Using evidence from Ghana's LEAP 1000 program, Transfer Project's Richard de Groot explores whether cash transfers targeted to children in the first 1,000 days of life can improve their nutritional status.
Presented as part of EPRC's What Works for Africa’s Poorest Children conference in Kampala, Uganda in September 2018.
1. unite for
children
Can cash transfers targeted to children in the
first 1,000 days of life improve nutritional
status? The impact of Ghana LEAP 1000 on
young child nutrition and its determinants
Richard de Groot, UNICEF Office of Research – Innocenti
Jennifer Yablonski, UNICEF Ghana
On Behalf of the LEAP 1000 Evaluation Team
What Works for Africa’s Poorest Children?
International Conference, 10-12 September 2018
Kampala, Uganda
2. 2
LEAP 1000 Evaluation Team
UNICEF Office of Research – Innocenti: Tia Palermo (co-Principal
Investigator), Richard de Groot, Elsa Valli;
Institute of Statistical, Social and Economic Research (ISSER), University
of Ghana: Isaac Osei-Akoto (co-Principal Investigator), Clement Adamba,
Joseph K. Darko, Robert Darko Osei, Francis Dompae and Nana Yaw;
Carolina Population Center, University of North Carolina at Chapel Hill:
Clare Barrington (co-Principal Investigator), Gustavo Angeles, Sudhanshu
Handa (co-Principal Investigator), Frank Otchere, Marlous de Miliano;
Navrongo Health Research Centre (NHRC): Akalpa J. Akaligaung (co-
Principal Investigator) and Raymond Aborigo.
3. 3
Background
• Over 155 million children under 5 are stunted around
the world, negatively affecting their development
• In Ghana, 19% of children under 5 are stunted and
levels of stunting are higher for children in rural areas,
from poorly educated mothers, and living in poor
households.
• The Northern region, one of the regions under study in
this paper, shows the highest prevalence of stunting with
a rate of 33%.
4. 4
Background
• What policies can help to alleviate the burden of
undernutrition?
• Social protection, in the form of cash transfers, has been
identified as a potential nutrition-sensitive intervention
• BUT, evidence to date shows inconclusive evidence of
a positive impact on child nutritional status and
pathways of impact are not clearly understood
5. 5
Contribution
1. Evaluate a CT programme that specifically targets
pregnant women & children <1 - critical window of
child growth and development.
2. Study not only the ultimate outcome (malnutrition), but
also the various underlying determinants of child
malnutrition to contextualize the findings
3. Examine the role of moderating factors (exposure
during first 1,000 days, access to health facilities,
shocks)
7. 7
The Ghana LEAP programme
• Ghana Livelihood Empowerment Against Poverty (LEAP)
– national programme reaching 213,000 HHs
• Pilot in two regions in Northern Ghana, app. 6,000
households with infants/pregnant woman
• Selection based on demographic eligibility and proxy
means test (PMT)
• GH₵38 – 53 (USD 9.50 – 13.25) per month plus free
enrolment in health insurance
• Approximately 14% of baseline consumption expenditures
8. 8
Methods: Evaluation Design and sample
2-year mixed method, quasi-
experimental, longitudinal study
8,058 households targeted by
government and 3,619 deemed eligible
PMT scores range: 6.1 – 8.7
Evaluation aimed to include 1,250
households + 10% on either side
of PMT cut-off: 7.0 – 7.3
Baseline (Jul-Sept 2015), Endline
(Jul-Sept 2017)
Final evaluation sample N=2,497
households (1,262 T and 1,235 C)
Final analysis sample: N=4,926
children 6 – 59 months
Districts: Yendi, Karaga, East
Mamprusi, Bongo Garu Tempane
9. 9
Indicators and indices
Outcome/
determinant
Indicators
Outcomes
Malnutrition HAZ, WAZ, WHZ, stunted, wasted, underweight
Immediate determinants
Food intake Infant and young child feeding (IYCF) index, breastfeeding,
diet diversity and meal frequency
Health Diarrhea, fever and acute respiratory infections (ARI)
Underlying determinants
Household food
security
Household food expenditure, household food security
(HFIAS), household diet diversity
Care for mothers Women’s agency, subjective health, stress, social support,
nutritional knowledge
Household health
environment
Source of water, sanitation facility, hand washing facility,
floor material
10. 10
Baseline situation and validity checks
• 31% stunted, 15% wasted and 20% underweight
• Poor dietary intake
• High levels of food insecurity, poor health environment
and low levels of care for mothers
• Strong balance on key indicators (except health)
• No differential attrition in T & C groups, but some
selective attrition
• No manipulation of eligibility status
11. 11
Impact analysis methodology
• Difference-in-difference (DD)
• Additional baseline controls: age and sex of the child,
household PMT score, household size, sex of the
household head, age of the household head, and
education level of the household head
• District fixed effects
• Estimate a local average treatment effect (LATE)
14. 14
Further details
• Immediate determinants
• Negative impact on meal frequency (-0.12** SD) and fever (-0.10*
SD)
• Underlying determinants
• Positive impact on food expenditures (0.15** SD) and diet
diversity (0.14** SD)
• For care, positive impact on social support (0.16** SD)
15. 15
Next, examine heterogeneous effects
• Does length of exposure during the first 1,000 days
matter?
