The document summarizes research comparing the impacts of providing assistance in the form of food or cash transfers in developing countries. It discusses a four-country experimental study conducted by IFPRI and WFP that directly compared the impacts of providing equivalent amounts of food or cash/vouchers in Ecuador, Niger, Uganda, and Yemen. The study found that across contexts, cash transfers generally improved food security outcomes slightly better than food transfers while costing significantly less to implement. The study provides evidence that cash transfers did not have adverse impacts and in some cases had additional beneficial impacts, though food transfers remain appropriate in some contexts. The discussion emphasizes moving beyond ideological debates to focus on objectives, context, and cost-effectiveness of different modalities.
1. Cash, Food or Vouchers? Evidence from a
Four-Country Experimental Study
February 12, 2019
Presented by Dan Gilligan
Research by Dan Gilligan, Melissa Hidrobo, John Hoddinott,
Amy Margolies, Amber Peterman, Shalini Roy, Susanna
Sandström, Benjamin Schwab, Joanna Upton
2. Motivation
What do we know about the relative merits of providing
assistance in the form of food or cash in developing countries?
Until recently, very little.
Many studies provide evidence on the impact of food transfers
or cash transfers but very few directly compare the impacts of
food and cash in the same setting.
There is also limited information on relative costs.
Often this issue is dominated by ideology and politics.
3. What are the arguments for food or cash transfers?
• Food transfers
– may be more likely to be spent on food (transaction costs)
– or on child goods (controlled by women)
– are not affected by price increases
• Cash transfers
– may be more fungible, allowing spending to occur where
needed with lower transaction costs
– may be cheaper to deliver
– less paternalistic
4. What does economic theory say about food or cash?
• Food transfers (or food vouchers) should
– reduce out-of-pocket food expenditures
– increase the value of food consumed
– increase the value of other goods consumed through
substitution
• If the amount of food transferred is inframarginal – no more
than what a household typically consumes – the effect on the
composition of spending should be equal between food and
cash (Hoynes and Schanzenbach 2009)
• But food rations are sometimes extramarginal for some food
items and there are sometimes substitution costs, as when
not all foods are purchased
5. Current evidence
• Recent literature has begun to provide rigorous empirical
evidence on the question of food versus cash
• Price effects: both food and cash transfers lead to increased
demand for (normal) goods, but food transfers also increase
supply which can depress prices
– One study found that food transfers reduced prices by 4% on
average while cash transfers led to small price increases. Price
declines were larger in more remote villages (Cunha et al 2019)
– Another study showed that cash transfers led to a large increase
in prices for perishable protein-rich foods (Filmer et al 2018)
• program beneficiaries: child nutrition
• program nonbeneficiaries : child nutrition
6. Cross-Country Study on Food and Cash Transfers
• How can we provide compelling evidence on the relative
impacts of food and cash?
– In the same setting, implement transfers keeping all aspects of the
intervention the same …
• Value of transfers
• Frequency of transfers
• Payment dates
– … except for the modality (cash, food, voucher) …
– which is randomly assigned …
– so that differences in impacts can be ascribed to modality and NOT to
other confounding factors …
– And do this in several countries
• Today I report on the results of a three year IFPRI-WFP study
that does just that
7. Cross Country Study Outline
• Describe the interventions
• Explain the evaluation design
• Describe core findings:
– Impacts
– Costs
• Note additional findings
• Summary and implications
9. Ecuador
• Unconditional transfers in urban areas in northern provinces of
Carchi and Sucumbíos with large concentration of Colombian
refugees and poor host Ecuadorians
• Beneficiaries randomized (at barrio level) to receive:
– $40 cash transfers accessed from ATMs using a debit card
– $40 in vouchers redeemable for specified foods in supermarkets
– $40 in food: rice (24 kg), vegetable oil (4ℓ), lentils (8 kg), and
canned sardines (8 cans of 0.425 kg)
• Transfers received monthly for a six month period
• All beneficiaries received nutrition sensitization
• Roughly 75% of beneficiaries were women
10. Niger
• Three months public works (all households) and three months
unconditional transfers (targeted households) in Mirriah
departement, Zinder region
• Beneficiaries randomized at worksite level to receive:
– Cash payment of 25000 FCFA ($55) per month. Cash dispensed
from mobile ATM brought into each village
– In-kind payment of 87.5kg cereals, 18kg of pulses and 3.5 kg
vegetable oil and salt. Food shipped into villages at beginning of
public works and stored in granary.
