Food for Education Project phase 2 
Implemented by Planet Aid International and ADPP Mozambique 
Funded by the US Department of Agriculture
Midterm Evaluation Report 
Author: Simone Doctors 
with Forcier Mozambique and NFER 
30 October 2019 
Agreement Number: FFE-656-2015/009-00
2
Table of contents
Acknowledgments..................................................................................................................................................5
List of abbreviations and acronyms........................................................................................................................6
Executive summary.................................................................................................................................................7
Introduction..........................................................................................................................................................14
Background.......................................................................................................................................................15
Overview of the FFE2 Midterm Evaluation.......................................................................................................17
Alignment with previous and future evaluations.........................................................................................17
Methodology ....................................................................................................................................................18
Information gathered for the midterm evaluation.......................................................................................18
Sampling for pupil-level data collection ...........................................................................................................20
Challenges, delays and lessons learned............................................................................................................26
Midterm findings..................................................................................................................................................29
Improved literacy of school-age children (SO1) ...............................................................................................29
Improved Quality of Literacy Instruction (MGD 1.1) ........................................................................................52
Access to School Supplies and Materials (MGD 1.1.2) .....................................................................................52
Literacy Instructional Materials (1.1.3).............................................................................................................56
Skills and Knowledge of Teachers (1.1.4) .........................................................................................................56
In-service training in literacy of primary school teachers.............................................................................57
Initial teacher education in teacher training colleges (EPFs)........................................................................79
Profiles of EPF trainers..................................................................................................................................81
Specialized subjects: trainer supply and demand........................................................................................83
Skills and Knowledge of School Administrators (1.1.5) ....................................................................................87
Attentiveness (1.2) ...........................................................................................................................................88
Short-term hunger (1.2.1) ................................................................................................................................91
Access to food (school feeding) (1.2.1.1) .........................................................................................................93
3
Student attendance (1.3)..................................................................................................................................99
School infrastructure (1.3.3)...........................................................................................................................103
Enrolment (1.3.4)............................................................................................................................................104
Capacity of government institutions (1.4.1)...................................................................................................104
Government support (1.4.3)...........................................................................................................................105
Engagement of Local Organizations and Community output (1.4.4) .............................................................105
Use of Health, Nutrition and Dietary Practices (SO2).....................................................................................106
Knowledge of Health and Hygiene Practice (2.1) ...........................................................................................106
Knowledge of Safe Food Preparation and Storage Practices (2.2) .................................................................109
Knowledge of Nutrition (2.3)..........................................................................................................................109
Access to Clean Water and Sanitation Services (2.4)......................................................................................121
Access to Clean Water................................................................................................................................121
Access to Sanitation Services......................................................................................................................123
Access to preventative health interventions (2.5)..........................................................................................125
Access to requisite food preparation and storage tools and equipment (2.6)...............................................126
Capacity of government institutions (2.7.1)...................................................................................................126
Engagement of Local Organizations and Community output (2.7.4) .............................................................127
Transversal themes.........................................................................................................................................128
Human Resources, capacity, collaboration and ownership........................................................................128
Transport ....................................................................................................................................................131
Administrative and financial systems and procedures...............................................................................131
Monitoring and Evaluation (M&E)..............................................................................................................132
Sustainability and relevance to the local and national school feeding policy and program environment.134
Conclusions.........................................................................................................................................................136
External evaluator’s concluding remarks ...................................................................................................138
Annexes ..............................................................................................................................................................139
Annex 1: Results Framework of the Food for Education Project in Mozambique..........................................139
Annex 2: McGovern-Dole Results Frameworks..............................................................................................140
4
Annex 3: Technical appendix 1 .......................................................................................................................144
5
Acknowledgments
The evaluators gratefully acknowledge the many individuals and institutions who have contributed to this 
midterm evaluation report. Thanks are due to all the Planet Aid, ADPP, Humana People to People and Food for 
Education project leadership and staff who have contributed their time, effort and knowledge, supporting 
fieldwork, providing information and answering numerous queries. In particular, the new project M&E team 
who “hit the ground running” during the midterm evaluation made immense efforts to support the process, 
accompany field visits and obtain information on request. The helpful information and support provided by
Cade Fields-Gardner, who supported the M&E team while key positions were vacant, is also gratefully 
acknowledged. The project leadership team were consistently supportive, helpful and skilled in finding solutions 
to problems and making things happen. Component leaders and staff were welcoming and flexible, providing
valuable information and detailed explanations of their work practices. The evaluators sincerely thank all the 
individuals who welcomed evaluation visits, taking part in interviews and focus groups and sharing their lived 
experience of the project. 
The contribution of all the representatives of the Government of Mozambique, at national, provincial and district 
levels, who supported the external data gathering process, made education and health sector data available and 
provided other invaluable inputs, is acknowledged with gratitude. 
Grateful thanks are also due to the representatives of the World Initiative for Soy in Human Health and 
Cambridge Education for their support in developing evaluation tools, training data gatherers, providing and 
cross-checking large amounts of information, often at short notice, and reviewing the draft report. 
Immense thanks are also due to Valerie Legg, for data verification. 
6
List of abbreviations and acronyms
ADPP Ajuda de Desenvolvimento de Povo para Povo Development Aid from People to People
ASA American Soybean Association
BMI Body Mass Index
CE Cambridge Education
CSB+ Corn Soy Blend-Plus
DNUSE Direção Nacional de Nutrição e Saúde Escolar National Directorate for Nutrition and School 
Health, MINEDH
DPEDH Direção Provincial de Educação e Desenvolvimento Humano Provincial Directorate of 
Education and Human Development 
EGRA Early Grade Reading Assessment
EPC Escola Primária Completa Complete primary school, which includes Grades 1-7
EPF Escolas de Professores do Futuro (teacher training colleges run by ADPP)
FFE Food for Education (project title)
FFK Food for Knowledge (alternative project title)
GoM Government of Mozambique 
GPS Global Positioning System
HGSFG Home Grown School Feeding Garden 
HIV/AIDS Human immunodeficiency virus infection/acquired immunodeficiency syndrome
IMF International Monetary Fund
INDE Instituto Nacional para Desenvolvimento da Educação National Institute for Education
Development
ISET/OWU Instituto Superior de Educação e Tecnologia One World University (ISET/OWU)
M&E Monitoring and Evaluation 
MGD McGovern-Dole International Food for Education and Child Nutrition Program
MINEDH Ministry of Education and Human Development 
NFER National Foundation for Education Research
ODK Open Data Kit (data collection software for resource-constrained environments)
PAI Planet Aid, Inc.
PRONAE Programa Nacional de Alimentação Escolar National School Feeding Program
QA Quality assurance
SDEJT Serviço Distrital de Educação, Juventude e Tecnologia District Office for Education, Youth and 
Technology 
SDSMAS Serviço Distrital de Saúde Mulher e Acção Social District Service for Women Health and Social 
Action 
SEN Special Educational Needs
SFC School Feeding Committee
TB Tuberculosis
ToC Theory of Change
USAID United States Agency for International Development
USDA-FAS United States Department of Agriculture - Foreign Agricultural Service
WFP World Food Program 
WHO World Health Organization
WISHH World Initiative for Soy in Human Health
WPM Words per minute
7
Executive summary
Phase Two of the 2015-2020 Food for Education project (FFE2) follows the 2012-2016 project of the same name 
(FFE1), supported and funded by the Foreign Agricultural Service (FAS) of the United States Department of 
Agriculture (USDA). A comprehensive school-feeding project, complemented by a major literacy component, it 
also includes child health, nutrition education, teacher training, water and sanitation components. The project 
is implemented in Mozambique by Planet Aid (PAI) and Ajuda de Desenvolvimento de Povo para o Povo (ADPP), 
Planet Aid’s local partner, with partners World Initiative for Soy in Human Health program (WISHH) of the 
American Soybean Association (ASA), Cambridge Education (CE) and Mozambique’s Ministry of Education and 
Human Development (MINEDH). 
This midterm evaluation offers a “photograph” of the project in March-April 2019 in order to measure the 
project’s “state of health”. Key objectives of this evaluation are to permit comparison against the baseline 
evaluation to assess and document changes since the beginning of phase two; report how recommendations 
and lessons from phase one are being implemented; and make recommendations in relation to planning and 
implementation for the remaining 16 months of the project. While education access has generally improved, 
Mozambique faces high dropout rates, especially for girls, in the context of the combined challenges of political, 
economic and environmental crises (two major cyclones occurred during the preparation of this report). In 
addition to providing accountability to donors and other stakeholders and allowing learning and improvement, 
this report presents further information requested by project staff and managers.
There are three major components of the project, implemented by three different agencies: i) 88,100 students 
in 271 target schools in 4 districts of Maputo Province, receive a daily meal, consisting of nutritious porridge 
made from corn soya blend-plus (CSB+). School gardens and eight large-scale Home Grown School Feeding 
Gardens (HGSFGs) seek to encourage diversification and sustainability of school feeding. These “core” areas of 
the project are implemented by the FFE project team; ii) the project literacy team, with technical assistance from
CE, is supporting literacy teaching in schools with materials developed to reinforce reading and writing skills for 
1
st, 2nd and 3rd grade children, teacher training and coaching; iii) a team managed by the WISHH is implementing 
a comprehensive nutrition education campaign in the beneficiary districts of Maputo province. 
The midterm evaluation of FFE2 was conducted by a consortium composed of: a) Dr Simone Doctors (evaluation 
lead, qualitative data collection and analysis, main report author), b) Forcier Consulting (quantitative data 
collection) and c) the National Foundation for Educational Research (NFER; quantitative data analysis). The three 
members worked collaboratively under the coordination of Simone Doctors. To allow comparisons over time 
and build on lessons learned during previous evaluations, the FFE2 baseline survey conducted in 2017 was based 
on a newly developed evaluation plan; this commenced a new cohort study but continued to collect data on 
many of the same indicators as used in FFE1. Where possible, the FFE2 plan follows up on many of the research 
8
questions used during FFE1, assessing these at three time points: baseline, midterm and end point. The pupil￾level data at FFE2 baseline was collected from a new cohort of pupils, who can be expected to remain in primary 
school for the duration of the project lifespan, using a three-ways sampling frame developed by NFER; teacher￾level data was collected from a new sample of teachers. The present midterm evaluation follows the progress 
of these cohorts of intervention and counterfactual pupils, and of the teachers and schools studied at baseline. 
The midterm evaluation used a mixed method approach to gather and analyze a combination of quantitative 
and qualitative information consisting of: i) existing information provided by ADPP / the Provincial Directorate 
of Education and Human Development (DPEDH), and reviewed, collated and analyzed by Simone Doctors, ii) 
quantitative data collected by Forcier Mozambique and analyzed by NFER, and iii) qualitative data collected and 
analyzed by Simone Doctors. This information included:
• Project documents, surveys of (PAI/ADPP-run teacher training colleges (EPFs), pre- and post- Portuguese
language tests of trainee teachers attending EPFs, DPEDH enrolment and dropout data, de-worming data;
• Literacy testing of 3,956 learners (from project schools and control/counterfactual schools), interviews
and anthropometric measurements (weight and height) of the same 3,956 learners, a school survey of 170
project schools, a teacher survey of 190 teachers from project schools, classroom observation of teaching
in 246 project and comparison schools, and statistical modelling (by NFER) to account for measured
background differences between target learners and the comparison group;
• Focus groups with key stakeholder groups, semi-structured interviews with key stakeholders, visits and
observation of a sample of schools, visits of the project headquarters and project warehouse, visits and
observation of three EPFs, observation of food preparation and distribution processes, visits and
observation of school gardens and HGSFGs, observation of teaching practice by EPF students, an internal
self-review by the project implementation team, a stakeholder review workshop.
The midterm report is structured according to the logic of the project Theory of Change (ToC), with particular 
attention given to the two new components which began with the second phase: i) the program to improve 
literacy instruction in primary schools, particularly literacy in local languages, and ii) the HGSFGs, which strive to 
make school feeding more sustainable.
The project aims to improve literacy of school-age children in Mozambique (Strategic Objective 1) and to 
increase the use of health and dietary practices (Strategic Objective 2). In line with the project’s ToC, 
interventions are designed to improve the quality of literacy instruction through training teachers in project 
schools in literacy, a significant addition in phase two of the project, giving learners improved access to school 
supplies and materials and other interventions to improve attendance. Interventions designed to improve health 
and dietary practice include nutrition education, school gardens and deworming. Achievement of both SO1 and 
9
SO2 depend upon increased engagement of local organizations and community groups and increased capacity 
of government institutions. Improving school infrastructure and increasing enrolment and attendance, as a 
result of the school feeding program and other interventions to increase enrolment, is expected to further 
contribute to improving child literacy. 
Assessing the literacy levels of students who have benefited from the literacy program supported by CE is key 
to evaluating the success of this intervention and of the wider project. At midterm, the target for students able 
to read and understand the meaning of grade level text by the end of two grades of primary schooling was 50%
achieved by project students (67% by students benefiting from the literacy program in Portuguese). 92,237 
individuals are benefiting directly from USDA-funded interventions, resulting in a midterm target 108% achieved. 
The number of individuals benefiting indirectly from USDA-funded interventions was 336,000, which achieves 
the project’s target. 
The cohort of pupils tested at baseline, when they were in grade 1, performed an Early Grade Reading 
Assessment (EGRA) comparable to that used at baseline. The results of each EGRA subtask were analyzed across 
the following comparisons: a) FFE + literacy (all) versus comparison (all); b) FFE + literacy (Portuguese) versus 
comparison (Portuguese); c) FFE + literacy (local languages) versus comparison (local languages); d) FFE + literacy 
(Portuguese) versus FFE only (Portuguese); e) FFE + literacy (Portuguese) versus FFE + literacy (local languages).
EGRA results demonstrate a low level of reading overall, with only 8% of the students benefiting from the literacy 
intervention meeting the national benchmark. However, taken overall, the results suggest that the literacy 
intervention is beginning to have an impact. Although the FFE + literacy group did not perform significantly 
better than the comparison group overall, in 8 of the 18 cases reported, of the students tested in local languages, 
the FFE + literacy group performed significantly better than the comparison group; there were no cases where 
the comparison (local languages) group performed significantly better than the FFE + literacy (local languages) 
group. 
In 12 of the 18 cases reported, the FFE + literacy group performed significantly better than the FFE only group 
amongst students tested in Portuguese, providing promising early evidence that the literacy intervention is 
having a positive impact on the literacy acquisition of students within FFE project schools. 
This finding provides promising early evidence of the impact of the literacy intervention, although more time 
may be needed for the effects of the intervention to translate into more systematic results. The final evaluation 
of FFE2 in 2020 will allow the initial achievements reported here to be tested after an additional year of the 
literacy intervention and cast more light on how the initial progress made translates into later literacy. For the 
students taught in local languages, the final evaluation will capture how literacy instruction begun in local 
languages then transitions into literacy in Portuguese, the national language of instruction. 
10
The project ToC encapsulates the assumption that the quality of literacy instruction will be improved by a) better 
access to school supplies and materials and b) increased skills and knowledge of teachers. At midterm, 71,932 
textbooks and other teaching and learning materials had been provided as a result of USDA assistance, resulting 
in 182% achievement of the final target. 264 schools had received school supplies and materials as a result of 
USDA assistance, resulting in 103% achievement of the final target. The project has contributed to teachers and 
pupils having improved access to teaching and learning materials.
Awards have been provided to both pupil and teachers since the beginning of the project to encourage both to 
make efforts in teaching and learning. The final target for awards to teachers has been surpassed, and the 
midterm target for awards to pupils has been 90% achieved. 
A range of literacy books and materials in local languages Xichangana and Xirhonga and in Portuguese has been 
produced by the project literacy team, and targets for distributing these have been surpassed. The midterm 
evaluation revealed widespread praise for these innovative materials, for the quality of their content and their 
culturally appropriate messages. 
Two major dimensions of the project interventions target improving the skills and knowledge of teachers: 586 
teachers in project schools have received training in teaching reading and writing using phonics - a significant 
innovation for Mozambique, where phonics have not been commonly used. Lessons were observed in 240 
project schools and comparison schools in order to attempt to evaluate whether teachers in project schools 
were using “new and quality teaching techniques or tools” as a result of the literacy training. FFE + literacy 
teachers were more likely to exhibit these characteristics than FFE only teachers or comparison group teachers; 
FFE only and comparison group teachers were almost equally likely to exhibit these characteristics; a higher 
percentage of teachers using local languages demonstrated the characteristics than did teachers teaching in 
Portuguese. 
Prior evaluations of the FFE project have underlined the important numerical and quality contribution the eleven 
ADPP-run teacher training colleges (EPFs) are making to Mozambique’s teacher pool. At the time of the midterm 
evaluation, 6,059 trained teachers had graduated during the current phase of the project, well exceeding the 
target. The EPF teacher training is providing significant added value in Portuguese language competence 
through remedial support to student teachers who arrive at the EPF with poor levels of Portuguese. The EPFs 
provide a culture in which a “different sort of teacher” may be formed, with a focus on the more holistic aspects 
of education, which provides trainees with skills, attributes and attitudes appropriate to the realities of teaching 
in low-resource rural schools. The target for the number of school administrators and officials in target schools 
who demonstrate use of new techniques or tools as a result of USDA assistance, has been 110% achieved; the 
target for the number of school administrators and officials trained or certified as a result of USDA assistance 
has been 161% achieved.
11
Short-term hunger leads to lack of attentiveness in the classroom, leading to problems in learning. At midterm, 
11% of teachers report increased attentiveness of students in the classroom and 6% of students report being 
more attentive in the classroom. 42% of pupils in intervention schools (FFE + literacy and FFE only combined) 
said they sometimes felt hungry in the classroom, compared with 58% of students in comparison schools. These 
levels of hunger are lower than the baseline levels for both groups. 23% of pupils in intervention schools (FFE + 
literacy and FFE only combined) said they felt hungry during the interview, compared with 33% of students in 
comparison schools.
The distribution of school meals is on target. At midterm, 28,162,260 daily school meals (breakfast, snack, lunch) 
were provided to school-age children as a result of USDA assistance, which is 104% of the midterm target. 
84,026 school-age children were receiving daily school meals, 114% of the midterm target. When students 
missed school meals, the principle reason for non-feeding was pupil absence. Of reasons attributable to the 
project itself, the main reasons for schools not feeding are lack of CSB+ and lack of cooks. 
During school visits for the midterm evaluation, as during other evaluations of the project, teachers, school 
directors, parents, government officials and other stakeholders consistently claimed that school feeding and the 
other project interventions had led to increases in student attendance. 
One project activity intended to promote increased school attendance is support to extra-curricular learning 
clubs. Of the schools surveyed at midterm, 89% reported having one or more learning clubs. 113% of the final 
target has been achieved for the number of after-school learning clubs active in project, and 109% of the final 
target for the number of school children participating. In accordance with a recommendation made in the 
baseline report, this evaluation found that in project schools, pedagogical directors and teachers responsible for 
clubs have been trained in the facilitation of extracurricular clubs, play-based and other activities to promote 
learning and in the use of provided materials. 
Targets for maintaining the school infrastructure provided by the project have been surpassed, with 1,002 
maintenance interventions performed in kitchens, storerooms, and firewood-saving stoves at midterm, or 127% 
of target.
The project ToC considers enrolment as a prerequisite to attendance and therefore to improved learning. 
Targets for enrolment of students have been surpassed, and informants overwhelmingly claimed that enrolment 
had increased, and dropouts decreased as a result of the projects’ activities, particularly school feeding. In was 
not possible to verify this claim using information provided by the DPEDH, due to inconsistencies in the data.
The project’s second Strategic Objective (SO2) concerns the use of health, nutrition and dietary practices. 
Targets for training in good hygiene practice, including training volunteer cooks in health and hygiene, have been 
greatly surpassed. In regard to the percentage of students that demonstrate acceptable knowledge of health 
12
and hygiene practices, the final target has been 170% achieved; the midterm target for food preparers at target 
schools trained in hand washing, safe food preparation and storage practices has been 118% achieved. FFE2 
targets for training individuals in child health and nutrition have been greatly surpassed. 
School gardens have been encouraged and supported since the inception of the project, and the final target for 
the number of school children benefiting from school gardens has been achieved. 8 new HGSFGs have been 
established; these are an innovation of FFE2 to address the question of sustainability of school feeding after the 
end of the project activities. Access to safe water continues to be a condition for school feeding, local production 
of food, health and wellbeing; despite the water team’s best efforts, this also continues to be a challenge for 
some project schools. The final target for the number of schools using an improved water source has been 
achieved; however, in many schools the absence of safe water some or all of the time is a major problem.
The evaluation provides further analysis of transversal themes which cut across the different components and 
activities and concern the overall operations, strategy and vision of the project. These are Human Resources, 
capacity, collaboration and ownership; Transport; administrative and financial systems and procedures; 
monitoring and evaluation; and sustainability and relevance to the local and national school feeding policy and 
program environment. 
Most activities planned have been implemented as intended, often in the face of considerable obstacles, and 
most midterm targets have been met or surpassed, with a number of final targets met at midterm; in the small 
percentage of cases where targets have not been met, these are likely to be met by the end point. The exception 
is the target concerning the percentage of school children meeting the literacy benchmark; it may be that the 
target set was just too ambitious to be achieved within such a short time, given the low baseline levels. 
Unlike many projects operating in the extremely challenging context of the Mozambican education system, FFE2 
is largely managing to do what it set out to do within the planned timeframe. It is recommended that project 
staff and leaders (and evaluators) remain aware of the reasons why given activities are conducted, what positive 
change they are intended to bring and how they are experienced by all those involved in their implementation, 
including but not limited to the beneficiaries. Project interventions are beginning to make a difference: teachers 
are beginning to use improved methods in their teaching; students are beginning to read better; community 
volunteers in the HGSFGs are beginning to apply new methods to produce large amounts of food; thanks to this 
produce, schools are beginning to diversify the menus consumed by students; school water committees are 
beginning to take responsibility for maintenance of their water system. While project interventions are 
beginning to make a difference, more time is needed in order for this initial change to be consolidated and 
integrated into durable behavioral changes. 
The project applies many of the principles of good international development practice and is providing services 
and skills which are highly relevant to both the beneficiary population and the country as a whole. While it is 
13
beginning to produce impact in the form of positive change, questions remain about the sustainability of these 
achievements in the current Mozambican context. Whether there is an extension to the current project, or the 
remaining period isspent implementing an exit plan to prioritize activities which can be continued after the close 
of the project, maximizing future sustainability is now imperative. 
Every effort should be made to prolong the FFE2 project to capitalize on the human and material investment 
made so far.
14
Introduction
The 2015-2020 Food for Education project (phase 2 of the 2012-2016 project of the same name, or FFE2)
12
is 
supported and funded by the Foreign Agricultural Service (FAS) of the United States Department of Agriculture 
(USDA). FFE2 is a comprehensive school-feeding project with child health, nutrition education, teacher training, 
water and sanitation components, complemented by a major literacy component. The project is implemented 
in Mozambique by Planet Aid (PAI) and Ajuda de Desenvolvimento de Povo para o Povo (ADPP), Planet Aid’s local 
partner in Mozambique, in partnership with the World Initiative for Soy in Human Health program (WISHH) of 
the American Soybean Association (ASA), Cambridge Education (CE) and the Ministry of Education and Human 
Development (MINEDH).The project is intended to serve as one of the models for the PRONAE, the national 
school feeding program being developed and implemented by the Government of Mozambique (GOM) with 
collaboration and assistance from the World Food Program (WFP) and other partners.
The midterm evaluation seeks to measure the “state of health” of the project at midterm, offering a 
“photograph” of phase 2 of the project in March-April 2019. This allows:
• comparison against the baseline evaluation to assess and document changes since the beginning of 
phase 2;
• monitoring of how recommendations and lessons from phase one are being implemented;
• recommendations to be made in relation to planning and implementation for the remaining 16 months of 
the project.
Like the baseline report for phase 2, this midterm evaluation report:
• accounts to the project donors and other stakeholders for how resources have been used;
• collects and analyzes information identified as important by project staff and managers to allow them to
make informed decisions about their operations;
• identifies lessons and makes recommendations for the remainder of the project and beyond, including in 
relation to sustainability. 
Despite progress in improving access to education, education quality remains a cause for concern in 
Mozambique. Dropout rates are high, with over half the girls who enroll in primary school dropping out by the 
fifth grade, only 11 percent continuing to secondary level, and 1 percent to college (USAID, 2019). Of those who 
2 often referred to internally as Food for Knowledge (FFK2) or Comida para o Saber, in Portuguese. 
15
finish primary school, “nearly two-thirds leave the system without basic reading, writing, and math skills” 
(USAID, 2019)3
. 
The overall context in Mozambique has deteriorated since the baseline report. Residents speak of a three-fold 
political, economic and environmental crisis. The ongoing saga of the illegal debt has led to the suspension of 
much international development aid. Armed conflict in certain areas of the country is leading to violence and 
population displacement. Extreme weather events disrupt daily life for many citizens: during the preparation of 
this report, between March and April 2019, cyclones Idai and Kenneth both led to widespread flooding, loss of 
life and infrastructure and population displacement. 
Background
The FFE 2 project aims to promote the health, wellbeing and education of school aged children, though holistic 
interventions combining school feeding with health, water, sanitation, literacy and nutrition components. Based 
on the Theory of Change captured in the McGovern Dole Results Framework (see Annexes1 and 2), the project’s 
Strategic Objectives are to improve literacy of school-age children in Mozambique (SO1) and to increase the use 
of health and dietary practices (SO2). In order to achieve these Strategic Objectives, interventions are structured 
according to a hierarchy of assumptions about the conditions leading to change. 
Within SO1:
• the quality of literacy instruction is improved by providing intensive, targeted literacy training to 
teachers in project schools;
• the skills and knowledge of primary school teachers are increased through on the spot coaching of in￾service teachers within primary schools4
;
• providing teachers and learners with improved access to school supplies and materials contributes to 
improved literacy instruction;
• school feeding reduces short-term hunger in the classroom, leading to improved student attentiveness 
and therefore learning;
• student attendance is improved through a combination of improved school infrastructure (including 
kitchens, water and sanitation);
3https://www.usaid.gov/mozambique/education
4This dimension was a significant addition to Phase 2 of the project. Evaluations of Phase 1 revealed that,
although attendance and retention rates and reported pupil concentration in the classroom had improved since 
the onset of school feeding, in the absence of improved teaching, educational outcomes were unchanged.
16
• increased enrolment results from students having increased access to food increased as a result of 
school feeding.
Within SO2:
Use of health and dietary practices increases through:
• improved knowledge of health and hygiene practices, such as hand washing;
• increased knowledge of safe food preparation and storage practices, for example through training 
volunteer cooks;
• increased knowledge of nutrition, through nutrition education and access to school gardens;
• increased access to clean water and sanitation (latrines);
• increased access to preventative health interventions, such as de-worming and TB screening;
Critical to the achievement of both SO1 and SO2are increased engagement of local organizations and community 
groups and increased capacity of government institutions.
In order to bring about the desired changes, the project consists of three major components, implemented by 
different agencies:
1.School feeding, water supply development, school gardens, and related activities
The project strives to provide 88,100 students in 271
5
target schools in 4 districts of Maputo Province, 
Mozambique (Moamba, Magude, Manhiça and Matutuine) with a daily meal, consisting of nutritious porridge 
made from corn and soya. School gardens and two large-scale Home Grown School Feeding Gardens per district 
aim to encourage diversification of students’ diet contribute to sustainability of school feeding after the end of 
the project. Schools are provided with a safe and adequate school water supply, latrines and hand-washing 
facilities. Students receive regular de-worming. Actions to provide an improved learning environment, 
enrolment and attendance for all pupils include after-school clubs and kits of educational and recreational 
materials. These “core” areas of the project are implemented by the FFE project team.
2. Early grade reading and writing intervention
FFE2 includes an innovative intervention to support literacy teaching, in both Portuguese and local languages 
(Xirhonga or Xichangana)6
. Schools benefiting from the literacy intervention receive specially developed
5
Significantly increased from the original proposal and contract with USDA to provide meals to 74,000 students 
in 264 schools.
6 Children in the beneficiary districts speak one of several local Mozambican languages when they enter school, 
but they have traditionally been taught to read in Portuguese, the official national language. Approaching early 
literacy in a language which children do not speak or understand (and which many teachers master imperfectly) 
creates a major obstacle in their acquisition of reading and writing skills as well as in their wider education. 
17
materials to reinforce reading and writing skills for 1st 2
ndand 3rd grade children. Teachers receive training in how 
to use the new methods and reading coaches provide ongoing support to the trained teachers, through 
classroom observation and lesson coaching. The literacy intervention is implemented by a dedicated team 
(referred to hereafter as “the project literacy team”) supported by CE. 
3. Nutrition education program
Since the onset of FFE1, the project has implemented a comprehensive nutrition campaign in the 4 beneficiary 
districts of Maputo province and the 11 teacher training colleges run by ADPP (the Escolas de Professores do 
Futuro: EPFs). Teachers currently teaching in primary schools are trained as Nutrition Educators, known as 
“coaches”, and train other primary school teachers in the schools in which they work. Nutrition education is an 
integral part of the teacher training provided by the EPFs. Nutrition and health educational materials and 
information documentation have been produced. The Nutrition Education program is implemented by a team 
based in the project office but managed by the World Initiative for Soy in Human Health (WISHH).
Overview of the FFE2 Midterm Evaluation
Alignment with previous and future evaluations
Regular evaluations of the FFE project have been conducted, in accordance with the project Evaluation Plan7
:
• a baseline study of FFE1 in May 2013 by project staff with support from an external evaluator; 
• a midterm evaluation of FFE1 in 2014 by an external evaluation team;
• a final evaluation of FFE1 in late 2016 (report dated July 2017) by an external evaluation team;
• a baseline study of FFE2 in 2017 (report dated July 2018) by project staff with support from an external 
evaluator (to avoid redundancy and in order to make the best use of resources, the baseline of FFE2 built 
on the final evaluation of FFE1).
The current evaluation is of the FFE2 project at midpoint, without direct reference to FFE1. 
The FFE2 evaluation plan builds on that of FFE1. For the sake of coherence, to allow comparisons over time and
build on lessons learned during previous evaluations, the FFE2 baseline survey was based on a newly developed 
evaluation plan, commencing a new cohort study but continuing to collect data on many of the same indicators 
as used in FFE1. Where possible, the FFE2 plan follows up on many of the research questions used during FFE1, 
assessing these at three time points: baseline, midterm and end point. Alignment between evaluations of FFE1 
Bilingual education seeks to introducing reading and writing in mother tongue, whilst simultaneously learning 
the language of instruction, into which teaching and learning transitions in grade 3. 
7The reports of these evaluations are available on request.
18
and FFE2 has contributed to the amassing of a considerable body of evaluation evidence which is relevant to the 
current project.
The pupil-level data at FFE2 baseline was collected from a new cohort of pupils, who can be expected to remain 
in primary school for the duration of the project lifespan, using a three-wayssampling frame developed by NFER; 
teacher-level data was collected from a new sample of teachers. 
The present midterm evaluation of FFE2 measures progress against the baseline in achieving objectives; it
formulates recommendations to improve effectiveness and efficiency for the remainder of the project and to 
ensure sustainability thereafter. The indicators and methods used are based on those of the FFE2 baseline, to
follow the progress of the cohorts of intervention and counterfactual pupils and of the teachers and schools 
studied at baseline.
A final external evaluation of the second phase will be conducted at the end point of the project in 2020, in order 
to assess the project’s relevance, effectiveness, efficiency, impact and sustainability. Using the same indicators, 
measures and methods, the final evaluation of FFE2 will assess whether the project objectives have been 
achieved, what has changed in the lives of the project beneficiaries and whether recommendations made 
following the midterm evaluation have been integrated and implemented. 
Methodology
The midterm evaluation used a mixed method approach to gather and analyze a combination of quantitative 
and qualitative information. The methods used built on those used at baseline (and, to some extent, during the 
evaluations of FFE1). These will also form the basis of the approach to be used for the final evaluation of FFE2, 
ensure comparability and consistency of the information gathered at the three time points and evaluate change 
over the duration of the project. 
Information gathered for the midterm evaluation
Three types of information were gathered and analyzed:
Existing information provided by ADPP / DPEDH, and reviewed, collated and analyzed by Simone Doctors 
• Project documents: project records, in the form of individual record level data in unanalyzed /non￾tabulated form (for example, participant lists from training courses); project documentation; training 
materials produced; reports (for example the semi-annual Logmon reports submitted to USDA) and other 
monitoring information (for example the monthly indicator tracking files and commodity tracking files), 
were provided by ADPP and WISHH to the external evaluator for analysis upon request. The external 
evaluator reviewed these sources extensively, extracting elements to analyze and engaging in an iterative 
conversation with the project leadership and M&E teams to discuss these and reconcile records as 
19
necessary. During visits to the project headquarters, she performed spot checks on records and evidence, 
asking M&E staff to produce different types of information and to demonstrate the systems for gathering 
and archiving these. 
• Surveys of EPFs (PAI/ADPP-run teacher training colleges) regarding infrastructure and equipment, profiles
of teaching staff and staffing levels relative to teaching needs. The instruments used at baseline were
updated as necessary by the external evaluator and sent to all EPF directors to complete, through the
ADPP EPF focal point.
• Results of pre- and post- Portuguese language tests of trainee teachers attending EPFs showing pre-test
and post-test scores for all trainee teachers in 6 of the EPFs, allowing their performance before training
began to be compared with that after one year’s training over the course of 2015; this information was
gathered by 6 EPFs at the beginning and end of the 2018 academic year and made available to the
external evaluator by the ADPP EPF focal point.
• Enrolment data provided by the DPEDH Maputo for all districts of Maputo province for 2012, 2013,
2014,2015, 2016, 2017 and 2018;
• Dropout data provided by the DPEDH Maputo for all districts of Maputo province for 2012, 2013,
2014,2015, 2016, 2017 and 2018;
• Data on de-worming provided by the district health services in the intervention districts, either directly to
the external evaluator or through the intermediary of the project district coordinators. This was collated
and analyzed by the external evaluator
Quantitative data collected by Forcier Mozambique and analyzed by NFER
• Literacy testing of 3,956 learners (from project schools and control/counterfactual school) using the Early
Grade Reading Assessment (EGRA) used at baseline with minor modifications;
• Interviews with the same 3,956 learners, using the instrument used at baseline with minor modifications;
• Anthropometric measurements (weight and height) of the same 3,956 learners to establish BMI z scores;
• School survey of 174 project schools, using a reduced version of the instrument used at baseline;
• Teacher survey of 190 teachers from project schools, using a slightly modified version of the instrument
used at baseline;
• Classroom observation of teaching in 246 project and comparison schools, using a standardized
observation sheet;
• Statistical modelling (NFER) to account for measured background differences between target learners and
the comparison group, allowing potential changes associated with the program to be measured.
Qualitative data collected and analyzed by the external evaluators 8
• Focus groups with key stakeholder groups: learners, teachers, parents, volunteer food preparers, school￾feeding committee members, student teachers, teacher trainers, project “professionals” (the project field
officers);
8 Ten of the focus groups (two focus groups each with female pupils, male pupils, parents, volunteer cooks, 
school feeding committee members) were conducted by Forcier since they required the use of local languages; 
20
• Semi-structured interviews with key stakeholders: school heads, community representatives, government
education officials (national, provincial and districts level), district health officers, EPF heads, project
component managers or team representatives, representatives of USDA, representatives of PAI,
representatives of WISHH, representatives of CE, representatives of other organizations providing school
feeding, other partners);
• Visits and observation of a sample of schools in all four project districts, using a standardized observation
framework and photographic recording of infrastructure;
• Visit of the project headquarters in Manhiça, including observation and discussion of working practices and
M&E systems;
• Visit of the main project warehouse in Manhiça;
• Visits and observation of three EPFs (Tete, Gaza and Maputo)9
;
• Observation of food preparation and distribution processes, using a standardized observation framework
and photographic recording;
• Visits and observation of school gardens and Home Grown School Feeding Gardens in all four project
districts;
• Observation of teaching practice by EPF students and of the support and feedback provided by EPF tutors
and school mentors;
• Results of an internal self-review by the project implementation team, facilitated by the external evaluator
on 22 February 2019 and attended by approximately 40 project leaders and staff, including the entire
leadership team;
• Results of a stakeholder review workshop to validate initial findings and suggest corrections or changes in
strategy for phase 2 of the project. This was held on 10 April 2019 and attended by approximately 50 project
leaders and staff, including the entire leadership team and ADPP’s Executive Director. The initial findings as
presented were recognized, owned and well received and generated invaluable comments and responses
from project leaders and staff.
Sampling for pupil-level data collection
The evaluation plan for FFE2 is conceived as a quasi-experimental cohort study at three time points. The 
collection of quantitative data at midterm was based on the stratified sampling frame of schools drawn up by 
experts in planning statistical analysis for education studies from NFER and already used at baseline. Since 
almost all schools in the four project districts (Manhiça, Magude, Moamba and Matutuine) are benefitting from 
the project interventions, the control group schools were selected from the four non-project districts 
(Namaacha, Matola, Marracuene and Boane). 
An explanation of the selection of the sample is summarized below; more details can be found in the FFE2 
baseline report.
all other qualitative information was collected by Simone Doctors using Portuguese, except in a small number 
of cases where respondents preferred to speak English. All translations of quotes from interviews are by Simone 
Doctors. 
9 The sample of EPFs visited aimed to achieve a geographical spread and balance between those EPFs considered 
by WISHH to be implementing the Nutrition education program well. The choice was constrained by logistical 
considerations and time constraints and by the disruption caused to two EPFs by cyclones in March and April 
2019.
21
• Schools in the project districts not receiving the intervention were excluded
• Schools in the comparison districts known to be receiving a different feeding program were excluded
• Xichangana and Xirhonga were combined into one “local language” group.
• The sampling frame contains five categories of schools:
o FFE+literacy (local language)
o FFE+literacy (Portuguese)
o FFE+literacy (local language and Portuguese)
o FFE only
o Comparison
• It was not possible to achieve perfect matching between the categories, since the sample was drawn based 
on the schools actually benefiting from the project, rather than based on their characteristics (as would be 
the case in a randomized control trial). The project districts were selected to benefit from the project on 
the basis of their relative poverty and food insecurity; the comparison districts are likely to have lower levels 
of socioeconomic deprivation than the intervention districts. 
Whereas at baseline, pupils in grade 1,2 and 3 were tested, the midterm only followed up on the cohort of 16 
pupils per school, who had been in grade 1 at baseline and, for the most part, were in grade 3 at midterm (where 
schools benefit from literacy interventions in both Portuguese and the local language, two cohorts of 16 pupils 
were tested: one from the class learning in Portuguese, the other from the class learning in the local language). 
At baseline, 4,150 grade 1 students were tested: 1,887 in the FFE + literacy group, 1,104 in the FFE only group 
and 1,159 in the comparison group10
. 115 students tested were removed from the sample at midterm, following 
verification of the intervention groups, which revealed that these students no longer qualified to be part of the 
sample, for example because they had been part of the FFE only group but their school had begun receiving the 
literacy intervention between baseline and midterm. At midterm, the FFE only (bilingual) group was removed 
from the study, since there are no longer project schools teaching the bilingual curriculum which are not 
receiving the literacy intervention. At midterm, therefore, there was one FEE only group, tested in Portuguese.
Of the remaining students, 2,819 were successfully recontacted at midterm: 1,180 FFE + literacy; 668 FFE only 
and 971 comparison. Data gatherers were unable to recontact 902 at midterm, for a variety of reasons, including 
leaving school, changing schools or residence, absence from school on the day of data gathering, and death: 
these were deemed “lost” to the project. This high level of attrition had been anticipated and the sample drawn 
accordingly. Where possible, the lost students were replaced with substitutes (628 FFE + literacy, 284 FFE only 
and 110 comparison), using a rigorous sampling protocol based on the sampling method used at baseline. This 
information is summarized in table 1 below. 
10 Throughout the report, the terms “FFE + literacy”, “FFE only” and “Comparison” are used, without quotation 
marks or italics, to differentiate teachers, students and schools belonging to the three groups. 
22
Students who participated at baseline and at midterm
Row Labels Lost Recontacted Replacement Grand Total
Baseline 902 2819 3721
Comparison (Bilingual) 89 89
Comparison (Portuguese) 165 882 1047
FFE + lit (Bilingual) 8 530 538
FFE + lit (Portuguese) 390 650 1040
FFE only 339 668 1007
Midline 2819 1022 3841
Comparison (Bilingual) 89 20 109
Comparison (Portuguese) 882 90 972
FFE + lit (Bilingual) 530 231 761
FFE + lit (Portuguese) 650 397 1047
FFE only 668 284 952
Table 1: Students who participated in the baseline and midterm evaluations
Schools within the sample were either classified as “bilingual”, meaning they teach in both a local language 
(Xichangana or Xirhonga) and Portuguese, or “Portuguese”, meaning they teach only in Portuguese. Students 
from the bilingual schools are tested in local languages and students from Portuguese schools are tested in 
Portuguese (cohorts of 16 per school, in each case). However, in some bilingual schools, both local language and 
Portuguese students are tested (so 2 cohorts of 16 students are tested, one in Portuguese, the other in local 
languages). For the purposes of the evaluation plan, bilingual students from Moamba, Magude and Manhiça are 
tested in Xichangana and those from Matutuine in Xirhonga11
. For details of the number of schools within each
group within the sample, see table 2 below. The groups are not balanced, due to the relative scarcity of schools 
delivering the bilingual program, compared to those teaching in Portuguese (see figure 1 below). 
11 This decision is based on the languages taught by the project in these districts. In reality, the distribution of 
the two languages does not follow the district boundaries but there is some overlap. Furthermore, Xichangana 
and Xirhonga are mutually intelligible and speakers of Xichangana frequently use words and expressions from 
Xirhonga and vice versa. 
23
Figure 1: distribution of bilingual and Portuguese schools within the sample 
Table 3 (below) show the distribution of students who participated in the midterm evaluation by language of 
instruction. Students were tested in their language of instruction. However, where students in the Portuguese 
group did not speak Portuguese sufficiently well to understand the instructions of the EGRA subtasks, for 
example, these were given in the local language, with the items themselves (names of object, words, letter 
sounds, etc.) presented in Portuguese. 
Midterm: breakdown of students by language of instruction
Row Labels Biligual Portuguese Grand Total
Comparison (Bilingual) 5 5
Comparison (Portuguese) 67 67
FFE + lit (Bilingual) 42 42
FFE + lit (Portuguese) 51 51
FFE only 71 71
Grand Total 48 189 236
 Table 3: numbers of students who participated in the midterm evaluation by language of instruction 
Groups Number of schools
Comparison (Bilingual) 5
Comparison (Portuguese) 67
FFE + lit (Bilingual) 42
FFE + lit (Portuguese) 51
FFE only 71
Grand Total 236
5
42
0
47
67
51
71
189
0
20
40
60
80
100
120
140
160
180
200
Comparison FFE + literacy FFE only Total schools
Number of Schools
Midterm: Bilingual and Portuguese Schools
Bilingual Portuguese
Table 2: Numbers of schools in the evaluation sample at midterm
24
This composition of the three groups is balanced in terms of gender. 50% of the bilingual students are female 
and 50% male. 49% of the Portuguese students are female and 51% male. Because students were selected at 
random from within the classes identified by the sampling plan, the gender distribution within the sample should 
reflect that of the school population. For details of the breakdown of each group by gender, see table 4 below.
Midterm: breakdown of students by gender
Intervention groups boy % girl % Grand Total %
Comparison (Bilingual) 57 3% 52 3% 109 3%
Comparison (Portuguese) 469 24% 503 27% 972 25%
FFE + lit (Bilingual) 379 19% 382 20% 761 20%
FFE + lit (Portuguese) 531 27% 516 28% 1047 27%
FFE only 531 27% 421 22% 952 25%
Grand Total 1967 1874 3841
Table 4: breakdown of student who participated in the midterm evaluation by gender
Management and quality assurance of the Midterm Evaluation
The midterm evaluation of FFE2 was conducted by a consortium of three members who responded to Planet 
Aid’s Request for Proposal of 27 November 2018 in December and were awarded the contract. The consortium 
is composed of: a) Dr Simone Doctors (evaluation lead, qualitative data collection and analysis, main report 
author), b) Forcier Consulting (quantitative data collection) and c) the National Foundation for Educational 
Research (quantitative data analysis). The three members worked collaboratively under the coordination of 
Simone Doctors. The midterm evaluation was designed and planned by Simone Doctors with extensive input 
from and collaboration with the other members. Throughout the planning and data collection and analysis, 
regular three-way Skype meetings were held to ensure these processes ran as smoothly as possible and to 
optimize communication between the members. 
The evaluation instruments were developed by Simone Doctors, based on those used at baseline, with input 
from members of the FFE team where necessary, so that the information collected would meet the needs and 
vision of the project team12
. These were scripted and coded in the Open Data Kit (ODK) application by the Forcier 
team, so data collection could take place using smartphones. Simone Doctors worked closely with the Forcier 
team during this process, testing the tools on the phones and requesting amendments to the script so that they 
worked as intended. 
12 supervisors and 48 enumerators were recruited by Forcier and trained in Maputo by Simone Doctors and the 
Forcier team. A team of 20 potential supervisors received 5 days intensive training, following which 12 were 
12 All evaluation instruments used are available on request.
25
selected; the following week, these 12 supervisors participated in and contributed to the enumerator training, 
which also lasted 5 days, including one day piloting of the instruments. Again, the 48 enumerators were selected 
from a group of 60 based on their performance during training and piloting. Groups consisting of 4 enumerators 
and one supervisor were formed; care was taken to ensure groups were balanced in terms of gender, experience 
and languages competence (Portuguese, Xichangana and Xirhonga). 
The evaluation instruments as coded in ODK were piloted in two schools using ODK-encoded smart phones13
. 
Simone Doctors and three members of the Forcier team observed the piloting and made recommendations; 
recommendations from the enumerators and supervisors themselves were also considered. Forcier uploaded 
and collated the results from the smartphones and Simone Doctors and the Forcier team analyzed these.
Following the pilot, some improvements were made to the instruments and an additional day’s enumerator 
training was held to address issues identified by the enumerators themselves and their supervisors.
NFER were responsible for the sampling framework (see Sampling for Pupil-level data collection above). Forcier 
planned fieldwork and logistics for the quantitative data collection using this, with considerable assistance from 
the FFE project’s M&E assistants. Forcier collected the quantitative data between 11 March and 16 April 2019, 
beginning with the comparison group schools, where there were fewer observations to conduct. Data collected 
using the ODK interface was uploaded to Forcier’s secure server by supervisors at the end of each day’s 
fieldwork. After cleaning, it was uploaded to NFER’s secure server. Forcier and NFER liaised extensively with the 
objective of ensuring the data was presented in suitable form for NFER to work with. 
Qualitative data were collected by Simone Doctors during two field visits to Mozambique (10 February – 3 March 
and 22 April – 11 May 2019). She was assisted in making logistical arrangements by the FFE2 management and 
M&E teams. She was accompanied in visits to the project districts to visit schools, HGSFGs and government 
officials by a member of the M&E team but all interviews were conducted without the presence of project 
representatives. When visiting EPFs, she travelled alone, for logistical reasons; however, she was accompanied 
on school visits to observe teaching practice by members of the EPF management and staff. Again, all interviews 
were conducted without EPF/ADPP staff being present. Semi-structured interviews and focus groups were 
recorded using a digital voice recorder and later transcribed by Forcier. These were then analyzed by Simone 
Doctors, using a version of framework analysis. 
Quality assurance protocols were in place to ensure the consistency of the data collection process and validity 
of the data. Random spot checks were conducted by Forcier to ensure consistency between enumerators and 
the validity of the data gathered. Forcier will provide more detailed information about this process on request. 
13 Two schools, one peri-urban, one rural, were selected. Both schools are receiving the FFE literacy program in 
local languages but do not form part of the evaluation sample. 
26
Spot checks of approximately 10% of the transcriptions of qualitative data were performed, comparing this 
against the original recordings: these revealed the transcription quality to be acceptable. 
Rigorous data security protocols were in place. These provided for all data to be uploaded to NFER’s secure 
server, so no data containing personal information was transmitted using other means. In particular, email was 
not used to transmit data containing personal information. In the field, all information was stored on encrypted 
computers, then uploaded to NFER’s secure server. Enumerators were not permitted to use the smart phones 
used for quantitative data collection for any other purpose (all enumerators had additional phones and were 
provided with airtime to allow them to make logistical arrangements and to ensure their own security). 
The internal self-review by project management and staff and the stakeholder review workshop were facilitated 
by Simone Doctors with assistance from a local researcher from Forcier. 
The draft report was prepared by Simone Doctors then sent to Planet Aid, ADPP and the different implementing 
partners for factual verification. The evaluators wished to ensure that the report fairly and accurately reflected 
the lived experience of the implementing teams and incorporated any information which provided context for 
the report’s findings. Simone Doctors and NFER responded to all comments and requests for clarification 
received as a result of this process and received a small number of follow-up comments. During this dialogue, 
the evaluators retained a strictly independent stance, dealing with each comment on a case-by-case basis and
deciding not to incorporate comments judged not to be relevant or supported by other sources of evidence. 
Following this dialogue, the final draft was prepared by Simone Doctors.
Challenges, delays and lessons learned
A number of challenges during field work and data analysis led to delays and lessons learned for the future: 
• The team of enumerators and supervisors found the logistics of accessing some of the more remote schools 
quite challenging, as suitable vehicles were not always available. Extreme weather impacted the fieldwork, 
with rain leaving some roads impassable: one school could not be visited, despite the enumerator team 
making three attempts. This was exacerbated by the fact that the enumerators and drivers were not familiar 
with the location of some of the schools, particularly within the comparison schools not under FFE2, 
although the FFE2 team had provided guidance and contact information for the school directors. The 
importance of having reliable GPS coordinates for each school and more information about the conditions 
of the roads, particularly leading to the comparison group schools, is a lesson learned when planning for the 
final evaluation. 
• Last minute edits to the ODK scripts to deal with problems identified during the piloting and initial field work 
were not fully tested due to pressure of time, leading to a small number of questions in the pupil 
27
questionnaire being skipped and some data relating to school clubs not being collected.14 Although 
regrettable, this loss of data is not actually critical, since information regarding school clubs was collected 
both from teachers and in the school survey. The evaluation is therefore able to report on the question of 
school clubs from two perspectives rather than three as intended.
• The data collected by Forcier and uploaded to NFER initially included a number of duplicate entries, likely 
caused by confusion amongst enumerators about the substitution process. In order to resolve this issue, 
the Forcier team manually verified the physical data tracking sheets to identify the observations, in distance 
liaison with Simone Doctors. In a small number of cases, where it was not possible to ascertain with certainty 
which entry was correct, both were deleted from the data set. This time-consuming process led to delays in 
the data analysis workplan. 
• Previous evaluations have documented the challenges caused by turnover and capacity challenges within 
the project M&E team. At the beginning of the evaluation process, there was only a skeleton M&E staff in 
post, and the external evaluators were supported with basic arrangements by the project leadership team 
and two M&E assistants. During the evaluation fieldwork, a new M&E manager and a new M&E specialist 
took up their functions. Both appear to be capable and competent; with support from the leadership team, 
they made considerable efforts to assist the external evaluators, accompany field visits and obtain 
information on request. Although M&E of the project has continued throughout the changes and gaps in 
personnel, these have inevitably led to some challenges in data collation and monitoring. It is promising 
that a fully-staffed M&E team is now in place. 
• Delays in NFER receiving useable data to analyze led to delays in the workplan overall and required 
previously unscheduled time to correct. 
• Some adjustments to the evaluation plan were necessary due to the two cyclones which devastated two 
regions of Mozambique. Two ADPP-run EPFs were affected, leading to one being evacuated and both being 
preoccupied with other vital matters during the period of evaluation. Some EPF information was therefore 
incomplete. Despite these events affecting many ADPP and FFE staff and their families, through involvement 
in the emergency response or because family members live in the affected areas, the ADPP leadership 
insisted that the evaluation should continue as planned and continued to provide all support necessary. 
• The discovery by the project team that the lists of schools supplied to the evaluators included some 
inaccuracies meant that, late in the report preparation process, it was necessary to verify the status of all 
schools in the sample, reassign certain schools to different intervention groups and rerun the statistical 
analysis of the pupil-level information. The other data sets were not affected. In the future, it should be 
noted that, at least for the purposes of the evaluation plan, it is preferable to avoid changes to the status of 
14Since these questions were only asked at the intervention schools and fieldwork began in the comparison 
schools, this problem was not picked up by Forcier’s usual QA processes.
28
schools, such as extending the literacy program to new schools. Where judged to be of such value to the 
schools as to outweigh other considerations, such changes should be rigorously noted for evaluation 
purposes, so the evaluation sample can be adjusted accordingly. 
29
Midterm findings
Previous evaluations have been structured around the three evaluation questions:
• Internal validity (is the project doing /did the project do what it said it would do?) 
• Initial impact (is it making /has it made a difference?) 
• Strategic relevance regarding effectiveness, efficiency, impact and sustainability (are/ were these the 
right things to do? what can be learned?) 
The current report does not follow this approach but is structured according to the logic of the project Theory 
of Change (see Annex 1), with findings presented in function of the project indicators, followed by a number of 
other transversal themes. Each section addresses the three evaluation questions to some extent. Given the 
intrinsic links between the areas covered by the indicators, the second and third evaluation questions regarding 
initial impact and strategic relevance are covered in more or less detail as appropriate, avoiding unnecessary 
repetition. Given that the project has already been evaluated several times, although the current report covers 
all project activities, it gives particular attention to the two new components which began with the second 
phase: i) the program to improve literacy instruction in primary schools, particularly literacy in local languages, 
and ii) the Home Grown School Feeding Gardens, which aim to make school feeding more sustainable, by 
promoting local production and diversification of food sources. Other project activities are often considered in 
less detail, unless there have been particular changes since baseline or the midterm data reveals something 
particular of note. The conclusion returns to the three evaluation questions listed above and addresses each of 
them in turn in the light of the midterm findings. 
Where data collected at midterm from students, teachers or schools are reported, it should be noted that these 
refer to a sample of those populations. Where appropriate, results are extrapolated to the whole population, as 
is the case where targets are based on indicators formulated in terms “percentage of students who…” or 
“number of schools which…”. In these cases, informed estimations have been made, based on the behavior
within a representative sample of the population. This should be borne in mind when, for example, results of 
the schools survey (conducted in 170 schools at midterm) expressed as percentages of those schools are 
compared with those of the baseline school survey (conducted in all 271 project schools at baseline) and 
expressed as percentages of all project schools. 
Improved literacy of school-age children (SO1)
The first of the two large Strategic Objectives of the project within the project ToC, (SO1) concerns Improved 
literacy of school-age children. Each project activity is presented following the structure and logic of the ToC: 
each section begins with boxes summarizing the relevant project results in relation to the specific indicators and 
30
targets associated with each activity. Midterm targets have been calculated as of March or April 2019 on the 
basis of targets for 201915
. 
Indicator 26 (Outcome): Percent of students who, by the end of two grades of primary schooling, 
demonstrate that they can read and understand the meaning of grade level text. Final target: 45%
Midterm target: 12% 
Midterm results (April 2019): 6% (Female: 7%; Male 5%)
Midterm target 50% achieved (Female: 67%; Male 42% achieved) 
Indicator 27 (Output): Number of individuals benefiting directly from USDA-funded interventions. 
Final target: 85,560 (Female: 45,450; Male: 40,110; Continuing: 78,748; New: 6,812) 
Midterm target: 77,088 (Female: 38,701; Male: 38,387; Continuing: 76,307; New: 781)16
Midterm results (March 2019): 92,237; Female: 48,119; Male: 44,118; Continuing: 92,081; New: 156
Midterm target 108% achieved (Female: 106%; Male: 110% achieved; Continuing: 121% achieved; 
New: 20% achieved) 
Indicator 28 (Output): Number of individuals benefiting indirectly from USDA-funded interventions. 
Final target: 336,000 
Midterm target: 336,000
Midterm results (March 2019): 336,104 
Final target 100% achieved 
Assessing the literacy levels of students who have benefited from the literacy program delivered by ADPP with 
technical assistance from CE is key to evaluating the success of this intervention and of the wider project. The 
report assesses the literacy program itself within the section entitled Improved Quality of Literacy Instruction 
(MGD 1.1) below. The cohort of pupils tested at baseline, when they were in grade 1, performed an Early Grade 
Reading Assessment (EGRA) comparable to that used at baseline. The EGRA used at baseline had been developed 
by a team of specialists from CE, language specialists from INDE, the FFE literacy team and the FFE M&E team, 
working closely with an international consultant. Three equivalent versions were developed in Portuguese, 
Xichangana and Xirhonga. The version used at midterm was improved by the external evaluator working closely 
with the FFE literacy team. The task was then coded using ODK and administered by enumerators using 
smartphones. 
The EGRA consists of 11 progressively ordered subtasks to assess pre-reading then reading skills, designed to 
reflect the progressive, cumulative stages of literacy acquisition. Pupils continue until they have reached the 
15 Midterm targets have been calculated on a pro rata basis, or as appropriate for the nature of the activity, 
taking the targets for 2019 as set out in the document entitled Attachment E in Amendment # 1: Performance 
Indicators Modification 1.
16New midterm target = 25% of 2019 new target. Continuing midterm target = all the rest in midterm target
31
threshold at which they can no longer perform the tasks; at this point, discontinuation filters stop them 
attempting tasks beyond their level of competence, to minimize student distress. However, all pupils attempt 
the writing task.
Each of the three language versions of EGRA consisted of the following subtasks (all of which are language 
specific, reflecting the phonological and orthographic rules of the target language):
• ST1. Oral vocabulary (name the object depicted in stimuli pictures – 10 items)
• ST2. Comprehension of an oral text (respond orally to comprehension questions after hearing a short text
– 4 items)
• ST3. Phonological awareness (indicate the picture of an object whose name begins with the same initial
sound as that of a stimulus object – 10 items)
• ST4. Concepts of print (perform tasks to demonstrate familiarity with how printed language functions – 10
items)
• ST5. Letter sounds (produce the letter sounds when shown printed lower and upper case letters – 100 items)
• ST6. Syllable recognition (read aloud syllables consisting of consonant and vowel combinations permitted
in the target language – 50 items)
• ST7. Reading simple words (read aloud a list of words progressively longer and less common - 30 items)
• ST8. Reading fluency (read a short passage – number of words read correctly within 60 seconds –70 items;
and Reading comprehension (respond to questions based on the passage just read – 4 items)
• ST9. Writing (correctly write first name and family name – 2 items); and Dictation - words (correctly write 5
words dictated – 5 items); Dictation – phrase (correctly write 5-word phrase - 5 items)
The 11 subtasks were analyzed as 9 subscales (reading fluency and reading comprehension are considered as 
one subscale; writing tasks were considered as one subscale). The results of each subtask were analyzed across 
the following comparisons: a) FFE + literacy (all) versus comparison (all); b) FFE + literacy (Portuguese) versus 
comparison (Portuguese); c) FFE + literacy (local languages) versus comparison (local languages); d) FFE + literacy 
(Portuguese) versus FFE only (Portuguese); e) FFE + literacy (Portuguese) versus FFE + literacy (local languages)17
. 
The results for each subtask are presented below (for more details, please see Technical Appendix 1, prepared 
by the statistical team from NFER). 
17 Following changes to the project since baseline, there is no longer a group of FFE only schools in project 
districts learning in local languages; all such schools are now benefiting from the literacy intervention. The 
sample and the evaluation design have been changed to reflect the fact that there is no longer an “FFE only 
(local languages)” intervention group. 
32
Oral vocabulary
The oral vocabulary subtask consists of ten rows of three pictures of everyday objects; pupils are asked to 
indicate the object named by the enumerator in their language of instruction (Portuguese, Xichangana or 
Xirhonga, as appropriate). 
a) Contrast 1: FFE + literacy (all) vs Comparison (all): overall, FFE + literacy students achieved significantly 
greater increases between baseline and midterm than did comparison students. 
b) Contrast 2: FFE + literacy (Portuguese) vs comparison (Portuguese): overall, FFE + literacy students 
studying and tested in Portuguese achieved significantly greater increases between baseline and 
midterm than did comparison students studying and tested in Portuguese.
c) Contrast 3: FFE + literacy (local languages) vs comparison (local languages): overall, FFE + literacy
students studying and tested in local languages (Xichangana or Xirhonga) achieved slightly smaller 
mean increases between baseline and midterm than students in the comparison group studying and 
tested in the same languages, although this difference was not significant. 
d) Contrast 4: FFE + literacy (Portuguese) vs FFE only (Portuguese): overall, FFE + literacy students 
studying and tested in Portuguese achieved significantly lesser increases between baseline and 
midterm than FFE only students studying and tested in Portuguese. 
e) Contrast 5: FFE + literacy (local languages) vs FFE + literacy (Portuguese): overall, FFE + literacy 
students studying and tested in Portuguese achieved significantly greater increases between baseline 
and midterm than did FFE + literacy students studying and tested in local languages. 
This is a straightforward task which would not be expected to cause difficulty for the majority of grade 3 pupils. 
In effect, project students performed well overall on this pre-reading task, with all groups achieving mean scores 
of over 9.5/10. Since the analysis looked at the difference between baseline and midterm, rather than the 
midterm results, it is the comparative performance since baseline which was measured, rather than the midterm 
result in isolation. For example, with regard to d) contrast 4 (FFE + literacy versus FFE only tested in Portuguese), 
although the FFE + literacy group outperformed the FFE only group at midterm, because they had outperformed 
them still further at baseline the FFE only group achieved significantly better results. 
The complete results are presented in Technical Appendix 1, pp 5 - 15. 
Oral comprehension
Pupils listened to a short story (48 words in length) read by the enumerator, then answered four simple 
questions relating to what they had heard. 
a) Contrast 1: FFE + literacy (all) vs Comparison (all): overall, comparison students achieved significantly 
greater increases between baseline and midterm than did FFE + literacy students. 
33
b) Contrast 2: FFE + literacy (Portuguese) vs comparison (Portuguese): overall, comparison students 
studying and tested in Portuguese achieved significantly greater increases between baseline and 
midterm than did FFE + literacy students studying and tested in Portuguese.
c) Contrast 3: FFE + literacy (local languages) vs comparison (local languages): overall, FFE + literacy 
students studying and tested in local languages (Xichangana or Xirhonga) achieved smaller mean 
increases between baseline and midterm than students in the comparison group studying and tested 
in the same languages, although this difference was not significant. 
d) Contrast 4: FFE + literacy (Portuguese) vs FFE only (Portuguese): overall, FFE + literacy students 
studying and tested in Portuguese achieved significantly greater increases between baseline and 
midterm than FFE only students studying and tested in Portuguese. 
e) Contrast 5: FFE + literacy (local languages) vs FFE + literacy (Portuguese): overall, FFE + literacy 
students studying and tested in local languages achieved smaller increases between baseline and 
midterm than did FFE + literacy students studying and tested in Portuguese, although this difference 
was not significant. 
Although results overall have increased since baseline, this remains a task which causes some difficulties to many 
pupils, both those tested in Portuguese and in local languages, with mean scores for all groups of below 3 out 
of 4, and mean scores for both the FFE only group and the comparison (local languages) of less than 2 out of 4, 
suggesting that some grade 3 pupils are struggling to process oral language effectively (in all groups, the 
minimum score was 0 and the maximum 4). 
The complete results are presented in Technical Appendix 1, pp 16 - 25. 
Phonological awareness 
Pupils were presented with a series of ten rows of three photos of everyday objects. For each row, they 
responded to a stimulus word pronounced by the enumerator by indicating the object whose name begins with 
the same initial consonant sound.
a) Contrast 1: FFE + literacy (all) vs Comparison (all): overall, comparison students progressed more 
between baseline and midterm than FFE + literacy students, although the difference was not significant. 
b) Contrast 2: FFE + literacy (Portuguese) vs comparison (Portuguese): overall, FFE + literacy students 
outperformed comparison students, although the difference was not significant. 
c) Contrast 3: FFE + literacy (local languages) vs comparison (local languages): overall, FFE + literacy 
students studying and tested in local languages (Xichangana or Xirhonga) achieved higher scores at 
midterm than students in the comparison group studying and tested in the same languages. 
Nevertheless, the comparison group made relatively more progress between baseline and midterm, 
although the difference was not statistically significant. 
34
d) Contrast 4: FFE + literacy (Portuguese) vs FFE only (Portuguese): overall, FFE + literacy students 
studying and tested in Portuguese achieved significantly greater increases between baseline and 
midterm than FFE only students studying and tested in Portuguese. 
e) Contrast 5: FFE + literacy (local languages) vs FFE + literacy (Portuguese): overall, FFE + literacy 
students studying and tested in Portuguese achieved slightly greater increases between baseline and 
midterm than did FFE + literacy students studying and tested in local languages, although this difference 
was not significant. 
All groups progressed in this activity between baseline and midterm, with most groups achieving mean midterm 
scores of between 6.3 and 7.8 out of 10 (in all groups, the minimum score was 0 and the maximum 10). 
The complete results are presented in Technical Appendix 1, pp 26 - 35. 
