Last but not least: The impact of the Farmer Input Subsidy Program on livestock sector in the rural areas of Malawi
1. Last but not least: The impact of the
Farmer Input Subsidy Program on
livestock sector in the rural areas of
Malawi
Alessandro Romeo, Janice Meerman and Mulat Demeke
Monitoring and Analyzing Food and Agricultural Policies (MAFAP)
Lilongwe, Malawi
June 4 2015
3. Motivation (1)
• Malawi: FISP was launched in 2005
– Aiming to intensify crop production while enabling food security
to smallholder farmers
– Provision of coupons to redeem fertilizers and seeds
– Targeted to the “productive” poor (Chibwana et al., 2014)
• FISP mitigated input market failures
• FISP “direct” effects:
– Increased level of crop income leading to higher levels of food
security
– Increased levels of crops used for purposes (small holders selling
crops or crop residue devoted to animal feeding)
4. Motivation (2)
• FISP “indirect” effects:
– “Pull” households toward agriculture diversification other than
crop harvesting
Diversification as an ex-ante risk management strategy to cope with
unpredictable events and market failures
– Good candidate to diversify agriculture activities: livestock and in
particular small livestock as less input intensive and more suitable
for smallholder farmers
• Livestock and diet diversification:
– Positive association between livestock ownership, diet
diversification and consumption of animal source foods
5. Empirical evidences
• Increased levels of crop production, yet, graduation is a
concern (Jayne et al., 2013)
• Increased adoption of improved maize and crop
diversification (Fizbein and Schady, 2014)
• Between 2005 and 2011 rural poverty dropped by
approximately 7pp (Pauw et al., 2014)
• “FISP has improved livestock at household levels as people
produce more than enough food thereby avoiding sale of
livestock in exchange for food” – MANA online, Thursday,
25 December 2014 07:41
• Agriculture diversification is associated with diet
diversification (Jones et al., 2014; Romeo et al.,
forthcoming)
6. Data and Methods
• Data:
– Integrated household survey collected by MNSO in 2010/2011
– Approximately 9000 farmers
– Data collected over socio demographic characteristics including agriculture
livelihood
• Outcome indicators:
– Crop income indicators, livestock ownership by livestock species, household
livestock count and livestock diversification and household diet
diversification
• Methods:
– FISP, crop income and livestock: Propensity score matching to restore
covariate balance and thus mitigating selection bias affecting FISP estimates
> Propensity score estimated making use of relevant socio economic
characteristics
– Association between livestock and diet diversification: Regressions
(Poisson and probit) controlling for relevant socio economic factors
7. Results (1)
• Hypothesis to be tested:
– FISP impacted crop income, revenues from crops
sold, and crops devoted to animal feeding
– FISP impacted livestock ownership, livestock
diversity, and livestock holding levels by livestock
species.
– Association between household diet diversification
and livestock diversification
– Gender of household head affected the association
between livestock holdings and household diet
diversity.
8. Results (2)
Table 1. FISP impacts on net MPC crop income outcomes (MWK)
(1) (2) (3)
Outcomes
Net crop
income
Revenues from
crops sold
Crop income
for animal
feeding
FISP 342.30 291.43 0.25
Control 233.94 251.71 0.04
Mean difference 108.36*** 39.72* 0.36**
SE (12.77) (0.08) (0.04)
Note: *** =1% level, **=5% level, *10% level. Propensity score is estimated
making use reference rainy season, household head characteristics,
demographic characteristics, household wealth, agriculture services and
infrastructures, regional dummies
10. Selected results (4)
Table 4. Household livestock diversification, diet diversification and consumption incidence
of animal source foods
(1) (2) (3) (4)
Household
diet
diversity
score
Meat
(1=yes)
Milk and
dairy
products
(1=yes)
Eggs
(1=yes)
Coeff. Coeff. Coeff. Coeff.
Household livestock count 0.01** 0.04*** 0.02*** 0.05***
Note: *** =1% level, **=5% level, *10% level. Control characteristics include crop count, month of the
interview to capture seasonality in food consumption, reference rainy season, household head
characteristics, demographic characteristics, household wealth, agriculture services and infrastructures,
regional dummies.
11. Selected results (5)
Table 3. Household livestock diversification, diet diversification and consumption incidence
of animal source foods
(1) (2) (3) (4)
Diet
diversity
score
Meat
(1=yes)
Milk and
dairy
products
(1=yes)
Eggs
(1=yes)
Coeff. Coeff. Coeff. Coeff.
Household livestock count 0.00 0.03*** 0.01*** 0.04***
Female head (1=yes) -0.28*** -0.09*** 0.00 -0.05**
Household livestock count * Female head
(1=yes) 0.27*** 0.03* 0.02*** 0.05***
Note: *** =1% level, **=5% level, *10% level. Control characteristics include crop count, month of the
interview to capture seasonality in consumption, reference rainy season, household head characteristics,
demographic characteristics, household wealth, agriculture services and infrastructures, regional dummies.
12. Conclusions
Direct effects:
• Significant increase in crop income indicators
Indirect effects:
• Increase in livestock ownership and diversification toward small
livestock
Diet diversification:
• Household livestock is associated with diet diversification and
consumption of animal source foods
Study limitations:
• Panel data to address time invariant unobserved heterogeneity and
further expand/refine the set of analyzed indicators
Thinking out of the box:
• Along with direct effects, subsidy programme might generate indirect
effects that could help better understanding overall progarmme
impacts (Jayne 2013)
14. Estimated propensity score and
reweighting the data
• Propensity score estimated making
use of relevant household
characteristics including:
> Reference rainy season
> Household head characteristics
> Demographic characteristics
> Household wealth
> Agriculture services and infrastructures
> Regional dummies
• Good overlap between the FISP
beneficiaries and non beneficiaries
• Matching reduced the number of
unbalanced household
characteristics from 16 to 7 out of
27
• After weighting the data with PS
standardize bias difference was
4.5pp below the recommended cut-
off point of 10pp
0 .2 .4 .6 .8 1
Propensity Score
Untreated: Off support Untreated: On support
Treated