Impact of School Resources on Educational Outcomes: Evidence from school dry
cereals in Burkina Faso
By
Pouirkèta Rita NIKIEMA
PhD. Candidate in Applied Economics
Inter University Graduate Program (PTCI)
University Cheick Anta Diop
Email: [email protected]
Abstract
This study uses a quasi-experimental method to evaluate the impact of a School Feeding
Program on educational outcomes in Northern primary schools in Burkina Faso. The
program targeted at school level is a doubled scheme. First, in all schools, at lunch time
students are provided with meal each school day. Second, in all rural schools where girls’
enrollment rate is below 40 per cent, all girls are provided with an additional scheme, take
home ration for household. Under this additional scheme, girls are provided with 10 kg of
dry cereals each month, conditional on 90 per cent attendance rate. At the end of 20112012 academic year, results show that school take home ration increased school attendance
(boys and girls) by 8.4 per cent. Also, the study finds that girls’ enrollment rate within
schools increased by 3.2 per cent. This is driven by the increase in number of new enrolled
girls more than boys. We conclude that take home ration have the potential to increase girl
educational attainment and gender equality within schools.
JEL Classification: D04, I21, I25, O15
Key words: school feeding program, take home ration, enrollment, attendance, Burkina
Faso
1. Introduction
Improving educational outcomes is one of the top priorities in most countries, especially in
the developing world, which lags behind high income countries with respect to many
educational indicators (Galiani and Perez-Truglia, 2011). This concern is partially driven
by the idea that the formation of human capital through education is one of the main
drivers of economic growth. In the late 1980s and early 1990s, the endogenous growth
theories (Romer, 1990; Aghion and Howitt, 1998) argued that differences in economic
growth over time and across countries comes mainly from differences in investment in
human capital including education, health and nutrition. According to human capital
theory, there is a positive relationship between education (human capital accumulation)
and economic growth (Schultz, 1961; Denison, 1962).
Indeed, since over the past two decades, Sub-Saharan Africa (SSA) countries have
embarked on the path of universal education: Education For All (EFA) by 2015. 1 Despite
1
The EFA not only calls for an invest in the duration of schooling, it also recommends that children acquire
skills that they need to improve the quality of their social and economic later life
1
these efforts to reach the second priority among the Millennium Development Goals
(MDGs), 67 million primary school-age children did not receive education in 2009
(UNESCO, 2011). In addition, more than half of these children who are still out of school
are in only fifteen countries, one of which being Burkina Faso.2 Most of those children
reaching school will drop out of school before they complete the full six-year cycle of
primary school. Once more, in developing countries, students who finish primary school
often perform poorly on academic tests (Michaelowa, 2000; Aturupane et al., 2006) and
the quality of education may be low.
As observed, countries spend billions of dollars each year on schools, educational materials
and teachers, but little is known about the effectiveness of the expenditures in increasing
students’ years of completed schooling, and the skills they learn while in school (Glewwe
et al., 2011). In Burkina Faso, each year on average 29 percent of the GDP per capita is
spend per student in primary education. Nevertheless, until 2012, only 59.5 per cent is
obtained as primary school completion rate (MENA, 2013a). Also, skills assessment show
that students in the last grade (sixth grade) acquired as knowledge level 43.2 per cent and
41.2 per cent respectively in French and mathematics (MENA, 2013b).3 However, efforts
led the enrollment rate from 45.9 per cent in 2000 to 81.3 per cent in 2012. In spite of the
increased primary school gross enrollment ratio, 13.7 per cent of students drop out before
attaining the last grade (MENA, 2013a).
At present, the remaining question is to know how to improve education outcomes,
especially enrollment rate, student retention and academic performance. One of the
answers is that countries must mobilize both material and financial resources to develop
strategies in order to attain the EFA. Thus, the greatest challenge achieving EFA in
developing countries, particularly in Burkina Faso, is not so much getting children into
school, but increasingly keeping them there (UNICEF, 2012) and the quality of learning
(measured in terms of test scores). A review by Glewwe et al., (2011) shows that in the last
thirty years, many of studies on the relationship between school resources and education
outcomes are based on the impact of school and teacher characteristics on students test
scores. These studies have aimed to assess the extent to which school and teacher
characteristics have a causal impact on student enrollment and learning.
Beside these studies, some authors found that food for education (FFE) programs are
appropriate to improve school attendance and enrollment particularly among poor
population. School food service is attractiveness for parents to enroll their children and see
that they attend school regularly (Adelman et al., 2008; Levinger, 1986). Indeed, FFE aims
at motivating poor households in their investments on education by subsidizing some of
the costs of school participation. Grantham-McGregor, Chang et al. (1998) found that,
because the provision of school meals reduces the parents’ costs of sending children to
school, it promotes early enrollment and improving attendance. School feeding programs
(SFPs) can thus be a powerful instrument for achieving many multi-sectoral benefits such
as education, gender equality, food security, poverty reduction, nutrition and health, and
agricultural development. SFPs are intended to alleviate short-term hunger, improve the
2
In a decreasing order : Nigeria, Pakistan, India, Ethiopia, Bangladesh, Niger, Kenya, Yemen, Philippines,
Burkina Faso, Mozambic, Ghana, Brazil, Thaïlande, South Africa (UNESCO, 2011)
3
Tests objective is to evaluate students’ skills at the end of the schooling year for each grade of primary
school
2
nutrition and cognition of children, and transfer income to families (Jomaa et al., 2011).
SFPs are implemented in two forms: school meals or on-site meal and take home ration
(THR). According to the feeding program objective, a school may benefit from one or both
forms. Under school meals program, breakfast and/or lunch (possibly fortified with
micronutrients) is served at the school every school day. The only requirement to have
access to the meal is that the student be present. Both boys and girls were eligible for the
school meals intervention. Under THR, a student receives a certain amount of food staples
each period conditional on maintaining a specified attendance rate during that period.
Attendance records were maintained by the school administration, according to the
standard policies applied by the Ministry of Education. THR are targeted to girls. In
general, THR stipulated each month, each female student girl receives 10 kg of dry cereal
for household, conditional on a 90 percent attendance rate. Hence, THR focus relatively
more on improving food security at the household level (Pollitt, 1995). For some authors,
on-site feeding coupled with THR seems to have a greater impact on girls' enrolment
compared with that of boys (Gelli et al. 2007; Kazianga et al. 2009).
Empirical evidence suggests that in-school feeding has a positive impact on school
participation in areas where initial indicators of school participation are low. In-school
meal programs have been shown to have small impacts on school attendance rates for
children already enrolled in school (Adelman et al., 2008). Ahmed and Ninno (2002) find
that THR were effective in increasing enrollment and attendance in Bangladesh, but
academic performance was lower than in schools that did not benefit from the program.4
Levinger (1986) concluded in a previous study that SFPs do indeed increase enrollment,
but the impact on academic performances is mixed and depends on the local conditions.
Kazianga et al. (2012) found in the Sahel region in Burkina Faso that food for education
programs can increase enrollment, but are not always as successful in improving
attendance and academic performance among enrolled children. As for He (2009), he
found that World Food Program (WFP) assisted School Feeding Program in Sri Lanka
does not increase enrollment at any level compared to control schools.
As a result, the pathway as to how school feeding develops or hinders educational
outcomes is still not very clear (Ghosh and Saha, 2014). This calls for more investigation
of the circumstances under which SFPs could increase enrollment, attendance and also
improve academic performances. This research attempts to address the gap by more
enlighten the role of a SFP on education outcomes. The present study, therefore, takes
account of these arguments and evaluates the significance of a particular School Feeding
Program in improving school participation among primary school children. In particular
this study assesses the impact of THR in comparison with on-site canteens called daily
meal (DM). This research will focus on issues of school feeding impact on education
outcomes in Burkina Faso. As the program included both DM and the provision of THR in
selected schools, this analysis is further able to consider specific aspect of SFP design.
More specifically, it aims to address the extent to which THR affect 1) students attendance
rate and 2) school enrollment particularly for girls.
School canteens were first introduced in the 1960s in Burkina Faso by the Catholic Relief
Service (CRS)/Cathwell during severe famine which affected the Sahel region of West
4
Cited by Kazianga et al.(2009)
3
Africa and several school feeding programs have since been implemented since this period.
