CRS Food For Education 
Baseline Study
PREPARED FOR: CATHOLIC RELIEF SERVICES BENIN 
PREPARED BY: DANICE BROWN & JUAN CARLOS GUZMÁN 
THE UNIVERSITY OF NOTRE DAME
NOTRE DAME INITIATIVE FOR GLOBAL DEVELOPMENT (NDIGD) 
926 GRACE HALL
NOTRE DAME, IN 46556 
WEB: WWW.NDIGD.ND.EDU
March 2015 
Baseline Evaluation for FY 2014 McGovern-Dole Food 
for Education Project in Benin
Program: McGovern-Dole International Food for Education and Child 
Nutrition
Agreement Number: FFE-680-2014/046-00
Funding Year: Fiscal Year 2014
Project Duration: 2014-2020
Implemented by: Catholic Relief Services
Evaluation Authored by: Notre Dame Initiative for Global Development (NDIGD)
 Danice Brown
 Carlos Guzman
DISCLAIMER: This evaluation was conducted by an independent third party firm. The views expressed 
in this report do not necessarily reflect the views of the United States Department of Agriculture or the 
United States Government. 
Accessibility Note: An accessible version of this report can be made available upon request by 
emailing FAS.monitoring.evaluation@usda.gov. 
2 
Contents 
Tables............................................................................................................................................... 4
List of Abbreviations ........................................................................................................................ 6
Executive Summary ......................................................................................................................... 7
Strategic Objective 1: Improved Literacy of School-Aged Children ............................................ 8
Recommendation: ................................................................................................................... 9
Strategic Objective 2: Increased Use of Health and Dietary Practices ........................................ 9
Recommendation: ................................................................................................................... 9
Students’ Literacy: EGRA........................................................................................................... 10
Recommendation: ................................................................................................................. 10
1. Introduction ........................................................................................................................... 11
2. Methodology ......................................................................................................................... 13
2.1 Sampling Design for Principals, Teachers, and Parents ................................................. 13
2.2 Sampling Design for EGRA Implementation .................................................................. 16
2.3 Sampling Design for Qualitative Component ................................................................ 18
2.4 Sampling Weights .......................................................................................................... 18
Adjustment Factors ............................................................................................................... 18
3. Results of Questionnaires and Observation .......................................................................... 20
3.1 Strategic Objective 1: Improved Literacy of School-Aged Children .............................. 20
3.1.1 Improved Quality of Literacy Instruction .................................................................. 20
3.1.2 Improved Attentiveness ............................................................................................ 24
3.1.3 Improved Student Attendance .................................................................................. 26
3.1.4 Foundational Results ................................................................................................. 29
3.2 Strategic Objective 2: Increased Use of Health and Dietary Practices .......................... 30
3.2.1 Improved Knowledge of Health and Hygiene Practices ............................................ 31
3.2.5 Increased Access to Preventative Health Interventions ............................................ 35
4. Results of EGRA ..................................................................................................................... 35
4.1 Vocabulary ........................................................................................................................... 35
4.2 Oral comprehension ............................................................................................................ 36
4.3 Knowledge of letter names ................................................................................................. 36
4.4 Identification of letter sounds ............................................................................................. 37
3 
4.5 Simple words reading .......................................................................................................... 37
4.6 Invented words reading....................................................................................................... 38
4.7 Text reading and comprehension ........................................................................................ 38
4.8 Oral comprehension ............................................................................................................ 39
4.9 Dictation .............................................................................................................................. 39
5. Conclusions and Recommendations ..................................................................................... 40
Strategic Objective 1: Improved Literacy of School-Aged Children ...................................... 40
Strategic Objective 2: Increased Use of Health and Dietary Practices .................................. 41
Students’ Literacy: EGRA....................................................................................................... 41
6. Appendices ................................................................................................................................ 44
6.1 Baseline Study Key Questions ............................................................................................. 44
6.2 Instruments ......................................................................................................................... 46
Principal Questionnaire ......................................................................................................... 46
Teacher Questionnaire.......................................................................................................... 49
Parent Questionnaire ............................................................................................................ 51
Classroom Observation Rubric .............................................................................................. 55
Focus Group Guide ................................................................................................................ 56
Interview Rubric for CCS ........................................................................................................ 58
6.3 Power Calculations.............................................................................................................. 59
6.4 Summary Statistics .............................................................................................................. 61
6.5 Calculation of Indicators ...................................................................................................... 63
6.6 Supplementary Tables ......................................................................................................... 65
4 
Tables 
Table 1: Summary of SO1 Indicators ............................................................................................... 8
Table 2: Summary of SO2 Indicators ............................................................................................... 9
Table 3: School Summary Statistics ............................................................................................... 14
Table 4: Principals Summary Statistics .......................................................................................... 15
Table 5: Teachers: Summary Statistics .......................................................................................... 15
Table 6: Parents Summary Statistics ............................................................................................. 16
Table 7: Number of Classrooms Observed .................................................................................... 16
Table 8: EGRA Student Summary Statistics, 2015 ......................................................................... 17
Table 9: Literacy Instruction .......................................................................................................... 21
Table 10: Teacher Attendance Rate .............................................................................................. 22
Table 11: Teacher Attendance Rate by Commune (Intervention Schools Only) ........................... 22
Table 12: National Curriculum ....................................................................................................... 23
Table 13: Student Attentiveness ................................................................................................... 24
Table 14: Eating and Hunger ......................................................................................................... 26
Table 15: Attendance Counts and Rates ....................................................................................... 28
Table 16: Health Related Absences, Counts and Percentages ...................................................... 28
Table 17: Importance of Education ............................................................................................... 29
Table 18: Parents' and Women's Associations .............................................................................. 29
Table 19: Minimum Acceptable Diet ............................................................................................. 31
Table 20: Diet Content and Frequency .......................................................................................... 31
Table 21: Knowledge of Good Health and Hygiene ....................................................................... 32
Table 22: Place in which youngest child's feces were disposed .................................................... 32
Table 23: Water products named .................................................................................................. 33
Table 24: Reasons to wash hands named ..................................................................................... 33
Table 25: Times to wash hands named ......................................................................................... 33
Table 26: Ways to make water better for drinking named ........................................................... 34
Table 27: Student Knowledge of when to wash hands ................................................................. 34
Table 28: Student Knowledge of hygiene practices ...................................................................... 34
Table 29: Number of Students Dewormed ................................................................................... 35
Table 30: Percentage of vocabulary known by grade and type of school: 2015 .......................... 36
Table 31: Percentage of actions correctly performed by grade and type of school: 2015 ........... 36
Table 32: Average number of letters correctly read per minute by grade and type of school: 2015
....................................................................................................................................................... 37
Table 33: Average percentage of sounds correctly identified by grade and type of school: 2015
....................................................................................................................................................... 37
Table 34: Average number of words correctly read per minute by grade and type of school: 2015
....................................................................................................................................................... 37
Table 35: Average number of invented words correctly read per minute by grade and type of 
school: 2015 .................................................................................................................................. 38
5 
Table 36: Average number of words read per minute and questions responded correctly by 
grade and type of school: 2015 ..................................................................................................... 38
Table 37: Average percentage of questions correctly answered by grade and type of school: 
2015 ............................................................................................................................................... 39
Table 38: Percentage of words correctly written by grade and type of school: 2015. ................. 39
Table 39: EGRA Students’ Previous Education .............................................................................. 61
Table 40: EGRA Students’ Languages spoken at home ................................................................. 61
Table 41: EGRA Students’ Reading materials................................................................................ 61
Table 42: EGRA Students’ Literacy help at home.......................................................................... 62
Table 43: Who helps with homework at home ............................................................................. 62
Table 44: After school activities .................................................................................................... 62
Table 45: Eating at school .............................................................................................................. 62
Table 46: Teacher Attendance in Treatment Schools ................................................................... 65
Table 47: Status of Kitchen and Water System in Intervention Schools ...................................... 67
Table 48: Supplies Available in Treatment Schools ....................................................................... 70
Table 49: Supplies Available in Treatment Schools, cont. ............................................................. 72
6 
List of Abbreviations 
CCS Head of School Districts (Chef de Circonscriptions Scolaires) 
CI Cours d’Initiation (first grade) 
CP Conseillers Pédagogiques (Pedagogical Advisors) 
CP Course Préparatoire (second grade) 
CRS Catholic Relief Services 
FAO Food and Agriculture Organization 
HIP USAID Hygiene Improvement Project 
EGRA Early Grade Reading Assessment 
IFPRI International Food Policy Research Institute 
INFRE Institute National Pour la Formation et la Recherche en Education 
LOA Life of Activity 
ND University of Notre Dame 
USDA US Department of Agriculture 
WFP World Food Programme 
WEI World Education, Inc. 
7 
Executive Summary 
Catholic Relief Services, Benin (CRS Benin) is implementing a four-year project to improve 
literacy and alleviate hunger. The project is aimed at strengthening the learning environment 
and improving the quality of instruction, with an explicit focus on community engagement. The 
project will improve literacy through provision of school lunches which will increase student 
attentiveness and reduce student absences. Trainings for teachers and school administrators, 
improved instructional materials, school facilities and infrastructure will also increase the quality 
of education. The project will also improve health and dietary practices by implementation of 
awareness campaigns for parents, trainings for teachers and food preparers on hygienic 
practices, improved infrastructure and sanitation facilities, and distribution of deworming pills. 
Overall, the project will benefit approximately 38,000 primary school-age children attending 141 
primary schools in four communes in the Alibori and Borgou Departments of Northeast Benin. 
This baseline study utilized both quantitative and qualitative tools to provide information 
necessary to understand the starting point of students, parents, teachers and principals in the 
intervention areas. The quantitative portion has two main components. The first component 
consists of surveys of principals, teachers and parents, as well as a classroom observation. The 
second component consists of an Early Grade Reading Assessment (EGRA). The qualitative 
component consists of interviews with key stakeholders such as the head of the school districts 
(Chef de Circonscription Scolaires, CCS) in Malanville, Gogounou, and Kandi communes. 
Additionally focus groups were conducted with both parents’ and mothers’ associations 
identified as being particularly active, as well as with parents who have decided not to enroll 
their children in school. 
Data were collected in schools which will receive the intervention, and in schools which, like 
intervention schools, do not have canteens but were identified by CRS as having mothers’
associations. The majority of these schools are located in communes neighboring the 
intervention areas, but nine schools are located in the same commune as intervention schools 
(Kalale Commune). These comparison schools are hoped to serve as a control group during 
midline and end line evaluation. Throughout this report, balance tests are performed to 
demonstrate the adequacy of the control group as a comparison. Statistically significant 
differences are noted when detected. The comparison schools are found to be useful as a 
control group. A few issues that should be dealt with using econometric models are listed in the 
conclusions section. 
Data has been weighted to represent the entire population of students, teachers, principals, or 
parents, within school clusters and strata of urban or rural schools. Results of the baseline study 
were mixed. It was observed that some indicators for the project have been surpassed prior to 
program implementation, while in other categories a clear need for improvement is needed. 
Results are discussed below by strategic objective. 


10
 Provide education to students and parents on health and hygienic practices. 
 Provide de-worming pills in intervention area. 
Students’ Literacy: EGRA
 Because the cut-off point of “reading and understanding the meaning of grade level text” has 
not yet been developed at the time of writing this baseline report, individual elements of the 
EGRA test are presented in this report. Compared to other sections, students demonstrated 
some proficiency in vocabulary identification and oral comprehension of commands. In other 
sections such as knowledge of letter names, identification of letter sounds, reading and 
comprehension of texts, and dictation, students performed at very low levels. 