– Payments every two weeks during public works
– monthly payments for unconditional transfers
• Transfers made to household head
– approx 75% male; 25% female
11. Uganda
• Transfers linked to children’s enrollment in Early Childhood
Development centers (preschools) in three districts of Karamoja – a
poor, rugged, post-conflict sub-region in Eastern Uganda with high
seasonal food insecurity
• For each child aged 3-5 years enrolled in preschool, beneficiaries
were randomized (at preschool level) to receive:
– 25,500 UGX ($10.25) in cash: added to mobile money cards
– 25,500 UGX of food: multiple-micronutrient fortified corn soya
blend (CSB), Vitamin-A fortified oil, sugar
– No transfer (control group)
• Transfers given every 6-8 weeks for 12 months
• 6 transfer cycles, but food delivered earlier
• Transfers made preferentially to child’s mother
12. Yemen
• Unconditional transfers in rural districts of two governorates with
high baseline levels of food insecurity
• Randomization at the Food Distribution Point (FDP) level
– Each FDP is a school serving a cluster of neighboring villages
• 3 transfers each:
– Cash Transfer: 10500 YER (≈$49)
• every two months beginning in Nov 2011
– Food Transfer:
• 50kg of fortified wheat flour, 5ℓ of oil
• Aug 2011, Oct 2011 & April 2012
• Beneficiaries: 19% female headed households
14. Evaluation design: Core questions addressed in all countries
Ecuador Niger Uganda Yemen
How do benefits of
cash(vouchers) compare
to food transfers?
Y Y Y Y
Does the delivery of
cash(vouchers) cost less
than food transfers?
Y Y Y Y
Which modality do
households prefer?
Y Y Y Y
15. Additional questions addressed in some countries
How does receipt of
cash(voucher) affect
Ecuador Niger Uganda Yemen
Intra-household
decisionmaking
Y
Economic, social relations
between households
Y Y
Who is selected to receive
transfers
Y
Anemia Y Y
Child anthropometric
status
Y
ECD outcomes Y
Purchase of agricultural
inputs
Y
Expenditures on
undesirable goods
Y Y
16. Evaluation design
• In Ecuador and Uganda, the evaluation design includes
treatment groups (food, cash or voucher (Ecuador only)) and a
control group (no transfer), so we can answer two questions:
– Impact of each modality relative to no transfer and
– Impact of cash(vouchers) relative to food
• In Niger and Yemen, the intervention design consists only of
treatment groups (food or cash), so we can answer only one
question:
– Impact of cash relative to food
18. Impact on household food security: Two measures
• Food Consumption Score (FCS): WFP’s principal food security
indicator
• FCS = (# times food group consumed in last seven days)
x (weight attached to food group)
18
Group
Food items
Food group Weight
1
Maize, rice, sorghum, millet, pasta, bread , other cereals
Staples 2
Cassava, potatoes and sweet potatoes, other tubers, plantains
2 Beans, peas, lentils, groundnuts and cashew nuts Pulses 3
3 Vegetables Vegetables 1
4 Fruit Fruit 1
5 Beef, goat, poultry, pork, eggs, and fish Meat and fish 4
6 Milk, yogurt, and other dairy Dairy 4
7 Sugar, sugar products, and honey Sugars 0.5
8 Oils, fats, and butter Fats and oils 0.5
19. Impact of household food security: Two measures
• Households are considered food insecure if FCS < 35 (28 in
Yemen) and have poor food security status if FCS < 21.