However little is known of the evidence of school feeding programs because there were no
effective evaluations of their impact on education outcomes and student learning in
particular, Burkina Faso. The first known attempt is Moore’s report.5 Moore (1994)
evaluated the impact of CRS’s school canteens on education outcomes. She found that the
presence of a school canteen was associated with increased enrollment, regular attendance,
decreased dropout rates, and increased score on national exams in Burkina Faso. More
recently, Kazianga et al., (2012) analyze under World Food Program the two schemes of
school feeding in the Burkina Faso Sahel region but results are not conclusive. Authors
find that SFPs improved enrollment but no improvement in attendance and academic
performance. In these studies, authors compare either on-site meal either on-site meal and
THR schools compare to schools which had no feeding program.
The present study differs from the previous by comparing THR schools (as an additional
intervention) with DM or on-site schools. Specifically the research is based on a context
where all schools receive DM and only few of them receive an additional intervention, the
THR. Thus, the present study aims is to evaluate the impact of this additional scheme on
educational outcomes. The study will contribute to orientate policy makers on the
effectiveness of school feeding programs by assessing their attainment in a poverty
context. Results also have broader implications for the literature on effort to improve
educational outcomes through the implementation of SFPs. The study uses data drawn
from the CRS’s School Feeding Program in Burkina Faso which provided assistance to
more than 130 000 students in primary schools (grade 1-6). Specifically, the program was
phased in across two provinces (of 3 in total) in the center northern region in the 20112012 school year. Two schemes were introduced from only DM to a complementary THR
for girls in schools where their enrollment rate is below 40 per cent. This allows us to make
comparison between the two schemes of program intervention.
We find that THR increased attendance for all students but while enrollment increased for
girls, it decreased for boys. It means that THR leads parents to send more girls in school
but pull up boys for labor because the latter do not benefit from THR. In addition more
girls mean more food for household. We perform robustness checks on the results
including regression on a second comparison group constructed with the third province in
the northern region where no school received a feeding program.
The remainder of the paper is organized as follows: Section 2 provides a review of the
literature surrounding theoretical and empirical research on school feeding program;
Section 3 gives a background of food for education and the feeding program design in
Burkina Faso; Section 4 summarizes the data and descriptive statistics; Section 5 reports
on econometrics methods used; Section 6 discusses the estimations results and robustness
checks; and Section 7 concludes.
2. Literature review
It has been claimed that School Feeding Programs (SFPs) increase school participation
among poor and food insecure group of people. Three goals are associated with SFPs as
5
This report is unpolished, so we do not have more detail on how author conducted her study
4
the pathways by which school meals could affect student learning (Levinger, 1986;
Kazianga et al., 2009; Bundy et al., 2009). Firstly, SFPs are a conditional transfer to
students. It may induce families and motivate parents to enroll their children, to enroll
them sooner or conditional on enrollment, to encourage regular attendance. Secondly, SFPs
improve the nutritional status of school age children over time, and alleviate short-term
hunger in malnourished or otherwise well-nourished schoolchildren. As malnurishment has
been shown to affect learning (Tara, 2005), SFPs can be expected to improve educational
outcomes. Thirdly, SFPs improve cognitive functions and academic performance via
reduced absenteeism and increased attention and concentration due to improved nutritional
status. Indirectly, by increasing the amount of food available to the household, SFPs could
improve the nutritional status of household members who are not in school, especially
when SFPs entail take home rations. In this way, SFPs are appealing because if properly
designed and implemented they lead to increased number of children being enrolled with
better academic performance (Kazianga et al., 2009).
In general, two schemes constitute school feeding program and each scheme has its
specific values. Meals served at schools go directly to the students who are supposed to
benefit from the program. However, parents could react by reallocating food in the
household away from these children. Food received by the household under THR is more
likely to be shared by other household members, possibly reaching children who may be in
as much or greater need of additional food. As For Kazianga et al., (2012) show, because
the nutritional benefits are diluted within the household, the impact of food on learning
outcome as academic performance may be lower in THR than with a school meals
program.
Many works have attempted to confirm school feeding goals on educational outcomes and
also on health status of school student. Studies consider generally the following elements
as school outcomes: enrollment, attendance, tardiness, classroom behavior, cognition,
grade repetition, attainment levels, and drop out. Previous empirical works have found
mixed evidence for the impact of school feeding at the same time on enrollment,
attendance and academic performance. Some evaluations of FFE programs have shown
that FFE programs can lead to increased access (of girls in particular), reduced dropout,
particularly in the lower primary school grades, and improved student’s learning (Drèze
and Kingdon, 2001; Ahmed, 2004; Taras, 2005; Vermeersch and Kremer, 2004;
Kristjansson et al., 2007).
Impact of food for education (FFE) programs on enrollment and attendance
Results are most compelling for school enrollment and attendance, particularly where
initial rates of participation are low. Ahmed and del Ninno (2002) used a non-experimental
design to assess the food for education (FFE) program set up in Bangladesh designed to
transfer food to the poorest households through THR programs in primary schools.
Authors found that the enrollment have increased by 35 percent over the two years period
between the program start and after its first year. This increase was driven by a 44 percent
increase in girl’s enrollment and by a 28 percent increase for boys. Ahmed and del Ninno
(2002) also looked at the dropout rate as affected by the program, and found that from
1999 to 2000, 15 percent of students from households who did not receive rations dropped
out while only 6 percent dropped-out among those receiving the THR ration. Under THR
program, food received is to be shared by other members in the household where students
5
live, so the using of means as method to evaluate the impact allow to control external
factors that can influence the effect of the program.
Using experimental design, Ahmed (2004) conducted a study in food insecure areas of
Bangladesh to see the impact of SFP on school participation. The author found that SFPs
have statistically significant positive impacts on both gross and net enrollment rates with
14.2 percent and 9.6 percent increases respectively. Also, the program increased class
attendance of participating students by 1.34 days per month. However, this finding does
not take account of other unobservable characteristics of households in the treatment area
that could affect household’s decision to enroll children. Therefore, it appears inconclusive
to claim that the difference in enrollment between treatment and control groups was the
result of the program without considering unobserved factors.
Afridi (2007) examined by non-experimental design the feeding program effects on school
enrollment and attendance. Using difference-in-difference estimation the author showed
that girls’ attendance increased by 10.5 percent in schools which implemented the SFP in
grade 1 in Madhya Pradesh (India). In Burkina Faso, Kazianga et al., (2009) found in their
study of “Girl-Friendly School” in BRIGHT schools that both THR and school meals
interventions had a statistically significant impact on the overall enrollment and the
enrollment of girls.6 The reviews by Bundy et al. (2009) also find that the provision of
FFE programs increases the access to learning and education for schoolchildren by
improving enrollment and attendance rates. However, Buttenheim et al. (2011) did not
found by a quasi-experimental design consistent effect of school feeding programs in Lao
PDR. Indeed, they found minimal evidence that the school feeding schemes increased
enrollment or improved children’s nutritional status. Cheung and Berlin (2014) using
difference-in-difference method found that school enrollment increased but the impact was
largest from the full program including on-site feeding, THR and de-worming.
Impact of food for education (FFE) on student performance
Generally, student performance is evaluated by cognitive tests. The scores are used to
assess level of knowledge students acquire at school comparing to the corresponding
grade. More specifically, cognitive tests are based on mathematics, reading and language.7
However, evidence of impact of school feeding on learning achievement and cognitive
function is also hard to find (Buttenheim et al., 2011). Studies have shown significant
impact in one but not multiple domains, e.g., increased math but not language scores or
vice versa (Kazianga et al., 2009; Kristjansson et al., 2007 and Ahmed, 2004; Tan et al.,
1999).
In the Bangladesh study, Ahmed (2004) tested grade five students from the treatment
groups to see the effects of fortified snacks on school performance. Total test scores
improved by 15.7 percent in the treatment group with mathematics improving by 28.5 per
cent in the treatment group over the control. In the same way, Meyers et al. (1989) by nonexperimental design found significantly greater improvements in total scale score and
language sub score, while marginal improvements were seen in mathematics and reading
6
Burkinabé Response to Improve Girls’ cHances to Succeed. The BRIGHT program placed relatively
wellresourced schools with a number of amenities directed at encouraging the enrollment of girls in
132 rural villages in Burkina Faso
7
French or English depending on the country official language or any local language in the country
6
sub scores. Also, Kristjansson et al. (2007) and Jomaa et al. (2011) find that mathematics
tests were consistently positively impacted by SFP. Tan et al. (1999) by using quasiexperimental design found significant effect in math and language. They saw that math and
English test scores are improved due to the school feeding programs. Afridi et al.,(2013)
by using difference-in-difference method analyzed in India the effects of school meal
intake on the cognitive effort of students within classroom. Authors found that the
provision of meals significantly improved the classroom effort of students in grade seven.