Because EGRA was conducted in the middle of the school year, students cannot be proficient at 
their grade level, particularly among first grade (CI) students. Performing the EGRA at the end of 
the school year will provide richer information on students’ literacy levels, particularly for 
second grade (CP) students. 
Recommendation: 
 Develop national cut-off for reading at grade level 
 Conduct EGRA assessment at intervention schools annually. 
11
1. Introduction 
Catholic Relief Services, Benin (CRS Benin) is implementing a four-year project to improve 
literacy and alleviate hunger. The project is aimed at strengthening the learning environment 
and improving the quality of instruction, with an explicit focus on community engagement. CRS 
Benin, in partnership with World Education, Inc (WEI) will ensure sustainability by leveraging 
community-based activities to engage parent-teacher associations (Associations des parents 
d’élèves) and school mothers’ associations (Associations des mères d’élèves) in the 
management of school canteens and other activities. The project also aims to increase 
government capacity and ownership by working with the Ministry of Education’s National 
Institute for Training and Research in Education (Institute National pour la Formation et la 
Recherche en Education – INFRE) in designing improved teacher training. The project will train 
pedagogical advisors known as “CPs” (conseillers pédagogiques), who will in turn directly train 
and monitor an estimated 212 teachers and 141 officials to assess literacy levels and share new 
and effective teaching methods. 
The project will improve literacy through provision of school lunches which will increase student 
attentiveness and reduce student absences. Trainings for teachers and school administrators, 
improved instructional materials, school facilities and infrastructure will also increase the quality 
of education. The project will also improve health and dietary practices by implementation of 
awareness campaigns for parents, trainings for teachers and food preparers on hygienic 
practices, improved infrastructure and sanitation facilities, and distribution of deworming pills. 
Overall, the project will benefit approximately 38,000 primary school-age children attending 141 
primary schools in four communes in the Alibori and Borgou Departments of Northeast Benin. 
This report presents the baseline information necessary to understand the starting point of 
students, parents, teachers and principals in the intervention areas. The methodology which 
was used to gather this information included both quantitative and qualitative tools. Data 
collected for this baseline study was gathered in January 2015. 
Additionally this report provides similar information on students, parents, teachers and 
principals in two neighboring communes. These populations serve as a comparison group in 
order to determine the impact of the program. 
The results of this baseline study demonstrate high levels of teacher and student attendance, 
and low rates of health-related absences for students. However, it demonstrates moderate 
levels of attention in classrooms and student-reported hunger. Finally, it demonstrates low 
levels of student literacy, and very few students who are receiving a minimum acceptable diet. 
Additionally, most parents do not have complete knowledge and correct behavior related to 
health and hygiene. Finally, schools lack physical resources such as buildings and books, and 
12
personnel resources such as sufficient numbers of teachers to deliver quality education to the 
community. 
This report is divided into 6 sections. The first is this introduction. The second section presents 
the methodology used to gather data. The third section presents the baseline statistics for the 
indicators of the project. The fourth section presents the baseline statistics for the EGRA test. 
The fifth section presents conclusions and recommendations. 
13
2. Methodology 
This baseline study utilized both quantitative and qualitative tools. The quantitative portion has 
two main components. The first component consists of surveys of principals, teachers and 
parents, as well as a classroom observation. The second component consists of an Early Grade 
Reading Assessment (EGRA). The qualitative component consists of interviews with key 
stakeholders such as the head of the school districts (Chef de Circonscription Scolaires, CCS) in 
each intervention commune, and parents’ and mothers’ associations which were identified by 
the quantitative component as particularly active. This section explains the sampling strategy 
for each component of the study, and how the analysis for this report has been completed. 
2.1 Sampling Design for Principals, Teachers, and Parents 
Sampling was determined based on expected impact to be detected at midterm or end line. The 
goal was to achieve sufficient statistical power to detect the impact we expect to find. School 
sampling was stratified by urban and rural schools, and was calculated proportional to the 
number of students in each school. In this way each student had an equal probability of 
selection into the sample, despite the different size in schools. Stratification was completed by 
classifying the schools as rural or urban, and selecting equal numbers from each category in 
both intervention and comparison groups. Data was stratified by urban and rural schools 
because it was assumed that urban and rural populations will differ on relevant indicators. 
Schools were then selected randomly within these strata. Treatment and control groups each 
consisted of seven urban schools, and 43 rural schools. Resulting data is stratified by the same 
categories, but individual indicators are not disaggregated by strata because none of the 
indicators in the PMP require this. The complete list of schools which were sampled included 
142 intervention schools and 98 comparison schools. 
The sampling strategy focused on outcome indicators for parents. The two main outcomes of 
interest for parents are the percentage of parents who can name at least three benefits of 
primary education, and the percentage of parents who receive a passing score on a test of good 
health and hygiene. For both indicators, the goal is 60%. Using a significance of 5%, a statistical 
power of 80%, it was determined that with 50 treatment and 50 control schools, and 10 parents 
surveyed per school, the study would have power to detect a minimum change from 46% to 
60%. Since this effect is fairly modest (a .2 effect size), we could reasonably expect for a change 
of this size to be observed. 
For selection of the schools based on these calculations, the gsample1
 command in STATA was 
used. Within each stratum, schools were randomly selected with a probability proportional to 
the size of the school. Since for some schools the probability was greater than the available 
selection probability, these probabilities were limited to 1/50. 
 
1
 Jann, B. (2006). gsample: Stata module to draw a random sample. Available from 
http://ideas.repec.org/c/boc/bocode/s456716.html. 
14
For selection of respondents, different methods were used per school. The universe of teachers 
and principals were planned to be surveyed at each school so no selection methodology was 
implemented. For parents, 10 parents were randomly selected using the school rosters. This 
ensured that the sample of parents interviewed in each school were representative of the 
parents of enrolled students in each community. This is consistent with CRS’s indicators, but it 
should be noted that it is not representative of parents in the community in general. 
Classrooms for observation were also selected randomly. At each school, two classrooms were 
randomly selected for observation. Within the classroom, ten students were again randomly 
selected to be observed and classified as attentive or inattentive at the time of observation. For 
this component of the study, only classrooms from 3rd to 6th grade were selected, as the 
students in 1st and 2nd grade were being pulled out for EGRA testing. Thus it was assumed that 
those classrooms would have sufficient disturbance to negatively bias the results, since students 
would be paying attention to the movement of their peers instead of the teacher. 
Open Data Kit was used on smart phones to collect information from the surveys. Open Data Kit 
allows data to be collected offline and downloaded to a server once the phone is again 
connected to the internet. More information on Open Data Kit can be found here: 
http://opendatakit.org/. In addition, a Random Number Generator (RNG) free app was used to 
randomly select classrooms and students. The App Smart Lock, a free application, was used to 
prevent the use of other applications in the cellphones and tablet that could potentially drain 
the battery. All phones and tablets used Android 4.1. 
Schools sampled in the treatment area were significantly larger than those in the control area. 
According to data from last academic year, which was used for sampling, selected schools in the 
treatment area have 282 students, while schools in the control area have 183 students. This 
difference is observed because there is a similarly statistically significant difference within the 
universe of treatment and comparison schools. Average number of teachers in treatment and 
control areas is the same. 
Table 3: School Summary Statistics 
Control Treatment Difference
Mean student population (#) 183 282 -99***
Mean number of teachers 3 3 -0.48
* p<0.1; ** p<0.05; *** p<0.01 
A total of 95 principals were interviewed—48 in the comparison area and 47 in the intervention 
area. The sample is balanced on percent female principals, and levels of experience. Principals 
are, on average, six years older in the treatment group than the control group, but the size of 
this difference is negligible, particularly because this difference is not reflected in amount of 
experience. 
15
Table 4: Principals Summary Statistics 
Control Treatment Difference
Percent Female 2.21 4.99 -2.79
Mean Age 40.50 46.56 -6.06***
Experience (yrs) 3.64 4.01 -0.38
Education Level:
Some High School 42.99 26.47 16.51
High School Graduate 24.36 44.06 -19.69
Some college 7.39 11.30 -3.91
College Graduate 9.17 1.51 7.66
Post-graduate or professional 16.08 16.66 -0.58
* p<0.1; ** p<0.05; *** p<0.01 
A total of 399 teachers were interviewed—221 in the intervention group and 178 in the 
comparison group. Again, this difference is due to the larger size of schools in the intervention 
area. On average, a higher proportion of female teachers were interviewed in the treatment 
area, and this difference is statistically significant. Conversely, teachers in the comparison group 
have less education (secondary cycle in “normal” track high school, or junior high school in the 
education track). More teachers have completed “education track” high school in the treatment 
areas, but this difference is only statistically significant to the 10% level. 
Table 5: Teachers: Summary Statistics 
Control Treatment Difference
Percent Female 12.76 28.99 -16.23***
Mean Age 30.57 31.77 -1.20
Experience (yrs) 5.38 5.65 -0.27
Education Level:
High School (Secondary Cycle) (%) 27.19 15.52 11.68*
BPC: Junior High in Education Track (%) 18.62 6.81 11.81**
CEAP: High School in Education Track (%) 21.14 32.39 -11.24*
Baccalauréat: High School Graduate (%) 14.54 23.15 -8.61
CAP: Teaching Certificate (%) 14.84 15.39 -0.54
Some University or higher (%) 3.66 5.31 -1.65
* p<0.1; ** p<0.05; *** p<0.01 
 A total of 556 parents participated in the questionnaire, 284 in the area of comparison and 272 
in the area of intervention. No statistically significant difference was observed between 
treatment and control groups in age, number of children, and household size. There was a 
sizable difference in the proportion of female respondents in the treatment area; however, this 
difference is only significant to the 10% level. 
16
Table 6: Parents Summary Statistics 
Control Treatment Difference
Percent Female 23.04 37.88 -14.84*
Mean Age of Parent 39.42 37.51 1.91
Mean Number of Children 7.71 6.34 1.37*
Mean Household Size 16.54 16.79 -0.25
* p<0.1; ** p<0.05; *** p<0.01 
A total of 1,420 students were observed, in 142 classrooms. An equal proportion of 3rd grade 
classes were observed in both treatment and control areas, but significantly more 4th and 6th
grade classrooms were observed in the treatment area. In some cases, classrooms were 
combined because of a lack of sufficient teaching staff. In this case, enumerators were 
instructed to observe students in only the randomly selected grade if the two grades could be 
distinguished from each other. For instance, teachers of combined classrooms assign one grade 
to sit on the right side of the room, and the other grade to sit on the left. In three instances this 
was not the case, meaning the class selected for observation was in a room where the students 
were mixed with students of another grade. This is listed as a 3rd and 4th grade or 5th and 6th 
classroom below. On average, the proportion of male and female students observed in the 
treatment and control area was equal (not reported). 
Table 7: Number of Classrooms Observed 
Control Treatment Difference
3rd grade 24 24 0
4th grade 13 26 -13**
5th grade 17 15 2
6th grade 5 16 -11**
3rd and 4th grade 0 2 -2
5th and 6th grade 1 0 1
* p<0.1; ** p<0.05; *** p<0.01 
 
2.2 Sampling Design for EGRA Implementation 
The sample size for the selection of students to complete EGRA was determined based on the 
indicator goal of students reading at grade level. This goal is 65%, meaning that there is a .65 
probability of success for each student at baseline. Again a clustered sampling strategy was 
employed, with students constituting the level of analysis but considering that students are 
nested within schools. Similar constraints of significance (5%), statistical power (80%), and 
number of clusters (50 treatment and 50 control) were utilized. Within these constraints, if 16 
students are tested per grade, the study has power to detect a change of at least 12 percentage 
points (from 53% to 65% of students reading at grade level). Again, this effect size is below .2 
and therefore considered to be a modest effect, meaning it can be reasonably expected that the 
intervention will result in an impact of this size. 
