These slides were presented on our research at the ANH Academy meeting in Accra, Ghana, in June 2018. We estimate food expenditure elasticities for different types of food (meat, fruits, vegetables).
ANH Academy slides on Food Consumption and Expenditures in Nigeria
1. Food Consumption and Household
Expenditure Variability: Evidence from
the Nigeria LSMS
Alan de Brauw and Sylvan Herskowitz
International Food Policy Research Institute
ANH Academy, June 2018
2. Research Questions
• As diets are changing and markets improving, what
can household expenditure data tell us about changes
in diets?
– Focus on Food Away From Home, Foods with more
micronutrient availability
• How do people adjust their consumption/expenditure
patterns in response to fluctuating household
resources?
3. Data and its Challenges
• Use six consumption expenditures modules from Nigeria
LSMS-ISA (Panel, 2011, 2014, 2016)
• LSMS-ISA are meant to assess poverty through expenditures
and measure agricultural production
• We ideally want to use diet shares, but ..
– How to aggregate? We choose value of consumption for classes of
foods
– Food away from home a black box
– Signal to noise ratio already high for food consumption;
– For components of diet may be even higher noise over time
4.
5.
6.
7. Regression analysis
• We estimate:
• Where i indexes food types, h households, g clusters, and r
survey rounds
• W is share of expenditures on food type i
• X is (log) total household expenditures
• Prices absorbed by PSU-time fixed effects
• Generates income elasticities if linear (semiparametric graphs
“look” really linear!)
𝑤𝑖ℎ𝑔𝑟 = 𝛼ℎ + 𝑓 𝑋ℎ𝑔𝑟 + 𝛿 𝑔𝑟 + 𝑢ℎ𝑔𝑟
8. Income Elasticities, by Food Type
0 0.2 0.4 0.6 0.8 1 1.2 1.4
Food Away From Home
Dairy
Grains
Meat
"Other"
Pulses
Vegetables
9.
10. Conclusions and Further Work
• Nigerians are purchasing more food away
from home as incomes rise (but a black
box)
• Vegetables fall as a share of the
consumption bundle as incomes rise
• Processed food (“other”) more unit elastic
• Would like to study market change, but
data in LSMS insufficient
• Conflicted on a move to calories rather
than values, so advice would be helpful