As to Adrogué and Orlicki (2011), they analyzed a SFP conducted in Argentina national
public schools. In their non- experimental design, they found positive impact of the SFP on
school scores for students in grade three. The language scores for students increased by
0.15, but there was no significant effect on mathematics scores. Jacoby et al., (1998) also
in non-experimental design found no significant effect in both mathematics and reading
test. But using the same design, Meyers et al., (1989) found significant impact in math and
reading, language and math. Likewise, Kazianga et al. (2009) found no significant impact
on performance in mathematics in their randomized control trial in Burkina Faso. In
Senegal, Diagne et al. (2014) using experimental design in rural primary schools find that
school canteens have a positive and significant impact only on the average score of student
in grade 2 in both mathematics and French. But, overall in French, they find that the
program increases the average score of all primary students by 5.6 percentage points.
However, the effect disappears when authors add control variables. As opposed to French,
the canteen significantly improves the average score of students in mathematics by 6.35
points without control and 6.32 points with control.
3.
Background of Food for Education Programs in Burkina Faso
School canteens were first introduced in 1962 in Burkina Faso by the Catholic Relief
Service (CRS)/Cathwell in the aftermath of severe famine which affected the Sahel region
of West Africa. Since this period, CRS has provided educational assistance and SFPs in
vulnerable areas by implementing several SFPs. Also, dry THR which is a more recent
intervention also initiated in Burkina Faso by the CRS/Cathwell. Students who attend
school on a regular basis receive a food ration (flour) that they can bring back home each
month. THR are targeted only to girls. However little is known of the evidence of school
feeding programs because there were no effective evaluations of their impact on education
outcomes and student learning in Burkina Faso. Our present study aims to address this gap.
Our study covers the region served by the CRS, and all schools which were listed in the
academic year 2011-2012. The focus of this study is the center northern region of Burkina
Faso. Northern Burkina Faso is an appropriate context to evaluate the impact of FFE
programs for two main reasons. First, the region has low primary school participation in
the country. On average only 53.5 per cent of school age children (6 to 11 years old) attend
school (MENA, 2012). Therefore there exists a large scope for increasing enrollment.
Second, income levels are very low and severe food shortages are frequent. Hence, the
value of the food offered should be a sufficient incentive to attract children to school.
Households are largely dependent upon subsistence agriculture, and malnutrition is
extremely high in the target area, with stunting in 40% of children under age five due to
7
diet, poor hygiene practices, and illness (ENAIM, 2009).8 The project is implemented in
two of the three provinces of the northern region: Bam and Sanmatenga.
The two provinces are characterized by periods of infrequent rain, which result in food
insecurity and increasing migration. Following the poor harvest of 2011, the Northern
region is declared food insecurity risk area. Additionally, Bam and Sanmatenga are
characterized by low levels of girls’ educational enrollment and achievement. In 2008,
school enrollment was 75% with gender disparities (81% boys and 70% girls) and
inequities between urban and rural areas (MENA, 2009). There are many challenges in
access to education, including prohibitive school distances, financial costs, cultural barriers
and the opportunity costs of sending girls to school who are expected to perform household
chores and look after other children (as siblings). These factors contribute to a high number
of dropouts at an early age. The grade five drop-out rate is 17 per cent in Bam and 15 per
cent in Sanmatenga, often due to early marriage, puberty (and lack of proper sanitation
facilities in schools) and work duties at home. In the last few years, digging for gold has
become widespread in the two provinces. This phenomenon has increased the pull of
children from school, and households from their crops. In year 2011, a new mine was
implemented in Bam province which can hinder SFP impact.
Through its latest project called “Beoog Biiga” (“Tomorrow Child” in local language),
CRS aims to respond to food insecurity and increase school access and continuation by
improving student health and the school environment in Burkina Faso. The project’s goal
is to improve food security through the sectors of education, health and capacity building.
This mutli-sectoral project was funded from September 30th, 2011 to December 31st, 2014
by USDA and is implemented in partnership with the government of Burkina Faso and
local development organizations.9 The implementation of the SFP and health initiatives is
in close collaboration with the Ministry of Primary Education and Literacy (MENA), the
Ministry of Health (MoH), the Ministry of Social Action and National Solidarity
(MASSN).
The project targets the provinces of Bam and Sanmatenga, and included in its first year 684
schools and 134,128 students of whom 62,442 girls. The project covers all schools in the
two provinces served by the CRS in the academic year 2011-2012.10 In this way, two main
activities are carried out as project schemes under the project objectives. First, CRS
distributed a DM to all students during all school year. Primary school students received a
daily ration of 136 grams (g) of soy-fortified bulgur, 27 g of lentils and 18 g of vegetable
oil per student, for a total of 726 kilocalories (kcal) and 31 g of protein per day. The second
main activity is the distribution of THR. CRS provided THR to improve girls’ enrollment
and attendance and decrease drop-out rates in Bam and Sanmatenga. In each school where
girls’ enrollment rate is under 40%, female students are given a food ration consisting of
10 kilograms (kg) of corn soy blend (CSB) for each month in which they have attendance
above 90%. According to current figures, approximately 150 schools (excluding schools in
the 2 big cities) have girls’ enrollment rates of less than 40% and are benefit from THR.
We want to stress at this point that the project “Beoog Biiga“is not a randomized
intervention. The schools were selected based on administrative criteria, which may
8
National Food Security Survey
United States Department of Agriculture (USDA)
10
The 2011-2012 school year is from October 2011 to June 2012
9
8
correlate with other characteristics potentially influencing school enrollment and student
attendance. Additionally, the local community was asked (not a compulsory duty) to
provide some wood or help preparing the food to complement the meals might also induce
self selection. As the program was implemented at school level, the potential biases will be
negligible (when a student brought or not wood, he/her received the DM at school).
4. Data and descriptive statistics of the SFP
4.1. Data
The data used in this study will come from two main sources: CRS/Burkina Faso and
Ministry of Primary Education and Literacy of Burkina Faso (MENA). Schools
characteristics data are drawn from the MENA yearly school survey (2010-2011 and 20112012 school years). The survey dataset includes information on schools location, status,
number of teachers by gender, and other school facilities. Through the project “Beoog
Biiga”, CRS provides data on education outcomes such as enrollment and attendance.
Attendance is measured by the average number of half days of courses not missed by
students in each school. Enrollment rate in particular girl enrollment is measured by the
percentage of female students in each school. CRS’s project data for baseline was collected
in 2011 prior the beginning of school year.
Schools are selected by steps. The step 1 consists to divide schools in two groups; the first
group is composed by all schools located in urban area and the second group by all schools
in rural area. At step 2, within rural schools, all schools where girls’ enrollment rate is
below 40 percent are selected to benefit from THR and the rest of schools received only
DM. The baseline is constructed using data from primary education ministry based on year
2010-2011 and 2011-2012 school years. This baseline consists to collect school
characteristics to complete data on enrollment and attendance already collected in CRS
database. Then we matched the two datasets by schools name at distric level. Table 1.1
reports all schools characteristics such as school facilities, location and school status. To
all these variables, we add an exogenous variable to capture its impact on schools
enrollment and students attendance. Indeed, as notice in previous Section 3, Bam and
Sanmatenga provinces are affected by gold digging which can hinder the program impact.
Taking account this factor would able us to not mislead the program impact estimation.
We identify our treatment and comparison group of schools on the basis of whether they
receive only DM or DM and THR at the 2011-2012 School Year. As describe below, after
step 2 of schools selection, all schools which received DM and THR for girls form our
treatment group while schools which received only DM form our comparison group of
schools. As in the CRS dataset, it is showed that SFP was functional in almost all schools
after 4 months of the the beginning of school year, these first four months constitute the
pre-treatment period and the rest of the school year constitutes the post-treatment period.11
4.2. Descriptive statistics in baseline
Table 1.1 summarizes at school level the key characteristics of schools and students at
11
The school year 2011-2012 began from October 2011 to July 2012.
9
baseline in all targeted schools. The summary statistics in Panel A show that schools are
characterized by low attendance rate (51 per cent). On average, girl enrollment rate is
45.67 per cent in each school showing that gender gap persists until now (about 0.840).
Also, 92 per cent of schools are located in rural area and 88 percent of them are public
schools. In addition, only 5.5 and 7 per cent of these schools have respectively library and
latrine. Pupil-teacher ratio is 58.33 showing that on average in each class in each school we
have 58 students. According to the education for all Fast-Track Initiative (EFA-FTI), the
standard ratio must be 40.12 As result, schools classes are oversize. Only 10 per cent of
schools have access to electricity and 5 per cent have running water for students. In Panel
A, 36 per cent of schools are affected by the mining and 74 per cent of them have around
an external restaurant meaning that prior to the SFP implementation some students can
have breakfast or lunch at school.