17
Among the schools which were selected using the above strategy, students were again selected 
randomly using school rosters. This ensures that the students selected to complete EGRA were 
representative of the first and second grade student population within the schools visited. Since 
these schools were also selected randomly, it can be inferred that the EGRA scores are 
representative of the first and second grade populations in all the intervention and comparison 
schools. 
Tangerine® was the program that was used to test EGRA. Tangerine® was developed by RTI 
International and is available under a General Public License (GNU). For more information on 
this program, please visit tangerinecentral.org. 
For more information on the sampling strategy, please see Appendix 7.3: Power Calculations 
Power calculations were conducted assuming a cluster size of 48 intervention and 48 
comparison schools. These power calculations are relevant to our final sample size of 50 
intervention and 50 comparison schools. 
A total of 2,677 students were tested with EGRA in 99 schools visited. The total number of 
students tested by grade was 1,346 for CI and 1,331 for CP. The distribution of students 
between treatment and control schools was similar: 1,254 students in control schools and 1,423 
students in treatment schools. The sample had an equal proportion of girls and boys. The 
average age of students was about 6 years old for CI and 7 years old for CP. However, the 
response rate for age was extremely low. Enumerators were instructed to ask the student or 
copy the age from the class roaster, but both sources of information did not provide more than 
32% of information on average. Students understood the question, since it was asked in the 
local language and not French. This is a common occurrence in rural communities where time 
and age have different cultural significance and people may not be aware of their age. 
The test took on average 10 minutes to complete with CI students and 16 minutes with CP 
students. More background information on the sample can be found in Appendix 6.4: Summary 
Statistics. 
Table 8: EGRA Student Summary Statistics, 2015
CI CP
Indicator Control Treatment Diff Control Treatment Diff
Girls (%) 49.0 52.2 -3.2 49.9 49.5 0.3
Mean age 6.1 6.0 0.1 7.0 7.2 -0.2
Age percent 
response
29.4 25.5 3.9 33.2 42.3 -9.1
* p<0.1; ** p<0.05; *** p<0.01 
18
2.3 Sampling Design for Qualitative Component 
The qualitative component focused on key stakeholders and key informants on project 
outcomes of interest. First, interviews were conducted with the head of school districts (Chef 
de circonscription scolaire, or CCS), in each intervention commune. Second, focus groups were 
conducted among parents’ and women’s associations (PME’s and AME’s) that were identified by 
the principals or CCS’s as most active, by their own interpretation. These associations were in 
turn asked to identify parents who have not enrolled their children in school. Interviews were 
conducted with those parents as well. In this way, the qualitative interview employed a 
snowball method, in which key informants identify other key informants. 
2.4 Sampling Weights 
In the schools, the goal was to interview all principals and all teachers in the 100 selected 
schools. However, because some teachers and principals were absent, a response rate for each 
sample was calculated. Similarly, the study aimed to interview ten parents at each school but 
often less than ten selected parents were present. Each sample was weighted by the inverse 
probability of selecting the school from the universe of schools provided, and by the inverse 
probability of selecting the respondent from the total respondents available within that school. 
In this way, sample weighting accounts for the selection of the school, but also for the selection 
within the school or classroom. Schools were randomly selected with probability proportional to 
size. The measure of size was the number of students enrolled in each school. Therefore the 
school weight is as follows: 
𝑊𝑗
𝑖 =
𝑀
𝑗𝑚𝑖
Where j is the number of schools sampled in each strata, mi
 is the number of students in each 
school and 
𝑀 = ∑ 𝑚𝑖
𝑁
𝑖=1
where N is the number of schools in the explicit stratum. The school weight was adjusted in 
specific cases. There were 7 schools in the intervention area classified as urban; these schools 
received a weight of 1. In order to avoid some weight being less than unity, the size of large 
schools (schools with sizes larger than the sampling interval given by M / n), was set equal to 50. 
As a result, these large schools were sampled with equal probability without having to use an 
explicit stratification approach. Finally, the sampling weight was adjusted for the rural control 
stratum since one of the schools was closed. 
Adjustment Factors 
Weights for principals and teachers were adjusted based on response rate. The adjustment 
factors are listed as follows: 
19
𝐴𝑗𝑡 =
𝑛𝑗
𝑡𝑗
Is the adjustment rate for teachers, where 𝑛𝑗
 is the number of teachers within school j and 𝑡𝑗
is 
the number of teachers who completed the survey in school j. Thus the final weight for the 
teachers is as follows: 
𝐹𝑊𝑗𝑡 = 𝑊𝑗 ∗ 𝐴𝑗𝑡
Similarly, the parents subsample was adjusted based on the probability of selecting the parents 
from the total available parents at school j. 
𝐴𝑗𝑝 = 
𝑛𝑗
𝑝𝑗
𝐴𝑗𝑝 is the adjustment factor for parents, where 𝑛𝑗
 is the total number of parents at school j, and 
𝑝𝑗
 is the number of parents from school j who participated in the survey. Thus the final weight 
for the parents is as follows: 
𝐹𝑊𝑗𝑝 = 𝑊𝑗 ∗ 𝐴𝑗𝑝
Classrooms were similarly adjusted for the probability of selection. The adjustment factor for 
classrooms is as follows: 
𝐴𝑗𝑐 =
𝑛𝑗
𝑐𝑗
Where 𝑛𝑗
 is the total number of classrooms eligible for observation in school j (note that only 
grades 3-6 were eligible for selection), and 𝑐𝑗
 is the number of classrooms that were observed in 
school j. Additionally, this sample is adjusted based on the probability of selecting a student 
within the classroom. This adjustment factor is as follows: 
𝐴𝑗𝑐𝑠 =
𝑛𝑗𝑐
𝑠𝑗𝑐
Where 𝑛𝑗𝑐 is the number of students in classroom c in school j, and 𝑠𝑗𝑐 is the number of students 
observed in classroom c in school j. Thus the final weight for classroom observation is as follows: 
𝐹𝑊𝑗𝑐𝑠 = 𝑊𝑗 ∗ 𝐴𝑗𝑐 ∗ 𝐴𝑗𝑐𝑠
Finally, the principal sample was adjusted by the response rate of principals within the strata. 
𝐴𝑟 =
𝑛
𝑟
20
Where 𝑛 is the total number of principals within the strata, and 𝑟 is the number of principals 
who completed the survey in that stratum. Thus the full weight calculation for principals is: 
𝐹𝑊𝑗𝑟 = 𝑊𝑗 ∗ 𝐴𝑟
For EGRA, the goal was to test 16 students in CI and CP grades from each school. The standard 
errors to test hypotheses were corrected because students were grouped or nested within 
classes and then nested with schools, as oppose to a random selection of students among all 
eligible students in the districts where the survey took place. In addition, because students have 
a different probability of selection depending on the total number of students in each class, a 
weight was calculated as follows: 
𝐹𝑊𝑗𝑖 =
𝑊𝑗∗𝑘𝑗𝑖
𝑠
𝑗𝑖
Where Wj
 is the school weight for school j, kji is the total number of students in grade i from 
school j, and sji is the number of students sampled from grade i in school j. When the number of 
students is less or equal to 16 the sample weight was equal to the school weight. 
The following section reports the results observed. It is organized by indicators for each result 
stream and strategic objective of the project. When necessary, data are disaggregated by grade, 
gender, commune or level of experience. 
3. Results of Questionnaires and Observation 
The analysis below considers the strategic objectives as laid out in CRS’s Project Monitoring Plan 
(PMP). Indicators which are not zero at baseline are focused on here. 
3.1 Strategic Objective 1: Improved Literacy of School-Aged Children 
3.1.1 Improved Quality of Literacy Instruction 
Percent of teachers who devote at least 45 minutes per day to literacy instruction 
Teachers of all grades were asked how many minutes they spent each day on literacy 
instruction. Literacy was defined as “teaching how to read and write.” Each teacher was then 
classified as meeting the indicator or not. Weights were applied to extrapolate the count of 
teachers to the population of teachers. Using this data, 600 (90.46%) teachers spend at least 45 
minutes a day on literacy instruction. On average, teachers spend 100 minutes per day teaching 
literacy. The LOA target of 70% has been surpassed. 
21
Table 9: Literacy Instruction 
Control Treatment Difference
Spend at least 45 minutes per day teaching 
literacy (#)
283.83 599.89 -316.06***
Spend at least 45 minutes/day teaching 
literacy (%)
90.35 90.46 -0.11
Minutes spent teaching literacy 92.97 100.78 -7.82
* p<0.1; ** p<0.05; *** p<0.01 
While it appears that the time spent teaching literacy is sufficient, the quality of that time is 
insufficient, as evidenced by both students’ literacy scores and reports from CCS’s. They report 
a lack of resources for teachers, and a lack of teachers. For example, the of Kandi, 
a larger town in the area of intervention, stated that 
“In Kandi, the school district is comprised of 100 schools today, exactly 100. And 
it has 471 classrooms but actually 246 teachers. You can already see the 
deficiency of teaching staff.”2
 Furthermore, the teachers who are working in the region do not have sufficient training. Over 
half the teachers surveyed have only completed some high school education or less. See Table 2 
for a break-down of teachers’ education levels.
Average Teacher Attendance Rates for each school and aggregated by district 
Data on teacher attendance rates were acquired from principals. Principals were asked the 
following question about each of their teachers: “How many days has [Teacher t] been absent 
last trimester?” The principal then referred to his or her register to report the number of days 
teacher t was absent. The reported number of days was subtracted from the total number of 
days last trimester (47) and divided by the total number of days to achieve the rate by teacher. 
Teacher attendance rates were then averaged by school. See Appendix 6.5 for a definition of 
this formula. 
Using this calculation, it was found that teachers are present at their schools 92% of the time, 
surpassing the LOA target of 75%. No statistical difference between teachers in control and 
treatment areas was observed on this indicator. This rate was verified by the teacher response 
rate in this data collection, which was 91.7% (not reported). Caution should be used in relying 
solely on this observational data, because of potential bias due to advance warning of 
enumerator visits, but it is provided here to supplement principal-reported data. See Tables 7 
and 8 for a breakdown of this indicator among schools by grade taught, teacher experience, and 
commune. Table 8 reports rates for teachers in treatment schools only by commune. 
Data for this indicator was only available from principals’ registers. However, this source is 
potentially biased for several reasons—a desire to please potential program implementers, a 
 
2
 Key Respondent Interview, , Kandi. . 
22
fear of retribution if data is reported to the Ministry of Education, and finally, a lack of adequate 
records. The first two risks of bias were mitigated by the consent form, which specifically states 
that data will only be reported at aggregate levels, and that participation in the survey has no 
impact on selection for program participation. The enumerators explained this to each principal 
before beginning the interview. However, little can be done about a school with inaccurate 
records. It is recommended that program partners verify information through unannounced 
visits during the planning and initial implementation of the program. 
Schools in Benin have two types of teachers—“permanent” teachers (those employed by the 
state), and “community” teachers (employed by the community). Because community teachers 
are paid by the hour, it could be possible that these teachers have a higher rate of attendance 
than community teachers. Therefore, it is recommended that data on teacher type is gathered 
at mid-line and end-line so data can be disaggregated by type of teacher. 