Concepts of print 
Within this sub task, pupils are asked to perform ten activities as evidence of their familiarity with 
printed materials. 
a) Contrast 1: FFE + literacy (all) vs Comparison (all): at midterm, FFE + literacy students achieved greater 
increases between baseline and midterm than comparison students overall, although the difference 
was not significant. 
b) Contrast 2: FFE + literacy (Portuguese) vs comparison (Portuguese): at midterm, comparison students 
achieved greater increases between baseline and midterm than FFE + literacy students, although the 
difference was not significant.
c) Contrast 3: FFE + literacy (local languages) vs comparison (local languages): at midterm, FFE + literacy 
students achieved greater increases between baseline and midterm than comparison students, 
although the difference was not significant. 
d) Contrast 4: FFE + literacy (Portuguese) vs FFE only (Portuguese): at midterm, FFE + literacy students 
achieved greater increases between baseline and midterm than FFE only students, although the 
difference was not significant. 
e) Contrast 5: FFE + literacy (local languages) vs FFE + literacy (Portuguese): overall, FFE + literacy 
students studying and tested in local languages achieved greater increases between baseline and 
midterm than did FFE + literacy students studying and tested in Portuguese, although this difference 
was not significant. 
All groups progressed in this activity between baseline and midterm, with all groups achieving mean midterm 
scores of between 6.2 and 7.6 out of 10 (in all groups, the minimum score was 0 and the maximum 10). 
The complete results are presented in Technical Appendix 1, pp 6 - 45. 
35
Letter sound recognition 
Pupils were asked to produce sounds of 100 upper- or lower-case letters, presented in ten rows of ten, producing 
as many as possible within 60 seconds. 
a) Contrast 1: FFE + literacy (all) vs Comparison (all): at midterm, FFE + literacy students progressed more
than comparison students overall although the difference was not significant.
b) Contrast 2: FFE + literacy (Portuguese) vs comparison (Portuguese): at midterm, FFE + literacy students
progressed more than comparison students overall although the difference was not significant.
c) Contrast 3: FFE + literacy (local languages) vs comparison (local languages): at midterm, FFE + literacy
students progressed significantly more than comparison students overall.
d) Contrast 4: FFE + literacy (Portuguese) vs FFE only (Portuguese): at midterm, FFE + literacy students
progressed significantly more than FFE only students overall.
e) Contrast 5: FFE + literacy (local languages) vs FFE + literacy (Portuguese): overall, FFE + literacy
students studying and tested in local languages achieved smaller increases between baseline and
midterm than did FFE + literacy students studying and tested in Portuguese, although this difference
was not significant.
Scores for this subtask were low overall (means of between 8.6 and 8.8 out of 100 in the project groups, with 
means of as low as 3.0 out of 100 in the comparison groups; the overall range of scores was between 0 and 100).
The low scores overall for this subtask reflect the fact that a large number of pupils failed to produce a single 
one of the first ten letter sounds correctly and therefore discontinued this activity (only 41% of the students
tested at midterm completed this activity). As noted at baseline, in Mozambique, due to the ways reading is 
traditionally taught, pupils shown a written letter are not generally able to produce the corresponding letter 
sound but tend to respond by naming that letter. Since teaching reading does not traditionally involve phonics 
or letter sound recognition, experience shows that teachers, trainee teachers and teacher trainers may also have 
difficulty producing letter sounds. Given this knowledge, considerable focus was given to preparing enumerators 
to administer this sub-task during enumerator training; it is nevertheless possible that either pupils’ 
performance of the task or enumerator grading of this may have been impacted by the enumerators’ relative 
lack of familiarity with the task. 
The complete results are presented in Technical Appendix 1, pp 46 - 55. 
Reading syllables
Pupils were asked to read 50 syllables constituting plausible vowel-consonant combinations of the languages of 
testing, reading as many as possible within 60 seconds. 
a) Contrast 1: FFE + literacy (all) vs Comparison (all): at midterm, FFE + literacy students progressed less
than comparison students overall although the difference was not significant.
36
b) Contrast 2: FFE + literacy (Portuguese) vs comparison (Portuguese): at midterm, FFE + literacy students 
progressed more than comparison students overall although the difference was not significant.
c) Contrast 3: FFE + literacy (local languages) vs comparison (local languages): at midterm, FFE + literacy 
students progressed more than comparison students overall although the difference was not 
significant. 
d) Contrast 4: FFE + literacy (Portuguese) vs FFE only (Portuguese): at midterm, FFE + literacy students 
progressed significantly more than FFE only students overall. 
e) Contrast 5: FFE + literacy (local languages) vs FFE + literacy (Portuguese): overall, FFE + literacy 
students studying and tested in Portuguese progressed significantly more than did FFE + literacy 
students studying and tested in local languages. 
All groups progressed overall in this activity between baseline and midterm and performed better overall than 
on the previous activity (reading letter sounds). However, low scores overall for this subtask (means of between 
4.0 and 9.7 out of 50 in the project groups, with means of as low as 2.5 out of 50 in the comparison groups; the 
overall range of scores was between 0 and 50) reflect the fact that only approximately 40% of pupils completed 
this activity, the others having discontinued after 60 seconds or when they failed to produce any one of the first 
ten syllables correctly. Of the 41% of all students who attempted this activity across all groups, 45% scored 
between 0 and 10 out of 50; 19% between 11 and 20; 13% between 21 and 30; 12% between 31 and 40 and 11% 
between 41 and 50 out of 50. 
The complete results are presented in Technical Appendix 1, pp 56 - 65. 
Reading words 
Pupils were asked to read a series of 30 progressively more difficult words aloud. This task was not timed.
a) Contrast 1: FFE + literacy (all) vs Comparison (all): at midterm, FFE + literacy students progressed more 
than comparison students overall although the difference was not significant. 
b) Contrast 2: FFE + literacy (Portuguese) vs comparison (Portuguese): at midterm, FFE + literacy students 
progressed more than comparison students overall although the difference was not significant.
c) Contrast 3: FFE + literacy (local languages) vs comparison (local languages): at midterm, FFE + literacy 
students progressed significantly more than comparison students overall. 
d) Contrast 4: FFE + literacy (Portuguese) vs FFE only (Portuguese): at midterm, FFE + literacy students 
progressed significantly more than FFE only students overall. 
e) Contrast 5: FFE + literacy (local languages) vs FFE + literacy (Portuguese): overall, FFE + literacy 
students studying and tested in local languages progressed more than did FFE + literacy students 
studying and tested in Portuguese, although the difference was not significant. 
37
Students who attempted this task performed relatively better than on the previous one (with means of between 
7.1 and 8.2 out of 30 in the project groups, and means of between 2.4 and 6.4 out of 30 in the comparison 
groups; the overall range of scores was between 0 and 30). The results are extremely polarized: of the 41% of 
all students who attempted this activity across all groups, 43% scored between 0 and 10; 17% between 11 and 
20; and 40% between 21 and 30. At either extreme, 16% scored 0, whereas 13% scored 30 out of 30. The 
complete results are presented in Technical Appendix 1, pp 66 – 75.
Reading fluency (words per minute)
Pupils were asked to read a short text (70 words in Portuguese or 53 words in Xichangana or Xirhonga: the length 
of the local language texts is equivalent to the Portuguese text but has fewer words due to the morphological 
structure of those languages). The task was timed and the number of words read correctly within one minute 
recorded to give a fluency rating; pupils then had a further two minutes to finish reading the text if needed. The 
results below relate to reading fluency (words per minute). 
a) Contrast 1: FFE + literacy (all) vs Comparison (all): at midterm, FFE + literacy students progressed less 
than comparison students overall, although the difference was not significant. 
b) Contrast 2: FFE + literacy (Portuguese) vs comparison (Portuguese): at midterm, FFE + literacy students 
progressed less than comparison students overall although the difference was not significant.
c) Contrast 3: FFE + literacy (local languages) vs comparison (local languages): at midterm, FFE + literacy 
students progressed more than comparison students overall although the difference was not 
significant. 
d) Contrast 4: FFE + literacy (Portuguese) vs FFE only (Portuguese): at midterm, FFE + literacy students 
progressed significantly more than FFE only students overall. 
e) Contrast 5: FFE + literacy (local languages) vs FFE + literacy (Portuguese): overall, FFE + literacy 
students studying and tested in local languages progressed lessthan did FFE + literacy students studying 
and tested in Portuguese, although the difference was not significant. 
Of the 41% of students who attempted this activity, mean scores for the number of words read correctly in 60 
seconds were between 6.7 and 8.3 in the project groups, and between 2.4 and 8.7 in the comparison groups; 
the overall range of scores was between 0 and 70 for students tested in Portuguese and between 0 and 53 for 
students tested in local languages. Of the students who attempted this activity, 66% read between 0 and 20 
words in 60 seconds; 22% read between 21 and 40 words; 13% read between 41 and 70 words. The relationship 
between fluency and accuracy is discussed on page 39 below.
The complete results are presented in Technical Appendix 1, pp 76 - 85.
38
Comparative reading fluency
In order to compare the performance of students tested in Xichangana and Xirhonga with those tested in 
Portuguese, overall fluency scores were divided by 70 for students tested in Portuguese and 53 for those tested 
in Xichangana and Xirhonga. 
a) Contrast 1: FFE + literacy (all) vs Comparison (all): at midterm, FFE + literacy students progressed more 
than comparison students overall although the difference was not significant. 
b) Contrast 2: FFE + literacy (Portuguese) vs comparison (Portuguese): at midterm, FFE + literacy students 
progressed less than comparison students overall although the difference was not significant.
c) Contrast 3: FFE + literacy (local languages) vs comparison (local languages): at midterm, FFE + literacy 
students progressed more than comparison students overall although the difference was not 
significant. 
d) Contrast 4: FFE + literacy (Portuguese) vs FFE only (Portuguese): at midterm, FFE + literacy students 
progressed significantly more than FFE only students overall. 
e) Contrast 5: FFE + literacy (local languages) vs FFE + literacy (Portuguese): overall, FFE + literacy 
students studying and tested in local languages progressed lessthan did FFE + literacy students studying 
and tested in Portuguese, although the difference was not significant. 
The complete results are presented in Technical Appendix 1, pp 86 - 95.
Overall reading ability
In order to assess students’ overall ability to read the text, regardless of timing, as opposed to fluency (words 
per minute), students who had not finished reading the text within 60 seconds continued reading for up to a 
further 2 minutes or until they completed the text or discontinued the activity. This condition did not apply at 
baseline, when all students discontinued reading the text after 60 seconds. It was included at midterm at the 
request of the literacy team in order to assess pupils’ reading of a text without pressure of time. 
a) Contrast 1: FFE + literacy (all) vs Comparison (all): at midterm, FFE + literacy students performed less 
well than comparison students overall although the difference was not significant. 
b) Contrast 2: FFE + literacy (Portuguese) vs comparison (Portuguese): at midterm, FFE + literacy students 
performed less well than comparison students overall although the difference was not significant.
c) Contrast 3: FFE + literacy (local languages) vs comparison (local languages): at midterm, FFE + literacy 
students performed significantly better than comparison students overall. 
d) Contrast 4: FFE + literacy (Portuguese) vs FFE only (Portuguese): at midterm, FFE + literacy students 
performed better than FFE only students overall although the difference was not significant. 
39
e) Contrast 5: FFE + literacy (local languages) vs FFE + literacy (Portuguese): overall, FFE + literacy 
students studying and tested in local languages performed less well than did FFE + literacy students 
studying and tested in Portuguese, although the difference was not significant. 
Due to the lack of baseline values, these differences cannot conclusively be attributed to the intervention.
The complete results are presented in Technical Appendix 1, pp 96 - 105.
It is interesting to compare the number of words students are able to read correctly when they are allowed 
three minutes to read the test (overall reading ability or accuracy), with the number of words they are able to 
read correctly in a minute (fluency). Overall, there is an inverse relationship between fluency and accuracy: 
accuracy increased as fluency decreased (see figure 2 below). Of the 41% of students who attempted the task, 
only 5% were able to correctly read between 61 – 70 words in a minute, whereas 24% were able to correctly 
read between 61 to 70 words in three minutes. 
Figure 2: Comparative fluency and accuracy scores reading a text (all students who attempted task)
Comparative overall reading ability
In order to compare the overall reading ability (total number of words correctly read) of students tested in 
Xichangana and Xirhonga with those tested in Portuguese, overall reading ability scores were divided by 70 for 
students tested in Portuguese and 53 for those tested in Xichangana and Xirhonga. 
a) Contrast 1: FFE + literacy (all) vs Comparison (all): at midterm, FFE + literacy students performed 
slightly better than comparison students overall, although the difference was not significant. 
Percent of 
students
40
b) Contrast 2: FFE + literacy (Portuguese) vs comparison (Portuguese): FFE + literacy students progressed 
slightly less well than comparison students between baseline and midterm, although the difference was 
not significant.
c) Contrast 3: FFE + literacy (local languages) vs comparison (local languages): at midterm, FFE + literacy 
students performed significantly better than comparison students overall. 
d) Contrast 4: FFE + literacy (Portuguese) vs FFE only (Portuguese): at midterm, FFE + literacy students 
performed better than FFE only students overall although the difference was not statistically significant. 
e) Contrast 5: FFE + literacy (local languages) vs FFE + literacy (Portuguese): overall, FFE + literacy 
students studying and tested in local languages performed better than did FFE + literacy students 
studying and tested in Portuguese, although the difference was not significant. 
The complete results are presented in Technical Appendix 1, pp 105 - 115.
Comprehension of reading 
To assess whether pupils understood what they had read, they were asked between one and four 
comprehension questions corresponding to the point in the text they had reached when reading.
a) Contrast 1: FFE + literacy (all) vs Comparison (all): at midterm, FFE + literacy students made more 
progress between baseline and midterm than comparison students overall, although the difference was 
not significant. 
b) Contrast 2: FFE + literacy (Portuguese) vs comparison (Portuguese): at midterm, FFE + literacy students 
performed less well than comparison students overall although the difference was not significant.
c) Contrast 3: FFE + literacy (local languages) vs comparison (local languages): at midterm, FFE + literacy 
students performed significantly better than comparison students overall. 
d) Contrast 4: FFE + literacy (Portuguese) vs FFE only (Portuguese): at midterm, FFE + literacy students 
performed significantly better than FFE only students overall. 
e) Contrast 5: FFE + literacy (local languages) vs FFE + literacy (Portuguese): overall, FFE + literacy 
students studying and tested in local languages performed less well than did FFE + literacy students 
studying and tested in Portuguese, although the difference was not significant. 
The comprehension results overall were low: overall the mean scores of the intervention groups were between 
0.49 and 0.57 out of a possible total of 4; overall the mean scores of the comparison groups were between 0.11
and 0.60 out of a possible total of 4; scores ranged between 0 and 4 overall. Comprehension results were 
polarized: of the 41% of students who attempted this activity, 37% scored 0 out of 4, whereas 25% scored 4 out 
of 4. 16% students scored 1 out of 4; 13% scored 2 out of 4; 8% scored 3 out of 4. It should be remembered that 
students only attempted the comprehension questions for those portions of the text they had finished reading. 
The complete results are presented in Technical Appendix 1, pp 115 - 125.
41
All pupils, regardless of whether they had completed the reading tasks or discontinued these, were asked to 
perform a series of short writing tasks: writing their first name, writing their family name, writing five discrete 
words as a dictation exercise and writing a short phrase of five words. Discontinuation protocols operated within 
the writing subtasks. 
Writing first name correctly 
a) Contrast 1: FFE + literacy (all) vs Comparison (all): at midterm, FFE + literacy students made less
progress between baseline and midterm than comparison students overall, although the difference was 
not significant. 
b) Contrast 2: FFE + literacy (Portuguese) vs comparison (Portuguese): at midterm, FFE + literacy students 
performed less well than comparison students overall although the difference was not significant.
c) Contrast 3: FFE + literacy (local languages) vs comparison (local languages): at midterm, FFE + literacy 
students performed better than comparison students overall although the difference was not 
significant. 
d) Contrast 4: FFE + literacy (Portuguese) vs FFE only (Portuguese): at midterm, FFE + literacy students 
performed significantly better than FFE only students overall. 
e) Contrast 5: FFE + literacy (local languages) vs FFE + literacy (Portuguese): overall, FFE + literacy 
students studying and tested in local languages made more progress between baseline and midterm
than FFE + literacy students studying and tested in Portuguese, although the difference was not 
significant. 
All participating students were asked to write their first and family names. Given that all students are in the third 
grade, it is striking that only 73% of the students from intervention groups and between 39% and 74% from 
comparison groups were able to perform this task correctly. 
The complete results are presented in Technical Appendix 1, pp 155 - 161. 
Writing family name correctly 
a) Contrast 1: FFE + literacy (all) vs Comparison (all): at midterm, FFE + literacy students performed less 
well than comparison students overall although the difference was not significant. 
b) Contrast 2: FFE + literacy (Portuguese) vs comparison (Portuguese): at midterm, FFE + literacy students 
performed less well than comparison students overall although the difference was not significant.
c) Contrast 3: FFE + literacy (local languages) vs comparison (local languages): at midterm, FFE + literacy 
students performed better than comparison students overall although the difference was not 
significant. 
d) Contrast 4: FFE + literacy (Portuguese) vs FFE only (Portuguese): at midterm, FFE + literacy students 
performed better than FFE only students overall although the difference was not significant. 
42
e) Contrast 5: FFE + literacy (local languages) vs FFE + literacy (Portuguese): overall, FFE + literacy 
students studying and tested in local languages performed less well than FFE + literacy students 
studying and tested in Portuguese, although the difference was not significant. 
Overall, pupils were less likely to be able to write their family names correctly than their first names: of the 
intervention groups, between 42% and 45% of students were able to write their family name correctly; within 
the comparison groups this ranged from between 23% to 47%. Again, this is a striking finding given the age of 
the students. 
The complete results are presented in Technical Appendix 1, pp 161 - 167.
Writing whole name correctly 
A further analysis of the same items (first name and family name) sought to assess whether students were able 
to write their complete name correctly. 
a) Contrast 1: FFE + literacy (all) vs Comparison (all): at midterm, FFE + literacy students performed less 
well than comparison students overall although the difference was not significant. 
b) Contrast 2: FFE + literacy (Portuguese) vs comparison (Portuguese): at midterm, FFE + literacy students 
performed less well than comparison students overall although the difference was not significant.
c) Contrast 3: FFE + literacy (local languages) vs comparison (local languages): at midterm, FFE + literacy 
students performed significantly better than comparison students overall. 
d) Contrast 4: FFE + literacy (Portuguese) vs FFE only (Portuguese): at midterm, FFE + literacy students 
performed better than FFE only students overall although the difference was not significant. 
e) Contrast 5: FFE + literacy (local languages) vs FFE + literacy (Portuguese): overall, FFE + literacy 
students studying and tested in local languages performed less well than FFE + literacy students 
studying and tested in Portuguese, although the difference was not significant. 
The complete results are presented in Technical Appendix 1, pp 167 - 173.
Writing dictation: individual words
At baseline, pupils had been asked to write ten progressively more difficult individual words as a dictation. At 
midterm, at the request of the literacy team, this task was divided into two discrete elements in order to 
investigate whether pupils would perform better if the words were presented in the context of a simple phrase: 
a dictation of five individual words followed by a dictation of a short phrase containing five words. Due to this 
change in the format of the task, it was not possible to compare the midterm results directly with those from 
baseline. 
43
a) Contrast 1: FFE + literacy (all) vs Comparison (all): at midterm, FFE + literacy students performed less 
well than comparison students overall although the difference was not significant. 
b) Contrast 2: FFE + literacy (Portuguese) vs comparison (Portuguese): at midterm, FFE + literacy students 
performed less well than comparison students overall although the difference was not significant.
c) Contrast 3: FFE + literacy (local languages) vs comparison (local languages): at midterm, FFE + literacy 
students performed significantly better than comparison students overall. 
d) Contrast 4: FFE + literacy (Portuguese) vs FFE only (Portuguese): at midterm, FFE + literacy students 
performed significantly better than FFE only students overall. 
e) Contrast 5: FFE + literacy (local languages) vs FFE + literacy (Portuguese): overall, FFE + literacy 
students studying and tested in local languages performed less well than FFE + literacy students 
studying and tested in Portuguese, although the difference was not significant. 
Of the students who performed this task, the mean results ranged between 0.93 to 1.07 out of 5 for the 
intervention groups and between 0.21 and 1.12 out of 5 for the comparison groups; the range was from 0 to 5 
out of 5. The complete results are presented in Technical Appendix 1, pp 136 - 145.
Writing dictation: phrase
In writing the short phrase consisting of five words:
a) Contrast 1: FFE + literacy (all) vs Comparison (all): at midterm, FFE + literacy students performed better 
than comparison students overall although the difference was not significant. 
b) Contrast 2: FFE + literacy (Portuguese) vs comparison (Portuguese): at midterm, FFE + literacy students 
performed better than comparison students overall although the difference was not significant.
c) Contrast 3: FFE + literacy (local languages) vs comparison (local languages): at midterm, FFE + literacy 
students performed significantly better than comparison students overall. 
d) Contrast 4: FFE + literacy (Portuguese) vs FFE only (Portuguese): at midterm, FFE + literacy students 
performed significantly better than FFE only students overall. 
e) Contrast 5: FFE + literacy (local languages) vs FFE + literacy (Portuguese): overall, FFE + literacy 
students studying and tested in local languages performed less well than FFE + literacy students 
studying and tested in Portuguese, although the difference was not significant. 
Of the students who performed this task, the mean results ranged between 1.06 to 1.09 out of 5 for the 
intervention groups and between 0.20 and 0.92 out of 5 for the comparison groups; the range was from 0 to 5 
out of 5. The students receiving the literacy intervention performed the writing task better when the five words 
formed a coherent phrase than when they were five discrete words; this was not the case for the students not 
receiving the literacy intervention (FFE only and comparison students). The complete results are presented in 
Technical Appendix 1, pp 145 - 155.
44
Total writing 
A total writing score was calculated based on the sum of scores for all the writing tasks (writing first name, 
writing family name, writing dictation words, writing dictation phrase). 
a) Contrast 1: FFE + literacy (all) vs Comparison (all): at midterm, FFE + literacy students performed less 
well than comparison students overall, although the difference was not statistically significant. 
b) Contrast 2: FFE + literacy (Portuguese) vs comparison (Portuguese): at midterm, FFE + literacy students 
performed significantly less well than comparison students overall.
c) Contrast 3: FFE + literacy (local languages) vs comparison (local languages): at midterm, FFE + literacy 
students performed better than comparison students overall, although the difference was not 
significant. 
d) Contrast 4: FFE + literacy (Portuguese) vs FFE only (Portuguese): at midterm, FFE + literacy students 
performed significantly better than FFE only students overall. 
e) Contrast 5: FFE + literacy (local languages) vs FFE + literacy (Portuguese): overall, FFE + literacy 
students studying and tested in Portuguese performed better than FFE + literacy students studying and 
tested in local languages, although the difference was not significant. 
The mean total writing results ranged between 2.10 to 2.22 out of 12 for the intervention groups and between 
0.92 and 3.05 out of 12 for the comparison groups; the range was from 0 to 10 out of 12 for all groups except 
for the comparison (Portuguese) group, which had a range of 0 - 12. Students in the comparison (Portuguese) 
performed significantly better in writing overall than the other groups. 
The complete results are presented in Technical Appendix 1, pp 126 - 135.
EGRA results: Discussion 
At baseline, the EGRA results for the cohort tested, then in grade 1, had revealed very low literacy outcomes 
overall, with almost no grade 1 learners able to proceed beyond the first four sub tasks (this was not surprising, 
given that, at the time of testing, they had only been attending school for three to four months). At midterm, 
the situation is much more nuanced: although many learners discontinued the reading part of the EGRA after 
the first five tasks, since they were not able to read any syllables, 41% continued and completed the subtasks 
on reading syllables, reading words, reading a text and reading comprehension. The results of these subtasks 
should therefore be considered in this light; for example, although the overall mean score for reading words in 
the FFE + literacy group was 7.13/30, the overall mean score of those who actually attempted the subtask was 
closer to 20/30. 
Overall, then, the pupils tested have progressed since baseline, although almost 60% are still not able to read 
letter sounds or syllables. Two questions must be answered:
45
1. How have the pupils tested progressed compared to expectations (national benchmarks)?
2. What has been the impact of the literacy intervention? Have the students in the FFE + literacy group 
progressed more than the other students tested? 
Progression in relation to national benchmarks 
The indicator “Percent of students who, by the end of two grades of primary schooling, demonstrate that they 
can read and understand the meaning of grade level text” assumes a nationally accepted benchmark, or general 
agreement as to what children of a particular grade should be able to do. In the absence of official national 
benchmarks, but in line with recommendations developed by the USAID Vamos Ler (Let’s Read) project18, the 
project literacy team has been working with the following definition: 30 words per minute (WPM) for 
Portuguese; 25 WPM for local languages (Xichangana/Xirhonga) and 3/4 (75%) comprehension questions correct 
in each case. This is the measure of “demonstrate that they can read and understand the meaning of grade level 
text” which is therefore adopted for the purposes of this evaluation. 
At midterm, pupils within the FFE + literacy group were compared with those in the FFE only and comparison 
groups, using the same 5 comparisons as for the individual EGRA subtasks, in order to investigate the role of the 
literacy intervention on their ability to meet the benchmark. 
a) Contrast 1: FFE + literacy (all) vs Comparison (all): at midterm, 6.7 % of FFE + literacy students met the 
benchmark compared with 8.3% of comparison students; FFE + literacy students performed less well 
than comparison students overall, although this difference was not significant. 
b) Contrast 2: FFE + literacy (Portuguese) vs comparison (Portuguese): at midterm, 7.8 % of FFE + literacy 
students met the benchmark compared with 9.1% of comparison students; FFE + literacy students 
performed less well than comparison students overall, although this difference was not significant.
c) Contrast 3: FFE + literacy (local languages) vs comparison (local languages): at midterm, 5.2% of FFE + 
literacy students met the benchmark, compared with 1.9% of comparison students; FFE + literacy 
students performed better than the comparison students, although the difference was not significant. 
d) Contrast 4: FFE + literacy (Portuguese) vs FFE only (Portuguese): at midterm, 7.8% of FFE + literacy 
students met the benchmark, compared with 4.5% of FFE only students; FFE + literacy students
performed better than FFE only students, although this difference was not statistically significant. 
e) Contrast 5: FFE + literacy (local languages) vs FFE + literacy (Portuguese): at midterm, 7.8% of FFE + 
literacy students studying and tested in Portuguese met the benchmark, compared with 5.2% of FFE + 
literacy students studying and tested in local languages; FFE + literacy students studying and tested in 
18 The five-year USAID-funded Mozambique Vamos Ler! Program, which began in 2016, develops bilingual 
education pedagogical tools and activities to improve national early grade literacy. 
46
Portuguese, performed better than FFE + literacy students studying and tested in local languages, 
although the difference was not significant . 
The complete results are presented in Technical Appendix 1, pp 173 - 178.
Progression in relation to national benchmarks: boys 
a) Contrast 1: FFE + literacy (all) vs Comparison (all): at midterm, 5.9 % of FFE + literacy students met the 
benchmark compared with 7.2% of comparison students; FFE + literacy students performed less well 
than comparison students overall, although this difference was not significant. 
b) Contrast 2: FFE + literacy (Portuguese) vs comparison (Portuguese): at midterm, 7.6 % of FFE + literacy 
students met the benchmark compared with 7.9% of comparison students; FFE + literacy students 
performed less well than comparison students overall, although this difference was not significant.
c) Contrast 3: FFE + literacy (local languages) vs comparison (local languages): at midterm, 3.4% of FFE + 
literacy students met the benchmark, compared with 1.8% of comparison students; FFE + literacy 
students performed better than the comparison students, although the difference was not significant. 
d) Contrast 4: FFE + literacy (Portuguese) vs FFE only (Portuguese): at midterm, 7.6% of FFE + literacy 
students met the benchmark, compared with 2.5% of FFE only students; FFE + literacy students 
performed significantly better than FFE only students. 
e) Contrast 5: FFE + literacy (local languages) vs FFE + literacy (Portuguese): at midterm, 7.6% of FFE + 
literacy students studying and tested in Portuguese met the benchmark, compared with 3.4% of FFE + 
literacy students studying and tested in local languages; FFE + literacy students studying and tested in 
Portuguese, performed better than FFE + literacy students studying and tested in local languages, 
although the difference was not significant.
The complete results are presented in Technical Appendix 1, pp 178 - 183.
Progression in relation to national benchmarks: girls 
a) Contrast 1: FFE + literacy (all) vs Comparison (all): at midterm, 7.6 % of FFE + literacy students met the 
benchmark compared with 9.4% of comparison students; FFE + literacy students performed less well 
than comparison students overall, although this difference was not significant. 
b) Contrast 2: FFE + literacy (Portuguese) vs comparison (Portuguese): at midterm, 8.1 % of FFE + literacy 
students met the benchmark compared with 10.1% of comparison students; FFE + literacy students 
performed less well than comparison students overall, although this difference was not significant.
c) Contrast 3: FFE + literacy (local languages) vs comparison (local languages): at midterm, 6.9% of FFE + 
literacy students met the benchmark, compared with 1.9% of comparison students; FFE + literacy 
students performed better than the comparison students, although the difference was not significant. 
47
d) Contrast 4: FFE + literacy (Portuguese) vs FFE only (Portuguese): at midterm, 8.1% of FFE + literacy 
students met the benchmark, compared with 7.1% of FFE only students; FFE + literacy students 
performed better than FFE only students, although this difference was not significant. 
e) Contrast 5: FFE + literacy (local languages) vs FFE + literacy (Portuguese): at midterm, 8.1% of FFE + 
literacy students studying and tested in Portuguese met the benchmark, compared with 6.9% of FFE + 
literacy students studying and tested in local languages; FFE + literacy students studying and tested in 
Portuguese, performed better than FFE + literacy students studying and tested in local languages, 
although the difference was not significant.
The complete results are presented in Technical Appendix 1, pp 183 - 188.
8% of students (8% of boys; 8% of girls) in the FFE + literacy (Portuguese) group met the benchmark, compared 
with 5% (2% of boys; 7% of girls) in the FFE only group and 5% (3% of boys; 7% of girls) in the FFE + literacy 
(bilingual) group (see table 5 below). Amongst project students, the percentage of those reaching the national 
benchmark is thus 6% (5% boys; 7% girls). The midterm target of 12% for this indicator has thus been 50% 
achieved (42% for boys; 58% for girls). This result should be considered both in the light of the overall results 
presented above, and of the differences discussed below. 
Intervention groups
Did not 
achieve 
benchmark
Achieved 
benchmark
Percent 
achieved 
benchmark Total
Comparison (Bilingual) 106 2 2% 108
Comparison (Portuguese) 884 88 9% 972
FFE + lit (Bilingual) 716 39 5% 755
FFE + lit (Portuguese) 956 81 8% 1037
FFE only 908 43 5% 951
Total project students 2,580 163 6% 2,743
Grand Total 3,570 253 7% 3,823
What has been the impact of the literacy intervention? Have the students in the FFE + literacy group 
progressed more than the other students tested? 
The analysis of the EGRA results shows a nuanced picture. One the one hand, the overall results demonstrate a 
very low level of reading overall, with only 6% of project students (7% of those benefiting from the literacy 
intervention) meeting the national benchmark, compared with 8% of comparison students overall. The results 
do not demonstrate a resounding impact of the literacy intervention to date, in which FFE + literacy children 
consistently outperform the comparison group children. However, taken overall, the results suggest that the 
literacy intervention is beginning to have an impact: Table 6 below summarizes the significant differences in 
results of EGRA subtasks between the intervention groups (all the results reported are included, meaning that 
Table 5: Students achieving national benchmarks 
48
Table 6: Statistically significant differences in results of EGRA subtasks between intervention groups
some are reported twice. For example, “writing first name correctly”, “writing family name correctly” and 
“writing whole name correctly” are presented on three separate lines, so some of the information presented 
Subtask Significant difference observed?
FFE+literacy 
(all) vs 
comparison 
(all) 
FFE+literacy 
(Portuguese)
vs 
comparison 
(Portuguese) 
FFE+literacy 
(local langs)
vs 
comparison 
(local langs)
FFE+literacy 
(Portuguese)
vs FFE only 
(Portuguese)
FFE+literacy 
(local langs) 
vs 
FFE+literacy 
(Portuguese)
Oral vocabulary + significant + significant --ns -- significant -- significant 
Oral 
comprehension
-- significant -- significant -- ns + significant -- ns
Phonological 
awareness
-- ns + ns --ns + significant -- ns
Concepts of print +ns --ns + ns + ns +ns
Letter sound 
recognition
+ ns + ns + significant + significant -- ns
Reading syllables -- ns + ns + ns + significant -- significant 
Reading words + ns + ns + significant + significant +ns
Reading fluency 
(words per minute)
--ns --ns + ns + significant -- ns
Comparative 
reading fluency 
+ ns -- ns + ns + significant -- ns
Overall reading 
ability 
--ns --ns + significant + ns -- ns
Comparative
overall reading 
+ ns = ns + significant + ns +ns
Comprehension of 
reading
+ns --ns + significant + significant -- ns
Writing first name 
correctly
--ns --ns + ns + significant +ns
Writing family 
name correctly 
--ns --ns + ns + ns -- ns
Writing whole 
name correctly
--ns --ns + significant + ns -- ns
Writing dictation: 
individual words
--ns --ns + significant + significant -- ns
Writing dictation: 
phrase
+ ns + ns + significant + significant -- ns
Total writing --ns -- significant + ns + significant -- ns
 KEY
+ significant
treatment group 
performed 
significantly better 
-- significant 
treatment group 
performed 
significantly worse
+ ns
treatment group 
performed better but 
difference not 
statistically significant 
--ns
treatment group 
performed worse but 
difference not 
statistically significant
treatment groups are 
in red in column 
heads
49
overlaps). Across all comparisons, in 16 of the 18 lines presented in Table 6, the FFE + literacy group performed 
significantly better than either the comparison group or the FFE only group, in one or more instances 19
. The 
FFE + literacy group did not perform significantly better than the comparison group overall. However, in 8 of the 
18 cases reported, of the students tested in local languages, the FFE + literacy group performed significantly 
better than the comparison group; there were no cases where the comparison (local languages) group 
performed significantly better than the FFE + literacy (local languages) group. 
There was only one instance where the FFE + literacy group performed significantly better than the comparison 
group when students were tested in Portuguese. In two cases, the comparison group performed significantly 
better than the FFE + literacy group when students were tested in Portuguese. Taken together, these findings 
provide tentative evidence of the efficacy of teaching literacy in local languages. This suggests that, while more 
time may be needed for the effects of the intervention to translate into improved literacy on a wide scale, the 
first signs are promising. 
In 12 of the 18 cases reported, the FFE + literacy group performed significantly better than the FFE only group 
overall, amongst students tested in Portuguese. This finding provides promising early evidence that the literacy 
intervention is having a positive impact on the emerging literacy skills of students within schools where the FFE 
project is operating. While more time may be needed for the effects of the intervention to translate into more 
consistent results, the first signs are encouraging. 
When the results of FFE + literacy (Portuguese) are compared with those of FFE + literacy (local languages), only 
two lines show a significant difference: students tested in Portuguese performed significantly better than those 
tested in local languages in the oral vocabulary subtask and the reading syllables subtask. 
Although the FFE + literacy students did not perform significantly better than the comparison students overall, 
two trends emerge clearly from this situation:
i) Amongst students taught and tested in local languages, the FFE + literacy students performed 
significantly better than the comparison students in 8 of the 18 lines; the FFE + literacy students 
performed better than the comparison students in a further 7 of the 18 lines, although these 
differences were not statistically significant.
ii) Amongst students taught and tested in Portuguese, the FFE + literacy students performed 
significantly better than the FFE only students in 12 of the 18 lines; the FFE + literacy students 
19 This discussion considers only results where a difference observed is statistically significant. In reality, there 
were other results where a difference was observed which suggested the FFE + literacy group of students 
outperformed one of the other groups; in some cases, these results would have been statistically significant if a 
lower threshold had been applied.
50
performed better than the FFE only students in a further 5 of the 8 lines, although these differences 
were not significant. 
One possible explanation for the fact that the FFE + literacy group did not perform significantly better than the 
comparison group overall might concern the characteristics of the comparison group. It had become clear at 
baseline that, despite the careful sampling plan drawn for this study, the comparison group consistently 
outperformed the two intervention groups in literacy. The baseline report speculated that the comparison group 
could not be considered truly comparable to the FFE + literacy and FFE only groups but differed in terms of other 
variables. Demographic and socio-economic factors were offered as possible explanations: two of the districts 
in the comparison group, Matola and Marracuene, are peri-urban, and have a higher socioeconomic profile than 
the four project districts. Indeed, the four districts of Maputo Province in which the project is implemented were 
chosen precisely because they have the lowest socioeconomic profiles and the highest rates of poverty, food 
insecurity and malnutrition of the districts in Maputo Province.
In order to probe further the possibility of differences in performance within the comparison group, the mean 
scores of the EGRA sub tasks for each district were combined in a single figure. Figure 3 below reveals that, in 
contrast to the assumption that students from Matola and Marracuene might be outperforming those from 
other districts, it is in fact students from Boane who out-perform those from other districts overall; students 
from Matutuine and Moamba also perform well overall on several subtasks. Magude has the lowest scores 
overall. Figure 4 below shows the same results, arranged to show overall EGRA scores by district (the subtask 
scores are still visible within the histogram columns). This allows the overall EGRA scores to be ranked by district: 
Boane performed best overall, followed by Moamba, Matutuine, Maracuenne, Matola, Manhiça, Namaacha, 
then Magude. This does not support the notion that the comparison group might be performing well overall 
because it includes the peri-urban districts of Matola and Marracuenne, which come fourth and fifth in the 
overall ranking. 
51
Figure 3: EGRA mean subtask scores by district
Figure 4: EGRA mean overall scores by district (subtask scores included)
The notion that the reason the FFE + literacy students did not out-perform the comparison students overall, 
because the comparison group performed better overall, due to extraneous variables, must then be rejected. 
How then to explain the fact that, despite having made progress through the literacy intervention, the FFE + 
intervention students are not overall performing significantly better than the comparison group? In the 
judgement of the external evaluator, this is likely to be due to two factors:
0
2
4
6
8
10
12
14
16
18
EGRA Mean Scores by District
Boane Magude Manhica Marracuene Matola Matutuine Moamba Namaacha
0
10
20
30
40
50
60
70
80
90
100
Boane Magude Manhica Marracuene Matola Matutuine Moamba Namaacha
Total EGRA Mean Scores by District
oral vocab oral comprehension phon awareness concepts print letter sounds read syllables
read words read fluency (wpm) compara fluency (wpm) read accuracy compara accuracy comprehension
write first name write family name write whole name total writing write words write phrase
52
• The progress which the literacy intervention is beginning to show is still at an early stage and needs to be 
consolidated. This is not surprising, given that this is only the third year the literacy intervention has been 
operating in schools and such programs take time to embed;
• It is impossible to know what the EGRA outcomes of the students in the FFE + literacy schools would have 
been if they had not benefited from the literacy intervention. It is possible – as strongly suggested by one 
of the project leaders - that without the project interventions, this group of students would have 
educationally fallen further behind the students from the more socioeconomically advanced districts in the 
comparison group. 
The final evaluation of FFE2 in 2020 should allow the initial achievements reported here to be tested after an 
additional year of the literacy intervention and cast more light on how the initial progress made translates into 
later literacy. In particular, for the students taught in local languages, the final evaluation will capture how 
literacy instruction begun in local languages then transitions into literacy in Portuguese, the national language 
of instruction. 
Improved Quality of Literacy Instruction (MGD 1.1)
The project ToC encapsulates the assumption that the quality of literacy instruction will be improved by a) better 
access to school supplies and materials and b) increased skills and knowledge of teachers. The project 
interventions which aim to address both of these dimensions are presented below. 
Access to School Supplies and Materials (MGD 1.1.2)
Indicator 2 (Output): Number of textbooks and other teaching and learning materials provided as a 
result of USDA assistance. Final target: 39,600
Midterm results (March 2019): 71,932
Final target 182% achieved
Number of schools receiving school supplies and materials as a result of USDA assistance. Final 
target: 264
Midterm target: 264 
Midterm results (March 2019): 271
Final target 103% achieved 
The project distributed textbooks and other instructional materials throughout phase one and continues to do 
so in phase two. These materials are presented as being intended to support extracurricular learning clubs
(although in reality schools have discretion over how they are used). Until recently, the project clubs component
organized annual training sessions in the use of these materials in the classroom for pedagogical directors and 
one additional teacher from each school. During these training sessions, participants were also consulted about 
53
the type of materials they would like to receive. In recognition of the limitations of this form of training, and that 
those teachers who received it were not necessarily “cascading” it down to their colleagues as anticipated, in
recent months the strategy has changed: training of teachers in the facilitation of learning clubs and the use of 
the materials supplied has been included in the in-service training for teachers offered by the project 
“professionals”. 
During phase two, analysis of the project monitoring records shows that targets for the distribution have been 
surpassed. Just over 80% of teachers in project schools reported their school had received materials to support
extracurricular learning clubs20. Of these:
• just over 30% said they had been consulted over the choice of materials;
• 93% of teachers in FFE + literacy schools and 100% in FFE only schools said they found the materials 
useful
• when asked whether they had used any of the material “today or yesterday”, 32% of teachers in FFE + 
literacy schools and 33% in FFE only schools said that they had; teachers of bilingual classes claimed to 
have do so slightly more often (38%, compared with 29% in Portuguese only schools);
• almost all (100% of FFE + literacy and 94% of FFE only) described the materials as either “good” or 
“excellent”.
When asked whether they now had access to sufficient teaching and learning resources, 52% of FFE + 
literacy teachers and 44% of FFE only teachers replied that they did. Teachers in bilingual schools were more 
likely to say they had access to sufficient resources (54%, compared with 45% of teachers in Portuguese 
only schools). This difference is probably a consequence of the FFE + literacy schools, to which most of the 
bilingual schools belong, having benefited from extra resources through the literacy program. The project 
should now ensure that all teachers are aware of the existence of the materials provided and encouraged 
or capacitated to use these. 
For more detail, see Technical Appendix 2, pp 28 – 40.
School directors expressed gratitude for the materials and claimed they made a significant difference to 
teachers’ ability to teach within their schools. One school director stated:
We received chalk, we received reams of paper, we received pens, pencils, colored pencils, which the 
children used in the clubs … this was linked to changes in their writing… One child was full of admiration 
20 During analysis of the midterm data collected from teachers, it became clear that there is confusion amongst 
teachers between these materials provided by the project to all participating schools and the materials 
developed by the literacy team and provided to schools taking part in the literacy program. This is discussed 
further in the presentation of support to teachers below.
54
to see the difference between what he wrote the first day and what he wrote more or less at the end [of 
the year]. We gave them to the children to motivate them and to show their families at home too. 
During school visits some of the materials, including syllable charts and other posters, were in evidence in 
some classrooms, suggesting they are indeed being used. In many cases, they can be seen in the directors’ 
offices: this is often because this is the most secure and clean place to store objects of value (it is to be 
hoped that they are in fact used for teaching and learning, and not merely kept safe). 
Lesson observations provided an opportunity to observe whether pupils had access to basic equipment such 
as textbooks, pens or pencils, exercise books and erasers. The results do not show a particular pattern of 
differences between FFE + literacy, FFE only and comparison group schools. During lesson observations, for 
the classes observed:
• 89% of FFE + literacy schools, 87% of FFE only schools and 81% comparison schools had a 
blackboard and chalk in acceptable condition;
• in 83% of FFE + literacy schools,84% of in FFE only schools and 79% comparison schools there were 
textbooks for all pupils in acceptable condition, compared with;
• in79% of FFE + literacy schools, 83% of FFE only schools and 81% of comparison schools, all pupils 
had either a pencil or a pen; 
• In 90% of FFE + literacy schools, 88% of FFE only schools and 97% of comparison schools, all pupils 
had an exercise book;
• In 27% of FFE + literacy schools, 17% of FFE only and 21% of comparison schools, all pupils in the 
class observed had an eraser.
More details can be found in Technical Appendix 4, pp 66-7521
.
The project has undoubtedly contributed to teachers and pupils having improved access to some teaching and 
learning materials. It is to be noted that they now appear to be at a similar level to pupils in the comparison 
schools, at least as far as the most basic classroom equipment and materials are concerned, as observed during 
the lesson observation sessions. It is not possible to know from the information gathered whether the project 
schools are actually better equipped than the comparison schools in terms of less basic teaching and learning 
materials, such as syllable charts and other books and equipment, although this is likely to be the case.
21 Technical Appendix 4 (classroom observation) is available in electronic form: pdf file entitled 
03_ADPE_TABLES_COBSERV_PDF__c
55
Number of awards given to students. Final target: 30,000
Midterm target: 19,500
Midterm results (March 2019): 17,552
Midterm target 90% achieved 
Number of awards given to teachers. Final target: 7,500
Midterm target: 4,875
Midterm results (March 2019): 7,845
Final target 105% achieved 
Providing awards to both pupil and teachers has been used since the beginning of the project as an incentive to 
encourage both to make efforts in teaching and learning. During the first phase of the project, awards were 
presented at high profile ceremonies and district-level meetings, involving large numbers of schools, known as 
the “Olympiads”. Despite the success of these large-scale events in the district capitals in raising the visibility of 
the project and promoting the importance of academic achievement, it was felt that local communities, parents 
and guardians were excluded from them and that they were also costly and time consuming to organize. 
Recently, the project has focused on more local events and competitions, which allow local schools and families 
to take part and are considered more sustainable. 
Rewards can vary from a pencil or a reading book to a bicycle in certain rare cases. Some of the awards 
encountered during fieldwork were poor quality plastic toys and school equipment, which seems regrettable, 
given current awareness of the environmental impact of limited use plastics and the potentially toxic 
composition of some low-grade plastics.
Analysis of the project monitoring records reveals that the final target for awards to teachers has been 
surpassed. The midterm target for awards to pupils has been 90% achieved. Of the wide-ranging subjects evoked 
by respondents during interviews and focus groups, awards were mentioned less frequently than during 
previous evaluations, suggesting that they are less of a focus than during the first phase of the project. However, 
they are still perceived by school directors and district education officials as being a valuable means of 
encouraging both teachers and students. 
Before distributing further awards, it would be useful to assess their usefulness in providing real incentives and 
also whether the types of awards currently distributed are the most appropriate for this purpose and reflect 
ADPP’s wider objectives of environmental protection and sustainability, or whether these might be replaced by 
more sustainable alternatives.
56
Literacy Instructional Materials (1.1.3)
Number of grade 1-3 children receiving literacy books. Final target: 22,300
Midterm target: 16,450
Midterm results (March 2019): 22,111
Midterm target 134% achieved 
Number of supplementary literacy materials produced and distributed to project schools. Final 
target:159,500
Midterm target: 81,950 
Midterm results (March 2019): 199,280
Final target 125% achieved 
Within the scope of the literacy intervention supported by CE, a range of literacy books and materials in local 
languages Xichangana and Xirhonga and in Portuguese has been produced by the project literacy team, in 
partnership with experts from INDE and with support from Cambridge’s reading experts. Targets for distribution 
of these have been surpassed. The production of the materials was an enormous undertaking, which mobilized 
the literacy team for the first months of the project. As part of this effort, the orthography of the two local 
languages was revised and standardized. The materials produced include textbooks, resource materials for 
teachers, reading books for students to take home and story books for teachers to read aloud to their class. The
midterm evaluation revealed widespread praise for these innovative materials, for the quality of their content 
and their culturally appropriate messages. They are recognized as filling a much-needed gap in the resources 
available to teachers. District education official and head teachers expressed great appreciation of these 
materials, as did teachers and reading coaches. Government officials express gratitude that the project and CE
have been able to provide these materials, which are currently beyond the scope of their stretched budgets. In 
particular, they are delighted that the project is promoting bilingual education, which has long been part of the
MINEDH’s plans but without resources being available to implement this. 
On examination, the books and other materials appear reasonably durable, without being excessively expensive. 
They are attractively presented and contain pictures and themes which reflect the daily lives of the pupils who 
will use them. There were a number of orthographic errors in the first printing which have been corrected in 
preparation for a second printing in the future. 
Skills and Knowledge of Teachers (1.1.4)
There are two major dimensions of the project interventions which target improving the skills and knowledge of 
teachers: i) in-service training (CPD) of primary school teachers in schools in the project districts as part of the 
literacy intervention coordinated and supported by CE and ii) the initial teacher education (pre-service teacher 
training) delivered by the 11 teacher training colleges run by ADPP in partnership with the MINEDH in each 
57
province of Mozambique. The skills and knowledge of primary school teachers in project schools are considered 
first. The skills and knowledge of trainee teachers in EPFs are addressed in the following section. 
In-service training in literacy of primary school teachers
Indicator 6 (Output): Number of teachers/educators/teaching assistants trained or certified 
(primary schools) as a result of USDA assistance. Final target: 879 
Midterm target: 586
Midterm results (March 2019): 557
Midterm target 95% achieved
Number of teachers who receive in-service training as a result of USDA assistance. Final target: 293
Midterm target: 293
Midterm results (March 2019): 557 
Final target 190% achieved
Indicator 5 (Outcome): Number of teachers/educators/teaching assistants in target schools who 
demonstrate use of new and quality teaching techniques or tools as a result of USDA assistance. 
Final target: 525 
Midterm target: 350
Midterm results (March 2019): 451 (estimate based on 93/240 classroom observations)
Final target 129%achieved
586 teachers in project schools have received training in phonics and their use of in teaching reading and writing. 
The project intends to train further teachers in 2020. This is a significant innovation for Mozambique, where 
phonics has not been commonly used. In complement to the training, 27 reading coaches have been recruited 
by the project to provide ongoing support in the classroom. The reading coaches are themselves experienced 
primary school teachers, some retired, some on career breaks from their teaching functions, who were identified 
through a competitive process as possessing the necessary skills and attributes for this role. They have received 
2.5 days’ training prior to commencing their role and ongoing training and support from Cambridge and the 
literacy team. The reading coaches were considered by most informants to bring considerable added value to 
the program. The reading coaches themselves reported observing positive changes within the classrooms they 
support. One reading coach explained:
These days we have pupils who are in the second or third grade who already read perfectly and write 
perfectly.…in Rhonga. We can see the difference through the assessments we do, because each term we 
do a mini-EGRA …, also the reports we receive at the district level in the schools where the reading 
coaches are working it’s different from the other schools, yes, the performance is better.
58
Informants regretted that there were not more reading coaches and supervisors since these are rather thinly 
spread, particularly in some districts; in particular more supervisors are needed in order to have more time in 
the field. 
The literacy team and reading coaches reported having a good working relationship with GoM officials at all 
levels; this was confirmed during interviews with national, provincial and district-level officials. One 
demonstration of such positive relationships is the fact that the District Education Office (SDEJT) focal points for 
the literacy program are invited to training sessions for reading coaches and teachers, where they participate as 
trainers. This is a rare case of good practice in working in partnership with government staff and building capacity 
within government institutions. The literacy team would like district education directors to also take part in 
these meetings, to build their awareness of pupil-centered, phonics-based reading methods, in the hope they 
will then serve as champions of these at strategic levels within the education system. 
The literacy component has an M&E plan and a full-time M&E officer who systematically collects monitoring 
information from EGRAs and mini-EGRAs regularly conducted by samples of the students in FFE + literacy schools 
and from observations of lessons delivered by the teachers receiving literacy coaching. Reports of these 
evaluations and lesson observations are produced regularly by the M&E officer who, despite being dedicated to 
the literacy project, is physically based in the M&E office which, at least in theory, allows integration with wider 
M&E activities. Given the gap in M&E management and provision over recent times, in reality she has frequently 
been left to work alone (this should change under the management of the newly appointed M&E coordinator). 
In contrast, the staff member responsible for M&E of the nutrition education project is currently based in the 
nutrition education team office, although the plan is to transition the full nutrition education M&E responsibility 
to the FFE2 M&E team, integrating the work and personnel22
. Whereas M&E of the nutrition education project 
is has been closely supported by implementing partner WISHH, the implementing partner of the literacy 
intervention, CE focusses on providing expert technical support but little M&E of the intervention. Providing 
some robust technical support to the literacy M&E officer would be useful, given her position between two 
operational units and relative lack of support recently, and allow the information in her reports to be more 
widely exploited for program improvements. 
85% of teachers surveyed declared they had received some form of training as part of the project. Overall, 94% 
of these considered the training to be either “good” or “excellent”. 72% of teachers from FFE + literacy schools 
and 64% from FFE only schools declared they had received visits to provide professional support visits whilst at 
22 WISHH representatives explain that, at the beginning of FFE2, the nutrition education component attempted 
to transfer responsibility for M&E of its activities to the FFE2 M&E team. However, given the lack of stability 
within the M&E team, WISHH maintained support until the M&E team had the capacity to take on this role, 
continuing to support M&E of nutrition education activities for longer than originally planned. The nutrition 
education sustainability plan specifically addresses integration of M&E into the wider project M&E processes: 
this has begun, under the new M&E coordinator and team.
59
school; 94% of these teachers considered the professional support to be either “good” or “excellent”. 40% of 
teachers in FFE + literacy schools and 12% in FFE only school claim to teach in Xichangana or Xirhonga.
For more details, refer to Technical Appendix 2, pp 40-49.
The survey of project teachers included several items designed to assess the knowledge of literacy techniques 
of those teachers benefiting from the literacy training. 
Of those teachers surveyed, 71% of those from FFE + literacy schools and 39% of those from FFE only schools 
stated they had received literacy training as part of the project. They overwhelmingly stated they had found the 
training useful (FFE + literacy 100%; FFE only 93%). It seems rather surprising that teachers from FFE only schools 
claimed to have received literacy training; it is possible these teachers were, in fact, referring to other forms of 
support such as training in the use of the materials provided for after-school clubs. 
Teachers’ knowledge of some basic concepts of literacy was evaluated by asking them to give definitions of five 
commonly-used terms using multiple choice questions. Table 7 below presents the results of this task for 
teachers in FFE + literacy schools and FFE only schools at midterm. The baseline results are also presented. 
However, at baseline, 65% of respondents did not reply to these questions, making comparisons problematic. 
At midterm, in 4 out of 5 cases, the FFE + literacy teachers performed better than the FFE only teachers, in some 
cases by a considerable margin. Only in the case of “comprehension” did the FFE only teachers do better. Overall, 
the FFE + literacy teachers have a better understanding of these terms. It is striking that even the FFE + literacy 
teachers did not perform particularly well on these tasks: in no case did more than 62%of either group choose 
the correct response. 
For more details, refer to Technical Appendix 2, pp 70-90.
Correct response Midterm FFE + literacy (% 
of correct responses)
Midterm FFE only (% of 
correct responses)
Baseline: all teachers (% of 
correct responses)
Phonological awareness 52% 32% 8% (23% of responses)
Phonetics 42% 23% 9% (26% of responses)
Fluency 62% 52% 30% (86% of responses)
Vocabulary 57% 52% 28% (80% of responses)
Comprehension 58% 62% 31% (89% of responses)
Table 7: teachers’ ability to define basic terms of literacy
When asked about the different stages of the instructional model “eu faço, nos fazemos, vocês fazem” (I do, we 
do, you do), taught to teachers as the basis of the literacy intervention, 52% of teachers from FFE + literacy 
60
schools were able to correctly define the stage “eu faço”, as opposed to 48% of those from FFE only schools
(compared with 16% overall at baseline). 70% of teachers from FFE + literacy schools were able to correctly 
define the stage “nos fazemos”, compared with 58% of those from FFE only schools (22% overall at baseline). 
62% of teachers from FFE + literacy schools were able to correctly define the stage “vocês fazem” compared 
with 53% of those from FFE only schools (23% overall at baseline) (see figure 5 below). 
For more details, refer to Technical Appendix 2, pp 90-97.
Figure 5: Teachers’ ability to define the stages of the Instructional Model at baseline and midterm
When asked about their experience of reading, 89% of teachers in FFE + literacy schools and 92% in FFE only 
schools claim to read every day to prepare their teaching (compared with 30% overall at baseline)whereas 9% 
of teachers in FFE + literacy schools and 8% in FFE only schools claim to read “about once a week” (3% overall at 
baseline) (see figure 6 below).49% of teachers in FFE + literacy schools and 61% in FFE only schools claim to read 
for pleasure every day (compared with 24% overall at baseline), with 42% of teachers in FFE + literacy schools 
and 31% in FFE only schools stating they do so “about once a week” (compared to 7% at baseline) (see figure 7
below). There appears to be a large increase in teachers stating they read to prepare their teaching and for 
pleasure between baseline and midterm.
23
23 It should be noted that i) only 32% of teachers surveyed responded to this question at baseline; ii) the survey 
was administered at midterm in the form of an interview, with the teachers’ responses captured by an 
enumerator on a smart phone, whereas at baseline teachers filled in a questionnaire themselves. It is probable 
that this difference influenced the teachers both to reply and to reply with the perceived correct or approved 
response at midterm. 
0%
10%
20%
30%
40%
50%
60%
70%
80%
Baseline Midterm: FFE + literacy Midterm: FFE only
Awareness of Stages of the Instructional Model
eu faço (I do) nos fazemos (we do) vocês fazem (you do)
61
Figure 6: Teachers’ reporting of reading for lesson preparation
Figure 7: Teachers’ reporting of reading for pleasure
When asked in which language they read for pleasure, Portuguese was the most frequently cited (80% FFE + 
literacy; 91% FFE only), followed by Xichangana (14% FFE + literacy; 4% FFE only) and Xirhonga (2% FFE + literacy; 
1% FFE only); 4% of each group stated they read for pleasure in other languages. Even allowing for a certain 
degree of probable overreporting, at midterm, teachers in FFE + literacy schools are significantly more likely to 
read for pleasure in local languages than those in FFE only schools. At baseline, the responses were Portuguese 
26%, Xichangana 4%, Xirhonga 1%, with 68% of nonresponses.
0%
10%
20%
30%
40%
50%
60%
70%
80%
90%
100%
Every day About 3 times
a week
About once a
week
About once a
month
Never No response
Frequency of Reading for Lesson Preparation
Baseline Midterm: FFE + literacy Midterm: FFE only
0%
10%
20%
30%
40%
50%
60%
70%
80%
Every day About 3
times a week
About once a
week
About once a
month
Never No response
Frequency of Reading for Pleasure
Baseline Midterm: FFE + literacy Midterm: FFE only
62
When asked to evaluate their own reading and comprehension ability in Portuguese, 64% of teachers in FFE + 
literacy schools and 71% in FFE only schools claim to be able to read fluently and understand everything, with 
35% of teachers in FFE + literacy schools and 27% in FFE only schools stating they can read fluently and have 
some difficulty understanding certain words. Only 1% of teachers overall claimed to have many difficulties 
understanding what they are reading and 0% stated that they read rarely and had many comprehension 
problems (see figure 8 below). When asked to evaluate their own reading and comprehension ability in 
Xichangana or Xirhonga, 29% of teachers in FFE + literacy schools and 26% teachers in FFE only schools claim to 
be able to read fluently and understand everything, with 48% of teachers in FFE + literacy schools and 38% in 
FFE only schools stating they can read fluently and have some difficult understanding certain words.13% of 
teachers in FFE + literacy schools and 22% of teachers in FFE only schools overall claimed to have many 
difficulties understanding what they are reading and 10% of teachers in FFE + literacy schools and 14% in FFE 
only stated that they read rarely and had many comprehension problems (see figure 9 below).
Figure 8: Teachers’ reporting of own reading ability in Portuguese
0%
10%
20%
30%
40%
50%
60%
70%
80%
Read fluently and
understand
everything
Read fluently
with some
difficulties
Read and have
difficulty
understanding
Read
rarely/never
Left blank (No
response)
Reading Ability in Portuguese
Baseline Midterm: FFE + literacy Midterm: FFE only
63
Figure 9: Teachers’ reporting of own reading ability in Xichangana/Xirhonga
When asked about a variety of literacy instruction materials distributed by the project to teachers, the following 
responses were obtained (see figures 10, 11 and 12 below):
• Materials to improve literacy: stated to have been received by 62% of teachers in FFE + literacy schools 
and 77% in FFE only schools; 91%of teachers in FFE + literacy schools and 96% of teachers in FFE only 
schools claim to have found them useful; 
• Framework to facilitate lesson planning: stated to have been received by 5% of teachers in FFE + literacy 
schools and 37% in FFE only schools; 95% of teachers in FFE + literacy schools and 100% of teachers in 
FFE only schools claim to have found them useful;
• Guide of methodological suggestions: 98% of teachers in FFE + literacy schools and 78% in FFE only 
schools claim to use this; 80% of teachers in FFE + literacy schools and 71% of teachers in FFE only 
schools claim to use it every day; 20% of teachers in FFE + literacy schools and 29% of teachers in FFE 
only schools claim to use it at least once a week;
• Letter- syllable- and word-cards: 88% of teachers in FFE + literacy schools and 78% in FFE only schools 
claim to use these in their teaching; 65% of teachers in FFE + literacy schools and 43% of teachers in 
FFE only schools claim to use them every day; 35% of teachers in FFE + literacy schools and 57% of 
teachers in FFE only schools claim to use them at least once a week;
• “Teacher read alouds” (stories for teachers to read aloud in class): 98% of teachers in FFE + literacy 
schools and 100% in FFE only schools claim to use these in their teaching; 73% of teachers in FFE + 
literacy schools and 67% of teachers in FFE only schools claim to read them aloud about 3 times a week; 
24% of teachers in FFE + literacy schools and 33% of teachers in FFE only schools claim to use them at 
least once a week;
0%
10%
20%
30%
40%
50%
60%
70%
80%
Read fluently and
understand
everything
Read fluently with
some difficulties
Read and have
difficulty
understanding
Read rarely/never Left blank (No
response)
Reading Ability in Xichangana/Xirhonga
Baseline Midterm: FFE + literacy Midterm: FFE only
64
• Reading books for pupils: 100% of teachers in FFE + literacy schools and 100% in FFE only schools claim 
to use these in their teaching; 88% of teachers in FFE + literacy schools and 100% of teachers in FFE 
only schools claim to use them every day; 12% of teachers in FFE + literacy schools claim to use them 
at least once a week.
For more details refer to Technical Appendix 2, pp 113- 139
Figure 10: Use of literacy materials provided by the project (FFE + literacy at midterm)
Figure 11: Use of literacy materials provided by the project (FFE only at midterm)
0%
10%
20%
30%
40%
50%
60%
70%
80%
90%
100%
Guide of methodological
suggestions
Letter, syllable and word
cards
Teacher read-alouds Reading books for pupils
Midterm FFE + Literacy: Use of Literacy Materials Provided by 
the Project
Every day About 3 times a week About once a week About once a month Never
0%
10%
20%
30%
40%
50%
60%
70%
80%
90%
100%
Guide of methodological
suggestions
Letter, syllable and word
cards
Teacher read-alouds Reading books for pupils
Midterm FFE only: Use of Literacy Materials Provided by the 
Project
Every day About 3 times a week About once a week About once a month Never
65
Figure 12: Use of literacy materials provided by the project (FFE + literacy and FFE only at baseline)
It should be noted that, in addition to the materials developed by the project literacy team and supplied to 
schools taking part in the literacy program, all project schools receive the teaching and learning materials 
supplied by the project to support extracurricular clubs. It is probable that, in responding to questions about 
materials received, teachers did not differentiate between these two types of materials; indeed, teachers 
in FFE only schools are likely to have assumed they were being questioned about the only materials they 
received, those designed to support the extracurricular learning clubs. With hindsight, the teacher 
questionnaire should have been designed so as to avoid this confusion (for example, by presenting images 
of the materials referred to in each question).
Although these responses are encouraging and suggest teachers are making good use of the materials
provided, it is not possible to know for sure whether their responses referred to the materials provided by 
the literacy intervention or the general materials supplied as support to learning clubs. 
Teachers were asked how much individual reading time they allowed their pupils in class; they were 
requested to select the response which best corresponded out of four options. Figure 13 shows their 
responses, compared with those from the baseline survey (relatively few teachers responded to this 
question at baseline). 
0%
10%
20%
30%
40%
50%
60%
70%
80%
90%
100%
Guide of methodological
suggestions
Letter, syllable and word
cards
Teacher read-alouds Reading books for pupils
Baseline: Use of Literacy Materials Provided by the Project
Every day About 3 times a week About once a week
About once a month Never No response
66
Figure 13: time pupils spend reading individually in class at baseline and midterm
No particular tendency appears to emerge from these results, except for the fact that, at midterm, “15 minutes 
per day or more” and “15 minutes per week or more” are by far the most frequently selected options (the range 
of options available may be considered not to constitute a clear continuum from more to less and may have 
confused some respondents). 
As at baseline, teachers were presented with a series of binary options, designed to test their familiarity with 
modern classroom practice and approaches to reading24
. Their responses are presented in table 8, along with 
those from baseline (when, unfortunately, there was a high percentage of non-responses). 
Table 8 below presents teachers’ responses at midterm and baseline. The options which are compatible with 
modern classroom practice and approaches (so are considered as “correct” for the present purposes) are 
shaded. The responses do not reveal consistent changes, either between baseline and midterm or between the 
teachers in FFE + literacy and FFE only schools. It is interesting that there is little change since baseline overall. 
The majority of teachers chose the “correct” response (modern approach) in five of the seven cases. However, 
they remain fairly evenly split on the questions of whether or not grade one pupils are too young to be 
responsible for handing out books and whether young pupils learn to read best in front of the blackboard or 
24 These were adapted from the Learning Gains Study used as part of the EGRA developed in Malawi by USAID 
Malawi and RTI, 2016.
0%
10%
20%
30%
40%
50%
60%
70%
80%
5 minutes per
day
15 minutes per
day
5 minutes per
week
15 minutes per
week
No response
Time Pupils Spend Reading Individually
Baseline Midterm: FFE + literacy Midterm: FFE only
67
when holding a book and at their own pace. It is notable that these “traditional” attitudes, which suggest a lack 
of exposure to modern pedagogical theory and practice, have not changed amongst the FFE + literacy teachers25
. 