Panel B shows the students characteristics in a sub-sample. The MENA randomly
interviewed in selected schools students in grade 6 and grade 3, for details on their socioeconomic characteristics such as parents’ occupation, parent literacy, the distance from
school and household chores. The panel B also shows that on average, students are living
at 1.62 km from school. More, Panel B shows that 82 per cent of students have a farmer
father and only 46 per cent of them are literate. It appears here that mothers are less
literate. Indeed, only 36.38 per cent of students have their mother who can read or write in
French or local language. Panel B suggest that more than 78 per cent of students declare
having household chores before and after school and 60 per cent of them have to keep
siblings at home after school.
Table 1.2 reports the average schools characteristics of the treatment and comparison
group at baseline. Prior to the treatment, schools are similar on some variables including
attendance level for boys and girls, pupil-teacher ratio. Also, before treatment, it appeared
that there is no significant difference between schools by the presence of electricity,
latrine, library, and external restaurant in schools. However, we observed significant
difference on girl enrollment, school status and location, number of female teachers. On
average, 36 per cent of girls are enrolled in THR schools against 48 per cent in comparison
group. This is an evidence because school received THR based on criteria on girl
enrollment rate (which is below 40 per cent). At the baseline 23.1 per cent of schools in
THR and 38.9 per cent in comparison group are affected by mining. More, Table 1.2 has
shown that at baseline all THR schools are located in rural area while 89.6 per cent of
comparison schools are located in the same area. We have 82.4 per cent of THR schools
against 82.4 per cent of comparison schools which are public.
5. Methodology
This study attempts to evaluate the impact of a particular SFP on education outcomes. The
primary purpose of impact evaluation is to determine whether a program has an impact on
a few key outcomes, and more specifically, to quantify how large that impact is. School
feeding impact evaluations aim to measure the differences in the outcomes attributable to
school feeding. This involves comparing the outcomes for beneficiaries of school feeding
to the outcomes from a control or comparison group not receiving the intervention. Any
12
FTI is established in 2002
Available on : http://www.oecd.org/dac/37819963.pdf
10
impact evaluation attempts essentially to answer a counterfactual question (Duflo and
Kremer, 2003): how would individuals who participated in the program have fared in
absence of the program? How would those who were not exposed to the program have
fared in the presence of the program? The difficulty with these questions is immediate: at a
given point in time, an individual is observed to be either exposed or not exposed to the
program.
The critical objective of impact evaluation is therefore to establish a credible comparison
group, a group of individuals who in the absence of the program would have had outcomes
similar to those who were exposed to the program (Duflo and Kremer, 2003). This group
should give us an idea of what would have happened to the members of the program group
if they had not been exposed, and thus allow us to obtain an estimate of the average impact
on the group in question. Assessing the impact of any intervention requires making an
inference about the outcomes that would have been observed for program participants if
they not participated (Smith and Todd, 2005).
Denote by Y1 outcome conditional on participation and by Y0 the outcome conditional on
non-participation, so that the impact of participating in the program is:
  Y1  Y0
For each individual, only Y1 or Y0 is observed, so ∆ is not observable. This missing data
problem lies at the heart of the evaluation problem. So, let T = 1 for the group of
individuals who applied and got accepted into the program for whom Y1 is observed and T
= 0 for individuals who do not enter the program for whom Y0 is observed. Let X denotes a
vector of observed individual characteristics used as conditioning variables. The most
common evaluation parameter of interest is the mean impact of treatment on the treated
(TT):
TT  E X , T  1  EY1  Y0 X , D  1  EY1 X , T  1  EY0 X , T  1
TT estimates the average impact of the program among those participating in it.
(1)
It is useful to stress at the point that our SFP was not a randomized intervention. THR
schools are selected based on administrative criteria which may correlate with other
schools characteristics potentially influencing enrollment. Based on these criteria, schools
are divided into two groups. The first group counts 134 schools where all students received
DM and girls in additional receive a THR constitute our treatment group. The second
group includes 550 schools where students only receive DM constitute our comparison
group.
According to the literature and the nature of data at hand, this research uses a quasiexperimental design to evaluate the impact of THR program education outcomes in
Burkina Faso. Indeed, the present study analyses a program which already took place and
for which school participation was not randomized as all schools in both provinces
participate either in at least one of SFP schemes. Because the program was offered at the
school level, we estimate the average intent to treat (AIT), which is the impact of the
program, on average of all students in a given school.
𝐴𝐼𝑇 = 𝐸(𝑦1𝑖 |𝑇𝑖 = 1) − 𝐸(𝐸(𝑦0𝑖 |𝑇𝑖 = 1)
Or
(2)
11
𝐸(𝑦1𝑖 |𝑇𝑖 = 1) = 𝐸(𝐸(𝑦0𝑖 |𝑇𝑖 = 0), so
𝐴𝐼𝑇 = 𝐸(𝑦1𝑖 |𝑇𝑖 = 1) − 𝐸(𝐸(𝑦0𝑖 |𝑇𝑖 = 0)
The aim of our study is to assess the impact of an additional scheme of school feeding on
educational outcomes. It consists to analyze the effect of THR on students’ enrolment and
attendance. Ideally, the school feeding formats would have been randomly assigned to
schools, so that any identified differences in outcomes could be attributed to the school
feeding programs (Heckman and Smith, 1995). However, the design of the present study
offers several challenges to the suitable identification between control and beneficiary
groups. Given these challenges, and the availability of pre- intervention (baseline) and
post-intervention (follow-up) data, we are able to use difference-in-difference to estimate
the THR impact on enrollment and attendance. Difference-in-difference (DID) estimation
on attendance and enrollment rate can be written as:
𝑌𝑖 = 𝛽0 + 𝛽1 𝑡𝑖 + 𝛽2 𝑇𝐻𝑅𝑖 + 𝛽3 𝑡𝑖 ∗ 𝑇𝐻𝑅𝑖 + 𝛽𝑘 𝑋𝑘,𝑖 +∪𝑝 + 𝑒𝑖
(3)
Where 𝑌𝑖𝑡 is the outcome of interest (attendance or enrollment) for school𝑖.
𝑡𝑖 takes value 1 for all schools if observation is in follow up and 0 for baseline. 𝑇𝐻𝑅𝑖
takes value 1 for all schools where all students received DM and girls received THR and 0
in schools where all students received only DM. 𝑋𝑘,𝑖 is a school characteristic, ∪𝑝 is
province specific factor. The interaction 𝑡𝑖 ∗ 𝑇𝐻𝑅𝑖 estimates the difference-in-difference
(DID) effect of THR on school attendance and enrollment.
The present study makes a contribution to the quasi-experimental literature in impact
evaluation in developing area. Indeed, the use of a retrospective analysis to evaluate a
school feeding program differ of previous study in a developing country like Burkina Faso
where (Kazianga et al., 2012; Kazianga et al., 2013) have been used randomized design.
6. Results
We now discuss our results from estimating equation (3). Table 1.3 and Table 1.4 show
respectively the effect of THR on attendance and enrollment including control for all
school characteristics. The coefficient of interest is the interaction term THR x Year1
which is the DID estimate of the effect of THR on school enrollment and attendance. For
all regressions, we estimate the effect of THR on attendance and enrollment with all
schools thus with only rural schools. Given that, as shown in descriptive statistics, THR
schools are all rural, the choice of the right comparison group follows this criterion in
order to avoid some biases in the program impact. Therefore, we restrict our interpretation
to rural schools only. But we present the results with all school and we observed that
results are similar.
6.1.Impact of THR on school attendance
Attendance is measured by the average number of half days of courses not missed by
students in each school as reported by the CRS’s survey. Table 1.3 presents the effect of
THR on attendance at school level. While columns 4 shows attendance rate for all
students, columns 5 and 6 report boys and girls attendance rate respectively. The DID
12
results prove that on average THR program has positive impact on students’ attendance
rate. Indeed, the attendance rate increased by 8.4 per cent more in THR group. When
estimating separately, both boys and girls attendance increased and boys’ attendance rate is
better-off than girls’. Girls’ attendance rate increased by 6 per cent against 8.4 per cent for
boys showing that school attendance in THR is driven by boys.
The main insight here is that attendance rate has a marginal positive effect on attendance.
