cows, missing milk markets and nutrition in rural ethiopia
DESCRIPTION
InteTRANSCRIPT
INTERNATIONAL FOOD POLICY RESEARCH INSTITUTE
Cows, missing milk markets and nutrition in rural Ethiopia
John Hoddinott, Derek Headey and Mekdim Dereje
2
Introduction
In rural areas, is child nutrition affected by which goods the household produces?
Conceptually, if there are complete markets, production and consumption decisions are separable
=> production decisions do not affect consumption=> agricultural assets & production decisions only affect nutrition via their contribution to income But if markets are missing, this will no longer be trueÞ production decisions can directly affect nutritionÞ There is an opportunity for agricultural policies to
influence nutrition outcomes directly, not just via income
3
Introduction, cont’d
Milk production is an instructive good to consider Markets often missing since milk is naturally perishable
• In Ethiopia, 85% of all milk produced is consumed by the producing household
• Another 7% seems to be trade/bartered locally• Domestically processed milk only started about a
decade ago and largely confined to urban areas Milk is important for growth in early life
• Cow’s milk contains and stimulates insulin-like growth factor I (IGF-I) which plays a key role in child growth
• Also important source of animal-sourced protein, amino acids, iron, B-12 and other micronutrients
4
Introduction, cont’d
Milk may also be important for other reasons:1. Relative to solids, it’s easily digestible for young children2. Dairy cattle often the responsibility of women, so more
maternal control of milk resources than other nutrient-rich crops
Factors above may imply more opportunities for snacking Milk consumption may also explain a long-noted paradox
in Ethiopia: some areas of high agricultural potential have relatively high rates of malnutrition (e.g SNNP); some areas of low potential have relatively low rates of stunting (e.g. Somali region) >>> could cow ownership explain this?
5
Introduction, cont’d Nutrition literature finds evidence of milk impacting linear
growth and other nutrition outcomes (Marquis, et al. 1997, McLean, et al. 2007, Murphy & Allen 2003, Neumann et al. 2002, Randolph, et al. 2007)
But not much evidence on the issue at the farm level: Small sample farm surveys from Kenya (Hoorweg,
Leegwater and Veerman 2000, Nicholson, et al. 2003), and Uganda (Vella, et al. 1995)
More experimental design from Rwanda (Pimkina et al. 2013) uses cow allocation to HHs in an NGO program
Sadler et al. (2012) use an RCT with supplementary feed in pastoralist Ethiopia (during drought)
Nearly all find some impact on child growth
6
Data
Primarily use a very new and large (n=7,930 HHs) baseline survey for GoE’s Agricultural Growth Programme (AGP)
Sample is drawn from 94 high agricultural potential woredas, with 304 villages (EAs) , 2011
Most of the sample pertains to highlands Although “high potential”, there is actually a lot of
variation in agroecologies and market access This variation should, in turn, produce the necessary
variation in farm sizes, asset ownership and access to markets
7
F1: AGP enumeration areas , major markets and woreda level population density
Source: http://www.gafspfund.org/content/ethiopia. Market towns (light circles) are from FEWSNET, and population density at the woreda level is from the 2007 National Census of Ethiopia. Notes: Population density categories (in persons per square kilometer) from lightest to darkest are 0-31, 31-101, 101-139, 139-195, 195-537, 537 and above.
8
Data
Data collected on agricultural assets, production, foods consumed by children under 2 years in previous seven days, anthropometry of children under 5 years• Around one quarter of children consumed milk in the
last 7 days; • one of the most important sources of protein for young
children, in a highly undiversified diet• About 64% of households own at least one cow• High level of stunting (47%)
9
Data
Cattle and other livestock ownership are also available in the Demographic Health Surveys (DHS) of 2000 and 2010
However, 2011 only includes 24 hr recall, and both surveys don’t distinguish between sex of cattle
Nevertheless, they are nationally representative high-quality surveys, so we test robustness to these data
10
Estimation strategy – very simple
Outcome variables:• Did child consume milk in last 7 days• Number of days child consumed milk in last 7 days• Anthropometry (HAZ, stunting, WHZ, wasting)
Use linear probability models and OLS. More complicated estimators (probit, count models) produce similar results
Include as controls: child sex and age; care givers education and age; characteristics of the head (age, education, sex); region dummy variables; and capital goods for agriculture - land operated by the household for cultivation and the ownership of at least one cow.
Standard errors are clustered at the woreda level.