Table 10: Teacher Attendance Rate 
Control Treatment Difference
Teacher Attendance Rate 93.33 93.29 0.05
Grade 1 93.26 97.29 -4.03
Grade 2 96.53 85.03 11.50
Grade 3 89.53 94.25 -4.72
Grade 4 91.57 96.10 -4.52
Grade 5 90.70 95.21 -4.50
Grade 6 89.75 89.06 0.70
Some High School 92.71 96.08 -3.37
High School Graduate 91.95 91.97 -0.03
Some college 97.62 94.24 3.39
College Graduate 94.86 95.74 -0.89
* p<0.1; ** p<0.05; *** p<0.01 
Table 11: Teacher Attendance Rate by Commune (Intervention Schools Only) 
Rate
Kalale 99.10
Gogounou 94.54
Kandi 86.44
Malanville 97.62
* p<0.1; ** p<0.05; *** p<0.01 
A full list of teacher attendance rates by school can be found in Appendix 7.5: Supplemental 
Tables. The interpretation of this indicator is consistent with that of the time spent on literacy 
instruction. While teachers are present at their schools for a sufficient amount of time, they do 
not have sufficient resources or training to lead quality instruction during that time. 
23
Percent of Teachers Using the National Curriculum 
In 88.3% of the schools, principals report that their schools are using the national curriculum. 
Extrapolating this rate to the universe of teachers in the intervention area implies that 587 
teachers use the national curriculum (see Table 12). Again, this already surpasses the LOA target 
of 141 teachers using the national curriculum. This could be because the program goal is 1 
teacher per school using the national curriculum. However, national curriculum is typically 
implemented by a school entirely, in which case many teachers use the curriculum; or not at all, 
in which case no teachers use the curriculum. A change in the definition of the target can help 
capture information more clearly. For example, if the goal is 100% of schools using the national 
curriculum, this target has not yet been met. 
For a deeper understanding of the national curriculum, we can note the response from the 
 in Kandi. He states, regarding the curricula 
“The program is well designed, the elements are well-designed, but the 
actors who need to direct these programs, to develop them, are they 
sufficiently equipped? This is the problem. It’s one thing is to have well￾designed programs, but the other thing is the knowledge that those who 
are in charge of driving the programs forward have the hoped-for profile. 
Do they have the necessary capacity and competence for developing these 
programs?” 
 In general, most school officials interviewed agreed that not only is the number of teachers in 
the intervention areas insufficient, but the teachers who were hired by the school district or the 
local community lacked necessary levels of education or certification. Parents also noted that 
schools suffer from a lack of resources. When asked what could be done to help them facilitate 
their children’s learning, the parents in Kalale asked for resources such as notebooks and math 
or French activity books “that the state does not provide.”3
 Without these resources and 
without sufficient training, it can be presumed that teachers would face difficulty in providing 
quality instruction. Further classroom observation by program implementers can verify quality 
of instruction, presence of resources, or teacher capacity when instructing using the national 
curriculum. 
Table 12: National Curriculum 
Control Treatment Difference
Schools Using the National Curriculum (%) 86.03 88.30 -2.27
Number of Teachers Using National Curriculum 295.58 587.19 -291.61***
* p<0.1; ** p<0.05; *** p<0.01 
 
3
 Focus group with Kalale PTA. 22 January 2015. 
24
3.1.2 Improved Attentiveness 
Percent of Students who are identified as Attentive during Classroom Instruction 
Attentiveness was identified through classroom observation. A classroom observation template 
was created based on validated instruments from Carnegie Mellon’s Eberly Center for Teaching 
Excellence, the London School of Education and Massachusetts Department of Education (see 
Appendix 6.2: Instruments). As described in Section 2.2, classrooms for observation were 
randomly selected from grades 3-6, because students from grades 1-2 were participating in the 
EGRA exam, and thus would not provide valid data on attentiveness. Enumerators were 
instructed to observe each of the ten randomly-selected students individually until that student 
could be classified, according to the rubric, as attentive or inattentive. An average of 
attentiveness was then determined for each classroom. See Appendix 6.5 for equation. 
To disaggregate average classroom attentiveness by gender, the gender ratio of the classroom 
was used. Gender ratio was calculated using the number of boys and girls in the observed 
classroom, as reported in the teacher survey. The proportion of girls in each classroom was 
determined from this data. However, it was assumed that the gender ratio of students observed 
was not always exactly proportional to the gender ratio of the classroom. So using this data a 
gender was assigned to each observed student in a classroom by generating a normally 
distributed set of binomial variants, centered on the gender ratio of the observed classroom. 
Using this method, each student was assigned a gender and mean values of attentiveness were 
calculated in the same manner described above, among male students and female students 
separately. 
Using this procedure, 74 % of students in intervention schools were classified as attentive. This 
is quite close to the LOA target of 75%. Again, there was no statistical difference between 
treatment and control schools on this indicator, demonstrating that the sample is balanced. 
Also the differences between male and female students within either treatment or control 
group are not statistically significant (not reported). This data could be confirmed through 
unannounced visits to the schools by program implementers if desired. 
Table 13: Student Attentiveness 
Control Treatment Difference
Average Attentiveness 70.77 73.99 -3.22
Female 73.78 79.87 -6.09
Male 68.29 65.78 2.52
* p<0.1; ** p<0.05; *** p<0.01 
Percent of parents whose children report being hungry during the school day 
Data on student hunger was gathered from parents. The World Food Programme (WFP), Food 
and Agriculture Organization (FAO), the US Department of Agriculture (USDA) and the 
International Food Policy Research Institute (IFPRI) all use parent-reported information on child 
25
hunger.4
 Furthermore, research which includes both child-reported and parent-reported child 
hunger demonstrates a correlation between the two.5
 In addition, pilot tests of the survey 
instrument resulted in the removal of the question of child hunger from the teacher 
questionnaire, because teachers indicated that they did not know if their students were hungry 
during the day. They believed that their students would not report this to them out of fear. For 
these reasons, it was decided that parents were the best sources for data on student hunger. 
Parents were asked about each of their children who were enrolled in the intervention school, 
“Has (name) reported being hungry during the school day last week?” The school day was 
defined as the time period from the beginning of the school day until the end. It should be 
noted that the school day in Benin is quite long (8:00-17:00), with a lengthy lunch break (12:00-
15:00). This entire time was included in the definition of school day to be consistent with USDA 
definition of school day. Parents were asked about each of their children individually. The 
indicator of interest was a binomial, with a parent being classified as “1” if they reported at least 
one hungry child. This means parents were classified as a “0” only if they responded that none 
of their children reported hunger. See Appendix 6.5 for a breakdown of this formula. 
Using this definition, 55% of parents stated that at least one of their children had reported being 
hungry during the school day (see table 11). Again, this is statistically indistinguishable from 
students in the comparison area. Improvement will be needed to reach the final target of only 
20% of parents with children who report hunger during the school day. 
Percent of students who consume daily meals during the school day 
Similarly, parents were asked about students’ consumption of daily meals during the school day. 
Again parents were asked about each of their children enrolled in the school “Did [name] eat 
during the school day yesterday?” Enumerators were prompted that if the prior day was not a 
school day, to ask specifically about the last school day. Again, a similar definition of school day 
was used, which included the long lunch break. In this case data calculated by student, and then 
averaged by parent. See Appendix 6.5 for a breakdown of the formula. 
According to these calculations, 98% of students in the schools of intervention ate something 
during the school day (see Table 14). Again no statistically significant difference was detected 
between the intervention and comparison schools on this indicator. 
This rate surpasses the LOA target of 95%. Because the Beninese school day is long, it is logical 
that most students might eat something during that time. But as can be seen in the indicator on 
minimum acceptable diet below, the content and frequency of students’ diets is insufficient 
according to USDA standards. 
 
4
 For an adaptation of one of the most commonly cited tools for assessing child hunger, the Community 
Child Hunger Identification Project, see the following: 
http://www.sahealthinfo.org/nutrition/food8hungerscale.pdf 
5 Murphy, J M et al. “Relationship between hunger and psychosocial functioning in low-income American 
children.” Journal of the American Academy of Child Adolescent Psychiatry. 1998 Feb;37(2):163-70.
26
Parents were not asked where their children were acquiring food during the day—whether they 
were going home or purchasing something near the school. Additionally, information on 
breakfast habits in the region was not gathered. This information can be gathered from AME’s 
before program implementation. However, regardless of where students are eating, the content 
of the food is clearly insufficient as evidenced both by the hunger reported and by the 
proportion of students who do not receive a minimum acceptable diet. 
Table 14: Eating and Hunger 
Control Treatment Difference
Percent of parents whose children report being 
hungry during the school day
58.06 54.74 3.32
Male 57.17 62.58 -5.40
Female 58.81 49.51 9.31
Percent of students who have eaten during the 
school day
96.99 98.22 -1.24
Male 97.16 98.26 -1.10
Female 96.83 98.20 -1.36
* p<0.1; ** p<0.05; *** p<0.01 
3.1.3 Improved Student Attendance 
Number of Students regularly (80% or more) attending USDA-supported classrooms 
Student attendance data were gathered from teacher registers. Teachers were asked to report, 
using their registers, the number of boys and girls in each of their classrooms who were missing 
more than 10 days last trimester. According to the Benin Department of Education, the first 
trimester in the 2014-2015 school year began on Thursday October 16th, and ended on Friday 
December 19th. This is a total of 47 school days. If a student is absent for 10 of these 47 days, 
their rate of attendance falls below 80% and they are considered to have irregular attendance. 
Students with absences of 9 days or less (more than 80% of the total school days in the 
trimester) are considered to be regularly attending. This number of students who were not 
regularly attending was subtracted from the total number of students in each grade. See 
Appendix 6.5 for a calculation of this indicator. 
Using this calculation and the weights described above to extrapolate these counts to the entire 
population, the number of students regularly attending intervention schools is 18,893 males and 
18,589 females. This does not meet the goals of 19,607 and 19,069 for male and female 
students respectively. Because this is a count, not a proportion, a statistically significant 
difference is detected from the comparison area. Percentages are included in this baseline 
report to provide context for these totals. On average, 95% of students are regularly attending 
intervention schools, and this proportion is not statistically distinguishable from the comparison 
area. 
27
It should be noted that the 80% cut-off is a low threshold for regular attendance—students 
could be absent for up to two weeks of instruction and still be considered to be attending school 
regularly. It is unlikely that sickness due to malnutrition would cause students to be absent this 
much. Instead, it is likely that they are regularly attending school, but without proper nutrition 
or in poor health (as evidenced by the indicators on hunger and minimum acceptable diet). 
Enrollment 
This baseline study did not gather community-level data on enrollment rates; these will be 
gathered through partnership with AME’s. However, focus groups were conducted in each of 
the four communes of intervention with parents who have not enrolled their children in school. 
They were asked why they have not enrolled their children in school, what they thought the 
cause was for this problem, and what could be done to solve it. Consistently parents answered 
that the cause was economic. Some focus groups cited cultural reasons for the lack of 
enrollment. For example, the mothers’ group in Kandi mentioned that in the past, parents sent 
the children to watch the cattle, and that they needed to do the same. But although this reason 
was interpreted as cultural by respondents, it could also be considered economic. More 
frequently, parents’ and mothers’ groups cited a lack of financial means as the reason for 
choosing to not enroll their children. They referenced the costs of registration, school clothes, 
books, and other supplies. The combination of these costs is prohibitive for the families. 