For more details refer to Technical Appendix 2, pp 141- 157.
Answer A Midter
m FFE + 
literacy
Midte
rm
FFE 
only
Baseline
%
Answer B Midter
m FFE + 
literacy
Midte
rm
FFE 
only
Baseline
%
N/R at 
baseli
ne%
It is important to allow 
pupils to take their 
books home
(baseline)/use books 
regularly(midterm)
26
92% 92% 23% (77% of 
responses)
Pupils should not take 
schoolbooks home
(baseline)/ books should 
be used occasionally s as 
to preserve them for the 
future (midterm). 
8% 8% 7% (23% 
of 
respons
es)
70%
Pupils cannot practice 
reading at home 
because the majority of 
parents cannot read
15% 14% 2% (6% of 
responses)
Pupils should practice 
reading at home, even if 
their parents cannot read
85% 86% 29%
(94% of 
respons
es)
69%
Grade one pupils are 
too young to be 
responsible for handing 
out books
47% 43% 12% (46% of 
responses)
Grade one pupils can 
learn to hand out books
53% 57% 14%
(54% of 
respons
es)
74%
Pupils should be seated 
in rows facing the front 
so they can take part in 
the class
26% 32% 7% (24% of 
responses)
The way pupils are seated 
should be adapted to suit 
different activities 
74% 68% 22%
(76% of 
respons
es)
71%
Preparing classes in 
advance is necessary 
for good classroom 
management 
97% 96% 27% (93% of 
responses)
Preparing classes in 
advance does not help 
much with classroom 
management 
3% 4% 2% (7% 
of 
respons
es)
71%
Young pupils learn best
when reading is done 
from the blackboard 
56% 44% 12% (41% of 
responses) 
Young pupils learn to 
read best when they are 
holding a book and can 
learn at their own pace
44% 56% 17%
(59% of 
respons
es)
71%
Independent practice is 
not useful for young 
pupils because they 
make mistakes
25% 18% 3% (11% of 
responses) 
Being able to make 
mistakes during 
independent practice is 
an important part of 
learning 
75% 82% 25%
(89% of 
respons
es)
72%
Table 8: Familiarity with modern classroom practice and approaches to reading: teachers’ responses
Teachers were asked how they monitor pupils’ progress in reading and how they record this, through a series of 
multiple-choice options; they were asked to choose all which applied. Their responses are presented in figures
14 and 15 respectively, along with those from baseline (when, unfortunately, there was a high percentage of 
non-responses). 
25 As pointed out in the baseline report, when questions are framed with focus on attitudes rather than actual 
practice, teachers may select what they perceive to be the desired response, rather than one which reflects their 
actual practice.
26 The formulation of this item (both A and B) was changed at midterm, at the request of the literacy team).
68
For more details refer to Technical Appendix 2, pp 386-394.
Figure 14: How teachers evaluate students’ progress in reading – midterm and baseline
Regarding their monitoring of students’ reading, teachers’ responses at midterm were fairly similar over the two 
groups, with the largest number claiming they listen to pupils read individually to help them and to monitor their 
progress, the most modern, “correct” response, which has been the method promoted by reading coaches and 
fewer claiming to use a test, the more “traditional” and less effective method. It is striking that greater difference 
was not found between the two groups. Comparisons with the baseline results are not meaningful: at baseline, 
once again, a large number of teachers (69%) did not respond. Furthermore, at baseline, teachers were asked 
to choose only one response.
For more details refer to Technical Appendix 2, pp 394-401.
0%
10%
20%
30%
40%
50%
60%
70%
80%
Test at end of
each trimester
Listen to pupils
read individually
Pupils read one
by one in class
Only evaluate
those who can
read
No response
How Teachers Evaluate Pupils' Reading Progress
Baseline Midterm: FFE + literacy Midterm: FFE only
69
Figure 15: How teachers record students’ progress in reading – midterm and baseline
Regarding their recording of students’ reading progress, significantly more FFE + literacy group teachers claimed 
to record the stages of reading attained by individual pupils on an ongoing basis, the “correct” method promoted 
by the literacy coaches. They were also less likely to claim to correct exercises books during the lesson (since the 
question was about monitoring reading this response suggests a certain confusion amongst teachers who chose 
this response) or to fill in the record book at the end of the year27
. Taken together, these results suggest that 
the literacy coaching has resulted in increased knowledge of effective recording methods amongst participating 
teachers. Once again, comparisons with the baseline results are not meaningful, for the reasons cited above.
Lessons were observed in 240 project schools and comparison schools in order to attempt to evaluate whether 
teachers in project schools were using the “new and quality teaching techniques or tools” as a result of the 
literacy training. The observation instrument was developed by Simone Doctors with support from the project 
literacy team and is designed to be used by non-specialist observers, who are not teacher educators. All lesson 
observation was performed by supervisors, who had been trained in this activity and selected for their capacity 
to perform it. Several of the supervisors were themselves trained teachers. During fieldwork in FFE + literacy 
schools, whenever possible a class taught by a teacher who had benefited from the literacy intervention was 
observed28
. The results of the lesson observation are presented below following the structure and order of the 
27Given that teachers were asked to select all which apply, the interaction between these choices is not clear 
(filling in a record book at the end of the school year is a very inadequate from of monitoring reading progress; 
however, it would be a useful complement to recording ongoing progresses and stages attained). With hindsight, 
it would be more informative to ask teachers to select only one response.
28 Enumerators were given instructions as to which teachers to observe in project and comparison schools, 
according to a structured protocol. Where possible, they were told to observe a teacher trained by the project 
0%
10%
20%
30%
40%
50%
60%
70%
80%
Record the
stages achieved
in record book
Give marks at
end of trimester
in record book
Correct exercise
books during
classes
Complete
register at the
end of year
No response
How Teachers Record Pupils' Reading Progress
Baseline Midterm: FFE + literacy Midterm: FFE only
70
items on the observation instrument (although the data was captured using ODK on smartphones, observers 
also had paper copies of the observation instruments, to facilitate their navigation though the instrument). 
Elements verified before the lesson began
• The classroom walls were decorated with T&L materials in the case of 75% of FFE + literacy teachers, 
56% of FFE only teachers and 60% of comparison school teachers gave reading exercises during the 
lesson; this was the case for 74% of teachers teaching in local languages and 61% of teachers using 
Portuguese;
• Pupils’ work was displayed on the walls in the case of 44% of FFE + literacy teachers, 26% of FFE only 
teachers and 23% of comparison school teachers; this was the case for 44% of teachers teaching in local
languages and 28% of teachers using Portuguese; 
• 96% of FFE + literacy teachers, 94% of FFE only teachers and 94% of comparison school teachers had a
written lesson plan; this was the case for 89% of teachers teaching in local languages and 96% of 
teachers using Portuguese; 
• 73% of FFE + literacy teachers, 75% of FFE only teachers and 67% of comparison school teachers 
explained their lesson plan to the enumerator; this was the case for 79% of teachers teaching in local 
languages and 70% of teachers using Portuguese; 
• 78% of FFE + literacy teachers, 78% of FFE only teachers and 73% of comparison school teachers had 
evidence of their continuous assessment of students (in the form of record books, assessment sheets, 
etc.) or other; this was the case for 88% of teachers teaching in local languages and 72% of teachers 
using Portuguese; 
• 96% of FFE + literacy teachers, 96% of FFE only teachers and 91% of comparison school teachers arrived 
in the classroom on time; this was the case for 98% of teachers teaching in local languages and 93% of 
teachers using Portuguese; 
Conditions in the classroom
• The classroom was “perfectly clean” in the case of 49% of FFE + literacy teachers, 52% of FFE only 
teachers and 40% of comparison school teachers; this was the case for 45% of teachers teaching in local 
languages and 48% of teachers using Portuguese; The classroom was “more or less clean but could have 
been better” in the case of 37% of FFE + literacy teachers, 39% of FFE only teachers and 47% of 
comparison school teachers; this was the case for 39% of teachers teaching in local languages and 41% 
of teachers using Portuguese; The classroom was “not clean” in the case of 14% of FFE + literacy 
teachers, 9% of FFE only teachers and 13% of comparison school teachers; this was the case for 15% of 
teachers teaching in local languages and 11% of teachers using Portuguese;
teaching in local languages; where this was not possible, they should observe a teacher trained by the project 
teaching in Portuguese. There were a series of further instructions in case neither option was possible.
71
• The desks were arranged in such a way that all pupils could participate in the lesson in the case of 80%
of FFE + literacy teachers, 73% of FFE only teachers and 77% of comparison school teachers; this was 
the case for 83% of teachers teaching in local languages and 74% of teachers using Portuguese;
• Pupils with SEN were placed appropriately in the case of 44% of FFE + literacy teachers, 32% of FFE only 
teachers and 53% of comparison school teachers; this was the case for 50% of teachers teaching in local 
languages and 40% of teachers using Portuguese;
During the lesson
• The lesson had an introduction, development and a conclusion in the case of 94% of FFE + literacy 
teachers, 96% of FFE only teachers and 97% of comparison school teachers; this was the case for 95% 
of both teachers teaching in local languages and teachers using Portuguese;
• The teacher corrected the students’ homework in the case of 76% of FFE + literacy teachers, 73% of FFE 
only teachers and 80% of comparison schoolteachers; this was the case for 82% of teachers teaching in 
local languages and 74% of teachers using Portuguese;
• The teacher was considered by the observer to use the instructional model “I do, we do, you do” in the 
case of 91% of FFE + literacy teachers, 95% of FFE only teachers and 91% of comparison school teachers; 
this was the case for 92% of teachers teaching in local languages and 93% of teachers using 
Portuguese;
29
• The teacher was considered by the observers to demonstrate knowledge and mastery of the content 
being taught in the case of 99% of FFE + literacy teachers, 99% of FFE only teachers and 99% of 
comparison school teachers; this was the case for 100% of teachers teaching in local languages and 98% 
of teachers using Portuguese;
• 83% of FFE + literacy teachers, 78% of FFE only teachers and 30% of comparison school teachers gave 
oral exercises during the lesson; this was the case for 85% of teachers teaching in local languages and 
59% of teachers using Portuguese;
• 77% of FFE + literacy teachers, 81% of FFE only teachers and 29% of comparison school teachers gave 
reading exercises during the lesson; this was the case for 67% of teachers teaching in local languages 
and 63% of teachers using Portuguese;
29 During supervisor training in administering the classroom observation, the instructional model was explained 
to the supervisors, and they were told criteria to look out for in their observations. However, they are not 
education specialists (although many are trained teachers) and had not had formal, sustained training in this 
instructional model. It is therefore possible that they applied the criteria quite loosely. The fact that over 90% 
all three groups of teachers were categorised as using the model, whereas in principle only those in the FFE + 
literacy group would have had direct access to it, suggests this. This highlights a limitation of observation 
checklists where the observer is asked to choose between two binary responses, with no possibility of a nuanced 
response.
72
• 75% of FFE + literacy teachers, 75% of FFE only teachers and 94% of comparison school teachers gave 
written exercises during the lesson; this was the case for 73% of teachers teaching in local languages 
and 84% of teachers using Portuguese;
• The teacher monitored the students’ activities to verify the lesson objective were being met in the case 
of 99% of FFE + literacy teachers, 96% of FFE only teachers and 76% of comparison school teachers; this 
was the case for 98% of teachers teaching in local languages and 92% of teachers using Portuguese;
• The majority of pupils (80%) participated actively in the lesson in the case of 89% of FFE + literacy 
teachers, 79% of FFE only teachers and 96% of comparison school teachers; this was the case for 88% 
of teachers teaching in local languages and 88% of teachers using Portuguese;
• The majority of pupils (80%) demonstrated they had understood what to do in the case of 92% of FFE 
+ literacy teachers, 84% of FFE only teachers and 89% of comparison school teachers; this was the case 
for 92% of teachers teaching in local languages and 87% of teachers using Portuguese;
Communication and language 
• The teacher showed mastery of Portuguese (when spoken) in the case of 99% of FFE + literacy teachers, 
99% of FFE only teachers and 100% of comparison school teachers; this was the case for 98% of teachers 
teaching in local languages and 99% of teachers using Portuguese;
• The teacher showed mastery of Xichangana (when spoken) in the case of 96% of FFE + literacy teachers, 
96% of FFE only teachers and 88% of comparison school teachers; this was the case for 97% of teachers 
teaching in local languages and 94% of teachers using Portuguese;
• The teacher showed mastery of Xirhonga (when spoken) in the case of 100% of FFE + literacy teachers, 
100% of FFE only teachers and 89% of comparison school teachers; this was the case for 93% of teachers 
teaching in local languages and 100% of teachers using Portuguese;
• The teachers’ language was appropriate for children, clear and precise, audible and friendly in the case 
of 96% of FFE + literacy teachers, 99% of FFE only teachers and 93% of comparison school teachers; this 
was the case for 98% of teachers teaching in local languages and 95% of teachers using Portuguese;
• The teachers refrained from using humiliating or abusive language in the classroom in the case of 97%
of FFE + literacy teachers, 97% of FFE only teachers and 90% of comparison school teachers; this was 
the case for 100% of teachers teaching in local languages and 93% of teachers using Portuguese;
• The teacher took account of students with SEN when communicating in the case of 54% of FFE + literacy 
teachers, 47% of FFE only teachers and 50% of comparison school teachers; this was the case for 59% 
of teachers teaching in local languages and 47% of teachers using Portuguese;
Use of teaching and learning materials and resources
Some of the information about teaching and learning materials is reported in the section on Access to School 
Supplies and Materials above.
73
• 96% of FFE + literacy teachers, 87% of FFE only teachers and 87% of comparison school teachers used 
appropriate T&L materials to support their teaching; 98% of teachers teaching in local languages did 
so, compared with 87% of teachers teaching in Portuguese;
• 98% of FFE + literacy teachers, 97% of FFE only teachers and 99% of comparison school teachers wrote 
on the blackboard neatly and without errors; 97% of teachers teaching in local languages did so, 
compared with 98% of teachers teaching in Portuguese; 
• 61% of FFE + literacy teachers, 58% of FFE only teachers and 56% of comparison school teachers used 
materials which were suitable for learnings with SEN; 68% of teachers teaching in local languages did 
so, compared with 55% of teachers teaching in Portuguese; 
Classroom management 
• 96% of FFE + literacy teachers, 92% of FFE only teachers and 89% of comparison school teachers 
maintained discipline while teaching without resorting to corporal punishment; 95% of teachers 
teaching in local languages did so, compared with 91% of teachers teaching in Portuguese; 
• In the classes of 92% of FFE + literacy teachers, 75% of FFE only teachers and 93% of comparison school 
the majority of pupils (at least 80%)participated and was involved in the lesson; this applied in the case 
of 89% of teachers teaching in local languages did so, compared with 86%of teachers teaching in 
Portuguese; 
• Classroom rules were in evidence in classes taught by 90% of FFE + literacy teachers, 90% of FFE only 
teachers and 87% of comparison schools; this was the case in classes taught by 91% of teachers teaching 
in local languages did so, compared with 89% of teachers teaching in Portuguese; 
• 92% of FFE + literacy teachers, 97% of FFE only teachers and 96% of comparison school teachers were 
perceived to notice and respond to the learning needs of students with SEN; this was the case for both 
95% of teachers teaching in local languages and in Portuguese;
• There was a balance between teacher talking time/presentation and student talking time/activities in 
94% of FFE + literacy teachers, 96% of FFE only teachers and 91% of comparison school teachers were 
perceived to notice and respond to the learning needs of students with SEN; this was the case for 95% 
of teachers teaching in local languages and 93% of teachers teaching in Portuguese
Group work
• 42% of FFE + literacy teachers, 39% of FFE only teachers and 46% of comparison school teachers got 
students to work in groups; this was the case for 44% of teachers teaching in local languages and 41% 
of teachers using Portuguese;
• Of these, 90% of FFE + literacy teachers, 77% of FFE only teachers and 88% of comparison school 
teachers had students well positioned for group work; this was the case for 90% of teachers teaching 
in local languages and 83% of teachers using Portuguese;
74
• When used, teachers managed to monitor students’ participation in group work in the case of 97% of 
FFE + literacy teachers, 97% of FFE only teachers and 100% of comparison school teachers ;this was the 
case for 100% of teachers teaching in local languages and 97% of teachers using Portuguese;
• When used, groups were well balanced in terms of gender, disability and SEN in the case of 87% of FFE 
+ literacy teachers, 90% of FFE only teachers and 78% of comparison school teachers had students well 
position for group work; this was the case for 83% of teachers teaching in local languages and 86% of 
teachers using Portuguese;
Questions and feedback
• 89% of FFE + literacy teachers, 89% of FFE only teachers and 74% of comparison school teachers 
involved both boys and girls and gave them equal opportunities to participate; this was the case for 
86% of teachers teaching in local languages and 83% of teachers using Portuguese;
• 87% of FFE + literacy teachers, 74% of FFE only teachers and 90% of comparison school teachers 
responded to students’ questions; this was the case for 89% of teachers teaching in local languages and 
81% of teachers using Portuguese;
• 82% of FFE + literacy teachers, 65% of FFE only teachers and 83% of comparison school teachers 
allowed pupils to express their own ideas in the classroom; this was the case for 80% of teachers 
teaching in local languages and 75% of teachers using Portuguese;
• 97% of FFE + literacy teachers, 96% of FFE only teachers and 99% of comparison school teachers gave 
appropriate, constructive feedback to girls and boys; this was the case for 97% of teachers teaching in 
both local languages and Portuguese;
• 98% of FFE + literacy teachers, 97% of FFE only teachers and 96% of comparison school teachers 
corrected exercises completed during the lesson; this was the case for 95% of teachers teaching in local 
languages and 98% of teachers using Portuguese;
Equality, equity and Special Education Needs
• 42% of FFE + literacy teachers, 38% of FFE only teachers and 59% of comparison school teachers 
adapted activities so as to be suitable for students with SEN; this was the case for 52% of teachers 
teaching in local languages and 43% of teachers using Portuguese;
• Students with SEN are able to participate actively in the case of 48% of FFE + literacy teachers, 42% of 
FFE only teachers and 50% of comparison school teachers; this was the case for 56% of teachers 
teaching in local languages and 43% of teachers using Portuguese;
• Students with SEN (including specially gifted pupils) are given differentiated tasks by 42% of FFE + 
literacy teachers, 31% of FFE only teachers and 39% of comparison school teachers; this was the case 
for 48% of teachers teaching in local languages and 33% of teachers using Portuguese;
75
• 96% of FFE + literacy teachers, 94% of FFE only teachers and 91% of comparison school teachers
encourage girls and boys to participate equally; this was the case for 94% of teachers both teaching in 
local languages and using Portuguese;
• 91% of FFE + literacy teachers, 92% of FFE only teachers and 90% of comparison school teachers treat 
all students (boys, girls, with/without disabilities, of all linguistic groups) equally; this was the case for 
94% of teachers teaching in local languages and 90% of teachers using Portuguese;
Conclusion and homework
• 78% of FFE + literacy teachers, 77% of FFE only teachers and 89% of comparison school teachers set 
homework involving reading and/or writing; this was the case for 82% of teachers teaching in local
languages and 80% of teachers using Portuguese;
• 99% of FFE + literacy teachers, 94% of FFE only teachers and 89% of comparison school teachers 
finished the class on time; this was the case for 100% of teachers teaching in local languages and 92% 
of teachers using Portuguese.
76
Observation criteria Teacher group Language of instruction
FFE + 
literacy
FFE only Comparison Local 
languages
Portuguese 
Before the lesson % % % % %
Walls decorated with T&L materials 75 56 60 74 61
Pupils work displayed on walls 44 26 23 44 28
Written lesson plan 96 94 94 89 96
Explained lesson plan orally 73 75 67 79 70
Evidence of continuous assessment (record books, sheets, etc) 78 78 73 88 72
Teacher arrived in classroom on time 96 96 91 98 93
Conditions in the classroom 
Classroom “perfectly clean” 49 52 40 45 48
Classroom More or less clean – could be better 37 39 47 39 41
Classroom not clean 14 9 13 15 11
Desks arranged so all pupils can participate 80 73 77 83 74
Pupils with SEN placed appropriately 44 32 53 50 40
During the lesson
Lesson had introduction, development and conclusion 94 96 97 95 95
Teacher corrected homework 76 73 80 82 74
Teacher used instructional model “I do, we do, you do” 91 95 91 92 93
Teacher showed knowledge and mastery of lesson content 99 99 99 100 98
Teacher gave students oral exercises 83 78 30 85 59
Teacher gave students reading exercises 77 81 29 67 63
Teacher gave students written exercises 75 75 94 73 84
Teacher monitored activities to see objectives being met 99 96 76 98 92
Majority of pupils (80%) participated actively in lesson 89 79 96 88 88
Majority of pupils (80%) knew what to do 92 84 89 92 87
Communication and language 
Teacher showed mastery of Portuguese (when spoken) 99 99 100 98 99
Teacher showed mastery of Xichangana (when spoken) 96 96 88 97 94
Teacher showed mastery of Xirhonga (when spoken) 100 100 89 93 100
Teachers’ language appropriate for children 96 99 93 98 95
Teacher avoided humiliating or abusive language 97 97 90 100 93
Teacher took account of SEN students when communicating 54 47 50 59 47
Use of T&L materials and resources
Teacher used appropriate T&L materials to support topic 96 87 87 98 87
Teacher wrote on blackboard neatly without errors 98 97 99 97 98
Teacher used materials suitable for learners with SEN 61 58 56 68 55
Classroom management 
Teacher maintained discipline without corporal punishment 96 92 89 95 91
Majority of pupils (80%) participated and involved in lessons 92 75 93 89 86
Classroom rules in evidence 90 90 87 91 89
Teacher perceived and responded to needs of SEN students 92 97 96 95 95
Balance of teacher presentation and student activities 94 96 91 95 93
Group work
Teacher got students to work in groups 42 39 46 44 41
When used, students well positioned for group work 90 77 88 90 83
If used, teacher monitored student participation in groups 97 97 100 100 97
When used, groups well balanced (gender, disability, SEN) 87 90 79 83 86
Questioning and feedback
Girls and boys involved - equal opportunities to participate 89 89 74 86 83
Teacher responded to students’ questions 87 74 90 89 81
Teacher allowed students to express own ideas in classroom 82 65 83 80 75
Appropriate, constructive feedback given to girls and boys 97 96 99 97 97
Teacher corrected exercises completed during the lesson 98 97 96 95 98
Equality, equity and special needs
Teacher adapted activities for students with SEN 42 38 59 52 43
Students with SEN able to participate actively 48 42 50 52 43
Students with SEN given appropriately differentiated tasks 42 31 39 48 33
Teacher encouraged girls and boys to participate equally 96 94 91 94 94
Teacher treated all students equally (girls, boys, disabled, SEN) 91 92 90 94 90
Conclusion and homework
Teacher set homework involving reading and/or writing 78 77 89 82 80
Lesson finished on time 99 94 89 100 92
Table 9: Summary of classroom observation findings
77
Table 9 (above) presents a summary of these results30. The 49 observation criteria observed, grouped in 10 
sections, are considered to cover various aspects, although not necessarily an exhaustive description, of “good 
practice” in primary classroom teaching and to reflect a significant number of the “improved methods and tools” 
which the literacy training seeks to transmit. Analysis of the 49 observation criteria reveals certain trends:
• FFE + literacy teachers were more likely to exhibit these characteristics than FFE only teachers; a 
higher percentage of FFE + literacy teachers than of FFE only teachers demonstrated these 
characteristics in 57% of cases; a lower percentage of FFE + literacy teachers than of FFE only teachers 
demonstrated them in 20% of cases; the two groups performed equally in 22% of cases31
. 
• FFE + literacy teachers were more likely to exhibit these characteristics than comparison group 
teachers; a higher percentage of FFE + literacy teachers than of comparison group teachers 
demonstrated them in 61% of cases; a lower percentage of FFE + literacy teachers than of comparison 
group teachers demonstrated them in 35% of cases; the two groups performed equally in 4% of cases.
• FFE only and comparison group teachers were almost equally likely to exhibit these characteristics; 
FFE only teachers scored higher than comparison teachers in 49% of cases, whereas comparison group 
scored higher than the FFE only group in 45% of cases, with equal numbers in 6% of cases;
• Teachers observed teaching in Xichangana and Xirhonga were far more likely to demonstrate the 
characteristics than those teaching in Portuguese: a higher percentage of teachers using local 
languages demonstrated the characteristics than did the teachers teaching in Portuguese in 71% of 
cases; in 18% of cases, a higher percentage of teachers teaching in Portuguese demonstrated the 
characteristics; in 10% of cases they scored equally;
• Scores overall are surprisingly high (many over 90%). This suggests that the observers using the 
observation instrument applied it rather generously and tended to err on the side of positive 
assessments32; 
• The notable exception to the overall high scores concerns all criteria pertaining to students with 
special educational needs (SEN), which typically have scores closer to 50% for all three groups. This 
30 The greyed lines are not included in the total. The item regarding cleanliness of the classroom had a three￾option response: only the first option (classroom “perfectly clean”) is included in the summary of good practice 
used for the present analysis. 
31 This descriptive method of presenting the findings is not based on a formal difference-in-difference analysis 
and does not claim to be statistically significant. 
32 This is probably due to the fact that the observers had limited experience of what good teaching using modern 
teaching methods looks like. It may also reflect, in the case of those who are trained teachers, a reluctance to 
be critical of ones peers. The author’s experience of observing classroom teaching in Mozambique does not on 
the whole support these optimistic assessments. However, it is hoped that, although the absolute assessment 
may be overstated, the comparative assessment of the difference groups still has some validity, since all groups 
were observed according to the same criteria and strict quality controls were applied. 
78
suggests that awareness and practice of inclusivity are still not widely developed throughout the three 
groups of teachers observed. 
Analysis of project records reveals that, as of the time of fieldwork in April 2019, 557 primary teachers had been 
trained by the project (of a final total objective of 879). Of the 240 lessons observed during lesson observations,
93 of these were taught by teachers in the FFE + literacy group. The scores obtained by these teachers on the 
49 criteria assessed are presented in table 10 below, grouped according to the thematic sections used in the 
classroom observation instrument, and as a total overall score. It should be stressed that these scores are 
obtained by summing the number of teachers who were judged to fulfil the criteria, in most cases using a single 
“yes/no” question. Table 10 presents the total scores against the total possible scores for each section. FFE + 
literacy teachers observed scored most highly overall on classroom management (93%) and questioning and 
feedback (91%) and least highly on conditions in the classroom (58%) and equality, equity and special needs 
(64%). Overall the 93 FFE + literacy teachers observed scored a total of 3,471 points out of a possible total of 
4,259 (81%). Considering that the 93 FFE + literacy teachers observed were a sample of the 557 teachers trained 
by the project to date, if this score of 81% is extrapolated the total group of 557 teachers trained to date, this 
would suggest that, if observed, 451 of these teachers would demonstrate the desired “new and quality teaching 
techniques or tools” referenced in MGD indicator 5. This estimation is therefore suggested as the midterm
outcome for this indicator. 
Observation criteria Total score 
of FFE + 
literacy 
teachers
Potential 
total score
%
Before the lesson 428 558 77%
Conditions in the classroom 161 279 58%
During the lesson 815 930 88%
Communication and language 369 422 87%
Use of T&L materials and resources 237 279 85%
Classroom management 432 465 93%
Group work 146 210 70%
Questioning and feedback 421 465 91%
Equality, equity and special needs 297 465 64%
Conclusion and homework 165 186 89%
Total score of FFE + literacy teachers overall 3,471 4,259 81%
 Table 10 scores of FFE + literacy teachers based on classroom observation criteria
79
Initial teacher education in teacher training colleges (EPFs)
Indicator 6 (Output): Number of teachers/educators/teaching assistants trained or certified (EPFs)
as a result of USDA assistance. Final target: 3,704
Midterm results (March 2019): 6,059
Final target 164% achieved
Percentage of a student-teachers who improve their Portuguese language literacy skills during pre￾service training. Final target: 80%
Midterm target (for year 2018): 80%
Midterm results (March 2019): 84%33
Midterm target 105% achieved
Number of teachers in training as a result of USDA assistance. Final target: 3,512
Midterm target:3,386 (accumulated)
Midterm results (March 2019): 6,895
Final target 204% achieved
Previous evaluations of the FFE project have underlined the important numerical and quality contribution the 
eleven ADPP-run teacher training colleges (EPFs) are making to Mozambique’s teacher pool. At the time of the 
midterm evaluation, 6,059 trained teachers had graduated during the life span of the current phase of the 
project, exceeding the target by a considerable margin. The EPF model and curriculum are designed to train
teachers able and motivated to work and teach in remote rural primary schools, to be enterprising, creative and 
adaptable, including finding solutions to teach in low resource environments. In addition to preparing future 
teachers for the reality of teaching in remote rural areas, the EPF program aims to teach practical skills, such as 
community development, building, small scale agriculture and husbandry. The midterm evaluation of FFE 
revealed that, although all EPFs are now required to teach the entirety of the national teacher training 
curriculum, they are managing to maintain their commitment to providing future teachers with the skills they 
need to function effectively in challenging environments. This is achieved thanks to an intensive program of 
additional input, their specific culture within which trainee teachers take responsibility for many of the practical 
dimensions of the EPF management, such as menu planning and routine maintenance, and their boarding 
system, whereby student teachers are present in the EPF during evenings and some weekends.
The Baseline report recommended that M&E of EPF activities should be more integrated in the overall M&E 
processes of the project. Given the low level of Portuguese language skills, including handwriting and spelling, 
of some trainee teachers on arrival in the EPF, remedial Portuguese support is provided in all the EPFs. Previous 
evaluations of the first phase of the FFE program recommended the efficacy of this remedial intervention should 
be consistently monitored, with the EPFs gathering data on the added value provided by the EPF remedial 
33 Based on data from 6 EPFs
80
Portuguese program, using a standardized diagnostic instrument in all EPFs as a pre- and post-test. The 
instrument had been developed by Portuguese language trainers from several of the EPFs, with support from 
two experts in evaluation and Portuguese language assessment respectively. This information was collected at 
the beginning and the end of the 2018 training year in six of the 11 EPFs: EPF Gaza, Inhambane, Nhamatanda, 
Tete, Nametil and Nacala. The tests were not conducted at the three EPFs operating the three-year program, 
where the academic year is not aligned with that of the other EPFs (the intake of new students arrives in late 
July). It is not clear why it was not collected in the remaining 2 colleges, although there was some suggestion 
that the EPFs did not consider the content of the tests to be relevant to their program. If this is the case, the 
EPFs collectively should devise and administer a test relevant to their needs in order to gather and use systematic 
monitoring information.
Table 11 below summarizes the results of the pre-and post-tests in the 6 EPFs for 2018.
EPF Average mark 
out of 130 -
pre-test
Average mark 
out of 130 -
post-test
Difference 
between pre￾and post-test
%average 
increase in 
score by EPF
% of students 
whose scores 
increased more 
than 10%
EPF Gaza 94.7 107.8 13.1 13.8% 60%
EPF Inhambane 79.7 83.4 3.7 4.6% 23%
EPF Nhamatanda 74.6 103.2 28.6 38.3% 96%
EPF Tete 57.9 63.2 5.3 9.2% 58%
EPF Nametil 79.75 102.3 22.6 28.3% 88%
EPF Nacala 78.3 110.9 32.6 41.6% 97%
Average 77.5 95.1 17.6 22.8% 84.4%
Table 11: Results of Portuguese language pre- and post-tests in 6 EPFs 2018
Comparison of these shows that:
• 84% of students’ scores improved by more than 10% between the beginning and the end of the academic 
year; 
• On average, the scores of trainee teachers improved by 23% between the pre- and post-tests;
• There is some variation between the EPFs, across all metrics of comparison, with Nhamatanda, Nametil and 
Nacala showing most improvement
These results demonstrate that the EPF teacher training is providing significant added value in Portuguese 
language competence, at least in the 6 colleges for which this information is available. It is striking that EPFs 
such as Nhamatanda, Nametil and Nacala managed to improve their scores significantly more than the three 
other EPFs for which data is available. Going forward, it is strongly recommended that the diagnostic tests, or, 
if preferred, a comparable test rigorously devised to interrogate the different areas of Portuguese language 
knowledge should be used each year in all EPFs to monitor improvements in Portuguese language attributable 
81
to the teacher training program. Internal investigation should be undertaken to ascertain why some EPFs 
perform so much better than others in this area and devise a plan to learn from and extend this good practice 
across all 11 teacher education colleges. 
The Baseline report presented information on the overall profiles of EPF trainers and on supply and demand for 
specialized trainers, which the EPFs themselves had identified as useful for their planning and operations. 
Updated information based on the same indicators was collected at midterm by asking the EPF directors to 
complete the same questionnaires. Information on the overall profiles of EPF trainers and on supply and demand 
for specialized trainers at midterm is compared with the same information at baseline and presented below. 
Due to unforeseen circumstances, this information was not made available by EPF Cabo Delgado, which was 
dealing with the aftermath of Hurricane Kenneth at the time of data gathering. Given that data from EPF Cabo 
Delgado was not available at midterm, the comparison also disregards the baseline information from that EPF. 
Profiles of EPF trainers
The eleven EPFs (data from ten are presented since Cabo Delgado is not included in this presentation) have 
between 11 and 24 trainers, with the mean being 18; this has not changed since Baseline. The ratio of female to 
male trainers remains the same as at Baseline, at 23%. Given the relatively low turnover in teacher trainers and 
the fact that the baseline and midterm data collections took place only one academic year apart, it is not 
surprising that there has been little change in this area. 
There has been a small improvement in the overall level of qualification of the trainers. 93% of trainers hold a 
Licenciatura as their highest qualification (compared with 90% at Baseline). 4% of trainers hold a Master’s degree
(compared with3% at Baseline). Only 3% of trainers now have a lower qualification than the Licenciatura 
(compared with 8% at Baseline). Of these, four hold a Bacharelato and two the “Medium” level (at Baseline, 
there were 6 trainers with Bacharelatos, 8 with “Medium” level and 2 with “Basic” level), suggesting that 
attempts to ensure all trainers upgrade their level of qualification to at least Licenciatura are gradually 
succeeding34
. 
42% of the trainers have a qualification in Education, Science of Education or School Management; 34% in the 
areas of the arts, languages or humanities and 23% in the areas of science, social sciences, mathematics or 
physical education; this is not significantly different from the situation at Baseline. Overall the level of 
34 The “Licenciatura” is the equivalent of a European or US Bachelor’s Honours degree; a “Bacharelato” is 
approximately equivalent to an Associate Degree, following two or three years of higher education, and has 
been phased out in recent years in favour of a three-year “Licenciatura” in alignment with other countries; 
“Medium level” is the teacher training qualification earned after graduating from 12th grade of secondary school; 
“Basic level”is the teacher training qualification earned after graduating from 10th grade of secondary school. In 
principle, in order to train teachers in Mozambique, trainers are now generally required to possess at minimum 
the level of “Licenciatura”, although this requirement has not always been consistently applied.
82
professional qualification of EPF trainers has improved slightly: 27% are graduates of One World University/ISET 
(27% at Baseline); 44% were trained at the Pedagogical University (40% at baseline); 25% at another university, 
generally either Universidade Eduardo Mondlane or the Catholic University (25% at Baseline); the number of less 
qualified trainers, whose highest qualification was earned at institutions such as the state-run teacher training 
colleges or the EPFs has fallen to 2% from 4% at Baseline. 99% of the trainers have received psycho-pedagogical 
training (98% at Baseline)35
.
The overall levels of experience of primary teaching of the EPF trainers have increased slightly since baseline: 
28% of the trainers have had two or fewer years’ experience teaching at primary level (compared with 30% at 
Baseline); 15% have had between 2 years 1 month and 5 years (17% at Baseline); 27% between 5 years 1 month 
and 10 years (25% at Baseline) and 30% have had more than 10 years 1 month (28% at Baseline). 
There has been more change in terms of the trainers’ experience of training student teachers within an EPF 
setting: 11% have had two or fewer years’ experience training student teachers (compared with 9% at baseline); 
24% between 2 years 1 month and 5 years (30% at Baseline); 35% between 5 years 1 month and 10 years (37% 
at Baseline) and 30% more than 10 years 1 month (24% at Baseline). 
There has been significant progress concerning the administrative categories of the EPF trainers. The percentage 
classified (and therefore paid) appropriately as teacher trainers (Instructor Técnico Pedagógico - ITP) has risen 
to 35% (compared with 15% at baseline). 34% are now classified in the category ITP-N1, compared with 14% at 
baseline; two trainers are in the category ITP-N2 (compared with one at Baseline). This is encouraging progress, 
on which the EPFs and the government are to be congratulated, although there is still considerable progress to 
be made in this respect. The most frequent category is still DN1, the highest category of “docente” (teacher), 
with 57% of teacher trainers still classified in this way (compared with 66% at baseline). The percentage of 
trainers classified in the lower DN categories of DN2, DN3, and DN4 is now 8%, compared with 19% at Baseline
(see figure 16 below). 
35 As explained above, that data from EPF Cabo Delgado was not available at midterm. This comparison also 
excludes the baseline information from that EPF. 
83
Figure 16: Administrative categories of EPF trainers
Specialized subjects: trainer supply and demand
Analysis of the overall relationship between the needs for specialized teacher trainers and the numbers of 
trainers employed per subject throughout the ten EPFs for which data is available reveals a situation fairly similar 
to that at baseline. Although the majority of positions are filled in many subjects, such as Psycho-pedagogy 
(16/17 posts filled, compared with 19/19 at baseline), Portuguese language teaching methods (16/17 posts 
filled, compared with 15/16 at baseline), Expression Techniques and Language(11/13 posts filled, compared with 
13/14 at baseline), School management and Organization (12/14 posts filled, compared with 13/14 at baseline), 
Social Sciences teaching methods (13/14 posts filled, compared with 13/15 at baseline) and Bantu languages 
and bilingual education (9/10 posts filled, compared with (9/11 at baseline), there are considerable numbers of 
vacant posts in certain other subjects. There are a large number of unfilled posts in Research methods/ICTs (5/11 
posts filled, compared with 5/12 at baseline), Visual education teaching methods (4/10 posts filled, compared 
with 4/11 at baseline), Building, Maintenance and School Production (4/9 posts filled, compared with 4/11 at 
baseline), Crafts teaching methods (2/10 posts filled; 2/11 at baseline), Music teaching methods (7/10 posts 
filled; 6/11 baseline) and Moral and Civic Education (5/9 posts filled; 4/11 baseline) (see figures 17 and 18
below). As pointed out in the baseline report, although the evaluation does not reveal why there should be such 
a gap between supply and demand in these subject areas, the situation is similar in other countries and sectors 
where labor market forces have led to a shortage of specialized teachers. 
0
20
40
60
80
100
120
140
160
DN1 DN2 DN3 DN4 ITP-N1 ITP-N2 ITP-N3
Number of Trainers
Administrative Category
Baseline Midterm
84
Figure 17: Specialized subjects: EPF trainer supply and demand at baseline
Figure 18: Specialized subjects: EPF trainer supply and demand at midterm
Finally, the percentage of the EPF trainers overall with at least five years’ experience of primary teaching has 
increased slightly to 64%, compared with 61% at baseline. Once again, there is considerable variation between 
the EPFs in this regard, with EPF Maputo, Nacala, Nametil and Tete having the best ratios36
. 
36This is significant in terms of conformity with the Ministry of Education and Human Development’s 
requirement for teacher trainers in government teacher training institutions. The requirement is based on a 
regulation regarding the required profile of teacher trainers, introduced in 2007 by the then Ministry of 
Education and Culture. Article 53 of this regulation states that the following people can apply for posts of 
instructor in one of the state-run teacher training colleges:
a) Instructors and teachers who have a higher education qualification in the disciplinary areas required to 
teach the curriculum and who have a minimum experience of 5 years of basic level primary teaching;
b) Professionals with a “medium” level qualification, who have undergone specific training and have the 
recognised professional competence and relevant professional experience of 7 years minimum.
0
5
10
15
20
25
Baseline: Specialised Trainers/Target
Supply Demand (unmet)
0
2
4
6
8
10
12
14
16
18
20
Midterm: Specialised Trainers/Target
Supply Demand (unmet)
85
Visits to EPFs and discussion with management, trainers and students, revealed several areas of progress since 
baseline were noted, along with some which should be monitored. 
The EPFs are still able to provide a culture in which a “different sort of teacher” may be formed, with a focus on 
the more holistic aspects of education, which provides trainees with the skills, attributes and attitudes which 
the realities of teaching in low-resource rural schools require. However, this is made more challenging by the 
requirement that all EPFs follow the national curriculum for teacher education, and the fact that all but three 
EPFs are still obliged to offer only a one-year teacher training program: only EPF Maputo, EPF Chimoio and EPF 
Cabo Delgado offer the 3-year course. Students who have low levels of Portuguese language or reading or writing 
on arrival at the EPF are given extra support, as one EPF director explained:
There are some students who have problems in writing and even in reading…/… when they arrive, the 
Portuguese language trainer gives them a diagnostic test …/… the aim is not to correct it and give a 
mark, the aim is to evaluate their level, how does this person write, what is their handwriting like, what 
is their spelling like …/… to select the people who have problems and identify those who need particular
support. If you noticed, our classrooms all have syllable charts so in their free time, study time, those 
who have handwriting problems are told to draw those letters, copy from the syllable chart. It’s 
effective, we see changes: people who wrote rather badly when they arrived can write better by the 
time they graduate. They know how to form the letters well, mostly cursive letters, which are the ones
they are recommended to teach. 
During visits to three EPF, the external evaluator was able to observe a number of EPF students conducting 
teaching practice and to assess the quality of support and feedback they received. The quality of the practice 
lessons observed was variable, not surprisingly given that he student teachers were close to the beginning of 
their training, only one month into their course. On the whole, they made good use of teaching aids, including 
items they had evidently produced themselves. The lessons were well structured and based on a written lesson
plan. Concepts and content were introduced before being practiced. There was little use of group or pair work, 
although one student teacher had pupils moving around the classroom and making patterns to illustrate a 
mathematics principle she was teaching, which was very effective and engaging for the pupils. One student 
teacher showed remarkable awareness of two disabled children in the class and managed to include them in the 
lessons, physically helping a child with reduced mobility to come to the blackboard to write an exercise and 
guiding the hand of a child with motor coordination challenges to hold a pencil correctly and form letters. All 
the student teachers had progress to make in terms of the pace of their lessons, providing students with 
opportunities for active practice of the concepts taught and monitoring the comprehension and participation of 
students in the class. However, the lessons observed were all of an acceptable level for near-novice trainee 
86
teachers; some showed considerable promise. The quality of observation and feedback from the EPF tutors and 
the school-based mentors was high. Feedback was balanced between positive comments and aspects to improve 
and was provided in a constructive manner. Suggestions for improvement were practical and realistic. The 
external evaluator was impressed by the conscientious and competent support provided to the student teachers 
observed (of course they were aware they were themselves being observed, so this may have conditioned their 
behavior). 
Trainee teachers in the EPFs have school-based teaching practice throughout their training, with support and 
feedback from their trainers and from school-based mentors. The coordination between the EPFs and practice 
schools appears to be good, at least in the EPFs visited. 
The EPFs visited had active gardens producing a side variety of vegetables, fruit and other produce, which is used 
for the meals served to students and staff, which are well-balanced and planned by students on the basis of 
principles learnt during nutrition education. They have also begun to raise small livestock (chickens, rabbits or 
ducks) for sale or consumption. Student teachers contribute to the garden work and husbandry, learning 
valuable skills to apply when they become teachers. The nutrition education component of the EPF curriculum
includes materials on basic nutrition, balanced diet, hygiene and sanitation and the use of school gardens, in 
addition to a nutrition album (teaching tool), pamphlets, posters and a cookbook developed by the nutrition 
education team. 
The infrastructure and equipment within the EPFs, at least those visited, is relatively good and has benefitted 
from a program of progressive upgrading over the course of FFE2. 
Finally, the recent evacuation of EPF Nhamatanda following Cyclone Idai must be mentioned. This occurred in 
early March, during the weeks following the first and just prior to the second visit to Mozambique of the external 
evaluator. When the devastating cyclone struck, the students were evacuated on foot from Nhamatanda, leaving 
all their possessions behind. All students and staff escaped alive, though shaken. They were subsequently sent 
to five different EPFs who welcomed them and allowed their training to continue with only 2 weeks delay. This 
prompt response and the relatively small interruption in studies is a testament to the efficiency and engagement 
of ADPP Mozambique.
One class of the student teachers evacuated from Nhamatanda was welcomed in the EPF Gaza, where the 
external evaluator met them during her visit. They were still visibly affected by their experience and sat huddled 
in donated coats, claiming they felt cold in Gaza.
87
Recommendations
Due to the short duration of the one-year teacher training courses, it is vital that EPFs promote active, teacher￾centered teaching methods even more vigorously. Additional transport is needed to allow trainee teachers to 
reach their practice schools on time and presentable to teach (many currently walk long distances, meaning they 
arrive dusty and less than fresh.
Systematic pre-and post-testing of students to measure the added value of the Portuguese literacy instruction 
should take place each year in all EPFs. 
The link between nutrition education and food production and meal planning and preparation in the EPFs could 
be made even stronger by explicitly linking theory and practice, using nutrition education to reinforce the 
practical food-based activities and vice versa. Introducing some assessment of the application of nutrition 
education in menu and food preparation activities might reinforce the status of nutrition education within the 
EPFs. 
Some needs for infrastructure and equipment upgrading remain. The accommodation for teaching staff at EPF 
Gaza is in extremely poor condition and, in some cases, unsafe for human habitation. Additional computers are 
needed for use by both students and trainers.
Trainer trainers should have access to continuous professional development in order to stay abreast of 
developments in their specialist subjects, in pedagogy and in assessment. Trainer trainers should have access to 
more resources and additional computers in the staff room. 
Students evacuated following Cyclone Idai should continue to receive appropriate support and possibly 
counselling in some cases.
Skills and Knowledge of School Administrators (1.1.5)
Indicator 3 (Outcome): Number of school administrators and officials in target schools who 
demonstrate use of new techniques or tools as a result of USDA assistance. Final target: 210
Midterm target: 150
Midterm results (March 2019): 165
Midterm target 110%achieved
Indicator 4 (Output): Number of school administrators and officials trained or certified as a result of 
USDA assistance. Final target: 264
Midterm results (March 2019): 426
Final target 161% achieved
88
Analysis of project records reveals that the target for training school administrators and officials in management 
and administration of the school feeding program in target schools has been surpassed as of midterm. School 
directors reported this training to have been useful. During school visits, the external evaluators were impressed 
with the level of knowledge and commitment to the project demonstrated by school leaders as they showed the 
visitors the installations and activities put in place by the project.
It was not possible to formally collect data demonstrating that school leaders demonstrating use of the new 
techniques or tools acquired. However, two items of the school survey, which was conducted in 170 project 
schools are considered as a proxy for the purposes of the midterm evaluation: enumerators were asked to 
observe the state of cleanliness of the storeroom and of the kitchen; In 106 of the 170 schools visited (62%), the 
storeroom was considered by the observer to be “perfectly clean”; 51 (30%) were judged “more or less clean” 
and 13 (8%) “not clean”). In 104 schools surveyed (61%), the kitchen was considered to be “perfectly clean”, 58 
kitchens (34%) were judged “more or less clean”, with 8 (5%) judged “not clean”. Where this infrastructure 
provided by the project and used continuously to support it is well maintained, this is considered as evidence 
that school leaders are performing their leadership role and applying the training they have received. 
Extrapolating from the figure of 61% of the schools sampled, it is estimated that the leaders of 61% of the total 
of 270 schools (165 schools) are likely to be demonstrating these characteristics, making an estimate of 165 
school administrators and officials who are expected to be applying the training they have received. 
Overall, the detailed responses of school directors to the school survey reveal a high degree of knowledge of
and commitment to the project. 
Attentiveness (1.2)
(Outcome): Percentage of teachers who report increased attentiveness of students in the classroom.
(Special study indicator). Final target: N/A
Midterm target: N/A
Midterm results (March 2019): 11%
(Outcome): Percentage of students in project schools who report increased attentiveness in the 
classroom. (Special study indicator). Final target: N/A
Midterm target:N/A
Midterm results (March 2019): 6%
Attentiveness in the classroom is an important link in the causal chain which the project ToC seeks to address: 
short-term hunger leads to lack of attentiveness in the classroom leading to problems in learning. The ToC 
postulates that reducing short-term hunger in the classroom should lead to fewer problems of attentiveness, 
89
leading in term to better learning. Many interview informants mentioned the improvement in students’ 
attentiveness since school feeding began. One school director explained: 
They cover large distances to come to school, some of them leave home without eating anything but 
when they arrive, they have the meal and they manage to be calm and attentive in their lessons.
A SDEJTs school feeding focal point was more explicit:
Because if I’m in class and I have something in my stomach, my attention is better than if my stomach’s 
rumbling and I’m physically present but not attentive. 
Previous evaluations showed the project school feeding interventions had reduced short-term hunger and 
problems with attentiveness; however, in the absence of improved teaching, these had not resulted in improved 
learning. At midterm, in order to offer some triangulation of these information obtained, both teachers and 
learners themselves were asked about their experience of students feeling hunger in the classroom37
. 
When asked whether they had problems paying attention in the classroom, 42% of the pupils in intervention 
schools (FFE + literacy and FFE only combined) said they did, compared with 48% in comparison schools. Both 
groups had reported slightly lower rates at baseline (38% for pupils in intervention schools, 36% for pupils in 
comparison schools). Although pupils in intervention schools are less likely to report problems of attentiveness 
in the classroom, differences in change between baseline and midterm for the two groups are not statistically 
significant (see figure 19 below and Technical Appendix 1, pp 197 - 198). 
37 Students are the primary beneficiaries of many of the project’s interventions and their voices and experiences 
are essential to the evaluation. In order to capture multiple experiences and mitigate the risk of bias or recall 
errors (or adults providing “official” responses they perceive to be correct or desired), wherever possible the 
same questions are asked of students and teacher. 
90
Figure 19: Pupils’ reporting of attention problems 
Teachers were also asked about their experience of students’ attentiveness in the classroom. At midterm, 79% 
of teachers surveyed (in FFE + literacy and FFE only schools combined) reported their students “often” or 
“sometimes” having attention problems in the classroom (compared with 88% at baseline) and 22% that they 
“never” did (compared with 11% at baseline) (see figure 20 below and Technical Appendix 2,pp 52-54)
38
. 
Figure 20: Teachers’ reporting of pupils’ attention problems
38 The “often” and “sometimes” responses were combined, due to their imprecision, compared with the more 
categorical “never”.
0%
10%
20%
30%
40%
50%
60%
Comparison Intervention
Pupils Reporting Attention Problems
Baseline Midterm
0%
20%
40%
60%
80%
100%
Often or
sometimes
Never No response
Teachers' Reporting of Pupils' Attention 
Problems
Baseline Midterm
91
Teachers were asked whether their pupils had difficulties learning. At midterm, 84% of teachers surveyed (in 
FFE + literacy and FFE only schools combined) reported their students “often” or “frequently” having attention 
problems in the classroom (compared with 82% at baseline); 15% reported students “never” had problems 
learning (16% at baseline) (see figure 21 below and Technical Appendix 2, pp 54-56). 
Figure 21: Teachers’ reporting of pupils’ learning problems 
On the basis of teachers’ responses to these questions, it appears that improvements in students’ attention in 
the classroom are not yet being translated into improved learning. 
Short-term hunger (1.2.1)
(Outcome): Percentage of teachers who report reduced short-term hunger of students in the 
classroom. Final target:N/A
Midterm target:N/A
Midterm results (March 2019): 10% increase in “never” since baseline
(Outcome): Percentage of students in project schools who report reduced short-term hunger in the 
classroom. Final target: N/A
Midterm target:N/A
Midterm results (March 2019): 18% decrease since baseline
When asked whether they sometimes felt hungry in the classroom, at midterm 42% of pupils in intervention 
schools (FFE + literacy and FFE only combined) said they did, compared with 58% of students in comparison 
schools. These levels of hunger are lower than the baseline levels for both groups: 60% for pupils in intervention 
schools, and 68% of students in comparison schools. Although considerably fewer students in project schools 
0%
10%
20%
30%
40%
50%
60%
70%
80%
90%
Often or
sometimes
Never No response
Pupils Have Problems Learning
Baseline Midterm
92
said they were sometimes hungry than those in comparison schools and considerably fewer students in 
intervention schools said they were sometimes hungry at midterm than at baseline, the differences in change 
between baseline and midterm for the two groups are not statistically significant (see figure 22 below and
Technical Appendix 1, pp 195 - 196). 
Figure 22: Pupils’ reporting “sometimes” hungry at school 
Students were also asked whether they felt hungry “now” i.e. at the time of the interview39
. At midterm 23% of 
pupils in intervention schools (FFE + literacy and FFE only combined) said they did, compared with 33% of 
students in comparison schools. Again, these levels of hunger are lower than the baseline levels for both groups: 
42% for pupils in intervention schools, and 66% of students in comparison schools. Once again, although 
considerably fewer students in project schools said they were hungry “now” than those in comparison schools 
and considerably fewer students in intervention schools said they were hungry “now” at midterm than they had 
at baseline, the differences in change between baseline and midterm for the two groups are not statistically 
significant (see figure 23 below and Technical Appendix 1, pp 196 - 197). 
To increase validity and allow triangulation of responses, pupils were asked two separate but related questions 
about their experience of hunger at school. It should be borne in mind that, due to fieldwork constraints, pupils 
were asked whether they felt hungry “now” at varying times of day; in some cases after they had already eaten 
the soya porridge, in others before. This constraint also applied during the baseline data gathering. Interview 
times were recorded in all cases.
0%
10%
20%
30%
40%
50%
60%
70%
80%
Comparison Intervention
Pupils Sometimes Hungry at School
Baseline Midterm
93
Figure 23: Pupils’ reporting hungry at time of interview
When asked about their perceptions of student’s hunger in the classroom, at midterm 67% of teachers stated 
that their pupils “often” or “sometimes” appeared to be hungry in class (76% at baseline) and 33% that they
“never” appeared hungry (23% at baseline) (see figure 24 below and Technical Appendix 2, pp 50-52). 
Figure 24: Teachers’ reporting pupils appear hungry at school 
Access to food (school feeding) (1.2.1.1)
Indicator 15 (Output): Number of daily school meals (breakfast, snack, lunch) provided to school￾age children as a result of USDA assistance. Final target: 48,100,000 
Midterm target: 27,056,250
Midterm results (March 2019): 28,162,260
Midterm target 104% achieved
0%
10%
20%
30%
40%
50%
60%
70%
Comparison Intervention
Pupils Feeling Hungry at Time of 
Interview
Baseline Midterm
0%
10%
20%
30%
40%
50%
60%
70%
80%
Often or
sometimes
Never No response
Pupils Appear to be Hungry at School
Baseline Midterm
94
Indicator 16 (Output): Number of school-age children receiving daily school meals (breakfast, snack, 
lunch) as a result of USDA assistance. Final target: 87,000; (000 new: 13,000; continuing: 74,000; 
female: 42,630; male: 44,370)
Midterm target: 74,000 (new: 2,000; continuing: 72,000; female: 36,260; male: 37,740)
Midterm results (March 2019): 84,026 (new: 787; continuing: 83,239; female: 42,141; male: 41,885)
Target 114% achieved (new: 39%; continuing: 116%; female: 116%; male: 111%)
Indicator 17 (Output): Number of social assistance beneficiaries participating in productive safety 
nets as a result of USDA assistance. Final target: 74,000 87,000; continuing: 69,000; new: 13,000; 
female: 42,630; male: 44,370)
Midterm target: 74,000(continuing: 72,000; new: 2,000; female: 36,260; male: 37,740) 
Midterm results (March 2019): 84,026 (new: 787; continuing: 83,239; female: 42,141; male: 41,885) 
Final target 114% achieved (new: 39%; continuing: 116%; female: 116% ; male: 111% ) 
Analysis of project records shows the distribution of school meals to be on target. In order to triangulate this 
information, the midterm evaluation surveys included questions for students, teachers and school leaders about 
the distribution of school meals. 
64% of schools surveyed reported that school feeding had been missed one or more times during the past two 
months. This is a significant increase since baseline, when 34% schools stated that in the past two months there 
had been one day or more when they had not been able to prepare the lunch. At midterm, the main reasons 
cited for not being able to feed children were running out of CSB+ (cited by 21% of schools who had had gaps in 
feeding, compared with 45% at baseline); cooks not turning up (cited by 52% of schools who had gaps in feeding, 
compared with 47% at baseline); lack of water (cited by 16% of schools who had gaps in feeding, compared with 
23% at baseline); lack of firewood (cited by 10% of schools who had gaps in feeding, compared with 11% at 
baseline). 
25 of the 170 schools surveyed (15%) reported running out of CSB+ during the past two months. 7 of these 
schools reported remaining without CSB+ for 1-3 days; 6 for 4 – 6 days; 4 for 7 – 9 days; 3 for 10-12 days and 3 
for 13 – 15 days. One school reported being without CSB+ for 20 days and one for 21 days. Project 
representatives reported that delays in a shipment of CSB from the United States early in April accounted for 
the number of schools which ran out of the commodity at this time. Although it is encouraging that 88% of 
schools did not run out of CSB+, for those schools unable to feed their students due to interruptions in the stock, 
the impact on their students will be considerable. Not to receive food when they have become used to doing so 
is disruptive and distressing. It may also have long-term negative impact on the advances made in terms of 
attendance and retention, if families lose trust in school feeding and, potentially, in the school in general for not 
being able to provide a promised service. This risk was mentioned several times during evaluation interviews.
Analysis of project records in the commodity tracking files for March and April 2019 (see tables 12 and 13 below), 
the same period as the school survey was conducted, reveal:
95
• The principle overall reason for non-feeding is pupil absence;
• Of reasons directly attributable to the project itself, the main reasons for schools not feeding are lack 
of CSB+ and lack of cooks;
• In March 2019, 126 schools (46%) failed to feed at least once;
• In March 2019, 435 feeding days (8%) were lost of a total of 5,420 potential days (271 x 20);
• In April 2019, 179 schools (68%) failed to feed at least once;
• In April 2019, 989 feeding days (19%) were lost of a total of 5,149 potential days (271 x 19)40;
• Infested CSB+ is still a problem in a small number of schools;
• No schools fail to feed because of lack of utensils;
• Only 4% of schools failed to feed due to lack of water, suggesting many schools without safe drinking 
water are managing to find a source of water to prepare the soya porridge (interview respondents 
explained that pupils bring water from home or river water is used). 
Magude (66) Moamba (65) Manhica (85) Matutuine (55) Total
Schools Days 
lost
Schools Days 
lost
Schools Days 
lost
Schools Days 
lost
Schools Days 
lost
No CSB+ 0 0 15 83 20 77 0 0 35 160
Infested CSB+ 3 7 0 0 1 7 0 0 4 14
No water 1 10 7 29 1 9 2 3 11 51
No firewood 1 1 1 1 3 15 0 0 5 17
No cooks 9 31 13 38 5 21 10 23 37 113
Other 2 2 2 2 1 1 5 10 10 15
No reason given 9 24 3 9 7 18 5 14 24 65
Total 25 75 41 162 38 148 22 50 126 435
Table 12: Schools not feeding for specific reasons and feeding days lost March 2019
Magude (66) Moamba (65) Manhica (85) Matutuine (55) Total
Schools Days 
lost
Schools Days 
lost
Schools Days 
lost
Schools Days 
lost
Schools Days 
lost
No CSB+ 18 63 26 160 59 466 9 61 112 750
Infested CSB+ 0 0 0 0 0 0 0 0 0 0
No water 1 1 4 13 3 24 1 1 9 39
No firewood 0 0 0 0 2 5 0 0 2 5
No cooks 4 6 7 13 7 16 1 6 19 41
Other 2 20 1 2 3 3 2 10 8 35
No reason given 12 54 0 0 7 14 10 51 29 119
Total 37 144 38 188 81 528 23 129 179 989
Table 13: Schools not feeding for specific reasons and feeding days lost - April 2019 
40 The relatively high number of feeding days lost due to schools running out of CSB in April is attributed to a 
delayed shipment of CSB+, rather than to failures in the distribution of the commodity to schools. 
96
It is clear that the two priorities in terms of improving the rate of school feeding are: i) ensuring there is CSB+ in 
all schools at all times and ii) ensuring there are sufficient volunteers cooks available each day.
Teachers rated the distribution of food in school to be “excellent” (27%) or “good” (67%). However, 66% of 
teachers surveyed stated the distribution could be improved, citing the following aspects which require 
improvement:
• Quality of CSB+ delivered to schools;
• Timely delivery of CSB+, so this does not run out;
• Absence / lateness of volunteer cooks;
• Lack of incentives for volunteer cooks and storeroom managers;
• Insufficient cooking utensils, bowls and spoons;
• Better organization of time so distribution of food does not excessively interrupt teaching time (for 
example, better planning, so as to reduce waiting time and feed pupils after/ before class or during 
the long break);
• Lack of water or firewood.
This list is very similar to the aspects revealed in the school survey and to those cited during previous evaluations. 
It is also rather concerning that there continue to be so many days without school feeding, especially where this 
is due to schools running out of CSB+. The final evaluation of FFE 1 had found there had been improvements to 
the distribution of CSB+, due to the use of a commodity tracking system (CTS) and improved processes and 
systems. 
Students who participated in focus groups explained that the soya porridge was often too hot when distributed, 
so children got burned. They also claimed to only enjoy the food when they were able to add sugar, brought 
from home. 
There has been a relatively high turnover within the project Logistics team. At the time of the midterm data 
collection, a new Logistics manager had recently been appointed and he appeared to have a clear vision of where 
changes were needed and how to implement these.
In addition to several positive aspects of the logistical organization of the project observed, some others require 
monitoring:
There have been improvements in unloading and Customs clearance processes and sanitary controls for CSB+
arriving from USA, so these procedures are now achieved more rapidly.
Information flow appears to have improved since previous evaluations, including greater cooperation from most 
storeroom managers who send information on weekly stock levels by SMS to their “professional”, who transmits 
97
this information to the logistics manager so CSB+ delivery can be planned. There are still some flaws in the 
information flow between schools and the logistics manager. Some storeroom managers lack commitment or 
are absent during deliveries. The storeroom managers are teachers and often do not perceive the management 
of the storeroom as part of their responsibilities as teachers. They have not received any form of incentive for 
managing the storerooms since the inception of the project.
During fieldwork, the external evaluator was told several times of CSB+ in school storerooms being found to be 
infested. The rate of infestation of CSB+ is low overall; where it occurs, the commodity is sifted to remove weevils 
then repackaged. 
Recommendations
Cooks should be made aware of the need to let the porridge cool down before serving, to avoid burning the 
students. 
Adding additional sugar is not recommended, as the nutritional content of the CSB+ is already well balanced to 
meet the nutritional needs of growing children.
The project’s internal organization of these processes is still in need of some improvement, as are other logistical 
processes. 
Consider offering storeroom managers an expression of thanks and appreciation in the form of a token gift of 
phone airtime in recognition of their important role in the project in order to improve morale and commitment. 
Systems must be improved so all infestation is eliminated. 
Height, weight and BMI z-scores
In order to see whether school feeding impacts on pupils’ physical development, two anthropometric measures: 
height and weight will be recorded at three time points, so that BMI z-scores can be calculated, according to
WHO guidelines. At midterm, BMI z-scores for the two groups receiving school feeding (FFE + literacy and FFE 
only) were compared with those from the comparison group. The FFE + literacy and FFE only group had a mean 
score of -0.514 (SD 1.24), compared with -0.437 (SD 1.72) at baseline. The comparison group had a mean score 
of -0.675 (SD 1.92) compared with -0.416 (SD 2.15) at baseline. The difference in rate of change from baseline 
to midterm between the two groups was therefore not significant. This result suggests that those children who 
received school feeding did not grow significantly more than those who did not. 
98
At baseline, calculating BMI Z-score scores proved problematic in many cases, since approximately one third of 
pupils did not have accurate dates of birth but a year of birth and 1 January as a default date41. It had been 
hoped that at midterm accurate dates of birth could be collected in order to rectify this. Enumerators were 
trained to probe several avenues to obtain correct dates of birth for students and the ODK data collection 
interface required them to indicate the source of the date of birth information. This effort was successful in 
yielding more dates of birth, although it was still not possible to obtains dates of birth for all students. Of the 
2,882 students for whom data was available at both time points, 1,935 did not have consistent dates of birth at 
both time points. When midterm dates of birth were available, these were considered to be more reliable and 
were therefore used for both the mid-term and baseline data, to allow appropriate comparisons.
In order to probe the FFE2 midterm results further, the height and weight gain of participating students were 
also analyzed42
. The rate of change from baseline to midterm was almost identical between the two groups. At 
midterm, the mean height of the FFE + literacy and FFE only group was 126cm (SD 7.83), compared with 114cm 
(SD 6.62) at baseline. The mean height of the comparison group was also 126cm (SD 8.75) compared with 114cm
(SD 6.29) at baseline. The rate of increase in weight from baseline to midterm was also very similar between the 
intervention and comparison groups. At midterm, the mean weight of the FFE + literacy and FFE only group was 
24.5 kg (SD 4.52), compared with 19.8 at baseline (SD 3.39). The mean weight of the comparison group was 24.2 
kg (SD 5.10), compared with 20.0 at baseline (SD 3.62). No significant difference in either height or weight was 
found between the two groups. 
The complete results are presented in Technical Appendix 1, pp 188 - 194.
Informants consistently commented on the physical improvements observed in students’ appearance since the 
onset of schools feeding: students looking “healthier” and noting improvements in the condition of their skin 
and hair, in addition to increased energy and dynamism.
