endline evaluation · cbo community-based organization cba cost-benefit analysis cea...

61
April 18, 2016 ENDLINE EVALUATION UNICEF KENYA’S IMPLEMENTATION OF INTEGRATED INTERVENTIONS UNDER THE CHILD FRIENDLY SCHOOLS FRAMEWORK IN ASAL COUNTIES

Upload: others

Post on 19-Apr-2020

1 views

Category:

Documents


0 download

TRANSCRIPT

Page 1: ENDLINE EVALUATION · CBO Community-Based Organization CBA Cost-benefit analysis CEA Cost-effectiveness analysis ... consistent with the approach that is promoted by the intervention,

April 18, 2016

ENDLINE EVALUATION

UNICEF KENYA’S IMPLEMENTATION OF INTEGRATED

INTERVENTIONS UNDER THE CHILD FRIENDLY SCHOOLS

FRAMEWORK IN ASAL COUNTIES

Page 2: ENDLINE EVALUATION · CBO Community-Based Organization CBA Cost-benefit analysis CEA Cost-effectiveness analysis ... consistent with the approach that is promoted by the intervention,

2

ACKNOWLEDGEMENTS

This report was prepared by the FHI 360 Education Policy and Data Center, authored by Audrey Moore,

Carina Omoeva, Charles Gale, and Wael Moussa. Peter Mwarogo and Racheal Kamau managed the field

component of the study. The evaluation was conducted by FHI 360, and commissioned by UNICEF with

funding and technical oversight from the UK Department for International Development. The project was

monitored by Daniel Baheta, Charles Olaka Kesa, and Shweta Sandilya of UNICEF Kenya Country

Office, and Sandra Barton and Caroline Wangeci of DfID provided technical review of the design and

analysis.

Page 3: ENDLINE EVALUATION · CBO Community-Based Organization CBA Cost-benefit analysis CEA Cost-effectiveness analysis ... consistent with the approach that is promoted by the intervention,

3

TABLE OF CONTENTS Table of contents ........................................................................................................................................... 3

List of Tables ................................................................................................................................................ 5

List of Figures ............................................................................................................................................... 5

Abbreviations ................................................................................................................................................ 6

Executive Summary ...................................................................................................................................... 7

Introduction ................................................................................................................................................. 10

Background ............................................................................................................................................. 10

UNICEF Interventions in the ASAL Region .......................................................................................... 10

Methodology ............................................................................................................................................... 11

Key data sources ..................................................................................................................................... 11

Constructing the comparison group ........................................................................................................ 12

Descriptive analysis............................................................................................................................ 12

Regression analysis ............................................................................................................................ 13

Description of the study population ........................................................................................................ 15

Characteristics of the sample .................................................................................................................. 16

Results ......................................................................................................................................................... 16

Key Findings .......................................................................................................................................... 16

Overview of changes in key outcomes ................................................................................................... 18

Enrollment trends over the intervention period .................................................................................. 19

Attendance rate trend over the intervention period ............................................................................ 19

Changes in retention ........................................................................................................................... 20

Indicative results of the WASH Interventions ........................................................................................ 21

Status of WASH interventions ........................................................................................................... 21

How effective are the WASH interventions for outcomes of interest? .............................................. 21

WASH regression analysis ................................................................................................................. 23

Challenges in implementation and monitoring of WASH interventions. ........................................... 24

Findings from focus groups ................................................................................................................ 25

Indicative Results of Solar-Power Interventions .................................................................................... 27

Status of Solar interventions .............................................................................................................. 27

How effective are the solar interventions for outcomes of interest? .................................................. 27

Solar regression analysis .................................................................................................................... 29

Challenges in implementation and function of solar power ............................................................... 30

Findings from focus groups ................................................................................................................ 31

Indicative Results of Child-Friendly Schools Interventions ................................................................... 33

How effective are the CFS interventions for outcomes of interest? ................................................... 33

Page 4: ENDLINE EVALUATION · CBO Community-Based Organization CBA Cost-benefit analysis CEA Cost-effectiveness analysis ... consistent with the approach that is promoted by the intervention,

4

Classroom Observation - Background ............................................................................................... 34

Academic and non-academic activities .............................................................................................. 34

CFS regression analysis ..................................................................................................................... 36

Findings from Focus Groups .............................................................................................................. 37

Indicative Results of C4D Interventions ................................................................................................. 37

How effective are the C4D interventions for outcomes of interest? .................................................. 37

C4D regression analysis ..................................................................................................................... 40

Findings from Focus Groups .............................................................................................................. 41

