impact of food transfers in zomba aberman
TRANSCRIPT
Evidence from a mixed-methods study in Malawi
Aulo Gelli, Noora-Lisa Aberman, Amy Margolies, Marco Santacroce, Bob Baulch
and Ephraim Chirwa
The impact of lean season food transfers on food security, diets and nutrition status
Study objectives• To estimate the impact and targeting of lean
season food transfers (MVAC) on households’ food security and children’s diets and nutrition
• To understand village-level norms on allocation of food transfers and other resources that may help explain those results
Study sites
• Data was collected from 60 communities randomly selected among a set of food-insecure villages in MVAC targeted region of Zomba district in southern Malawi
Study methods: Quantitative Data
• Longitudinal study based on two rounds of surveys undertaken (as part of a cluster randomized controlled trial of a pre-school based agriculture and nutrition intervention): – Baseline survey undertaken in the post-harvest
season (September 2015) – Follow-up undertaken during the peak lean season
(February 2016) after scale-up of food transfers – Rich data set including ~1200 households, over
1,500 children
Study methods: Qualitative Data
• Qualitative data is made up of 45 in-depth time-line interviews in the same communities in Zomba, with women, men and adolescent girls (March 2016) – Translated and transcribed, then thematically
coded
Evaluation strategy• First estimated a probit model to assess the
probability of targeting criteria to predict program participation – using range of household and community level
characteristics on sample of MVAC beneficiaries and non-beneficiaries
• Then we evaluated the impact of MVAC by combining propensity score matching and difference in difference (DID) methods
Outcomes
• Two levels of outcomes:1. Alternative measures of household food security
estimated from consumption and expenditure modules using adult equivalents
2. Measures of diets and nutrition status of young children
MVAC programme characteristics
• Eligibility criteria for MVAC food assistance included households headed by women, the elderly or children, or households including orphans, the chronically ill or households that had lost their main source of income due to chronic illness
• Household screening criteria also included asset holdings (including livestock, land and small durables), participation in other social assistance programs (including social cash transfers, inputs subsidy programme and school meals)
Food transfers
• Food rations were to be provided to households on a monthly basis and included maize (50kg), legumes (10kg) and fortified vegetable oil (1.84 kg)
• At endline, 175 (15%) out of 1,191 households had received MVAC in the survey population
• Households consuming <1800 calories per capital per day: 36% at baseline and 46% during lean season
0.000.050.100.150.200.250.300.350.400.450.50
Scho
ol fe
edin
g pr
ogra
mPr
e-sc
hool
feed
ing
prog
ram
MVA
CSe
eds
Voca
tiona
l tra
inin
gFo
od fo
r wor
kCa
sh tr
ansf
erW
ater
pur
ifica
tion
for h
ome
use
Heal
th o
utre
ach/
trai
ning
FISP
Oth
er fr
ee a
gric
ultu
ral
Assis
tanc
e fr
om c
omm
unity
…Re
mitt
ance
s fro
m w
ithin
or…
Agric
ultu
re to
ols
Mos
quito
net
sCr
edit
from
mic
rofin
ance
age
ncie
s…Ed
ucati
onal
ass
istan
ceVo
catio
nal t
rain
ing
othe
rSp
ecia
lized
kin
ds o
f foo
ds fo
r…O
ther
Cash
forw
ork
Inpu
ts fo
r wor
kN
APSA
/Soc
ial S
ecui
rty/
Pen
sion…
Scho
lars
hip
for s
tudy
/tra
inin
gCl
othe
s/sh
oes
Wor
kman
’s c
ompe
nsati
on …
Shar
e of
HH.
Cov
ered
Social assistance coverage in survey population (lean season)
Share of HH. receiving assistance -33% of HHs receive no prog.-33% of HHs receive 1 prog.-20% of HHs Receive 2 progs.…
Effects of MVAC Food Transfer
• During the lean season, households in the sample experienced substantive declines in household food security
• Compared to control, MVAC food recipients were better off:– substantive positive impact on household food
consumption in 7 day recall period – substantive positive impact on young children’s
diets and nutrition outcomes
Treatment effects: Household level
• Substantive positive impact on household food consumption in 7 day recall period– Per capita food expenditures + 19% / 35MK pppd– Daily acquisition of iron +16% / 3.92mg pppd
