Farzana Afridi
Indian Statistical Institute, Economics Department, Delhi
ADB, Manila (July 11-12, 2012)
School meal programs – An impact
assessment
School meal program in India
Every public primary school mandated to implement program in 1995.
All enrolled children provided with a free meal, cooked out of 100 grams of wheat or rice grains on each school day on school premises.
Those unable to due to financial constraints, distributed free grain rations (rural areas) per month subject to a minimum monthly attendance of 80% per student or served ready to eat meals (viz. Delhi) on each school day on school premises.
Nation-wide transition to cooked meals completed in 2003.
Largest school feeding program in the world in terms of number of potential beneficiaries.
Research questions 1. Impact of cooked meals on
short-term nutrition (daily food consumption)
“Child Welfare Programs and Child Nutrition: Evidence from a Mandated School Meal Program.” Journal of Development Economics, 92(2):152-165, July 2010
student partcipation in school (enrollment and
attendance) “The Impact of School Meals on School Participation in Rural India.” Journal of Development Studies, 47(11): 1636-1656, November 2011
2. Impact of the design of program on student participation in school
“The role of design in school subsidy programs: Evidence from mid day meals in India” (with Bidisha Barooah and Rohini Somanathan)
1. Impact on short-term nutrition -
Identification of Program Impact
Randomly chosen date of village interview yielded primary school age children in some villages recalling their previous day’s food intake for a school day while in other villages the previous day was a non-school day.
Sub sample of these children revisited to obtain their food consumption data on both school and non-school days.
total daily intake of nutrient i by child j in village k
school attendance on reference day
nutrient intake from a school meal
sum of school meal transfer of nutrient i to siblings s in the primary school age group
1. Impact on short-term nutrition -Methodology
A
jD
0 1 2 3 4
1
NT A M M
ijk j ijk iks j k ijk
s
C D C C
η
T
ijkC
M
ijkC
1
NM
iks
s
C
Instrumental Variables Analysis Instruments for: - whether the reference day was school day - average consumption of nutrients in a school on
the last school day x reference day was school day
Methodology (cont’d.)
A
jD
T
ijkC
Impact of School Meal Program on Total Daily Nutrient Intake: Cross-
sectional Analysis
1. Impact on short-term nutrition - Results
(1)IV: School day and Average nutrient consumed in school x attendance on reference day
(2)IV: School day and School meal offer x dummy for school day
(3)IV: School day and Month of interview x dummy for school day
Note: Robust standard errors in parentheses. * Significant at 10 per cent, ** 5 per cent and ***1 per cent
Specification Coefficient on Quantity of Nutrient Intake from School Meal
Calories Carbohydrates Protein Calcium Iron N
(1) 2SLS 0.49***
(0.163)
0.52***
(0.154)
0.58***
(0.157)
0.69
(0.445)
0.96***
(0.232)
901
(2) 2SLS 0.70**
(0.272)
0.73***
(0.260)
0.81***
(0.279)
0.55
(0.648)
1.15***
(0.324)
898
(3) 2SLS 0.66**
(0.269)
0.75***
(0.251)
0.75***
(0.271)
0.25
(0.603)
0.77*
(0.386)
901
Impact of School Meal Program on Total Daily Nutrient Intake: Community Fixed Effects
Specification Coefficient on Quantity of Nutrient Intake from School Meal
Calories Carbohydrates Protein Calcium Iron N
(4) 2SLS-FE 0.86**
(0.336)
1.01***
(0.299)
0.61*
(0.304)
-2.74
(1.663)
-0.58
(0.536)
243
(5) 2SLS-FE 1.12**
(0.470)
1.24**
(0.425)
1.09*
(0.528)
-3.45
(2.638)
0.84
(0.535)
243
Results (cont’d.)
community dummies (11 villages within 5 communities)
(4) school day and Average nutrient consumed in school x dummy for school day
(5) school day and School meal offer x dummy for school day
Note: Robust standard errors in parentheses. * Significant at 10 per cent, ** 5 per cent and ***1 per cent
Specification Coefficient on Quantity of Nutrient Intake from School Meal
Calories Carbohydrates Protein Calcium Iron N
(6) OLS-FE 0.76*
(0.404)
0.66
(0.399)
0.62
(0.443)
0.15
(0.671)
0.43
(0.342)
546
Results (cont’d.)