Cost-Benefit Analysis ................................................................................................................................. 43

Assumptions ........................................................................................................................................... 43

Results of the benefit-cost analysis......................................................................................................... 45

Limitations of the Benefit-cost model .................................................................................................... 46

Concluding remarks for cost-benefit analyses ........................................................................................ 47

Discussion ................................................................................................................................................... 47

Understanding Dosage, Duration and Enabling Environment ........................................................... 47

Duration .............................................................................................................................................. 48

Dosage ................................................................................................................................................ 49

Enabling environment ........................................................................................................................ 50

References ................................................................................................................................................... 52

Annexes ...................................................................................................................................................... 54

Annex A: Supplemental Methodology Information ............................................................................... 54

Instruments and Key Data Sources ......................................................................................................... 55

Hypotheses to be analyzed in the observational study: .......................................................................... 56

Variables and analyses ............................................................................................................................ 57

Qualitative analysis ............................................................................................................................ 57

Changes to the data collection and analysis process from baseline to endline ....................................... 57

Annex B: Computation of Net Present Value in Cost-Benefit Analysis ................................................ 59

Annex C: Additional Focus Group Results ............................................................................................ 60

Page 5: ENDLINE EVALUATION · CBO Community-Based Organization CBA Cost-benefit analysis CEA Cost-effectiveness analysis ... consistent with the approach that is promoted by the intervention,

5

LIST OF TABLES Table 1 - Number of schools by type of intervention ................................................................................. 15 Table 2 - Enrollment changes over the intervention period ........................................................................ 19 Table 3 - Highlights of the classroom observation ..................................................................................... 35 Table 4 – Changes in enrollment, C4D status ............................................................................................. 38 Table 5 – Estimated enrollment benefits for WASH, Solar, and CFS ........................................................ 43 Table 6 – Grade promotion, dropout, and repetition rates – Kenya, 2014 .................................................. 44 Table 7 – Increased enrollment and grade attainment, by intervention ...................................................... 45 Table 8 – Costs per school, per student, and total, by intervention ............................................................ 45 Table 9 – Lifetime benefit-to-cost ratios per intervention, by discount rate ............................................... 46 Annex table 1 – Number of schools by reported treatment status .............................................................. 54 Annex table 2 – Benefits-to-cost ratios for WASH, Solar, and CFS .......................................................... 59

LIST OF FIGURES Figure 4 – Map of Kenya ............................................................................................................................ 10 Figure 5 - Propensity Score Distribution, by Treatment and Pre-Designated Comparison Groups ........... 13 Figure 6 - Propensity Score Distribution, by Treatment and Constructed Comparison Groups ................. 14 Figure 7 - Number of schools reached by county ....................................................................................... 16 Figure 8 - Pupil-teacher ratio ...................................................................................................................... 16 Figure 9 – Attendance rate changes over the intervention period ............................................................... 20 Figure 10 – Changes in retention over the intervention period ................................................................... 20 Figure 11 – WASH changes in key outcomes ............................................................................................ 22 Figure 12 – Head teacher assessment of WASH outcomes ........................................................................ 22 Figure 13 – Changes in availability of latrines ........................................................................................... 23 Figure 14 – Margins plot of WASH outcomes ........................................................................................... 24 Figure 15 - WASH challenges in implementation, as reported by Head Teachers ..................................... 25 Figure 16 – Solar: type of power ................................................................................................................ 27 Figure 17 – Solar: UNICEF and Non-UNICEF .......................................................................................... 27 Figure 18 – Solar: differences in evening study.......................................................................................... 28 Figure 19 – Hours of study ........................................................................................................................ 28 Figure 20 – Solar: head teacher assessment ................................................................................................ 29 Figure 21 - Margins plot of Solar outcomes ............................................................................................... 29 Figure 22 - Margins plot of Solar effect on frequency of studying ............................................................ 30 Figure 23 – Solar challenges in implementation ......................................................................................... 31 Figure 24 – CFS: Changes in key outcomes ............................................................................................... 33 Figure 25 – Number of schools in which classroom observations were conducted.................................... 34 Figure 26 – Percentage of class time devoted to academic activities ......................................................... 35 Figure 27 – Percentage of class time spent on nonacademic activities ....................................................... 35 Figure 28 – Margins plot of CFS outcomes ................................................................................................ 36 Figure 29 – Changes in attendance, C4D status .......................................................................................... 38 Figure 30 – Changes in promotion, C4D status .......................................................................................... 38 Figure 31 – Has received C4D training ...................................................................................................... 39 Figure 32 – OOSC communication plan ..................................................................................................... 39 Figure 33 – C4D: Engaged in advocacy or social mobilization .................................................................. 40 Figure 34 – C4D: Head teachers’ assessment ............................................................................................. 40 Figure 35 – Margins plots of C4D outcomes ............................................................................................. 41 Figure 36 – Age-earnings profile by educational attainment – Kenya, 2013 ............................................. 44 Figure 37 – Cost-Benefit simulation results – NPV of cumulative taxes paid over lifecycle ..................... 46