Additional research could investigate the cut-off point of household income where the cost of 
school enrollment becomes prohibitive, or differences between families with equal incomes that 
send their children to school and those that do not. As no household level data was collected 
from families of unenrolled students, this report cannot provide additional information on these 
behaviors. 
Percent of students who report a decrease in health-related absences 
Data on student health-related absences were gathered in the same manner as data on student 
regular attendance. It was confirmed by local education partners that teachers record not only 
absences but the reason for absences in their registers. Teachers were asked how many boys 
and girls in each of their classes were absent for health-related reasons in the last full month of 
classes. During the time of data collection, the last full month of classes was November. 
Teachers referred to their registers and reported the number of boys and girls. Those students 
were totaled and divided by the total students in each grade. See Appendix 6.5 for the 
calculation of health-related absences at the school level. 
This indicator is defined as “percent of students who report a decrease in health-related 
absences.” The number reported here is the total number of students who were absent for 
health-related reasons. Percentage out of total students is also provided for context. At mid￾line, the difference between total students absent for health-related reasons can be calculated 
to determine the percent change in this indicator. 
Using this calculation, 5% of students in the area of intervention experienced health-related 
absences. Again no statistical difference was observed between the intervention and 
28
comparison areas on this indicator. It should be noted that the percentage is quite low so the 
target of a 10% decrease in this indicator will be hard to detect. 
This rate and the high rate of regular attendance demonstrate that students may be attending 
school while still lacking proper nutrition. This hypothesis is consistent with students’ low 
scores in literacy (presuming that if a student attends school while not receiving adequate 
nutrition, s/he will not be learning well). 
Table 15: Attendance Counts and Rates 
Control Treatment Difference
Students Regularly Attending School(#) 13,793 37,481 -23,688***
Percentage of total students 92.12 94.76 -2.64
Male (#) 7,038 18,893 -11,855***
Percent of total male students 92.49 95.31 -2.82
Female (#) 6,755 18,589 -11,833***
Percent of total female students 91.75 94.21 -2.46
* p<0.1; ** p<0.05; *** p<0.01 
Table 16: Health Related Absences, Counts and Percentages 
Control Treatment Difference
Students with Health-Related Absences (#) 855.66 1,664.46 -808.80*
Percentage of total students 6.44 4.97 1.47
Male (#) 440.22 890.74 -450.52**
Percent of total male students 6.40 5.35 1.05
Female (#) 415.44 773.72 -358.28*
Percent of total female students 6.56 4.60 1.96
* p<0.1; ** p<0.05; *** p<0.01 
Percent of parents in target communities who can name at least 3 benefits of primary 
education
Data on this indicator was gathered by asking parents two questions—first “Do you consider 
your children’s education important?” If parents answered yes, they were asked why. 
Enumerators could classify parents’ answers to categories listed or write in other answers that 
could not clearly fit into one of the categories. Written-in answers were later checked and re￾classified if possible. Each parent was classified as having met the goal (by naming three or 
more reasons), or not (by naming less than three reasons or by saying that his or her child’s 
education is not important). 
Based on these calculations, 39% of parents can name at least 3 benefits of education in the 
intervention areas. Improvement will be needed to meet the LOA target of 60%. Again, no 
statistically significant difference was detected between intervention and comparison groups on 
this indicator. In the intervention area, 47% of male respondents are able to name 3 reasons for 
the importance of education, while only 26% of female respondents are able to do so. This 
difference between men and women in the treatment group is significant to the 5% level (not 
29
reported). However, despite the magnitude of difference between the responses of females in 
treatment and control group, this difference is statistically indistinguishable from zero. No 
difference can be detected between the treatment and control groups, on the percent of 
women who named 3 reasons that education is important. 
Table 17: Importance of Education 
Control Treatment Difference
Can name 3 reasons that education is important (%) 46.72 38.81 7.91
Male 48.32 46.61 1.71
Female 41.42 26.03 15.39
* p<0.1; ** p<0.05; *** p<0.01 
3.1.4 Foundational Results 
Percent of schools in target communities with active PTA’s or other school government 
structures 
Data on parents’ and mothers’ associations were gathered from principals. Principals were 
asked if their schools had parents’ and women’s associations. If they answered yes, they were 
asked how active the association is. The enumerator classified the principal’s answer as “very 
active,” “moderately active,” or “not at all active.” Principals were able to answer based on their 
opinions of what defined these categories. If the principal’s answer was one of the first two 
options, that school’s parents’ or mothers’ association was classified as “active.” Counts of 
active PTA’s and AME’s were applied to the entire population of schools according to the 
weights described previously, and the results can be seen below in Table 12. 122 schools in the 
intervention area have active PTA’s, which is approximately 92% of the schools. Improvement 
will be needed to meet the LOA target of 100%. There is no statistically significant difference 
between number of active PTA’s in the treatment and control schools. 
Only 54% percent of the treatment schools have an active AME, as compared with 15% of the 
control schools. The difference observed between the treatment and control groups in this 
variable is statistically significant to the 99th percentile. Among intervention schools without 
active AMEs, the AME was either inactive (17.11%), or non-existent (26.93%). 
Table 18: Parents' and Women's Associations 
Control Treatment Difference
Schools with Active PTAs (#) 91 122 -31**
Percent 89.57 92.39 -2.82
Schools with Active AMEs (#) 15 72 -58***
Percent 14.90 54.06 -39.16***
* p<0.1; ** p<0.05; *** p<0.01 
Among female parent respondents, the presence of an active AME in a community is positively 
and significantly correlated with knowledge of times to wash hands, and with children receiving 
a minimum acceptable diet (not reported). Of the 556 parents interviewed, only 21 stated that 
30
they were members of their AME. Therefore this correlation could be considered a spillover 
effect of the presence of an active AME in the community. However, this data is not 
experimental so endogenous factors probably exist that contribute to these correlations. No 
such correlations were found between active AME’s and attendance rates, sick rates, or ability 
to name benefits of education. 
3.2 Strategic Objective 2: Increased Use of Health and Dietary 
Practices 
Percent of Students Receiving a Minimum Acceptable Diet 
Data on students’ diets were gathered from parents. The parent was asked to report all the 
items that his or her child ate yesterday. Parents were asked about each of their children 
enrolled in the target school individually, and it was specified that the parent should consider 
food eaten during the day or the evening, and at school, at home or outside the home. The 
parent was asked about food in groups of food items (see Appendix 7.2 Instruments for these 
food groups). The specific foods listed were discussed in the training and pilot study, and edited 
to include local fruits, vegetables and dishes as appropriate. Foods were classified into the 
seven food groups as specified by USDA: 
 Grains, roots and tubers 
• Legumes and nuts 
• Dairy products (milk, yogurt, cheese) 
• Flesh foods (meat, fish, poultry and liver/organ meats) 
• Eggs 
• Vitamin-A enriched foods, including vegetable oil, fruits and vegetables 
• Other fruits and vegetables 
Parents were also asked how often their children ate the day before. Any answer over 7 was 
classified as 7 times. A minimum acceptable diet is defined as eating 4 or more of these food 
groups, and a minimum frequency of 3 times a day. Each child was classified as receiving a 
minimum acceptable diet if the parent reported consumption of food items that fit into at least 
4 of the categories listed above, and if the parent reported that the child ate at least three times 
yesterday. 
According to this definition, 67% of the students in the area of intervention are receiving a 
minimum acceptable diet. Again the comparison area does not exhibit a statistically significant 
difference from the intervention area. Again, improvement is needed in order to reach the LOA 
target of 95%. 
Diet content and frequency are included in Table 17 for additional information. The most 
commonly consumed food groups are #1 (Grains, roots and tubers) and #7(other fruits and 
vegetables). Sauce is classified into this category because many typical Beninese sauces contain 
tomatoes and onions. Eggs and dairy products are the categories of foods consumed by the least 
number of respondents. 
31
In focus groups, parents were asked about what foods they would feed their children if they had 
the means. Most parents mentioned pasta (“macaroni”), rice, couscous and sorghum. Some 
parents mentioned beans, fresh tomatoes, okra, meat or fish, and milk to put in the tea. One 
parent also mentioned a lack of oil for cooking. In all communes, parents cited a lack of means 
as the reason that they were not able to provide these foods to their children. Instead they 
cited the need to feed their children pâte and stretch meals to last for more than 1 day. From 
these focus groups, it can be concluded that parents lacked both complete knowledge of a 
balanced diet and the financial means to provide that diet.6
Table 19: Minimum Acceptable Diet 
Control Treatment Difference
Students Receiving a Min. Acceptable Diet (%) 64.24 67.14 -2.89
Male 63.33 68.04 -4.71
Female 65.00 66.53 -1.53
* p<0.1; ** p<0.05; *** p<0.01 
Table 20: Diet Content and Frequency 
Control Treatment Difference
1. Grains, roots and tubers (%) 92.48 98.31 -5.83**
2. Legumes and nuts (%) 46.53 51.97 -5.43
3. Dairy products (%) 25.51 36.73 -11.21
4. Flesh foods (%) 55.93 77.28 -21.34***
5. Eggs (%) 4.83 11.97 -7.13
6. Vitamin A Rich foods (%) 96.82 95.41 1.41
7. Other fruits and vegetables (%) 100.00 99.79 0.21
Eats three or more times a Day (%) 77.59 78.08 -0.48
* p<0.1; ** p<0.05; *** p<0.01 
3.2.1 Improved Knowledge of Health and Hygiene Practices 
Percent of Parents in target communities who achieve a passing score on a test of good 
health and hygiene practices 
Data on parent knowledge of health and hygiene practices was gathered from parents. Parents 
were asked questions related to USAID’s Hygiene Improvement Project (HIP). Specifically, 
questions from this section were taken from a verified instrument used to determine baseline 
indicators for a USAID HIP in Madagascar.7
 
As a “passing score” threshold was not yet determined by CRS at the time of writing this report, 
summary statistics are reported instead. Some indicators reflect parents’ knowledge, such as 
the ability to name practices or products that can make water safe to drink, and reasons or 
 
6
Focus groups with PTA’s and AME’s in Kandi, Malanville, Kalale, Gogounou. Jan 20-23 2015. 
7
 To see this report and the full instrument as developed by FHI360, please visit: 
http://www.hip.fhi360.org/file/16498/HIP%20Madagascar%20Baseline%20Report%20%20Dec.%2008.pdf 
32
times to wash hands. Other indicators reflect parents’ practices, such as the toilet that the 
family uses, or the place in which they disposed of their youngest child’s feces. These indicators 
can be combined and weighted according to the planned awareness campaigns, to develop a 
health and hygiene score. From that score, a cut-off for passing can be determined. 
In most indicators, intervention and comparison areas are balanced, meaning that there is no 
detectable difference between the two groups. However, 42% of respondents in treatment 
areas reported open defecation (defined as defecation in the field, nature, or otherwise outside 
the home but not in a latrine of any kind), while 77% of respondents in the comparison area 
reported this. The difference of 36% is statistically significant at the 1% level. In the case of 
toilet use, the control group does not provide an adequate comparison for the treatment group. 