Impact of School Meal Program on Total Daily Nutrient Intake: Individual
Fixed Effects
Note: Robust standard errors in parentheses. * Significant at 10 per cent, ** 5 per cent and ***1 per cent
2. Impact on school participation – Identification of program impact
Transition to cooked meals (from foodgrain distribution) began in July 2003 in the rural survey area
Some schools transitioned in July, 2003
Remaining schools transitioned between July and December 2003
D1 = Average participation in December – Average participation in July
in schools whose participation status changed between July and December
D2 = Average participation in December – Average participation in July
in schools whose participation status did NOT change between July and December
Difference-in-differences= D1 – D2
2. Impact on school participation –
Methodology
2. Impact on school participation – Results
Note: Includes controls for baseline school characteristics interacted with December dummy. Standard errors corrected for clustering on the school reported in parentheses. Missing data for no girl enrollment in grade five. * Significant at 10 per cent, ** 5 per cent and ***1 per cent
3. Impact of Program Design - Program background
Nearly 80% of all public primary schools in Delhi are run by the Municipal Corporation of Delhi (MCD), accounting for almost 85% of all public primary school enrollments in the city.
Prior to 2002-03, ready-to-eat snacks was served in MCD schools such as biscuits, roasted grams, wheat puffs and fruity bread (but mostly biscuits).
Provision of cooked meals started in July 2003.
Phased implementation of cooked meal scheme
Phase 1: between July and August 2003
Phase 2: between September and November 2003
In Phase 1, local service providers were selected on the basis of availability of infrastructure and distance from schools.
In 2004-05 quality of cooked meals improved by weeding out sub-standard service providers.
3. Impact of Program Design – Program background: Costs
Providers re-imbursed at the rate of Rs. 2 per child by Municipal Corporation of Delhi (MCD) under both cooked and ready to eat program.
The raw foodgrains (or foodgrain content of ready to eat items) and transport costs related to foodgrain procurement were fully subsidized by the federal government under both programs.
Schools were mandated to provide atleast 300 kcal per child per day (or 100 grams of raw wheat) in cooked meals and approximatey 265 kcal (or 75.6 grams of raw wheat) in the regime of packaged foods.
10 paise or less than 5% difference in per day per child costs. Annual costs per child are comparable if regularity of both programs is similar.
3. Impact of Program Design – Methodology
Use the staggered implementation of cooked meals.
Almost 50% of the sampled schools implemented cooked
meals program before September 2003 and the other
half, after September 2003.
Treatment group: schools which transitioned before
September 2003 (in July or August 2003) .
Control group: schools which transitioned after
September 2003 (in October or November 2003).
3. Impact of Program Design – Estimation strategy
Aicjmt = α + β0 Cookedjmt + β1Tt + β2 Xm + β3 µict + δi + εicjmt
A icjmt : total number of classes attended by student i in grade c in school j in month m and academic year t divided by the total number of classes that the student could have potentially attended in school j at m, t
Cookedjmt : 0 for all m,t before j transitioned to cooked meals and 1 for all m, t after the cooked meals were initiated
Tmt : 1 if observation is recorded for months in 2003-04 academic year
Xm : 1 if observation is recorded for the month of February, 0 for September
µict : student i’s grade c in academic year t
δi : unobservable, time-invariant individual characteristics
3. Impact of Program Design - Results
3. Impact of Program Design - Results
3. Impact of Program Design - Results
3. Impact of Program Design - Results
Conclusions
On-site school meal program can be more effective in improving nutritional intakes and daily participation rates of intended beneficiaries than an off-site program. Distribution of foodgrains allowed for sharing of the transfer
among family members; on-site cooked meals program was better targeted, potentially reducing leakage of benefits to non-enrolled children and adults in household
Change in program design itself can lead to significant improvement in average outcomes
Food grain distribution was designed to be conditional on a minimum monthly attendance rate of a student but such conditionalities are rarely imposed by public schools. Transitioning to a program that by design required regular attendance to receive program benefits automatically induced an attendance conditionality
Less portability (2002-03), better taste and more variety of meals (2004-05) increased program benefits within limited budget
Conclusions
Need to build in rigorous impact evaluation into program design.
Non-randomized program implementation requires innovative approaches to estimate program impact.
Natural experiments
Instrumental variables
Data constraints