Page 6: ENDLINE EVALUATION · CBO Community-Based Organization CBA Cost-benefit analysis CEA Cost-effectiveness analysis ... consistent with the approach that is promoted by the intervention,

6

ABBREVIATIONS

ASAL Arid and Semi-Arid Lands

C4D Communication for Development

CBO Community-Based Organization

CBA Cost-benefit analysis

CEA Cost-effectiveness analysis

CFS Child Friendly Schools

DFID Department for International Development

DRR Disaster Risk Reduction

GOK Government of Kenya

LCPBS Low Cost Primary Boarding School

MOEST Ministry of Education Science and Technology

QED Quasi-experimental Design

RCT Randomized Control Trial

BoM School Board of Management

WASH Water Sanitation and Hygiene

Page 7: ENDLINE EVALUATION · CBO Community-Based Organization CBA Cost-benefit analysis CEA Cost-effectiveness analysis ... consistent with the approach that is promoted by the intervention,

7

EXECUTIVE SUMMARY Description of interventions. In 2013, UNICEF Kenya partnered with the Keya Ministry of Education

Science and Technology (MOEST) to launch a series of program interventions with the goal of

strengthening the quality of educational experience for children in the Kenyan Arid and Semi-Arid Lands

(ASAL). The United Kingdom’s Department for International Development (DFID) provided funding and

technical oversight for the program, which benefits an estimated 66,000 children (UNICEF, 2014). The

program consists of four discrete school-based interventions:

1) Water, sanitation, and hygiene (WASH) package, which involves installation of WASH

infrastructure and training;

2) Installation of solar panels to increase availability of electric lighting in classrooms, WASH

facilities, and dormitories;

3) Child-friendly schools (CFS) training and capacity development of head teachers, teachers and

BOM;

4) Communication for development (C4D) efforts, aimed at reducing dropout and encouraging school

enrollment among out-of-school children

Evaluation design. FHI 360 was contracted to carry out an external evaluation of the program in 2014. Due

to the nature of selection for interventions, and the prior presence of similar program interventions in the

target areas, a quasi-experimental methodology (matching on the propensity score) was selected as the

primary method for this evaluation. In addition, the study factored in the timing and duration of program

interventions, and was able to account for the differential effects of the interventions resulting from longer

duration. The study draws on data from 2,136 student interviews, 321 head teacher interviews, and 26

student/community focus group discussions, to examine program effects, implementation challenges, and

perceptions of the beneficiaries on program effectiveness. In addition, classroom observations were

conducted at 54 CFS and C4D intervention schools provided a gauge of time use and the application of

child-friendly instructional practices. The endline reached 321 out of 349 schools (92%) across all eight

counties, as some schools were dropped due to security concerns.

Results This evaluation showed mostly positive outcomes for schools supported by the program.

Enrollment gains were the greatest among schools that received WASH interventions. We exploit variation

in the reported exposure to the treatment, with some schools reporting having received an intervention prior

to the UNICEF ASAL program period (starting in June 2014), to identify strong and substantively large

positive enrollment effects of longer duration of interventions. This is true in the case of both WASH and

CFS interventions, that were reported present since prior to June 2014, while the Solar intervention was

implemented only after June 2014. Lastly, we find that WASH, Solar power, and CFS exhibit benefits that

exceed the costs of implementation with CFS being the most cost-effective. Brief summaries of results for

each treatment arm follow.

WASH – Both the descriptive and regression analyses

show that schools that implemented WASH the longest

(since prior to June 2014) exhibited the largest gains in

enrollment relative to all other schools. Specifically, we

estimate that enrollment among schools that

implemented WASH the longest increased by

approximately 20 percent relative to similar schools in

the constructed comparison group that never received

the WASH intervention. Attendance rates are also

highest among WASH schools with the longest

duration.