As Table 1.2 reports, at baseline attendance is at a low level and not different between the
THR and DM group of schools. But at the end of school year, attendance rate increased in
both group with THR. While controlling for schools characteristics, we find that pupilteacher ratio has a negative and significant effect on overall and boys’ attendance. Public
school has a positive impact on attendance by increasing all students’ attendance by 9 per
cent. But its effect is not significant on girls’ attendance. At province level, we control for
the presence of an additional exogenous factor which affect student and their household
particularly vulnerable ones. Indeed, mining newly opened can lead to absenteeism and
thus drop out. More results show that this factor impacts negatively attendance by
decreasing respectively boys’ and girls’ attendance by 25 per cent and 14 per cent.
6.2.Impact of THR on school enrollment
Table 1.4 reports the impact of THR on enrollment. Columns 9 and 10 report respectively
girls and boys enrollment rate. By girl enrollment, we mean female enrollment rate within
schools after program implementation. Initially, the THR program objective is to increase
girl enrollment rate within school. Results revealed that girls’ enrollment rate increased
significantly with THR by 3.2 per cent. However, the impact is negative on boys’
enrollment. Indeed, boys’ enrollment rate decreased by 3.2 per cent in THR group. We
note that girls’ enrollment may increase due to the fact that their number increased or boys
number decrease. To check with this, we also compute by the change in the students’
number within schools. Columns 4, 5 and 6 report respectively the change in the number of
students within school, the change in girls’ number and the change in boys’ number.
Nevertheless, these numbers will be interpreted with caution. Indeed, in some case,
enrollment numbers cannot be trusted because the schools might have incentives to boost
them in order to receive more funds. On the other hand, it is possible for a child to attend
without being enrolled, perhaps because of incomplete school records (Cheung and Berlin,
2014). Column 4 shows that students’ number increased on average by 11 students per
school. While girls’ number increased significantly, boys’ change in number is not
significant. More, summarized statistics in Table 1.5 (figures in columns 2 and 5) show us
that we have mixed effect. While enrolled girls number increased in THR, enrolled boys
decreased at the same time. Thus we conclude that girls’ increased enrollment rate is
driven more by the increasing of their number within schools.
In Table 1.4, figures suggest a certain substitution between girls’ enrollment and boys’
enrollment. When girls enrollment rate increased by an amount, boys’ enrollment rate
decreased by the same amount. Among control variables, public schools have a positive
and significant impact on number of enrollment. However, while the impact is positive and
significant on girls’ enrollment rate, the effect is negative on boys’ enrollment rate. For
pupil-teacher ratio, the effect is positive on number of enrollment but the impact is not
significant on enrollment rate. A high ratio may indicate that number of students per class
13
is very great and thus can discourage some students from participating and thus avoid new
enrollment. Female teachers have a positive and significant impact on girls’ enrollment
rate while the effect is negative for boys. When there are more female teachers in school, it
leads more enrollments of girls. In presence of female teachers, parents have more
confidence to send their girls in school. A female teacher may be considered as mother and
thus will take care of students. Also in future, girls will become like this teacher i.e. more
literate and get a good job. Another result is that mining presence decreases the number of
boys by 3 students per school and thus boys’ enrollment rate decreased in the same time by
2 per cent.
Overall, the results can be explained by the fact that THR can be considered as a
reallocation in food between girls and their household members. Household receive more
food to be share with all members and this could lead parents to enroll other girls not yet in
school in order to increase the food ration. At the same time, parents may retain boys for
labor either in crop or in mining in order to increase household resources. These results
have been shown previously (Levinger, 1986; Kazianga et al. 2012). In the same way,
Ahmed and Ninno (2002) find that THR were effective in increasing enrollment and
attendance in Bangladesh. Authors found that the enrollment have increased by 35 per cent
over the two year period between the program start and after its first year. This increase
was driven by a 44 per cent increase in girl’s enrolment and by a 28 per cent increase for
boys. Kazianga et al., (2009), found that THR increased enrollment for girls by 6
percentage points. Cheung and Berlin (2014) find that THR boosted the school enrollment
in the short run by 5 per cent.
6.3.Robustness
Given that as describe earlier, all schools in our sample received at least one scheme of
SFP, we use in addition an external group of comparison to give robustness to our results.
Indeed, the program which was implemented in two of the three provinces of the Northern
region, we can use the third province where no school received the program as another
comparison group, the non feeding (NF) schools. Prior to use the third province, we have
to ensure that this province can be considered as similar as the other provinces in order to
form a good NF comparison group. On the one hand, the administrative zoning in Burkina
Faso consists to form regions with provinces based on their geographic and socioeconomic characteristics. This suggests that the third province is comparable to the other.
Table R0 on the other hand, shows that on average the third province is similar to the
targeted ones. Columns 2 and 3 show that girls’ enrollment rate, proportion of rural
schools, and the number of schools where girls’ enrollment is low (below 40 per cent as
defined in program criteria) are similar. In this section we use in two different ways the NF
schools as second comparison group.
Firstly, we had all the NF schools with the previous DM group and run the DID estimation
on the number of new enrolled students. Results in Table R1 show that we obtain the same
results as our main results in Table 1.4 (with only DM as comparison group). This means
that taking account the NF schools of the third province do not change our results. So THR
has no effect on NF schools in the external province. The previous findings reflect well the
program impact on enrollment in terms of new enrolled. Indeed, students’ number
increased significantly by 12. While the change in number for girls is significant (6.53), the
change in boys’ number is not significant. Also, public schools increased the number by 24
14
new students (against 22 in Table 1.4).
Secondly, our robustness check consists to compare THR schools and NF schools where
girls’ enrollment rate is below 40 per cent. In Table R2, results show that girls enrollment
rate increased by 3.8 per cent (against 3.2 in Table 1.4) while for boys it decreased by the
same value. Controlling for public school, we find that public schools increased girls’
enrollment rate by 1.8 per cent and boys’ enrollment rate decreased by 1.8 per cent. As
shown in Table 1.4, THR program has the same results on enrollment. So, the value of the
coefficient of interest THR x year1 we found does not differ from our main results and the
impact of THR on educational outcomes is causal.
More, for robustness check, we also compare only schools where girls’ enrollment is
below 40% i.e. THR schools in rural with DM in urban zone (which did not receive THR
for girls because located in urban area). Table R3 shows that the results do not vary widely
from results found in Table 1.4. Indeed, girl’s enrollment rate increased by 2.4 per cent
(against 3.2 in Table 1.4) while boys’ enrollment rate decreased by 2.4 per cent. Overall,
NF group of school allow us to corroborate the study main results on enrollment: THR
program improve school enrollment.
7. Conclusion
This study provides an ex-post evaluation of a food for education (FFE) program
implemented in Burkina Faso. It is an insight on the impact of an additional scheme of
feeding on educational outcome. Specifically, the study evaluates the impact of take home
ration (THR) on school attendance and girls’ enrollment in Northern primary schools. THR
is targeted only for girls where their enrollment rate is below 40 per cent conditional on 90
per cent attendance. As we rely on a baseline and follow up, we use difference-indifference (DID) regression to estimate the impact of the THR program. Because we have
not experimental data we control for schools and province level characteristics to find an
estimated impact which can be interpreted as causal. We find that attendance rate within
schools increased by 8.4 per cent more in THR group (6 per cent for girls and 8.4 per cent
for boys). Also, results show us that girls’ enrollment rate increased by 3.2 per cent but
boys’ enrollment is decreased by the same percentage suggesting a substitution effect
between boys and girls in THR schools. More, our results suggest that schools
characteristics influence the extent to which THR improves school attendance and girls’
enrollment. Students in schools that had more female teachers and in public schools gained
significantly more from the program.
Overall, our results show that school feeding through THR programs in a specific context
of food insecurity can increase school attendance and girls’ enrollment. However, the
impact on nutrition and health of this program remains to be investigated. Moreover, given
that the THR is targeted at school level, this calls for more investigation at household and
individual level of the circumstances under which THR impacts on attendance and
enrollment. These are open questions for future research. The findings of this research
have policy relevance- THR for girls can improve school attendance and enrollment within
schools. This carries implication for the long term gender equality in school and
educational attainment of girls.
