals a day,
  access to credit and an index of asset holdings

14
Government Targeting Results
15
Overview of Bandhans Targeting the Ultra Poor
Program

• Identification
• To identify the Ultra Poor, Bandhan . . .
• Conducts PRAs
• Conducts follow up survey among those ranked most
  poor (rank 5 or 6) in the PRA
• Identified households receive a final
  verification visit from the Ultra Poor Project
  coordinator

16
PRA Process

• Social Mapping
• Location of each household in hamlet demarcated
  on map
• Name of household head recorded on index card
• Wealth Ranking
• Residents define what constitutes poverty in
  their community
• Index cards sorted into piles corresponding to
  socio-economic status
• Sorting of households into ranked piles (richest
  to poorest) based on participatory discussion by
  hamlet residents
• Concern that PRA may not generate accurate
  ranking
• May not be sufficient participation (too few
  people present)
• May be that influential hamlet members dominate
  process
• Could manipulate ranking in the expectation that
  highest (poorest) ranked households will receive
  aid
• Our study assesses reliability of ranking

17
Evaluation of the PRA Process
18
Evaluation of the PRA Process contd
19
Evaluation of the PRA Process contd
20
Determinates of PRA Rank

• In addition to evaluating whether PRA wealth
  rankings accord with statistical measures of
  poverty, we assess what determines whether
  households are considered poor by their peers

21
Evaluation of Bandhans Verification Process

• Generally, PRA and Bandhans process identify
  similar sub-populations
• Those ranked poor in PRA and those identified as
  Ultra Poor by Bandhan have less land, fewer
  assets, less education and lack formal credit
  access
• What does Bandhans verification process add
  above and beyond the PRA ranking?
• How does it further narrow the targeted
  population?

22
Evaluation of Bandhans Verification Process
contd
23
Evaluation of Bandhans Verification Process
contd
24
Expenditure Puzzle
25
Expenditure Puzzle contd
26
Expenditure Puzzle contd
27
Conclusion

• Effective targeting of a particular
  sub-population depends crucially on the
  identification mechanism used
• Censuses, similar to those used for targeting of
  government aid, do not appear to identify the
  most disadvantaged population
• Peer wealth rankings gathered in PRAs can provide
  statistically reliable information about which
  households are most poor
• More detailed household interviews, as used by
  Bandhan, provide a way to further narrow the
  identified population and target more precisely