41 There are several possible explanations for this: early grades pupils may not know their date of birth; school 
records may in incomplete or inaccurate; the enumerators may lack the experience and expertise to be able to 
elicit this information. 
4242 The final evaluation of FFE1 had found that students receiving school feeding grew on average 2 cm more 
than students in the comparison group, although increases in their BMI Z-scores were not significantly different. 
99
Student attendance (1.3)
Indicator 1 (Outcome): Number of students regularly (80%) attending USDA supported 
classrooms/schools. Final target: 59,200 (female: 29,008; male: 30,192)43
Midterm target: 59,200 (female: 29,008; male: 30,192)
Midterm results (March 2019): 84,026 (female: 42,141; male: 41,885) [if 80% used: 67,220 (female: 
33,712; male: 33,508)]
Final target 142% achieved (female: 145%; male: 139%) [if 80% used: 114% achieved (female: 116%; 
male: 111%)]
During school visits for the midterm evaluation, as during other evaluations of the project, teachers, school 
directors, parents, government officials and other stakeholders consistently claimed that school feeding and the 
other project interventions had led to increases in student attendance. One SDEJT focal point reported: 
As soon as the meals were introduced, we saw fewer children missing school – before they missed a lot 
– and the pedagogical results improved. 
A parent explained this slightly differently:
The children don’t miss school anymore because as soon as they remember that they’ll get food there 
they’re motivated …/… they get there early to be sure of receiving the food, even though they know it 
will be served later on. 
Analysis of project monitoring data reveal that targets for school attendance have been surpassed at midterm. 
There are however two caveats to be born in mind: firstly, the targets for 80% attendance are calculated by the 
project as 80% of the numbers of students receiving food. However, they ought to reflect the numbers of 
students who attend school at least 80% of the time (see footnote below). In an attempt to partially address this 
issue, calculations based on 80% of the project numbers are included in square brackets in the box above; 
however, without actual daily attendance data it is not possible to accurately calculate the number of students 
who attend 80% of the time).
Secondly, the attendance numbers reported by the project are actually based on school feeding data. Since the 
inception of the project, there have been challenges for the project to obtain consistent, reliable pupil-level 
attendance data from schools, due to lack of rigor in recording pupil attendance and teachers’ reluctance to 
complete an additional daily attendance register for the purposes of monitoring school feeding, in addition to 
43The targets for this indicator as set out in the contractual Attachment E in Amendment #1 are based on 80% 
of the target for students benefiting from school feeding; however, results are reported based on 100% of the 
results of students benefiting from school feeding, which makes the target appear to have been surpassed by 
more than is actually the case. It is recommended that the targets and measurement of this indicator should be 
amended from 80% of students attending school to the numbers of student who achieve 80% school attendance, 
to allow the project to report on this indicator as intended.
100
the school register. After considerable effort, towards the end of FFE1, the project leadership managed to obtain 
authorization from the DPEDH to have specially designed school registers produced which included carbon paper 
so the register could be filled in once by the teachers producing two copies: one for the school, the other for the 
project. The midterm evaluation visits revealed that, although this has led in some cases to attendance 
information in paper form being successfully collected for the project, this information is not currently being 
collated electronically, due to other priorities for M&E staff and to the fact that the information received from 
schools is incomplete since not all schools are filling in the registers systematically. For this reason, the numbers 
used by the project to record pupil attendance are the same as those used by the project tracking system to 
track and record the number of students receiving school feeding. These are based on school level enrolment 
data provided at the beginning of the school year by the district education offices (SDEJT), which does not always 
reflect mid-trimester changes in pupil numbers. These are not therefore an accurate record of attendance rates
and do not allow verification of the claim made by school directors and education officials that school attendance 
rates have improved since the introduction of school feeding. 
Number of after-school learning clubs active in project schools. Final target: 600
Midterm target: 600
Midterm results (March 2019): 675
Final target 113% achieved 
Number of school children participating in clubs. Final target: 11,300
Midterm target: 11,300
Midterm results (March 2019): 12,300
Final target 109% achieved 
One of the project activities intended to promote increase school attendance is support to extra-curricular
learning clubs. Of the schools surveyed at midterm, 89% reported having one or more learning club (compared 
with 98% at baseline). Of those schools which report having clubs, 98% report having reading clubs (97% at 
baseline) schools. 87% report having mathematics clubs (84% at baseline). 28% reported having clubs for 
“recreational arts and dance” (50% at baseline). 32% report having school garden clubs (26% at baseline). 34% 
report having other types of club (7% at baseline)44
. 
94% schools report seeing “some differences” in the children who participate in the clubs (94% at baseline). 95% 
report that they are more active in regular lessons (93% at baseline). 100% schools report pupils have improved 
44 Project staff reported that between baseline and midterm, the method for counting clubs changed: only 
literacy numeracy, gardens, science and sanitation clubs were counted; sports, recreational arts and dance were 
no longer counted for reporting purposes, although these continue to be supported by the project. This change 
may have impacted on the responses given by headteachers to the school survey. 
101
learning outcomes (93% at baseline). 95% report improvements in the pupils’ communication and expression
(87% at baseline). For more information, see Technical Appendix 3 (school survey), pp 69 – 8945
.
When asked about the challenges encountered in implementing clubs, the rate of challenges reported appears 
to have substantially diminished since baseline. At midterm, 44% schools report low levels of pupil participation
(73% at baseline), 55% report difficulties finding time for club activities (69% at baseline), 19% report a lack of 
interest amongst teachers in implementing clubs (55% at baseline), 12% report lacking information about 
methodologies or activities which can be used in the clubs (57% at baseline), 30% report lacking the materials 
needed (48% at baseline). For more information, see Technical Appendix 3, pp 247 – 259.
Teachers were also asked about extra-curricular clubs. 83% of teachers reported that clubs had been created in 
their schools (92% at baseline). 95% of teachers reported knowing what activities are offered by the clubs (85% 
at baseline). 95% reported that the clubs had a positive effect on pupils’ schoolwork (88% at baseline) (see 
Technical Appendix 2, pp 22 – 28)46
.
During the first phase of the project, in the absence of substantive support and training of teachers, the project’s 
support to school clubs and provision of kits of educational materials was the main activity designed to improve 
teaching and learning. The baseline report of FFE2 pointed out that, now the project includes an intensive 
literacy intervention, at least in some schools, “the role of school clubs must be clarified, so clubs do not simply 
repeat what happens in lessons. The project needs to rethink its support to extra-curricular clubs, to allow them 
to make more of an impact on learning. Club leaders should receive clear guidance and training in using play￾based and other alternative activities to promote learning and the value and potential of clubs should be 
communicated to government, school leaders and teachers”. The midterm evaluation found some evidence of 
progress in this direction: in all project schools, pedagogical directors and teachers responsible for clubs have 
been trained in the facilitation of extracurricular clubs, play-based and other activities to promote learning and 
use of the materials provided by the project. Within the recent specialization of the project “professionals”, 
several of these now specialize in providing support to clubs and can be called upon by schools who require 
more help in that area. 
Unfortunately, at midterm information on club attendance was not collected from students, due to a technical 
error, which led to the relevant questions being excluded from the instrument during fieldwork. The baseline 
report noted that “The reporting of clubs’ existence by schools does not necessarily mean that these are 
particularly active or dynamic or that pupils’ attendance is regular”. There is some indication from the qualitative 
45 Technical Appendix 3 (school survey) is available in electronic form: pdf file entitled 
04_ADPE_TABLES_SCHOOL_PDF__c
46 Technical Appendix 2 (teacher survey) is available in electronic form: pdf file entitled 
05_ADPE_TABLES_TEACHERS_PDF__c
102
information gathered at midterm that this situation has improved since baseline. The project should continue 
to monitor the quality and activities of extra-curricular clubs and ensure that pupils are participating actively, 
that the activities proposed are useful, practical and well-facilitated, so that the clubs are able to actually make 
a substantive contribution to student learning and outcomes. 
School directors reported positive innovations, such as learning clubs taking place off the school premises:
Last year we had clubs functioning in the neighborhoods, because of the distance separating the children 
from the school - some children live three kilometers away … so we took advantage of the children in 
the neighborhoods who could already read … and were able to teach the others. We made groups of 
neighbors and those older ones, we gave them materials and a monitor responsible for literacy and 
numeracy went to the community. 
A number of positive developments were noted at midterm, along with some aspects to be monitored:
Overall, the club sector appears to be more dynamic. Some basic rules have been applied, such as the posting 
of attendance times for clubs.
Links with other components, such as literacy, school gardens and nutrition are stronger at midterm, in part due 
to the role of the “professionals” who have specialized in these areas and act as specialist focal points.
The principles and mission of clubs appears to be better understood than during previous evaluations. There is 
also more flexibility in their organization, so clubs can take place outside of the school premises where necessary 
or be facilitated by older pupils if appropriate. 
Recommendations
The system for gathering evidence of club activities and criteria for measuring these needs to be strengthened: 
a list of participants may prove clubs exist but is not proof that they provide useful activities. 
The clubs coordinator does not have a dedicated team and would benefit from more support.
Principles of active, learner-centered learning and practical, active activities need to be continually promoted, 
rather than traditional teacher-based methods and reciting lessons by heart.
103
School infrastructure (1.3.3)
Indicator 7 (Output): Number of kitchens, storerooms and firewood saving stoves maintained as a 
result of USDA assistance. Final target: 792 (kitchens: 264; storerooms: 264; firewood saving stoves: 
264) 
Midterm target: 792 (kitchens: 264; storerooms: 264; firewood saving stoves: 264) 
Midterm results (March 2019): 1,002 maintenance activities (kitchens: 271; storerooms: 271; 
firewood saving stoves: 460)47
Final target 127% achieved (kitchens: 103%; storerooms: 103%; firewood saving stoves: 174%)
Analysis of project records reveals that the targets for maintaining the school infrastructure provided by the 
project (in terms of the number of interventions to maintain kitchens, storerooms and firewood-saving stoves) 
have been surpassed. 
The midterm school survey included inspection visits of the infrastructure to verify its state of maintenance and 
cleanliness. In 62% of schools visited, the storeroom was found to be “perfectly clean”; in 30% of schools visited 
the storeroom was found to be “more or less clean”; in 8% “not clean”.77% of schools reported that the 
storeroom was cleaned “daily” (79% at baseline), 16% “weekly” (18% at baseline), 4% “twice a week” (1% at 
baseline), 1% “from time to time” (1% at baseline) and 2% gave another response. 
In 61% of schools visited the kitchen was found to be “perfectly clean”; in 34% of schools visited the kitchen was 
found to be “more or less clean”; in 5% “not clean”. 88% of schools reported that the kitchen was cleaned “daily”
(98% at baseline), 9% “weekly” (5% at baseline), 1% “twice a week”, 1% “from time to time” and 1% gave another 
response. 
27% of schools report having a maintenance plan (39% at baseline). Of these, 87% say their maintenance plan is 
followed (see Technical Appendix 3, pp. 7 – 19). In schools where this is the case, school council members 
contribute to maintenance in 54% of schools, teachers and school leaders in 30%, project staff linked to the 
school in 57%, the director in 35%, the pedagogical director in 15% and other people in 9%. 
In 85% schools, volunteer cooks are amongst those who clean the kitchens (96% at baseline); in 42% schools 
pupils contribute to cleaning the kitchen (3% at baseline), the teacher who is storeroom supervisor contributes 
in24% schools (1% at baseline) and a member of the school council in one school. The storeroom is cleaned by 
volunteer cooks in 208 (77%) schools, by pupils in 39 (14%) schools, by the teacher who is the storeroom 
47 These numbers refer to maintenance interventions performed, not to the number of installations maintained 
(the project has a total of 270 kitchens, 268 storerooms and 271 firewood saving stoves). 
104
supervisor in 17 (6%) schools and school council membersin 4% schools(1% at baseline) (see Technical Appendix 
3, pp. 97 – 131).
Recommendation
In the interests of sustainability, schools should be encouraged to put in place maintenance plans and cleaning 
plans and to ensure these are followed. 
Enrolment (1.3.4)
Indicator 8 (Outcome): Number of students enrolled in school receiving USDA assistance. Final 
target: 74,000 (female: 36,260; male: 37,740)
Midterm results (March 2019): 88,100 (female: 43,180; male: 44,920)
Final target 119% achieved (female: 119%; male: 119%)
Enrolment in project schools is one of the steps considered in the project ToC as a prerequisite to attendance 
and therefore to improved learning. Analysis of project records reveals that the targets for enrolment of 
students have been surpassed. Interview and focus group informants overwhelmingly claimed that enrolment 
had increased, and dropouts decreased as a result of the projects’ activities, particularly school feeding. 
During previous evaluations, analysis of the monitoring data gathered by the DPEDH in respect of enrolments 
and dropouts had provided some degree of support to these claims. It had been hoped to extend this analysis 
for the purposes of the midterm evaluation of FFE2. Unfortunately, the DPEDH was not able to provide 
information of sufficient quality to allow this analysis48
. 
Capacity of government institutions (1.4.1)
This indicator appears with different MGD numbers under both SO1 and SO2. For discussion of capacity of 
government institutions, please refer to the section Capacity of government institutions (2.7.1) below.
48 The DPEDH did make the information requested available to the evaluation team. Unfortunately, due to 
inconsistencies in the way it had been complied, this could not be used. The evaluators requested that the data 
be checked and corrected on several occasions. However, due to a shortage of DPEDH personnel, this was not 
possible. 
105
Government support (1.4.3)
Indicator 11 (Outcome): Value of new public and private sector investments leveraged as a result of 
USDA assistance. Final target: 80,000 
Midterm target: 40,000
Midterm results (March 2019): 0
Target 0% achieved 
To date, the project has not demonstrated the leveraging of new public and private sector investments. 
Discussions are ongoing within the leadership and M&E teams and through the bi-annual reporting process as 
to which current or future project activities may be counted towards this indicator. 
Engagement of Local Organizations and Community output (1.4.4)
Indicator 9 (Output): Number of Parent-Teacher Associations (PTAs) or similar “school” 
governance structures supported as a result of USDA assistance. Final target: 264
Midterm target: 264
Midterm results (March 2019): 271
Final target 103% achieved 
Indicator 10 (Output): Number of public-private partnerships formed as a result of USDA 
assistance. Final target: 2 
Midterm target: 2
Midterm results (March 2019): 0
Target 0% achieved 
At the outset of the project, School Feeding Committees (SFC) were created in project schools. In many schools, 
these were extensions of the existing School Councils, which have an important role in all primary schools in 
Mozambique and wield considerable influence (they often have close links to other political structures). It 
became apparent during previous evaluations that the model of SFCs originally planned by the project had been 
transformed into a more locally appropriate model; this development was welcomed as a sign of local ownership 
of the project. Focus groups with SFCs revealed a high level of support for and approval of the project and a high 
level of collaboration between the SFCs and the project. The SFCs see themselves as the owners or managers of 
the project and the main decision makers. This was explained during a focus group with SFC members:
The school feeding committee which exists in each neighborhood decides the program of how and what 
will be cooked during the week and who will cook it, together with the school council members. The 
support the school director and the teachers to inform each neighborhood ahead of time who are the 
106
mums who must go and cook, in coordination with the “secretários dos bairros” (neighborhood
secretaries).
The project has provided training and reference manuals to SFCs.
The project has not formed public-private partnerships. The project leadership explained to the evaluator that 
they had initially believed that the renovation of premises owned and made available to the project by the 
municipality in Manhiça constituted such a partnership. However, they had subsequently discovered that this 
did not meet the criteria for a public-private partnership. 
Use of Health, Nutrition and Dietary Practices (SO2)
The second large Strategic Objective (SO2) of the project concerns the use of health, nutrition and dietary 
practices and interacts directly with the first in that it feeds into Improved Student Attendance and therefore 
Improved Literacy of School-aged children within the project ToC. 
Knowledge of Health and Hygiene Practice (2.1)
(Outcome): Percentage of students that demonstrate acceptable knowledge of health and hygiene 
practices. Final target: 50%
Midterm target: 31% (female 31%; male 31%)
Midterm results (March 2019): 85% (female 84%; male 85%)
Final target 170% achieved 
Number of volunteer cooks receiving health and hygiene training. Final target: 8,468
Midterm target: 6,620
Midterm results (March 2019): 9,749
Final target 115% achieved 
Number of people trained in good hygiene practices. Final target: 8,400
Midterm target: 6,275
Midterm results (March 2019): 9,394
Final target 112% achieved 
Analysis of the project records demonstrate that the targets for training in good hygiene practice, including 
training volunteer cooks in health and hygiene, have been greatly surpassed. 82% of project schools reported 
that the volunteer cooks responsible for school feeding within that school had received training in good hygiene 
practices from project staff during the previous year (compared with 89% at baseline). 
107
During a focus group with volunteer cooks, one participant explained:
Sr M taught us a lot of good things, first to sweep the floor before lighting the fire, then while the fire is 
catching we wash the pan before to taking it to the fire, then we put the water in, and finally we put the 
soya in: that’s all different from the way we used to do it. 
At midterm of FFE 2, as during previous evaluations, pupils were asked a series of questions to assess their 
knowledge of good health and hygiene practices49.At midterm, 89% of the FFE + literacy and the FFE only 
students combined claimed (without being prompted) to wash their hands before eating, compared with 84% 
of the comparison group. At baseline, 85% of the FFE + literacy students and 71% of the comparison students 
had claimed to wash their hands before eating (see figure 25 below). Although, at midterm, students from the 
intervention groups are more likely to claim to wash their hands before eating than the comparison group, the 
differences between the respective changes since baseline of the two groups are not significant. 
Figure 25: Students’ reporting of hand washing before eating 
At midterm 81%of the FFE + literacy and the FFE only students combined claimed (without being prompted) to 
wash their hands after going to the toilet, compared with 77% of the comparison group. At baseline, 84% of the 
FFE + literacy students and 66% of the comparison students had claimed to wash their hands after using the 
toilet (see figure 26 below). At midterm, the differences between the respective changes since baseline of the 
49 Pupils were asked what they did after using the toilet (enumerators were trained not to mention hand washing 
in their questioning, so as not to prompt this response). Pupils were also asked what they did before eating 
(likewise, enumerators were trained not to mention hand washing in their questioning, so as not to prompt this 
response). 
0%
10%
20%
30%
40%
50%
60%
70%
80%
90%
100%
Comparison Intervention
Hand Washing Before Eating
Baseline Midterm
108
two groups is significant, offering support for the intervention having impacted on the outcome, despite the 
intervention groups being slightly less likely to wash their hands at midterm than at baseline. 
Figure 26: Students’ reporting of hand washing after using the toilet 
In each case, whereas previous evaluations have found significant differences between intervention and 
comparison students in both these respects, it seems that these are now plateauing out; for the grade 3 children 
in the intervention groups, since the project hygiene interventions began six years ago in most cases, their whole 
school career is likely to have been spent in a school where handwashing was strongly promoted, meaning that 
they have not experienced anything else, which would explain the relatively small change between baseline and 
midterm(see Technical Appendix 1 pp198 – 200 for more information). Nevertheless, taken overall, 85% of pupils 
in intervention schools demonstrate an acceptable knowledge of health and hygiene practices, which is a
gratifying result for the project. 
Teachers were asked whether they had seen differences in pupils’ hygiene practices since the beginning of the 
project; 92% claimed that they had. Regarding their own knowledge of hygiene, teachers were asked to name 
three of the stages of handwashing, as taught during the hygiene training. 89% of them were able to name three 
stages correctly. This compares with the baseline survey where 66% of teachers were able to name three stages 
of hand washing correctly (see Technical Appendix 2, pp 58-60 and 366 – 77 for more detail). 
0%
10%
20%
30%
40%
50%
60%
70%
80%
90%
Comparison Intervention
Hand Washing After Using the Toilet
Baseline Midterm
109
Knowledge of Safe Food Preparation and Storage Practices (2.2)
Indicator 20 (Output): Number of food preparers at target schools trained in hand washing, safe 
food preparation and storage practices. Final target: 10,560 (female: 10,032; male: 528) 
Midterm target: 8,250 (female: 7,837; male: 412) 
Midterm results (March 2019): 9,749 (female: 9,588; male: 161)
Midterm target 118% achieved (female 122%; male: 39%)
Analysis of the project records demonstrate that the targets for training volunteer cooks in handwashing and 
safe preparation and storage of food have been surpassed. During the school survey, 82% of project schools 
reported that the volunteer cooks responsible for school feeding within that school had received training in 
handwashing and safe food preparation and storage from project staff during the previous year (compared with 
89% at baseline).
Knowledge of Nutrition (2.3)
Indicator 19 (Outcome): Number of people trained in child health and nutrition as a result of USDA 
assistance. Final target: 10,300 (female: 7,519; male: 2,781) 
Midterm target: 8,175 (female: 5,967; male: 2,207) 
Midterm results (March 2019): 19,257 (female: 14,003; male: 5,254)
Final target 187% achieved (female 186; male: 189%)
Since the onset of the project, the nutrition education program has been implemented by a dedicated team 
managed and supported by WISHH and based in the project headquarters. Analysis of project records reveals 
that FFE2 targets for training individuals in child health and nutrition have been greatly surpassed. 
71% of teachers reported having received nutrition training as part of the project (compared with 81% at 
baseline)50
. Of the modules which make up the training, 71%teachers reported receiving training in hygiene (76%
at baseline); 64% in food groups (72% at baseline) ; 74% in Hygiene and food storage (69% at baseline); 55% in 
Handwashing (72% at baseline); 53% in Balanced diet (70% at baseline) (see Technical Appendix 2, pp 240 – 150). 
50 The goal (and assumed achievement) of the nutrition education of teachers was for 100% of all teachers in 
the 4 districts to receive training. The discrepancy between this and the 71% who reported receiving training 
may be accounted for by the nutrition education team failing to reach all teachers in all schools or, alternatively, 
by surveyed teachers failing to recall receiving nutrition education or failing to recognise the training received 
as nutrition education training (possibly because they consider “training” to be a more formal activity than the 
one they received.
110
Teachers were asked with whom they had shared the information learned during nutrition training. 87% claimed 
to have shared it with their pupils (compared with 76% at baseline); 29% with parents and guardians (43% at 
baseline); 41% with their own families (47% at baseline); 67% with the school council (37% at baseline); 34% 
with the community (34% at baseline) (see figure 27 below and Technical Appendix 2,pp 250 – 61). 
Figure 27: Teachers’ reporting of sharing nutrition education 
98% of those who replied to the question considered that the training had brought about changes in their own 
lives; 90% considered it to have brought about changes in the lives of their pupils. 
89% schools (81% at baseline) reported receiving nutrition posters from the project; 81% schools (6% at baseline) 
reported receiving pamphlets; 87% reported receiving school feeding manuals (0% at baseline); 3% of school 
reported not receiving any nutrition materials. These results reflect the strong push by the nutrition team to 
distribute materials to schools as part of their exit strategy and to promote sustainability of the nutrition 
education component before their imminent withdrawal. Schools overwhelming reported that these materials 
were being used: 91% of schools reported teachers using them in their lessons (84% at baseline); 56% that they 
were displayed on the wall of the kitchen, storeroom or teachers’ room (79% at baseline); 26% by being 
distributed to pupils (72% at baseline) and 10% that they were being distributed to the community (54% at 
baseline). These results reflect the shift in emphasis during the final stage of the nutrition education project to 
promoting use of materials by teachers in the classroom. During school visits, the evaluators frequently saw 
nutrition education materials distributed by the project either displayed on walls or being used in the classroom. 
The materials were of good quality, attractively presented with clear messages; they were robustly produced on 
laminated cardboard for durability. One item not referred to in the survey but observed during school visits was 
0%
10%
20%
30%
40%
50%
60%
70%
80%
90%
100%
Pupils Your family Parents &
guardians
School
council
Community No response
With Whom Teachers Shared Nutrition Information
Baseline Midterm
111
an attractively produced recipe book containing pictures and simple instructions, giving simple, healthy, 
culturally appropriate recipes using recognizable, locally grown products. 
The teacher survey included questions designed to evaluate teachers’ knowledge of nutrition. Teachers were
asked to provide examples of the different food groups used within the nutrition education program (base foods, 
protein-rich foods, foods rich in vitamins/minerals, energy-rich foods). Overall teachers performed well on this 
task, revealing a good level of assimilation of the nutrition education content. In the case of base foods, 90% of 
teachers were able to give two appropriate examples; 9% were able to give one appropriate example and 1% 
were not able to give any appropriate examples. In the case of protein-rich foods, 80% of teachers were able to 
give two appropriate examples, 16% to give one and 4% to give 0 appropriate examples. In the case of foods rich 
in vitamins/minerals, 85% of teachers were able to give two appropriate examples, 10% to give one and 4% to 
give 0 appropriate examples; in the case of energy-rich foods, 75% of teachers were able to give two appropriate 
examples, 17% one example and 8% to give 0 appropriate examples. In summary, between 75% and 90% of 
teachers surveyed were able to provide two examples for each food group; at baseline, this was the case in 
between 61% and 85% of cases (see figure 28 below).For more details, see Technical Appendix 2, pp 196 – 214 
and 262 – 366. 
Figure 28: Teachers’ ability to provide examples of different food groups
In order to triangulate the assertion by teachers that they share the nutrition training they have received with 
their students, pupils were asked whether their teachers talk about nutrition during school. Of the intervention 
group pupils, 9% claimed their teachers “often” talked about nutrition during classes (7% at baseline); 27% 
claimed they “sometimes” did (18% at baseline); 58% claimed they did not (53% at baseline); and 6% did not 
respond to this question (22% at baseline) (see figure 29 below). Although students were slightly more likely to 
0%
20%
40%
60%
80%
100%
120%
Baseline:
Basic
foodstuff
Midterm:
Basic
foodstuff
Baseline:
Protein￾rich
Midterm:
Protein￾rich
Baseline:
Vit/min￾rich
Midterm:
Vit/min￾rich
Baseline:
Energy-rich
Midterm:
Energy-rich
Teachers Able to Provide Examples of Different Food Groups
2 examples 1 example No examples No response
112
report their teacher talking about nutrition at midterm than at baseline, there is something of a discrepancy 
between these results and teachers’ reporting that 87% shared nutrition education information with their pupils. 
Figure 29: Pupils’ reporting of teachers talking about nutrition during school 
Where training content is shared, this may be rather formal and theoretical, rather than practical. One school 
director, when asked how the information received in nutrition training was shared replied:
First they go to the seminar and then we meet and they teach us that…/… we sit down together in the 
classroom and they start to teach us what they learned there. 
Positive developments in the nutrition education identified in pervious evaluations appear to be sustained at 
midterm of the second phase (better collaboration and information sharing between the nutrition education
team and the rest of the project team, increased capacity, confidence, record keeping and reporting skills within 
the nutrition education team, more alignment between M&E of the nutrition education and the wider project). 
As mentioned above, the recent materials produced by the nutrition education team are of good quality and 
have been well received and appreciated by teachers and school directors. The midterm evaluation revealed a 
coherent exit strategy, with project staff, schools and communities being prepared to take responsibility for 
sustaining the work and achievements of the nutrition component and priorities being identified for the 
remaining months of the intervention, in order to maximize sustainability of the achievements of the nutrition 
education program once this has closed. 
The school feeding focal point of one of the SDEJT described the training received in how to use the nutrition 
materials provided to schools:
0%
10%
20%
30%
40%
50%
60%
70%
Often Sometimes Never No response
Pupils Report Teachers Talking About Nutrition
Baseline Midterm
113
This year we were trained in the use of the manual which has been produced, the manual to support 
nutrition education in primary schools. It was very interesting, it’s more connected to the pedagogical 
area because there is a connection between the subjects in the manual and those in the curriculum. That 
manual can be consulted by the teacher according to the thematic unit they want to teach in class… 
there is a lot for the teacher to read and to reproduce, implement in their classroom, in their lessons.
Evaluation of the impact of the nutrition education program within the EPFs was rather less reassuring and there 
is some doubt about its sustainability, at least within some EPFs, after the end of the program. Interviews with 
the teacher trainers responsible for nutrition education and with the directors and pedagogical directors in the 
EPFs visited revealed a certain lack of dynamism, which was confirmed by the nutrition education team 
themselves. EPF teachers appeared to consider that, with the formal end of the nutrition education program, it 
will no longer be possible to continue to maintain the culture and content the program has sought to disseminate 
within the EPFs51
. In particular, the expectation that there would be no further workshops or face-to-face contact 
with the nutrition team left some EPF teachers and nutrition focal points feeling abandoned and disempowered
since they would no longer have resources provided by the project to work with. 
This is in marked contrast to the vision expressed by WISHH, that nutrition education should hereafter be an 
integrated part of the EPF curriculum delivered to all trainee teachers, so that they are able to incorporate good 
quality nutrition education into their own classroom practice. As discussed with the nutrition education team, it 
is recommended that there should be a final series of face-to-face meetings with the EPF focal points and 
pedagogical directors with the explicit objective of transmitting methods and processes to sustain the nutrition 
education programs within the EPFs once the program closes. In particular, there should be a focus on practical, 
knowledge-based activities and methods, which do not require expensive resources to implement, and on 
inculcating within the EPFs a culture of promoting nutrition education, integrated with nutritional behavior, so 
that the theoretical and the practical dimensions reinforce one another.
51 The nutrition education team report that this issue has been addressed and the EPFs have agreed to integrate 
nutrition education into their teacher training programs across the board. However, this did not occur until after 
the midterm evaluation was completed.
114
Recommendations
The nutrition education program under WISHH managed team members is shortly to close, with the nutrition 
education program due to continue to the end of FFE2 through transfer of management to the FFE2 leadership. 
A number of recommendations are offered for the time under WISHH management and thereafter: 
Ensure that, in addition to knowledge of nutrition, school focal points have practical examples of how to share 
this in order to provoke positive changes in nutritional behavior.
Hold a final training session with EPF focal points under WISHH management aimed at ensuring continuity and 
ongoing transmission of information, through low-cost, knowledge-based activities and the creation of a 
dynamic culture. Dispel the myth that knowledge transmission requires financial resources.
Ensure that any remaining training is focused on practical application and producing change, rather than
theoretical knowledge.
To promote the use of the excellent nutrition education materials designed, produced and distributed in schools, 
show focal points, head teachers and others how the materials can be used in schools. Give very practical advice 
on how they can be incorporated into lessons and the life of the school or community. 
Percentage of school-age children receiving a minimum acceptable diet. Final target: 39% (female: 
39%; male: 39%)
Midterm target: 34% (female: 34%; male 34%)
Midterm results (March 2019): 93% (female: 93%; male 93%)
Final target 274% achieved 
As part of the student survey, students were asked what they had eaten the day before, in an attempt to 
establish how many can be considered to receive a minimum acceptable diet52
. The nutrition education team 
were asked to code their responses as 2 “well balanced, i.e. all four food groups were mentioned”, 1 “reasonably 
well balanced but lacking in at least one essential food group” or 0 “not balanced/only one food group
mentioned”. No students’ responses were coded as 2; 93% of students’ responses were coded as 1; 7% were 
52 Establishing whether students are receiving a balanced diet through a simple survey is challenging. It was 
decided that the best way to gather information relative to this indicator was to ask the students what they 
ate the day before. This is not necessarily a reliable way of gathering information about diet, since recall may 
be poor (informants do not necessarily remember what they ate the previous day); furthermore, ideally 
responses should be followed up to prompt more complete recall, using follow up questions or pictures. 
However, in the present case, given the limited time available and the range of information to be collected, 
questioning was kept to the simplest form: “Can you tell me what you ate yesterday?” 
115
coded as 053
. Students whose responses were coded as “reasonably well balanced but lacking in at least one 
essential food group” were therefore considered to have a minimum acceptable diet. However, caveats apply 
(see discussion in footnotes). 
Number of school gardens further developed and maintained. Final target: 60
Midterm target: 60
Midterm results (March 2019): 60
Final target 100% achieved 
Number of school children benefiting from school gardens. Final target: 20,000
Midterm target: 20,000
Midterm results (March 2019): 19,929
Final target 100% achieved 
School gardens have been encouraged and supported since the inception of the project. In addition to the 60 
school gardens officially supported by the project plan, other project schools have school gardens.
The school survey asked schools whether they had a school garden or larger “machamba” without 
differentiating. Overall, the responses show a small but fairly consistent degree of progress since baseline 
(assuming that the sample of 170 schools surveyed in representative of the total number within the project). 72 
of the 170 schools surveyed (42%) reported having either a school garden or “machamba.” Of these, 31% (16%
at baseline) report growing food once a year, 10% (7% at baseline) twice a year, 14% (10% at baseline) all year 
around and 4% (1% at baseline) three times a year. 42% (23% at baseline) report only growing food during the 
rainy season. The productive area varies: 4% of schools (1% at baseline) report having between 3 and 5 hectares, 
13% (3% at baseline) having between 1 and 2 hectares, 51% of schools (20% at baseline) having between 0.25 
and one hectare and 32% (29% at baseline) having less than 0.25 hectares.
In some cases, the food grown regularly contributed to the school feeding. 40% (46% at baseline) of schools 
reported being able to show evidence of how the food produced was consumed; 49% (54% at baseline)report
53During nutrition education, participants are taught that a balanced diet should contain items from each of 
four essential food groups. Students’ responses were only coded as 2, if they mentioned an item from each of 
the four food groups. A flaw in this system is that only three of the four food groups are “foods” per se (base –
grains/starches, protein sources, and micronutrient sources – fruits/vegetables). The fourth food group is 
“energy” and is considered extra as it represents oil, other fats, and sugars. Unsurprisingly, students recalling 
what they ate the previous day did not mention items in the fourth group, meaning that none of them were 
coded as having an “adequate diet” in this scheme: no students scored as “2” because none mentioned 4 food 
groups. It is recommended that at end point, a more realistic coding system would be: 2 for responses that 
mention 3 food groups; 1 for responses mentioning only 2 food groups; 0 for responses mentioning zero or 
one food group. If this system is used at end point, the midterm responses will need to be recoded to ensure 
comparability over the two time points. 
116
they have no evidence and 11% do not know. Where evidence exists it may be in the form of: a register of school 
production (38%, compared with 38% at baseline), a document provided by the SDEJT (10%, compared with 33% 
at baseline), it may be registered in the CSB+ distribution report (21%,compared with 6% at baseline), in a garden 
record book (10%, 0% at baseline) or another form (21%, compared with 10% at baseline).
Of the schools with gardens, 61% report receiving seeds from the project (62% at baseline). Of these, 16% 
received one type of seed, 16% received two types, 25% received three types and 43% received more than three 
types. 60% of schools with gardens (42% at baseline) reported receiving garden tools, such as hoes, rakes and 
machetes: 12% report receiving one type of tool, 40% two types, 30% three types and 19% report receiving more
than three types of garden tool. 
7% of the gardens are irrigated using river water (18% at baseline); 33% using rainwater (34% at baseline); 49% 
using water from a water fountain (42% at baseline) and 11% from another type of water source, such as a 
manual pump or tap water (7% at baseline).In 65% of cases this water source is available all year round (52% at 
baseline); in the other 35% it is seasonal (48% at baseline). Of the schools where water is only available in some 
seasons, 80% only cultivate food during those seasons: 20% manage to grow food all year around. 61% of the 
school gardens are fenced (compared with 46% at baseline). 
Only 13% (compared with 6% at baseline) of schools with gardens have so far obtained a DUAT, the 
administrative document authorizing them to cultivate the land. 
As far as community involvement in school gardens is concerned, in 31% of cases the community helps regularly 
(at least once weekly) with preparation of the ground and production of food (24% at baseline); in 42% of cases 
the community helps several times per year when there are large-scale activities (39% at baseline) ; in 26% of 
cases the community is not involved (37% at baseline). (See Technical Appendix 3, pp 34 -65).
51% of schools report growing legumes (31% at baseline), 78% report growing vegetables (41% at baseline), 49% 
report growing tubers and root crops (29% at baseline), 7% growing oil seeds (4% at baseline), 25% growing 
cereals (25% at baseline) and 13% fruit trees (10% at baseline), with 7% reporting other crops are grown. 
(See Technical Appendix 3, pp 205 - 219).
The school survey included questions about the types and quantities of food grown in the past two years. The 
responses are summarized in table 14 below and compared with those reported at baseline54. Although at 
54 Since the questions asked about food grown “in the past two years” it is possible that there may be some 
double reporting of food grown: the baseline survey covered years 2016 and 2017, whereas for the midterm 
data collection, which was conducted in April 2019, crops grown during the second half of 2017 may have been 
counted again. This does not necessarily greatly affect the patterns observed, since all schools were surveyed 
under the same conditions, and responded to the same questions. 
117
midterm only 170 schools were surveyed, 72 of which reported having school gardens (compared with 156 of 
the total of 270 surveyed at baseline) comparison of the two time points suggests that:
• more schools are growing a more diverse range of products;
• the average quantities of these products grown by schools is increasing.
Although the number of schools producing quantities of food which can make some contribution to school 
feeding and diversification of the menu, at least for occasional meals is increasing, school gardens do not have 
the capacity to contribute to daily school feeding on a regular and sustainable basis.
Schools 
report 
growing 
(%)
Schools 
report 
growing 
(%)
Meanqua
lity (kg)
Mean 
quantity 
(kg)
Number of schools 
producing 1-50 kg
Number of schools 
producing 51-100 kg
Number of schools 
producing 101-500 
kg
Midterm Baseline Midterm Baseline Midterm Baseline Midterm Baseline Midterm Baseline
Legumes 51% 31% 102 kg 52kg 31% 40% 10% 7% 11% 7%
Vegetables 78% 41% 109 kg 70kg 47% 40% 7% 20% 24% 10%
Tubers / roots 49% 29% 102 kg 75kg 26% 36% 13% 5% 10% 8%
Oil seeds 7% 4% 55 kg 47kg 4% 7% 1% 0% 1% 1%
Cereals 25% 25% 95 kg 70kg 15% 25% 6% 10% 4% 5%
Fruit trees 13% 10% 68 kg 114kg 8% 7% 1% 4% 3% 6%
Table 14: school reporting of produce grown in the past two years at Midterm and Baseline
Visits to schools during field work confirmed that school gardens are extremely dynamic in some cases, 
contributing to school feeding, nutrition education and other aspects of education such as natural sciences and 
mathematics. Those schools with active school gardens invariably have a director, pedagogical director or named 
teachers responsible for the garden activity, who is passionate about food production and gardening and highly 
motivated. One such school director explained that the secret of having a successful school garden in to involve 
everyone in both the work and the consumption: 
When we began our little garden, the things we grew were for everyone, for us all. We sat down 
together, and we said we’re going to increase our production. Everybody is going to be involved. When 
we have a harvest it’s not only for the leadership, it’s for everyone. The teachers got something to eat, 
everyone got something to eat, and they all felt involved. That was what motivated them, inviting 
everyone to participate. 
These individuals reported that produce from the garden is used to supplement the CSB+ to improve the meals 
given to students, particularly on special occasions and festivals, when festive meals are prepared for the whole 
school or for the school and wider community. In a small number of cases they reported being able to regularly 
use manioc or sweet potatoes instead of CSB+, for example once a week. One school director explained:
118
Yes, as well as the soya we have cornmeal porridge (xima), from time to time with curry, we have, now 
we’re going to have manioc with lettuce salad, yes, we’ve cooked maize, the mums are going to pound 
the maize to make cornmeal porridge.
They also enthusiastically reported receiving seeds from the larger HGSFGs and technical advice from the 
technicians attached to these. These visits confirmed the role school gardens can play in school feeding and 
nutrition education; however, it is clear that the relatively small-scale gardens in most schools, cultivated by 
teachers and pupils, do not have the potential to produce enough food to feed students regularly. 
Number of Home-Grown School Feeding Gardens established. Final target: 8
Midterm target: 5
Midterm results (March 2019): 7
Midterm target 140% achieved 
Number of school children benefitting from Home-Grown School Feeding Gardens. Final target: 
6,600
Midterm target: 2,750
Midterm results (March 2019): 1,792
Midterm target 65% achieved 
Home Grown School Feeding Gardens (HGSFG) are an innovation of FFE2 and aim directly to address the 
question of sustainability of school feeding into the future, after the end of the project activities. They are based 
on a vision of local production on a sufficiently large scale to support up to four schools and the principle that 
the surplus production can be sold to finance technical help and the purchase of foodstuffs such as salt and 
cooking oil needed to produce meals from the food produced. They aim to combine production of staple foods 
such as manioc and sweet potato with legumes, vegetables, salad and fruit; each HGSFG raises small animals 
(currently ducks). 
At the time of midterm field visits, 7 HGSFGs were in operation; the external evaluator visited 6 of these55. It had
originally been intended that all eight HGSFGs should be operational by 2019. However, their establishment was 
more challenging than anticipated. One HGSFG had had to be moved from its original location due to deviation 
of the river on which it depended for irrigation by unregulated extraction of sand from the river bank, with the 
result that it was no longer possible to irrigate the plot. Having to concede that the garden was not viable was a 
great blow to the school and community which had invested in its initial preparation; at the time of fieldwork 
the motor pump and other resources were being removed to be installed in another site linked to another 
school. 
55 As of October 2019, the evaluator is informed that all 8 HGSFGs are now in operation.
119
During field visits, the evaluator was impressed by the scale of the HGSFGs and the extent of progress already 
achieved. Considerable areas have been cleared and planted, with the help of community members, who 
contribute to production for the school and in return receive an area benefiting from the same irrigation system 
to cultivate for their own consumption. At the time of field visits (April 2019), diverse crops were being 
produced: base foods, such as manioc, maize, plantains and sweet potatoes; different types of bean, legumes 
and tubers; vegetables and salads: lettuce, tomatoes, onions, cabbage, peppers; and fruit trees: mango, oranges, 
grapefruit, papaya, bananas. Each HGSFG visited also had a recently arrived flock of ducks, which were being 
carefully raised and tended. 
Each HGSFG is based in a school and is the primary responsibility of the head teacher, with the technical work 
being supervised and coordinated by agricultural technicians. In each HGSFG visited, the school director and the 
technician were interviewed. The external evaluator was extremely impressed by the knowledge, commitment 
and dynamism of the technicians. These individuals had previously worked as agricultural extensionists in other 
ADPP projects, such as farmers’ clubs. They demonstrated extensive knowledge of both agricultural techniques 
and community development. They work cooperatively alongside the community volunteers, showing them 
improved agricultural methods and practices but also being prepared to use traditional methods where these 
are preferred. One technician showed how he was engaged in a friendly competition with the community 
volunteers: the community had planted several rows of vegetables using their traditional approach and he had 
planted several rows alongside them using a modern method assumed to be more efficient; the harvest would 
demonstrate which was the winner. These knowledgeable individuals work long hours alongside community 
workers, leading by example, in addition to planning, keeping records, and liaising with the school. Their role 
appears key to motivating and engaging the community volunteers. In each HGSFG they are training one 
community member in particular to become the future coordinator. The technicians are also supporting 
neighboring school gardens, providing seeds from their production and technical advice. 
The school directors demonstrated the record keeping methods used to record and account for the production. 
The records appear well kept, although it is not clear whether they include all the food produced. This raises one 
important issue regarding the management and governance of the HGSFGs. Head teachers, who have been 
trained to manage a school (in the best of cases - not all school directors in Mozambique have actually been 
trained as such)are not necessarily well qualified and equipped to manage fairly large scale agricultural
production and to account for it rigorously. This requires significant skills, time, probity and influence within the 
community. It is not always easy to account for food produced; when large qualities of, say, lettuces are 
produced, as observed during field visits, who is to say who can take a lettuce (or several) home? This question 
is all the more pertinent in a context of relative food insecurity, where many families struggle to eat, let alone 
to eat a balanced diet. It is, however, a question which will need to be addressed if the HGSFGs are to succeed
and maintain community adhesion and support. The evaluator learned of one school where a crop of potatoes 
120
cultivated by the community and the school had been taken to market and sold without the profits being 
returned to the school. This had demotivated the community members who had worked to grow them, who 
were no longer participating in the work of the HGSFG. It is clear that rules and systems governing the utilization 
and consumption of the produce need to be strengthened and the recording of this needs to be improved.
Evaluation fieldwork revealed many positive aspects of the new HGSFGs, as well as some which will need to be 
closely monitored if the initiative is to be successful in the long term.
The creation of large-scale, completely fenced market gardens, with motor pumps, irrigation systems and a 
reliable water source is a huge achievement in a challenging environment, as is the fact that these are already 
yielding considerable quantities of produce and vegetables and, in some cases, contributing to the diversification 
of school feeding. 
Recommendations
In the future, once the project is no longer distributing CSB+, an alternative base food, possibly sweet potatoes,
will need to be produced or procured, in sufficient quantities to continue daily school feeding. 
The existing plans and systems need to be reinforced to ensure the HGSFGs are well managed and sustainable. 
The food produced needs to be consistently recorded and accounted for, using transparent processes.
The agricultural technicians who currently play such an active role in coordinating and doing much of the work 
will either need to be retained and paid or their place taken by the community members currently being 
prepared to replace them. In this case, these individuals will need to develop considerable competence, 
including in production management, maintenance of the motor pump and other equipment, and the ability to 
mobilize their fellow community members and coordinate their activities. 
Current levels of commitment from schools and communities will need to be sustained and, in some cases, 
increased.
The inclusion of small livestock rearing in the HGSFGs is an impressive innovation and offers the potential for 
greater diversification of diet and income generation to finance some overheads, such as fuel for motor pumps, 
fencing, tools, salaries, etc. Those responsible for the livestock should be made aware of principles of animal 
welfare. Ducks, for example, need constant access to water. 
Overall, at present, the HGSFGs are evidently making a significant contribution to the diversification and 
sustainability of school feeding, at least in the school where they are based and which they support. Once the 
support from the project finishes, they will need to be self-sustaining. This will be a considerable challenge, one 
which is certainly not impossible to meet, but which will require considerable leadership and good management 
in order to succeed. 
121
Access to Clean Water and Sanitation Services (2.4)
The project recently divided the water and sanitation component into two; the sanitation activity now comes 
under the responsibility of the construction coordinator. This leaves the water team available to address the 
challenges and demands of this delicate domain. The report first addresses the activities of the water sector, 
before reporting on sanitation activities. 
Access to Clean Water
Indicator 22 (Output): Number of schools using an improved water source. Final target: 264
Midterm target: 226
Midterm results (March 2019): 270
Final target 102% achieved
Number of school children benefitting from clean water sources. Final target: 54,000
Midterm target: 54,500
Midterm results (March 2019): 87,847
Final target 163% achieved 
Number of water tanks secured. Final target: 330
Midterm target: 330
Midterm results (March 2019): 295
Final target 89% achieved 
Access to safe water continues to be a condition for school feeding, local production of food, health and 
wellbeing; despite the water team’s best efforts, this also continues to be a challenge for some project schools, 
due to the depth of the water table, the saline quality of much of the subterranean water and the distance of 
many communities from the existing municipal piped water network. In many schools, the absence of safe water 
some or all of the time is a major problem. One school director interviewed painted a desperate picture of a 
school without water, where children and teachers cannot drink during the day, there is no water for 
handwashing after using the latrines and teachers who spend the night at the school cannot wash before work. 
Of course, the preparation of the soya porridge also requires water. One school director explained:
The big challenge is water, because children have to bring water from home [to make the soya porridge]. 
Often they don’t have water at home either. 
The water team is now stable, compared to phase one of the project where it was difficult to recruit and retain 
water technicians; all four water technician posts are now filled, and the same committed and competent 
coordinator leads the team. In partnership with the State, some schools who can afford to pay water bills are 
being connected to the piped water network. Progress has been made in decentralizing funds for small water 
interventions, which are now administered at district level, allowing more rapid repairs and maintenance. At 
122
this stage in the project, with earlier organizational problems resolved, it seems that the challenges to do with 
ensuring schools have access to clean water are more to do with the intrinsic difficulties of finding technical 
solutions in the challenging environment in which the project operates. Unfortunately, a number of contracts 
for drilling boreholes have not been honored, meaning new contractors need to be identified and new contracts 
issued. The water coordinator explained that some of the theoretical solutions to providing schools with water, 
such as connecting them to the mains supply or delivering water by tanker trucks to fill up the water tanks which 
all project schools have received are not possible within the current budget. Furthermore, having water standing 
in tanks for long periods of time is a health and hygiene risk. 
There continues to be a disconnect within the reporting of water team activities between the numbers of 
interventions reported and the number of functioning water systems. This is confirmed by comparing the figures 
form project records, which show targets as surpassed with the number of respondents to the midterm school 
survey who stated their school did not have a water system or did not always have safe water. 
Recommendations of earlier evaluations to develop clearer criteria for considering water installations as 
completed and therefore “countable” and more focus on outcomes (functioning water systems) have not led to 
this reporting discrepancy being completely resolved. 
The FF2 baseline report carried out a thorough inventory of the different types of water installation at each 
school, which it was not considered necessary to repeat at midterm. Of the schools surveyed at midterm, 74% 
report having have some sort of system to supply water: piped mains water, electric water pumps, manual water 
pumps and/or rainwater harvesting (compared with 94% at baseline); 26% do not (6% at baseline).Of those that 
reported having a water system, 73% reported the system functioned “every day this year” 15% that it 
functioned “most of the time”, 7% that it functions “from time to time”, and 4% that it “doesn’t work”.
57% of these schools reported having a plan for the routine maintenance and cleaning of their water system
(58% at baseline). This plan was known variously to the school council (54%), named teachers (61%), project 
staff not connected to the school (31%), the school director (38%), the pedagogical director (24%) and/or other 
people (24%). 
68% reported having a water committee or designated people responsible for looking after the water system
(69% at baseline). The composition of the water committee included the teacher responsible for the storeroom 
(43%), teachers from the school (33%), project staff not connected to the school (19%), the school director (33%), 
the pedagogical director (20%) and/or other people (38%). Those who actually performed the maintenance 
included: school council members (47%), members of the neighboring community through the school council 
(37%), designated teachers (37%), project staff not connected with the school (29%), pupils from the school 
(10%) and/or others (10%). For more details, see Technical Appendix 3, pp - 131 - 172.
123
When asked how often the water system was cleaned, responses varied between daily (25%, compared with 
27% at baseline), weekly (17%, compared with 29% at baseline), twice weekly (6%, compared with 10% at 
baseline), “from time to time” (40%,compared with 35% at baseline) and “other” (13%). 
A source of clean water continues to have a transformative effect on the schools and communities to which it is 
brought. Unfortunately, several schools still do not have such a water source or only have water during part of 
the year when rainwater harvesting is possible. The way the indicator is formulated may hide this reality. 
Access to Sanitation Services
Indicator 23 (Output): Number of schools with improved sanitation facilities. Final target: 264
Midterm target: 264
Midterm results (March 2019): 246
Final target 93% achieved 
Number of hand washing facilities rehabilitated. Final target: 171
Midterm target: 171
Midterm results (March 2019): 140
Final target 82% achieved 
Number of latrines and hand washing facilities constructed. Final target: 94
Midterm target: 94
Midterm results (March 2019): 108
Final target 115% achieved 
Number of educational facilities (i.e. school buildings, classrooms and latrines) 
rehabilitated/constructed as a result of USDA assistance. Final target: 808 (kitchens 264; 
storerooms 264; latrines 280 [94 constructed + 186 renovated])
Midterm results (March 2019): 862 (kitchens 194; storerooms 410; latrines 235 [108 constructed + 
127 renovated])
Final target 107% achieved 
Number of educational facilities (i.e. school buildings, classrooms and latrines) 
rehabilitated/constructed as a result of USDA assistance. Final target: latrines 94 constructed + 
186 renovated = 280
Midterm results (March 2019): latrines 235 [108 constructed and 127 renovated]
Target 84% achieved 
Number of school children benefiting from latrines and hand washing facilities. Final target: 
48,000
Midterm target: 48,000
Midterm results (March 2019): 78,311
Final target 163% achieved 
124
During FFE2, the installations which were constructed in all schools during the first phase (kitchens with firewood 
saving stoves and storerooms) are being maintained and renovated, including improving their ventilation. A
large program to build and renovate improved models of latrines, all with handwashing facilities is also
underway. 
Analysis of project records reveals that the target for construction of handwashing facilities and of latrines has 
already been reached, and the rehabilitation of existing handwashing facilities and latrines is slightly behind 
schedule. The building component which is now responsible for construction and rehabilitation of latrines and 
handwashing facilities was without a manager for some months, which led to delays in implementation of the 
plan. At the time of fieldwork, a new manager had just been appointed a week previously and was still learning 
the details of the component. He had previously been the coordinator of one of the project districts, so had 
good knowledge of the component but was still learning some of the management processes and aspects such 
as procurement. Like the other new appointments, he will need to be closely supported and mentored in this 
role initially and possibly receive specific training. With a new manager in place, it should be possible to catch 
up with the delays in latrine renovation and meet the targets without too much difficulty by the end of the 
project.
A major priority for the remainder of the project is to ensure communities and schools are able to maintain the 
infrastructure after the close of the project: 
We have always had close contact with the school council and the school leadership and we remind 
them that the assets need to be looked after and maintained, because that infrastructure doesn’t belong 
to ADPP but to the school and the community for the benefit of the children. There is a continuous 
mobilization to get the message across that after the project finishes, it is they who will need to continue 
and maintain the infrastructure and everyone needs to play their part.
An example of this need was encountered during visits to some schools, where the external evaluator was told 
by school directors that some of the latrines renovated by the project several years previously were now full and 
could no longer be used. This is a good example of the need to look beyond targets and respond to situations 
encountered. 
Recommendation
A plan is needed to identify those latrines which are likely to become full in the near future and work with the 
schools involved to take steps to either cover them over and replace them with others or to empty them where 
this is possible. Even where latrines will not become full for some time, it would be good practice as part of the 
sustainability plan to raise this issue with school councils and provide training so that they can manage the 
situation when it does arise. 
125
Access to preventative health interventions (2.5)
Indicator 24 (Output): Number of students receiving deworming medication(s). Final target: 74,000
Midterm target: 74,000
Midterm results (March 2019): 91,103
Final target 123% achieved 
Number of cooks tested for TB. Final target: 5,280
Midterm target: 5,280
Midterm results (March 2019): 8,402
Final target 159% achieved 
The project has provided logistical support to the local health services to provide de-worming tablets to all pupils 
once annually. Analysis of project records reveals that this target has been surpassed as of midterm. 
When asked whether they had undergone deworming during the current school year, 65% of students in project 
schools reported that they had (compared with 77% of the cohort at baseline). This lower result at midterm may 
be due to the fact that the midterm survey was administered towards the beginning of the school year (March 
– April 2019). 37% of comparison group students reported having received deworming tablets during the current 
school year (compared with 63% at baseline) (see figure 30 below). The lower figure at midterm than at baseline 
for the comparison group may also be due to the timing of the midterm evaluation. However, given the 
significant difference between the rates reported by students in project schools and in comparison schools, it is 
likely that this is impacted by the fact that the deworming activity is subsidized in project districts, by 
contributions made by the project to the fuel costs; in some cases, project staff also accompany the deworming 
visits and promote participation in this activity to students and their families. 
Figure 30: students’ reporting of deworming during the current school year
0%
10%
20%
30%
40%
50%
60%
70%
80%
90%
Comparison Intervention
Deworming During the Current School Year
Baseline Midterm
126
The testing of volunteer cooks for tuberculosis was introduced during the second phase of the project in order 
to protect students from transmission of the disease. Final targets for this activity have been exceeded as of 
midterm. 
Health officials from the SDSMAS (district health services) welcomed the collaboration between the health 
services and the project and, in particular, the fact that volunteer cooks were undergoing testing for TB. 
Access to requisite food preparation and storage tools and equipment (2.6)
Number of community volunteers receiving laundry soap. Final target: 5,280
Midterm target: 5,280
Midterm results (March 2019): 5,420
Final target 103% achieved 
Number of schools receiving dish washing soap. Final target: 264
Midterm target: 264
Midterm results (March 2019): 271
Final target 103% achieved 
Distribution of laundry soap to volunteers was a recommendation made by evaluations of FFE1, following 
repeated requests from volunteers for soap to help them keep their uniforms clean. Likewise, volunteers 
requested dish washing soap to improve the cleanliness and hygiene of the CSB+ preparation and subsequent 
washing of dishes and utensils. Analysis of project records reveals that the targets for these activities have been 
met. During fieldwork, volunteer cooks expressed satisfaction at having been given soap, which is seen as more 
than an incentive and rather as a necessity for their work and therefore a recognition of their status within the 
project and contribution to the task of school feeding. The motivation and commitment which result from this 
relatively small incentive can be considered good value for money. 
Capacity of government institutions (2.7.1)
Number of government officials trained in nutrition. Final target: 100
Midterm target: 65
Midterm results (March 2019): 302
Final target 302% achieved 
Number of participants in school feeding conferences (district, provincial and national). Final 
target: 180
Midterm target: 102
Midterm results (March 2019): 1,208
Final target 671% achieved 
127
Number of seminars and meetings conducted at the local, regional and national level. Final target: 
4
Midterm target: 2
Midterm results (March 2019): 18
Final target 450% achieved 
As part of the commitment to build local capacity and to sustain school feeding and other activities after the 
closure of the project, the project has trained government officials at national, provincial and district levels in 
school feeding. The target for this activity has been greatly surpassed. Government officials interviewed during 
fieldwork were enthusiastic about the training received and stated it had helped them play their role effectively. 
The school feeding focal point of one SDEJT explained:
The trainings are good, they are good because after receiving the training we feel a difference, we 
compare the before and after and we feel when we go into schools that something is improving thanks 
to the training or the seminar. 
The project has also held a number of meetings and conferences to allow actors involved in school feeding in 
Mozambique, including the national school feeding program (PRONAE) and various other initiatives supported 
by NGOs and other actors, to meet and exchange experiences. 
Engagement of Local Organizations and Community output (2.7.4)
Number of local leaders and school council members attending school feeding project 
management training. Final target: 2,000 (female: 500; male: 1,500)
Midterm target: 1886 (female: 471; male: 1,414)
Midterm results (March 2019): 2,498 (female: 689; male: 1,320)
Final target 125% achieved (female: 138%; male: 88%
Since its inception the project has made efforts to train local leaders and school council members. During focus 
groups with community leaders and volunteers, participants stated they had been trained and appreciated the 
value of that training:
We’ve all received training.
Thank you for teaching us and helping us prepare food for the children. We’re very grateful. Come back 
and teach us more any time! 
128
This is important to i) promote ownership of the project by the communities in which it operates ; ii) ensure 
interventions are culturally appropriate and correspond to local needs and objectives; and iii) foster 
sustainability, by ensuring that, after the close of the formal project interventions, local actors have the capacity 
and vision to continue with those activities judged necessary and appropriate. At midterm, analysis of project 
records reveals that targets for training local leaders and school council members have been surpassed. During 
evaluation fieldwork, evaluators were impressed by the level of ownership of the project by local leadership 
structures, who had adopted their own systems and methods for selecting and organizing the rosters of 
volunteer cooks, water committee members and volunteers to work in school gardens and HGSFGs. Interviews 
and informal discussions with community leaders showed a high level of commitment to the project and its 
activities, a determination that it should continue and considerable knowledge and understanding of the 
technical methods needed. Although they do not of themselves guaranty sustainability, these factors are among 
the prerequisites to ensuring the project activities continue after the formal closure of the project.
Transversal themes
In addition to the presentation structured according to the MGD indicators, a number of transversal themes
deserve attention; these cut across the different components and activities and concern the overall operations, 
strategy and vision of the project. They are: human resources, capacity, collaboration and ownership; transport; 
administrative and financial systems and procedures; monitoring and evaluation; and sustainability and 
relevance to the local and national school feeding policy and program environment. Within each of these 
themes, the evaluation revealed both positive aspects and areas of progress since the FFE2 baseline and areas 
which are still evolving, or which should be monitored and given additional support as needed. 
Human Resources, capacity, collaboration and ownership
With the FFE program now in its seventh year of functioning, some staff now have significant experience and 
have demonstrated significant professional development; many staff have been promoted to functions with 
more responsibility. In particular, several of the “professionals” have been promoted to positions of district 
leaders, component manager or leadership. These welcome changes will need to be monitored and 
accompanied to ensure young, newly-promoted staff receive appropriate support and training. 
Progress has been made in terms of gender balance within the project team. Three of the four district 
coordinators are now women; some components are headed by female managers, notably M&E and literacy. 
Some components still lack gender balance; in particular the leadership team is composed of four men, following 
the departure in 2018 of the only female member, running the risk of giving the impression of a “glass ceiling” 
to project staff. 
129
Overall, the “professionals” have greatly grown in capacity, competence and professionalism. In many cases the 
novice professionals recruited seven years ago have grown into impressive, experienced professionals. Various 
“professionals” have developed areas of specialization and are called to apply these in different schools. Their 
work has been reorganized so that, amongst the “professionals” there are focal points for specific areas, such 
as school clubs and school gardens, leading to more efficient and skilled support to schools. 
It is clear, however, that not all “professionals” have the same level of commitment to their work; the evaluator 
was told that some “professionals” do not regularly visit the schools. A well-planned, sensitive performance 
management system might be helpful to motivate the demotivated and spread the workload more equitably. 
The “professionals”’ personal and professional situation has changed over the past seven years56. They have 
families to support and many have to maintain a household at distance, only returning some weekends to see 
their spouses and children. Over the same period, the cost of living has spiraled in Mozambique. Many of the 
“professionals” explained that they are struggling to feed and clothe their children and wider families. Their 
salaries and conditions should be reviewed and improved to reflect this reality, in combination with a 
performance management scheme, so they are able to focus on their professional commitments in the 
knowledge their dependents are properly taken care of. 
Communication and coordination between project staff at all levels (from leadership to “professionals”) with 
the GoM at all levels (national, provincial and district) and with the local communities is excellent. During 
evaluation interviews, GoM and community informants consistently praised the quality of the relationship, 
communication and collaboration with the project, comparing it favorably to that with other partners and 
projects. However, communication and coordination within the project team could be improved in some cases; 
alignment of workplans and working practices, defining work system and procedures and better communication 
between components would improve the efficiency and quality of some operations.
The level of commitment, engagement and effectiveness of the volunteers has improved overall. In general, 
volunteers appear to have accepted the culture of not providing monetary incentives and embraced the 
opportunity to improve their children’s diet and education. At the same time, innovations such as providing dish 
soap and laundry soap have been appreciated and taken as an expression of dignity and a recognition of the
important role played by volunteers. To date, the teachers who play the role of storeroom managers have not 
received any such incentives. 
In order to have as much information as possible about the school feeding environment in Mozambique, the 
evaluation sought to include other organizations currently working in school feeding in the country. One model 
56 A remarkably high number of the “professionals” have been in that role since the beginning of the project. 
This indicative of both the supportive culture of ADPP and the project and the barren employment prospects in 
Mozambique for many young professionals. 
130
being implemented by JAM (Joint Aid Management), in the spirit of their slogan: “Helping Africans to Help 
Themselves” provides an interesting example and food for thought, particularly in light of the discussion of 
ownership and incentives above. JAM is implementing a school feeding program in 48 schools in Sofala and 
Inhambane provinces, which also uses CSB+ as the base food. The 48 schools develop school gardens to teach 
students about nutrition and grow produce to diversify their diet. The biggest difference with the FFE2 project 
is that JAM’s project functions as a value chain. Corn and soya grown by farmers according to JAM’s protocols 
and following training from JAM are sold to a factory in Dondo, Sofala which processes it to produce CSB+ 
(fortified with a premix imported from Holland) which is in turn sold to schools for school feeding. In the initial 
stages, the schools are assisted by JAM with the funds to purchase the CSB+; however, in the long term, it is 
planned that schools should be able to fund these purchases using the proceeds of other income generating 
activities, including the school gardens. There are also plans to sell CSB+ in bags of one kilogram on the open 
market for non-school use. The project is in the early stages but there are plans to scale up to include more 
schools. JAM also provides water and sanitation using solar panels or hand pumps, mobilizes parents in PTA 
structures and nutrition and food security training activities. JAM runs a model farm and learning center for 
farmers, researchers and extensionists and also provides training in financial management. According to the 
project representative interviewed, both the farms growing the corn and soya and the factory transforming 
them into CSB+ are “commercially sustainable”, employing their staff and making a profit. This model of a 
commercial activity, initially supported and subsidized by a supportive partner, offers answers to some of the 
challenges encountered by the FFE2 project, such as the tension between the principle of working with 
volunteers and the very real need of families living in poverty to receive income in exchange for their work and 
time.
Recommendations
Newly promoted staff need to receive appropriate support, training, professional development and orientation.
All appointments and promotions should use equitable processes which promote diversity. When making 
appointments to the leadership team, efforts should be made to identify and encourage female candidates. 
Look at the possibility of introducing a well-designed performance management system for the “professionals” 
to increase motivation and spread the workload more equitably. 
Review and the salaries and conditions of the “professionals” with a view to improving these in conjunction with 
a performance management scheme. 
Improve communication and coordination within the project team, including alignment of workplans and 
working practices, better definition of work system and procedures and communication between components.
131
To increase commitment and motivation amongst storeroom managers, they should receive an expression of 
thanks and some recognition for the central role they play. For example, they could be provided with phone 
credit as compensation for the weekly SMS they send to report the weekly stock levels of CSB+. 
Transport
The acquisition of additional vehicles and motorbikes and the strategy of providing transport to the 
“professionals” and supporting them to obtain driving licenses recommended in previous evaluations has been 
successful. Most of the “professionals” now have driving licenses and their performance and productivity has 
increased accordingly. 
The project still does not have sufficient, appropriate vehicles for all its activities. The components responsible 
for CSB+ delivery, water and construction activities have to juggle their activities in order to share the available 
vehicles, leading to inefficiencies. Hired transporters are unreliable and often their vehicles are not fit for 
purpose. 