During focus groups, parents were asked how the health and hygiene situation can be improved 
in the home. Some parents answered that education was necessary about healthy food 
preparation, latrine use, water purification and hand-washing.8
 In most focus group discussions, 
the lack of both potable water and clean water for washing was mentioned. The focus group in 
Kandi also mentioned the possibility of a community garden for the women to grow fruit.9
 In 
Malanville, parents reported that people with means are digging wells and selling water suitable 
for washing. These new water sources reduced their trips to the river for water.10
Table 21: Knowledge of Good Health and Hygiene 
Control Treatment Difference
Can name 1 water sanitization product (%) 43.82 73.19 -29.37***
Can name 2 water sanitization products (%) 10.25 25.96 -15.71*
Can name 3 water sanitization products (%) 0.53 0.00 0.53
Can name 1 reason for hand washing (%) 99.65 100.00 -0.35
Can name 2 reasons for hand washing (%) 40.49 49.15 -8.66
Can name 3 reasons for hand washing (%) 16.53 13.66 2.86
Can name 1 time to wash hands (%) 98.49 99.62 -1.14
Can name 2 times to wash hands (%) 73.99 75.81 -1.81
Can name 3 times to wash hands (%) 27.51 48.27 -20.76**
Open Defecation (%) 77.42 41.53 35.89***
* p<0.1; ** p<0.05; *** p<0.01 
Table 22: Place in which youngest child's feces were disposed 
Control Treatment Difference
In the latrine (%) 17.84 52.75 -34.92***
Pit or garbage (%) 15.96 16.13 -0.17
In the road (%) 0.30 0.54 -0.24
Outside (%) 60.67 25.03 35.63***
In a public latrine (%) 0.41 1.44 -1.03
 
8
 Focus Group Discussion, Kalale and Malanville PTA. 23 January 2015. 
9
 Focus Group Discussion, Kandi PTA. 23 January 2015. 
10 Focus Group Discussion, Malanville PTA. 23 January 2015. 
33
In the sink or the tub (%) 0.35 0.00 0.35
Down a drain (%) 2.93 0.42 2.51
Don't Know (%) 1.54 3.69 -2.15
* p<0.1; ** p<0.05; *** p<0.01 
Table 23: Water products named 
Control Treatment Difference
Liquid chlorine (Sur ’Eau) (%) 6.81 21.81 -15.00*
Chlorine tablets (Aqua tabs) (%) 27.05 42.84 -15.80*
Chlorine flocculants (%) 0.30 5.09 -4.80
Filter (PuR, Watermaker) (%) 1.33 5.13 -3.80
Other chlorine products (%) 18.41 20.90 -2.49
Iodine (tablets or liquids) (%) 0.00 2.62 -2.62
Permanganate (%) 0.70 0.74 -0.05
Nothing (%) 16.92 3.92 12.99***
* p<0.1; ** p<0.05; *** p<0.01
Table 24: Reasons to wash hands named 
Control Treatment Difference
Prevention of diarrhea (%) 23.40 33.37 -9.97
Prevention of other illnesses (%) 82.99 86.65 -3.67
Removing germs (%) 33.75 29.55 4.21
To prevent dirt from entering the mouth (%) 12.22 8.98 3.23
To prevent dirt from entering food (%) 7.39 6.18 1.21
Good health (%) 0.68 0.33 0.35
* p<0.1; ** p<0.05; *** p<0.01 
Table 25: Times to wash hands named 
Control Treatment Difference
After Defecation (%) 51.89 65.82 -13.93
Before Eating (%) 93.70 90.56 3.14
After washing children or diapers (%) 1.84 10.91 -9.07***
After washing latrines (%) 1.89 10.32 -8.43*
After washing pots (%) 1.57 1.97 -0.41
Before preparing food (%) 2.24 7.65 -5.42
Before feeding children (%) 3.79 5.83 -2.03
After eating (%) 48.32 48.24 0.08
* p<0.1; ** p<0.05; *** p<0.01
34
Table 26: Ways to make water better for drinking named 
Control Treatment Difference
Boiling (%) 22.38 27.06 -4.68
Add Javel (%) 6.40 14.09 -7.70
Add chlorine solutions (Sur’ Eau) (%) 2.22 7.16 -4.94
Add chlorine tablets (Aqua tabs) (%) 10.74 12.52 -1.78
Pass it through a fabric (%) 10.74 10.57 0.17
Let it sit (%) 3.13 4.05 -0.92
Use a ceramic filter (%) 0.30 0.00 0.30
Use a bio-Sable filter (%) 0.00 0.00 0.00
Solar disinfection (%) 0.00 0.00 0.00
Keep it in a covered container (%) 23.53 12.20 11.33*
* p<0.1; ** p<0.05; *** p<0.01
Percent of Children in target communities who achieve a passing score on a test of good 
health and hygiene practices 
Table 27: Student Knowledge of when to wash hands 
CI CP
Indicator Control Treatment Diff Control Treatment Diff
After defecate (%) 1.6 0.4 1.3 0.8 1.6 -0.8
Before eating (%) 32.4 34.8 -2.4 45.7 49.4 -3.7
After taking care of baby (%) 0.0 0.0 0.0 0.0 0.0 0.0
After cleaning latrines (%) 0.0 0.0 0.0 0.0 0.8 -0.8
After cleaning the toilet (%) 0.0 0.0 0.0 0.0 0.0 0.0
Before preparing food (%) 0.7 0.1 0.5 0.7 0.1 0.7
Before giving food to 
children (%)
0.8 0.0 0.8 0.5 0.0 0.5
After eating (%) 12.7 15.2 -2.5 18.2 25.3 -7.1
None (%) 67.0 64.7 2.3 53.1 49.5 3.6
* p<0.1; ** p<0.05; *** p<0.01 
Table 28: Student Knowledge of hygiene practices 
CI CP
Indicator Control Treatment Diff Control Treatment Diff
Personal hygiene (%) 0.0 0.0 0.0 0.8 1.0 -0.1
Wash hands (%) 0.0 0.0 0.0 6.9 9.8 -2.9
Bathing every day (%) 0.0 0.0 0.0 11.1 18.6 -7.4
Clean environment (%) 0.0 0.0 0.0 0.7 0.0 0.7
Clean School (%) 0.0 0.0 0.0 0.9 1.3 -0.5
Cleaning classroom (%) 0.0 0.0 0.0 0.8 1.5 -0.8
Brush teeth/ mouth hygiene 
(%)
0.0 0.0 0.0 0.1 0.3 -0.2
Healthy nutrition (%) 0.0 0.0 0.0 0.0 0.1 -0.1
None (%) 100.0 100.0 0.0 81.9 76.8 5.0
* p<0.1; ** p<0.05; *** p<0.01 
35
3.2.5 Increased Access to Preventative Health Interventions 
Number of students receiving de-worming medication 
Data on number of students receiving de-worming medication were gathered from principals. 
Principals were asked if any students from their schools received de-worming pills from any 
campaigns this year. If they answered yes, the principals were asked how many students 
benefitted from the campaign. Twelve principals in the intervention area noted that deworming 
campaigns had benefitted some of the students in their schools. The reported numbers below 
reflect the weighting of the schools sampled within the intervention and comparison areas. 
Thus, among all the intervention schools, it is estimated that 5,876 students have benefitted 
from de-worming campaigns this year. Program activities such as provision of deworming pills 
should help to meet the LOA target of 43,804 students within the program. 
Table 29: Number of Students Dewormed 
Control Treatment Difference
Number of Students Dewormed 3,504.43 5,876.05 -2,371.62
Percent 20.54 26.00 -5.46
4. Results of EGRA 
Apart from an initial section that collected student environment information, the test used in 
this study had 9 sections: 1) vocabulary 2) oral comprehension 3) knowledge of letter names 4) 
identification of letter sounds 5) simple words reading 6) invented words reading 7) reading and 
comprehension 8) Oral comprehension and 9) Dictation. The results of each section are shown 
here. The test was based on a model prepared by World Education Benin. We adapted the test 
in consideration of the purpose of the test and the place in the curriculum where students 
should be in the academic year: 1. Question on households assets were eliminated 2. An open 
ended question requesting examples on what reading material is available at home was 
eliminated. 3. Two questions about hygiene knowledge and practices were added, these were 
reported above. 4. The section on producing letter sounds was not used, and 5. Only one of the 
sentences was used for dictation. These changes were done in consideration to the level of 
knowledge of the students in the grades that were taking the EGRA. Specifically, because all 
students had completed one trimester of classes, they were not expected to have the reading 
and comprehension that at student at the end of the school year might have. This was 
particularly true for students in the first grade, or Cours d’Initiation (CI) level. Also, the 
introduction of the letter sounds is a new development in the national curriculum that is 
beginning to be implemented. 
4.1 Vocabulary 
During the vocabulary section, the enumerator asked the students to point at different items, in 
total there were 17 items. Each item was asked once to the student and it was marked as 
correct or incorrect according to the student’s response. The items were body parts (head, knee, 
36
ears, foot, stomach), school items (eraser, blackboard, notebook, chalk), things from nature (the 
sun, a stone, a bowl and a tree), and spatial terms (a door, the top of the table, a rag, the corner 
of the table). The following table presents the average percentage of objects correctly identified 
by each student. 
Table 30: Percentage of vocabulary known by grade and type of school: 2015 
CI CP
Indicator Control Treatment Diff Control Treatment Diff
Mean 39.0 41.8 -2.8 60.4 61.2 -0.8
* p<0.1; ** p<0.05; *** p<0.01 
Students in first grade, or CI were able to identify about 40 % of the items that were requested, 
while students in second grade, or Course Préparatoire (CP) were able to identify 60% of all the 
items that the enumerators ask for. There is no statistically significant difference between 
students from control and treatment schools. 
4.2 Oral comprehension 
The oral comprehension section is an extension to the vocabulary section. The enumerator 
asked students to perform some actions using a pencil or a rag. In total the students were asked 
four groups of actions. In Group A, the enumerator asked students to pick a pencil and put it on 
top of a sheet of paper, behind them, on the soil, under the paper, by their right side, and in a 
box. In Group B the enumerator ask students to take the pencil with their hand and put it on the 
table on their right side. In Group C, the enumerator asks students to take a chalk with their 
right hand and put it under the table. Finally, in Group D, the enumerator asks students to take 
an eraser with their right hand and put it under the chair. The total number of actions of all 
these requests added to 19. The following table shows the average percent of actions correctly 
done by students. 
Table 31: Percentage of actions correctly performed by grade and type of school: 2015 
CI CP
Indicator Control Treatment Diff Control Treatment Diff
Mean 44.6 52.1 -7.5** 67.2 67.0 0.2
* p<0.1; ** p<0.05; *** p<0.01 
About half of the students in CI were able to complete their tasks correctly, and more than two 
thirds of students in CP were able to complete their tasks correctly. There is no statistically 
significant difference between students in treatment and control groups. 
4.3 Knowledge of letter names 
In the section on knowledge of letters, the students were asked to read as many letters as they 
could in a minute. The enumerator presented a grid containing 100 letters organized in table of 
10 rows by 10 columns. Letters were capital or lower case and some had accents. If students 
37
could not read any of the letters in the first row correctly they were asked to stop this section of 
the test and move to the next one. The following table presents the number of letters correctly 
read per minute. 
Table 32: Average number of letters correctly read per minute by grade and type of school: 2015 
CI CP
Indicator Control Treatment Diff Control Treatment Diff
Mean 1.8 0.8 1.1*** 11.8 9.0 2.7
* p<0.1; ** p<0.05; *** p<0.01 
The number of letters per minute by CI students is very low, about 1 letter per minute, while for 
CP students the average was about 10 letters per minute. There is a statistically significant 
difference in CI level but the difference is not large in magnitude. There is no statistically 
significant difference at the CP level. 
