G.R. Chintala, NABARD, Bangladesh, Partnerships that Build Bridges to New Fro...
Nathanael Goldberg, A Deeper Look at Programs that Work with the Ultra-Poor: From Safety Net Programs to Other Innovations
1. Ultra Poor Graduation Pilots:
Spanning the gap between charity &
microfinance
Nathanael Goldberg, Policy Director
Valladolid, November 14, 2011
2. Ultra Poor Graduation Model
Program designed by BRAC
to target ultra poor
households
Holistic approach: support
household needs for savings,
financial education, skills
training, healthcare
Targeting
Use PWR to identify ultra
poor households
4. Why not just microfinance?
Substantial numbers of clients do not take up traditional
products
COUNTRY FIELD PARTNER PRODUCT TAKE-UP RATE
(%)
Mexico Compartamos Credit 4
Peru Arariwa Credit 7.9
South Africa Credit Indemnity Credit 8.7
Ghana Opportunity International Credit 1.8
Philippines Green Bank Insurance 27
Philippines Green Bank Insurance 46
Kenya K-Rep Development Agency Savings 55
Philippines First Isabela Cooperative Bank Savings 2.5
Philippines First Valley Bank Savings 23
Philippines Green Bank Savings 28
5. Who are the Ultra Poor:
Consumption
In past 12 months, did HH members get enough food?
Large share of expenses
spent on food and fuel: Yemen
No
Pakistan: 54% of total 18.6 %
budget
India: 72% Some Months
Peru: 70% 45 %
Yemen: 69% of stipend
Majority have at least 2 All Year
Not Enough
Some Months
I Don't Know
meals a day
India: 62% Pakistan No
11% Yes
Pakistan: 75% 33%
Yemen: 87.5%
Ethiopia: 83% Some
But, seasonality Months
56%
6. Who are the Ultra Poor:
Income-Generating Activities
Pakistan
Type of activities:
ACTIVITY OF THE HEAD OF HOUSEHOLD Percent
India:
Farming Land 2.15
12% report agricultural labor tending animals 1.43
as main source of income Selfemployed in business 3.06
Fishing 2.99
1% cultivate land owned by Housework 6.05
Agricultural labor 6.84
household Daily labor fishing 15.49
Pakistan: Daily Labor Other
Formal Employment Salaried
40.04
3.45
Does not work 13.15
Avg 1.4 activities Other 5.34
13% no activities Total 100
Main activity is daily labor in
agriculture or fishing
Only 2% cultivate land they
own
7. Who are the Ultra Poor:
School Enrollment
Percent Enrolled
0.99
1 0.94
0.76 0.79
0.8
0.6
0.4
0.4 0.33
0.2
0
India Pakistan Peru (6-12) Peru (13-16) Honduras (6- Honduras
12) (13-16)
8. Who are the Ultra Poor:
Health
Shocks
India: 29% of selected HH
experience 1 or more health
shock in past year
Pakistan: Medical expenses
12% of HH budget
Access to care
More than 1 hour to nearest
health center
Peru: 24-30%
Honduras: 28-58%
9. Replication at 10 Sites
Replication sponsored by CGAP and Ford Foundation: Haiti,
Honduras, Peru, Ghana, Ethiopia, Yemen, Pakistan, India(3)
Randomized evaluation at 8 sites
10. Why Evaluate?
What is the impact of the
Graduation model on the
ultra poor?
Impact evaluation measures:
How have the lives of clients
changed compared to how
they would have changed in the
absence of the program
Note this is different from
“How have their lives changed”
11. Impact: What is it?
Intervention
Primary Outcome
Impact
Time
12. Why do we need a control group?
What does it mean when your clients face:
Macroeconomic changes (+/-) My program is good!
Agricultural shocks (+/-) My program is bad!
Price shocks (-) My program is bad!
New road (+) My program is good!
Government subsidies (+) My program is good!
The control group gives us the counterfactual
A randomized control group ensures we compare
similar types of people
13. Study Design
Collect data on eligible households
Baseline Some sites: also collect data from non-ultra
poor households
Consumption Many sites have quarterly surveys to monitor
surveys changes
Survey re-applied to baseline households
Endline after 2 years
Follow-up 1 year after Endline- track more long-term
impacts
survey
15. Cost per client
Cost Includes:
Country Cost/client (US$)
Consumption support
India $330-$650
Asset transfer
Haiti $1,900
Staff costs
Head office overhead
Proportion of cost spent on each program component varies
by country
Which parts of the program create the most impact?
Need evaluations to separate contributions of individual components
16. Ghana Ultra Poor Pilot Design
241 Communities (3,981 households)
Asset Control:
Core Module: Savings Only:
Only:
76
78 communities 77 communities
10 comm. comm.
control asset only Pure
Regular Regular
Matched HH HH control
Susu UP only Susu
control HH Savings HH
Savings HH Savings
HH
& UP HH HH
Randomize communities, then households
Measure spillovers of program by comparing ‘pure’ control
households with control households in treatment communities
Next year: randomize employment program
17. Initial Results: Bandhan
Increase in food
consumption
mean difference of Rs. 64
per person per month
represents 15% of control
group mean
greater than 25% for
excluding those who
refused participation
little indication of increase
in non-food consumption
18. Initial Results: Bandhan
Increased Food Security
less likely to skip or reduce meals, especially among
adults
Health
Increase in health knowledge (hand washing, etc.)
Decreased emotional stress and increased life
satisfaction
Little impact on physical health
19. Initial Results: SKS
No significant increases in
consumption
Shift in income: agriculture
livestock
8% less likely to receive
“below the poverty line”
rations
No impact on children in school
Decreased amounts of
outstanding loans by 84% of
initial average amount
Less likely to borrow from
money lender
20. Initial Results: SKS
More likely to be saving,
impact on amount is
uncertain
Health
Reported being sick less: lost
2 fewer work days (in month
prior to interview) to sickness
(which is 56% of baseline
number of days lost)
12% less likely to have gone
to doctor/hospital in past
year
21. Take-away points
Integrated approach: Ultra Poor Graduation Model
takes holistic approach to addressing needs of
households in extreme poverty
Evaluation is essential: Does the graduation program
help ultra poor to generate income, build up assets,
increase consumption? RCTs answer these questions.
Early impacts mixed but promising: Initial results
tentatively suggest model is working, but long-term
results will be key
Improving the model: Which program components are
most useful?
Ghana evaluation will explore this question by randomizing
program components
Suggest most cost-effective way to scale program