The project drivers have a good knowledge of the schools in the four districts and the roads between these and 
play a valuable role in supporting logistical operations. Delays in processing and paying drivers’ per diems and 
expenses puts drivers in difficult situations; where they go out for the day without cash to cover their incidental 
costs, such as road tolls, and meals, they have to resort to borrowing money from their passengers. 
Recommendations
The project needs to acquire additional vehicles in order to undertake its activities in due form. 
Systems to process and pay drivers’ per diems and expenses should be improved to avoid putting them in 
embarrassing situations. 
Administrative and financial systems and procedures
The influence of a strategically focused financial manager with a vision of how systems and procedures can be 
improved has led to improvements in administrative and financial systems and processes and streamlined some 
previously onerous workplace tasks. For example, the creation of a procurement office/purchasing service has 
led to all purchases being processed by dedicated staff. The introduction of a reception desk within the 
Administrative Section through which all administrative and financial requests have to pass, has reduced waiting 
time and made processing requests more efficient. Recruitment and training of additional staff has allowed more 
specialization of activities. Administrative systems could be made even more efficient by further improving 
processes, recruiting additional staff and training existing staff. Staff require continual training to improve work 
132
quality and reduce errors. Additional recruitment would be desirable, for example the project would benefit 
from an HR manager based in the Manhiça headquarters. 
The standard and quality of documentation has improved as staff better understand the need for consistency 
and rigor in preparing and processing documents and that all expenses or requests for funds must be supported 
by the correct documentation. One dedicated member of staff now checks all documentation, leading to more 
consistency. Still more improvements are needed, so that preparing and submitting documentation which is 
acceptable the first time and does not need to be sent back for improvement becomes the norm. 
Some administrative processes and the management of some funds have been decentralized to district office 
level, leading to greater flexibility and responsiveness and reducing delays and bureaucratic bottlenecks. For 
example, the district water technicians have funds for small parts and repairs, so these do not have to be 
processed and authorized centrally, leading to more rapid interventions and flexibility. Staff in the districts 
required ongoing support, supervision and professional development in order for these decentralized systems 
to work effectively. Internal communication and coordination still need improvement; for example, workplans 
need to be aligned and district or school visits coordinated. 
Recommendations
Look for ways to further improve administrative processes, recruit additional staff and train existing staff, for 
example consider employing an HR manager based in the Manhiça headquarters. 
Continue to improve the standard and quality of documentation.
Provide ongoing support, supervision and professional development to staff based in the districts.
Continue to improve internal communication and coordination for example, by aligning workplans and 
coordinating school visits. 
Monitoring and Evaluation (M&E)
High turnover of M&E management and staff throughout the life of the project to date have led to a lack of 
overall vision and continuity in M&E processes and procedures, some gaps in data and a lack of institutional 
memory of previous M&E situations and steps taken. This challenging M&E context has been mitigated to some 
extent by continuous support from the project leadership team and from the distance advisor based at Humana 
People to People Headquarters in Zimbabwe, Stanley Kudzibatira, particularly in preparing semi-annual reports 
and supporting evaluations, and by support from the consultant Cade Fields-Gardner. As of March 2019, a new 
133
M&E coordinator and a new M&E manager have taken their functions; although it is early days there is every 
sign that these individuals have the necessary skills and attributes lead the M&E department effectively and 
gradually make up for the gaps in provision over the years. 
Some M&E activities have been decentralized to the districts; each district now has a data clerk and an 
administrative assistant to perform them. The information flow is now better defined and clearer: information 
from schools is transmitted by the “professionals” to the district staff for data entry. This information is included 
in district coordinators’ reports and sent by them to the project M&E department and project leadership 
simultaneously. Data received from the four districts is processed as necessary by the M&E department and 
transmitted as necessary to Planet Aid; Planet Aid submits semi-annual reports to USDA. 
Despite these improved processes, some of the data collected and entered are still not being used. For example, 
school attendance data is collected from schools using the carbon copies from the specially designed school 
registers; it is then digitized by the district data clerks. However, as of the midterm field visits, this information 
was not being analyzed or used in any way: project school feeding information was being used to report on 
attendance. 
There is still a lack of clarity in the measurement and/or formulation of some indicators. For example, as 
indicated on page 122 above, there is a mismatch between the indicators measuring water system installations 
and those measuring schools with access to safe water systems. 
The imminent arrival of the long-awaited database – the second database commissioned by the project, which 
has been subject to successive delays - should allow better management and vision of information overall. It is 
imperative to invest the necessary time and human resources to get the database up and running as quickly as 
possible, so that this investment can begin to bring real benefits before the end of the project. 
Field visits revealed coherent, orderly and systematic practices of filing and archiving written information / paper 
records by members of the M&E team, who were able to locate and produce information and records as 
requested by the external evaluator as part of her series of spot checks of M&E records.
In addition to this commendable rigorous adherence to procedures, M&E staff need to have a vision of and 
understand the whole process: they should know why information is being collected, what it will be used for 
and how. Component managers and staff also need to have a wider vision of the M&E of their activities: they 
need to know where they are position in relation to project indicators and targets and to be able to use this 
information to plan and prioritize their operational tasks. The project leadership team should also, of course, 
have an overall vision of the project’s real-time situation in relation to the different indicators and targets. 
134
Recommendations
All data collected and digitized should be used in some way; the project should stop collecting and processing 
information which is not being used. 
Review indicators and how these are measured to ensure they accurately capture the information intended.
As a priority, ensure the new database is up and running as quickly as possible.
Ensure M&E staff have a vision and understanding of the whole M&E process (why information is being 
collected, what it will be used for and how); component managers and staff have a wider vision of the M&E of 
their activities (where they are position in relation to project indicators and targets) and can use this information 
to plan and prioritize their operational tasks; project leadership team have an overall vision of the project’s real￾time situation in relation to indicators and targets.
Sustainability and relevance to the local and national school feeding policy and program environment
So close to the closure of the project, the theme of sustainability was a priority: it informed every aspect of the 
midterm evaluation and was an integral part of all instruments used for qualitative data collection and all 
interviews, focus groups and conversations with informants at all levels. The question of sustainability of the 
FFE2 project is intimately connected with the national school feeding program (PRONAE) and also with the 
overall political, economic and social context in Mozambique. 
The National School Feeding Program (PRONAE), which was piloted in a small number of schools Gaza, Manhica, 
Tete and Nampula from 2013, has since been expanded to 70 schools, with support from the WFP.A debt from 
Mozambique to Russia was recently converted into funding for development, within which the PRONAE will 
expand to 300 schools by 2021. This is a welcome development but still only covers a fraction of the national 
demand. The MINEDH representatives responsible for the PRONAE express great appreciation of the FFE2 
project and all it is achieving; they also welcome the openness of the project leadership and staff to working in 
partnership to plan future expansion and reinforcement of the PRONAE and have welcomed the opportunities 
for sharing of experiences and learning provided by the school feeding conferences and workshops organized
by the FFE2 project. It is clear, however that there are many challenges associated with the local procurement 
model embraced by the PRONAE, including the technical and bureaucratic obstacles to local production, 
procurement and transport of food; the difficulty of incentivizing volunteers to work to produce food for school 
feeding without some corresponding reward; and the scale of the food production needed in a context of 
increasingly unpredictable climate, by populations who are ill equipped to adapt to increasing climate variability. 
135
Despite the fact that the duration of the current FFE2 project, due to finish at the end of 2020, is known to all 
project partners, during evaluation interviews, representatives of SDEJTs, the DPEDH and national government 
expressed dismay at the prospect of the project coming to an end and invariably asked whether there was any 
possibility it could continue. While this response might be construed as an unwillingness to take responsibility 
and ownership of a program which has always been destined to be handed over to the GoM and the local schools 
and communities, it also reflects a hard reality. At this moment in time, in additional to the challenges already 
mentioned, with international cooperation aid reduced and high levels of national debt, there is no obvious 
source of funding for a national school feeding program57
. 
At the level of the project itself, the question of sustainability is also complex. On the one hand, the activities 
planned to increase capacity and hand over a series of functioning activities to be continued by schools, 
communities and government institutions are largely on track and have been successful. On the other hand, the 
gaps which will be left once the significant financial contributions and donations of CSB+ by USDA terminate will 
be very hard to fill in the current context. Initiatives such as the HGSFGs are still at an early stage and, while this 
model may in the future make a substantial contribution to the food required to feel pupils in 271 schools once 
each school day, even the most optimistic informants predicted that the current HGSFGs might be able to feed 
children in a small percentage of the project schools and at best two or three times per week. The HGSFGs may 
in the future, with good management, generate enough income from the sale of excess crops to fund the 
purchase of large quantities of a base food, such as sweet potato or manioc, to replace the corn soya blend 
currently playing this role, at least for the schools in their immediate vicinity. However, this income will also be 
needed to pay overheads and possibly salaries. It is clear that, while the HGSFGs may be part of the solution, 
they will not constitute the whole solution. 
Likewise, the literacy program, which has been successful and achieved promising results within the limited 
scope of the initial two- year intervention, is not likely either to be robust enough to be taken over by the 
MINEDH by the end of 2020, nor to be sustainable beyond that period if the financial gap left by the closure of 
the project is not filled, so that the reading coaches can continue to be employed and the other costs of the 
program supported. Despite definite achievements and advances made, other aspects of the project, such as 
the food preparation by volunteers also appear precarious: given the number of feeding days currently lost 
through the absence of cooks, one can only speculate how many of the existing volunteer cooks would continue 
to volunteer regularly, without the regular small incentives and other support the project provides. 
57 Mozambique was obliged to borrow from the IMF to fund the response to the recent cyclones; the IMF did 
not donate funds for this purpose, given the ongoing stalemate surrounding the illegal debt crisis (Hanlon, 2019). 
Furthermore, the government’s ability to respond to the recent Cyclone Idai was a reminder of the limited 
capacity of the GoM to respond to the urgent needs of its citizens. 
136
The picture is therefore complex. On the one hand, the project is proving successful in meeting its targets and 
performing the activities as planned and is producing visible and needed change. On the other hand, this in itself 
no longer appears to be sufficient to achieve sustainability beyond the end of the current project contract. Given 
this reality, two scenarios must be considered: an extension to the current project with an even greater emphasis 
on sustainability, which from the beginning works with the future implementers to ensure a staggered handover, 
including identifying future funding sources and a costed business model able to be self-sustaining ; or an exit 
plan which uses the remaining sixteen months of the project to work closely with the future implementers of all 
components to establish priorities and determine what can realistically be continued and put plans in place to 
achieve this, working within greatly reduced financial parameters. 
Conclusions
Internal validity (did the project do what it said it would do?)
The midterm evaluation revealed that the majority of activities planned have been implemented as intended, 
often in the face of considerable obstacles, and most midterm targets have been met or surpassed, with a 
number of final targets already met a midterm. This is a testimony to the hard work and organizational ability of 
a large number of staff and leaders from Planet Aid, ADPP Mozambique and the FFE2 project. In the relatively 
small percentage of cases where targets have not been met, the report analyzes why and, in most cases, suggests 
that these are likely to be met by the end point. The one exception to this is the target concerning the percentage 
of school children meeting the literacy benchmark; it may be that the target set was just too ambitious to be 
achieved within such a short time, given the low baseline levels. Implementing literacy programs requires a 
wholesale change of culture and practices within a school system; such systemic changes are possible, but they 
take time to embed. 
Like previous evaluations of FFE1 and FFE2, the current evaluation found that, unlike many projects operating 
in the extremely challenging context of the Mozambican education system, FFE2 is largely managing to do what 
it set out to do within the planned timeframe. In terms of targets and indicators then, most of the boxes placed 
at the beginning of each section of the Results chapter of this report make for positive reading. However, as the 
external evaluator pointed out to project staff and leadership at the stakeholder review workshop, activities 
which concern human lives and children’s education cannot be evaluated only on the basis of activities, targets 
and indicators: it is important that project staff and leaders (and evaluators) maintain and promote awareness
of the reasons why given activities are conducted, what the positive change they are intended to bring and how 
they are experienced by all those involved in their implementation, including but not limited to the beneficiaries. 
At times, the fieldwork revealed a preoccupation with numerical targets, sometimes obscuring the question of 
137
whether these are translating into positive change (this was commented on by the project leader at the final 
validation workshop). A project such as this, which is well organized enough and sufficiently embedded to be 
meeting the majority of its targets should be capable of reflecting on whether the activities are making a 
difference and whether they are the right things to do, whether they could be done differently and what can be 
learned from the experience. 
Initial impact (has it made a difference?)
This report findsthat many of the activities presented are beginning to make a difference: teachers are beginning 
to use improved methods in their teaching; students are beginning to read better; community volunteers in the 
HGSFGs are beginning to apply new methods to produce large amounts of food; thanks to this produce, schools 
are beginning to diversity the menus consumed by students; school water committees are beginning to take 
responsibility for maintenance of their water systems. However, in all these cases, more time is needed for 
these initial changes to be considered embedded and irreversible. Demonstrating a behavior in which one has 
recently been trained does not constitute changing the underlying system of reflexes so that behavioral change 
takes place. Phenomena such as classroom practice, literacy, school culture, agricultural practices and 
community decision making are all complex and deeply culturally embedded, so that changes occur 
incrementally over time and rarely in a consistent, linear manner. Furthermore, formal training may transmit 
information effectively; however, there are further steps necessary for this information to become knowledge 
or skills, which are a prerequisite to real capacity increase (and behavior change). This is all the more necessary 
in a context such as Mozambique, where levels of poverty and disempowerment and poor-quality education 
constitute systemic barriers to change. 
The overall finding of the midterm evaluation, then, is that the project interventions are beginning to make a 
difference but that more time is needed in order for this initial change to be consolidated and integrated into 
behavioral changes. The project has 14 months still to run and this time will be helpful in consolidating the initial 
changes achieved. However, in order to achieve long-term, truly sustainable change, it is probable that both the 
project activities and the technical and expert support the project is providing to those implementing them 
would need to be sustained over a longer period. 
Strategic relevance regarding effectiveness, efficiency, impact and sustainability (were these the right things 
to do? what can be learned?)
The project applies many of the principles of good international development practice, such as: working in 
partnership with both national and local government and local communities at every stage, building capacity 
rather than just “doing”, building commitment from beneficiaries and partners through working together rather 
than by paying large per diems and other incentives. It is providing services and skills which are highly relevant 
138
to both the beneficiary population and the country as a whole. It has a good record and robust processes in 
place in terms of efficiency and value-for-money and uses the resources donated by the taxpayers of a partner 
country wisely and responsibly. It is beginning to produce impact, in the form of positive change, at national, 
district, school and individual level; this change cannot yet be considered embedded, however, and requires 
consolidation. Certain aspects of the project activities, such as school feeding, have become part of daily life and 
would be badly missed were they to cease: the current cohort of primary school children in the project schools 
has never known school without the daily soya porridge. And yet, major questions remain about the 
sustainability of these achievements in the current context in Mozambique. 
The evaluation revealed significant questions about sustainability, in terms of both government- and 
community-/school-level ability to sustain the activities after the end of the project. Other school feeding 
projects, such as the model followed by JAM (see page 129 - 30 above) may offer lessons to help the current 
FFE2 project be more sustainable. However, the bottom line is that school feeding (and the other activities 
developed by the project) requires resources, whether these are generated by project activities or provided by 
government, development partners, communities or families, or a combination of these. There is, as yet, no 
clear indication of where the necessary resources will come from to allow the project activities to continue. 
External evaluator’s concluding remarks
During the evaluation field visits, the external evaluator was very aware of the approaching close of the current 
phase of the project and that many of the achievements and positive changes observed may not be able to be 
maintained. In particular, the prospect that school children who are currently receiving food in school and 
benefiting from this, in terms of their health and education, might soon no longer receive food, is sobering. The 
context of Mozambique has deteriorated considerably since the external evaluator was first involved with the 
baseline of FFE1 in 2012. If that had not been the case, this report might be concluding on a more optimistic 
note, celebrating the achievements and noting their prospects of long-term sustainability. However, in the 
current context, it must end by celebrating the achievements and noting that, in the current reality, more time 
is needed for these to be sustained, although there is currently little prospect of the necessary resources being 
made available to allow this. The external evaluator considers that this is a valuable project, which is being well 
run and is largely achieving its objectives and beginning to produce the hoped-for changes. In her professional 
judgement, the FFE2 project merits being prolonged in order to have the opportunity to capitalize on all the 
investment made so far, both human and material. 
139
Annexes
Annex 1: Results Framework of the Food for Education Project in Mozambique, capturing 
the Theory of Change upon which the project is based
140
Annex 2: McGovern-Dole Results Frameworks
144
Annex 3: Technical appendix 1
FFE: Impact Evaluation
Analysis including substitute students
NFER (Jose Liht)
Contents
1 EGRA Numerical Variables 5
1.1 EGRA_ST1: Oral vocabulary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
1.1.1 Contrast 1: FFE + lit (all) vs Comparison (all) . . . . . . . . . . . . . . . . . . . . . . 5
1.1.2 Contrast 2: FFE + lit (Portuguese) vs Comparison (Portuguese) . . . . . . . . . . . . 8
1.1.3 Contrast 3: FFE + lit (Bilingual) vs Comparison (Bilingual) . . . . . . . . . . . . . . 10
1.1.4 Contrast 4: FFE + lit (Portuguese) vs FFE only (Portuguese) . . . . . . . . . . . . . 12
1.1.5 Contrast 5: FFE + lit (Bilingual) vs FFE + lit (Portuguese) . . . . . . . . . . . . . . 14
1.2 EGRA_ST2: Oral comprehension . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16
1.2.1 Contrast 1: FFE + lit (all) vs Comparison (all) . . . . . . . . . . . . . . . . . . . . . . 16
1.2.2 Contrast 2: FFE + lit (Portuguese) vs Comparison (Portuguese) . . . . . . . . . . . . 18
1.2.3 Contrast 3: FFE + lit (Bilingual) vs Comparison (Bilingual) . . . . . . . . . . . . . . 20
1.2.4 Contrast 4: FFE + lit (Portuguese) vs FFE only (Portuguese) . . . . . . . . . . . . . 22
1.2.5 Contrast 5: FFE + lit (Bilingual) vs FFE + lit (Portuguese) . . . . . . . . . . . . . . 24
1.3 EGRA_ST3: Phonological awareness . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26
1.3.1 Contrast 1: FFE + lit (all) vs Comparison (all) . . . . . . . . . . . . . . . . . . . . . . 26
1.3.2 Contrast 2: FFE + lit (Portuguese) vs Comparison (Portuguese) . . . . . . . . . . . . 28
1.3.3 Contrast 3: FFE + lit (Bilingual) vs Comparison (Bilingual) . . . . . . . . . . . . . . 30
1.3.4 Contrast 4: FFE + lit (Portuguese) vs FFE only (Portuguese) . . . . . . . . . . . . . 32
1.3.5 Contrast 5: FFE + lit (Bilingual) vs FFE + lit (Portuguese) . . . . . . . . . . . . . . 34
1.4 EGRA_ST4: Concepts of print . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36
1.4.1 Contrast 1: FFE + lit (all) vs Comparison (all) . . . . . . . . . . . . . . . . . . . . . . 36
1.4.2 Contrast 2: FFE + lit (Portuguese) vs Comparison (Portuguese) . . . . . . . . . . . . 38
1.4.3 Contrast 3: FFE + lit (Bilingual) vs Comparison (Bilingual) . . . . . . . . . . . . . . 40
1.4.4 Contrast 4: FFE + lit (Portuguese) vs FFE only (Portuguese) . . . . . . . . . . . . . 42
1.4.5 Contrast 5: FFE + lit (Bilingual) vs FFE + lit (Portuguese) . . . . . . . . . . . . . . 44
1.5 EGRA_ST5: Knowledge of letter sounds . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 46
1.5.1 Contrast 1: FFE + lit (all) vs Comparison (all) . . . . . . . . . . . . . . . . . . . . . . 46
1.5.2 Contrast 2: FFE + lit (Portuguese) vs Comparison (Portuguese) . . . . . . . . . . . . 48
1.5.3 Contrast 3: FFE + lit (Bilingual) vs Comparison (Bilingual) . . . . . . . . . . . . . . 50
1
1.5.4 Contrast 4: FFE + lit (Portuguese) vs FFE only (Portuguese) . . . . . . . . . . . . . 52
1.5.5 Contrast 5: FFE + lit (Bilingual) vs FFE + lit (Portuguese) . . . . . . . . . . . . . . 54
1.6 EGRA_ST6: Knowledge of syllables . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 56
1.6.1 Contrast 1: FFE + lit (all) vs Comparison (all) . . . . . . . . . . . . . . . . . . . . . . 56
1.6.2 Contrast 2: FFE + lit (Portuguese) vs Comparison (Portuguese) . . . . . . . . . . . . 58
1.6.3 Contrast 3: FFE + lit (Bilingual) vs Comparison (Bilingual) . . . . . . . . . . . . . . 60
1.6.4 Contrast 4: FFE + lit (Portuguese) vs FFE only (Portuguese) . . . . . . . . . . . . . 62
1.6.5 Contrast 5: FFE + lit (Bilingual) vs FFE + lit (Portuguese) . . . . . . . . . . . . . . 64
1.7 EGRA_ST7: Reading of words . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 66
1.7.1 Contrast 1: FFE + lit (all) vs Comparison (all) . . . . . . . . . . . . . . . . . . . . . . 66
1.7.2 Contrast 2: FFE + lit (Portuguese) vs Comparison (Portuguese) . . . . . . . . . . . . 68
1.7.3 Contrast 3: FFE + lit (Bilingual) vs Comparison (Bilingual) . . . . . . . . . . . . . . 70
1.7.4 Contrast 4: FFE + lit (Portuguese) vs FFE only (Portuguese) . . . . . . . . . . . . . 72
1.7.5 Contrast 5: FFE + lit (Bilingual) vs FFE + lit (Portuguese) . . . . . . . . . . . . . . 74
1.8 EGRA_ST8_1A: Reading fluency . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 76
1.8.1 Contrast 1: FFE + lit (all) vs Comparison (all) . . . . . . . . . . . . . . . . . . . . . . 76
1.8.2 Contrast 2: FFE + lit (Portuguese) vs Comparison (Portuguese) . . . . . . . . . . . . 78
1.8.3 Contrast 3: FFE + lit (Bilingual) vs Comparison (Bilingual) . . . . . . . . . . . . . . 80
1.8.4 Contrast 4: FFE + lit (Portuguese) vs FFE only (Portuguese) . . . . . . . . . . . . . 82
1.8.5 Contrast 5: FFE + lit (Bilingual) vs FFE + lit (Portuguese) . . . . . . . . . . . . . . 84
1.9 EGRA_ST8_1AS: Comparative reading fluency . . . . . . . . . . . . . . . . . . . . . . . . . 86
1.9.1 Contrast 1: FFE + lit (all) vs Comparison (all) . . . . . . . . . . . . . . . . . . . . . . 86
1.9.2 Contrast 2: FFE + lit (Portuguese) vs Comparison (Portuguese) . . . . . . . . . . . . 88
1.9.3 Contrast 3: FFE + lit (Bilingual) vs Comparison (Bilingual) . . . . . . . . . . . . . . 90
1.9.4 Contrast 4: FFE + lit (Portuguese) vs FFE only (Portuguese) . . . . . . . . . . . . . 92
1.9.5 Contrast 5: FFE + lit (Bilingual) vs FFE + lit (Portuguese) . . . . . . . . . . . . . . 94
1.10 EGRA_ST8_1B: Overall reading ability . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 96
1.10.1 Contrast 1: FFE + lit (all) vs Comparison (all) . . . . . . . . . . . . . . . . . . . . . . 96
1.10.2 Contrast 2: FFE + lit (Portuguese) vs Comparison (Portuguese) . . . . . . . . . . . . 97
1.10.3 Contrast 3: FFE + lit (Bilingual) vs Comparison (Bilingual) . . . . . . . . . . . . . . 99
1.10.4 Contrast 4: FFE + lit (Portuguese) vs FFE only (Portuguese) . . . . . . . . . . . . . 101
1.10.5 Contrast 5: FFE + lit (Bilingual) vs FFE + lit (Portuguese) . . . . . . . . . . . . . . 103
1.11 EGRA_ST8_1BS: Comparative overall reading ability . . . . . . . . . . . . . . . . . . . . . . 105
1.11.1 Contrast 1: FFE + lit (all) vs Comparison (all) . . . . . . . . . . . . . . . . . . . . . . 105
1.11.2 Contrast 2: FFE + lit (Portuguese) vs Comparison (Portuguese) . . . . . . . . . . . . 107
1.11.3 Contrast 3: FFE + lit (Bilingual) vs Comparison (Bilingual) . . . . . . . . . . . . . . 109
2
1.11.4 Contrast 4: FFE + lit (Portuguese) vs FFE only (Portuguese) . . . . . . . . . . . . . 111
1.11.5 Contrast 5: FFE + lit (Bilingual) vs FFE + lit (Portuguese) . . . . . . . . . . . . . . 113
1.12 EGRA_ST8_2: Comprehension of reading . . . . . . . . . . . . . . . . . . . . . . . . . . . . 115
1.12.1 Contrast 1: FFE + lit (all) vs Comparison (all) . . . . . . . . . . . . . . . . . . . . . . 115
1.12.2 Contrast 2: FFE + lit (Portuguese) vs Comparison (Portuguese) . . . . . . . . . . . . 118
1.12.3 Contrast 3: FFE + lit (Bilingual) vs Comparison (Bilingual) . . . . . . . . . . . . . . 120
1.12.4 Contrast 4: FFE + lit (Portuguese) vs FFE only (Portuguese) . . . . . . . . . . . . . 122
1.12.5 Contrast 5: FFE + lit (Bilingual) vs FFE + lit (Portuguese) . . . . . . . . . . . . . . 124
1.13 EGRA_ST9_2A: Total writing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 126
1.13.1 Contrast 1: FFE + lit (all) vs Comparison (all) . . . . . . . . . . . . . . . . . . . . . . 126
1.13.2 Contrast 2: FFE + lit (Portuguese) vs Comparison (Portuguese) . . . . . . . . . . . . 128
1.13.3 Contrast 3: FFE + lit (Bilingual) vs Comparison (Bilingual) . . . . . . . . . . . . . . 130
1.13.4 Contrast 4: FFE + lit (Portuguese) vs FFE only (Portuguese) . . . . . . . . . . . . . 132
1.13.5 Contrast 5: FFE + lit (Bilingual) vs FFE + lit (Portuguese) . . . . . . . . . . . . . . 134
1.14 EGRA_ST9_2B: Writing words . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 136
1.14.1 Contrast 1: FFE + lit (all) vs Comparison (all) . . . . . . . . . . . . . . . . . . . . . . 136
1.14.2 Contrast 2: FFE + lit (Portuguese) vs Comparison (Portuguese) . . . . . . . . . . . . 137
1.14.3 Contrast 3: FFE + lit (Bilingual) vs Comparison (Bilingual) . . . . . . . . . . . . . . 139
1.14.4 Contrast 4: FFE + lit (Portuguese) vs FFE only (Portuguese) . . . . . . . . . . . . . 141
1.14.5 Contrast 5: FFE + lit (Bilingual) vs FFE + lit (Portuguese) . . . . . . . . . . . . . . 143
1.15 EGRA_ST9_2C: Writing phrase . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 145
1.15.1 Contrast 1: FFE + lit (all) vs Comparison (all) . . . . . . . . . . . . . . . . . . . . . . 145
1.15.2 Contrast 2: FFE + lit (Portuguese) vs Comparison (Portuguese) . . . . . . . . . . . . 147
1.15.3 Contrast 3: FFE + lit (Bilingual) vs Comparison (Bilingual) . . . . . . . . . . . . . . 149
1.15.4 Contrast 4: FFE + lit (Portuguese) vs FFE only (Portuguese) . . . . . . . . . . . . . 151
1.15.5 Contrast 5: FFE + lit (Bilingual) vs FFE + lit (Portuguese) . . . . . . . . . . . . . . 153
2 EGRA Categorical Variables 155
2.1 EGRA_ST9_1A: Writing first name . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 155
2.1.1 Contrast 1: FFE + lit (all) vs Comparison (all) . . . . . . . . . . . . . . . . . . . . . . 155
2.1.2 Contrast 2: FFE + lit (Portuguese) vs Comparison (Portuguese) . . . . . . . . . . . . 157
2.1.3 Contrast 3: FFE + lit (Bilingual) vs Comparison (Bilingual) . . . . . . . . . . . . . . 158
2.1.4 Contrast 4: FFE + lit (Portuguese) vs FFE only (Portuguese) . . . . . . . . . . . . . 159
2.1.5 Contrast 5: FFE + lit (Bilingual) vs FFE + lit (Portuguese) . . . . . . . . . . . . . . 160
2.2 EGRA_ST9_1B: Writing family name . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 161
2.2.1 Contrast 1: FFE + lit (all) vs Comparison (all) . . . . . . . . . . . . . . . . . . . . . . 161
2.2.2 Contrast 2: FFE + lit (Portuguese) vs Comparison (Portuguese) . . . . . . . . . . . . 163
3
2.2.3 Contrast 3: FFE + lit (Bilingual) vs Comparison (Bilingual) . . . . . . . . . . . . . . 164
2.2.4 Contrast 4: FFE + lit (Portuguese) vs FFE only (Portuguese) . . . . . . . . . . . . . 165
2.2.5 Contrast 5: FFE + lit (Bilingual) vs FFE + lit (Portuguese) . . . . . . . . . . . . . . 166
2.3 EGRA_ST9_1C: Writing name overall . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 167
2.3.1 Contrast 1: FFE + lit (all) vs Comparison (all) . . . . . . . . . . . . . . . . . . . . . . 167
2.3.2 Contrast 2: FFE + lit (Portuguese) vs Comparison (Portuguese) . . . . . . . . . . . . 169
2.3.3 Contrast 3: FFE + lit (Bilingual) vs Comparison (Bilingual) . . . . . . . . . . . . . . 170
2.3.4 Contrast 4: FFE + lit (Portuguese) vs FFE only (Portuguese) . . . . . . . . . . . . . 171
2.3.5 Contrast 5: FFE + lit (Bilingual) vs FFE + lit (Portuguese) . . . . . . . . . . . . . . 172
3 Literacy Benchmark 173
3.1 Literacy Benchmark: Boys and Girls . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 173
3.1.1 Contrast 1: FFE + lit (all) vs Comparison (all) . . . . . . . . . . . . . . . . . . . . . . 173
3.1.2 Contrast 2: FFE + lit (Portuguese) vs Comparison (Portuguese) . . . . . . . . . . . . 174
3.1.3 Contrast 3: FFE + lit (Bilingual) vs Comparison (Bilingual) . . . . . . . . . . . . . . 175
3.1.4 Contrast 4: FFE + lit (Portuguese) vs FFE only (Portuguese) . . . . . . . . . . . . . 176
3.1.5 Contrast 5: FFE + lit (Bilingual) vs FFE + lit (Portuguese) . . . . . . . . . . . . . . 177
3.2 Literacy Benchmark: Boys only . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 178
3.2.1 Contrast 1: FFE + lit (all) vs Comparison (all) . . . . . . . . . . . . . . . . . . . . . . 178
3.2.2 Contrast 2: FFE + lit (Portuguese) vs Comparison (Portuguese) . . . . . . . . . . . . 179
3.2.3 Contrast 3: FFE + lit (Bilingual) vs Comparison (Bilingual) . . . . . . . . . . . . . . 180
3.2.4 Contrast 4: FFE + lit (Portuguese) vs FFE only (Portuguese) . . . . . . . . . . . . . 181
3.2.5 Contrast 5: FFE + lit (Bilingual) vs FFE + lit (Portuguese) . . . . . . . . . . . . . . 182
3.3 Literacy Benchmark: Girls only . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 183
3.3.1 Contrast 1: FFE + lit (all) vs Comparison (all) . . . . . . . . . . . . . . . . . . . . . . 183
3.3.2 Contrast 2: FFE + lit (Portuguese) vs Comparison (Portuguese) . . . . . . . . . . . . 184
3.3.3 Contrast 3: FFE + lit (Bilingual) vs Comparison (Bilingual) . . . . . . . . . . . . . . 185
3.3.4 Contrast 4: FFE + lit (Portuguese) vs FFE only (Portuguese) . . . . . . . . . . . . . 186
3.3.5 Contrast 5: FFE + lit (Bilingual) vs FFE + lit (Portuguese) . . . . . . . . . . . . . . 187
4 BMI 188
4.1 Height . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 188
4.2 Weight . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 191
4.3 BMI Z Score . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 193
4
5 Questions regarding hunger, attention and hygiene 195
5.1 HUNGER: Hungry in school . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 195
5.2 HUNGERNOW: Hungry now . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 196
5.3 ATTENTION: Attention problems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 197
5.4 TOILET: Wash after toilet . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 198
5.5 EATING: Wash before eating . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 199
6 Teacher questions regarding hunger, attention and difficulty in learning 201
6.1 Students are hungry . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 201
6.2 Students have attention problems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 202
6.3 Students have learning problems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 203
1 EGRA Numerical Variables
1.1 EGRA_ST1: Oral vocabulary
1.1.1 Contrast 1: FFE + lit (all) vs Comparison (all)
Table 1: Oral vocabulary
Baseline Midline
Intervention Mean SD n min max Mean SD n min max
Control 9.450 1.636 1136 0 10 9.779 0.621 1081 3 10
Treatment 9.026 1.967 1578 0 10 9.672 0.747 1808 1 10
5
0.0
2.5
5.0
7.5
10.0
Control Treatment
CINTERVENTION
Mean
PERIOD
Baseline
Midline
Oral vocabulary
6
0.0
2.5
5.0
7.5
10.0
Control Treatment
CINTERVENTION
Oral vocabulary
PERIOD
Baseline
Midline
Oral vocabulary
As shown in the table above, for the the Oral vocabulary EGRA subtask, the mean for the Control (Com￾parison (all)) condition at baseline was 9.449824 (SD = 1.636355) and the mean for the Treatment (FFE +
lit (all)) condition at baseline was 9.025982 (SD = 1.967384). The difference between Control (Comparison
(all)) and Treatment (FFE + lit (all)) at baseline was thus of 0.4238417 points. The p-value for this differ￾ence was 0.0003841511. The mean for the Control (Comparison (all)) condition at midline was 9.778908 (SD
= 0.6207516) and the mean for the Treatment (FFE + lit (all)) condition at midline was 9.672013 (SD =
0.7470673). The difference between Control (Comparison (all)) and Treatment (FFE + lit (all)) at midline
was thus of 0.1068951 points. The p-value for this difference was 0.00774526. The change from the baseline
to the midline of 0.3290845 points for the Control (Comparison (all)) condition establishes the conterfactual
against which the change for the Treatment (FFE + lit (all)) can be assessed against. The change for the
Treatment (FFE + lit (all)) from the baseline to the midline was of 0.646031 points. Consequently, the
change for the Control (Comparison (all)) relative to the change for the Treatment (FFE + lit (all)) condi￾tion was of 0.3169465 points. The p-value for this difference was 0.007771889. Accordingly, the difference
was statistically significant (p < 0.01 corrected for 4 additional contrasts) and thus there is evidence of a
different degree of change from baseline to midline in the Treatment (FFE + lit (all)) when compared to the
Control (Comparison (all)) condition. This provides evidence that the Treatment (FFE + lit (all)) (or some
other unobserved process) impacted on the Oral vocabulary EGRA subtask and thus the rate of change of
the two conditions relative to their baselines differed.
7
1.1.2 Contrast 2: FFE + lit (Portuguese) vs Comparison (Portuguese)
Table 2: Oral vocabulary
Baseline Midline
Intervention Mean SD n min max Mean SD n min max
Control 9.441 1.665 1047 0 10 9.773 0.629 972 3 10
Treatment 8.721 2.215 1040 0 10 9.611 0.816 1047 1 10
0.0
2.5
5.0
7.5
10.0
Control Treatment
CINTERVENTION
Mean
PERIOD
Baseline
Midline
Oral vocabulary
8
0.0
2.5
5.0
7.5
10.0
Control Treatment
CINTERVENTION
Oral vocabulary
PERIOD
Baseline
Midline
Oral vocabulary
As shown in the table above, for the the Oral vocabulary EGRA subtask, the mean for the Control (Com￾parison (Portuguese)) condition at baseline was 9.441261 (SD = 1.665112) and the mean for the Treatment
(FFE + lit (Portuguese)) condition at baseline was 8.721154 (SD = 2.214906). The difference between Con￾trol (Comparison (Portuguese)) and Treatment (FFE + lit (Portuguese)) at baseline was thus of 0.7201069
points. The p-value for this difference was 1.973785e-06. The mean for the Control (Comparison (Por￾tuguese)) condition at midline was 9.772634 (SD = 0.6294784) and the mean for the Treatment (FFE + lit
(Portuguese)) condition at midline was 9.61127 (SD = 0.8161767). The difference between Control (Com￾parison (Portuguese)) and Treatment (FFE + lit (Portuguese)) at midline was thus of 0.1613634 points. The
p-value for this difference was 0.0008852814. The change from the baseline to the midline of 0.331373 points
for the Control (Comparison (Portuguese)) condition establishes the conterfactual against which the change
for the Treatment (FFE + lit (Portuguese)) can be assessed against. The change for the Treatment (FFE +
lit (Portuguese)) from the baseline to the midline was of 0.8901164 points. Consequently, the change for the
Control (Comparison (Portuguese)) relative to the change for the Treatment (FFE + lit (Portuguese)) con￾dition was of 0.5587435 points. The p-value for this difference was 0.0002102456. Accordingly, the difference
was statistically significant (p < 0.01 corrected for 4 additional contrasts) and thus there is evidence of a dif￾ferent degree of change from baseline to midline in the Treatment (FFE + lit (Portuguese)) when compared
to the Control (Comparison (Portuguese)) condition. This provides evidence that the Treatment (FFE +
lit (Portuguese)) (or some other unobserved process) impacted on the Oral vocabulary EGRA subtask and
thus the rate of change of the two conditions relative to their baselines differed.
9
1.1.3 Contrast 3: FFE + lit (Bilingual) vs Comparison (Bilingual)
Table 3: Oral vocabulary
Baseline Midline
Intervention Mean SD n min max Mean SD n min max
Control 9.551 1.252 89 1 10 9.835 0.536 109 6 10
Treatment 9.615 1.161 538 0 10 9.756 0.631 761 5 10
0.0
2.5
5.0
7.5
10.0
Control Treatment
CINTERVENTION
Mean
PERIOD
Baseline
Midline
Oral vocabulary
10
0.0
2.5
5.0
7.5
10.0
Control Treatment
CINTERVENTION
Oral vocabulary
PERIOD
Baseline
Midline
Oral vocabulary
As shown in the table above, for the the Oral vocabulary EGRA subtask, the mean for the Control (Com￾parison (Bilingual)) condition at baseline was 9.550562 (SD = 1.252373) and the mean for the Treatment
(FFE + lit (Bilingual)) condition at baseline was 9.615242 (SD = 1.160616). The difference between Control
(Comparison (Bilingual)) and Treatment (FFE + lit (Bilingual)) at baseline was thus of 0.06467984 points.
The p-value for this difference was 0.7338745. The mean for the Control (Comparison (Bilingual)) condition
at midline was 9.834862 (SD = 0.5359962) and the mean for the Treatment (FFE + lit (Bilingual)) condition
at midline was 9.755585 (SD = 0.6309337). The difference between Control (Comparison (Bilingual)) and
Treatment (FFE + lit (Bilingual)) at midline was thus of 0.07927763 points. The p-value for this difference
was 0.1083091. The change from the baseline to the midline of 0.2843006 points for the Control (Comparison
(Bilingual)) condition establishes the conterfactual against which the change for the Treatment (FFE + lit
(Bilingual)) can be assessed against. The change for the Treatment (FFE + lit (Bilingual)) from the baseline
to the midline was of 0.1403431 points. Consequently, the change for the Control (Comparison (Bilingual))
relative to the change for the Treatment (FFE + lit (Bilingual)) condition was of -0.1439575 points. The
p-value for this difference was 0.4222522. Accordingly, the difference was not statistically significant (p <
0.01 corrected for 5 additional contrasts) and therefore the statistical test provides no conclusive evidence
in regards to there being a difference in the rate of change between the baseline and the midline for the
Oral vocabulary EGRA subtask across the Control (Comparison (Bilingual)) and the Treatment (FFE + lit
(Bilingual)) conditions beyond that expected from sampling error.
11
1.1.4 Contrast 4: FFE + lit (Portuguese) vs FFE only (Portuguese)
Table 4: Oral vocabulary
Baseline Midline
Intervention Mean SD n min max Mean SD n min max
Control 8.141 2.212 1007 0 10 9.558 0.868 952 3 10
Treatment 8.721 2.215 1040 0 10 9.611 0.816 1047 1 10
0.0
2.5
5.0
7.5
10.0
Control Treatment
CINTERVENTION
Mean
PERIOD
Baseline
Midline
Oral vocabulary
12
0.0
2.5
5.0
7.5
10.0
Control Treatment
CINTERVENTION
Oral vocabulary
PERIOD
Baseline
Midline
Oral vocabulary
As shown in the table above, for the the Oral vocabulary EGRA subtask, the mean for the Control (FFE only
(Portuguese)) condition at baseline was 8.141013 (SD = 2.211927) and the mean for the Treatment (FFE +
lit (Portuguese)) condition at baseline was 8.721154 (SD = 2.214906). The difference between Control (FFE
only (Portuguese)) and Treatment (FFE + lit (Portuguese)) at baseline was thus of 0.5801409 points. The
p-value for this difference was 0.002241314. The mean for the Control (FFE only (Portuguese)) condition at
midline was 9.557773 (SD = 0.8681915) and the mean for the Treatment (FFE + lit (Portuguese)) condition
at midline was 9.61127 (SD = 0.8161767). The difference between Control (FFE only (Portuguese)) and
Treatment (FFE + lit (Portuguese)) at midline was thus of 0.05349719 points. The p-value for this difference
was 0.3538011. The change from the baseline to the midline of 1.41676 points for the Control (FFE only
(Portuguese)) condition establishes the conterfactual against which the change for the Treatment (FFE +
lit (Portuguese)) can be assessed against. The change for the Treatment (FFE + lit (Portuguese)) from
the baseline to the midline was of 0.8901164 points. Consequently, the change for the Control (FFE only
(Portuguese)) relative to the change for the Treatment (FFE + lit (Portuguese)) condition was of -0.5266437
points. The p-value for this difference was 0.006323686. Accordingly, the difference was statistically signifi￾cant (p < 0.01 corrected for 4 additional contrasts) and thus there is evidence of a different degree of change
from baseline to midline in the Treatment (FFE + lit (Portuguese)) when compared to the Control (FFE
only (Portuguese)) condition. This provides evidence that the Treatment (FFE + lit (Portuguese)) (or some
other unobserved process) impacted on the Oral vocabulary EGRA subtask and thus the rate of change of
the two conditions relative to their baselines differed. It should be pointed out that the difference in the
rate of change of the Treatment (FFE + lit (Portuguese)) relative to the Control (FFE only (Portuguese))
was negative and thus suggests that students in the Control (FFE only (Portuguese)) schools performed
significantly better than those in the Treatment (FFE + lit (Portuguese)) .
13
1.1.5 Contrast 5: FFE + lit (Bilingual) vs FFE + lit (Portuguese)
Table 5: Oral vocabulary
Baseline Midline
Intervention Mean SD n min max Mean SD n min max
Control 8.721 2.215 1040 0 10 9.611 0.816 1047 1 10
Treatment 9.615 1.161 538 0 10 9.756 0.631 761 5 10
0.0
2.5
5.0
7.5
10.0
Control Treatment
CINTERVENTION
Mean
PERIOD
Baseline
Midline
Oral vocabulary
14
0.0
2.5
5.0
7.5
10.0
Control Treatment
CINTERVENTION
Oral vocabulary
PERIOD
Baseline
Midline
Oral vocabulary
As shown in the table above, for the the Oral vocabulary EGRA subtask, the mean for the Control (FFE +
lit (Portuguese)) condition at baseline was 8.721154 (SD = 2.214906) and the mean for the Treatment (FFE
+ lit (Bilingual)) condition at baseline was 9.615242 (SD = 1.160616). The difference between Control (FFE
+ lit (Portuguese)) and Treatment (FFE + lit (Bilingual)) at baseline was thus of 0.8940878 points. The
p-value for this difference was 3.994351e-09. The mean for the Control (FFE + lit (Portuguese)) condition
at midline was 9.61127 (SD = 0.8161767) and the mean for the Treatment (FFE + lit (Bilingual)) condition
at midline was 9.755585 (SD = 0.6309337). The difference between Control (FFE + lit (Portuguese)) and
Treatment (FFE + lit (Bilingual)) at midline was thus of 0.1443145 points. The p-value for this difference
was 0.002332906. The change from the baseline to the midline of 0.8901164 points for the Control (FFE + lit
(Portuguese)) condition establishes the conterfactual against which the change for the Treatment (FFE + lit
(Bilingual)) can be assessed against. The change for the Treatment (FFE + lit (Bilingual)) from the baseline
to the midline was of 0.1403431 points. Consequently, the change for the Control (FFE + lit (Portuguese))
relative to the change for the Treatment (FFE + lit (Bilingual)) condition was of -0.7497733 points. The
p-value for this difference was 3.744629e-07. Accordingly, the difference was statistically significant (p < 0.01
corrected for 4 additional contrasts) and thus there is evidence of a different degree of change from baseline
to midline in the Treatment (FFE + lit (Bilingual)) when compared to the Control (FFE + lit (Portuguese))
condition. This provides evidence that the Treatment (FFE + lit (Bilingual)) (or some other unobserved
process) impacted on the Oral vocabulary EGRA subtask and thus the rate of change of the two conditions
relative to their baselines differed. It should be pointed out that the difference in the rate of change of the
Treatment (FFE + lit (Bilingual)) relative to the Control (FFE + lit (Portuguese)) was negative and thus
suggests that students in the Control (FFE + lit (Portuguese)) schools performed significantly better than
those in the Treatment (FFE + lit (Bilingual)) .
15
1.2 EGRA_ST2: Oral comprehension
1.2.1 Contrast 1: FFE + lit (all) vs Comparison (all)
Table 6: Oral comprehension
Baseline Midline
Intervention Mean SD n min max Mean SD n min max
Control 0.888 1.159 1136 0 4 2.854 1.245 1081 0 4
Treatment 1.271 1.438 1578 0 4 2.521 1.255 1808 0 4
0
1
2
Control Treatment
CINTERVENTION
Mean
PERIOD
Baseline
Midline
Oral comprehension 
16
0
1
2
3
4
Control Treatment
CINTERVENTION
Oral comprehension 
PERIOD
Baseline
Midline
Oral comprehension 
As shown in the table above, for the the Oral comprehension EGRA subtask, the mean for the Control
(Comparison (all)) condition at baseline was 0.8882042 (SD = 1.159068) and the mean for the Treatment
(FFE + lit (all)) condition at baseline was 1.271229 (SD = 1.437763). The difference between Control
(Comparison (all)) and Treatment (FFE + lit (all)) at baseline was thus of 0.3830252 points. The p-value
for this difference was 0.001128986. The mean for the Control (Comparison (all)) condition at midline
was 2.853839 (SD = 1.244583) and the mean for the Treatment (FFE + lit (all)) condition at midline was
2.521018 (SD = 1.255249). The difference between Control (Comparison (all)) and Treatment (FFE + lit
(all)) at midline was thus of 0.3328213 points. The p-value for this difference was 0.0009810953. The change
from the baseline to the midline of 1.965635 points for the Control (Comparison (all)) condition establishes
the conterfactual against which the change for the Treatment (FFE + lit (all)) can be assessed against.
The change for the Treatment (FFE + lit (all)) from the baseline to the midline was of 1.249788 points.
Consequently, the change for the Control (Comparison (all)) relative to the change for the Treatment (FFE
+ lit (all)) condition was of -0.7158465 points. The p-value for this difference was 2.260209e-06. Accordingly,
the difference was statistically significant (p < 0.01 corrected for 4 additional contrasts) and thus there is
evidence of a different degree of change from baseline to midline in the Treatment (FFE + lit (all)) when
compared to the Control (Comparison (all)) condition. This provides evidence that the Treatment (FFE +
lit (all)) (or some other unobserved process) impacted on the Oral comprehension EGRA subtask and thus
the rate of change of the two conditions relative to their baselines differed. It should be pointed out that
the difference in the rate of change of the Treatment (FFE + lit (all)) relative to the Control (Comparison
(all)) was negative and thus suggests that students in the Control (Comparison (all)) schools performed
significantly better than those in the Treatment (FFE + lit (all)) .
17
1.2.2 Contrast 2: FFE + lit (Portuguese) vs Comparison (Portuguese)
Table 7: Oral comprehension
Baseline Midline
Intervention Mean SD n min max Mean SD n min max
Control 0.895 1.165 1047 0 4 2.951 1.221 972 0 4
Treatment 1.235 1.426 1040 0 4 2.555 1.309 1047 0 4
0
1
2
3
Control Treatment
CINTERVENTION
Mean
PERIOD
Baseline
Midline
Oral comprehension 
18
0
1
2
3
4
Control Treatment
CINTERVENTION
Oral comprehension 
PERIOD
Baseline
Midline
Oral comprehension 
As shown in the table above, for the the Oral comprehension EGRA subtask, the mean for the Control
(Comparison (Portuguese)) condition at baseline was 0.8949379 (SD = 1.164502) and the mean for the
Treatment (FFE + lit (Portuguese)) condition at baseline was 1.234615 (SD = 1.425991). The difference
between Control (Comparison (Portuguese)) and Treatment (FFE + lit (Portuguese)) at baseline was thus of
0.3396775 points. The p-value for this difference was 0.009005426. The mean for the Control (Comparison
(Portuguese)) condition at midline was 2.950617 (SD = 1.22101) and the mean for the Treatment (FFE
+ lit (Portuguese)) condition at midline was 2.554919 (SD = 1.309106). The difference between Control
(Comparison (Portuguese)) and Treatment (FFE + lit (Portuguese)) at midline was thus of 0.3956985
points. The p-value for this difference was 0.000293822. The change from the baseline to the midline of
2.055679 points for the Control (Comparison (Portuguese)) condition establishes the conterfactual against
which the change for the Treatment (FFE + lit (Portuguese)) can be assessed against. The change for the
Treatment (FFE + lit (Portuguese)) from the baseline to the midline was of 1.320303 points. Consequently,
the change for the Control (Comparison (Portuguese)) relative to the change for the Treatment (FFE +
lit (Portuguese)) condition was of -0.7353759 points. The p-value for this difference was 5.591441e-06.
Accordingly, the difference was statistically significant (p < 0.01 corrected for 4 additional contrasts) and
thus there is evidence of a different degree of change from baseline to midline in the Treatment (FFE + lit
(Portuguese)) when compared to the Control (Comparison (Portuguese)) condition. This provides evidence
that the Treatment (FFE + lit (Portuguese)) (or some other unobserved process) impacted on the Oral
comprehension EGRA subtask and thus the rate of change of the two conditions relative to their baselines
differed. It should be pointed out that the difference in the rate of change of the Treatment (FFE +
lit (Portuguese)) relative to the Control (Comparison (Portuguese)) was negative and thus suggests that
students in the Control (Comparison (Portuguese)) schools performed significantly better than those in the
Treatment (FFE + lit (Portuguese)) .
19
1.2.3 Contrast 3: FFE + lit (Bilingual) vs Comparison (Bilingual)
Table 8: Oral comprehension
Baseline Midline
Intervention Mean SD n min max Mean SD n min max
Control 0.809 1.096 89 0 4 1.991 1.118 109 0 4
Treatment 1.342 1.459 538 0 4 2.474 1.176 761 0 4
0.0
0.5
1.0
1.5
2.0
2.5
Control Treatment
CINTERVENTION
Mean
PERIOD
Baseline
Midline
Oral comprehension 
20
0
1
2
3
4
Control Treatment
CINTERVENTION
Oral comprehension 
PERIOD
Baseline
Midline
Oral comprehension 
As shown in the table above, for the the Oral comprehension EGRA subtask, the mean for the Control
(Comparison (Bilingual)) condition at baseline was 0.8089888 (SD = 1.096237) and the mean for the Treat￾ment (FFE + lit (Bilingual)) condition at baseline was 1.342007 (SD = 1.45898). The difference between
Control (Comparison (Bilingual)) and Treatment (FFE + lit (Bilingual)) at baseline was thus of 0.5330187
points. The p-value for this difference was 0.03396653. The mean for the Control (Comparison (Bilingual))
condition at midline was 1.990826 (SD = 1.117996) and the mean for the Treatment (FFE + lit (Bilingual))
condition at midline was 2.474376 (SD = 1.176385). The difference between Control (Comparison (Bilin￾gual)) and Treatment (FFE + lit (Bilingual)) at midline was thus of 0.4835501 points. The p-value for this
difference was 0.002500106. The change from the baseline to the midline of 1.181837 points for the Control
(Comparison (Bilingual)) condition establishes the conterfactual against which the change for the Treatment
(FFE + lit (Bilingual)) can be assessed against. The change for the Treatment (FFE + lit (Bilingual)) from
the baseline to the midline was of 1.132368 points. Consequently, the change for the Control (Comparison
(Bilingual)) relative to the change for the Treatment (FFE + lit (Bilingual)) condition was of -0.04946854
points. The p-value for this difference was 0.8882155. Accordingly, the difference was not statistically signif￾icant (p < 0.01 corrected for 5 additional contrasts) and therefore the statistical test provides no conclusive
evidence in regards to there being a difference in the rate of change between the baseline and the midline
for the Oral comprehension EGRA subtask across the Control (Comparison (Bilingual)) and the Treatment
(FFE + lit (Bilingual)) conditions beyond that expected from sampling error.
21
1.2.4 Contrast 4: FFE + lit (Portuguese) vs FFE only (Portuguese)
Table 9: Oral comprehension
Baseline Midline
Intervention Mean SD n min max Mean SD n min max
Control 1.242 1.390 1007 0 4 1.931 1.446 952 0 4
Treatment 1.235 1.426 1040 0 4 2.555 1.309 1047 0 4
0
1
2
Control Treatment
CINTERVENTION
Mean
PERIOD
Baseline
Midline
Oral comprehension 
22
0
1
2
3
4
Control Treatment
CINTERVENTION
Oral comprehension 
PERIOD
Baseline
Midline
Oral comprehension 
As shown in the table above, for the the Oral comprehension EGRA subtask, the mean for the Control (FFE
only (Portuguese)) condition at baseline was 1.242304 (SD = 1.390424) and the mean for the Treatment
(FFE + lit (Portuguese)) condition at baseline was 1.234615 (SD = 1.425991). The difference between Con￾trol (FFE only (Portuguese)) and Treatment (FFE + lit (Portuguese)) at baseline was thus of 0.007688488
points. The p-value for this difference was 0.9560903. The mean for the Control (FFE only (Portuguese))
condition at midline was 1.930672 (SD = 1.445623) and the mean for the Treatment (FFE + lit (Por￾tuguese)) condition at midline was 2.554919 (SD = 1.309106). The difference between Control (FFE only
(Portuguese)) and Treatment (FFE + lit (Portuguese)) at midline was thus of 0.6242465 points. The p-value
for this difference was 2.647787e-08. The change from the baseline to the midline of 0.6883684 points for
the Control (FFE only (Portuguese)) condition establishes the conterfactual against which the change for
the Treatment (FFE + lit (Portuguese)) can be assessed against. The change for the Treatment (FFE +
lit (Portuguese)) from the baseline to the midline was of 1.320303 points. Consequently, the change for
the Control (FFE only (Portuguese)) relative to the change for the Treatment (FFE + lit (Portuguese))
condition was of 0.631935 points. The p-value for this difference was 0.0007014284. Accordingly, the differ￾ence was statistically significant (p < 0.01 corrected for 4 additional contrasts) and thus there is evidence
of a different degree of change from baseline to midline in the Treatment (FFE + lit (Portuguese)) when
compared to the Control (FFE only (Portuguese)) condition. This provides evidence that the Treatment
(FFE + lit (Portuguese)) (or some other unobserved process) impacted on the Oral comprehension EGRA
subtask and thus the rate of change of the two conditions relative to their baselines differed.
23
1.2.5 Contrast 5: FFE + lit (Bilingual) vs FFE + lit (Portuguese)
Table 10: Oral comprehension
Baseline Midline
Intervention Mean SD n min max Mean SD n min max
Control 1.235 1.426 1040 0 4 2.555 1.309 1047 0 4
Treatment 1.342 1.459 538 0 4 2.474 1.176 761 0 4
0
1
2
Control Treatment
CINTERVENTION
Mean
PERIOD
Baseline
Midline
Oral comprehension 
24
0
1
2
3
4
Control Treatment
CINTERVENTION
Oral comprehension 
PERIOD
Baseline
Midline
Oral comprehension 
As shown in the table above, for the the Oral comprehension EGRA subtask, the mean for the Control (FFE
+ lit (Portuguese)) condition at baseline was 1.234615 (SD = 1.425991) and the mean for the Treatment
(FFE + lit (Bilingual)) condition at baseline was 1.342007 (SD = 1.45898). The difference between Control
(FFE + lit (Portuguese)) and Treatment (FFE + lit (Bilingual)) at baseline was thus of 0.1073921 points.
The p-value for this difference was 0.5485581. The mean for the Control (FFE + lit (Portuguese)) condition
at midline was 2.554919 (SD = 1.309106) and the mean for the Treatment (FFE + lit (Bilingual)) condition
at midline was 2.474376 (SD = 1.176385). The difference between Control (FFE + lit (Portuguese)) and
Treatment (FFE + lit (Bilingual)) at midline was thus of 0.08054299 points. The p-value for this difference
was 0.4197186. The change from the baseline to the midline of 1.320303 points for the Control (FFE + lit
(Portuguese)) condition establishes the conterfactual against which the change for the Treatment (FFE + lit
(Bilingual)) can be assessed against. The change for the Treatment (FFE + lit (Bilingual)) from the baseline
to the midline was of 1.132368 points. Consequently, the change for the Control (FFE + lit (Portuguese))
relative to the change for the Treatment (FFE + lit (Bilingual)) condition was of -0.187935 points. The
p-value for this difference was 0.3774959. Accordingly, the difference was not statistically significant (p <
0.01 corrected for 5 additional contrasts) and therefore the statistical test provides no conclusive evidence in
regards to there being a difference in the rate of change between the baseline and the midline for the Oral
comprehension EGRA subtask across the Control (FFE + lit (Portuguese)) and the Treatment (FFE + lit
(Bilingual)) conditions beyond that expected from sampling error.
25
1.3 EGRA_ST3: Phonological awareness
1.3.1 Contrast 1: FFE + lit (all) vs Comparison (all)
Table 11: Phonological awareness
Baseline Midline
Intervention Mean SD n min max Mean SD n min max
Control 2.245 3.257 1136 0 10 6.925 2.607 1081 0 10
Treatment 3.054 3.462 1578 0 10 7.614 2.399 1808 0 10
0
2
4
6
Control Treatment
CINTERVENTION
Mean
PERIOD
Baseline
Midline
Phonological awareness 
26
0.0
2.5
5.0
7.5
10.0
Control Treatment
CINTERVENTION
Phonological awareness 
PERIOD
Baseline
Midline
Phonological awareness 
As shown in the table above, for the the Phonological awareness EGRA subtask, the mean for the Control
(Comparison (all)) condition at baseline was 2.244718 (SD = 3.256707) and the mean for the Treatment
(FFE + lit (all)) condition at baseline was 3.053866 (SD = 3.462126). The difference between Control
(Comparison (all)) and Treatment (FFE + lit (all)) at baseline was thus of 0.8091473 points. The p-value
for this difference was 0.02043418. The mean for the Control (Comparison (all)) condition at midline was
6.925069 (SD = 2.607136) and the mean for the Treatment (FFE + lit (all)) condition at midline was
7.613938 (SD = 2.39864). The difference between Control (Comparison (all)) and Treatment (FFE + lit
(all)) at midline was thus of 0.6888687 points. The p-value for this difference was 2.963609e-05. The change
from the baseline to the midline of 4.680351 points for the Control (Comparison (all)) condition establishes
the conterfactual against which the change for the Treatment (FFE + lit (all)) can be assessed against.
The change for the Treatment (FFE + lit (all)) from the baseline to the midline was of 4.560072 points.
Consequently, the change for the Control (Comparison (all)) relative to the change for the Treatment (FFE
+ lit (all)) condition was of -0.1202787 points. The p-value for this difference was 0.7427397. Accordingly,
the difference was not statistically significant (p < 0.01 corrected for 5 additional contrasts) and therefore
the statistical test provides no conclusive evidence in regards to there being a difference in the rate of change
between the baseline and the midline for the Phonological awareness EGRA subtask across the Control
(Comparison (all)) and the Treatment (FFE + lit (all)) conditions beyond that expected from sampling
error.
27
1.3.2 Contrast 2: FFE + lit (Portuguese) vs Comparison (Portuguese)
Table 12: Phonological awareness
Baseline Midline
Intervention Mean SD n min max Mean SD n min max
Control 2.355 3.309 1047 0 10 6.955 2.617 972 0 10
Treatment 2.764 3.238 1040 0 10 7.452 2.456 1047 0 10
0
2
4
6
Control Treatment
CINTERVENTION
Mean
PERIOD
Baseline
Midline
Phonological awareness 
28
0.0
2.5
5.0
7.5
10.0
Control Treatment
CINTERVENTION
Phonological awareness 
PERIOD
Baseline
Midline
Phonological awareness 
As shown in the table above, for the the Phonological awareness EGRA subtask, the mean for the Con￾trol (Comparison (Portuguese)) condition at baseline was 2.355301 (SD = 3.309386) and the mean for the
Treatment (FFE + lit (Portuguese)) condition at baseline was 2.764423 (SD = 3.238258). The difference
between Control (Comparison (Portuguese)) and Treatment (FFE + lit (Portuguese)) at baseline was thus
of 0.4091222 points. The p-value for this difference was 0.268457. The mean for the Control (Comparison
(Portuguese)) condition at midline was 6.954733 (SD = 2.61738) and the mean for the Treatment (FFE +
lit (Portuguese)) condition at midline was 7.451767 (SD = 2.456372). The difference between Control (Com￾parison (Portuguese)) and Treatment (FFE + lit (Portuguese)) at midline was thus of 0.4970344 points. The
p-value for this difference was 0.00867946. The change from the baseline to the midline of 4.599432 points
for the Control (Comparison (Portuguese)) condition establishes the conterfactual against which the change
for the Treatment (FFE + lit (Portuguese)) can be assessed against. The change for the Treatment (FFE
+ lit (Portuguese)) from the baseline to the midline was of 4.687344 points. Consequently, the change for
the Control (Comparison (Portuguese)) relative to the change for the Treatment (FFE + lit (Portuguese))
condition was of 0.08791223 points. The p-value for this difference was 0.8213137. Accordingly, the difference
was not statistically significant (p < 0.01 corrected for 5 additional contrasts) and therefore the statistical
test provides no conclusive evidence in regards to there being a difference in the rate of change between
the baseline and the midline for the Phonological awareness EGRA subtask across the Control (Comparison
(Portuguese)) and the Treatment (FFE + lit (Portuguese)) conditions beyond that expected from sampling
error.
29
1.3.3 Contrast 3: FFE + lit (Bilingual) vs Comparison (Bilingual)
Table 13: Phonological awareness
Baseline Midline
Intervention Mean SD n min max Mean SD n min max
Control 0.944 2.181 89 0 10 6.661 2.51 109 0 10
Treatment 3.613 3.799 538 0 10 7.837 2.30 761 0 10
0
2
4
6
8
Control Treatment
CINTERVENTION
Mean
PERIOD
Baseline
Midline
Phonological awareness 
30
0.0
2.5
5.0
7.5
10.0
Control Treatment
CINTERVENTION
Phonological awareness 
PERIOD
Baseline
Midline
Phonological awareness 
As shown in the table above, for the the Phonological awareness EGRA subtask, the mean for the Con￾trol (Comparison (Bilingual)) condition at baseline was 0.9438202 (SD = 2.181323) and the mean for the
Treatment (FFE + lit (Bilingual)) condition at baseline was 3.613383 (SD = 3.799339). The difference
between Control (Comparison (Bilingual)) and Treatment (FFE + lit (Bilingual)) at baseline was thus of
2.669563 points. The p-value for this difference was 6.276787e-07. The mean for the Control (Comparison
(Bilingual)) condition at midline was 6.66055 (SD = 2.510055) and the mean for the Treatment (FFE +
lit (Bilingual)) condition at midline was 7.837057 (SD = 2.299827). The difference between Control (Com￾parison (Bilingual)) and Treatment (FFE + lit (Bilingual)) at midline was thus of 1.176506 points. The
p-value for this difference was 1.406602e-05. The change from the baseline to the midline of 5.71673 points
for the Control (Comparison (Bilingual)) condition establishes the conterfactual against which the change
for the Treatment (FFE + lit (Bilingual)) can be assessed against. The change for the Treatment (FFE +
lit (Bilingual)) from the baseline to the midline was of 4.223674 points. Consequently, the change for the
Control (Comparison (Bilingual)) relative to the change for the Treatment (FFE + lit (Bilingual)) condition
was of -1.493057 points. The p-value for this difference was 0.02097117. Accordingly, the difference was not
statistically significant (p < 0.01 corrected for 5 additional contrasts) and therefore the statistical test pro￾vides no conclusive evidence in regards to there being a difference in the rate of change between the baseline
and the midline for the Phonological awareness EGRA subtask across the Control (Comparison (Bilingual))
and the Treatment (FFE + lit (Bilingual)) conditions beyond that expected from sampling error.