Page 8: ENDLINE EVALUATION · CBO Community-Based Organization CBA Cost-benefit analysis CEA Cost-effectiveness analysis ... consistent with the approach that is promoted by the intervention,

8

Solar power – Regression results show that enrollments are only modestly higher - by about 8.6 percent

among UNICEF Solar schools than similar schools in the constructed comparison group. These schools

also exhibited greater study time, which was one of the intended outcomes of the Solar intervention. All

else being equal, the estimated likelihood that a given student reports studying at least once a day is 13

percent higher among UNICEF Solar schools than similar schools that never received solar panels. The

effect of duration could not be estimated for the solar intervention, as it was limited to the period of program

performance.

CFS – Regression results suggest that enrollment growth among schools that implemented CFS prior to

June 2014 as well as those that implemented after June 2014 is 8.6 percent higher than the constructed

comparison group. However, the schools that reported having the CFS intervention, but were not supported

through the UNICEF/DFID program, that exhibited the largest gains in enrollment. Although, it is likely

that these differences are driven by differences in the level of implementation. Classroom observations

suggest that teachers in schools receiving the CFS intervention are more likely to exhibit characteristics

consistent with the approach that is promoted by the intervention, e.g. ask probing questions such as how

and why, ask students to explain in class discussions, and ask students to apply understanding to relevant

life experience. However, teachers in schools receiving the CFS intervention are slightly less likely to

integrate gender sensitive material into pedagogy.

C4D – The C4D intervention was largely implemented in conjunction with other interventions. Only six

schools report receiving only C4D from UNICEF, therefore, we are unable to isolate the impacts of C4D

from that of other interventions. Further, none of the schools report receiving C4D prior to June 2014. As

a result, we are unable to account for the duration of treatment effect. Overall, the descriptive analyses show

a positive effect on enrollment and attendance, but not grade promotion.

Combination Schools – Schools that receive two or more UNICEF interventions were more likely to show

positive effects on attendance and enrollment. Schools that received Solar power and C4D exhibited 10.5

percent higher enrollment relative to the constructed comparison group. The same combination of

interventions yielded a 46.5 percentage point increase in attendance rates and a 14.8 percent increase in the

likelihood of students studying more than once a day. Schools implementing C4D in conjunction with CFS

are estimated to have an attendance rate that is 28 percentage points higher than that of the constructed

comparison group. Lastly, schools that implemented both CFS and WASH also exhibited gains in

enrollment. Although, the gains achieved by schools with the CFS and WASH combination were relatively

smaller at 9.4 percent. Further, the estimated gain for the CFS and WASH combination was not statistically

significant.

Cost-Benefit Analysis – We compute the benefits of WASH, installation of solar lighting, and CFS

interventions as additional taxes paid over a lifetime as a result of increased educational attainment from

each intervention.

CFS yields the highest return on investment according to our simulation model. Initial investments

in CFS are returned in full after 4-5 years and the lifetime benefits-to-cost ratios for CFS range

between 4.0 and 12.5 per US Dollar invested. This means that an initial investment of 112,206 USD

yields approximately between USD 432,000 and USD 1,395,000 in increased tax revenues over 36

years.

WASH and Solar may be returned in full 10 to 21 years post-implementation depending on the

discount rate, and we calculate lifetime benefit-to-cost ratios between 1.2 and 4.2 per US Dollar

invested. This translates to an initial investment of USD 505,357 in WASH yields a return between

USD 602,000 and USD 1,944,000, and an initial investment of USD 111,667 in Solar panels

produces a return between USD 136,000 and USD 440,000 over 36 years.

Page 9: ENDLINE EVALUATION · CBO Community-Based Organization CBA Cost-benefit analysis CEA Cost-effectiveness analysis ... consistent with the approach that is promoted by the intervention,

9

Recommendations We offer the following conclusions and recommendations based on the findings.

Dosage and duration of the treatments play a significant role when determining intervention

impacts. In this report, we provide evidence that interventions that were implemented for a longer

duration exhibited larger effects on enrollment when compared to schools that implemented the

same intervention for a shorter duration. This is especially true with the WASH and CFS

interventions. This also supports the hypothesis, that certain programmatic effects require some

time to manifest themselves in education outcomes.

Further research is required to ascertain the level of program implementation fidelity in each

intervention. When examining the impact of the CFS or C4D interventions, we found that overall

results were either relatively small in magnitude and/or provided some evidence that similar

interventions supported by other organizations had larger gains. In both cases, the findings point

to a certain amount of untapped potential gains in enrollment that could be had. As such, we

recommend that context-specific elements of implementations should be investigated to

determine where and how certain practices could be improved from a programming standpoint.

For instance, some of the focus group discussion results pointed toward community members and

other stakeholders not being aware of the C4D intervention. Additional efforts into ensuring

community involvement and awareness could have possibly benefitted the schools more.