** p<0.05.
MKw
/day
Treatment effects: Child level
• Substantive positive impact on young children’s diets (DDS +15% & FVS +13%)
• And nutrition outcomes: weight-for-height z-scores (+14%)
Children 36-72m Children 6-59m
***
*** **
*** p<0.01; ** p<0.05.
MVAC targeting
• Findings indicate that MVAC targeting criteria are not good predictors of program participation
• Data on MVAC participation also suggests that ~20% of most food secure households (by quintile) received transfers
• Positive effects on food expenditures and children’s diets are concentrated among the poorer households
• SCTP recipients appear to be excluded from receiving MVAC food
MVAC participation by socioeconomic and food security status
MVAC participation by official poverty line
Household expenditures
poor non-poor
MVAC 10% 5%
no MVAC 55% 31%
Qualitative Results
Resilience and Coping: tracking the ups and downs
• Ups and downs: major life events cause shocks, as well as annual lean season shock
• All households face dietary shifts during the lean season, decreasing amount of food or shifting to less preferred (though not always less nutrient-dense) foods– Few manage lean season without negative coping
(diets, schooling, assets)
Social norms, targeting & favoritism • Village heads play a significant role in determining who received
MVAC and other social support programs.– Primarily through control over beneficiary selection, also through
decrees about sharing• Villages vary in terms of perceptions of extent of consultation in
the targeting process• Perceptions of the extent (and definition) of favoritism also
varies• Some complaints about chief intervention, e.g., inclusion errors
(favoritism) and forced sharing, but frequently this is viewed as unavoidable and part of village norms
• Beyond some complaints about chief intervention, the primary complaint about targeting is “not enough benefit” related to feeling that “everyone should get something”
Social norms and sharing• About half the time sharing is dictated by the chief. When
it’s not, HHs decide to share on their own due to kinship obligations, social pressure, and hope for reciprocation (often described as a moral or humanitarian requirement).
• Some people complained about forced sharing and community pressure to share. No one complained about sharing with relatives => social requirement.
• Sharing is required even if targeted recipient is objectively poorer than those with whom they share (relative wealth does not seem to be considered in sharing decisions)
• Sharing norms related to cash transfers may be different => possibly less sharing, fear of tracking
Female MVAC recipients in Zomba in favor of forced sharing:Interviewer: Did you think this sharing was beneficial?Respondent 1: Yes, it was beneficial because it could be you next year not in the program, and your friends would help you. But the owners of the program say not to share, this only happens in the village to just help each other.
Interviewer: Alright. Was the sharing fair?Respondent 2: I can say that on this side of the village it was fair. Because this village has two tribes; those from the chief and those not related to the chief…So if they need 8 people then they will take 4 from each side.
Female MVAC recipients in Zomba unhappy with targeting and forced sharing:
I: So what criteria were they using to select beneficiaries?R: They were choosing people who had nothing to eat…But at times they recorded names of people who had food but those who lacked food were also being skipped…As per village level problems, the chief said, “This maize should be shared amongst you. You will see how you can share.” So people could share two [households] per bag…
Male non-recipients in Zomba describing the unfair targeting processes:R: I should just give an example of a certain year, where I was really touched [hurt] in my heart. I received a coupon that I should be receiving maize. After three months I discovered that somebody was using my coupon to get my maize. Somebody with a higher position in the village. The person came and offered me 12 kilograms, and I said all the people were receiving 25 kilograms, why should I get 12? No I cannot accept that, you will consider me the next time.
Conclusions: Effects of MVAC
• Quantitative data suggests that MVAC food transfers are effective in protecting food security and nutrition status during the lean season.– Evidence of protective impact on household
food consumption, and on dietary diversity (of 3-6yr olds) and weight for height z-scores of young children (0-5yrs)
Conclusions: Targeting and Coverage
• Targeting and coverage of MVAC:– Overall coverage of transfers was low in the survey
population (~15% of HH)– Data on targeting criteria are not good predictors of
program participation– Evidence suggests that ~20% of the most food secure
households received transfers• Effects on food expenditures and diets are greater for
poorer households (better targeting=>more efficient)• Data also shows that those receiving SCTP are excluded
from receiving MVAC, even though SCTP recipients should be among the poorest and most food insecure.
Conclusions: Sharing and Favoritism
• Insufficient public social support reinforces high dependence on kinship networks and community support
• Favoritism in community-based targeting seen as unavoidable (villagers cannot contest/ it’s chief’s prerogative)
• A more objective targeting system (e.g., UBR) could improve targeting to some extent, but pressure to reallocate once transfer arrive in the village are likely to remain
Policy Implications• Putting in place a targeting system separate from village politics and
norms would likely increase impact. But how?• Most people face dietary shocks each year: suggesting that other social
support mechanisms (productive and protective) must be scaled up to meet the current need.