15
Tables
Table 1.1: Summary statistics of schools characteristics at baseline in targeted
provinces
Variables
Mean
SD
Public schools
0.881
0.322
Rural schools
0.917
0.275
Student attendance
0.506
0.218
Enrolled students
216.68
140.97
Girl enrollment rate
0.456
0.079
Gender gap
0.881
0.306
Teachers
3.53
2.33
Female teachers
1.35
1.64
Pupil/teacher ratio
58.33
19.29
Schools facilities
Electricity
0.099
0.299
Running water
0.542
0.498
Library
0.055
0.229
Latrine
0.644
0.448
Parents association
0.902
0.297
External restaurant
0.742
0.437
Mining area in 2011
0.358
0.479
Students characteristics (sub-sample )
Distance from schools 1.625
0.668
Repeaters
0.376
0.515
Keeping child at home 0.604
0.489
Household chores
0.785
0.410
Farther is farmer
0.820
0.384
Father is literate
0.463
0.499
Mother is literate
0.293
0.456
Min
0
0
0
18
0.18
0.22
0
0
11.5
Max
1
1
0
0
0
0
0
0
0
1
1
1
1
1
1
1
1
0
0
0
0
0
0
3
1
1
1
1
1
1
755
0.80
4
11
7
154
Notes: Panel A shows summary statistics for the 684 targeted schools where are enrolled
134128 students of whom 62442 females. Panel B shows students characteristics obtained
for sub-sample of 876 students (grades 3 and 6) in 24 randomly selected schools by
MENA.
16
Table 1.2: Average schools characteristics by treatment status
School characteristics
Public schools
Rural location
Students attendance
Girl attendance
Boy attendance
Girl enrollment rate
Pupil/teacher ratio
Female teachers
Drilling
Electricity
Latrine
Library
External Restaurant
New mining area
DM
THR
N=550
(1)
0.896
0.896
0.509
0.767
0.510
0.480
57.827
1.520
0.625
0.119
0.724
0.061
0.826
0.389
N=134
(2)
0.824
1.000
0.496
0.774
0.496
0.362
60.256
0.687
0.512
0.078
0.674
0.062
0.806
0.231
Difference
(2) – (1)
-0.072 **
0.104***
-0.014
0.007
-0.014
-0.119***
2.429
- 0.833***
-0.113**
-0.041
-0.050
0.001
-0.020
-0.158***
Notes: Summary statistics for schools targeted in the 2011-2012 school year.
Standard errors not presented. *** Significant at 1% ** significant at 5% * significant at
10%.
17
Table 1.3: Program impact on attendance
Baseline
THR
THR*year1
Public school
Rural zone
Pupil/teacher
ratio
Female teacher
Electricity
Latrine
Running water
Library
Ext. restaurant
Girl enrol.<40
Mining
R-squared
All schools
(1)
All
0.648**
(0.041)
-0.045*
(0.046)
0.088***
(0.028)
0.081***
(0.023)
0.033
(0.025)
-0.001***
(0.000)
-0.000
(0.004)
0.028
(0.019)
-0.021
(0.014)
-0.003
(0.013)
-0.016
(0.025)
-0.083***
(0.018)
-0.012
(0.019)
-0.245***
(0.012)
0.5636
(2)
Girls
0.820***
(0.036)
-0.008
(0.042)
0.064**
(0.026)
-0.004
(0.021)
0.045**
(0.023)
-0.000
(0.000)
-0.001
(0.004)
0.030*
(0.017)
-0.010
(0.0183)
0.004
(0.012)
-0.008
(0.023)
-0.011
(0.020)
0.020
(0.121)
-0.146***
(0.011)
0.1910
(3)
Boys
0.647***
(0.041)
-0.045*
(0.024)
0.088***
(0.028)
0.082***
(0.023)
0.034
(0.025)
-0.001***
(0.000)
-0.000
(0.004)
0.029
(0.019)
-0.021
(0.014)
-0.003
(0.013)
-0.015
(0.025)
-0.082***
(0.018)
-0.012
(0.019)
-0.245***
(0.012)
0.5638
Rural schools
(4)
(5)
All
Girls
0.668***
0.863***
(0.020)
(0.034)
-0.046*
-0.008
(0.006)
(0.022)
0.084***
0.060**
(0.028)
(0.026)
0.089***
-0.006
(0.025)
(0.022)
(6)
Boys
0.668***
(0.037)
-0.046*
(0.025)
0.084***
(0.028)
0.090***
(0.025)
- 0.001***
(0.000)
-0.002
(0.004)
0.032
(0.020)
-0.027*
(0.015)
0.000
(0.013)
-0.033
(0.026)
-0.071***
(0.019)
-0.010
(0.019)
-0.250**
(0.013)
0.5734
-0.001***
(0.000)
-0.002
(0.004)
0.033
(0.020)
-0.027*
(0.015)
0.000
(0.013)
-0.032
(0.026)
-0.071 ***
(0.019)
-0.009
(0.019)
-0.252***
(0.013)
0.5734
-0.000
(0.000)
-0.001
(0.004)
0.032*
(0.019)
-0.010
(0.001)
0.004
(0.012)
-0.019
(0.024)
-0.006
(0.017)
-0.019
(0.018)
-0.148***
(0.012)
0.1963
Notes: Robust standard errors in parentheses.
*** Significant at 1% **significant at 5% * significant at 10%.
The dependents variable is school average attendance rate.
Average attendance = number of days attended by students divide by the total number of
courses days. Regressions control for school characteristics
18
Table 1.4: Program impact on enrollment (new enrolled and enrollment rate)
Baseline
THR
THR*year1
Public school
Rural zone
Pupil/teacher
ratio
Teachers
Female
teacher
Electricity
Latrine
Running
water
Library
New enrolled
All schools
(1)
(2)
All
Girls
-202.7*** -85.78***
(7.629)
(4.243)
-5.261
-8.469***
(4.891)
(2.721)
10.712*
5.600*
(5.734)
(3.189)
26.471*** 13.516***
(4.028)
(2.240)
-10.535** -9.263***
(4.671)
(2.598)
3.065***
1.409***
(0.060)
(0.033)
59.285*** 27.316***
(1.018)
(0.566)
-3.091*** 0.984
(1.196)
(0.665)
8.907**
6.958***
(3.835)
(2.133)
-6.085**
-4.230**
(2.980)
(1.657)
2.057
-0.034
(2.648)
(1.473)
15.205*** 6.016**
(5.044)
(2.805)
(3)
Boys
-116.9***
(4.591)
3.207
(2.943)
5.053
(3.450)
12.955***
(2.424)
-1.272
(2.811)
1.656***
(0.036)
31.969***
(0.613)
-4.076***
(0.720)
1.949
(2.308)
-1.856
(1.793)
2.091
(1.593)
9.190***
(3.035)
Rural schools
(4)
(5)
All
Girls
-199.6*** -86.35***
(6.782)
(3.756)
-6.166
-9.665***
(4.723)
(2.615)
10.842**
5.701*
(5.507)
(3.049)
22.448*** 10.249***
(4.185)
(2.318)
(6)
Boys
-113.33***
(4.182)
3.500
(2.912)
5.142
(0.130)
12.199***
(2.581)
2.943***
(0.059)
59.417***
(1.021)
-4.187***
(1.205)
11.423***
(4.025)
-7.581**
(2.937)
1.944
(2.635)
17.377***
(5.163)
1.612***
(0.036)
32.278***
(0.630)
-4.646***
(0.743)
2.606
(2.482)
-2.591
(1.811)
1.699
(1.625)
9.472***
(3.183)
1.331***
(0.033)
27.140***
(0.566)
0.459
(0.667)
8.817***
(2.229)
-4.989***
(1.626)
0.245
(1.459)
7.905***
(2.859)
Enrollment rate
All schools
(7)
(8)
Girls
Boys
0.431***
0.569***
(0.011)
(0.011)
-0.103*** 0.103***
(0.006)
(0.012)
0.033***
-0.033***
(0.009)
(0.009)
0.036***
-0.036***
(0.006)
(0.006)
-0.001
0.001
(0.007)
(0.007)
-0.000
0.000
(0.000)
(0.000)
-0.001
0.001
(0.002)
(0.002)
0.010***
-0.010***
(0.002)
(0.002)
0.006
-0.006
(0.006)
(0.006)
-0.010**
0.010**
(0.005)
(0.005)
0.002
-0.002
(0.004)
(0.004)
-0.013*
0.013*
(0.008)
(0.008)
Rural schools
(9)
(10)
Girls
Boys
0.439*** 0.561***
(0.011)
(0.011)
-0.104*** 0.104***
(0.006)
(0.006)
0.032*** -0.032***
(0.009)
(0.009)
0.027*** -0.027***
(0.007)
(0.007)
-0.000
(0.000)
-0.001
(0.002)
0.009***
(0.002)
0.009
(0.006)
-0.010**
(0.005)
0.003
(0.004)
-0.007
(0.008)
0.000
(0.000)
0.001
(0.002)
-0.009***
(0.002)
-0.009
(0.006)
0.010**
(0.005)
-0.003
(0.004)
0.007
(0.008)
19
Ext.