• Does the impact vary by distance to and quality of the
nearest health facility?
• Does the experience of covariate shocks affect the impact
of LEAP 1000?
16. 16
Length of exposure
1. Estimate treatment effects for subgroups of 6 months.
• Results: no consistent impacts on nutritional status for younger
age groups. Some positive impacts on underlying determinants
(food security and care) for 6-11 months
2. Using endline data, interact months of exposure with
treatment
• Results: Interaction effect not significant for all 3 nutritional
indices
19. 19
Importance of health facilities
• Use a triple difference model (DDD) with five indicators of
distance and quality:
• Distance to nearest health facility in km
• Health facility within 5 kilometers (dummy)
• Doctor present at nearest health facility (dummy)
• Health facility within 5 kilometers and doctor present (dummy)
• Health facility quality index (factor analysis)
20. 20
Results for health facilities
• No differential effect for any of the nutritional outcomes
• Treatment effect on food security is between 0.34 – 0.37
SD larger for children who live nearest to a health facility
where a doctor is present and in case the health facility
with the doctor is within five kilometers.
• The effect on the care index is significantly larger for
children living within five kilometres of a health facility by
0.26 SD
21. 21
Shocks
• Use covariate shocks collected at the community level:
• Negative shock
• Drought in 2017
• Flood in 2017
• Crop disease/pest in 2017
• Livestock disease in 2017
• Interruption water supply in 2017
• Sharp change in prices in 2017
• Positive shock
• New road/transportation in 2017
• Development programme in 2017
• Estimate same DDD model
22. 22
Results for shocks
• The LEAP 1000 treatment effect on WHZ (-0.42 SD) and
WAZ (-0.24 SD) is significantly lower for children in
households living in communities that experienced crop
diseases or pests
• The treatment effect is larger on the health index (0.50
SD) and food security (0.38 SD) in communities that have
additional development programmes
23. 23
Discussion – what about other CTs?
Zambia CGP Malawi SCTP Zimbabwe
HSCT
Ethiopia
Tigray
SCTPP
Kenya HSNP
Impacts
HAZ No No No No No (stunting)
WHZ No No No No No (wasting)
WAZ No No No n/a No
(underweight)
IYCF Yes (meal
frequency)
Yes (meal
frequency)
n/a Yes (children
< 12)
n/a
Health No No Negative
impact
n/a No (children <
18)
Food
security
Yes Yes Yes (diet
diversity)
Yes Yes
Care No Yes (stress) n/a No (health
and stress)
n/a
Health
environment
Yes (toilet and
cement floors)
n/a n/a No (housing
quality)
n/a
24. 24
Discussion
• Transfer size is relatively low (14%)
• Impacts might take longer to occur (e.g. CCT in Indonesia
after 6 years)
• Links to additional programming need to be considered
(e.g. Bangladesh TMRI), but highly context-specific to key
drivers of malnutrition (behavior change communication
may not be priority need; e.g. sanitation, health service
quality/access)
25. 25
Conclusion
• Child malnutrition is a complex process with multiple
determinants
• LEAP1000 had a strong impact on household food
consumption, modest impacts on care for women and no
impact on health environment
• No impact on child health and food intake and no impact on
child malnutrition
• Heterogeneity analysis: no differential impacts by age group;
quality of health facility important for food security and care;
and worse outcomes in case of shocks and better health and
food security with other development programmes
26. 26
Conclusion
• Future research should focus on effective and context-
specific linkages between cash transfer programmes
and additional social services, that together are able to
improve underlying and immediate determinants of
malnutrition.
• In Ghana, additional linkages with the Ghana Health
Services, apart from the NHIS, are currently being
explored
27. 27
Acknowledgements
We are grateful for the support of the Government of Ghana for the implementation
of this evaluation, in particular William Niyuni, Mawutor Ablo and Richard Adjetey
from the Ministry of Gender, Children and Social Protection. In addition, the
UNICEF Ghana team was instrumental to the success of this report: Sara
Abdoulayi, Luigi Peter Ragno, Jennifer Yablonski, Sarah Hague, Maxwell Yiryele
Kuunyem, Tayllor Spadafora, Christiana Gbedemah and Jonathan Nasonaa
Zakaria.
We would also like to acknowledge the hard-working field teams of ISSER and
NHRC, who conducted the data collection for this study to the highest standards.
Funding for the evaluation was generously provided by the United States Agency
for International Development (USAID) and the Canadian International
Development Agency (CIDA). Additional funding to include intimate partner violence
modules in the evaluation and to produce this paper was received from an
Anonymous donor and the American World Jewish Services by the UNICEF Office
of Research—Innocenti via the US Fund for UNICEF. We thank Laura Meucci and
Michelle Kate Godwin for grant administrative support.
All indicators are standardized against the control group
This will allows us to compare effect sizes across determinant groups and other studies
Results on food security for children 6-11 driven by food security score (HFIAS)
Results for care again driven by social support
The main focus of this graph is the see whether the slopes are different. They are not
Results driven by food consumption expenditures and food security score (HFIAS)
Results for care again driven by social support
But no differential effects on immediate or underlying determinants for crop diseases/pests
For additional development programmes, results for health are driven by ARI and diarrhea. Results for food security driven by diet diversity