11
First some descriptives . . .No milk 1 to 3 times 4 to 6 times Daily
-2.20
-2.10
-2.00
-1.90
-1.80
-1.70
-1.60
-1.50
-2.07
-2.00-2.03
-1.73
Consumption milk, last 7 days
Child
Haz
scor
es
12
No milk 1 to 3 times 4 to 6 times Daily
0.00
0.10
0.20
0.30
0.40
0.50
0.600.55
0.520.55
0.45
Consumption milk, last 7 days
Prob
abili
ty o
f bei
ng st
unte
d
13
Basic results
Household owns at least one cow: Impact on:(se’s in parentheses)
Marginal impacts on Anthropometry: 6-24m
Anthropometry: 12-24m
Any milk in last 7 days
# milk days last week
HAZ Stunted HAZ Stunted
0.225*** 1.263*** 0.214* -0.055* 0.324*** -0.099***
(0.024) (0.140) (0.112) (0.028) (0.117) (0.028)
14
Extensions and robustness checks (1): Including additional controlsImpact of cow ownership on # days milk consumed in last 7 days
1.263*** 1.005*** 1.212*** 0.983*** 1.289*** 0.818***Other assets N Y N Y Y Y
Ag income N N Y Y Y Y
Village factors N N N N Y Y
Woreda dummy N N N N N Y
Impact of cow ownership on stunting, children 12-24m
-0.099*** -0.088*** -0.102*** -0.091*** -0.086*** -0.069***
Other assets N Y N Y Y Y
Ag income N N Y Y Y Y
Village factors N N N N Y Y
Woreda dummy N N N N N Y
15
Extensions and robustness checks (2): “Placebo test”
Marginal effect on any cow ownership on other types of food consumption in last seven days
Pulses Vegetables Leafy vegetables
Fruit Meat Meat organs
Eggs
0.002 -0.015 0.022 0.012 0.000 0.003 0.006
(0.027) (0.019) (0.024) (0.018) (0.016) (0.005) (0.023)
If cow ownership represents wealth rather than dairy consumption, we would expect it to have a significant impact on consumption of other high value foods
This placebo test suggests no evidence of this
16
Extensions and robustness checks (3): Looking for the missing markets
Food Market in village No food market in village Any milk
consumption# days milk
consumed
Stunted12-24m
Any milk consumpti
on
# days milk
consumed
Stunted12-24m
HH owns cow 0.189*** 0.702** 0.114 0.222*** 1.275*** -0.119*** (0.050) (0.309) (0.106) (0.026) (0.152) (0.033)
Any milk consumption
# days milk consumed
HAZ12 – 24m
Stunted12-24m
HH owns cow 0.225*** 0.20*** 1.26*** 1.10*** 0.32*** 0.23* -0.099 -0.08***
(0.024) (0.024) (0.140) (0.126) (0.117) (0.130) (0.023) (0.029)Village cows - 0.058** - 0.362**
*- 0.22 -0.032
(log) (0.028) (0.152) (0.150) (0.030)
17
Extension to 2000 Ethiopian DHS
Any milk consumption
Milk every day
Stunted12-24m
Child HAZ
Stunted12-24m
Child HAZ
HH owns cow
0.20*** 0.20** -0.06*** 0.22***
or camel (0.02) (0.02) (0.02) (0.06)Milk daily 0.10*** 0.28***
(0.02) (0.06)
Controls
Maternal Yes Yes Yes Yes Yes YesWealth Yes Yes Yes Yes Yes YesRegion Yes Yes Yes Yes Yes Yes
N 6150 6150 5430 5430 5430 5430
18
Extension to 2000 Ethiopian DHS
Any milk consumption
Milk every day
Stunted12-24m
Child HAZ
Stunted12-24m
Child HAZ
HH owns cow
0.20*** 0.20** -0.06*** 0.22***
or camel (0.02) (0.02) (0.02) (0.06)Milk daily 0.10*** 0.28***
(0.02) (0.06)
Controls
Maternal Yes Yes Yes Yes Yes YesWealth Yes Yes Yes Yes Yes YesRegion Yes Yes Yes Yes Yes Yes
N 6150 6150 5430 5430 5430 5430
19
Conclusions
Ownership of cows is associated with higher milk consumption by children 6-24m (especially children 12-24m), improvements in HAZ and reductions in stunting
Magnitudes of effects are large – reduction in stunting of approximately 7-10 percentage points
Impact is much larger than many other variables Some evidence that the existence of food markets can
partially substitute for own production
20
Conclusions
Need to be cautious; not an experimental study, though results are robust to a number of checks and alternative model specifications, and even very different data
Main concern is that cows are not randomly allocated; unobservables could be associated with cattle
Particular concern is that cattle represents wealth In future work we are exploring instrumentation
strategies. Conceivably, livestock feed constraints might impact cow ownership but not nutrition directly
May also try propensity score matching
21
Conclusions
Interesting and important policy implications In subsistence settings, ownership of cows at household
level is key, though some externalities to ownership at village level too
But ultimately cow ownership may decline with increasing feed and water constraints, and mechanization
This decline is substituted, however, by greater market access
For GoE and partners, there may be large returns to increasing investment in dairy sector, which currently gets a tiny fraction of agricultural budget
22
Conclusions
What sorts of investments? Ethiopia has large livestock herd, but very little use of
improved breeds Yields are very low: about half of neighbouring Kenya Modern processing of dairy products very limited and
confined to urban areas Potentially, important lessons to be learned from
international dairy success stories, such as India Perhaps also potential for dairy supplements, though
these need to target young children