31
1.3.4 Contrast 4: FFE + lit (Portuguese) vs FFE only (Portuguese)
Table 14: Phonological awareness
Baseline Midline
Intervention Mean SD n min max Mean SD n min max
Control 3.180 3.233 1007 0 10 6.294 2.757 952 0 10
Treatment 2.764 3.238 1040 0 10 7.452 2.456 1047 0 10
0
2
4
6
Control Treatment
CINTERVENTION
Mean
PERIOD
Baseline
Midline
Phonological awareness 
32
0.0
2.5
5.0
7.5
10.0
Control Treatment
CINTERVENTION
Phonological awareness 
PERIOD
Baseline
Midline
Phonological awareness 
As shown in the table above, for the the Phonological awareness EGRA subtask, the mean for the Control
(FFE only (Portuguese)) condition at baseline was 3.179742 (SD = 3.232918) and the mean for the Treatment
(FFE + lit (Portuguese)) condition at baseline was 2.764423 (SD = 3.238258). The difference between
Control (FFE only (Portuguese)) and Treatment (FFE + lit (Portuguese)) at baseline was thus of 0.4153187
points. The p-value for this difference was 0.1928234. The mean for the Control (FFE only (Portuguese))
condition at midline was 6.294118 (SD = 2.756819) and the mean for the Treatment (FFE + lit (Portuguese))
condition at midline was 7.451767 (SD = 2.456372). The difference between Control (FFE only (Portuguese))
and Treatment (FFE + lit (Portuguese)) at midline was thus of 1.157649 points. The p-value for this
difference was 2.845414e-08. The change from the baseline to the midline of 3.114376 points for the Control
(FFE only (Portuguese)) condition establishes the conterfactual against which the change for the Treatment
(FFE + lit (Portuguese)) can be assessed against. The change for the Treatment (FFE + lit (Portuguese))
from the baseline to the midline was of 4.687344 points. Consequently, the change for the Control (FFE
only (Portuguese)) relative to the change for the Treatment (FFE + lit (Portuguese)) condition was of
1.572968 points. The p-value for this difference was 9.488215e-05. Accordingly, the difference was statistically
significant (p < 0.01 corrected for 4 additional contrasts) and thus there is evidence of a different degree of
change from baseline to midline in the Treatment (FFE + lit (Portuguese)) when compared to the Control
(FFE only (Portuguese)) condition. This provides evidence that the Treatment (FFE + lit (Portuguese))
(or some other unobserved process) impacted on the Phonological awareness EGRA subtask and thus the
rate of change of the two conditions relative to their baselines differed.
33
1.3.5 Contrast 5: FFE + lit (Bilingual) vs FFE + lit (Portuguese)
Table 15: Phonological awareness
Baseline Midline
Intervention Mean SD n min max Mean SD n min max
Control 2.764 3.238 1040 0 10 7.452 2.456 1047 0 10
Treatment 3.613 3.799 538 0 10 7.837 2.300 761 0 10
0
2
4
6
8
Control Treatment
CINTERVENTION
Mean
PERIOD
Baseline
Midline
Phonological awareness 
34
0.0
2.5
5.0
7.5
10.0
Control Treatment
CINTERVENTION
Phonological awareness 
PERIOD
Baseline
Midline
Phonological awareness 
As shown in the table above, for the the Phonological awareness EGRA subtask, the mean for the Control
(FFE + lit (Portuguese)) condition at baseline was 2.764423 (SD = 3.238258) and the mean for the Treatment
(FFE + lit (Bilingual)) condition at baseline was 3.613383 (SD = 3.799339). The difference between Control
(FFE + lit (Portuguese)) and Treatment (FFE + lit (Bilingual)) at baseline was thus of 0.8489598 points.
The p-value for this difference was 0.04410332. The mean for the Control (FFE + lit (Portuguese)) condition
at midline was 7.451767 (SD = 2.456372) and the mean for the Treatment (FFE + lit (Bilingual)) condition
at midline was 7.837057 (SD = 2.299827). The difference between Control (FFE + lit (Portuguese)) and
Treatment (FFE + lit (Bilingual)) at midline was thus of 0.3852896 points. The p-value for this difference
was 0.02207021. The change from the baseline to the midline of 4.687344 points for the Control (FFE + lit
(Portuguese)) condition establishes the conterfactual against which the change for the Treatment (FFE + lit
(Bilingual)) can be assessed against. The change for the Treatment (FFE + lit (Bilingual)) from the baseline
to the midline was of 4.223674 points. Consequently, the change for the Control (FFE + lit (Portuguese))
relative to the change for the Treatment (FFE + lit (Bilingual)) condition was of -0.4636703 points. The
p-value for this difference was 0.3129217. Accordingly, the difference was not statistically significant (p <
0.01 corrected for 5 additional contrasts) and therefore the statistical test provides no conclusive evidence
in regards to there being a difference in the rate of change between the baseline and the midline for the
Phonological awareness EGRA subtask across the Control (FFE + lit (Portuguese)) and the Treatment
(FFE + lit (Bilingual)) conditions beyond that expected from sampling error.
35
1.4 EGRA_ST4: Concepts of print
1.4.1 Contrast 1: FFE + lit (all) vs Comparison (all)
Table 16: Concepts of print
Baseline Midline
Intervention Mean SD n min max Mean SD n min max
Control 3.589 2.675 1136 0 10 7.107 2.333 1081 0 10
Treatment 3.864 3.163 1578 0 10 7.512 2.208 1808 0 10
0
2
4
6
Control Treatment
CINTERVENTION
Mean
PERIOD
Baseline
Midline
Concepts of print
36
0.0
2.5
5.0
7.5
10.0
Control Treatment
CINTERVENTION
Concepts of print
PERIOD
Baseline
Midline
Concepts of print
As shown in the table above, for the the Concepts of print EGRA subtask, the mean for the Control
(Comparison (all)) condition at baseline was 3.588908 (SD = 2.674899) and the mean for the Treatment
(FFE + lit (all)) condition at baseline was 3.864385 (SD = 3.162977). The difference between Control
(Comparison (all)) and Treatment (FFE + lit (all)) at baseline was thus of 0.2754768 points. The p-value
for this difference was 0.2499473. The mean for the Control (Comparison (all)) condition at midline was
7.107308 (SD = 2.333244) and the mean for the Treatment (FFE + lit (all)) condition at midline was
7.512168 (SD = 2.208267). The difference between Control (Comparison (all)) and Treatment (FFE + lit
(all)) at midline was thus of 0.4048601 points. The p-value for this difference was 0.008410878. The change
from the baseline to the midline of 3.5184 points for the Control (Comparison (all)) condition establishes
the conterfactual against which the change for the Treatment (FFE + lit (all)) can be assessed against.
The change for the Treatment (FFE + lit (all)) from the baseline to the midline was of 3.647783 points.
Consequently, the change for the Control (Comparison (all)) relative to the change for the Treatment (FFE
+ lit (all)) condition was of 0.1293832 points. The p-value for this difference was 0.6381239. Accordingly,
the difference was not statistically significant (p < 0.01 corrected for 5 additional contrasts) and therefore
the statistical test provides no conclusive evidence in regards to there being a difference in the rate of
change between the baseline and the midline for the Concepts of print EGRA subtask across the Control
(Comparison (all)) and the Treatment (FFE + lit (all)) conditions beyond that expected from sampling
error.
37
1.4.2 Contrast 2: FFE + lit (Portuguese) vs Comparison (Portuguese)
Table 17: Concepts of print
Baseline Midline
Intervention Mean SD n min max Mean SD n min max
Control 3.585 2.676 1047 0 10 7.205 2.290 972 0 10
Treatment 3.855 3.143 1040 0 10 7.419 2.219 1047 0 10
0
2
4
6
Control Treatment
CINTERVENTION
Mean
PERIOD
Baseline
Midline
Concepts of print
38
0.0
2.5
5.0
7.5
10.0
Control Treatment
CINTERVENTION
Concepts of print
PERIOD
Baseline
Midline
Concepts of print
As shown in the table above, for the the Concepts of print EGRA subtask, the mean for the Control (Com￾parison (Portuguese)) condition at baseline was 3.584527 (SD = 2.67598) and the mean for the Treatment
(FFE + lit (Portuguese)) condition at baseline was 3.854808 (SD = 3.143209). The difference between Con￾trol (Comparison (Portuguese)) and Treatment (FFE + lit (Portuguese)) at baseline was thus of 0.2702805
points. The p-value for this difference was 0.3261805. The mean for the Control (Comparison (Portuguese))
condition at midline was 7.204733 (SD = 2.290053) and the mean for the Treatment (FFE + lit (Por￾tuguese)) condition at midline was 7.419293 (SD = 2.21867). The difference between Control (Comparison
(Portuguese)) and Treatment (FFE + lit (Portuguese)) at midline was thus of 0.2145607 points. The p-value
for this difference was 0.2067414. The change from the baseline to the midline of 3.620205 points for the
Control (Comparison (Portuguese)) condition establishes the conterfactual against which the change for the
Treatment (FFE + lit (Portuguese)) can be assessed against. The change for the Treatment (FFE + lit (Por￾tuguese)) from the baseline to the midline was of 3.564486 points. Consequently, the change for the Control
(Comparison (Portuguese)) relative to the change for the Treatment (FFE + lit (Portuguese)) condition was
of -0.05571976 points. The p-value for this difference was 0.8602412. Accordingly, the difference was not sta￾tistically significant (p < 0.01 corrected for 5 additional contrasts) and therefore the statistical test provides
no conclusive evidence in regards to there being a difference in the rate of change between the baseline and
the midline for the Concepts of print EGRA subtask across the Control (Comparison (Portuguese)) and the
Treatment (FFE + lit (Portuguese)) conditions beyond that expected from sampling error.
39
1.4.3 Contrast 3: FFE + lit (Bilingual) vs Comparison (Bilingual)
Table 18: Concepts of print
Baseline Midline
Intervention Mean SD n min max Mean SD n min max
Control 3.640 2.677 89 0 10 6.239 2.538 109 0 10
Treatment 3.883 3.204 538 0 10 7.640 2.189 761 0 10
0
2
4
6
8
Control Treatment
CINTERVENTION
Mean
PERIOD
Baseline
Midline
Concepts of print
40
0.0
2.5
5.0
7.5
10.0
Control Treatment
CINTERVENTION
Concepts of print
PERIOD
Baseline
Midline
Concepts of print
As shown in the table above, for the the Concepts of print EGRA subtask, the mean for the Control
(Comparison (Bilingual)) condition at baseline was 3.640449 (SD = 2.676697) and the mean for the Treatment
(FFE + lit (Bilingual)) condition at baseline was 3.8829 (SD = 3.203706). The difference between Control
(Comparison (Bilingual)) and Treatment (FFE + lit (Bilingual)) at baseline was thus of 0.2424502 points.
The p-value for this difference was 0.7076607. The mean for the Control (Comparison (Bilingual)) condition
at midline was 6.238532 (SD = 2.538223) and the mean for the Treatment (FFE + lit (Bilingual)) condition
at midline was 7.639947 (SD = 2.188891). The difference between Control (Comparison (Bilingual)) and
Treatment (FFE + lit (Bilingual)) at midline was thus of 1.401415 points. The p-value for this difference was
1.156373e-06. The change from the baseline to the midline of 2.598083 points for the Control (Comparison
(Bilingual)) condition establishes the conterfactual against which the change for the Treatment (FFE + lit
(Bilingual)) can be assessed against. The change for the Treatment (FFE + lit (Bilingual)) from the baseline
to the midline was of 3.757048 points. Consequently, the change for the Control (Comparison (Bilingual))
relative to the change for the Treatment (FFE + lit (Bilingual)) condition was of 1.158965 points. The
p-value for this difference was 0.06620163. Accordingly, the difference was not statistically significant (p <
0.01 corrected for 5 additional contrasts) and therefore the statistical test provides no conclusive evidence
in regards to there being a difference in the rate of change between the baseline and the midline for the
Concepts of print EGRA subtask across the Control (Comparison (Bilingual)) and the Treatment (FFE +
lit (Bilingual)) conditions beyond that expected from sampling error.
41
1.4.4 Contrast 4: FFE + lit (Portuguese) vs FFE only (Portuguese)
Table 19: Concepts of print
Baseline Midline
Intervention Mean SD n min max Mean SD n min max
Control 3.680 3.092 1007 0 10 6.348 2.453 952 0 10
Treatment 3.855 3.143 1040 0 10 7.419 2.219 1047 0 10
0
2
4
6
Control Treatment
CINTERVENTION
Mean
PERIOD
Baseline
Midline
Concepts of print
42
0.0
2.5
5.0
7.5
10.0
Control Treatment
CINTERVENTION
Concepts of print
PERIOD
Baseline
Midline
Concepts of print
As shown in the table above, for the the Concepts of print EGRA subtask, the mean for the Control (FFE only
(Portuguese)) condition at baseline was 3.680238 (SD = 3.091551) and the mean for the Treatment (FFE +
lit (Portuguese)) condition at baseline was 3.854808 (SD = 3.143209). The difference between Control (FFE
only (Portuguese)) and Treatment (FFE + lit (Portuguese)) at baseline was thus of 0.1745694 points. The
p-value for this difference was 0.5574752. The mean for the Control (FFE only (Portuguese)) condition at
midline was 6.347689 (SD = 2.452903) and the mean for the Treatment (FFE + lit (Portuguese)) condition
at midline was 7.419293 (SD = 2.21867). The difference between Control (FFE only (Portuguese)) and
Treatment (FFE + lit (Portuguese)) at midline was thus of 1.071604 points. The p-value for this difference
was 4.766178e-10. The change from the baseline to the midline of 2.667451 points for the Control (FFE
only (Portuguese)) condition establishes the conterfactual against which the change for the Treatment (FFE
+ lit (Portuguese)) can be assessed against. The change for the Treatment (FFE + lit (Portuguese)) from
the baseline to the midline was of 3.564486 points. Consequently, the change for the Control (FFE only
(Portuguese)) relative to the change for the Treatment (FFE + lit (Portuguese)) condition was of 0.8970348
points. The p-value for this difference was 0.01683061. Accordingly, the difference was not statistically
significant (p < 0.01 corrected for 5 additional contrasts) and therefore the statistical test provides no
conclusive evidence in regards to there being a difference in the rate of change between the baseline and
the midline for the Concepts of print EGRA subtask across the Control (FFE only (Portuguese)) and the
Treatment (FFE + lit (Portuguese)) conditions beyond that expected from sampling error.
43
1.4.5 Contrast 5: FFE + lit (Bilingual) vs FFE + lit (Portuguese)
Table 20: Concepts of print
Baseline Midline
Intervention Mean SD n min max Mean SD n min max
Control 3.855 3.143 1040 0 10 7.419 2.219 1047 0 10
Treatment 3.883 3.204 538 0 10 7.640 2.189 761 0 10
0
2
4
6
8
Control Treatment
CINTERVENTION
Mean
PERIOD
Baseline
Midline
Concepts of print
44
0.0
2.5
5.0
7.5
10.0
Control Treatment
CINTERVENTION
Concepts of print
PERIOD
Baseline
Midline
Concepts of print
As shown in the table above, for the the Concepts of print EGRA subtask, the mean for the Control (FFE
+ lit (Portuguese)) condition at baseline was 3.854808 (SD = 3.143209) and the mean for the Treatment
(FFE + lit (Bilingual)) condition at baseline was 3.8829 (SD = 3.203706). The difference between Control
(FFE + lit (Portuguese)) and Treatment (FFE + lit (Bilingual)) at baseline was thus of 0.02809194 points.
The p-value for this difference was 0.9391503. The mean for the Control (FFE + lit (Portuguese)) condition
at midline was 7.419293 (SD = 2.21867) and the mean for the Treatment (FFE + lit (Bilingual)) condition
at midline was 7.639947 (SD = 2.188891). The difference between Control (FFE + lit (Portuguese)) and
Treatment (FFE + lit (Bilingual)) at midline was thus of 0.2206542 points. The p-value for this difference
was 0.2185148. The change from the baseline to the midline of 3.564486 points for the Control (FFE + lit
(Portuguese)) condition establishes the conterfactual against which the change for the Treatment (FFE + lit
(Bilingual)) can be assessed against. The change for the Treatment (FFE + lit (Bilingual)) from the baseline
to the midline was of 3.757048 points. Consequently, the change for the Control (FFE + lit (Portuguese))
relative to the change for the Treatment (FFE + lit (Bilingual)) condition was of 0.1925623 points. The
p-value for this difference was 0.646684. Accordingly, the difference was not statistically significant (p <
0.01 corrected for 5 additional contrasts) and therefore the statistical test provides no conclusive evidence
in regards to there being a difference in the rate of change between the baseline and the midline for the
Concepts of print EGRA subtask across the Control (FFE + lit (Portuguese)) and the Treatment (FFE +
lit (Bilingual)) conditions beyond that expected from sampling error.
45
1.5 EGRA_ST5: Knowledge of letter sounds
1.5.1 Contrast 1: FFE + lit (all) vs Comparison (all)
Table 21: Knowledge of letter sounds
Baseline Midline
Intervention Mean SD n min max Mean SD n min max
Control 0.278 3.271 1136 0 100 6.355 11.876 1081 0 100
Treatment 0.341 4.572 1578 0 97 8.645 14.183 1808 0 100
0.0
2.5
5.0
7.5
Control Treatment
CINTERVENTION
Mean
PERIOD
Baseline
Midline
Knowledge of letter sounds
46
0
25
50
75
100
Control Treatment
CINTERVENTION
Knowledge of letter sounds
PERIOD
Baseline
Midline
Knowledge of letter sounds
As shown in the table above, for the the Knowledge of letter sounds EGRA subtask, the mean for the Control
(Comparison (all)) condition at baseline was 0.278169 (SD = 3.270762) and the mean for the Treatment
(FFE + lit (all)) condition at baseline was 0.3409379 (SD = 4.572018). The difference between Control
(Comparison (all)) and Treatment (FFE + lit (all)) at baseline was thus of 0.06276888 points. The p-value
for this difference was 0.7037663. The mean for the Control (Comparison (all)) condition at midline was
6.355227 (SD = 11.87612) and the mean for the Treatment (FFE + lit (all)) condition at midline was
8.644912 (SD = 14.18262). The difference between Control (Comparison (all)) and Treatment (FFE + lit
(all)) at midline was thus of 2.289685 points. The p-value for this difference was 0.04166406. The change
from the baseline to the midline of 6.077058 points for the Control (Comparison (all)) condition establishes
the conterfactual against which the change for the Treatment (FFE + lit (all)) can be assessed against.
The change for the Treatment (FFE + lit (all)) from the baseline to the midline was of 8.303974 points.
Consequently, the change for the Control (Comparison (all)) relative to the change for the Treatment (FFE
+ lit (all)) condition was of 2.226916 points. The p-value for this difference was 0.04633249. Accordingly,
the difference was not statistically significant (p < 0.01 corrected for 5 additional contrasts) and therefore
the statistical test provides no conclusive evidence in regards to there being a difference in the rate of change
between the baseline and the midline for the Knowledge of letter sounds EGRA subtask across the Control
(Comparison (all)) and the Treatment (FFE + lit (all)) conditions beyond that expected from sampling
error.
47
1.5.2 Contrast 2: FFE + lit (Portuguese) vs Comparison (Portuguese)
Table 22: Knowledge of letter sounds
Baseline Midline
Intervention Mean SD n min max Mean SD n min max
Control 0.298 3.404 1047 0 100 6.726 12.243 972 0 100
Treatment 0.137 2.145 1040 0 65 8.559 14.534 1047 0 100
0.0
2.5
5.0
7.5
Control Treatment
CINTERVENTION
Mean
PERIOD
Baseline
Midline
Knowledge of letter sounds
48
0
25
50
75
100
Control Treatment
CINTERVENTION
Knowledge of letter sounds
PERIOD
Baseline
Midline
Knowledge of letter sounds
As shown in the table above, for the the Knowledge of letter sounds EGRA subtask, the mean for the
Control (Comparison (Portuguese)) condition at baseline was 0.2979943 (SD = 3.404112) and the mean for
the Treatment (FFE + lit (Portuguese)) condition at baseline was 0.1365385 (SD = 2.145033). The difference
between Control (Comparison (Portuguese)) and Treatment (FFE + lit (Portuguese)) at baseline was thus
of 0.1614558 points. The p-value for this difference was 0.2308036. The mean for the Control (Comparison
(Portuguese)) condition at midline was 6.726337 (SD = 12.24287) and the mean for the Treatment (FFE
+ lit (Portuguese)) condition at midline was 8.558739 (SD = 14.53446). The difference between Control
(Comparison (Portuguese)) and Treatment (FFE + lit (Portuguese)) at midline was thus of 1.832402 points.
The p-value for this difference was 0.1684851. The change from the baseline to the midline of 6.428343 points
for the Control (Comparison (Portuguese)) condition establishes the conterfactual against which the change
for the Treatment (FFE + lit (Portuguese)) can be assessed against. The change for the Treatment (FFE
+ lit (Portuguese)) from the baseline to the midline was of 8.422201 points. Consequently, the change for
the Control (Comparison (Portuguese)) relative to the change for the Treatment (FFE + lit (Portuguese))
condition was of 1.993858 points. The p-value for this difference was 0.1320772. Accordingly, the difference
was not statistically significant (p < 0.01 corrected for 5 additional contrasts) and therefore the statistical
test provides no conclusive evidence in regards to there being a difference in the rate of change between the
baseline and the midline for the Knowledge of letter sounds EGRA subtask across the Control (Comparison
(Portuguese)) and the Treatment (FFE + lit (Portuguese)) conditions beyond that expected from sampling
error.
49
1.5.3 Contrast 3: FFE + lit (Bilingual) vs Comparison (Bilingual)
Table 23: Knowledge of letter sounds
Baseline Midline
Intervention Mean SD n min max Mean SD n min max
Control 0.045 0.424 89 0 4 3.046 7.108 109 0 50
Treatment 0.736 7.228 538 0 97 8.763 13.692 761 0 99
0.0
2.5
5.0
7.5
Control Treatment
CINTERVENTION
Mean
PERIOD
Baseline
Midline
Knowledge of letter sounds
50
0
25
50
75
100
Control Treatment
CINTERVENTION
Knowledge of letter sounds
PERIOD
Baseline
Midline
Knowledge of letter sounds
As shown in the table above, for the the Knowledge of letter sounds EGRA subtask, the mean for the
Control (Comparison (Bilingual)) condition at baseline was 0.04494382 (SD = 0.4239992) and the mean
for the Treatment (FFE + lit (Bilingual)) condition at baseline was 0.7360595 (SD = 7.228193). The
difference between Control (Comparison (Bilingual)) and Treatment (FFE + lit (Bilingual)) at baseline
was thus of 0.6911157 points. The p-value for this difference was 0.04152365. The mean for the Control
(Comparison (Bilingual)) condition at midline was 3.045872 (SD = 7.10814) and the mean for the Treatment
(FFE + lit (Bilingual)) condition at midline was 8.763469 (SD = 13.69236). The difference between Control
(Comparison (Bilingual)) and Treatment (FFE + lit (Bilingual)) at midline was thus of 5.717598 points. The
p-value for this difference was 4.525924e-05. The change from the baseline to the midline of 3.000928 points
for the Control (Comparison (Bilingual)) condition establishes the conterfactual against which the change
for the Treatment (FFE + lit (Bilingual)) can be assessed against. The change for the Treatment (FFE +
lit (Bilingual)) from the baseline to the midline was of 8.02741 points. Consequently, the change for the
Control (Comparison (Bilingual)) relative to the change for the Treatment (FFE + lit (Bilingual)) condition
was of 5.026482 points. The p-value for this difference was 0.0004872545. Accordingly, the difference was
statistically significant (p < 0.01 corrected for 4 additional contrasts) and thus there is evidence of a different
degree of change from baseline to midline in the Treatment (FFE + lit (Bilingual)) when compared to
the Control (Comparison (Bilingual)) condition. This provides evidence that the Treatment (FFE + lit
(Bilingual)) (or some other unobserved process) impacted on the Knowledge of letter sounds EGRA subtask
and thus the rate of change of the two conditions relative to their baselines differed.
51
1.5.4 Contrast 4: FFE + lit (Portuguese) vs FFE only (Portuguese)
Table 24: Knowledge of letter sounds
Baseline Midline
Intervention Mean SD n min max Mean SD n min max
Control 0.052 0.514 1007 0 6 4.263 8.859 952 0 84
Treatment 0.137 2.145 1040 0 65 8.559 14.534 1047 0 100
0.0
2.5
5.0
7.5
Control Treatment
CINTERVENTION
Mean
PERIOD
Baseline
Midline
Knowledge of letter sounds
52
0
25
50
75
100
Control Treatment
CINTERVENTION
Knowledge of letter sounds
PERIOD
Baseline
Midline
Knowledge of letter sounds
As shown in the table above, for the the Knowledge of letter sounds EGRA subtask, the mean for the
Control (FFE only (Portuguese)) condition at baseline was 0.05163853 (SD = 0.5135488) and the mean
for the Treatment (FFE + lit (Portuguese)) condition at baseline was 0.1365385 (SD = 2.145033). The
difference between Control (FFE only (Portuguese)) and Treatment (FFE + lit (Portuguese)) at baseline
was thus of 0.08489993 points. The p-value for this difference was 0.2218159. The mean for the Control (FFE
only (Portuguese)) condition at midline was 4.262605 (SD = 8.85853) and the mean for the Treatment (FFE
+ lit (Portuguese)) condition at midline was 8.558739 (SD = 14.53446). The difference between Control
(FFE only (Portuguese)) and Treatment (FFE + lit (Portuguese)) at midline was thus of 4.296134 points.
The p-value for this difference was 6.7168e-05. The change from the baseline to the midline of 4.210967
points for the Control (FFE only (Portuguese)) condition establishes the conterfactual against which the
change for the Treatment (FFE + lit (Portuguese)) can be assessed against. The change for the Treatment
(FFE + lit (Portuguese)) from the baseline to the midline was of 8.422201 points. Consequently, the change
for the Control (FFE only (Portuguese)) relative to the change for the Treatment (FFE + lit (Portuguese))
condition was of 4.211234 points. The p-value for this difference was 9.450926e-05. Accordingly, the difference
was statistically significant (p < 0.01 corrected for 4 additional contrasts) and thus there is evidence of a
different degree of change from baseline to midline in the Treatment (FFE + lit (Portuguese)) when compared
to the Control (FFE only (Portuguese)) condition. This provides evidence that the Treatment (FFE + lit
(Portuguese)) (or some other unobserved process) impacted on the Knowledge of letter sounds EGRA subtask
and thus the rate of change of the two conditions relative to their baselines differed.
53
1.5.5 Contrast 5: FFE + lit (Bilingual) vs FFE + lit (Portuguese)
Table 25: Knowledge of letter sounds
Baseline Midline
Intervention Mean SD n min max Mean SD n min max
Control 0.137 2.145 1040 0 65 8.559 14.534 1047 0 100
Treatment 0.736 7.228 538 0 97 8.763 13.692 761 0 99
0.0
2.5
5.0
7.5
Control Treatment
CINTERVENTION
Mean
PERIOD
Baseline
Midline
Knowledge of letter sounds
54
0
25
50
75
100
Control Treatment
CINTERVENTION
Knowledge of letter sounds
PERIOD
Baseline
Midline
Knowledge of letter sounds
As shown in the table above, for the the Knowledge of letter sounds EGRA subtask, the mean for the
Control (FFE + lit (Portuguese)) condition at baseline was 0.1365385 (SD = 2.145033) and the mean for
the Treatment (FFE + lit (Bilingual)) condition at baseline was 0.7360595 (SD = 7.228193). The difference
between Control (FFE + lit (Portuguese)) and Treatment (FFE + lit (Bilingual)) at baseline was thus of
0.599521 points. The p-value for this difference was 0.08007998. The mean for the Control (FFE + lit
(Portuguese)) condition at midline was 8.558739 (SD = 14.53446) and the mean for the Treatment (FFE +
lit (Bilingual)) condition at midline was 8.763469 (SD = 13.69236). The difference between Control (FFE
+ lit (Portuguese)) and Treatment (FFE + lit (Bilingual)) at midline was thus of 0.2047299 points. The
p-value for this difference was 0.8862875. The change from the baseline to the midline of 8.422201 points
for the Control (FFE + lit (Portuguese)) condition establishes the conterfactual against which the change
for the Treatment (FFE + lit (Bilingual)) can be assessed against. The change for the Treatment (FFE +
lit (Bilingual)) from the baseline to the midline was of 8.02741 points. Consequently, the change for the
Control (FFE + lit (Portuguese)) relative to the change for the Treatment (FFE + lit (Bilingual)) condition
was of -0.3947912 points. The p-value for this difference was 0.7887049. Accordingly, the difference was
not statistically significant (p < 0.01 corrected for 5 additional contrasts) and therefore the statistical test
provides no conclusive evidence in regards to there being a difference in the rate of change between the
baseline and the midline for the Knowledge of letter sounds EGRA subtask across the Control (FFE + lit
(Portuguese)) and the Treatment (FFE + lit (Bilingual)) conditions beyond that expected from sampling
error.
55
1.6 EGRA_ST6: Knowledge of syllables
1.6.1 Contrast 1: FFE + lit (all) vs Comparison (all)
Table 26: Knowledge of syllables
Baseline Midline
Intervention Mean SD n min max Mean SD n min max
Control 0.074 1.717 1136 0 50 8.404 14.072 1081 0 50
Treatment 0.080 1.633 1578 0 44 7.323 12.905 1808 0 50
0
2
4
6
8
Control Treatment
CINTERVENTION
Mean
PERIOD
Baseline
Midline
Knowledge of syllables
56
0
10
20
30
40
50
Control Treatment
CINTERVENTION
Knowledge of syllables
PERIOD
Baseline
Midline
Knowledge of syllables
As shown in the table above, for the the Knowledge of syllables EGRA subtask, the mean for the Control
(Comparison (all)) condition at baseline was 0.07394366 (SD = 1.717438) and the mean for the Treatment
(FFE + lit (all)) condition at baseline was 0.07984791 (SD = 1.633499). The difference between Control
(Comparison (all)) and Treatment (FFE + lit (all)) at baseline was thus of 0.005904247 points. The p￾value for this difference was 0.9274075. The mean for the Control (Comparison (all)) condition at midline
was 8.404255 (SD = 14.07192) and the mean for the Treatment (FFE + lit (all)) condition at midline was
7.323009 (SD = 12.90529). The difference between Control (Comparison (all)) and Treatment (FFE + lit
(all)) at midline was thus of 1.081246 points. The p-value for this difference was 0.3311634. The change
from the baseline to the midline of 8.330312 points for the Control (Comparison (all)) condition establishes
the conterfactual against which the change for the Treatment (FFE + lit (all)) can be assessed against.
The change for the Treatment (FFE + lit (all)) from the baseline to the midline was of 7.243161 points.
Consequently, the change for the Control (Comparison (all)) relative to the change for the Treatment (FFE
+ lit (all)) condition was of -1.087151 points. The p-value for this difference was 0.3283146. Accordingly,
the difference was not statistically significant (p < 0.01 corrected for 5 additional contrasts) and therefore
the statistical test provides no conclusive evidence in regards to there being a difference in the rate of change
between the baseline and the midline for the Knowledge of syllables EGRA subtask across the Control
(Comparison (all)) and the Treatment (FFE + lit (all)) conditions beyond that expected from sampling
error.
57
1.6.2 Contrast 2: FFE + lit (Portuguese) vs Comparison (Portuguese)
Table 27: Knowledge of syllables
Baseline Midline
Intervention Mean SD n min max Mean SD n min max
Control 0.080 1.789 1047 0 50 9.062 14.499 972 0 50
Treatment 0.073 1.573 1040 0 44 9.713 14.938 1047 0 50
0.0
2.5
5.0
7.5
10.0
Control Treatment
CINTERVENTION
Mean
PERIOD
Baseline
Midline
Knowledge of syllables
58
0
10
20
30
40
50
Control Treatment
CINTERVENTION
Knowledge of syllables
PERIOD
Baseline
Midline
Knowledge of syllables
As shown in the table above, for the the Knowledge of syllables EGRA subtask, the mean for the Control
(Comparison (Portuguese)) condition at baseline was 0.08022923 (SD = 1.788871) and the mean for the
Treatment (FFE + lit (Portuguese)) condition at baseline was 0.07307692 (SD = 1.572883). The difference
between Control (Comparison (Portuguese)) and Treatment (FFE + lit (Portuguese)) at baseline was thus
of 0.007152303 points. The p-value for this difference was 0.9222329. The mean for the Control (Comparison
(Portuguese)) condition at midline was 9.061728 (SD = 14.49906) and the mean for the Treatment (FFE
+ lit (Portuguese)) condition at midline was 9.712512 (SD = 14.93781). The difference between Control
(Comparison (Portuguese)) and Treatment (FFE + lit (Portuguese)) at midline was thus of 0.6507835 points.
The p-value for this difference was 0.6286879. The change from the baseline to the midline of 8.981499 points
for the Control (Comparison (Portuguese)) condition establishes the conterfactual against which the change
for the Treatment (FFE + lit (Portuguese)) can be assessed against. The change for the Treatment (FFE
+ lit (Portuguese)) from the baseline to the midline was of 9.639435 points. Consequently, the change for
the Control (Comparison (Portuguese)) relative to the change for the Treatment (FFE + lit (Portuguese))
condition was of 0.6579358 points. The p-value for this difference was 0.6247483. Accordingly, the difference
was not statistically significant (p < 0.01 corrected for 5 additional contrasts) and therefore the statistical
test provides no conclusive evidence in regards to there being a difference in the rate of change between
the baseline and the midline for the Knowledge of syllables EGRA subtask across the Control (Comparison
(Portuguese)) and the Treatment (FFE + lit (Portuguese)) conditions beyond that expected from sampling
error.
59
1.6.3 Contrast 3: FFE + lit (Bilingual) vs Comparison (Bilingual)
Table 28: Knowledge of syllables
Baseline Midline
Intervention Mean SD n min max Mean SD n min max
Control 0.000 0.000 89 0 0 2.541 7.180 109 0 46
Treatment 0.093 1.746 538 0 39 4.035 8.378 761 0 50
0
1
2
3
4
Control Treatment
CINTERVENTION
Mean
PERIOD
Baseline
Midline
Knowledge of syllables
60
0
10
20
30
40
50
Control Treatment
CINTERVENTION
Knowledge of syllables
PERIOD
Baseline
Midline
Knowledge of syllables
As shown in the table above, for the the Knowledge of syllables EGRA subtask, the mean for the Control
(Comparison (Bilingual)) condition at baseline was 0 (SD = 0) and the mean for the Treatment (FFE
+ lit (Bilingual)) condition at baseline was 0.0929368 (SD = 1.74616). The difference between Control
(Comparison (Bilingual)) and Treatment (FFE + lit (Bilingual)) at baseline was thus of 0.0929368 points.
The p-value for this difference was 0.2070935. The mean for the Control (Comparison (Bilingual)) condition
at midline was 2.541284 (SD = 7.179616) and the mean for the Treatment (FFE + lit (Bilingual)) condition
at midline was 4.03548 (SD = 8.377919). The difference between Control (Comparison (Bilingual)) and
Treatment (FFE + lit (Bilingual)) at midline was thus of 1.494195 points. The p-value for this difference
was 0.1210532. The change from the baseline to the midline of 2.541284 points for the Control (Comparison
(Bilingual)) condition establishes the conterfactual against which the change for the Treatment (FFE + lit
(Bilingual)) can be assessed against. The change for the Treatment (FFE + lit (Bilingual)) from the baseline
to the midline was of 3.942543 points. Consequently, the change for the Control (Comparison (Bilingual))
relative to the change for the Treatment (FFE + lit (Bilingual)) condition was of 1.401258 points. The
p-value for this difference was 0.1464555. Accordingly, the difference was not statistically significant (p <
0.01 corrected for 5 additional contrasts) and therefore the statistical test provides no conclusive evidence
in regards to there being a difference in the rate of change between the baseline and the midline for the
Knowledge of syllables EGRA subtask across the Control (Comparison (Bilingual)) and the Treatment
(FFE + lit (Bilingual)) conditions beyond that expected from sampling error.
61
1.6.4 Contrast 4: FFE + lit (Portuguese) vs FFE only (Portuguese)
Table 29: Knowledge of syllables
Baseline Midline
Intervention Mean SD n min max Mean SD n min max
Control 0.000 0.000 1007 0 0 5.828 11.903 952 0 50
Treatment 0.073 1.573 1040 0 44 9.713 14.938 1047 0 50
0.0
2.5
5.0
7.5
10.0
Control Treatment
CINTERVENTION
Mean
PERIOD
Baseline
Midline
Knowledge of syllables
62
0
10
20
30
40
50
Control Treatment
CINTERVENTION
Knowledge of syllables
PERIOD
Baseline
Midline
Knowledge of syllables
As shown in the table above, for the the Knowledge of syllables EGRA subtask, the mean for the Control
(FFE only (Portuguese)) condition at baseline was 0 (SD = 0) and the mean for the Treatment (FFE + lit
(Portuguese)) condition at baseline was 0.07307692 (SD = 1.572883). The difference between Control (FFE
only (Portuguese)) and Treatment (FFE + lit (Portuguese)) at baseline was thus of 0.07307692 points. The
p-value for this difference was 0.1348755. The mean for the Control (FFE only (Portuguese)) condition at
midline was 5.827731 (SD = 11.90347) and the mean for the Treatment (FFE + lit (Portuguese)) condition
at midline was 9.712512 (SD = 14.93781). The difference between Control (FFE only (Portuguese)) and
Treatment (FFE + lit (Portuguese)) at midline was thus of 3.884781 points. The p-value for this difference
was 0.001668309. The change from the baseline to the midline of 5.827731 points for the Control (FFE
only (Portuguese)) condition establishes the conterfactual against which the change for the Treatment (FFE
+ lit (Portuguese)) can be assessed against. The change for the Treatment (FFE + lit (Portuguese))
from the baseline to the midline was of 9.639435 points. Consequently, the change for the Control (FFE
only (Portuguese)) relative to the change for the Treatment (FFE + lit (Portuguese)) condition was of
3.811704 points. The p-value for this difference was 0.002023932. Accordingly, the difference was statistically
significant (p < 0.01 corrected for 4 additional contrasts) and thus there is evidence of a different degree of
change from baseline to midline in the Treatment (FFE + lit (Portuguese)) when compared to the Control
(FFE only (Portuguese)) condition. This provides evidence that the Treatment (FFE + lit (Portuguese))
(or some other unobserved process) impacted on the Knowledge of syllables EGRA subtask and thus the
rate of change of the two conditions relative to their baselines differed.
63
1.6.5 Contrast 5: FFE + lit (Bilingual) vs FFE + lit (Portuguese)
Table 30: Knowledge of syllables
Baseline Midline
Intervention Mean SD n min max Mean SD n min max
Control 0.073 1.573 1040 0 44 9.713 14.938 1047 0 50
Treatment 0.093 1.746 538 0 39 4.035 8.378 761 0 50
0.0
2.5
5.0
7.5
10.0
Control Treatment
CINTERVENTION
Mean
PERIOD
Baseline
Midline
Knowledge of syllables
64
0
10
20
30
40
50
Control Treatment
CINTERVENTION
Knowledge of syllables
PERIOD
Baseline
Midline
Knowledge of syllables
As shown in the table above, for the the Knowledge of syllables EGRA subtask, the mean for the Control
(FFE + lit (Portuguese)) condition at baseline was 0.07307692 (SD = 1.572883) and the mean for the
Treatment (FFE + lit (Bilingual)) condition at baseline was 0.0929368 (SD = 1.74616). The difference
between Control (FFE + lit (Portuguese)) and Treatment (FFE + lit (Bilingual)) at baseline was thus of
0.01985988 points. The p-value for this difference was 0.8216617. The mean for the Control (FFE + lit
(Portuguese)) condition at midline was 9.712512 (SD = 14.93781) and the mean for the Treatment (FFE
+ lit (Bilingual)) condition at midline was 4.03548 (SD = 8.377919). The difference between Control (FFE
+ lit (Portuguese)) and Treatment (FFE + lit (Bilingual)) at midline was thus of 5.677032 points. The
p-value for this difference was 7.226164e-08. The change from the baseline to the midline of 9.639435 points
for the Control (FFE + lit (Portuguese)) condition establishes the conterfactual against which the change
for the Treatment (FFE + lit (Bilingual)) can be assessed against. The change for the Treatment (FFE +
lit (Bilingual)) from the baseline to the midline was of 3.942543 points. Consequently, the change for the
Control (FFE + lit (Portuguese)) relative to the change for the Treatment (FFE + lit (Bilingual)) condition
was of -5.696892 points. The p-value for this difference was 6.542178e-08. Accordingly, the difference
was statistically significant (p < 0.01 corrected for 4 additional contrasts) and thus there is evidence of a
different degree of change from baseline to midline in the Treatment (FFE + lit (Bilingual)) when compared
to the Control (FFE + lit (Portuguese)) condition. This provides evidence that the Treatment (FFE + lit
(Bilingual)) (or some other unobserved process) impacted on the Knowledge of syllables EGRA subtask and
thus the rate of change of the two conditions relative to their baselines differed. It should be pointed out that
the difference in the rate of change of the Treatment (FFE + lit (Bilingual)) relative to the Control (FFE
+ lit (Portuguese)) was negative and thus suggests that students in the Control (FFE + lit (Portuguese))
schools performed significantly better than those in the Treatment (FFE + lit (Bilingual)) .
65
1.7 EGRA_ST7: Reading of words
1.7.1 Contrast 1: FFE + lit (all) vs Comparison (all)
Table 31: Reading of words
Baseline Midline
Intervention Mean SD n min max Mean SD n min max
Control 0.035 0.710 1136 0 18 6.027 10.347 1081 0 30
Treatment 0.034 0.814 1578 0 28 7.570 11.225 1808 0 30
0
2
4
6
Control Treatment
CINTERVENTION
Mean
PERIOD
Baseline
Midline
Reading of words
66
0
10
20
30
Control Treatment
CINTERVENTION
Reading of words
PERIOD
Baseline
Midline
Reading of words
As shown in the table above, for the the Reading of words EGRA subtask, the mean for the Control
(Comparison (all)) condition at baseline was 0.03521127 (SD = 0.7102717) and the mean for the Treatment
(FFE + lit (all)) condition at baseline was 0.03422053 (SD = 0.8144821). The difference between Control
(Comparison (all)) and Treatment (FFE + lit (all)) at baseline was thus of 0.0009907353 points. The p￾value for this difference was 0.9728826. The mean for the Control (Comparison (all)) condition at midline
was 6.026827 (SD = 10.3466) and the mean for the Treatment (FFE + lit (all)) condition at midline was
7.56969 (SD = 11.22461). The difference between Control (Comparison (all)) and Treatment (FFE + lit
(all)) at midline was thus of 1.542863 points. The p-value for this difference was 0.06638155. The change
from the baseline to the midline of 5.991616 points for the Control (Comparison (all)) condition establishes
the conterfactual against which the change for the Treatment (FFE + lit (all)) can be assessed against.
The change for the Treatment (FFE + lit (all)) from the baseline to the midline was of 7.53547 points.
Consequently, the change for the Control (Comparison (all)) relative to the change for the Treatment (FFE
+ lit (all)) condition was of 1.543854 points. The p-value for this difference was 0.06591212. Accordingly,
the difference was not statistically significant (p < 0.01 corrected for 5 additional contrasts) and therefore
the statistical test provides no conclusive evidence in regards to there being a difference in the rate of
change between the baseline and the midline for the Reading of words EGRA subtask across the Control
(Comparison (all)) and the Treatment (FFE + lit (all)) conditions beyond that expected from sampling
error.
67
1.7.2 Contrast 2: FFE + lit (Portuguese) vs Comparison (Portuguese)
Table 32: Reading of words
Baseline Midline
Intervention Mean SD n min max Mean SD n min max
Control 0.038 0.740 1047 0 18 6.438 10.635 972 0 30
Treatment 0.035 0.887 1040 0 28 7.128 11.039 1047 0 30
0
2
4
6
Control Treatment
CINTERVENTION
Mean
PERIOD
Baseline
Midline
Reading of words
68
0
10
20
30
Control Treatment
CINTERVENTION
Reading of words
PERIOD
Baseline
Midline
Reading of words
As shown in the table above, for the the Reading of words EGRA subtask, the mean for the Control
(Comparison (Portuguese)) condition at baseline was 0.03820439 (SD = 0.7397947) and the mean for the
Treatment (FFE + lit (Portuguese)) condition at baseline was 0.03461538 (SD = 0.8866206). The difference
between Control (Comparison (Portuguese)) and Treatment (FFE + lit (Portuguese)) at baseline was thus of
0.003589009 points. The p-value for this difference was 0.9193661. The mean for the Control (Comparison
(Portuguese)) condition at midline was 6.438272 (SD = 10.6351) and the mean for the Treatment (FFE
+ lit (Portuguese)) condition at midline was 7.127985 (SD = 11.0389). The difference between Control
(Comparison (Portuguese)) and Treatment (FFE + lit (Portuguese)) at midline was thus of 0.6897131
points. The p-value for this difference was 0.4770892. The change from the baseline to the midline of
6.400067 points for the Control (Comparison (Portuguese)) condition establishes the conterfactual against
which the change for the Treatment (FFE + lit (Portuguese)) can be assessed against. The change for the
Treatment (FFE + lit (Portuguese)) from the baseline to the midline was of 7.093369 points. Consequently,
the change for the Control (Comparison (Portuguese)) relative to the change for the Treatment (FFE + lit
(Portuguese)) condition was of 0.6933021 points. The p-value for this difference was 0.4745508. Accordingly,
the difference was not statistically significant (p < 0.01 corrected for 5 additional contrasts) and therefore
the statistical test provides no conclusive evidence in regards to there being a difference in the rate of
change between the baseline and the midline for the Reading of words EGRA subtask across the Control
(Comparison (Portuguese)) and the Treatment (FFE + lit (Portuguese)) conditions beyond that expected
from sampling error.
69
1.7.3 Contrast 3: FFE + lit (Bilingual) vs Comparison (Bilingual)
Table 33: Reading of words
Baseline Midline
Intervention Mean SD n min max Mean SD n min max
Control 0.000 0.000 89 0 0 2.358 6.206 109 0 30
Treatment 0.033 0.654 538 0 15 8.177 11.455 761 0 30
0
2
4
6
8
Control Treatment
CINTERVENTION
Mean
PERIOD
Baseline
Midline
Reading of words
70
0
10
20
30
Control Treatment
CINTERVENTION
Reading of words
PERIOD
Baseline
Midline
Reading of words
As shown in the table above, for the the Reading of words EGRA subtask, the mean for the Control
(Comparison (Bilingual)) condition at baseline was 0 (SD = 0) and the mean for the Treatment (FFE +
lit (Bilingual)) condition at baseline was 0.03345725 (SD = 0.6535931). The difference between Control
(Comparison (Bilingual)) and Treatment (FFE + lit (Bilingual)) at baseline was thus of 0.03345725 points.
The p-value for this difference was 0.2300851. The mean for the Control (Comparison (Bilingual)) condition
at midline was 2.357798 (SD = 6.205617) and the mean for the Treatment (FFE + lit (Bilingual)) condition
at midline was 8.177398 (SD = 11.45466). The difference between Control (Comparison (Bilingual)) and
Treatment (FFE + lit (Bilingual)) at midline was thus of 5.8196 points. The p-value for this difference was
7.35086e-06. The change from the baseline to the midline of 2.357798 points for the Control (Comparison
(Bilingual)) condition establishes the conterfactual against which the change for the Treatment (FFE + lit
(Bilingual)) can be assessed against. The change for the Treatment (FFE + lit (Bilingual)) from the baseline
to the midline was of 8.143941 points. Consequently, the change for the Control (Comparison (Bilingual))
relative to the change for the Treatment (FFE + lit (Bilingual)) condition was of 5.786143 points. The
p-value for this difference was 8.351949e-06. Accordingly, the difference was statistically significant (p <
0.01 corrected for 4 additional contrasts) and thus there is evidence of a different degree of change from
baseline to midline in the Treatment (FFE + lit (Bilingual)) when compared to the Control (Comparison
(Bilingual)) condition. This provides evidence that the Treatment (FFE + lit (Bilingual)) (or some other
unobserved process) impacted on the Reading of words EGRA subtask and thus the rate of change of the
two conditions relative to their baselines differed.
71
1.7.4 Contrast 4: FFE + lit (Portuguese) vs FFE only (Portuguese)
Table 34: Reading of words
Baseline Midline
Intervention Mean SD n min max Mean SD n min max
Control 0.000 0.000 1007 0 0 3.780 8.608 952 0 30
Treatment 0.035 0.887 1040 0 28 7.128 11.039 1047 0 30
0
2
4
6
Control Treatment
CINTERVENTION
Mean
PERIOD
Baseline
Midline
Reading of words
72
0
10
20
30
Control Treatment
CINTERVENTION
Reading of words
PERIOD
Baseline
Midline
Reading of words
As shown in the table above, for the the Reading of words EGRA subtask, the mean for the Control (FFE
only (Portuguese)) condition at baseline was 0 (SD = 0) and the mean for the Treatment (FFE + lit (Por￾tuguese)) condition at baseline was 0.03461538 (SD = 0.8866206). The difference between Control (FFE
only (Portuguese)) and Treatment (FFE + lit (Portuguese)) at baseline was thus of 0.03461538 points. The
p-value for this difference was 0.2085951. The mean for the Control (FFE only (Portuguese)) condition at
midline was 3.780462 (SD = 8.608259) and the mean for the Treatment (FFE + lit (Portuguese)) condition
at midline was 7.127985 (SD = 11.0389). The difference between Control (FFE only (Portuguese)) and
Treatment (FFE + lit (Portuguese)) at midline was thus of 3.347523 points. The p-value for this difference
was 0.0001868595. The change from the baseline to the midline of 3.780462 points for the Control (FFE
only (Portuguese)) condition establishes the conterfactual against which the change for the Treatment (FFE
+ lit (Portuguese)) can be assessed against. The change for the Treatment (FFE + lit (Portuguese)) from
the baseline to the midline was of 7.093369 points. Consequently, the change for the Control (FFE only
(Portuguese)) relative to the change for the Treatment (FFE + lit (Portuguese)) condition was of 3.312907
points. The p-value for this difference was 0.0002150606. Accordingly, the difference was statistically signifi￾cant (p < 0.01 corrected for 4 additional contrasts) and thus there is evidence of a different degree of change
from baseline to midline in the Treatment (FFE + lit (Portuguese)) when compared to the Control (FFE
only (Portuguese)) condition. This provides evidence that the Treatment (FFE + lit (Portuguese)) (or some
other unobserved process) impacted on the Reading of words EGRA subtask and thus the rate of change of
the two conditions relative to their baselines differed.
73
1.7.5 Contrast 5: FFE + lit (Bilingual) vs FFE + lit (Portuguese)
Table 35: Reading of words
Baseline Midline
Intervention Mean SD n min max Mean SD n min max
Control 0.035 0.887 1040 0 28 7.128 11.039 1047 0 30
Treatment 0.033 0.654 538 0 15 8.177 11.455 761 0 30
0
2
4
6
8
Control Treatment
CINTERVENTION
Mean
PERIOD
Baseline
Midline
Reading of words
74
0
10
20
30
Control Treatment
CINTERVENTION
Reading of words
PERIOD
Baseline
Midline
Reading of words
As shown in the table above, for the the Reading of words EGRA subtask, the mean for the Control (FFE + lit
(Portuguese)) condition at baseline was 0.03461538 (SD = 0.8866206) and the mean for the Treatment (FFE
+ lit (Bilingual)) condition at baseline was 0.03345725 (SD = 0.6535931). The difference between Control
(FFE + lit (Portuguese)) and Treatment (FFE + lit (Bilingual)) at baseline was thus of 0.001158136 points.
The p-value for this difference was 0.9761761. The mean for the Control (FFE + lit (Portuguese)) condition
at midline was 7.127985 (SD = 11.0389) and the mean for the Treatment (FFE + lit (Bilingual)) condition
at midline was 8.177398 (SD = 11.45466). The difference between Control (FFE + lit (Portuguese)) and
Treatment (FFE + lit (Bilingual)) at midline was thus of 1.049413 points. The p-value for this difference
was 0.3010063. The change from the baseline to the midline of 7.093369 points for the Control (FFE + lit
(Portuguese)) condition establishes the conterfactual against which the change for the Treatment (FFE + lit
(Bilingual)) can be assessed against. The change for the Treatment (FFE + lit (Bilingual)) from the baseline
to the midline was of 8.143941 points. Consequently, the change for the Control (FFE + lit (Portuguese))
relative to the change for the Treatment (FFE + lit (Bilingual)) condition was of 1.050572 points. The
p-value for this difference was 0.3004288. Accordingly, the difference was not statistically significant (p <
0.01 corrected for 5 additional contrasts) and therefore the statistical test provides no conclusive evidence
in regards to there being a difference in the rate of change between the baseline and the midline for the
Reading of words EGRA subtask across the Control (FFE + lit (Portuguese)) and the Treatment (FFE +
lit (Bilingual)) conditions beyond that expected from sampling error.
75
1.8 EGRA_ST8_1A: Reading fluency
1.8.1 Contrast 1: FFE + lit (all) vs Comparison (all)
Table 36: Reading fluency
Baseline Midline
Intervention Mean SD n min max Mean SD n min max
Control 0.077 1.675 1136 0 52 8.098 16.495 1080 0 70
Treatment 0.049 1.397 1578 0 52 7.605 13.853 1791 0 70
0
2
4
6
8
Control Treatment
CINTERVENTION
Mean
PERIOD
Baseline
Midline
Reading fluency
76
0
20
40
60
Control Treatment
CINTERVENTION
Reading fluency
PERIOD
Baseline
Midline
Reading fluency
As shown in the table above, for the the Reading fluency EGRA subtask, the mean for the Control (Com￾parison (all)) condition at baseline was 0.07746479 (SD = 1.674684) and the mean for the Treatment (FFE
+ lit (all)) condition at baseline was 0.04942966 (SD = 1.39665). The difference between Control (Compar￾ison (all)) and Treatment (FFE + lit (all)) at baseline was thus of 0.02803513 points. The p-value for this
difference was 0.6447485. The mean for the Control (Comparison (all)) condition at midline was 8.098148
(SD = 16.49483) and the mean for the Treatment (FFE + lit (all)) condition at midline was 7.605248 (SD
= 13.8533). The difference between Control (Comparison (all)) and Treatment (FFE + lit (all)) at midline
was thus of 0.4928997 points. The p-value for this difference was 0.6686538. The change from the baseline
to the midline of 8.020683 points for the Control (Comparison (all)) condition establishes the conterfactual
against which the change for the Treatment (FFE + lit (all)) can be assessed against. The change for the
Treatment (FFE + lit (all)) from the baseline to the midline was of 7.555819 points. Consequently, the
change for the Control (Comparison (all)) relative to the change for the Treatment (FFE + lit (all)) con￾dition was of -0.4648646 points. The p-value for this difference was 0.6859271. Accordingly, the difference
was not statistically significant (p < 0.01 corrected for 5 additional contrasts) and therefore the statistical
test provides no conclusive evidence in regards to there being a difference in the rate of change between the
baseline and the midline for the Reading fluency EGRA subtask across the Control (Comparison (all)) and
the Treatment (FFE + lit (all)) conditions beyond that expected from sampling error.
77
1.8.2 Contrast 2: FFE + lit (Portuguese) vs Comparison (Portuguese)
Table 37: Reading fluency
Baseline Midline
Intervention Mean SD n min max Mean SD n min max
Control 0.084 1.744 1047 0 52 8.737 17.047 972 0 70
Treatment 0.053 1.615 1040 0 52 8.223 15.261 1039 0 70
0.0
2.5
5.0
7.5
Control Treatment
CINTERVENTION
Mean
PERIOD
Baseline
Midline
Reading fluency
78
0
20
40
60
Control Treatment
CINTERVENTION
Reading fluency
PERIOD
Baseline
Midline
Reading fluency
As shown in the table above, for the the Reading fluency EGRA subtask, the mean for the Control (Compar￾ison (Portuguese)) condition at baseline was 0.08404967 (SD = 1.744317) and the mean for the Treatment
(FFE + lit (Portuguese)) condition at baseline was 0.05288462 (SD = 1.615043). The difference between Con￾trol (Comparison (Portuguese)) and Treatment (FFE + lit (Portuguese)) at baseline was thus of 0.03116505
points. The p-value for this difference was 0.6718402. The mean for the Control (Comparison (Portuguese))
condition at midline was 8.736626 (SD = 17.04723) and the mean for the Treatment (FFE + lit (Por￾tuguese)) condition at midline was 8.223292 (SD = 15.26094). The difference between Control (Comparison
(Portuguese)) and Treatment (FFE + lit (Portuguese)) at midline was thus of 0.5133339 points. The p-value
for this difference was 0.7056136. The change from the baseline to the midline of 8.652576 points for the
Control (Comparison (Portuguese)) condition establishes the conterfactual against which the change for the
Treatment (FFE + lit (Portuguese)) can be assessed against. The change for the Treatment (FFE + lit (Por￾tuguese)) from the baseline to the midline was of 8.170407 points. Consequently, the change for the Control
(Comparison (Portuguese)) relative to the change for the Treatment (FFE + lit (Portuguese)) condition was
of -0.4821688 points. The p-value for this difference was 0.722265. Accordingly, the difference was not sta￾tistically significant (p < 0.01 corrected for 5 additional contrasts) and therefore the statistical test provides
no conclusive evidence in regards to there being a difference in the rate of change between the baseline and
the midline for the Reading fluency EGRA subtask across the Control (Comparison (Portuguese)) and the
Treatment (FFE + lit (Portuguese)) conditions beyond that expected from sampling error.
79
1.8.3 Contrast 3: FFE + lit (Bilingual) vs Comparison (Bilingual)
Table 38: Reading fluency
Baseline Midline
Intervention Mean SD n min max Mean SD n min max
Control 0.000 0.000 89 0 0 2.352 8.333 108 0 51
Treatment 0.043 0.826 538 0 19 6.751 11.587 752 0 53
0
2
4
6
Control Treatment
CINTERVENTION
Mean
PERIOD
Baseline
Midline
Reading fluency
80
0
20
40
Control Treatment
CINTERVENTION
Reading fluency
PERIOD
Baseline
Midline
Reading fluency
As shown in the table above, for the the Reading fluency EGRA subtask, the mean for the Control (Com￾parison (Bilingual)) condition at baseline was 0 (SD = 0) and the mean for the Treatment (FFE + lit
(Bilingual)) condition at baseline was 0.04275093 (SD = 0.8255879). The difference between Control (Com￾parison (Bilingual)) and Treatment (FFE + lit (Bilingual)) at baseline was thus of 0.04275093 points. The
p-value for this difference was 0.2270531. The mean for the Control (Comparison (Bilingual)) condition at
midline was 2.351852 (SD = 8.333375) and the mean for the Treatment (FFE + lit (Bilingual)) condition
at midline was 6.75133 (SD = 11.58736). The difference between Control (Comparison (Bilingual)) and
Treatment (FFE + lit (Bilingual)) at midline was thus of 4.399478 points. The p-value for this difference
was 0.01060688. The change from the baseline to the midline of 2.351852 points for the Control (Comparison
(Bilingual)) condition establishes the conterfactual against which the change for the Treatment (FFE + lit
(Bilingual)) can be assessed against. The change for the Treatment (FFE + lit (Bilingual)) from the baseline
to the midline was of 6.708579 points. Consequently, the change for the Control (Comparison (Bilingual))
relative to the change for the Treatment (FFE + lit (Bilingual)) condition was of 4.356727 points. The
p-value for this difference was 0.01145581. Accordingly, the difference was not statistically significant (p <
0.01 corrected for 5 additional contrasts) and therefore the statistical test provides no conclusive evidence
in regards to there being a difference in the rate of change between the baseline and the midline for the
Reading fluency EGRA subtask across the Control (Comparison (Bilingual)) and the Treatment (FFE + lit
(Bilingual)) conditions beyond that expected from sampling error.
81
1.8.4 Contrast 4: FFE + lit (Portuguese) vs FFE only (Portuguese)
Table 39: Reading fluency
Baseline Midline
Intervention Mean SD n min max Mean SD n min max
Control 0.000 0.000 1007 0 0 4.916 12.433 951 0 70
Treatment 0.053 1.615 1040 0 52 8.223 15.261 1039 0 70
0
2
4
6
8
Control Treatment
CINTERVENTION
Mean
PERIOD
Baseline
Midline
Reading fluency
82
0
20
40
60
Control Treatment
CINTERVENTION
Reading fluency
PERIOD
Baseline
Midline
Reading fluency
As shown in the table above, for the the Reading fluency EGRA subtask, the mean for the Control (FFE
only (Portuguese)) condition at baseline was 0 (SD = 0) and the mean for the Treatment (FFE + lit
(Portuguese)) condition at baseline was 0.05288462 (SD = 1.615043). The difference between Control (FFE
only (Portuguese)) and Treatment (FFE + lit (Portuguese)) at baseline was thus of 0.05288462 points. The
p-value for this difference was 0.2920071. The mean for the Control (FFE only (Portuguese)) condition at
midline was 4.915878 (SD = 12.43346) and the mean for the Treatment (FFE + lit (Portuguese)) condition
at midline was 8.223292 (SD = 15.26094). The difference between Control (FFE only (Portuguese)) and
Treatment (FFE + lit (Portuguese)) at midline was thus of 3.307414 points. The p-value for this difference
was 0.005798436. The change from the baseline to the midline of 4.915878 points for the Control (FFE
only (Portuguese)) condition establishes the conterfactual against which the change for the Treatment (FFE
+ lit (Portuguese)) can be assessed against. The change for the Treatment (FFE + lit (Portuguese))
from the baseline to the midline was of 8.170407 points. Consequently, the change for the Control (FFE
only (Portuguese)) relative to the change for the Treatment (FFE + lit (Portuguese)) condition was of
3.254529 points. The p-value for this difference was 0.006581735. Accordingly, the difference was statistically
significant (p < 0.01 corrected for 4 additional contrasts) and thus there is evidence of a different degree of
change from baseline to midline in the Treatment (FFE + lit (Portuguese)) when compared to the Control
(FFE only (Portuguese)) condition. This provides evidence that the Treatment (FFE + lit (Portuguese))
(or some other unobserved process) impacted on the Reading fluency EGRA subtask and thus the rate of
change of the two conditions relative to their baselines differed.
83
1.8.5 Contrast 5: FFE + lit (Bilingual) vs FFE + lit (Portuguese)
Table 40: Reading fluency
Baseline Midline
Intervention Mean SD n min max Mean SD n min max
Control 0.053 1.615 1040 0 52 8.223 15.261 1039 0 70
Treatment 0.043 0.826 538 0 19 6.751 11.587 752 0 53
0
2
4
6
8
Control Treatment
CINTERVENTION
Mean
PERIOD
Baseline
Midline
Reading fluency
84
0
20
40
60
Control Treatment
CINTERVENTION
Reading fluency
PERIOD
Baseline
Midline
Reading fluency
As shown in the table above, for the the Reading fluency EGRA subtask, the mean for the Control (FFE
+ lit (Portuguese)) condition at baseline was 0.05288462 (SD = 1.615043) and the mean for the Treatment
(FFE + lit (Bilingual)) condition at baseline was 0.04275093 (SD = 0.8255879). The difference between
Control (FFE + lit (Portuguese)) and Treatment (FFE + lit (Bilingual)) at baseline was thus of 0.01013369
points. The p-value for this difference was 0.8687699. The mean for the Control (FFE + lit (Portuguese))
condition at midline was 8.223292 (SD = 15.26094) and the mean for the Treatment (FFE + lit (Bilingual))
condition at midline was 6.75133 (SD = 11.58736). The difference between Control (FFE + lit (Portuguese))
and Treatment (FFE + lit (Bilingual)) at midline was thus of 1.471962 points. The p-value for this difference
was 0.2100281. The change from the baseline to the midline of 8.170407 points for the Control (FFE + lit
(Portuguese)) condition establishes the conterfactual against which the change for the Treatment (FFE + lit
(Bilingual)) can be assessed against. The change for the Treatment (FFE + lit (Bilingual)) from the baseline
to the midline was of 6.708579 points. Consequently, the change for the Control (FFE + lit (Portuguese))
relative to the change for the Treatment (FFE + lit (Bilingual)) condition was of -1.461828 points. The
p-value for this difference was 0.2135764. Accordingly, the difference was not statistically significant (p <
0.01 corrected for 5 additional contrasts) and therefore the statistical test provides no conclusive evidence
in regards to there being a difference in the rate of change between the baseline and the midline for the
Reading fluency EGRA subtask across the Control (FFE + lit (Portuguese)) and the Treatment (FFE + lit
(Bilingual)) conditions beyond that expected from sampling error.