restaurant
Girl
enroll.<40
Mining
R-squared
10.594***
(3.513)
0.410
(3.671)
-4.916**
(2.492)
0.9188
4.080**
(1.954)
-16.918***
(2.042)
-1.532
(1.386)
0.9033
6.515***
(2.114)
17.328***
(2.209)
-3.384**
(1.499)
0.8886
8.788**
(3.528)
0.704
(3.583)
-3.091
(2.462)
0.9160
3.964**
(1.954)
-15.957***
(1.984)
-0.276
(1.364)
0.8980
4.824**
(2.175)
16.661***
(2.209)
-2.815*
(1.518)
0.8835
0.007
(0.005)
-0.007
(0.005)
0.012**
(0.006)
0.020***
(0.004)
0.3692
-0.020*** 0.020***
(0.004)
(0.004)
0.3692
0.3678
-0.012**
(0.006)
-0.020***
(0.004)
0.3678
Notes: Robust standard errors in parentheses * Significant at 10%; **significant at 5%, *** significant at 1%
Dependents variable are attendance and enrollment rates. Regressions control for school and province specific characteristics
20
Table 1.5: Some descriptive statistics on enrolled students in baseline and follow-up
Enrolled girls
Enrolled boys
Enrolled students
Baseline
(1)
DM
111.583
(73.654)
118.93
(75.124)
230.52
(146.687)
(2)
THR
58.56
(38.608)
103.44
(62.699)
162.00
(98.635)
(3)
NF
78.37
(53.435)
88.88
(53.020)
167.25
(103.583)
Follow-up
(4)
DM
110.26
(73.601)
73.60
(73.185)
226.53
(144.507)
Notes: DM= daily meal; THR= take home ration; NF = non feeding
Standard errors in parentheses
21
(5)
THR
68.32
(42.425)
101.300
(56.918)
169.62
(96.062)
(6)
NF
78.68
(50.853)
86.91
(50.319)
165.60
(98.305)
Difference
(4)-(1)
(5)-(2)
DM
THR
-1.32
9.763
(6)-(3)
NF
0.31
-2.66
-2.14
-1.97
-3.99
7.62
-1.65
Robustness check tables
Table R0: Key school variables between provinces
School
characteristics
Number of schools
All
provinces
N=927
(1)
Public schools
0.902
Rural schools
0.922
Girl enrollment rate 0.460
Girl enrollment <40 0.216
Enrolled students
202.26
Pupil/teacher ratio
52.62
Female teachers
1.39
Schools facilities ²
1.435
Electricity
0.096
Running water
0.586
Latrine
0.706
Library
0.046
Parents association 0.946
Mining area in year
2011^
Bam
& Namentenga Bam
Sanmatenga
Sanmatenga
N=684
(2)
0.881
0.917
0.456
0.217
216.68
58.33
1.35
1.34
0.099
0.542
0.644
0.055
0.902
N=439
(5)
0.892
0.916
0.448
0.271
218.89
59.63
1.47
1.33
0.097
0.551
0.635
0.052
0.874
No
N=243
(3)
0.962
0 .934
0.458
0.271
167.25
44.59
1.17
1.58
0.082
0.679
0.802
0.024
1
No
N=245
(4)
0.864
0.917
0.470
0.122
212.76
56.20
1.13
1.35
0.102
0.526
0.661
0.061
0.951
Yes
Notes: Bam & Sanmatenga = targeted provinces
Namentenga = third external province (as a second comparison group of schools)
² school facilities = electricity + running water + latrine + library
^ Mining area = Yes if newly opened in 2011 and mining area = No if mining opened
before the year 2011
22
Table R1: Program impact on enrollment THR group vs. (DM+NF) group of schools
Enrolled students
(1)
(2)
All schools
All
Girls
-210.53*** -90.723***
(6.800)
(3.687)
Boys
-119.81***
(4.088)
(4)
(5)
(6)
Rural schools
All
Girls
Boys
-206.51*** -90.573*** -115.94***
(6.093)
(3.270)
(3.775)
THR
-6.400
(4.434)
-9.893***
(2.404)
3.493
(2.666)
-5.696
(4.209)
-9.822***
(2.259)
4.126
(2.607)
THR*year1
11.839**
(5.587)
6.618**
(3.030)
5.221
(3.360)
11.993**
(5.303)
6.533**
(2.846)
5.459*
(3.285)
13.671***
(2.026)
-7.470***
(2.229)
1.485***
(0.028)
24.991***
(0.464)
2.092***
(0.552)
4.423**
(1.847)
-3.235**
(1.421)
0.004
(0.012)
6.287**
(2.546)
6.825***
(1.733)
-14.501***
(1.531)
2.239*
(1.243)
0.8937
11.842***
(2.247)
-0.180
(2.472)
1.760***
(0.031)
28.519***
(0.515)
-1.815***
(0.612)
1.607
(2.049)
-0.259
(1.576)
1.737
(1.364)
9.635***
(2.824)
8.981***
(1.922)
17.897***
(1.698)
1.156
(1.378)
0.8717
24.317***
(3.892)
11.89***
(2.089)
12.419***
(2.411)
3.105***
(0.051)
53.501***
(0.851)
-0.038
(1.017)
8.661**
(3.544)
-5.695**
(2.558)
2.681
(2.226)
16.004***
(4.772)
13.843***
(3.218)
2.899
(2.712)
4.618**
(2.249)
0.9039
1.397***
(0.027)
24.796***
(0.456)
1.938***
(0.546)
6.044***
(1.902)
-4.439***
(1.373)
1.057
(1.195)
6.833***
(2.561)
6.422***
(1.727)
-14.339***
(1.456)
3.193***
(1.207)
0.8903
1.708***
(0.031)
28.705***
(0.527)
-1.975***
(0.630)
2.617
(2.196)
-1.257
(1.585)
1.624
(1.379)
9.171***
(2.956)
7.421***
(1.993)
17.237***
(1.680)
1.425
(1.393)
0.8657
Baseline
Public
school
Rural zone
25.513 ***
(3.736)
-7.650*
(4.111)
Pupil3.245***
teacher ratio (0.052)
Teachers
53.511***
(0.856)
Female
0.276
teacher
(1.018)
Electricity
6.029*
(3.407)
Latrine
-3.494
(2.621)
Running
2.023
water
(2.268)
Library
15.922***
(4.696)
External
15.807***
restaurant
(3.196)
Girl
3.396
enroll.<40
(2.824)
Mining
3.395
(2.293)
R-squared 0.9062
(3)
Notes: Robust standard errors in parentheses.
***significant at 1% ** significant at 5% * significant at 10%
23
Table R2: Program impact on school enrollment rate: THR vs. NF schools in rural
area (control for school where girl enrollment below 40%)
Baseline
THR
THR*year1
Public school
R-squared
School enrollment rate
Enrolled students
(1)
Girls
0.350***
(0.011)
0.016
(0.011)
0.038***
(0.014)
0.018*
(0.010)
0.1521
(3)
Girls
18.055**
(6.918)
18.370***
(6.460)
5.542
(8.509)
27.452***
(6.255)
0.1300
(2)
Boys
0.686***
(0.019 )
-0.010
(0.011)
-0.038***
(0.014)
-0.018*
(0.010)
0.1521
(4)
Boys
44.641***
(9.026)
18.723**
(8.429)
-2.905
(11.102)
32.497***
(8.161)
0.0873
Notes: Robust standard errors in parentheses
***significant at 1% **significant at 5% * significant at 10%
NB: covariates are not significant, but including them give the same results.
24
Table R3: urban vs. rural where girl enrollment below 40% (in only targeted schools
THR vs. DM)
Baseline
THR
THR*year1
Public school
R-squared
School enrollment rate
(1)
(2)
Girls
Boys
0.341***
0.659***
(0.009)
(0.009 )
-0.028***
-0.028***
(0.009)
(0.009)
0.024*
-0.024*
(0.013)
(0.013)
0.035***
-0.035***
(0.007)
(0.007)
0.0967
0.097
Enrolled students
(3)
(4)
Girls
Boys
18.055**
44.641***
(6.918)
(9.026)
18.370*** 18.723**
(6.460)
(8.429)
5.542
-2.905
(8.509)
(11.102)
27.452*** 32.497***
(6.255)
(8.161)
0.0875
0.0805
Notes: Robust standard errors in parentheses
*** Significant at 1% ** significant at 5% * significant at 10%
Covariates not presented in the table
25
References
1. Adelman, S., D. Gilligan, and K. Lehrer, 2008. How Effective are Food for Education
Programs?: A Critical Assessment of the Evidence from Developing Countries.