4.4 Identification of letter sounds 
In the identification of letter sounds section, the enumerator read ten words twice. The student 
was asked to identify the initial sound—not letter name—of each word which was read. The 
enumerator stopped the section if the student couldn’t identify the first five words. The 
following table presents the average percent of sounds correctly identified by the students. 
Table 33: Average percentage of sounds correctly identified by grade and type of school: 2015 
CI CP
Indicator Control Treatment Diff Control Treatment Diff
Mean 0.8 0.7 0.0 8.2 7.5 0.6
* p<0.1; ** p<0.05; *** p<0.01 
The percentage of sounds identified by CI students is less than one, while for CP students were 
able to identify about 8 percent of the words. There is no statistically significant difference 
between treatment and control schools at either level. 
4.5 Simple words reading 
In this section, the enumerator presented a grid with 50 words and asked student to read as 
many words as possible in one minute. The majority of the words have one or two syllables. The 
section is stopped if the student cannot read the first five words correctly. The following table 
presents the average number of words correctly read per minute. 
Table 34: Average number of words correctly read per minute by grade and type of school: 2015
CI CP
Indicator Control Treatment Diff Control Treatment Diff
Mean 1.0 1.1 -0.0 1.6 1.9 -0.4
* p<0.1; ** p<0.05; *** p<0.01 
38
The results of this section show that there is little difference between CI and CP in terms of 
words read. CI students are able to read a word per minute, while CP students can read close to 
2. There is no statistically significant difference between treatment and control. 
4.6 Invented words reading 
Similar to the previous section, the enumerator presented a grid with invented words to the 
student who should read as many as possible in a minute. Most words have one or two syllables 
and have no meaning in the French language. The section is stopped if the student cannot read 
the first five words correctly. The following table presents the number of words correctly read 
per minute. 
Table 35: Average number of invented words correctly read per minute by grade and type of school: 2015 
CI CP
Indicator Control Treatment Diff Control Treatment Diff
Mean 0.9 0.1 0.8 1.3 0.7 0.6
* p<0.1; ** p<0.05; *** p<0.01 
The number of words read per minute is low for both CI and CP. CI students are able to read less 
than a word per minute, while CP students are able to read about a word per minute. There is 
no statistically significant difference between treatment and control schools. 
4.7 Text reading and comprehension 
This section had two parts. In the first part, students were asked to read a short text, or as 
many words as possible in a minute. In the second part, a series of questions are asked to the 
student on what they were able to read. There are a total of 5 questions, but students are 
asked the questions in proportion to how much they were able to read. The following table 
presents the number of words read per minute and questions correctly answered. 
Table 36: Average number of words read per minute and questions responded correctly by grade and type of 
school: 2015 
CI CP
Indicator Control Treatment Diff Control Treatment Diff
Number of words
read per minute
3.4 3.4 0.1 6.4 4.2 2.1
Number of 
questions answered 
correctly
0.0 0.0 0.0 0.0 0.0 0.0**
* p<0.1; ** p<0.05; *** p<0.01 
The number of words read per minute is low—CI students were able to read 3.4 words per 
minute, while CP students were able to read about 5 words per minute. The minimum number 
of words that student needs to read for the enumerator to ask the first question is 13, so given 
the low number of words read, it is not surprising that on average the number of questions 
answered is close to zero. There is a statistically significant difference between treatment and 
39
control at the CP level, but given that the number is so small it has been rounded to zero, the 
magnitude of this difference is negligible. 
4.8 Oral comprehension 
In the section of oral comprehension, the enumerator read a story twice in French and then 
asked 5 questions about the story. The following table presents the average percentage of 
questions correctly answered by students. 
Table 37: Average percentage of questions correctly answered by grade and type of school: 2015 
CI CP
Indicator Control Treatment Diff Control Treatment Diff
Mean 0.0 0.0 0.0 0.1 0.1 -0.0
* p<0.1; ** p<0.05; *** p<0.01 
There is not much difference between CI and CP students, the percentage of answers correctly 
answered by the students is close to zero, and there is not a statistically significant difference 
between treatment and control schools. 
4.9 Dictation 
In this section, the enumerator read a sentence three times. The first time students listened; the 
second time the enumerator read slowly and students wrote as best as they can; the third time, 
the enumerator read the text so students could correct what they wrote. The enumerator then 
graded the text by classifying each word as correct, partially incorrect or incorrect. A correctly 
written word is worth twice as a partially correct one. The students did not get any credit if they 
wrote the word incorrectly or nothing at all. Punctuation and spacing were also measured 
separately. The following table shows the results in percentages according to each category. 
Table 38: Percentage of words correctly written by grade and type of school: 2015.
CI CP
Indicator Control Treatment Diff Control Treatment Diff
Students that do not 
write (%)
80.0 85.2 -5.2 39.6 52.5 -12.9*
Percentage of words 
correctly written
0.7 0.2 0.6 8.4 5.6 2.8*
Percentage of words 
partially correctly 
written
0.9 0.5 0.4 5.7 3.2 2.5*
Percentage of words 
incorrectly written
91.8 95.6 -3.8* 76.8 81.8 -5.0
* p<0.1; ** p<0.05; *** p<0.01 
The percentage of students that cannot write is high in CI close to 82%, while in CP the 
percentage is close to 46 percent. Among the CI students that were able to write, more than 90 
percent wrote the words incorrectly. Among CP students, this percentage was in the upper 70 
40
percent. There were statistically significant differences between treatment and control schools. 
At the CI level, the percentage of words incorrectly written is higher in treatment schools. 
However, the difference is negligible. At the CP level, there is a difference of almost 13 
percentage points between students that cannot write—treatment schools have a higher 
percentage of those students than control schools. In addition, the percentage of words 
correctly written and partially correctly written is higher in control schools. These differences 
are negligible. However, within the sample of schools selected, there is no known reason why 
control schools would be better at writing than treatment schools. 
5. Conclusions and Recommendations 
Strategic Objective 1: Improved Literacy of School-Aged Children 
Approximately 90% of teachers in the intervention schools dedicate at least 45 minutes per day 
on literacy instruction. Teachers’ attendance rates are 92%, and 88% of the schools are using 
the national curriculum. However, key informant interviews reveal that there is a significant 
lack of educated teachers and materials. Thus, while teachers are present and have been told 
to use the national curriculum, they likely do not have sufficient training or resources to provide 
quality instruction. In addition, some are be called upon to teach two classes simultaneously, 
further reducing quality of instruction. 
On average, a classroom in the intervention area had 74% attentive students. 48% of students 
report being hungry during the school day to their parents, despite the fact that 98% of them 
had eaten during the school day. Some parents indicated needing to stretch meals over 
multiple days and being limited to serving ‘pate’. 95% of students in the treatment area are 
defined as regularly attending school, and only 5% of them experience health-related absences. 
While it appears that a lack of nutrition has not lead to absences, providing proper nutrition 
could cause fewer students to complain of hunger, and more students to be attentive during the 
school day. 
Only 40% of parents of enrolled students can name three reasons that education is important. 
It can be presumed that parents who did not enroll their children in school could name even 
fewer reasons for the importance of education, meaning that this estimate is positively biased. 
Efforts to improve awareness of the importance of education could improve student enrollment 
and performance. 
93% of the schools in the treatment area had active parents’ associations; however, only 54% 
had active mothers’ associations. The rest are non-existent (17%) or inactive (27%). Providing 
trainings could potentially increase the number of active women’s associations, which may lead 
to improved student enrollment and diet, as well as community knowledge of health and 
hygiene. Analysis demonstrates a correlation between active AME’s and proportion of children 
who receive a minimum acceptable diet, as well as knowledge of health and hygiene practices. 
Further research is necessary to understand current effectiveness of these associations, as well 
41
as needs for trainings and capacity-building. Local partners, upon working with these 
associations, can gather this information from association members. 
Recommendation: 
Theories of change about the impact of FFE programs demonstrate that the effect of an FFE 
program can be ambiguous if it increases enrollment and reduces farming labor in the 
community.11 In the case of Benin, schools lack an adequate supply of well-trained teachers 
prior to program implementation so increased enrollment could reduce effectiveness of the 
program on indicators such as literacy scores or quality of instruction. To address this, the 
program may choose to focus on increasing teaching staff and providing necessary resources for 
quality instruction. Indicators related to this activity could include student-to-teacher ratio or 
resources available to teachers. In addition, monitoring quality of literacy instruction could 
provide richer information. If quality of instruction is to be monitored, implementing partners 
would need to develop a rubric for assessing quality of literacy education. 
Because many schools do not have an active AME, the program should focus on supporting 
AME’s that require revitalization or initiation. Community level trainings on health, hygiene, and 
the importance of education can be implemented with the help of the AME. By working with the 
AME’s, program implementers can also gather information about local enrollment rates and 
percent of parents of unenrolled students who can name the benefits of education. 
Strategic Objective 2: Increased Use of Health and Dietary Practices 
67% of students in the intervention area are receiving a minimum acceptable diet as defined by 
the USDA. Approximately 40% of parents in the treatment areas indicated that their households 
usually practice open defecation. 8% of parents in treatment areas demonstrated knowledge of 
washing hands before preparing food, and approximately 1% of students demonstrated 
knowledge of the need to wash hands after defecation. These and other indicators related to 
health and hygiene can be improved. Approximately 5,876 students (26%) in the intervention 
area have benefitted from de-worming campaigns. Improvement is needed in the number of 
students receiving a minimum acceptable diet, students receiving de-worming pills, and student 
and parent knowledge of health, nutrition, and hygiene/sanitation practices. 
Recommendation: 
More students need to receive minimum acceptable diet. Provision of meals and take-home 
rations will improve this indicator, but education and outreach to parents can help as well. 
Education is needed on health and hygiene practices for both students and parents. 
Students’ Literacy: EGRA 
 Because the threshold which will define “reading and comprehension at grade level” has not 
yet been developed at the time of writing this baseline report, individual element of the EGRA is 
 
11 Braun, J. (2009). Ch. 23: How Effective Are Food-for-Education Programs? A Critical Reassessment. In The poorest 
and hungry: Assessment, analyses, and actions. Washington, D.C.: International Food Policy Research Institute. <
http://www.ifpri.org/sites/default/files/publications/oc63ch23.pdf>.
42
addressed here. Compared to other sections, students demonstrated some proficiency in 
vocabulary identification and oral comprehension of commands. In other sections such as 
knowledge of letter names, identification of letter sounds, reading and comprehension of texts, 
and dictation, students performed at very low levels. 
Recommendation on Timing of Future EGRA Implementation 
Because EGRA was conducted in the middle of the school year, the tool was not sufficient for 
detecting students’ literacy levels, particularly among first grade (CI) students. Performing the 
EGRA at the end of the school year will provide richer information on students’ literacy levels, 
particularly for second grade (CP) students. Furthermore, the indicator of interest specifically 
refers to “the end of two grades of primary schooling.” For this reason, we recommend 
repeating the EGRA at the end of this school year. After less than half a year of program 
implementation, data gathered at the end of the first year could still be considered a baseline 
indicator for academic outcomes. 
The EGRA should be implemented at treatment schools at the end of each academic year of the 
program for monitoring purposes. However, the EGRA results reported here can serve as 
baseline data in terms of impact evaluation. The difference-in-difference model allows for the 
calculation of impact even if the assessment is implemented at a different time of the year, 
because with a comparison group, that difference can be subtracted, or “differenced out.” 
Because of this flexibility, we recommend implementing the EGRA at the end of the academic 
year for any impact evaluation. 