85
1.9 EGRA_ST8_1AS: Comparative reading fluency
1.9.1 Contrast 1: FFE + lit (all) vs Comparison (all)
Table 41: Comparative reading fluency
Baseline Midline
Intervention Mean SD n min max Mean SD n min max
Control 0.001 0.024 1136 0 0.743 0.117 0.237 1080 0 1
Treatment 0.001 0.021 1578 0 0.743 0.122 0.218 1791 0 1
0.000
0.025
0.050
0.075
0.100
0.125
Control Treatment
CINTERVENTION
Mean
PERIOD
Baseline
Midline
Comparative reading fluency
86
0.00
0.25
0.50
0.75
1.00
Control Treatment
CINTERVENTION
Comparative reading fluency
PERIOD
Baseline
Midline
Comparative reading fluency
As shown in the table above, for the the Comparative reading fluency EGRA subtask, the mean for the
Control (Comparison (all)) condition at baseline was 0.00110664 (SD = 0.02392406) and the mean for the
Treatment (FFE + lit (all)) condition at baseline was 0.0007729256 (SD = 0.0208169). The difference
between Control (Comparison (all)) and Treatment (FFE + lit (all)) at baseline was thus of 0.0003337143
points. The p-value for this difference was 0.7049342. The mean for the Control (Comparison (all)) condition
at midline was 0.1167655 (SD = 0.2374994) and the mean for the Treatment (FFE + lit (all)) condition at
midline was 0.1216281 (SD = 0.2182686). The difference between Control (Comparison (all)) and Treatment
(FFE + lit (all)) at midline was thus of 0.004862558 points. The p-value for this difference was 0.7742896.
The change from the baseline to the midline of 0.1156589 points for the Control (Comparison (all)) condition
establishes the conterfactual against which the change for the Treatment (FFE + lit (all)) can be assessed
against. The change for the Treatment (FFE + lit (all)) from the baseline to the midline was of 0.1208551
points. Consequently, the change for the Control (Comparison (all)) relative to the change for the Treat￾ment (FFE + lit (all)) condition was of 0.005196272 points. The p-value for this difference was 0.758846.
Accordingly, the difference was not statistically significant (p < 0.01 corrected for 5 additional contrasts)
and therefore the statistical test provides no conclusive evidence in regards to there being a difference in the
rate of change between the baseline and the midline for the Comparative reading fluency EGRA subtask
across the Control (Comparison (all)) and the Treatment (FFE + lit (all)) conditions beyond that expected
from sampling error.
87
1.9.2 Contrast 2: FFE + lit (Portuguese) vs Comparison (Portuguese)
Table 42: Comparative reading fluency
Baseline Midline
Intervention Mean SD n min max Mean SD n min max
Control 0.001 0.025 1047 0 0.743 0.125 0.244 972 0 1
Treatment 0.001 0.023 1040 0 0.743 0.117 0.218 1039 0 1
0.00
0.04
0.08
0.12
Control Treatment
CINTERVENTION
Mean
PERIOD
Baseline
Midline
Comparative reading fluency
88
0.00
0.25
0.50
0.75
1.00
Control Treatment
CINTERVENTION
Comparative reading fluency
PERIOD
Baseline
Midline
Comparative reading fluency
As shown in the table above, for the the Comparative reading fluency EGRA subtask, the mean for the
Control (Comparison (Portuguese)) condition at baseline was 0.00120071 (SD = 0.02491882) and the mean
for the Treatment (FFE + lit (Portuguese)) condition at baseline was 0.0007554945 (SD = 0.02307205).
The difference between Control (Comparison (Portuguese)) and Treatment (FFE + lit (Portuguese)) at
baseline was thus of 0.000445215 points. The p-value for this difference was 0.6718402. The mean for the
Control (Comparison (Portuguese)) condition at midline was 0.1248089 (SD = 0.2435319) and the mean
for the Treatment (FFE + lit (Portuguese)) condition at midline was 0.1174756 (SD = 0.2180134). The
difference between Control (Comparison (Portuguese)) and Treatment (FFE + lit (Portuguese)) at midline
was thus of 0.007333341 points. The p-value for this difference was 0.7056136. The change from the baseline
to the midline of 0.1236082 points for the Control (Comparison (Portuguese)) condition establishes the
conterfactual against which the change for the Treatment (FFE + lit (Portuguese)) can be assessed against.
The change for the Treatment (FFE + lit (Portuguese)) from the baseline to the midline was of 0.1167201
points. Consequently, the change for the Control (Comparison (Portuguese)) relative to the change for the
Treatment (FFE + lit (Portuguese)) condition was of -0.006888126 points. The p-value for this difference was
0.722265. Accordingly, the difference was not statistically significant (p < 0.01 corrected for 5 additional
contrasts) and therefore the statistical test provides no conclusive evidence in regards to there being a
difference in the rate of change between the baseline and the midline for the Comparative reading fluency
EGRA subtask across the Control (Comparison (Portuguese)) and the Treatment (FFE + lit (Portuguese))
conditions beyond that expected from sampling error.
89
1.9.3 Contrast 3: FFE + lit (Bilingual) vs Comparison (Bilingual)
Table 43: Comparative reading fluency
Baseline Midline
Intervention Mean SD n min max Mean SD n min max
Control 0.000 0.000 89 0 0.000 0.044 0.157 108 0 0.962
Treatment 0.001 0.016 538 0 0.358 0.127 0.219 752 0 1.000
0.00
0.05
0.10
Control Treatment
CINTERVENTION
Mean
PERIOD
Baseline
Midline
Comparative reading fluency
90
0.00
0.25
0.50
0.75
1.00
Control Treatment
CINTERVENTION
Comparative reading fluency
PERIOD
Baseline
Midline
Comparative reading fluency
As shown in the table above, for the the Comparative reading fluency EGRA subtask, the mean for the Con￾trol (Comparison (Bilingual)) condition at baseline was 0 (SD = 0) and the mean for the Treatment (FFE
+ lit (Bilingual)) condition at baseline was 0.0008066213 (SD = 0.01557713). The difference between Con￾trol (Comparison (Bilingual)) and Treatment (FFE + lit (Bilingual)) at baseline was thus of 0.0008066213
points. The p-value for this difference was 0.2270531. The mean for the Control (Comparison (Bilingual))
condition at midline was 0.04437456 (SD = 0.1572335) and the mean for the Treatment (FFE + lit (Bilin￾gual)) condition at midline was 0.1273653 (SD = 0.218636). The difference between Control (Comparison
(Bilingual)) and Treatment (FFE + lit (Bilingual)) at midline was thus of 0.08299074 points. The p-value
for this difference was 0.01062726. The change from the baseline to the midline of 0.04437456 points for
the Control (Comparison (Bilingual)) condition establishes the conterfactual against which the change for
the Treatment (FFE + lit (Bilingual)) can be assessed against. The change for the Treatment (FFE + lit
(Bilingual)) from the baseline to the midline was of 0.1265587 points. Consequently, the change for the
Control (Comparison (Bilingual)) relative to the change for the Treatment (FFE + lit (Bilingual)) condition
was of 0.08218412 points. The p-value for this difference was 0.01147754. Accordingly, the difference was
not statistically significant (p < 0.01 corrected for 5 additional contrasts) and therefore the statistical test
provides no conclusive evidence in regards to there being a difference in the rate of change between the
baseline and the midline for the Comparative reading fluency EGRA subtask across the Control (Compari￾son (Bilingual)) and the Treatment (FFE + lit (Bilingual)) conditions beyond that expected from sampling
error.
91
1.9.4 Contrast 4: FFE + lit (Portuguese) vs FFE only (Portuguese)
Table 44: Comparative reading fluency
Baseline Midline
Intervention Mean SD n min max Mean SD n min max
Control 0.000 0.000 1007 0 0.000 0.071 0.181 951 0 1.321
Treatment 0.001 0.023 1040 0 0.743 0.117 0.218 1039 0 1.000
0.00
0.03
0.06
0.09
0.12
Control Treatment
CINTERVENTION
Mean
PERIOD
Baseline
Midline
Comparative reading fluency
92
0.0
0.5
1.0
Control Treatment
CINTERVENTION
Comparative reading fluency
PERIOD
Baseline
Midline
Comparative reading fluency
As shown in the table above, for the the Comparative reading fluency EGRA subtask, the mean for the
Control (FFE only (Portuguese)) condition at baseline was 0 (SD = 0) and the mean for the Treatment
(FFE + lit (Portuguese)) condition at baseline was 0.0007554945 (SD = 0.02307205). The difference be￾tween Control (FFE only (Portuguese)) and Treatment (FFE + lit (Portuguese)) at baseline was thus of
0.0007554945 points. The p-value for this difference was 0.2920071. The mean for the Control (FFE only
(Portuguese)) condition at midline was 0.07120013 (SD = 0.1807224) and the mean for the Treatment (FFE
+ lit (Portuguese)) condition at midline was 0.1174756 (SD = 0.2180134). The difference between Control
(FFE only (Portuguese)) and Treatment (FFE + lit (Portuguese)) at midline was thus of 0.04627547 points.
The p-value for this difference was 0.00734767. The change from the baseline to the midline of 0.07120013
points for the Control (FFE only (Portuguese)) condition establishes the conterfactual against which the
change for the Treatment (FFE + lit (Portuguese)) can be assessed against. The change for the Treatment
(FFE + lit (Portuguese)) from the baseline to the midline was of 0.1167201 points. Consequently, the change
for the Control (FFE only (Portuguese)) relative to the change for the Treatment (FFE + lit (Portuguese))
condition was of 0.04551997 points. The p-value for this difference was 0.008307649. Accordingly, the dif￾ference was statistically significant (p < 0.01 corrected for 4 additional contrasts) and thus there is evidence
of a different degree of change from baseline to midline in the Treatment (FFE + lit (Portuguese)) when
compared to the Control (FFE only (Portuguese)) condition. This provides evidence that the Treatment
(FFE + lit (Portuguese)) (or some other unobserved process) impacted on the Comparative reading fluency
EGRA subtask and thus the rate of change of the two conditions relative to their baselines differed.
93
1.9.5 Contrast 5: FFE + lit (Bilingual) vs FFE + lit (Portuguese)
Table 45: Comparative reading fluency
Baseline Midline
Intervention Mean SD n min max Mean SD n min max
Control 0.001 0.023 1040 0 0.743 0.117 0.218 1039 0 1
Treatment 0.001 0.016 538 0 0.358 0.127 0.219 752 0 1
0.00
0.05
0.10
Control Treatment
CINTERVENTION
Mean
PERIOD
Baseline
Midline
Comparative reading fluency
94
0.00
0.25
0.50
0.75
1.00
Control Treatment
CINTERVENTION
Comparative reading fluency
PERIOD
Baseline
Midline
Comparative reading fluency
As shown in the table above, for the the Comparative reading fluency EGRA subtask, the mean for the
Control (FFE + lit (Portuguese)) condition at baseline was 0.0007554945 (SD = 0.02307205) and the mean
for the Treatment (FFE + lit (Bilingual)) condition at baseline was 0.0008066213 (SD = 0.01557713). The
difference between Control (FFE + lit (Portuguese)) and Treatment (FFE + lit (Bilingual)) at baseline was
thus of 5.11268e-05 points. The p-value for this difference was 0.9582923. The mean for the Control (FFE +
lit (Portuguese)) condition at midline was 0.1174756 (SD = 0.2180134) and the mean for the Treatment (FFE
+ lit (Bilingual)) condition at midline was 0.1273653 (SD = 0.218636). The difference between Control (FFE
+ lit (Portuguese)) and Treatment (FFE + lit (Bilingual)) at midline was thus of 0.009889706 points. The
p-value for this difference was 0.6060341. The change from the baseline to the midline of 0.1167201 points
for the Control (FFE + lit (Portuguese)) condition establishes the conterfactual against which the change
for the Treatment (FFE + lit (Bilingual)) can be assessed against. The change for the Treatment (FFE +
lit (Bilingual)) from the baseline to the midline was of 0.1265587 points. Consequently, the change for the
Control (FFE + lit (Portuguese)) relative to the change for the Treatment (FFE + lit (Bilingual)) condition
was of 0.009838579 points. The p-value for this difference was 0.6083137. Accordingly, the difference was
not statistically significant (p < 0.01 corrected for 5 additional contrasts) and therefore the statistical test
provides no conclusive evidence in regards to there being a difference in the rate of change between the
baseline and the midline for the Comparative reading fluency EGRA subtask across the Control (FFE + lit
(Portuguese)) and the Treatment (FFE + lit (Bilingual)) conditions beyond that expected from sampling
error.
95
1.10 EGRA_ST8_1B: Overall reading ability
1.10.1 Contrast 1: FFE + lit (all) vs Comparison (all)
Table 46: Overall reading ability
Baseline Midline
Intervention Mean SD n min max Mean SD n min max
Control NA NA 0 NA NA 12.509 22.893 1078 0 70
Treatment NA NA 0 NA NA 12.204 20.815 1759 0 70
0
4
8
12
Control Treatment
CINTERVENTION
Mean
PERIOD
Midline
Overall reading ability
96
0
20
40
60
Control Treatment
CINTERVENTION
Overall reading ability
PERIOD
Midline
Overall reading ability
As shown in the table above, for the Overall reading ability EGRA subtask, the mean for the Control
(Comparison (all)) condition at midline was 12.50928 (SD = 22.89327) and the mean for the Treatment
(FFE + lit (all)) condition at midline was 12.20409 (SD = 20.81483). The difference for the Overall reading
ability across the Control (Comparison (all)) and Treatment (FFE + lit (all)) conditions at midline was
thus -0.3051832 points (there was no baseline measurement for this variable). The p-value for this difference
was 0.8609184. Accordingly, the difference was not statistically significant and therefore, the statistical test
provides no conclusive evidence in regards to the there being a difference in the levels of Overall reading
abilitybetween the Control (Comparison (all)) and Treatment (FFE + lit (all)) conditions at midline beyond
that which could result from sampling error.
1.10.2 Contrast 2: FFE + lit (Portuguese) vs Comparison (Portuguese)
Table 47: Overall reading ability
Baseline Midline
Intervention Mean SD n min max Mean SD n min max
Control NA NA 0 NA NA 13.466 23.617 970 0 70
Treatment NA NA 0 NA NA 12.455 22.366 1017 0 70
97
0
5
10
Control Treatment
CINTERVENTION
Mean
PERIOD
Midline
Overall reading ability
98
0
20
40
60
Control Treatment
CINTERVENTION
Overall reading ability
PERIOD
Midline
Overall reading ability
As shown in the table above, for the Overall reading ability EGRA subtask, the mean for the Control (Com￾parison (Portuguese)) condition at midline was 13.46598 (SD = 23.6169) and the mean for the Treatment
(FFE + lit (Portuguese)) condition at midline was 12.45526 (SD = 22.36627). The difference for the Over￾all reading ability across the Control (Comparison (Portuguese)) and Treatment (FFE + lit (Portuguese))
conditions at midline was thus -1.010719 points (there was no baseline measurement for this variable). The
p-value for this difference was 0.6293656. Accordingly, the difference was not statistically significant and
therefore, the statistical test provides no conclusive evidence in regards to the there being a difference in the
levels of Overall reading abilitybetween the Control (Comparison (Portuguese)) and Treatment (FFE + lit
(Portuguese)) conditions at midline beyond that which could result from sampling error.
1.10.3 Contrast 3: FFE + lit (Bilingual) vs Comparison (Bilingual)
Table 48: Overall reading ability
Baseline Midline
Intervention Mean SD n min max Mean SD n min max
Control NA NA 0 NA NA 3.917 11.891 108 0 53
Treatment NA NA 0 NA NA 11.860 18.487 742 0 53
99
0.0
2.5
5.0
7.5
10.0
Control Treatment
CINTERVENTION
Mean
PERIOD
Midline
Overall reading ability
100
0
20
40
Control Treatment
CINTERVENTION
Overall reading ability
PERIOD
Midline
Overall reading ability
As shown in the table above, for the Overall reading ability EGRA subtask, the mean for the Control
(Comparison (Bilingual)) condition at midline was 3.916667 (SD = 11.89056) and the mean for the Treatment
(FFE + lit (Bilingual)) condition at midline was 11.85984 (SD = 18.48731). The difference for the Overall
reading ability across the Control (Comparison (Bilingual)) and Treatment (FFE + lit (Bilingual)) conditions
at midline was thus 7.943172 points (there was no baseline measurement for this variable). The p-value for
this difference was 0.001604491. Accordingly, the difference was statistically significant (p < 0.01 corrected
for 4 additional contrasts). Nevertheless, due to the absence of baseline values and lack of random allocation
to the treatment and control conditions, a causal attribution of this effect to the treatment is not possible
1.10.4 Contrast 4: FFE + lit (Portuguese) vs FFE only (Portuguese)
Table 49: Overall reading ability
Baseline Midline
Intervention Mean SD n min max Mean SD n min max
Control NA NA 0 NA NA 9.120 20.498 950 0 70
Treatment NA NA 0 NA NA 12.455 22.366 1017 0 70
101
0
4
8
12
Control Treatment
CINTERVENTION
Mean
PERIOD
Midline
Overall reading ability
102
0
20
40
60
Control Treatment
CINTERVENTION
Overall reading ability
PERIOD
Midline
Overall reading ability
As shown in the table above, for the Overall reading ability EGRA subtask, the mean for the Control (FFE
only (Portuguese)) condition at midline was 9.12 (SD = 20.49843) and the mean for the Treatment (FFE +
lit (Portuguese)) condition at midline was 12.45526 (SD = 22.36627). The difference for the Overall reading
ability across the Control (FFE only (Portuguese)) and Treatment (FFE + lit (Portuguese)) conditions at
midline was thus 3.335261 points (there was no baseline measurement for this variable). The p-value for
this difference was 0.1007097. Accordingly, the difference was not statistically significant and therefore,
the statistical test provides no conclusive evidence in regards to the there being a difference in the levels of
Overall reading abilitybetween the Control (FFE only (Portuguese)) and Treatment (FFE + lit (Portuguese))
conditions at midline beyond that which could result from sampling error.
1.10.5 Contrast 5: FFE + lit (Bilingual) vs FFE + lit (Portuguese)
Table 50: Overall reading ability
Baseline Midline
Intervention Mean SD n min max Mean SD n min max
Control NA NA 0 NA NA 12.455 22.366 1017 0 70
Treatment NA NA 0 NA NA 11.860 18.487 742 0 53
103
0
4
8
12
Control Treatment
CINTERVENTION
Mean
PERIOD
Midline
Overall reading ability
104
0
20
40
60
Control Treatment
CINTERVENTION
Overall reading ability
PERIOD
Midline
Overall reading ability
As shown in the table above, for the Overall reading ability EGRA subtask, the mean for the Control (FFE
+ lit (Portuguese)) condition at midline was 12.45526 (SD = 22.36627) and the mean for the Treatment
(FFE + lit (Bilingual)) condition at midline was 11.85984 (SD = 18.48731). The difference for the Overall
reading ability across the Control (FFE + lit (Portuguese)) and Treatment (FFE + lit (Bilingual)) conditions
at midline was thus -0.5954223 points (there was no baseline measurement for this variable). The p-value
for this difference was 0.7526363. Accordingly, the difference was not statistically significant and therefore,
the statistical test provides no conclusive evidence in regards to the there being a difference in the levels of
Overall reading abilitybetween the Control (FFE + lit (Portuguese)) and Treatment (FFE + lit (Bilingual))
conditions at midline beyond that which could result from sampling error.
1.11 EGRA_ST8_1BS: Comparative overall reading ability
1.11.1 Contrast 1: FFE + lit (all) vs Comparison (all)
Table 51: Comparative overall reading ability
Baseline Midline
Intervention Mean SD n min max Mean SD n min max
Control NA NA 0 NA NA 0.181 0.330 1078 0 1
Treatment NA NA 0 NA NA 0.197 0.333 1759 0 1
105
0.00
0.05
0.10
0.15
0.20
Control Treatment
CINTERVENTION
Mean
PERIOD
Midline
Comparative overall reading ability
106
0.00
0.25
0.50
0.75
1.00
Control Treatment
CINTERVENTION
Comparative overall reading ability
PERIOD
Midline
Comparative overall reading ability
As shown in the table above, for the Comparative overall reading ability EGRA subtask, the mean for
the Control (Comparison (all)) condition at midline was 0.180502 (SD = 0.3296675) and the mean for the
Treatment (FFE + lit (all)) condition at midline was 0.1972422 (SD = 0.3328691). The difference for the
Comparative overall reading ability across the Control (Comparison (all)) and Treatment (FFE + lit (all))
conditions at midline was thus 0.01674024 points (there was no baseline measurement for this variable). The
p-value for this difference was 0.5162973. Accordingly, the difference was not statistically significant and
therefore, the statistical test provides no conclusive evidence in regards to the there being a difference in the
levels of Comparative overall reading abilitybetween the Control (Comparison (all)) and Treatment (FFE +
lit (all)) conditions at midline beyond that which could result from sampling error.
1.11.2 Contrast 2: FFE + lit (Portuguese) vs Comparison (Portuguese)
Table 52: Comparative overall reading ability
Baseline Midline
Intervention Mean SD n min max Mean SD n min max
Control NA NA 0 NA NA 0.192 0.337 970 0 1
Treatment NA NA 0 NA NA 0.178 0.320 1017 0 1
107
0.00
0.05
0.10
0.15
0.20
Control Treatment
CINTERVENTION
Mean
PERIOD
Midline
Comparative overall reading ability
108
0.00
0.25
0.50
0.75
1.00
Control Treatment
CINTERVENTION
Comparative overall reading ability
PERIOD
Midline
Comparative overall reading ability
As shown in the table above, for the Comparative overall reading ability EGRA subtask, the mean for
the Control (Comparison (Portuguese)) condition at midline was 0.1923711 (SD = 0.3373843) and the
mean for the Treatment (FFE + lit (Portuguese)) condition at midline was 0.1779323 (SD = 0.3195182).
The difference for the Comparative overall reading ability across the Control (Comparison (Portuguese))
and Treatment (FFE + lit (Portuguese)) conditions at midline was thus -0.01443884 points (there was no
baseline measurement for this variable). The p-value for this difference was 0.6293656. Accordingly, the
difference was not statistically significant and therefore, the statistical test provides no conclusive evidence
in regards to the there being a difference in the levels of Comparative overall reading abilitybetween the
Control (Comparison (Portuguese)) and Treatment (FFE + lit (Portuguese)) conditions at midline beyond
that which could result from sampling error.
1.11.3 Contrast 3: FFE + lit (Bilingual) vs Comparison (Bilingual)
Table 53: Comparative overall reading ability
Baseline Midline
Intervention Mean SD n min max Mean SD n min max
Control NA NA 0 NA NA 0.074 0.224 108 0 1
Treatment NA NA 0 NA NA 0.224 0.349 742 0 1
109
0.00
0.05
0.10
0.15
0.20
Control Treatment
CINTERVENTION
Mean
PERIOD
Midline
Comparative overall reading ability
110
0.00
0.25
0.50
0.75
1.00
Control Treatment
CINTERVENTION
Comparative overall reading ability
PERIOD
Midline
Comparative overall reading ability
As shown in the table above, for the Comparative overall reading ability EGRA subtask, the mean for the
Control (Comparison (Bilingual)) condition at midline was 0.07389937 (SD = 0.2243503) and the mean for
the Treatment (FFE + lit (Bilingual)) condition at midline was 0.2237088 (SD = 0.3488274). The difference
for the Comparative overall reading ability across the Control (Comparison (Bilingual)) and Treatment (FFE
+ lit (Bilingual)) conditions at midline was thus 0.1498094 points (there was no baseline measurement for
this variable). The p-value for this difference was 0.001613896. Accordingly, the difference was statistically
significant (p < 0.01 corrected for 4 additional contrasts). Nevertheless, due to the absence of baseline values
and lack of random allocation to the treatment and control conditions, a causal attribution of this effect to
the treatment is not possible
1.11.4 Contrast 4: FFE + lit (Portuguese) vs FFE only (Portuguese)
Table 54: Comparative overall reading ability
Baseline Midline
Intervention Mean SD n min max Mean SD n min max
Control NA NA 0 NA NA 0.132 0.297 950 0 1.321
Treatment NA NA 0 NA NA 0.178 0.320 1017 0 1.000
111
0.00
0.05
0.10
0.15
Control Treatment
CINTERVENTION
Mean
PERIOD
Midline
Comparative overall reading ability
112
0.0
0.5
1.0
Control Treatment
CINTERVENTION
Comparative overall reading ability
PERIOD
Midline
Comparative overall reading ability
As shown in the table above, for the Comparative overall reading ability EGRA subtask, the mean for the
Control (FFE only (Portuguese)) condition at midline was 0.1319594 (SD = 0.2969652) and the mean for the
Treatment (FFE + lit (Portuguese)) condition at midline was 0.1779323 (SD = 0.3195182). The difference
for the Comparative overall reading ability across the Control (FFE only (Portuguese)) and Treatment (FFE
+ lit (Portuguese)) conditions at midline was thus 0.04597287 points (there was no baseline measurement for
this variable). The p-value for this difference was 0.1161117. Accordingly, the difference was not statistically
significant and therefore, the statistical test provides no conclusive evidence in regards to the there being a
difference in the levels of Comparative overall reading abilitybetween the Control (FFE only (Portuguese))
and Treatment (FFE + lit (Portuguese)) conditions at midline beyond that which could result from sampling
error.
1.11.5 Contrast 5: FFE + lit (Bilingual) vs FFE + lit (Portuguese)
Table 55: Comparative overall reading ability
Baseline Midline
Intervention Mean SD n min max Mean SD n min max
Control NA NA 0 NA NA 0.178 0.320 1017 0 1
Treatment NA NA 0 NA NA 0.224 0.349 742 0 1
113
0.00
0.05
0.10
0.15
0.20
Control Treatment
CINTERVENTION
Mean
PERIOD
Midline
Comparative overall reading ability
114
0.00
0.25
0.50
0.75
1.00
Control Treatment
CINTERVENTION
Comparative overall reading ability
PERIOD
Midline
Comparative overall reading ability
As shown in the table above, for the Comparative overall reading ability EGRA subtask, the mean for the
Control (FFE + lit (Portuguese)) condition at midline was 0.1779323 (SD = 0.3195182) and the mean for the
Treatment (FFE + lit (Bilingual)) condition at midline was 0.2237088 (SD = 0.3488274). The difference for
the Comparative overall reading ability across the Control (FFE + lit (Portuguese)) and Treatment (FFE
+ lit (Bilingual)) conditions at midline was thus 0.04577648 points (there was no baseline measurement for
this variable). The p-value for this difference was 0.1376108. Accordingly, the difference was not statistically
significant and therefore, the statistical test provides no conclusive evidence in regards to the there being a
difference in the levels of Comparative overall reading abilitybetween the Control (FFE + lit (Portuguese))
and Treatment (FFE + lit (Bilingual)) conditions at midline beyond that which could result from sampling
error.
1.12 EGRA_ST8_2: Comprehension of reading
1.12.1 Contrast 1: FFE + lit (all) vs Comparison (all)
Table 56: Comprehension of reading
Baseline Midline
Intervention Mean SD n min max Mean SD n min max
Control 0.031 0.275 1136 0 4 0.546 1.180 1081 0 4
Treatment 0.016 0.201 1578 0 4 0.535 1.143 1808 0 4
115
0.0
0.2
0.4
Control Treatment
CINTERVENTION
Mean
PERIOD
Baseline
Midline
Comprehension of reading
116
0
1
2
3
4
Control Treatment
CINTERVENTION
Comprehension of reading
PERIOD
Baseline
Midline
Comprehension of reading
As shown in the table above, for the the Comprehension of reading EGRA subtask, the mean for the Control
(Comparison (all)) condition at baseline was 0.03080986 (SD = 0.2751398) and the mean for the Treatment
(FFE + lit (all)) condition at baseline was 0.01647655 (SD = 0.2007778). The difference between Control
(Comparison (all)) and Treatment (FFE + lit (all)) at baseline was thus of 0.01433331 points. The p-value
for this difference was 0.3039468. The mean for the Control (Comparison (all)) condition at midline was
0.5457909 (SD = 1.180075) and the mean for the Treatment (FFE + lit (all)) condition at midline was
0.5348451 (SD = 1.14277). The difference between Control (Comparison (all)) and Treatment (FFE + lit
(all)) at midline was thus of 0.0109458 points. The p-value for this difference was 0.8947624. The change
from the baseline to the midline of 0.5149811 points for the Control (Comparison (all)) condition establishes
the conterfactual against which the change for the Treatment (FFE + lit (all)) can be assessed against.
The change for the Treatment (FFE + lit (all)) from the baseline to the midline was of 0.5183686 points.
Consequently, the change for the Control (Comparison (all)) relative to the change for the Treatment (FFE
+ lit (all)) condition was of 0.003387505 points. The p-value for this difference was 0.967264. Accordingly,
the difference was not statistically significant (p < 0.01 corrected for 5 additional contrasts) and therefore
the statistical test provides no conclusive evidence in regards to there being a difference in the rate of change
between the baseline and the midline for the Comprehension of reading EGRA subtask across the Control
(Comparison (all)) and the Treatment (FFE + lit (all)) conditions beyond that expected from sampling
error.
117
1.12.2 Contrast 2: FFE + lit (Portuguese) vs Comparison (Portuguese)
Table 57: Comprehension of reading
Baseline Midline
Intervention Mean SD n min max Mean SD n min max
Control 0.032 0.285 1047 0 4 0.595 1.223 972 0 4
Treatment 0.011 0.178 1040 0 4 0.571 1.201 1047 0 4
0.0
0.2
0.4
0.6
Control Treatment
CINTERVENTION
Mean
PERIOD
Baseline
Midline
Comprehension of reading
118
0
1
2
3
4
Control Treatment
CINTERVENTION
Comprehension of reading
PERIOD
Baseline
Midline
Comprehension of reading
As shown in the table above, for the the Comprehension of reading EGRA subtask, the mean for the
Control (Comparison (Portuguese)) condition at baseline was 0.03247373 (SD = 0.2848902) and the mean
for the Treatment (FFE + lit (Portuguese)) condition at baseline was 0.01057692 (SD = 0.1779026). The
difference between Control (Comparison (Portuguese)) and Treatment (FFE + lit (Portuguese)) at baseline
was thus of 0.02189681 points. The p-value for this difference was 0.1397576. The mean for the Control
(Comparison (Portuguese)) condition at midline was 0.5946502 (SD = 1.222974) and the mean for the
Treatment (FFE + lit (Portuguese)) condition at midline was 0.5711557 (SD = 1.20088). The difference
between Control (Comparison (Portuguese)) and Treatment (FFE + lit (Portuguese)) at midline was thus
of 0.02349452 points. The p-value for this difference was 0.8094379. The change from the baseline to the
midline of 0.5621765 points for the Control (Comparison (Portuguese)) condition establishes the conterfactual
against which the change for the Treatment (FFE + lit (Portuguese)) can be assessed against. The change
for the Treatment (FFE + lit (Portuguese)) from the baseline to the midline was of 0.5605788 points.
Consequently, the change for the Control (Comparison (Portuguese)) relative to the change for the Treatment
(FFE + lit (Portuguese)) condition was of -0.001597711 points. The p-value for this difference was 0.9868452.
Accordingly, the difference was not statistically significant (p < 0.01 corrected for 5 additional contrasts)
and therefore the statistical test provides no conclusive evidence in regards to there being a difference in the
rate of change between the baseline and the midline for the Comprehension of reading EGRA subtask across
the Control (Comparison (Portuguese)) and the Treatment (FFE + lit (Portuguese)) conditions beyond that
expected from sampling error.
119
1.12.3 Contrast 3: FFE + lit (Bilingual) vs Comparison (Bilingual)
Table 58: Comprehension of reading
Baseline Midline
Intervention Mean SD n min max Mean SD n min max
Control 0.011 0.106 89 0 1 0.110 0.515 109 0 3
Treatment 0.028 0.239 538 0 3 0.485 1.056 761 0 4
0.0
0.1
0.2
0.3
0.4
0.5
Control Treatment
CINTERVENTION
Mean
PERIOD
Baseline
Midline
Comprehension of reading
120
0
1
2
3
4
Control Treatment
CINTERVENTION
Comprehension of reading
PERIOD
Baseline
Midline
Comprehension of reading
As shown in the table above, for the the Comprehension of reading EGRA subtask, the mean for the Control
(Comparison (Bilingual)) condition at baseline was 0.01123596 (SD = 0.1059998) and the mean for the
Treatment (FFE + lit (Bilingual)) condition at baseline was 0.02788104 (SD = 0.2386406). The difference
between Control (Comparison (Bilingual)) and Treatment (FFE + lit (Bilingual)) at baseline was thus of
0.01664509 points. The p-value for this difference was 0.3408817. The mean for the Control (Comparison
(Bilingual)) condition at midline was 0.1100917 (SD = 0.5153109) and the mean for the Treatment (FFE
+ lit (Bilingual)) condition at midline was 0.4848883 (SD = 1.056357). The difference between Control
(Comparison (Bilingual)) and Treatment (FFE + lit (Bilingual)) at midline was thus of 0.3747966 points.
The p-value for this difference was 0.000555817. The change from the baseline to the midline of 0.09885579
points for the Control (Comparison (Bilingual)) condition establishes the conterfactual against which the
change for the Treatment (FFE + lit (Bilingual)) can be assessed against. The change for the Treatment (FFE
+ lit (Bilingual)) from the baseline to the midline was of 0.4570073 points. Consequently, the change for the
Control (Comparison (Bilingual)) relative to the change for the Treatment (FFE + lit (Bilingual)) condition
was of 0.3581515 points. The p-value for this difference was 0.001274507. Accordingly, the difference was
statistically significant (p < 0.01 corrected for 4 additional contrasts) and thus there is evidence of a different
degree of change from baseline to midline in the Treatment (FFE + lit (Bilingual)) when compared to
the Control (Comparison (Bilingual)) condition. This provides evidence that the Treatment (FFE + lit
(Bilingual)) (or some other unobserved process) impacted on the Comprehension of reading EGRA subtask
and thus the rate of change of the two conditions relative to their baselines differed.
121
1.12.4 Contrast 4: FFE + lit (Portuguese) vs FFE only (Portuguese)
Table 59: Comprehension of reading
Baseline Midline
Intervention Mean SD n min max Mean SD n min max
Control 0.000 0.000 1007 0 0 0.280 0.879 952 0 4
Treatment 0.011 0.178 1040 0 4 0.571 1.201 1047 0 4
0.0
0.2
0.4
Control Treatment
CINTERVENTION
Mean
PERIOD
Baseline
Midline
Comprehension of reading
122
0
1
2
3
4
Control Treatment
CINTERVENTION
Comprehension of reading
PERIOD
Baseline
Midline
Comprehension of reading
As shown in the table above, for the the Comprehension of reading EGRA subtask, the mean for the Control
(FFE only (Portuguese)) condition at baseline was 0 (SD = 0) and the mean for the Treatment (FFE + lit
(Portuguese)) condition at baseline was 0.01057692 (SD = 0.1779026). The difference between Control (FFE
only (Portuguese)) and Treatment (FFE + lit (Portuguese)) at baseline was thus of 0.01057692 points. The
p-value for this difference was 0.08336649. The mean for the Control (FFE only (Portuguese)) condition at
midline was 0.2804622 (SD = 0.8786022) and the mean for the Treatment (FFE + lit (Portuguese)) condition
at midline was 0.5711557 (SD = 1.20088). The difference between Control (FFE only (Portuguese)) and
Treatment (FFE + lit (Portuguese)) at midline was thus of 0.2906935 points. The p-value for this difference
was 0.001308768. The change from the baseline to the midline of 0.2804622 points for the Control (FFE
only (Portuguese)) condition establishes the conterfactual against which the change for the Treatment (FFE
+ lit (Portuguese)) can be assessed against. The change for the Treatment (FFE + lit (Portuguese)) from
the baseline to the midline was of 0.5605788 points. Consequently, the change for the Control (FFE only
(Portuguese)) relative to the change for the Treatment (FFE + lit (Portuguese)) condition was of 0.2801166
points. The p-value for this difference was 0.001853958. Accordingly, the difference was statistically signifi￾cant (p < 0.01 corrected for 4 additional contrasts) and thus there is evidence of a different degree of change
from baseline to midline in the Treatment (FFE + lit (Portuguese)) when compared to the Control (FFE
only (Portuguese)) condition. This provides evidence that the Treatment (FFE + lit (Portuguese)) (or some
other unobserved process) impacted on the Comprehension of reading EGRA subtask and thus the rate of
change of the two conditions relative to their baselines differed.
123
1.12.5 Contrast 5: FFE + lit (Bilingual) vs FFE + lit (Portuguese)
Table 60: Comprehension of reading
Baseline Midline
Intervention Mean SD n min max Mean SD n min max
Control 0.011 0.178 1040 0 4 0.571 1.201 1047 0 4
Treatment 0.028 0.239 538 0 3 0.485 1.056 761 0 4
0.0
0.2
0.4
Control Treatment
CINTERVENTION
Mean
PERIOD
Baseline
Midline
Comprehension of reading
124
0
1
2
3
4
Control Treatment
CINTERVENTION
Comprehension of reading
PERIOD
Baseline
Midline
Comprehension of reading
As shown in the table above, for the the Comprehension of reading EGRA subtask, the mean for the Control
(FFE + lit (Portuguese)) condition at baseline was 0.01057692 (SD = 0.1779026) and the mean for the
Treatment (FFE + lit (Bilingual)) condition at baseline was 0.02788104 (SD = 0.2386406). The difference
between Control (FFE + lit (Portuguese)) and Treatment (FFE + lit (Bilingual)) at baseline was thus of
0.01730412 points. The p-value for this difference was 0.2648872. The mean for the Control (FFE + lit
(Portuguese)) condition at midline was 0.5711557 (SD = 1.20088) and the mean for the Treatment (FFE +
lit (Bilingual)) condition at midline was 0.4848883 (SD = 1.056357). The difference between Control (FFE
+ lit (Portuguese)) and Treatment (FFE + lit (Bilingual)) at midline was thus of 0.08626738 points. The
p-value for this difference was 0.366519. The change from the baseline to the midline of 0.5605788 points
for the Control (FFE + lit (Portuguese)) condition establishes the conterfactual against which the change
for the Treatment (FFE + lit (Bilingual)) can be assessed against. The change for the Treatment (FFE +
lit (Bilingual)) from the baseline to the midline was of 0.4570073 points. Consequently, the change for the
Control (FFE + lit (Portuguese)) relative to the change for the Treatment (FFE + lit (Bilingual)) condition
was of -0.1035715 points. The p-value for this difference was 0.2842989. Accordingly, the difference was
not statistically significant (p < 0.01 corrected for 5 additional contrasts) and therefore the statistical test
provides no conclusive evidence in regards to there being a difference in the rate of change between the
baseline and the midline for the Comprehension of reading EGRA subtask across the Control (FFE + lit
(Portuguese)) and the Treatment (FFE + lit (Bilingual)) conditions beyond that expected from sampling
error.
125
1.13 EGRA_ST9_2A: Total writing
1.13.1 Contrast 1: FFE + lit (all) vs Comparison (all)
Table 61: Total writing
Baseline Midline
Intervention Mean SD n min max Mean SD n min max
Control 0.012 0.288 1136 0 9 2.837 3.935 1081 0 12
Treatment 0.046 0.658 1578 0 10 2.170 3.653 1808 0 10
0
1
2
Control Treatment
CINTERVENTION
Mean
PERIOD
Baseline
Midline
Total writing
126
0.0
2.5
5.0
7.5
10.0
12.5
Control Treatment
CINTERVENTION
Total writing
PERIOD
Baseline
Midline
Total writing
As shown in the table above, for the the Total writing EGRA subtask, the mean for the Control (Comparison
(all)) condition at baseline was 0.01232394 (SD = 0.2875193) and the mean for the Treatment (FFE + lit
(all)) condition at baseline was 0.04626109 (SD = 0.658404). The difference between Control (Comparison
(all)) and Treatment (FFE + lit (all)) at baseline was thus of 0.03393715 points. The p-value for this
difference was 0.06935504. The mean for the Control (Comparison (all)) condition at midline was 2.837188
(SD = 3.934574) and the mean for the Treatment (FFE + lit (all)) condition at midline was 2.170354 (SD
= 3.653165). The difference between Control (Comparison (all)) and Treatment (FFE + lit (all)) at midline
was thus of 0.6668338 points. The p-value for this difference was 0.01980337. The change from the baseline
to the midline of 2.824864 points for the Control (Comparison (all)) condition establishes the conterfactual
against which the change for the Treatment (FFE + lit (all)) can be assessed against. The change for
the Treatment (FFE + lit (all)) from the baseline to the midline was of 2.124093 points. Consequently,
the change for the Control (Comparison (all)) relative to the change for the Treatment (FFE + lit (all))
condition was of -0.700771 points. The p-value for this difference was 0.01444353. Accordingly, the difference
was not statistically significant (p < 0.01 corrected for 5 additional contrasts) and therefore the statistical
test provides no conclusive evidence in regards to there being a difference in the rate of change between the
baseline and the midline for the Total writing EGRA subtask across the Control (Comparison (all)) and the
Treatment (FFE + lit (all)) conditions beyond that expected from sampling error.
127
1.13.2 Contrast 2: FFE + lit (Portuguese) vs Comparison (Portuguese)
Table 62: Total writing
Baseline Midline
Intervention Mean SD n min max Mean SD n min max
Control 0.013 0.299 1047 0 9 3.052 4.026 972 0 12
Treatment 0.047 0.679 1040 0 10 2.224 3.641 1047 0 10
0
1
2
3
Control Treatment
CINTERVENTION
Mean
PERIOD
Baseline
Midline
Total writing
128
0.0
2.5
5.0
7.5
10.0
12.5
Control Treatment
CINTERVENTION
Total writing
PERIOD
Baseline
Midline
Total writing
As shown in the table above, for the the Total writing EGRA subtask, the mean for the Control (Comparison
(Portuguese)) condition at baseline was 0.01337154 (SD = 0.2994782) and the mean for the Treatment (FFE
+ lit (Portuguese)) condition at baseline was 0.04711538 (SD = 0.6787659). The difference between Control
(Comparison (Portuguese)) and Treatment (FFE + lit (Portuguese)) at baseline was thus of 0.03374385
points. The p-value for this difference was 0.136983. The mean for the Control (Comparison (Portuguese))
condition at midline was 3.052469 (SD = 4.026217) and the mean for the Treatment (FFE + lit (Por￾tuguese)) condition at midline was 2.224451 (SD = 3.641204). The difference between Control (Comparison
(Portuguese)) and Treatment (FFE + lit (Portuguese)) at midline was thus of 0.8280183 points. The p-value
for this difference was 0.009876457. The change from the baseline to the midline of 3.039098 points for the
Control (Comparison (Portuguese)) condition establishes the conterfactual against which the change for the
Treatment (FFE + lit (Portuguese)) can be assessed against. The change for the Treatment (FFE + lit (Por￾tuguese)) from the baseline to the midline was of 2.177335 points. Consequently, the change for the Control
(Comparison (Portuguese)) relative to the change for the Treatment (FFE + lit (Portuguese)) condition
was of -0.8617622 points. The p-value for this difference was 0.006910867. Accordingly, the difference was
statistically significant (p < 0.01 corrected for 4 additional contrasts) and thus there is evidence of a different
degree of change from baseline to midline in the Treatment (FFE + lit (Portuguese)) when compared to
the Control (Comparison (Portuguese)) condition. This provides evidence that the Treatment (FFE + lit
(Portuguese)) (or some other unobserved process) impacted on the Total writing EGRA subtask and thus
the rate of change of the two conditions relative to their baselines differed. It should be pointed out that the
difference in the rate of change of the Treatment (FFE + lit (Portuguese)) relative to the Control (Compar￾ison (Portuguese)) was negative and thus suggests that students in the Control (Comparison (Portuguese))
schools performed significantly better than those in the Treatment (FFE + lit (Portuguese)) .
129
1.13.3 Contrast 3: FFE + lit (Bilingual) vs Comparison (Bilingual)
Table 63: Total writing
Baseline Midline
Intervention Mean SD n min max Mean SD n min max
Control 0.000 0.000 89 0 0 0.917 2.220 109 0 10
Treatment 0.045 0.618 538 0 10 2.096 3.671 761 0 10
0.0
0.5
1.0
1.5
2.0
Control Treatment
CINTERVENTION
Mean
PERIOD
Baseline
Midline
Total writing
130
0.0
2.5
5.0
7.5
10.0
Control Treatment
CINTERVENTION
Total writing
PERIOD
Baseline
Midline
Total writing
As shown in the table above, for the the Total writing EGRA subtask, the mean for the Control (Comparison
(Bilingual)) condition at baseline was 0 (SD = 0) and the mean for the Treatment (FFE + lit (Bilingual))
condition at baseline was 0.04460967 (SD = 0.6177531). The difference between Control (Comparison (Bilin￾gual)) and Treatment (FFE + lit (Bilingual)) at baseline was thus of 0.04460967 points. The p-value for
this difference was 0.1183386. The mean for the Control (Comparison (Bilingual)) condition at midline was
0.9174312 (SD = 2.219978) and the mean for the Treatment (FFE + lit (Bilingual)) condition at midline was
2.095926 (SD = 3.670651). The difference between Control (Comparison (Bilingual)) and Treatment (FFE
+ lit (Bilingual)) at midline was thus of 1.178495 points. The p-value for this difference was 0.008536447.
The change from the baseline to the midline of 0.9174312 points for the Control (Comparison (Bilingual))
condition establishes the conterfactual against which the change for the Treatment (FFE + lit (Bilingual))
can be assessed against. The change for the Treatment (FFE + lit (Bilingual)) from the baseline to the mid￾line was of 2.051317 points. Consequently, the change for the Control (Comparison (Bilingual)) relative to
the change for the Treatment (FFE + lit (Bilingual)) condition was of 1.133886 points. The p-value for this
difference was 0.01217518. Accordingly, the difference was not statistically significant (p < 0.01 corrected for
5 additional contrasts) and therefore the statistical test provides no conclusive evidence in regards to there
being a difference in the rate of change between the baseline and the midline for the Total writing EGRA
subtask across the Control (Comparison (Bilingual)) and the Treatment (FFE + lit (Bilingual)) conditions
beyond that expected from sampling error.
131
1.13.4 Contrast 4: FFE + lit (Portuguese) vs FFE only (Portuguese)
Table 64: Total writing
Baseline Midline
Intervention Mean SD n min max Mean SD n min max
Control 0.001 0.032 1007 0 1 1.289 2.920 952 0 10
Treatment 0.047 0.679 1040 0 10 2.224 3.641 1047 0 10
0.0
0.5
1.0
1.5
2.0
Control Treatment
CINTERVENTION
Mean
PERIOD
Baseline
Midline
Total writing
132
0.0
2.5
5.0
7.5
10.0
Control Treatment
CINTERVENTION
Total writing
PERIOD
Baseline
Midline
Total writing
As shown in the table above, for the the Total writing EGRA subtask, the mean for the Control (FFE
only (Portuguese)) condition at baseline was 0.0009930487 (SD = 0.03151267) and the mean for the Treat￾ment (FFE + lit (Portuguese)) condition at baseline was 0.04711538 (SD = 0.6787659). The difference
between Control (FFE only (Portuguese)) and Treatment (FFE + lit (Portuguese)) at baseline was thus of
0.04612234 points. The p-value for this difference was 0.02613085. The mean for the Control (FFE only
(Portuguese)) condition at midline was 1.288866 (SD = 2.919812) and the mean for the Treatment (FFE +
lit (Portuguese)) condition at midline was 2.224451 (SD = 3.641204). The difference between Control (FFE
only (Portuguese)) and Treatment (FFE + lit (Portuguese)) at midline was thus of 0.9355853 points. The
p-value for this difference was 0.0009320036. The change from the baseline to the midline of 1.287872 points
for the Control (FFE only (Portuguese)) condition establishes the conterfactual against which the change
for the Treatment (FFE + lit (Portuguese)) can be assessed against. The change for the Treatment (FFE
+ lit (Portuguese)) from the baseline to the midline was of 2.177335 points. Consequently, the change for
the Control (FFE only (Portuguese)) relative to the change for the Treatment (FFE + lit (Portuguese))
condition was of 0.8894629 points. The p-value for this difference was 0.001531693. Accordingly, the differ￾ence was statistically significant (p < 0.01 corrected for 4 additional contrasts) and thus there is evidence
of a different degree of change from baseline to midline in the Treatment (FFE + lit (Portuguese)) when
compared to the Control (FFE only (Portuguese)) condition. This provides evidence that the Treatment
(FFE + lit (Portuguese)) (or some other unobserved process) impacted on the Total writing EGRA subtask
and thus the rate of change of the two conditions relative to their baselines differed.
133
1.13.5 Contrast 5: FFE + lit (Bilingual) vs FFE + lit (Portuguese)
Table 65: Total writing
Baseline Midline
Intervention Mean SD n min max Mean SD n min max
Control 0.047 0.679 1040 0 10 2.224 3.641 1047 0 10
Treatment 0.045 0.618 538 0 10 2.096 3.671 761 0 10
0.0
0.5
1.0
1.5
2.0
Control Treatment
CINTERVENTION
Mean
PERIOD
Baseline
Midline
Total writing
134
0.0
2.5
5.0
7.5
10.0
Control Treatment
CINTERVENTION
Total writing
PERIOD
Baseline
Midline
Total writing
As shown in the table above, for the the Total writing EGRA subtask, the mean for the Control (FFE + lit
(Portuguese)) condition at baseline was 0.04711538 (SD = 0.6787659) and the mean for the Treatment (FFE
+ lit (Bilingual)) condition at baseline was 0.04460967 (SD = 0.6177531). The difference between Control
(FFE + lit (Portuguese)) and Treatment (FFE + lit (Bilingual)) at baseline was thus of 0.002505719 points.
The p-value for this difference was 0.943553. The mean for the Control (FFE + lit (Portuguese)) condition
at midline was 2.224451 (SD = 3.641204) and the mean for the Treatment (FFE + lit (Bilingual)) condition
at midline was 2.095926 (SD = 3.670651). The difference between Control (FFE + lit (Portuguese)) and
Treatment (FFE + lit (Bilingual)) at midline was thus of 0.1285244 points. The p-value for this difference
was 0.6731649. The change from the baseline to the midline of 2.177335 points for the Control (FFE + lit
(Portuguese)) condition establishes the conterfactual against which the change for the Treatment (FFE + lit
(Bilingual)) can be assessed against. The change for the Treatment (FFE + lit (Bilingual)) from the baseline
to the midline was of 2.051317 points. Consequently, the change for the Control (FFE + lit (Portuguese))
relative to the change for the Treatment (FFE + lit (Bilingual)) condition was of -0.1260187 points. The
p-value for this difference was 0.6817533. Accordingly, the difference was not statistically significant (p <
0.01 corrected for 5 additional contrasts) and therefore the statistical test provides no conclusive evidence
in regards to there being a difference in the rate of change between the baseline and the midline for the
Total writing EGRA subtask across the Control (FFE + lit (Portuguese)) and the Treatment (FFE + lit
(Bilingual)) conditions beyond that expected from sampling error.
135
1.14 EGRA_ST9_2B: Writing words
1.14.1 Contrast 1: FFE + lit (all) vs Comparison (all)
Table 66: Writing words
Baseline Midline
Intervention Mean SD n min max Mean SD n min max
Control NA NA 0 NA NA 1.03 1.776 1081 0 5
Treatment NA NA 0 NA NA 1.01 1.759 1808 0 5
0.00
0.25
0.50
0.75
1.00
Control Treatment
CINTERVENTION
Mean
PERIOD
Midline
Writing words
136
0
1
2
3
4
5
Control Treatment
CINTERVENTION
Writing words
PERIOD
Midline
Writing words
As shown in the table above, for the Writing words EGRA subtask, the mean for the Control (Comparison
(all)) condition at midline was 1.029602 (SD = 1.77562) and the mean for the Treatment (FFE + lit (all))
condition at midline was 1.009956 (SD = 1.759443). The difference for the Writing words across the Control
(Comparison (all)) and Treatment (FFE + lit (all)) conditions at midline was thus -0.01964647 points (there
was no baseline measurement for this variable). The p-value for this difference was 0.878366. Accordingly, the
difference was not statistically significant and therefore, the statistical test provides no conclusive evidence
in regards to the there being a difference in the levels of Writing wordsbetween the Control (Comparison
(all)) and Treatment (FFE + lit (all)) conditions at midline beyond that which could result from sampling
error.
1.14.2 Contrast 2: FFE + lit (Portuguese) vs Comparison (Portuguese)
Table 67: Writing words
Baseline Midline
Intervention Mean SD n min max Mean SD n min max
Control NA NA 0 NA NA 1.121 1.829 972 0 5
Treatment NA NA 0 NA NA 1.069 1.776 1047 0 5
137
0.0
0.3
0.6
0.9
Control Treatment
CINTERVENTION
Mean
PERIOD
Midline
Writing words
138
0
1
2
3
4
5
Control Treatment
CINTERVENTION
Writing words
PERIOD
Midline
Writing words
As shown in the table above, for the Writing words EGRA subtask, the mean for the Control (Comparison
(Portuguese)) condition at midline was 1.121399 (SD = 1.828844) and the mean for the Treatment (FFE
+ lit (Portuguese)) condition at midline was 1.068768 (SD = 1.775941). The difference for the Writing
words across the Control (Comparison (Portuguese)) and Treatment (FFE + lit (Portuguese)) conditions
at midline was thus -0.05263127 points (there was no baseline measurement for this variable). The p-value
for this difference was 0.7195677. Accordingly, the difference was not statistically significant and therefore,
the statistical test provides no conclusive evidence in regards to the there being a difference in the levels
of Writing wordsbetween the Control (Comparison (Portuguese)) and Treatment (FFE + lit (Portuguese))
conditions at midline beyond that which could result from sampling error.
1.14.3 Contrast 3: FFE + lit (Bilingual) vs Comparison (Bilingual)
Table 68: Writing words
Baseline Midline
Intervention Mean SD n min max Mean SD n min max
Control NA NA 0 NA NA 0.211 0.840 109 0 5
Treatment NA NA 0 NA NA 0.929 1.734 761 0 5
139
0.00
0.25
0.50
0.75
Control Treatment
CINTERVENTION
Mean
PERIOD
Midline
Writing words
140
0
1
2
3
4
5
Control Treatment
CINTERVENTION
Writing words
PERIOD
Midline
Writing words
As shown in the table above, for the Writing words EGRA subtask, the mean for the Control (Comparison
(Bilingual)) condition at midline was 0.2110092 (SD = 0.8396802) and the mean for the Treatment (FFE +
lit (Bilingual)) condition at midline was 0.9290407 (SD = 1.734392). The difference for the Writing words
across the Control (Comparison (Bilingual)) and Treatment (FFE + lit (Bilingual)) conditions at midline
was thus 0.7180316 points (there was no baseline measurement for this variable). The p-value for this
difference was 1.588162e-06. Accordingly, the difference was statistically significant (p < 0.01 corrected for
4 additional contrasts). Nevertheless, due to the absence of baseline values and lack of random allocation to
the treatment and control conditions, a causal attribution of this effect to the treatment is not possible
1.14.4 Contrast 4: FFE + lit (Portuguese) vs FFE only (Portuguese)
Table 69: Writing words
Baseline Midline
Intervention Mean SD n min max Mean SD n min max
Control NA NA 0 NA NA 0.623 1.424 952 0 5
Treatment NA NA 0 NA NA 1.069 1.776 1047 0 5
141
0.0
0.3
0.6
0.9
Control Treatment
CINTERVENTION
Mean
PERIOD
Midline
Writing words
142
0
1
2
3
4
5
Control Treatment
CINTERVENTION
Writing words
PERIOD
Midline
Writing words
As shown in the table above, for the Writing words EGRA subtask, the mean for the Control (FFE only
(Portuguese)) condition at midline was 0.6228992 (SD = 1.424445) and the mean for the Treatment (FFE
+ lit (Portuguese)) condition at midline was 1.068768 (SD = 1.775941). The difference for the Writing
words across the Control (FFE only (Portuguese)) and Treatment (FFE + lit (Portuguese)) conditions at
midline was thus 0.4458687 points (there was no baseline measurement for this variable). The p-value for
this difference was 0.001099579. Accordingly, the difference was statistically significant (p < 0.01 corrected
for 4 additional contrasts). Nevertheless, due to the absence of baseline values and lack of random allocation
to the treatment and control conditions, a causal attribution of this effect to the treatment is not possible
1.14.5 Contrast 5: FFE + lit (Bilingual) vs FFE + lit (Portuguese)
Table 70: Writing words
Baseline Midline
Intervention Mean SD n min max Mean SD n min max
Control NA NA 0 NA NA 1.069 1.776 1047 0 5
Treatment NA NA 0 NA NA 0.929 1.734 761 0 5
143
0.0
0.3
0.6
0.9
Control Treatment
CINTERVENTION
Mean
PERIOD
Midline
Writing words
144
0
1
2
3
4
5
Control Treatment
CINTERVENTION
Writing words
PERIOD
Midline
Writing words
As shown in the table above, for the Writing words EGRA subtask, the mean for the Control (FFE + lit
(Portuguese)) condition at midline was 1.068768 (SD = 1.775941) and the mean for the Treatment (FFE +
lit (Bilingual)) condition at midline was 0.9290407 (SD = 1.734392). The difference for the Writing words
across the Control (FFE + lit (Portuguese)) and Treatment (FFE + lit (Bilingual)) conditions at midline was
thus -0.1397272 points (there was no baseline measurement for this variable). The p-value for this difference
was 0.3384802. Accordingly, the difference was not statistically significant and therefore, the statistical test
provides no conclusive evidence in regards to the there being a difference in the levels of Writing wordsbetween
the Control (FFE + lit (Portuguese)) and Treatment (FFE + lit (Bilingual)) conditions at midline beyond
that which could result from sampling error.
1.15 EGRA_ST9_2C: Writing phrase
1.15.1 Contrast 1: FFE + lit (all) vs Comparison (all)
Table 71: Writing phrase
Baseline Midline
Intervention Mean SD n min max Mean SD n min max
Control NA NA 0 NA NA 0.851 1.806 1081 0 5
Treatment NA NA 0 NA NA 1.074 1.941 1808 0 5
145
0.0
0.3
0.6
0.9
Control Treatment
CINTERVENTION
Mean
PERIOD
Midline
Writing phrase
146
0
1
2
3
4
5
Control Treatment
CINTERVENTION
Writing phrase
PERIOD
Midline
Writing phrase
As shown in the table above, for the Writing phrase EGRA subtask, the mean for the Control (Comparison
(all)) condition at midline was 0.8510638 (SD = 1.805604) and the mean for the Treatment (FFE + lit
(all)) condition at midline was 1.074115 (SD = 1.940748). The difference for the Writing phrase across
the Control (Comparison (all)) and Treatment (FFE + lit (all)) conditions at midline was thus 0.2230512
points (there was no baseline measurement for this variable). The p-value for this difference was 0.06842061.
Accordingly, the difference was not statistically significant and therefore, the statistical test provides no
conclusive evidence in regards to the there being a difference in the levels of Writing phrasebetween the
Control (Comparison (all)) and Treatment (FFE + lit (all)) conditions at midline beyond that which could
result from sampling error.
1.15.2 Contrast 2: FFE + lit (Portuguese) vs Comparison (Portuguese)
Table 72: Writing phrase
Baseline Midline
Intervention Mean SD n min max Mean SD n min max
Control NA NA 0 NA NA 0.924 1.864 972 0 5
Treatment NA NA 0 NA NA 1.086 1.923 1047 0 5
147
0.0
0.3
0.6
0.9
Control Treatment
CINTERVENTION
Mean
PERIOD
Midline
Writing phrase
148
0
1
2
3
4
5
Control Treatment
CINTERVENTION
Writing phrase
PERIOD
Midline
Writing phrase
As shown in the table above, for the Writing phrase EGRA subtask, the mean for the Control (Comparison
(Portuguese)) condition at midline was 0.9238683 (SD = 1.864174) and the mean for the Treatment (FFE
+ lit (Portuguese)) condition at midline was 1.08596 (SD = 1.923057). The difference for the Writing
phrase across the Control (Comparison (Portuguese)) and Treatment (FFE + lit (Portuguese)) conditions
at midline was thus 0.1620916 points (there was no baseline measurement for this variable). The p-value
for this difference was 0.2546944. Accordingly, the difference was not statistically significant and therefore,
the statistical test provides no conclusive evidence in regards to the there being a difference in the levels
of Writing phrasebetween the Control (Comparison (Portuguese)) and Treatment (FFE + lit (Portuguese))
conditions at midline beyond that which could result from sampling error.
1.15.3 Contrast 3: FFE + lit (Bilingual) vs Comparison (Bilingual)
Table 73: Writing phrase
Baseline Midline
Intervention Mean SD n min max Mean SD n min max
Control NA NA 0 NA NA 0.202 0.941 109 0 5
Treatment NA NA 0 NA NA 1.058 1.966 761 0 5
149
0.00
0.25
0.50
0.75
1.00
Control Treatment
CINTERVENTION
Mean
PERIOD
Midline
Writing phrase
150
0
1
2
3
4
5
Control Treatment
CINTERVENTION
Writing phrase
PERIOD
Midline
Writing phrase
As shown in the table above, for the Writing phrase EGRA subtask, the mean for the Control (Comparison
(Bilingual)) condition at midline was 0.2018349 (SD = 0.9406441) and the mean for the Treatment (FFE +
lit (Bilingual)) condition at midline was 1.057819 (SD = 1.965979). The difference for the Writing phrase
across the Control (Comparison (Bilingual)) and Treatment (FFE + lit (Bilingual)) conditions at midline
was thus 0.8559838 points (there was no baseline measurement for this variable). The p-value for this
difference was 1.358622e-05. Accordingly, the difference was statistically significant (p < 0.01 corrected for
4 additional contrasts). Nevertheless, due to the absence of baseline values and lack of random allocation to
the treatment and control conditions, a causal attribution of this effect to the treatment is not possible
1.15.4 Contrast 4: FFE + lit (Portuguese) vs FFE only (Portuguese)
Table 74: Writing phrase
Baseline Midline
Intervention Mean SD n min max Mean SD n min max
Control NA NA 0 NA NA 0.628 1.533 952 0 5
Treatment NA NA 0 NA NA 1.086 1.923 1047 0 5
151
0.0
0.3
0.6
0.9
Control Treatment
CINTERVENTION
Mean
PERIOD
Midline
Writing phrase
152
0
1
2
3
4
5
Control Treatment
CINTERVENTION
Writing phrase
PERIOD
Midline
Writing phrase
As shown in the table above, for the Writing phrase EGRA subtask, the mean for the Control (FFE only
(Portuguese)) condition at midline was 0.6281513 (SD = 1.532805) and the mean for the Treatment (FFE
+ lit (Portuguese)) condition at midline was 1.08596 (SD = 1.923057). The difference for the Writing
phrase across the Control (FFE only (Portuguese)) and Treatment (FFE + lit (Portuguese)) conditions at
midline was thus 0.4578086 points (there was no baseline measurement for this variable). The p-value for
this difference was 0.001876018. Accordingly, the difference was statistically significant (p < 0.01 corrected
for 4 additional contrasts). Nevertheless, due to the absence of baseline values and lack of random allocation
to the treatment and control conditions, a causal attribution of this effect to the treatment is not possible
1.15.5 Contrast 5: FFE + lit (Bilingual) vs FFE + lit (Portuguese)
Table 75: Writing phrase
Baseline Midline
Intervention Mean SD n min max Mean SD n min max
Control NA NA 0 NA NA 1.086 1.923 1047 0 5
Treatment NA NA 0 NA NA 1.058 1.966 761 0 5
153
0.0
0.3
0.6
0.9
Control Treatment
CINTERVENTION
Mean
PERIOD
Midline
Writing phrase
154
0
1
2
3
4
5
Control Treatment
CINTERVENTION
Writing phrase
PERIOD
Midline
Writing phrase
As shown in the table above, for the Writing phrase EGRA subtask, the mean for the Control (FFE + lit
(Portuguese)) condition at midline was 1.08596 (SD = 1.923057) and the mean for the Treatment (FFE + lit
(Bilingual)) condition at midline was 1.057819 (SD = 1.965979). The difference for the Writing phrase across
the Control (FFE + lit (Portuguese)) and Treatment (FFE + lit (Bilingual)) conditions at midline was thus
-0.02814123 points (there was no baseline measurement for this variable). The p-value for this difference
was 0.8598707. Accordingly, the difference was not statistically significant and therefore, the statistical
test provides no conclusive evidence in regards to the there being a difference in the levels of Writing
phrasebetween the Control (FFE + lit (Portuguese)) and Treatment (FFE + lit (Bilingual)) conditions at
midline beyond that which could result from sampling error.
2 EGRA Categorical Variables
2.1 EGRA_ST9_1A: Writing first name
2.1.1 Contrast 1: FFE + lit (all) vs Comparison (all)
Table 76: Writing first name
Baseline Midline
Intervention % n % n
Control 0.057 1136 0.709 1081
Treatment 0.081 1578 0.722 1808
155
0.0
0.2
0.4
0.6
Control Treatment
CINTERVENTION
Mean
PERIOD
Baseline
Midline
Writing first name
As shown in the table above, for the the Writing first name EGRA subtask, the percentage correct for the
Control (Comparison (all)) condition at baseline was 0.05721831 and the percentage correct for the Treatment
(FFE + lit (all)) condition at baseline was 0.08111534. The difference between Control (Comparison (all))
and Treatment (FFE + lit (all)) at baseline was thus of 0.02389703 percentage points. The p-value for this
difference was 0.1636942. The percentage correct for the Control (Comparison (all)) condition at midline
was 0.7086031 and the percentage correct for the Treatment (FFE + lit (all)) condition at midline was
0.721792. The difference between Control (Comparison (all)) and Treatment (FFE + lit (all)) at midline
was thus of 0.01318889 points. The p-value for this difference was 0.6214181. The change from the baseline
to the midline of 0.6513848 points for the Control (Comparison (all)) condition establishes the conterfactual
against which the change for the Treatment (FFE + lit (all)) can be assessed against. The change for the
Treatment (FFE + lit (all)) from the baseline to the midline was of 0.6406767 points. Consequently, the
change for the Control (Comparison (all)) relative to the change for the Treatment (FFE + lit (all)) condition
was of -0.01070814 points. The p-value for this difference was 0.2805932. Accordingly, the difference was
not statistically significant (p < 0.01 corrected for 5 additional contrasts) and therefore the statistical test
provides no conclusive evidence in regards to there being a difference in the rate of change between the
baseline and the midline for the Writing first name EGRA subtask across the Control (Comparison (all))
and the Treatment (FFE + lit (all)) conditions beyond that expected from sampling error.