– Baseline rate of food insecurity ranged from 10% (Ecuador) to
53% (Uganda)
• In Ecuador, Uganda, and Yemen we fielded food consumption
modules that recorded details of quantities consumed over
the seven days prior to the survey. This allows us to calculate a
second food security measure, household caloric acquisition
• While FCS and caloric acquisition can increase together:
– Caloric acquisition can rise without much change in FCS (eg
increased consumption of quantities of staples)
– FCS can change without much change in caloric acquisition
(eg increased consumption of vegetables)19
20. Impact of transfers relative to control households:
Change in FCS score
20
4.95
5.75
9.04
0.00
3.00
0
1
2
3
4
5
6
7
8
9
10
Ecuador food Ecuador, cash Ecuador, vouchers Uganda, food Uganda, cash
21. Impact of cash transfers (or vouchers) relative to food:
Percentage change in FCS
21
-11
-9.6
0.6
5.6
9.2
10.1
-15
-10
-5
0
5
10
15
Niger, post-
harvest season
Niger, hungry
season
Ecuador, cash Ecuador,
vouchers
Yemen Uganda
22. Impact of transfers relative to control households:
Percent change in caloric acquisition
22
16.0
6.0
11.0
1.7
19.6
0.0
5.0
10.0
15.0
20.0
25.0
Ecuador, food Ecuador, cash Ecuador,
vouchers
Uganda, food Uganda, cash
24. Costing transfer modalities
• Exclude value of transfer itself (But we return to this later)
• Focus on modality specific costs (staff time, goods, services) that are
specific to the delivery modality chosen:
– Food: Staff and monetary costs associated with in-country transport, ration
preparation and distribution
– Cash: Costs associated with contract preparation; cost of debit cards; bank
fees for administering transfers
– Vouchers: Costs associated with supermarket selection; printing vouchers;
staff costs associated with liquidating vouchers
• Common costs that are incurred in program implementation
(planning costs, targeting, sensitization, nutrition training etc) are
allocated proportionately across modalities or are excluded.
How much does it cost to make a cash transfer relative to a food transfer?
25. Dollar cost of a cash transfer relative to a food transfer
25
-8.47 -8.91
-2.96
-6.28
-8.20
-30.00
-25.00
-20.00
-15.00
-10.00
-5.00
0.00
5.00
Ecuador Niger Uganda Yemen Ecuador,
voucher
26. Costing transfer modalities: Implications
• In Niger, the monthly transfer was worth $55
– The modality specific cost of a cash transfer was $4.00
– The modality specific cost of a food transfer was $12.91
• It costs $354.00 to make six cash transfers to a beneficiary
$354.00 = 6 x ($55 + $4.00)
• It costs $407.46 to make six food transfers to a beneficiary
$407.46 = 6 x ($55 + $12.91)
• Abstracting from other costs, with a $2,000,000 budget, we
could include:
– 5,649 households if cash were given ($2,000,000 / $354.00)
– 4,908 households if food were given ($2,000,000 / $407.46)
• By switching from food to cash, you could include another
741 households or 5,041 people (avg household size = 6.8)
27. Number of additional beneficiaries gained by
switching from food to cash transfers
27
4841 5041
13858
9062
32802
0
5000
10000
15000
20000
25000
30000
35000
Ecuador Niger Uganda Yemen Total
28. Costing transfer modalities: Final comments
• Our approach assumes that if WFP transfers $40 in cash or a food
basket worth $40 that the cost of the transfer itself to WFP is $40
• We obtained detailed data on procurement costs for the
commodities used in the Ecuador study. The $40 food basket cost
WFP approximately $40 to procure suggesting that this assumption
is reasonable
• With further experience and some standardization of procedures,
modality specific cash costs could fall substantially
– Eg, cost of delivering cash transfers was 6% of transfer value in Niger
28
30. Preferences for food, cash and vouchers
(Percent)
30
55
7 8
73
50
1 0
35
12
7
77
20
3
16
8 6
49
79
18
2
56
0
10
20
30
40
50
60
70
80
90
all food all cash all voucher
31. Additional Results
• Nutrition, Uganda
– Stunting/wasting: No robust impacts of cash or of food transfers
– Anemia: food transfers have no impact; cash transfers reduced anemia among