International Food Policy Research Institute, Washington, DC.
2. Adrogué, C., and M. E. Orlicki, 2011. Do in-School Feeding Programs Have Impact on
Academic Performance and Dropouts? The Case of Public Argentine Schools.
Unpublished manuscript, Universidad de San Andrés, Buenos Aires, Argentina.
3. Afridi, F., 2007. The Impact of School Meals on School Participation: Evidence from
Rural India. Working Paper, Syracuse University.
4. Afridi, F., B. Barooah, and R. Somanathan, 2013. School meals and classroom effort:
Evidence from India. Working Paper
5. Aghion, P., and P. Howitt, 1998. Endogenous Growth Theory. Cambridge, MA: MIT Press
6. Ahmed, A. U., 2004. Impact of Feeding Children in School: Evidence from Bangladesh.
International Food Policy Research Institute, Washington, DC.
7. Ahmed, A. U., and C. del Ninno, 2002. The Food for Education Program in Bangladesh:
an Evaluation of its Impact on Educational Attainment and Food Security. FCND
Discussion Paper No. 138, International Food and Policy Research Institute, Washington,
DC.
8. Bundy, D., C. Burbano, M. Grosh, A. Gelli, M. Jukes, and L. Drake, 2009. Rethinking
School Feeding: Social Safety Nets, Child Development, and the Education Sector. World
Bank/World Food Programme. Washington, DC:
9. Buttenheim, A., H. Alderman, and J. Friedman, 2011. Impact Evaluation of School
Feeding Programs in Lao PDR. World Bank Working Paper n°5518
10. Cheung, M. and Berlin, M. P. (2015), The Impact of a Food for Education Program on
Schooling in Cambodia. Asia & the Pacific Policy Studies, 2: 44–57. doi: 10.1002/app5.21
11. Denison, E. F., 1962. The sources of Economic Growth in the United State and the
Alternatives Before Us. The Economic Journal, Vol.72, No.288, pp.935-938
12. Diagne, A., M. M. Lô, O. Sokhna, and F. L. Diallo, 2014. Evaluation of the impact of
school canteens programs on internal efficiency of schools, cognitive acquisitions and
learning capacities of students in rural primary schools in Senegal. Working paper
13. Drèze, J. and G. G. Kingdon, 2001. School Participation in Rural India. Review of
Development Economics, 5(1), 1-24.
14. Duflo, E. & Kremer, M. (2003). Use of Randomization in the Evaluation of Development
Effectiveness. World Bank Operations Evaluation Department (OED), Washington, D.C
15. Galiani, S., and R. Perez-Truglia, 2011. School Management in Developing Countries.
Educational Policy in Developing Countries: What Do We Know and What Should We Do
to Understand What We Don’t Know?, University of Minnesota.
16. Gelli, A., U. Meir, and F. Espejo (2007). Does Provision of Food in School Increase Girls’
Enrollment? Evidence from Schools in Sub-Saharan Africa. Food and Nutrition Bulletin
28 (2), 149–55.
17. Ghosh, S., and H. Saha, 2014. The role of Adequate Nutrition on Academic Performance
of College Students in North Tripura. The Online Journal of New Horizons in
Education,Volume 3, Issue 3, pp41-53
18. Glewwe, P. W., E. A. Hanushek, S. D. Humpage, and R. Ravina, 2011. School Resources
and Educational Outcomes in Developing Countries: A Review of the Literature from 1990
to 2010. NBER Working Paper No. 17554.
19. Grantham-McGregor, S. M., S. Chang, S. P. Walker, 1998. Evaluation of School Feeding
Programs: Some Jamaican Examples. The American Journal of Clinical Nutrition,
26
67:785S– 9S.
20. He, F., 2009. School Feeding Programs and Enrollment: Evidence from Sri Lanka.
21. Heckman, J. J., and J. A. Smith. 1995. Assessing the Case for Social Experiments. Journal
of Economic Perspectives 9(2): 85-110.
22. Jacoby, E. R., S. Cueto, and E. Pollitt, 1998. When Science and Politics Listen to Each
Other: Good Prospects from a New School Breakfast Program in Peru. The American
Journal of Clinical Nutrition, 67: 4 795S-797S.
23. Jomaa, L.H., E. McDonnell, and C. Probart, 2011. School Feeding Programs in
Developing Countries: Impacts on Children’s Health and Educational Outcomes. Nutrition
Review, 69, 83–98.
24. Kazianga, H., D. de Walque, and H. Alderman, 2009. Educational and Health Impacts of
Two School Feeding Schemes: Evidence from a Randomizes Trial in Rural Burkina Faso.
World Bank Policy Research Working Paper 4976.
25. Kazianga, H., D. de Walque, and H. Alderman, 2012. Educational and Child Labor
Impacts of Two Food for Education Schemes: Evidence from a Randomizes Trial in Rural
Burkina Faso. World Bank/ World Food Program.
26. Kazianga, H., R., D. de Walque, and R. Akresh, 2013. Cash Transfer and Child Schooling:
Evidence from Randomized Evaluation of the Role of Conditionality. World Bank Policy
Research Working Paper 6340
27. Khandker, S. R., G. B. Koolwal and H. A. Samad, 2010. Handbook on Impact Evaluation:
Quantitative Methods and Practices, The World Bank,Washington, DC.
28. Kristjansson, E., V. Robinson, M. Petticrew, B. MacDonald, J. Krasevec, L. Janzen, T.
Greenhalgh, G. Wells, J. MacGowan, A. Farmer, B. J. Shea, A. Mayhew, and P. Tugwell,
2007. School Feeding for Improving the Physical and Psychosocial Health of
Disadvantaged Elementary School Children. Cochrane Database of Systematic Reviews.
29. Levinger, B., 1986. School Feeding Programs in Developing Countries: An Analysis of
Actual and Potential Impact. U. S. AID Evaluation Special Study N0. 30. U. S Agency for
International Development. Washington, D. C
30. MENA (2009). Annuaire Statistiques de 2008/2009.
31. MENA, 2012. Annuaires Statistiques de l’Education Nationale 2011/2012.
32. MENA., 2013.a Synthèse des données statistiques de l’éducation de base 2012-2013.
Direction des Etudes et de la Planification
33. MENA., 2013b. Evaluation des Acquis Scolaires (E. A. S) 2011-2012. Direction des
Etudes et de la Planification
34. Meyers, A. F., A.E. Sampson, M. Weitzman, B. L. Rogers, and H. Kayne, 1989. School
Breakfast Program and School Performance, American Journal of Diseases of Children
143:1234–1239.
35. Michaelowa, K., 2000. Dépenses d’éducation, qualité de l’éducation et pauvreté : exemple
de cinq pays d’Afrique francophone. Document Technique n°157 OCDE
36. Moore, E., 1994. Evaluation of the Burkina Faso School Feeding Program. Baltimore, MD:
Catholic Relief Services. Consultant Report (Unpublished).
37. Pollitt, E. (1995). Does breakfast make a difference in school? Journal of the American
Dietetic Association 95 (10), 1134–1139.
38. Romer, P., 1990. Endogenous technological change. Journal of Political Economy, Part 2,
98(5), S71–102
39. Schultz, T.W., 1961. Investment in Human Capita. The American Economic Review
Vol.51, No.1, 1-17
40. Smith, J. A. and P. Todd, 2005. Does Matching Overcome LaLonde’s Critique of Nonexperimental Estimators? Journal of Econometrics, 125: 305–53.
27
41. Tan, J. P., J. Lane, and G. Lassibille, 1999. Student Outcomes in Philippine Elementary
Schools: An Evaluation of Four Experiments. World Bank Economic Review, 13(3): 493–
508.
42. Tara, H. 2005. Nutrition and Student Performance at School. Journal of School Health.
Vol. 75, No. 6, pp.199-213.
43. UNESCO., 2011. Rapport mondial de suivi sur l’EPT. Aperçu régional : Afrique
subsaharienne, Paris, UNESCO
44. UNICEF., 2012. Schools for Burkina Faso: Investing in the future. Schools for
Africa/Burkina Faso.
45. Vermeersch, C. and M. Kremer 2004. School Meals, Educational Achievement and School
Competition: Evidence from a Randomized Evaluation. Policy Research Working Paper
No. 3523. Washington, D.C.: World Bank
28