156
2.1.2 Contrast 2: FFE + lit (Portuguese) vs Comparison (Portuguese)
Table 77: Writing first name
Baseline Midline
Intervention % n % n
Control 0.061 1047 0.744 972
Treatment 0.095 1040 0.732 1047
0.0
0.2
0.4
0.6
Control Treatment
CINTERVENTION
Mean
PERIOD
Baseline
Midline
Writing first name
As shown in the table above, for the the Writing first name EGRA subtask, the percentage correct for the
Control (Comparison (Portuguese)) condition at baseline was 0.06112703 and the percentage correct for the
Treatment (FFE + lit (Portuguese)) condition at baseline was 0.09519231. The difference between Control
(Comparison (Portuguese)) and Treatment (FFE + lit (Portuguese)) at baseline was thus of 0.03406528
percentage points. The p-value for this difference was 0.09135377. The percentage correct for the Control
(Comparison (Portuguese)) condition at midline was 0.7438272 and the percentage correct for the Treatment
(FFE + lit (Portuguese)) condition at midline was 0.7316141. The difference between Control (Comparison
(Portuguese)) and Treatment (FFE + lit (Portuguese)) at midline was thus of 0.01221302 points. The p￾value for this difference was 0.6539434. The change from the baseline to the midline of 0.6827001 points
for the Control (Comparison (Portuguese)) condition establishes the conterfactual against which the change
for the Treatment (FFE + lit (Portuguese)) can be assessed against. The change for the Treatment (FFE
+ lit (Portuguese)) from the baseline to the midline was of 0.6364218 points. Consequently, the change for
the Control (Comparison (Portuguese)) relative to the change for the Treatment (FFE + lit (Portuguese))
condition was of -0.0462783 points. The p-value for this difference was 0.07974952. Accordingly, the difference
was not statistically significant (p < 0.01 corrected for 5 additional contrasts) and therefore the statistical
test provides no conclusive evidence in regards to there being a difference in the rate of change between
157
the baseline and the midline for the Writing first name EGRA subtask across the Control (Comparison
(Portuguese)) and the Treatment (FFE + lit (Portuguese)) conditions beyond that expected from sampling
error.
2.1.3 Contrast 3: FFE + lit (Bilingual) vs Comparison (Bilingual)
Table 78: Writing first name
Baseline Midline
Intervention % n % n
Control 0.011 89 0.394 109
Treatment 0.054 538 0.708 761
0.0
0.2
0.4
0.6
Control Treatment
CINTERVENTION
Mean
PERIOD
Baseline
Midline
Writing first name
As shown in the table above, for the the Writing first name EGRA subtask, the percentage correct for
the Control (Comparison (Bilingual)) condition at baseline was 0.01123596 and the percentage correct for
the Treatment (FFE + lit (Bilingual)) condition at baseline was 0.05390335. The difference between Con￾trol (Comparison (Bilingual)) and Treatment (FFE + lit (Bilingual)) at baseline was thus of 0.04266739
percentage points. The p-value for this difference was 0.1000596. The percentage correct for the Control
(Comparison (Bilingual)) condition at midline was 0.3944954 and the percentage correct for the Treatment
(FFE + lit (Bilingual)) condition at midline was 0.7082786. The difference between Control (Comparison
(Bilingual)) and Treatment (FFE + lit (Bilingual)) at midline was thus of 0.3137832 points. The p-value
for this difference was 4.158437e-06. The change from the baseline to the midline of 0.3832595 points for
the Control (Comparison (Bilingual)) condition establishes the conterfactual against which the change for
the Treatment (FFE + lit (Bilingual)) can be assessed against. The change for the Treatment (FFE + lit
(Bilingual)) from the baseline to the midline was of 0.6543752 points. Consequently, the change for the
158
Control (Comparison (Bilingual)) relative to the change for the Treatment (FFE + lit (Bilingual)) condition
was of 0.2711158 points. The p-value for this difference was 0.7380733. Accordingly, the difference was not
statistically significant (p < 0.01 corrected for 5 additional contrasts) and therefore the statistical test pro￾vides no conclusive evidence in regards to there being a difference in the rate of change between the baseline
and the midline for the Writing first name EGRA subtask across the Control (Comparison (Bilingual)) and
the Treatment (FFE + lit (Bilingual)) conditions beyond that expected from sampling error.
2.1.4 Contrast 4: FFE + lit (Portuguese) vs FFE only (Portuguese)
Table 79: Writing first name
Baseline Midline
Intervention % n % n
Control 0.139 1007 0.600 952
Treatment 0.095 1040 0.732 1047
0.0
0.2
0.4
0.6
Control Treatment
CINTERVENTION
Mean
PERIOD
Baseline
Midline
Writing first name
As shown in the table above, for the the Writing first name EGRA subtask, the percentage correct for
the Control (FFE only (Portuguese)) condition at baseline was 0.1390268 and the percentage correct for
the Treatment (FFE + lit (Portuguese)) condition at baseline was 0.09519231. The difference between
Control (FFE only (Portuguese)) and Treatment (FFE + lit (Portuguese)) at baseline was thus of 0.0438345
percentage points. The p-value for this difference was 0.05972446. The percentage correct for the Control
(FFE only (Portuguese)) condition at midline was 0.5997899 and the percentage correct for the Treatment
(FFE + lit (Portuguese)) condition at midline was 0.7316141. The difference between Control (FFE only
(Portuguese)) and Treatment (FFE + lit (Portuguese)) at midline was thus of 0.1318242 points. The p-value
for this difference was 1.907784e-05. The change from the baseline to the midline of 0.4607631 points for
159
the Control (FFE only (Portuguese)) condition establishes the conterfactual against which the change for
the Treatment (FFE + lit (Portuguese)) can be assessed against. The change for the Treatment (FFE + lit
(Portuguese)) from the baseline to the midline was of 0.6364218 points. Consequently, the change for the
Control (FFE only (Portuguese)) relative to the change for the Treatment (FFE + lit (Portuguese)) condition
was of 0.1756587 points. The p-value for this difference was 3.997563e-05. Accordingly, the difference was
statistically significant (p < 0.01 corrected for 4 additional contrasts) and thus there is evidence of a different
degree of change from baseline to midline in the Treatment (FFE + lit (Portuguese)) when compared to
the Control (FFE only (Portuguese)) condition. This provides evidence that the Treatment (FFE + lit
(Portuguese)) (or some other unobserved process) impacted on the Writing first name EGRA subtask and
thus the rate of change of the two conditions relative to their baselines differed.
2.1.5 Contrast 5: FFE + lit (Bilingual) vs FFE + lit (Portuguese)
Table 80: Writing first name
Baseline Midline
Intervention % n % n
Control 0.095 1040 0.732 1047
Treatment 0.054 538 0.708 761
0.0
0.2
0.4
0.6
Control Treatment
CINTERVENTION
Mean
PERIOD
Baseline
Midline
Writing first name
As shown in the table above, for the the Writing first name EGRA subtask, the percentage correct for
the Control (FFE + lit (Portuguese)) condition at baseline was 0.09519231 and the percentage correct
for the Treatment (FFE + lit (Bilingual)) condition at baseline was 0.05390335. The difference between
Control (FFE + lit (Portuguese)) and Treatment (FFE + lit (Bilingual)) at baseline was thus of 0.04128896
percentage points. The p-value for this difference was 0.08149245. The percentage correct for the Control
160
(FFE + lit (Portuguese)) condition at midline was 0.7316141 and the percentage correct for the Treatment
(FFE + lit (Bilingual)) condition at midline was 0.7082786. The difference between Control (FFE + lit
(Portuguese)) and Treatment (FFE + lit (Bilingual)) at midline was thus of 0.02333555 points. The p￾value for this difference was 0.4414178. The change from the baseline to the midline of 0.6364218 points
for the Control (FFE + lit (Portuguese)) condition establishes the conterfactual against which the change
for the Treatment (FFE + lit (Bilingual)) can be assessed against. The change for the Treatment (FFE
+ lit (Bilingual)) from the baseline to the midline was of 0.6543752 points. Consequently, the change
for the Control (FFE + lit (Portuguese)) relative to the change for the Treatment (FFE + lit (Bilingual))
condition was of 0.01795341 points. The p-value for this difference was 0.1673764. Accordingly, the difference
was not statistically significant (p < 0.01 corrected for 5 additional contrasts) and therefore the statistical
test provides no conclusive evidence in regards to there being a difference in the rate of change between
the baseline and the midline for the Writing first name EGRA subtask across the Control (FFE + lit
(Portuguese)) and the Treatment (FFE + lit (Bilingual)) conditions beyond that expected from sampling
error.
2.2 EGRA_ST9_1B: Writing family name
2.2.1 Contrast 1: FFE + lit (all) vs Comparison (all)
Table 81: Writing family name
Baseline Midline
Intervention % n % n
Control 0.013 1136 0.445 1081
Treatment 0.022 1578 0.438 1808
161
0.0
0.1
0.2
0.3
0.4
Control Treatment
CINTERVENTION
Mean
PERIOD
Baseline
Midline
Writing family name
As shown in the table above, for the the Writing family name EGRA subtask, the percentage correct for the
Control (Comparison (all)) condition at baseline was 0.01320423 and the percentage correct for the Treatment
(FFE + lit (all)) condition at baseline was 0.02154626. The difference between Control (Comparison (all))
and Treatment (FFE + lit (all)) at baseline was thus of 0.008342036 percentage points. The p-value for this
difference was 0.2252622. The percentage correct for the Control (Comparison (all)) condition at midline
was 0.4449584 and the percentage correct for the Treatment (FFE + lit (all)) condition at midline was
0.4375. The difference between Control (Comparison (all)) and Treatment (FFE + lit (all)) at midline was
thus of 0.007458372 points. The p-value for this difference was 0.8121453. The change from the baseline to
the midline of 0.4317541 points for the Control (Comparison (all)) condition establishes the conterfactual
against which the change for the Treatment (FFE + lit (all)) can be assessed against. The change for the
Treatment (FFE + lit (all)) from the baseline to the midline was of 0.4159537 points. Consequently, the
change for the Control (Comparison (all)) relative to the change for the Treatment (FFE + lit (all)) condition
was of -0.01580041 points. The p-value for this difference was 0.2086559. Accordingly, the difference was
not statistically significant (p < 0.01 corrected for 5 additional contrasts) and therefore the statistical test
provides no conclusive evidence in regards to there being a difference in the rate of change between the
baseline and the midline for the Writing family name EGRA subtask across the Control (Comparison (all))
and the Treatment (FFE + lit (all)) conditions beyond that expected from sampling error.
162
2.2.2 Contrast 2: FFE + lit (Portuguese) vs Comparison (Portuguese)
Table 82: Writing family name
Baseline Midline
Intervention % n % n
Control 0.013 1047 0.469 972
Treatment 0.026 1040 0.450 1047
0.0
0.1
0.2
0.3
0.4
Control Treatment
CINTERVENTION
Mean
PERIOD
Baseline
Midline
Writing family name
As shown in the table above, for the the Writing family name EGRA subtask, the percentage correct for
the Control (Comparison (Portuguese)) condition at baseline was 0.01337154 and the percentage correct
for the Treatment (FFE + lit (Portuguese)) condition at baseline was 0.02596154. The difference between
Control (Comparison (Portuguese)) and Treatment (FFE + lit (Portuguese)) at baseline was thus of 0.01259
percentage points. The p-value for this difference was 0.1339904. The percentage correct for the Control
(Comparison (Portuguese)) condition at midline was 0.4691358 and the percentage correct for the Treatment
(FFE + lit (Portuguese)) condition at midline was 0.4498567. The difference between Control (Comparison
(Portuguese)) and Treatment (FFE + lit (Portuguese)) at midline was thus of 0.01927907 points. The p￾value for this difference was 0.5487912. The change from the baseline to the midline of 0.4557643 points
for the Control (Comparison (Portuguese)) condition establishes the conterfactual against which the change
for the Treatment (FFE + lit (Portuguese)) can be assessed against. The change for the Treatment (FFE
+ lit (Portuguese)) from the baseline to the midline was of 0.4238952 points. Consequently, the change for
the Control (Comparison (Portuguese)) relative to the change for the Treatment (FFE + lit (Portuguese))
condition was of -0.03186907 points. The p-value for this difference was 0.09599616. Accordingly, the
difference was not statistically significant (p < 0.01 corrected for 5 additional contrasts) and therefore
the statistical test provides no conclusive evidence in regards to there being a difference in the rate of
163
change between the baseline and the midline for the Writing family name EGRA subtask across the Control
(Comparison (Portuguese)) and the Treatment (FFE + lit (Portuguese)) conditions beyond that expected
from sampling error.
2.2.3 Contrast 3: FFE + lit (Bilingual) vs Comparison (Bilingual)
Table 83: Writing family name
Baseline Midline
Intervention % n % n
Control 0.011 89 0.229 109
Treatment 0.013 538 0.420 761
0.0
0.1
0.2
0.3
0.4
Control Treatment
CINTERVENTION
Mean
PERIOD
Baseline
Midline
Writing family name
As shown in the table above, for the the Writing family name EGRA subtask, the percentage correct
for the Control (Comparison (Bilingual)) condition at baseline was 0.01123596 and the percentage correct
for the Treatment (FFE + lit (Bilingual)) condition at baseline was 0.01301115. The difference between
Control (Comparison (Bilingual)) and Treatment (FFE + lit (Bilingual)) at baseline was thus of 0.001775197
percentage points. The p-value for this difference was 0.8855256. The percentage correct for the Control
(Comparison (Bilingual)) condition at midline was 0.2293578 and the percentage correct for the Treatment
(FFE + lit (Bilingual)) condition at midline was 0.4204993. The difference between Control (Comparison
(Bilingual)) and Treatment (FFE + lit (Bilingual)) at midline was thus of 0.1911415 points. The p-value
for this difference was 0.03141115. The change from the baseline to the midline of 0.2181218 points for
the Control (Comparison (Bilingual)) condition establishes the conterfactual against which the change for
the Treatment (FFE + lit (Bilingual)) can be assessed against. The change for the Treatment (FFE +
lit (Bilingual)) from the baseline to the midline was of 0.4074882 points. Consequently, the change for
164
the Control (Comparison (Bilingual)) relative to the change for the Treatment (FFE + lit (Bilingual))
condition was of 0.1893663 points. The p-value for this difference was 0.5372766. Accordingly, the difference
was not statistically significant (p < 0.01 corrected for 5 additional contrasts) and therefore the statistical
test provides no conclusive evidence in regards to there being a difference in the rate of change between
the baseline and the midline for the Writing family name EGRA subtask across the Control (Comparison
(Bilingual)) and the Treatment (FFE + lit (Bilingual)) conditions beyond that expected from sampling error.
2.2.4 Contrast 4: FFE + lit (Portuguese) vs FFE only (Portuguese)
Table 84: Writing family name
Baseline Midline
Intervention % n % n
Control 0.030 1007 0.307 952
Treatment 0.026 1040 0.450 1047
0.0
0.1
0.2
0.3
0.4
Control Treatment
CINTERVENTION
Mean
PERIOD
Baseline
Midline
Writing family name
As shown in the table above, for the the Writing family name EGRA subtask, the percentage correct for the
Control (FFE only (Portuguese)) condition at baseline was 0.02979146 and the percentage correct for the
Treatment (FFE + lit (Portuguese)) condition at baseline was 0.02596154. The difference between Control
(FFE only (Portuguese)) and Treatment (FFE + lit (Portuguese)) at baseline was thus of 0.003829921
percentage points. The p-value for this difference was 0.6835565. The percentage correct for the Control
(FFE only (Portuguese)) condition at midline was 0.3067227 and the percentage correct for the Treatment
(FFE + lit (Portuguese)) condition at midline was 0.4498567. The difference between Control (FFE only
(Portuguese)) and Treatment (FFE + lit (Portuguese)) at midline was thus of 0.143134 points. The p-value
for this difference was 0.0001087363. The change from the baseline to the midline of 0.2769312 points for
165
the Control (FFE only (Portuguese)) condition establishes the conterfactual against which the change for
the Treatment (FFE + lit (Portuguese)) can be assessed against. The change for the Treatment (FFE +
lit (Portuguese)) from the baseline to the midline was of 0.4238952 points. Consequently, the change for
the Control (FFE only (Portuguese)) relative to the change for the Treatment (FFE + lit (Portuguese))
condition was of 0.146964 points. The p-value for this difference was 0.03874017. Accordingly, the difference
was not statistically significant (p < 0.01 corrected for 5 additional contrasts) and therefore the statistical
test provides no conclusive evidence in regards to there being a difference in the rate of change between
the baseline and the midline for the Writing family name EGRA subtask across the Control (FFE only
(Portuguese)) and the Treatment (FFE + lit (Portuguese)) conditions beyond that expected from sampling
error.
2.2.5 Contrast 5: FFE + lit (Bilingual) vs FFE + lit (Portuguese)
Table 85: Writing family name
Baseline Midline
Intervention % n % n
Control 0.026 1040 0.45 1047
Treatment 0.013 538 0.42 761
0.0
0.1
0.2
0.3
0.4
Control Treatment
CINTERVENTION
Mean
PERIOD
Baseline
Midline
Writing family name
As shown in the table above, for the the Writing family name EGRA subtask, the percentage correct for
the Control (FFE + lit (Portuguese)) condition at baseline was 0.02596154 and the percentage correct
for the Treatment (FFE + lit (Bilingual)) condition at baseline was 0.01301115. The difference between
Control (FFE + lit (Portuguese)) and Treatment (FFE + lit (Bilingual)) at baseline was thus of 0.01295039
percentage points. The p-value for this difference was 0.1866927. The percentage correct for the Control
166
(FFE + lit (Portuguese)) condition at midline was 0.4498567 and the percentage correct for the Treatment
(FFE + lit (Bilingual)) condition at midline was 0.4204993. The difference between Control (FFE + lit
(Portuguese)) and Treatment (FFE + lit (Bilingual)) at midline was thus of 0.02935739 points. The p￾value for this difference was 0.4254192. The change from the baseline to the midline of 0.4238952 points
for the Control (FFE + lit (Portuguese)) condition establishes the conterfactual against which the change
for the Treatment (FFE + lit (Bilingual)) can be assessed against. The change for the Treatment (FFE
+ lit (Bilingual)) from the baseline to the midline was of 0.4074882 points. Consequently, the change for
the Control (FFE + lit (Portuguese)) relative to the change for the Treatment (FFE + lit (Bilingual))
condition was of -0.016407 points. The p-value for this difference was 0.2840003. Accordingly, the difference
was not statistically significant (p < 0.01 corrected for 5 additional contrasts) and therefore the statistical
test provides no conclusive evidence in regards to there being a difference in the rate of change between
the baseline and the midline for the Writing family name EGRA subtask across the Control (FFE + lit
(Portuguese)) and the Treatment (FFE + lit (Bilingual)) conditions beyond that expected from sampling
error.
2.3 EGRA_ST9_1C: Writing name overall
2.3.1 Contrast 1: FFE + lit (all) vs Comparison (all)
Table 86: Writing name overall
Baseline Midline
Intervention % n % n
Control 0.011 1136 0.426 1081
Treatment 0.016 1578 0.416 1808
167
0.0
0.1
0.2
0.3
0.4
Control Treatment
CINTERVENTION
Mean
PERIOD
Baseline
Midline
Writing name overall
As shown in the table above, for the the Writing name overall EGRA subtask, the percentage correct for the
Control (Comparison (all)) condition at baseline was 0.01056338 and the percentage correct for the Treatment
(FFE + lit (all)) condition at baseline was 0.01647655. The difference between Control (Comparison (all))
and Treatment (FFE + lit (all)) at baseline was thus of 0.005913172 percentage points. The p-value for this
difference was 0.3554128. The percentage correct for the Control (Comparison (all)) condition at midline
was 0.4255319 and the percentage correct for the Treatment (FFE + lit (all)) condition at midline was
0.4159292. The difference between Control (Comparison (all)) and Treatment (FFE + lit (all)) at midline
was thus of 0.009602711 points. The p-value for this difference was 0.7573427. The change from the baseline
to the midline of 0.4149685 points for the Control (Comparison (all)) condition establishes the conterfactual
against which the change for the Treatment (FFE + lit (all)) can be assessed against. The change for the
Treatment (FFE + lit (all)) from the baseline to the midline was of 0.3994527 points. Consequently, the
change for the Control (Comparison (all)) relative to the change for the Treatment (FFE + lit (all)) condition
was of -0.01551588 points. The p-value for this difference was 0.3152659. Accordingly, the difference was
not statistically significant (p < 0.01 corrected for 5 additional contrasts) and therefore the statistical test
provides no conclusive evidence in regards to there being a difference in the rate of change between the
baseline and the midline for the Writing name overall EGRA subtask across the Control (Comparison (all))
and the Treatment (FFE + lit (all)) conditions beyond that expected from sampling error.
168
2.3.2 Contrast 2: FFE + lit (Portuguese) vs Comparison (Portuguese)
Table 87: Writing name overall
Baseline Midline
Intervention % n % n
Control 0.011 1047 0.449 972
Treatment 0.020 1040 0.430 1047
0.0
0.1
0.2
0.3
0.4
Control Treatment
CINTERVENTION
Mean
PERIOD
Baseline
Midline
Writing name overall
As shown in the table above, for the the Writing name overall EGRA subtask, the percentage correct for the
Control (Comparison (Portuguese)) condition at baseline was 0.01146132 and the percentage correct for the
Treatment (FFE + lit (Portuguese)) condition at baseline was 0.02019231. The difference between Control
(Comparison (Portuguese)) and Treatment (FFE + lit (Portuguese)) at baseline was thus of 0.00873099
percentage points. The p-value for this difference was 0.2690187. The percentage correct for the Control
(Comparison (Portuguese)) condition at midline was 0.4485597 and the percentage correct for the Treatment
(FFE + lit (Portuguese)) condition at midline was 0.4297994. The difference between Control (Comparison
(Portuguese)) and Treatment (FFE + lit (Portuguese)) at midline was thus of 0.01876024 points. The p￾value for this difference was 0.5658451. The change from the baseline to the midline of 0.4370984 points
for the Control (Comparison (Portuguese)) condition establishes the conterfactual against which the change
for the Treatment (FFE + lit (Portuguese)) can be assessed against. The change for the Treatment (FFE
+ lit (Portuguese)) from the baseline to the midline was of 0.4096071 points. Consequently, the change for
the Control (Comparison (Portuguese)) relative to the change for the Treatment (FFE + lit (Portuguese))
condition was of -0.02749123 points. The p-value for this difference was 0.2054075. Accordingly, the difference
was not statistically significant (p < 0.01 corrected for 5 additional contrasts) and therefore the statistical
test provides no conclusive evidence in regards to there being a difference in the rate of change between
169
the baseline and the midline for the Writing name overall EGRA subtask across the Control (Comparison
(Portuguese)) and the Treatment (FFE + lit (Portuguese)) conditions beyond that expected from sampling
error.
2.3.3 Contrast 3: FFE + lit (Bilingual) vs Comparison (Bilingual)
Table 88: Writing name overall
Baseline Midline
Intervention % n % n
Control 0.000 89 0.220 109
Treatment 0.009 538 0.397 761
0.0
0.1
0.2
0.3
0.4
Control Treatment
CINTERVENTION
Mean
PERIOD
Baseline
Midline
Writing name overall
As shown in the table above, for the the Writing name overall EGRA subtask, the percentage correct for the
Control (Comparison (Bilingual)) condition at baseline was 0 and the percentage correct for the Treatment
(FFE + lit (Bilingual)) condition at baseline was 0.00929368. The difference between Control (Comparison
(Bilingual)) and Treatment (FFE + lit (Bilingual)) at baseline was thus of 0.00929368 percentage points.
The p-value for this difference was 3.017859e-138. The percentage correct for the Control (Comparison
(Bilingual)) condition at midline was 0.2201835 and the percentage correct for the Treatment (FFE + lit
(Bilingual)) condition at midline was 0.3968463. The difference between Control (Comparison (Bilingual))
and Treatment (FFE + lit (Bilingual)) at midline was thus of 0.1766628 points. The p-value for this difference
was 0.03840806. The change from the baseline to the midline of 0.2201835 points for the Control (Comparison
(Bilingual)) condition establishes the conterfactual against which the change for the Treatment (FFE + lit
(Bilingual)) can be assessed against. The change for the Treatment (FFE + lit (Bilingual)) from the baseline
to the midline was of 0.3875526 points. Consequently, the change for the Control (Comparison (Bilingual))
170
relative to the change for the Treatment (FFE + lit (Bilingual)) condition was of 0.1673691 points. The
p-value for this difference was 9.558092e-53. Accordingly, the difference was statistically significant (p <
0.01 corrected for 4 additional contrasts) and thus there is evidence of a different degree of change from
baseline to midline in the Treatment (FFE + lit (Bilingual)) when compared to the Control (Comparison
(Bilingual)) condition. This provides evidence that the Treatment (FFE + lit (Bilingual)) (or some other
unobserved process) impacted on the Writing name overall EGRA subtask and thus the rate of change of
the two conditions relative to their baselines differed.
2.3.4 Contrast 4: FFE + lit (Portuguese) vs FFE only (Portuguese)
Table 89: Writing name overall
Baseline Midline
Intervention % n % n
Control 0.01 1007 0.282 952
Treatment 0.02 1040 0.430 1047
0.0
0.1
0.2
0.3
0.4
Control Treatment
CINTERVENTION
Mean
PERIOD
Baseline
Midline
Writing name overall
As shown in the table above, for the the Writing name overall EGRA subtask, the percentage correct for
the Control (FFE only (Portuguese)) condition at baseline was 0.009930487 and the percentage correct
for the Treatment (FFE + lit (Portuguese)) condition at baseline was 0.02019231. The difference between
Control (FFE only (Portuguese)) and Treatment (FFE + lit (Portuguese)) at baseline was thus of 0.01026182
percentage points. The p-value for this difference was 0.1579548. The percentage correct for the Control
(FFE only (Portuguese)) condition at midline was 0.2815126 and the percentage correct for the Treatment
(FFE + lit (Portuguese)) condition at midline was 0.4297994. The difference between Control (FFE only
171
(Portuguese)) and Treatment (FFE + lit (Portuguese)) at midline was thus of 0.1482868 points. The p￾value for this difference was 3.8535e-05. The change from the baseline to the midline of 0.2715821 points
for the Control (FFE only (Portuguese)) condition establishes the conterfactual against which the change
for the Treatment (FFE + lit (Portuguese)) can be assessed against. The change for the Treatment (FFE
+ lit (Portuguese)) from the baseline to the midline was of 0.4096071 points. Consequently, the change for
the Control (FFE only (Portuguese)) relative to the change for the Treatment (FFE + lit (Portuguese))
condition was of 0.138025 points. The p-value for this difference was 0.8948978. Accordingly, the difference
was not statistically significant (p < 0.01 corrected for 5 additional contrasts) and therefore the statistical
test provides no conclusive evidence in regards to there being a difference in the rate of change between
the baseline and the midline for the Writing name overall EGRA subtask across the Control (FFE only
(Portuguese)) and the Treatment (FFE + lit (Portuguese)) conditions beyond that expected from sampling
error.
2.3.5 Contrast 5: FFE + lit (Bilingual) vs FFE + lit (Portuguese)
Table 90: Writing name overall
Baseline Midline
Intervention % n % n
Control 0.020 1040 0.430 1047
Treatment 0.009 538 0.397 761
0.0
0.1
0.2
0.3
0.4
Control Treatment
CINTERVENTION
Mean
PERIOD
Baseline
Midline
Writing name overall
As shown in the table above, for the the Writing name overall EGRA subtask, the percentage correct for
the Control (FFE + lit (Portuguese)) condition at baseline was 0.02019231 and the percentage correct
for the Treatment (FFE + lit (Bilingual)) condition at baseline was 0.00929368. The difference between
172
Control (FFE + lit (Portuguese)) and Treatment (FFE + lit (Bilingual)) at baseline was thus of 0.01089863
percentage points. The p-value for this difference was 0.2044722. The percentage correct for the Control
(FFE + lit (Portuguese)) condition at midline was 0.4297994 and the percentage correct for the Treatment
(FFE + lit (Bilingual)) condition at midline was 0.3968463. The difference between Control (FFE + lit
(Portuguese)) and Treatment (FFE + lit (Bilingual)) at midline was thus of 0.03295317 points. The p￾value for this difference was 0.361577. The change from the baseline to the midline of 0.4096071 points
for the Control (FFE + lit (Portuguese)) condition establishes the conterfactual against which the change
for the Treatment (FFE + lit (Bilingual)) can be assessed against. The change for the Treatment (FFE
+ lit (Bilingual)) from the baseline to the midline was of 0.3875526 points. Consequently, the change
for the Control (FFE + lit (Portuguese)) relative to the change for the Treatment (FFE + lit (Bilingual))
condition was of -0.02205454 points. The p-value for this difference was 0.3028833. Accordingly, the difference
was not statistically significant (p < 0.01 corrected for 5 additional contrasts) and therefore the statistical
test provides no conclusive evidence in regards to there being a difference in the rate of change between
the baseline and the midline for the Writing name overall EGRA subtask across the Control (FFE + lit
(Portuguese)) and the Treatment (FFE + lit (Bilingual)) conditions beyond that expected from sampling
error.
3 Literacy Benchmark
3.1 Literacy Benchmark: Boys and Girls
3.1.1 Contrast 1: FFE + lit (all) vs Comparison (all)
Table 91: Literacy Benchmark
Midline
Intervention % n
Control 8.333 1080
Treatment 6.696 1792
173
0
2
4
6
8
Control Treatment
CINTERVENTION
%
Literacy Benchmark
As shown in the table above, for the Literacy Benchmark, the percentage of students that achieved the
benchmark for the Control (Comparison (all)) condition at midline was 8.333333 and the percentage achieving
it for the Treatment (FFE + lit (all)) condition at midline was 6.696429. The difference for the Literacy
Benchmark across the Control (Comparison (all)) and Treatment (FFE + lit (all)) conditions at midline was
thus -1.636905 points (there was no baseline measurement for this variable). The p-value for this difference
was 0.2983881. Accordingly, the difference was not statistically significant and therefore, the statistical test
provides no conclusive evidence in regards to the there being a difference in the percentage achieving the
Literacy Benchmark between the Control (Comparison (all)) and Treatment (FFE + lit (all)) conditions at
midline beyond that which could result from sampling error.
3.1.2 Contrast 2: FFE + lit (Portuguese) vs Comparison (Portuguese)
Table 92: Literacy Benchmark
Midline
Intervention % n
Control 9.053 972
Treatment 7.811 1037
174
0.0
2.5
5.0
7.5
Control Treatment
CINTERVENTION
%
Literacy Benchmark
As shown in the table above, for the Literacy Benchmark, the percentage of students that achieved the
benchmark for the Control (Comparison (Portuguese)) condition at midline was 9.053498 and the per￾centage achieving it for the Treatment (FFE + lit (Portuguese)) condition at midline was 7.810993. The
difference for the Literacy Benchmark across the Control (Comparison (Portuguese)) and Treatment (FFE
+ lit (Portuguese)) conditions at midline was thus -1.242505 points (there was no baseline measurement for
this variable). The p-value for this difference was 0.511773. Accordingly, the difference was not statistically
significant and therefore, the statistical test provides no conclusive evidence in regards to the there being
a difference in the percentage achieving the Literacy Benchmark between the Control (Comparison (Por￾tuguese)) and Treatment (FFE + lit (Portuguese)) conditions at midline beyond that which could result
from sampling error.
3.1.3 Contrast 3: FFE + lit (Bilingual) vs Comparison (Bilingual)
Table 93: Literacy Benchmark
Midline
Intervention % n
Control 1.852 108
Treatment 5.166 755
175
0
1
2
3
4
5
Control Treatment
CINTERVENTION
%
Literacy Benchmark
As shown in the table above, for the Literacy Benchmark, the percentage of students that achieved the
benchmark for the Control (Comparison (Bilingual)) condition at midline was 1.851852 and the percentage
achieving it for the Treatment (FFE + lit (Bilingual)) condition at midline was 5.165563. The difference for
the Literacy Benchmark across the Control (Comparison (Bilingual)) and Treatment (FFE + lit (Bilingual))
conditions at midline was thus 3.313711 points (there was no baseline measurement for this variable). The
p-value for this difference was 0.2811655. Accordingly, the difference was not statistically significant and
therefore, the statistical test provides no conclusive evidence in regards to the there being a difference in the
percentage achieving the Literacy Benchmark between the Control (Comparison (Bilingual)) and Treatment
(FFE + lit (Bilingual)) conditions at midline beyond that which could result from sampling error.
3.1.4 Contrast 4: FFE + lit (Portuguese) vs FFE only (Portuguese)
Table 94: Literacy Benchmark
Midline
Intervention % n
Control 4.522 951
Treatment 7.811 1037
176
0
2
4
6
8
Control Treatment
CINTERVENTION
%
Literacy Benchmark
As shown in the table above, for the Literacy Benchmark, the percentage of students that achieved the
benchmark for the Control (FFE only (Portuguese)) condition at midline was 4.521556 and the percentage
achieving it for the Treatment (FFE + lit (Portuguese)) condition at midline was 7.810993. The difference for
the Literacy Benchmark across the Control (FFE only (Portuguese)) and Treatment (FFE + lit (Portuguese))
conditions at midline was thus 3.289437 points (there was no baseline measurement for this variable). The
p-value for this difference was 0.06589293. Accordingly, the difference was not statistically significant and
therefore, the statistical test provides no conclusive evidence in regards to the there being a difference in the
percentage achieving the Literacy Benchmark between the Control (FFE only (Portuguese)) and Treatment
(FFE + lit (Portuguese)) conditions at midline beyond that which could result from sampling error.
3.1.5 Contrast 5: FFE + lit (Bilingual) vs FFE + lit (Portuguese)
Table 95: Literacy Benchmark
Midline
Intervention % n
Control 7.811 1037
Treatment 5.166 755
177
0
2
4
6
8
Control Treatment
CINTERVENTION
%
Literacy Benchmark
As shown in the table above, for the Literacy Benchmark, the percentage of students that achieved the
benchmark for the Control (FFE + lit (Portuguese)) condition at midline was 7.810993 and the percentage
achieving it for the Treatment (FFE + lit (Bilingual)) condition at midline was 5.165563. The difference for
the Literacy Benchmark across the Control (FFE + lit (Portuguese)) and Treatment (FFE + lit (Bilingual))
conditions at midline was thus -2.64543 points (there was no baseline measurement for this variable). The
p-value for this difference was 0.1593429. Accordingly, the difference was not statistically significant and
therefore, the statistical test provides no conclusive evidence in regards to the there being a difference in the
percentage achieving the Literacy Benchmark between the Control (FFE + lit (Portuguese)) and Treatment
(FFE + lit (Bilingual)) conditions at midline beyond that which could result from sampling error.
3.2 Literacy Benchmark: Boys only
3.2.1 Contrast 1: FFE + lit (all) vs Comparison (all)
Table 96: Literacy Benchmark
Midline
Intervention % n
Control 7.238 525
Treatment 5.850 906
178
0
2
4
6
Control Treatment
CINTERVENTION
%
Literacy Benchmark
As shown in the table above, for the Literacy Benchmark, the percentage of students that achieved the
benchmark for the Control (Comparison (all)) condition at midline was 7.238095 and the percentage achieving
it for the Treatment (FFE + lit (all)) condition at midline was 5.84989. The difference for the Literacy
Benchmark across the Control (Comparison (all)) and Treatment (FFE + lit (all)) conditions at midline was
thus -1.388206 points (there was no baseline measurement for this variable). The p-value for this difference
was 0.4108179. Accordingly, the difference was not statistically significant and therefore, the statistical test
provides no conclusive evidence in regards to the there being a difference in the percentage achieving the
Literacy Benchmark between the Control (Comparison (all)) and Treatment (FFE + lit (all)) conditions at
midline beyond that which could result from sampling error.
3.2.2 Contrast 2: FFE + lit (Portuguese) vs Comparison (Portuguese)
Table 97: Literacy Benchmark
Midline
Intervention % n
Control 7.889 469
Treatment 7.561 529
179
0
2
4
6
8
Control Treatment
CINTERVENTION
%
Literacy Benchmark
As shown in the table above, for the Literacy Benchmark, the percentage of students that achieved the
benchmark for the Control (Comparison (Portuguese)) condition at midline was 7.889126 and the per￾centage achieving it for the Treatment (FFE + lit (Portuguese)) condition at midline was 7.561437. The
difference for the Literacy Benchmark across the Control (Comparison (Portuguese)) and Treatment (FFE
+ lit (Portuguese)) conditions at midline was thus -0.3276891 points (there was no baseline measurement
for this variable). The p-value for this difference was 0.8773095. Accordingly, the difference was not statis￾tically significant and therefore, the statistical test provides no conclusive evidence in regards to the there
being a difference in the percentage achieving the Literacy Benchmark between the Control (Comparison
(Portuguese)) and Treatment (FFE + lit (Portuguese)) conditions at midline beyond that which could result
from sampling error.
3.2.3 Contrast 3: FFE + lit (Bilingual) vs Comparison (Bilingual)
Table 98: Literacy Benchmark
Midline
Intervention % n
Control 1.786 56
Treatment 3.448 377
180
0
1
2
3
Control Treatment
CINTERVENTION
%
Literacy Benchmark
As shown in the table above, for the Literacy Benchmark, the percentage of students that achieved the
benchmark for the Control (Comparison (Bilingual)) condition at midline was 1.785714 and the percentage
achieving it for the Treatment (FFE + lit (Bilingual)) condition at midline was 3.448276. The difference for
the Literacy Benchmark across the Control (Comparison (Bilingual)) and Treatment (FFE + lit (Bilingual))
conditions at midline was thus 1.662562 points (there was no baseline measurement for this variable). The
p-value for this difference was 0.5128991. Accordingly, the difference was not statistically significant and
therefore, the statistical test provides no conclusive evidence in regards to the there being a difference in the
percentage achieving the Literacy Benchmark between the Control (Comparison (Bilingual)) and Treatment
(FFE + lit (Bilingual)) conditions at midline beyond that which could result from sampling error.
3.2.4 Contrast 4: FFE + lit (Portuguese) vs FFE only (Portuguese)
Table 99: Literacy Benchmark
Midline
Intervention % n
Control 2.453 530
Treatment 7.561 529
181
0
2
4
6
Control Treatment
CINTERVENTION
%
Literacy Benchmark
As shown in the table above, for the Literacy Benchmark, the percentage of students that achieved the
benchmark for the Control (FFE only (Portuguese)) condition at midline was 2.45283 and the percentage
achieving it for the Treatment (FFE + lit (Portuguese)) condition at midline was 7.561437. The difference for
the Literacy Benchmark across the Control (FFE only (Portuguese)) and Treatment (FFE + lit (Portuguese))
conditions at midline was thus 5.108606 points (there was no baseline measurement for this variable). The
p-value for this difference was 0.004860343. Accordingly, the difference was statistically significant (p < 0.01
corrected for 4 additional contrasts). Nevertheless, due to the absence of baseline values and lack of random
allocation to the treatment and control conditions, a causal attribution of this effect to the Treatment (FFE
+ lit (Portuguese)) is not possible
3.2.5 Contrast 5: FFE + lit (Bilingual) vs FFE + lit (Portuguese)
Table 100: Literacy Benchmark
Midline
Intervention % n
Control 7.561 529
Treatment 3.448 377
182
0
2
4
6
Control Treatment
CINTERVENTION
%
Literacy Benchmark
As shown in the table above, for the Literacy Benchmark, the percentage of students that achieved the
benchmark for the Control (FFE + lit (Portuguese)) condition at midline was 7.561437 and the percentage
achieving it for the Treatment (FFE + lit (Bilingual)) condition at midline was 3.448276. The difference for
the Literacy Benchmark across the Control (FFE + lit (Portuguese)) and Treatment (FFE + lit (Bilingual))
conditions at midline was thus -4.113161 points (there was no baseline measurement for this variable). The
p-value for this difference was 0.04996474. Accordingly, the difference was not statistically significant and
therefore, the statistical test provides no conclusive evidence in regards to the there being a difference in the
percentage achieving the Literacy Benchmark between the Control (FFE + lit (Portuguese)) and Treatment
(FFE + lit (Bilingual)) conditions at midline beyond that which could result from sampling error.
3.3 Literacy Benchmark: Girls only
3.3.1 Contrast 1: FFE + lit (all) vs Comparison (all)
Table 101: Literacy Benchmark
Midline
Intervention % n
Control 9.369 555
Treatment 7.562 886
183
0.0
2.5
5.0
7.5
Control Treatment
CINTERVENTION
%
Literacy Benchmark
As shown in the table above, for the Literacy Benchmark, the percentage of students that achieved the
benchmark for the Control (Comparison (all)) condition at midline was 9.369369 and the percentage achieving
it for the Treatment (FFE + lit (all)) condition at midline was 7.562077. The difference for the Literacy
Benchmark across the Control (Comparison (all)) and Treatment (FFE + lit (all)) conditions at midline was
thus -1.807293 points (there was no baseline measurement for this variable). The p-value for this difference
was 0.3772846. Accordingly, the difference was not statistically significant and therefore, the statistical test
provides no conclusive evidence in regards to the there being a difference in the percentage achieving the
Literacy Benchmark between the Control (Comparison (all)) and Treatment (FFE + lit (all)) conditions at
midline beyond that which could result from sampling error.
3.3.2 Contrast 2: FFE + lit (Portuguese) vs Comparison (Portuguese)
Table 102: Literacy Benchmark
Midline
Intervention % n
Control 10.139 503
Treatment 8.071 508
184
0.0
2.5
5.0
7.5
10.0
Control Treatment
CINTERVENTION
%
Literacy Benchmark
As shown in the table above, for the Literacy Benchmark, the percentage of students that achieved the
benchmark for the Control (Comparison (Portuguese)) condition at midline was 10.13917 and the per￾centage achieving it for the Treatment (FFE + lit (Portuguese)) condition at midline was 8.070866. The
difference for the Literacy Benchmark across the Control (Comparison (Portuguese)) and Treatment (FFE
+ lit (Portuguese)) conditions at midline was thus -2.068299 points (there was no baseline measurement for
this variable). The p-value for this difference was 0.3848931. Accordingly, the difference was not statistically
significant and therefore, the statistical test provides no conclusive evidence in regards to the there being
a difference in the percentage achieving the Literacy Benchmark between the Control (Comparison (Por￾tuguese)) and Treatment (FFE + lit (Portuguese)) conditions at midline beyond that which could result
from sampling error.
3.3.3 Contrast 3: FFE + lit (Bilingual) vs Comparison (Bilingual)
Table 103: Literacy Benchmark
Midline
Intervention % n
Control 1.923 52
Treatment 6.878 378
185
0
2
4
6
Control Treatment
CINTERVENTION
%
Literacy Benchmark
As shown in the table above, for the Literacy Benchmark, the percentage of students that achieved the
benchmark for the Control (Comparison (Bilingual)) condition at midline was 1.923077 and the percentage
achieving it for the Treatment (FFE + lit (Bilingual)) condition at midline was 6.878307. The difference for
the Literacy Benchmark across the Control (Comparison (Bilingual)) and Treatment (FFE + lit (Bilingual))
conditions at midline was thus 4.95523 points (there was no baseline measurement for this variable). The
p-value for this difference was 0.1724789. Accordingly, the difference was not statistically significant and
therefore, the statistical test provides no conclusive evidence in regards to the there being a difference in the
percentage achieving the Literacy Benchmark between the Control (Comparison (Bilingual)) and Treatment
(FFE + lit (Bilingual)) conditions at midline beyond that which could result from sampling error.
3.3.4 Contrast 4: FFE + lit (Portuguese) vs FFE only (Portuguese)
Table 104: Literacy Benchmark
Midline
Intervention % n
Control 7.126 421
Treatment 8.071 508
186
0
2
4
6
8
Control Treatment
CINTERVENTION
%
Literacy Benchmark
As shown in the table above, for the Literacy Benchmark, the percentage of students that achieved the
benchmark for the Control (FFE only (Portuguese)) condition at midline was 7.125891 and the percentage
achieving it for the Treatment (FFE + lit (Portuguese)) condition at midline was 8.070866. The difference for
the Literacy Benchmark across the Control (FFE only (Portuguese)) and Treatment (FFE + lit (Portuguese))
conditions at midline was thus 0.9449754 points (there was no baseline measurement for this variable). The
p-value for this difference was 0.7122201. Accordingly, the difference was not statistically significant and
therefore, the statistical test provides no conclusive evidence in regards to the there being a difference in the
percentage achieving the Literacy Benchmark between the Control (FFE only (Portuguese)) and Treatment
(FFE + lit (Portuguese)) conditions at midline beyond that which could result from sampling error.
3.3.5 Contrast 5: FFE + lit (Bilingual) vs FFE + lit (Portuguese)
Table 105: Literacy Benchmark
Midline
Intervention % n
Control 8.071 508
Treatment 6.878 378
187
0
2
4
6
8
Control Treatment
CINTERVENTION
%
Literacy Benchmark
As shown in the table above, for the Literacy Benchmark, the percentage of students that achieved the
benchmark for the Control (FFE + lit (Portuguese)) condition at midline was 8.070866 and the percentage
achieving it for the Treatment (FFE + lit (Bilingual)) condition at midline was 6.878307. The difference for
the Literacy Benchmark across the Control (FFE + lit (Portuguese)) and Treatment (FFE + lit (Bilingual))
conditions at midline was thus -1.192559 points (there was no baseline measurement for this variable). The
p-value for this difference was 0.6024658. Accordingly, the difference was not statistically significant and
therefore, the statistical test provides no conclusive evidence in regards to the there being a difference in the
percentage achieving the Literacy Benchmark between the Control (FFE + lit (Portuguese)) and Treatment
(FFE + lit (Bilingual)) conditions at midline beyond that which could result from sampling error.
4 BMI
4.1 Height
Table 106: Height
Baseline Midline
Intervention Mean SD n Mean SD n
Class 1: FFE + literacy and FFE only 114 6.62 2991 126 7.83 2873
Class 1: Comparison 114 6.29 1159 126 8.75 1081
188
0
50
100
Class 1: FFE + literacy and FFE only Class 1: Comparison
INTERVENTION2
Mean
PERIOD
Baseline
Midline
Height
189
90
120
150
180
Class 1: FFE + literacy and FFE only Class 1: Comparison
INTERVENTION2
HEIGHT
PERIOD
Baseline
Midline
Height
For the Height the FFE + Literacy and FFE only conditions have been collapsed into a single contrast
to evaluate the impact of both of the intervention conditions with a nutritional component against the
Comparison. As shown in the table above, for the the Height, the mean for the Comparison condition at
baseline was 114 (SD = 6.29) and the mean for the FFE + Literacy and FFE only condition at baseline was
114 (SD = 6.62). The difference between the Comparison and FFE + Literacy and FFE only at baseline
was thus of 0.0107 points. The p-value for this difference was 0.98. The mean for the Comparison condition
at midline was 126 (SD = 8.75) and the mean for the FFE + Literacy condition and FFE only at midline
was 126 (SD = 7.84). The difference between the Comparison and FFE + Literacy and FFE only at midline
was thus of 0.586 points. The p-value for this difference was 0.187. The change from the baseline to the
midline of 117 points for the Comparison condition establishes the conterfactual against which the change
for the FFE + Literacy and FFE only can be assessed against. The change for the FFE + Literacy and
FFE only from the baseline to the midline was of 11.9 points. Consequently, the change for the Comparison
relative to the change for the FFE + Literacy and FFE only condition was of 0.575 points. The p-value for
this difference was 0.288. Accordingly, the difference was not statistically significant (p < 0.017 corrected
for 3 additional comparisons) and therefore, the statistical test provides no conclusive evidence in regards to
there being a difference in the rate of change between the baseline and the midline for the Height across the
Comparison and the FFE + Literacy and FFE only conditions beyond that expected from sampling error.
All statistical tests accounted for the schools’ intraclass correlation effect on the outcome.
190
4.2 Weight
Table 107: Weight
Baseline Midline
Intervention Mean SD n Mean SD n
Class 1: FFE + literacy and FFE only 19.8 3.39 2991 24.5 4.52 2873
Class 1: Comparison 20.0 3.62 1159 24.2 5.10 1083
0
5
10
15
20
25
Class 1: FFE + literacy and FFE only Class 1: Comparison
INTERVENTION2
Mean
PERIOD
Baseline
Midline
Weight
191
20
40
60
Class 1: FFE + literacy and FFE only Class 1: Comparison
INTERVENTION2
WEIGHT
PERIOD
Baseline
Midline
Weight
For the Weight the FFE + Literacy and FFE only conditions have been collapsed into a single contrast
to evaluate the impact of both of the intervention conditions with a nutritional component against the
Comparison. As shown in the table above, for the the Weight, the mean for the Comparison condition at
baseline was 20 (SD = 3.62) and the mean for the FFE + Literacy and FFE only condition at baseline was
19.8 (SD = 3.39). The difference between the Comparison and FFE + Literacy and FFE only at baseline
was thus of 0.194 points. The p-value for this difference was 0.37. The mean for the Comparison condition
at midline was 24.2 (SD = 5.1) and the mean for the FFE + Literacy condition and FFE only at midline was
24.5 (SD = 4.52). The difference between the Comparison and FFE + Literacy and FFE only at midline was
thus of 0.343 points. The p-value for this difference was 0.17. The change from the baseline to the midline of
19.1 points for the Comparison condition establishes the conterfactual against which the change for the FFE
+ Literacy and FFE only can be assessed against. The change for the FFE + Literacy and FFE only from
the baseline to the midline was of 4.76 points. Consequently, the change for the Comparison relative to the
change for the FFE + Literacy and FFE only condition was of 0.537 points. The p-value for this difference
was 0.0861. Accordingly, the difference was not statistically significant (p < 0.017 corrected for 3 additional
comparisons) and therefore, the statistical test provides no conclusive evidence in regards to there being a
difference in the rate of change between the baseline and the midline for the Weight across the Comparison
and the FFE + Literacy and FFE only conditions beyond that expected from sampling error. All statistical
tests accounted for the schools’ intraclass correlation effect on the outcome.
192
4.3 BMI Z Score
Table 108: BMI Z Score
Baseline Midline
Intervention Mean SD n Mean SD n
Class 1: FFE + literacy and FFE only -0.437 1.72 2951 -0.514 1.24 1891
Class 1: Comparison -0.416 2.15 1150 -0.675 1.92 965
−0.6
−0.4
−0.2
0.0
Class 1: FFE + literacy and FFE only Class 1: Comparison
INTERVENTION2
Mean
PERIOD
Baseline
Midline
BMI Z Score
193
−10
0
Class 1: FFE + literacy and FFE only Class 1: Comparison
INTERVENTION2
BMI_Z
PERIOD
Baseline
Midline
BMI Z Score
For the BMI Z Score the FFE + Literacy and FFE only conditions have been collapsed into a single contrast
to evaluate the impact of both of the intervention conditions with a nutritional component against the
Comparison. As shown in the table above, for the the BMI Z Score, the mean for the Comparison condition
at baseline was -0.416 (SD = 2.15) and the mean for the FFE + Literacy and FFE only condition at baseline
was -0.437 (SD = 1.72). The difference between the Comparison and FFE + Literacy and FFE only at
baseline was thus of 0.0214 points. The p-value for this difference was 0.899. The mean for the Comparison
condition at midline was -0.675 (SD = 1.92) and the mean for the FFE + Literacy condition and FFE only
at midline was -0.514 (SD = 1.24). The difference between the Comparison and FFE + Literacy and FFE
only at midline was thus of 0.162 points. The p-value for this difference was 0.164. The change from the
baseline to the midline of -2.59 points for the Comparison condition establishes the conterfactual against
which the change for the FFE + Literacy and FFE only can be assessed against. The change for the FFE
+ Literacy and FFE only from the baseline to the midline was of -0.0763 points. Consequently, the change
for the Comparison relative to the change for the FFE + Literacy and FFE only condition was of 0.183
points. The p-value for this difference was 0.367. Accordingly, the difference was not statistically significant
(p < 0.017 corrected for 3 additional comparisons) and therefore, the statistical test provides no conclusive
evidence in regards to there being a difference in the rate of change between the baseline and the midline
for the BMI Z Score across the Comparison and the FFE + Literacy and FFE only conditions beyond that
expected from sampling error. All statistical tests accounted for the schools’ intraclass correlation effect on
the outcome.
194
5 Questions regarding hunger, attention and hygiene
5.1 HUNGER: Hungry in school
Table 109: Hungry in school
Baseline Midline
Intervention n % SD n % SD
FFE + literacy and FFE only 2956 59.6 9.03 2872 41.9 9.21
Comparison 1121 68.2 13.90 1083 58.3 14.98
0
20
40
60
FFE + literacy and FFE only Comparison
INTERVENTION
%
PERIOD
Baseline
Midline
Hungry in school
For the Hungry in school significance test, the FFE + Literacy and FFE only have been combined and
contrasted against the Comparison so that the impact of the interventions with a food and hygiene component
could be compared against the Comparison which received neither. As shown in the table above, for the the
Hungry in school, the percentage of success for the Comparison condition at baseline was 68.2 (SD = 13.9)
and the percentage for the FFE + Literacy and FFE only condition at baseline was 59.6 (SD = 9.03). The
difference between Comparison and FFE + Literacy and FFE only at baseline was thus of 8.69 percentage
points. The p-value for this difference was 0.202. The percentage for the Comparison condition at midline
was 58.3 (SD = 15) and the percentage for the FFE + Literacy and FFE only condition at midline was
41.9 (SD = 9.21). The difference between Comparison and FFE + Literacy and FFE only at midline was
thus of 16.4 points. The p-value for this difference was 1.53e-13. The change in percentage of success for
the Comparison condition from baseline to midline was of -9.91 points and establishes the conterfactual
against which the change for the FFE + Literacy and FFE only is assessed. The percentage change for
the FFE + Literacy and FFE only condition from baseline to midline was of -17.6 points. Consequently,
the change for the Comparison relative to the change for the FFE + Literacy and FFE only condition
195
amounted to -7.74 percentage points. The p-value for this difference was 0.354. Accordingly, the difference
was not statistically significant and therefore the statistical test provides no conclusive evidence in regards to
there being a difference in the rate of change between the baseline and the midline for the Hungry in school
outcome across the Comparison and the FFE + Literacy and FFE only conditions beyond that expected from
sampling error. All statistical tests accounted for the schools’ intraclass correlation effect on the outcome.
5.2 HUNGERNOW: Hungry now
Table 110: Hungry now
Baseline Midline
Intervention n % SD n % SD
FFE + literacy and FFE only 2956 42.1 9.08 2872 22.9 7.84
Comparison 1121 65.5 14.20 1083 32.6 14.25
0
20
40
60
FFE + literacy and FFE only Comparison
INTERVENTION
%
PERIOD
Baseline
Midline
Hungry now
For the Hungry now significance test, the FFE + Literacy and FFE only have been combined and contrasted
against the Comparison so that the impact of the interventions with a food and hygiene component could be
compared against the Comparison which received neither. As shown in the table above, for the the Hungry
now, the percentage of success for the Comparison condition at baseline was 65.5 (SD = 14.2) and the
percentage for the FFE + Literacy and FFE only condition at baseline was 42.1 (SD = 9.08). The difference
between Comparison and FFE + Literacy and FFE only at baseline was thus of 23.4 percentage points.
The p-value for this difference was 0.00057. The percentage for the Comparison condition at midline was
32.6 (SD = 14.2) and the percentage for the FFE + Literacy and FFE only condition at midline was 22.9
(SD = 7.84). The difference between Comparison and FFE + Literacy and FFE only at midline was thus
of 9.7 points. The p-value for this difference was 0.00000241. The change in percentage of success for the
196
Comparison condition from baseline to midline was of -32.9 points and establishes the conterfactual against
which the change for the FFE + Literacy and FFE only is assessed. The percentage change for the FFE +
Literacy and FFE only condition from baseline to midline was of -19.2 points. Consequently, the change for
the Comparison relative to the change for the FFE + Literacy and FFE only condition amounted to 13.7
percentage points. The p-value for this difference was 0.111. Accordingly, the difference was not statistically
significant and therefore the statistical test provides no conclusive evidence in regards to there being a
difference in the rate of change between the baseline and the midline for the Hungry now outcome across the
Comparison and the FFE + Literacy and FFE only conditions beyond that expected from sampling error.
All statistical tests accounted for the schools’ intraclass correlation effect on the outcome.
5.3 ATTENTION: Attention problems
Table 111: Attention problems
Baseline Midline
Intervention n % SD n % SD
FFE + literacy and FFE only 2956 37.9 8.92 2872 42.2 9.21
Comparison 1121 36.1 14.34 1083 48.3 15.19
0
10
20
30
40
50
FFE + literacy and FFE only Comparison
INTERVENTION
%
PERIOD
Baseline
Midline
Attention problems
For the Attention problems significance test, the FFE + Literacy and FFE only have been combined and
contrasted against the Comparison so that the impact of the interventions with a food and hygiene component
could be compared against the Comparison which received neither. As shown in the table above, for the
the Attention problems, the percentage of success for the Comparison condition at baseline was 36.1 (SD
= 14.3) and the percentage for the FFE + Literacy and FFE only condition at baseline was 37.9 (SD =
8.92). The difference between Comparison and FFE + Literacy and FFE only at baseline was thus of 1.84
197
percentage points. The p-value for this difference was 0.651. The percentage for the Comparison condition at
midline was 48.3 (SD = 15.2) and the percentage for the FFE + Literacy and FFE only condition at midline
was 42.2 (SD = 9.21). The difference between Comparison and FFE + Literacy and FFE only at midline
was thus of 6.15 points. The p-value for this difference was 0.0263. The change in percentage of success
for the Comparison condition from baseline to midline was of 12.2 points and establishes the conterfactual
against which the change for the FFE + Literacy and FFE only is assessed. The percentage change for the
FFE + Literacy and FFE only condition from baseline to midline was of 4.25 points. Consequently, the
change for the Comparison relative to the change for the FFE + Literacy and FFE only condition amounted
to -7.99 percentage points. The p-value for this difference was 0.113. Accordingly, the difference was not
statistically significant and therefore the statistical test provides no conclusive evidence in regards to there
being a difference in the rate of change between the baseline and the midline for the Attention problems
outcome across the Comparison and the FFE + Literacy and FFE only conditions beyond that expected from
sampling error. All statistical tests accounted for the schools’ intraclass correlation effect on the outcome.
5.4 TOILET: Wash after toilet
Table 112: Wash after toilet
Baseline Midline
Intervention n % SD n % SD
FFE + literacy and FFE only 2956 83.8 6.78 2872 81.1 7.3
Comparison 1121 66.0 14.15 1083 76.5 12.9
0
20
40
60
80
FFE + literacy and FFE only Comparison
INTERVENTION
%
PERIOD
Baseline
Midline
Wash after toilet
For the Wash after toilet significance test, the FFE + Literacy and FFE only have been combined and
contrasted against the Comparison so that the impact of the interventions with a food and hygiene component
198
could be compared against the Comparison which received neither. As shown in the table above, for the the
Wash after toilet, the percentage of success for the Comparison condition at baseline was 66 (SD = 14.2)
and the percentage for the FFE + Literacy and FFE only condition at baseline was 83.8 (SD = 6.78). The
difference between Comparison and FFE + Literacy and FFE only at baseline was thus of 17.8 percentage
points. The p-value for this difference was 1.41e-08. The percentage for the Comparison condition at midline
was 76.5 (SD = 12.9) and the percentage for the FFE + Literacy and FFE only condition at midline was
81.1 (SD = 7.3). The difference between Comparison and FFE + Literacy and FFE only at midline was
thus of 4.65 points. The p-value for this difference was 0.028. The change in percentage of success for the
Comparison condition from baseline to midline was of 10.5 points and establishes the conterfactual against
which the change for the FFE + Literacy and FFE only is assessed. The percentage change for the FFE +
Literacy and FFE only condition from baseline to midline was of -2.65 points. Consequently, the change for
the Comparison relative to the change for the FFE + Literacy and FFE only condition amounted to -13.1
percentage points. The p-value for this difference was 0.000602. Accordingly, the difference was statistically
significant and thus there is evidence of a different degree of change from baseline to midline in the FFE
+ Literacy and FFE only when compared to the Comparison condition. This provides evidence that the
intervention (or some other unobserved process) impacted on the Wash after toilet outcome and thus the
rate of change of the two conditions relative to their baselines differed. All statistical tests accounted for the
schools’ intraclass correlation effect on the outcome.
5.5 EATING: Wash before eating
Table 113: Wash before eating
Baseline Midline
Intervention n % SD n % SD
FFE + literacy and FFE only 2956 84.8 6.61 2872 89.1 5.81
Comparison 1121 70.6 13.61 1083 83.9 11.16
199
0
25
50
75
FFE + literacy and FFE only Comparison
INTERVENTION
%
PERIOD
Baseline
Midline
Wash before eating
For the Wash before eating significance test, the FFE + Literacy and FFE only have been combined and
contrasted against the Comparison so that the impact of the interventions with a food and hygiene component
could be compared against the Comparison which received neither. As shown in the table above, for the the
Wash before eating, the percentage of success for the Comparison condition at baseline was 70.6 (SD = 13.6)
and the percentage for the FFE + Literacy and FFE only condition at baseline was 84.8 (SD = 6.61). The
difference between Comparison and FFE + Literacy and FFE only at baseline was thus of 14.2 percentage
points. The p-value for this difference was 0.00000499. The percentage for the Comparison condition at
midline was 83.9 (SD = 11.2) and the percentage for the FFE + Literacy and FFE only condition at midline
was 89.1 (SD = 5.81). The difference between Comparison and FFE + Literacy and FFE only at midline
was thus of 5.17 points. The p-value for this difference was 0.00172. The change in percentage of success
for the Comparison condition from baseline to midline was of 13.3 points and establishes the conterfactual
against which the change for the FFE + Literacy and FFE only is assessed. The percentage change for the
FFE + Literacy and FFE only condition from baseline to midline was of 4.35 points. Consequently, the
change for the Comparison relative to the change for the FFE + Literacy and FFE only condition amounted
to -8.99 percentage points. The p-value for this difference was 0.0793. Accordingly, the difference was not
statistically significant and therefore the statistical test provides no conclusive evidence in regards to there
being a difference in the rate of change between the baseline and the midline for the Wash before eating
outcome across the Comparison and the FFE + Literacy and FFE only conditions beyond that expected from
sampling error. All statistical tests accounted for the schools’ intraclass correlation effect on the outcome.
200
6 Teacher questions regarding hunger, attention and difficulty in
learning
6.1 Students are hungry
Table 114: Students are hungry
Baseline Midline
Often 7.3 14.9
Sometimes 69.9 52.5
Never 22.8 32.6
Table 115: Students are hungry
Baseline Midline
Often 33 27
Sometimes 316 95
Never 103 59
0
20
40
60
Often Sometimes Never
Hungry
%
Period
Baseline
Midline
Students are hungry
As can be seen in the table above, in regards to pupils being hungry at baseline , 22.8 of the teacher
responded ‘Never’, 69.9 responded ‘Sometimes’ and 7.3 responded ‘Often.’ At midline,32.6 of the teacher
responded ‘Never’, 52.5 responded ‘Sometimes’ and 14.9 responded ‘Often.’ A hypothesis test for differeces
in ordinal distribution across period (baseline vs midline) did not yield a significant result (p = 0.335) and
thus there is no evidence that the distribution of responses differed across period.
201
6.2 Students have attention problems
Table 116: Students have attention prolems
Baseline Midline
Often 2.42 9.94
Sometimes 86.34 68.51
Never 11.23 21.55
Table 117: Students have attention prolems
Baseline Midline
Often 11 18
Sometimes 392 124
Never 51 39
0
25
50
75
Often Sometimes Never
Attention
%
Period
Baseline
Midline
Students have attention problems
As can be seen in the table above, in regards to pupils being hungry at baseline , 11.2 of the teacher
responded ‘Never’, 86.3 responded ‘Sometimes’ and 2.42 responded ‘Often.’ At midline,21.6 of the teacher
responded ‘Never’, 68.5 responded ‘Sometimes’ and 9.94 responded ‘Often.’ A hypothesis test for differeces
in ordinal distribution across period (baseline vs midline) did not yield a significant result (p = 0.635) and
thus there is no evidence that the distribution of responses differed across period.
202
6.3 Students have learning problems
Table 118: Students have learning problems
Baseline Midline
Often 1.34 13.3
Sometimes 82.59 71.3
Never 16.07 15.5
Table 119: Students have learning problems
Baseline Midline
Often 6 24
Sometimes 370 129
Never 72 28
0
20
40
60
80
Often Sometimes Never
Learning
%
Period
Baseline
Midline
Students have learning problems
As can be seen in the table above, in regards to pupils having problems to learn at baseline , 16.1 of the
teacher responded ‘Never’, 82.6 responded ‘Sometimes’ and 1.34 responded ‘Often.’ At midline,15.5 of the
teacher responded ‘Never’, 71.3 responded ‘Sometimes’ and 13.3 responded ‘Often.’ A hypothesis test for
differeces in ordinal distribution across period (baseline vs midline) yielded a significant result (p = 0.0172).
This provides evidence that the difference in the distribution of percentages across baseline and midline
could not be the result of sampling error.
203