• However, all program targeting mechanisms must consider sharing & reallocation.
• Possible approaches:– Increased village-level transparency about targeting criteria– Parallel village-level institutions to assist/monitor targeting, distinct from village
governing structures could bypass village norms and politics– Whole village targeting, when feasible– Universal (vulnerable sub-group) targeting, as in Ntchisi, to all families with under-
fives, or to all elderly• Educating villagers on good local governance practices, may slowly begin to
alter norms that yield exclusion and inclusion errors, but weak social support system reinforces these practices.
Open Questions
• Next round of data collection ongoing• Will examine any effects of MVAC on stunting.• Will explore village head’s perceptions of
favoritism and their role in targeting. • Are targeting errors due to sharing or
targeting process? • Do sharing norms differ for different types of
programs/transfers? (e.g. cash)?
Annexes
Post-harvest and lean season summary statistics
Survey population
• Baseline household characteristics
mean sd Obs. Houeshold size 5.33 1.8 1,199 Number of children 0-36 months 0.51 0.58 1,199 Number of children 36-72 months 1.13 0.36 1,199 Number of children 6-14 years 1.45 1.17 1,199 Number of adults 14-65 years 2.25 0.98 1,199 Number of elders 65+ years 0.04 0.23 1,199 Dependency Ratio* 1.58 1 1,199 HH head completed primary education? 0.35 0.48 1,199 Household head's age 36.5 10.5 1,199 Polygamous households 0.03 0.17 1,199 Female headed household 0.28 0.45 1,199 Assets owned: -Large livestock (Oxen or cattle) 0.04 0.43 1,199 -Small livestock (Goats, pigs, rabbits) 0.8 2.43 1,199 -Fowl (Chickens, guinea fowl…) 2.57 5.25 1,199 -Bees/Beehives 0 0 1,199 -Fish pond or fishing equipment 0 0 1,199 -Farm equipment (Non-mechanized) 3.12 2.74 1,199 -Farm equipment (Mechanized) 0.01 0.17 1,199 -Non-farm business equipment 0.16 2.94 1,199 -Houses or other structures 0.32 0.88 1,199 -Large consumer durables 0.23 1.14 1,199 -Small consumer durables (Radio…) 14 30.19 1,199 -Cell phones 0.62 0.75 1,199 -Other land not used for agriculture 0.05 0.32 1,199 -Means of transportation (bicycle…) 0.54 0.7 1,199 -Total asset count 22.5 33.5 1,199 Total expenditure (pp, day)* 242 188 1,139 Food expenditure (pp, day)* 186 150 1,139 Share of food expenditure on total exp.* 0.78 0.17 1,139 Chidlren (6-59 months) -Height-for-age (z-score) -1.74 5.87 1,383 -Weight-for-height (z-score) 0.09 4.29 1,402
Prop. Obs. -Prevalence of Stunting (6-59m) 42% 1,379 -Prevalence of Wasting (6-59m) 1% 1,402 Notes: *Table excludes outliers for food consumption and total expenditure
MVAC Coverage by Cluster
0
5
10
15
20
25
30
0.00
0.05
0.10
0.15
0.20
0.25
0.30
0.35
0.40
0.45
0.50
Kasim
u
Utw
e
Sunu
zi
Thab
wan
i
St P
ius
Golo
gota
Nam
akun
…
Mak
oka
Tsab
ola
Chik
ala
Chan
da
Mpa
ta
ST.A
ntho
n
Mac
here
ni
Nal
ikuk
ut
Naz
ithim
b
chim
wal
ir
Nsa
la
Nac
hisw
e
Nse
njer
e
Coverage (% of HH in cluster)Coverage (HH covered in cluster)
MVAC=175 Cash=69
FFW=52
Total=1,192 households
6
24
Programme participation (probit) model
MVAC targeting
0.00
0.25
0.50
0.75
1.00
Cov
erag
e of
Vul
nera
ble
(Sen
sitiv
ity)
0.00 0.25 0.50 0.75 1.00Inclusion of Non-Vulnerable (1 - Specificity)
Area under ROC curve = 0.4957
Households with Female, Child or Elderly Head or Orphans