preschool children in the target age groups
• Nutrition, Ecuador
– Anemia: neither food , cash or vouchers had any effect on anemia
• Other results
– Virtually no evidence that beneficiaries sell their food rations (all countries)
– No meaningful evidence that cash transfers are used for undesirable purposes
such as buying beer (Uganda) or qat (Yemen)
– No evidence that different modalities have differential impacts on intra-
household decision making or on social tensions between host and refugee
communities (Ecuador)
– Cash transfers improved child cognitive development across several domains:
visual reception, receptive language, expressive language (Uganda)
– Differences in private costs of obtaining transfers reflected program decisions
regarding placement of payment points
33. Summary
33
1) Across a range of interventions in a variety of settings, cash transfers
generally but not always proved more effective in improving WFP’s core
food security indicator – FCS – at significantly less cost
2) We found little evidence that cash had adverse impacts as measured by
creation of social tensions, changes in intrahousehold decisionmaking,
or purchase of intoxicants; we did find evidence of other beneficial
impacts of cash, such as on children’s cognitive development
3) This does NOT imply that cash is always “better” than food
4) Policy and intervention design discussions need to shift from
their current ideological and political focus to one that
emphasizes
– Greater precision in intervention objectives
– More nuanced understanding of context
– Greater emphasis on costing
34. INTERNATIONAL FOOD POLICY RESEARCH INSTITUTE
UGANDA: IMPACTS ON COGNITIVE DEVELOPMENT
FOR PRESCHOOL CHILDREN
34
35. Uganda: food/cash transfers for children in ECD centers
• Could a cash transfer linked to preschool improve early childhood
development, by promoting investments in nutrition and/or
stimulation?
• Cash transfers can be a nutrition-sensitive intervention
– Increasingly considered a preferred form of “food assistance”
– Can have similar/better effects to food transfers on dietary diversity, at less cost
to implementers (e.g., Hidrobo et al, 2013, in Ecuador)
• Conditioning transfers on preschool may increase preschool
participation
– Preschool shown to be an important source of early stimulation but
participation tends to be low (Engle et al, 2007)
36. Key questions
• In this study from Karamoja, Uganda, we focus on the following:
– Do cash transfers linked to preschool improve measures of early
childhood development for children aged 3-5 years?
– If so, what are plausible mechanisms? (Improvements in nutrition?
Increases in stimulation?)
Preschool
participation
Child
development
Nutrition
Diet quality
Health/
sanitation
Cash linked
to preschool
Stimulation
37. Study setting
• Informal UNICEF-supported preschools (ECD centers) in Karamoja, Uganda
Karamoja, Uganda:
• Rural, remote, rugged sub-region
in Eastern Uganda
• Very poor (~$80/month mean HH
consumption at baseline)
• Diet relies heavily on starches
ECD centers:
• In Karamoja since 2007, informal, usually under a tree
• Volunteer caregivers trained by CSCD, under UNICEF
• No food provided, very few materials
• Community supposed to support, rarely did at baseline
• Target age: 3-5 years
38. Study design
• WFP, UNICEF, and IFPRI collaboration to link cash transfers to ECD
center enrollment using a randomized controlled trial design
– 31 ECD centers to cash group, 32 ECD centers to control group
– Household eligible if had a child age 3-5 years, enrolled in ECD center
assigned to cash
– $10.25 (25,500 UGX) transfer for each eligible child per cycle: ~13% of
monthly household expenditure
– Seven 6-8 week cycles, from August 2011-August 2012
– Cash added electronically to a card given to mothers, retrieved from
mobile money agent
39. Data collection
• Longitudinal survey of households with at least one ECD-enrolled child aged 3-5
years at baseline
– Baseline survey in 2010; endline survey in 2012
– 1889 households in sample