Recommendation: On the Adequacy of the Comparison Group 
Statistically significant differences between treatment and control group were noted when 
observed throughout this report. In order for a group to provide an adequate comparison, it 
should not differ from the intervention group on both observable characteristics and 
characteristics that the program will not impact—such as gender, age, experience and education 
level. First, schools in the control area were significantly smaller than those in the treatment 
area. This causes concern because larger schools may be fundamentally different from smaller 
ones and thus not comparable. 
Several differences were observed in regards to gender. The intervention area has significantly 
more female teachers (a difference of 16 percentage points). More female parents were 
interviewed in the treatment area (15 percentage points), and principals reported a larger 
number of active mothers’ associations in the treatment areas (39 percentage points). This is 
especially of concern given that the presence of an active mothers’ association appears to be 
correlated with outcome indicators such as receiving a minimum acceptable diet and knowledge 
of health and hygiene practices. 
In most outcome-level indicators, we observed very few differences between treatment and 
control areas that were both statistically significant and large enough in magnitude to be of 
concern. One such difference was open defecation; however, it is unlikely that this program will 
43
impact this behavior greatly, since it is not providing new infrastructure at the community level 
such as latrines. 
Contamination could be an issue among the schools that exist within the intervention area. 
While most program activities will be occurring within the school, some components of the 
program could affect the entire community. Because Kalale commune is large, distance from 
the control schools and implementation schools can be calculated to determine if this is a 
concern. If the schools are sufficiently close to each other, outreach activities and enrollment 
campaigns could affect members of both the treatment and control group if they are 
implemented on a large scale. If these components are implemented within the context of the 
intervention school (for example, if education on health and hygiene is provided to parents’ 
association members only), contamination will be minimized. In this way, program 
implementation will affect concerns of contamination. 
We recommend addressing the above concerns in econometric models, which can account for 
differences in school size, gender proportions, and potential contamination. Mid-line 
evaluations should use the same comparison group but pay careful attention to indicators that 
need to be disaggregated by gender, making attempts to gather responses from equal portions 
of males and females at each level of data collection. Finally, respondents in the comparison 
area should be asked if any similar intervention that has been implemented during the program 
implementation. Any school which has experienced a similar intervention can no longer provide 
an adequate counterfactual. 

45
Questionnaire
2.5 Increased 
Access to 
preventative 
health 
interventions
How many students currently receive de-worming 
medication?
Principal 
Questionnaire
Key Questions by Source: 
Students: 
 EGRA Literacy tests 
 test of good health and hygiene 
Parents: 
 Children’s health-related absences 
 Benefits of primary education 
 Minimum Acceptable Diet of child 
 Test of good health and hygiene 
 Priorities for improving school infrastructure 
 Percent of students who report being hungry during the day 
Teachers: 
 Amount of time (daily) spent on literacy lessons 
 Students who regularly attend school 
 Percent of students who experience health related absences 
Principals: 
 Teacher attendance rate 
 Existence an active levels of a PTA 
 Current de-worming campaigns in the schools 
Classroom Observation: 
 Percent of students that are attentive 


61
6.4 Summary Statistics 
Table 39: EGRA Students’ Previous Education 
CI CP
Indicator Control Treatment Diff Control Treatment Diff
Attended CI (%) 12.1 10.4 1.7 79.5 82.1 -2.6
Attended CP (%) 0.6 2.5 -1.9 20.1 17.6 2.5
Not at school (%) 87.3 87.1 0.2 0.4 0.2 0.1
Attended kindergarten (%) 22.2 33.6 -11.4 27.3 30.9 -3.6
* p<0.1; ** p<0.05; *** p<0.01 
Table 40: EGRA Students’ Languages spoken at home 
CI CP
Indicator Control Treatment Diff Control Treatment Diff
Baatonou (%) 5.9 0.4 5.4 4.3 2.4 1.9
Bariba (%) 67.9 39.3 28.6** 74.8 41.4 33.4**
Dendi (%) 0.4 24.3 -23.9** 0.4 26.7 -26.3***
Mokole (%) 0.1 5.4 -5.3** 0.0 4.5 -4.5*
Peulh (%) 14.4 5.7 8.7* 9.7 4.0 5.7
Boo (%) 5.3 8.3 -3.0 7.1 5.7 1.4
Zarma (%) 0.4 1.5 -1.1 0.0 1.4 -1.4*
Djerma (%) 0.9 8.7 -7.8 0.4 6.9 -6.5*
Don't know (%) 0.7 0.2 0.4 1.0 0.0 0.9
Other (%) 4.5 6.6 -2.2 3.1 7.4 -4.3
* p<0.1; ** p<0.05; *** p<0.01 
Table 41: EGRA Students’ Reading materials 
CI CP
Indicator Control Treatment Diff Control Treatment Diff
Reading book or manual (%) 32.0 33.4 -1.3 47.1 44.1 3.0
Don't know (%) 13.4 13.0 0.4 5.7 4.7 1.0
Read books, journals, periodicals at 
home (%)
16.1 18.8 -2.7 22.6 27.3 -4.7
Don't know (%) 20.2 20.4 -0.2 10.7 4.5 6.2**
* p<0.1; ** p<0.05; *** p<0.01 
62
Table 42: EGRA Students’ Literacy help at home 
CI CP
Indicator Control Treatment Diff Control Treatment Diff
Father (%) 22.7 32.2 -9.5* 33.8 48.7 -14.9**
Mother (%) 13.0 18.3 -5.3 13.8 21.5 -7.8**
Brother (%) 23.0 30.6 -7.6 34.4 49.8 -15.4***
Sister (%) 13.1 22.3 -9.3* 23.0 29.6 -6.6
Other (%) 3.2 5.5 -2.3 4.3 3.6 0.7
Someone reads at home (%) 40.4 47.8 -7.4 55.9 71.4 -15.4**
Don't know (%) 23.6 23.8 -0.2 11.3 10.0 1.3
* p<0.1; ** p<0.05; *** p<0.01 
Table 43: Who helps with homework at home 
CI CP
Indicator Control Treatment Diff Control Treatment Diff
Father(%) 7.8 15.1 -7.3* 11.9 12.9 -1.0
Mother(%) 3.8 9.0 -5.2*** 5.0 9.6 -4.6
Brother(%) 14.9 13.1 1.8 19.8 28.1 -8.2
Sister(%) 7.1 7.5 -0.5 9.1 10.6 -1.4
Other(%) 2.8 7.3 -4.5 5.7 12.7 -7.1
* p<0.1; ** p<0.05; *** p<0.01 
Table 44: After school activities 
CI CP
Indicator Control Treatment Diff Control Treatment Diff
Go to the fields (%) 21.5 16.4 5.1 23.9 22.7 1.2
Household chores (%) 22.9 33.3 -10.4 33.2 33.2 -0.1
Sell things with parents (%) 1.8 2.3 -0.5 1.4 7.1 -5.7
* p<0.1; ** p<0.05; *** p<0.01 
Table 45: Eating at school 
CI CP
Indicator Control Treatment Diff Control Treatment Diff
Eat during recess (%) 7.7 10.3 -2.6 3.7 3.6 0.0
Eat at school kitchen (%) 16.7 22.7 -6.0 10.9 8.4 2.6
Times per week eat at school 
kitchen (mean)
1.0 2.0 -1.0* 1.3 2.5 -1.2
* p<0.1; ** p<0.05; *** p<0.01 
63
6.5 Calculation of Indicators 
Below are the equations for all indicators in this study. 
Formulas below describe the calculation of the indicator to the level of school j. Means of all indicators 
within the specified strata were then calculated, and weights were applied as described in the section 
on weighting. The notation follows a similar pattern: j refers to school, p refers to parents, c refers to 
classrooms, s refers to students, and t refers to teachers. 
Indicator Source Description Calculation
Average 
Teacher 
Attendance 
Rate
Principal 
Survey
The principal reported the 
number of days (𝒂𝒋𝒕 ) 
teacher t was absent. The 
reported number of days 
was subtracted from the 
total number of days last 
trimester (47) and divided 
by the total number of days 
to achieve the rate 
reported 𝒓𝒋𝒕 (see the 
following equation). 
Teacher attendance rates 
were then averaged by 
school by summing all the 
rates and dividing by the 
number of teachers 
surveyed in school (𝒏𝒋𝒕).
𝒓𝒋𝒕 =
𝟒𝟕 − 𝒂𝒋𝒕
𝟒𝟕
𝒓𝒋 =
∑ 𝒓𝒋𝒕
𝒏𝒋𝒕
𝒕=𝟏
𝒏𝒋𝒕
Improved 
Attentiveness
Classroom 
Observation
𝒂𝒋𝒄𝒊𝟏
is the attention for 
student i in classroom c1, 
and 𝒔𝒄𝟏
is the number of 
students observed in 
classroom c1. Likewise 
𝒂𝒋𝒄𝒊𝟐
is the attention for 
student i in classroom c2
and 𝒔𝐜𝟐
is the number of 
students observed in 
classroom c2. 𝒂𝒋
is the 
average of the two average 
attentiveness rates in the 
two classrooms observed.
𝒂𝒋 =
∑ 𝒂𝒋𝒄𝒊𝟏
𝒏𝒄𝟏
𝒊=𝟏
𝒔𝒄𝟏
+
∑ 𝒂𝒋𝒄𝒊𝟐
𝒏𝒄𝟐
𝒊=𝟏
𝒔𝒄𝟐
𝟐
Students who 
Indicate 
Hunger
Parent Survey Percent of parents who
indicate that their child was
hungry (𝒉𝒋) at school j was 
calculated where 𝒉𝒋𝒑 is a 
binomial indicator of 
𝒉𝒋 = ∑
𝒉𝒋𝒑
𝒑𝒋
𝒑𝒋
𝒊=𝟏
64
whether or not parent p 
responded that any of his 
or her children were hungry 
during the school day last 
week, and 𝒑𝒋
is the number 
of parents on which parent 
p provided data. 
Students who 
consume 
daily meals 
during the 
school day
Parent Survey Percent of students who
consumed a meal during 
the day (𝒆𝒋𝒑) was 
calculated as a mean of 
means. 𝒆𝒋𝒑𝒊 is a binomial 
indicator of whether or not 
student i reported hunger 
during the school day last 
week, and 𝒔𝒋𝒑 is the 
number of children on 
which parent p provided 
data. 
𝒆𝒋𝒑 = ∑
𝒆𝒋𝒑𝒊
𝒔𝒋𝒑
𝒔𝒋𝒑
𝒊=𝟏
Number of 
Students 
Regularly 
Attending 
School
Teacher 
Survey
The number of students 
regularly attending, 𝒈𝒋 was 
calculated by subtracting 
𝒊𝒋𝒄 , the number of students 
irregularly attending from 
𝒔𝒋𝒄 the total number of 
students in class c. This 
number was summed 
across all classrooms c in 
school j.
𝒈𝒋 = ∑(𝒔𝒋𝒄 − 𝒊𝒋𝒄
𝟔
𝒋𝒄=𝟏
)
Percent of 
Students who 
report a 
decrease in 
health￾related 
absences
Teacher 
Survey
𝒉𝒋𝒄is the number of 
students in class c of school 
j who had health-related 
absences. 𝒔𝒋𝒄 is the total 
number of students in class 
c of school j. This rate was 
summed for all classes in 
school j, and divided by𝒄𝒋
,
the number of classes in 
school j to create a school￾level rate of health-related 
absences. 
𝒅𝒋 =
∑ (
𝟔
𝒋𝒄=𝟏
𝒉𝒋𝒄
𝒔𝒋𝒄
)
𝒄𝒋