Post on 30-Jun-2015
transcript
School Meals and Classroom
Performance: Evidence from India
Farzana Afridi, Bidisha Barooah and Rohini Somanathan
Delhi School of Economics, 2nd August 2013
MotivationThree possible mechanisms by which school
meals improve learning improved enrolment and attendancehigher nutritional status and long-term cognitiongreater attention/concentration in classroom tasks
We study the effect of school meals on attention in classroom task as opposed to cognition
School meals and classroom attention
Possible positive effects Attention span and short-term memory Behavior- aggressive, attention-deficit,
hyperactivity Hunger alleviation
Possible negative effects Pressure on school resources (teacher time, space)Post-lunch dip
Mid-day Meals in DelhiTime Grades Program Features
1997-2003 1 to 5 Packaged food (biscuits)
July 2003 1 to 5 Cooked meals. 300 calories and 8-10 grams of protein per day per child
October 2006 1 to 5 450 calories and 8-10 grams of protein per day per child
29th Sep 2009 Extended to 6-8
Cooked meals. Exact date of implementation unanticipated. Norms (each child per day):1-5: 450 calories and 8-10 grams of protein6-8: 700 calories and 9-12 grams of protein
Study Design- I
Conducted tests of attention in grades 5 and 7 in 18 randomly selected public schools in Delhi.
Two phases: Aug- Nov 2009; Feb- Apr 2010
Two sessions of solving mazes in one school day- beginning and end of day.
Students were given to solve the same mazes in both phases.
Study Design-IIPhase 1 dates of survey
Grade
Phase 1 Meal
status
Phase 2 Meal
status
Treatment status
1st August to 8th September, 2009
7 No Yes Treatment
5 Yes Yes Placebo
8th October- 3rd November, 2009
7 Yes Yes Control
5 Yes Yes Placebo
Study Design- III: MazesMaze Number Difficulty level for Grade 5
1 (Practice) 1
2 (Practice) 1
3 2
4 3
5 4
Used in experiments to measure effort (Gneezy et al, Nierdle et al)
Requires no long-term cognitive skills such as verbal ability or problem solving or reading and writing skills
Data
1213 students of grade 5 and 7 solved maze puzzles in Phase 1
Panel data of scores in mazes for 827 students
Score = # of mazes correct/ 3
Other controls: heights and weights, curriculum based test on math and language , dietary intake of that day, socio-economic indicators of a subsample of 10 students per grade
Summary statistic (Full Sample)
Uptake of School Meals
Baseline school summary statistics
Control Schools (N=7)
Treatment School (N=10)
(1) (2)Enrolment in Grade 5 (all sections) 59.57 59.30
(9.413) (9.459)Enrolment in Grade 7 (all sections) 249.29 150.60
(56.018) (15.995)Attendance Rate in Grade 5 (all sections) 0.89 0.82
(0.023) (0.040)Attendance Rate in Grade 7 (all sections) 0.84 0.81
(0.030) (0.021)Grade 5 Score in Math and Language 3.13 2.88
(0.223) (0. 121)Grade 7 Score in Math and Language 4.19 3.93
(0.131) (0.341)Students able to solve IQ question 0.20 0.18
(0.019) (0.018)Time taken in distribution of MDM (in minutes) 56.42 29
(9.923) (6.741)Recess Duration (in minutes) 24.28 22.5
(2.02) (1.53)
Baseline Grade 7 student summary statistics
Baseline Grade 7 student summary statistics
Maze Score (averaged over two sessions)
Empirical Strategy
Identification of treatment and control groups based on Unanticipated introduction of school mealsRandom date of first visit to schools
Intention to Treat EstimatorTreatment status is by schools and not individuals
Estimating Equation
MeanScoreijp = β0 + β1 Phase2p+ β2(Treatj*Phase2p) + Xi + µj + eijp
MeanScoreijp is the average of two sessions of the score of child i in Grade 7 of school j in Phase p.
Phase 2p is the dummy for Phase 2
Treatj is the dummy variable for schools that were surveyed before 29th September.
β2 is the treatment effect.
µj is the school FE. We also use individual FE.
Grade 5 acts as a placebo test
Effect of school meals on average scores
Robust SE in parenthesis. Bootstrapped SE in square bracketsSchool FE has controls such as BMI for age, gender, age and ability to solve IQ questionResults unchanged for unbalanced panel school FE
Effect of school meals by sessions
Session 1 is the session conducted before school meals were served on the day of survey. Session 2 is the session conducted after the lunch break was over. The average gap between the 2 sessions is 1 ½ hours.
Effect on Nutritional Status and Maze Score by Gender
Child Fixed Effects
Student Characteristics by Gender
DiscussionOverall positive effect on tests of
attention/concentration due to school meals
Running the same analysis for all students, results remain unchanged
Positive effect in pre-meal sessions in all schools
Positive effect in post-lunch break session in schools with higher scores in math and language
Significant gains in BMI for girls