– Balancing and attrition tests show treatment arms very similar on average at baseline
– Impacts estimated using single-difference (robustness checks using ANCOVA)
– This presentation focuses on children with “maximum exposure” to transfers: within
eligible age range of 3-5 years (i.e., 36-71 months) throughout study period
Age of Target Child
at…
Baseline Endline
(Sep-Nov 2010) (Mar-May 2012)
36-53 months 54-71 months
40. Measurement of key outcome: Child development
• Items drawn from Mullen
Scales of Early Learning
• Simple games played with
child by trained enumerator
• Adapted to Karamoja by a
clinical psychologist at
Makerere University
• Domains:
• Visual reception
• Fine motor
• Receptive language
• Expressive language
41. Core result: Impacts on Child Development
• Cash linked to preschool significantly increased child development measures,
overall and across several distinct domains
0.33 **
0.31 **
0.15
0.39 **
0.45 ***
0.0
0.1
0.2
0.3
0.4
0.5
TOTAL Visual
reception
Fine motor Receptive
language
Expressive
language
Child Development
Standarddeviations
Impacts of Cash on Child Development
(ages 36-53 months at baseline)
Cash v Control
42. Evidence for nutrition pathway: Impacts on Anemia
• Cash significantly reduced anemia, particularly moderate to severe anemia (WHO
standards, adjusted for altitude)
• High anemia prevalence in control group: 42.9% have anemia, 24.5% have moderate/severe
anemia
• Plausible evidence for nutrition pathway of child development impacts
-10.0% * -9.6% **
-12%
-10%
-8%
-6%
-4%
-2%
0%
Any anemia Moderate/severe
Anemia
Percentagepointchange
Impacts of Cash on Anemia
(ages 36-53 months at baseline)
Cash v Control
43. Evidence for nutrition pathway: Impacts on Child Diet
• Cash linked to preschool significantly increased frequency of children’s
consumption of starches, meat/eggs, dairy in past 7 days
• Relative to control group, reflects 66% increase in meat/eggs, 100% increase in dairy
• Plausible evidence for diet pathway of anemia impacts
0.55 ***
0.17
0.51 ***
0.33 *
0.03
0.21
0.1
-0.02-0.1
0
0.1
0.2
0.3
0.4
0.5
0.6 Starches
Leafygreens
Meat/eggs
Dairy
Orangefruit/veg
Otherveg
Otherfruit
Corn-soya(CSB)
Child Diet
Daysinpast7days
Impacts of Cash on Child Diet
(ages 36-53 months at baseline)
Cash v Control
44. Evidence for nutrition pathway: Impacts on Sanitation
• Cash significantly improved sanitation at ECD centers (through household gifts)
and significantly reduced children’s diarrhea
• Also plausible evidence for sanitation pathway of anemia impacts
Cash Control
Gave a cash gift to ECD caregiver (%) 31.1 *** 13.5
Value of gifts to the ECD caregiver (UGX) 980.4 *** 318.2
ECD center has a shelter (%) 86.1 *** 65.5
ECD center has access to a latrine (%) 88.7 *** 60.5
ECD center has hand-washing facilities (%) 38.2 ** 22.0
ECD center has other materials (%) 13.0 ** 3.9
Child had diarrhea in past 15 days (%) 3.5 *** 7.6
45. Evidence for stimulation pathway: Impacts on ECD Participation
• Cash significantly increased number of days ECD centers were open and number
of days children attended
• As with CCTs conditioned on primary or secondary school, increase in participation
• Plausible evidence for stimulation pathway of child development impacts
2.35 ***
1.87 ***
0
0.5
1
1.5
2
2.5
Days open in past 7 days Days attended in past 7 days
ECD Center
Days Impacts of Cash on ECD Center Participation
(ages 36-53 months at baseline)
Cash v Control
46. Conclusions
• Cash transfers linked to preschool in Karamoja, Uganda
– Significantly improve child development for age 3-5 years: 0.33 standard deviations
– Plausible nutrition mechanism
• Significantly reduce anemia (particularly moderate/severe)
• Plausibly through
– Improved diet quality (particularly meat/eggs, dairy)
– Improved sanitation/health
– Plausible stimulation mechanism
• Significantly increase ECD center participation (similar to conventional CCTs)
• Takeaway: A cash transfer conditioned on preschool can effectively improve child
development among children age 3-5 years, through plausible mechanisms of
promoting nutrition and stimulation investments