The investments in WASH, CFS, and Solar panels yield positive returns on investment over the

long-term and to varying degrees. It is important to be cognizant of the assumptions of the cost-

benefit analysis, especially with the discount rate, which acts as a proxy for the value of future

money. In addition, wage based returns are estimated and do not account for any unforeseen

shocks to the economy at least in the ASAL region. However, even with the underlying

assumptions of the CBA, we are able to inform program implementers on the relative cost-

effectiveness of the different interventions, where CFS is the most cost-effective. On the other

hand, WASH and Solar power both yield returns that recover all of the initial investment, but

after at least 10-11 years. This means that the depreciation rate of physical capital in terms of

when reinvestment in WASH facilities and Solar panels is necessary.

Page 10: ENDLINE EVALUATION · CBO Community-Based Organization CBA Cost-benefit analysis CEA Cost-effectiveness analysis ... consistent with the approach that is promoted by the intervention,

10

INTRODUCTION This report presents the results of an observational study conducted by FHI 360 for UNICEF/ Kenya,

examining the implementation and early outcomes of the different interventions implemented in the Arid

and Semi-Arid Lands (ASAL) of Kenya. The study examines schools affected by four different types of

interventions, including water, hygiene and sanitation (WASH), solar panels, child-friendly schools (CFS),

and community involvement through communication for development (C4D), as well as their

combinations.

Background The ASAL region forms 84 percent of Kenya’s land mass and are characterized by low rainfall (see Figure

4). These regions are sparsely populated making it difficult to provide infrastructure and social services

within reasonable distances. The economic mainstay of

these areas is pastoralism. The regions have low

enrollment rates compared to other areas in the country.

According to the 2014 Education Statistical Yearbook,

the bottom 7 counties in the country in terms of primary

gross and net enrollment rates were in the ASAL region,

with only Isiolo (14th lowest, but still below the national

average) above this (MoEST, 2014). Further, an

analysis conducted to map schooling levels of all

persons aged 6 and above indicates that only 32 percent

of the population in Northern Kenya have ever enrolled

in school compared to the national average of 77

percent (KNBS, 2012). The low levels of primary and

secondary education attainment are attributed to early

or forced marriages; child labor, and other socio-

cultural practices that place a low value on formal

education. To overcome some of these challenges, the

Government of Kenya (GOK) has invested in a series

of initiatives to promote education in the ASALs. These

initiatives are complemented by international

development partners, nongovernmental organizations

(NGOs) and community-based organizations (CBOs),

including: DFID, UNICEF, the World Food Program, and the World Bank.

UNICEF Interventions in the ASAL Region In 2013-2014, UNICEF partnered with the Ministry of Education Science and Technology (MOEST) and

the United Kingdom’s Department for International Development (DFID) to establish a policy and

institutional framework to provide quality and child-friendly education services for nomadic children.

Today, this program benefits an estimated 3.5 million boys and girls aged 4–18 years in arid counties in

Kenya. The program consists of four discrete school-based interventions:

1. Water, sanitation, and hygiene (WASH), including constructing or connecting improved water

sources to schools for hand washing, drinking, bathing, and sanitation; building gender-sensitive

school latrines with hand washing facilities; conducting hygiene promotion training for teachers

and school children; and building capacity of Schools Boards of Management (BoM) to ensure the

financial and operational sustainability of services.

2. Solar lighting for two dormitories (boys and girls), ablution facilities, and two upper primary

classrooms in each school, to maximize afterschool time and improve security.

Figure 1 – Map of Kenya

Page 11: ENDLINE EVALUATION · CBO Community-Based Organization CBA Cost-benefit analysis CEA Cost-effectiveness analysis ... consistent with the approach that is promoted by the intervention,

11

3. Child-Friendly Schools (CFS) capacity development training developed teacher skills in child-

centered pedagogy; addressed issues of gender inclusiveness; developed teaching and learning

materials; and trained primary school head teachers to improve school management. These

activities were undertaken through a CFS monitoring toolkit, action research, and a mobile phone-

based school data monitoring and evaluation system. The program also established children’s

governance groups to enhance student leadership, participation, and empowerment in school

decision making.

4. Communication for Development (C4D) aims to build community capacity to meaningfully engage

and participate in analyzing challenges and barriers and identify priority solutions for education

access and quality issues in schools. The intervention used local channels to communicate and

promote understanding of the legal framework of education governance. Community members

were inducted in CFS principles and school management and disaster risk reduction (DRR), and

engaged parents, teachers, and pupils together in school-based change.

5. Combination schools received a combination of more than one of the above interventions, or

received all four.

METHODOLOGY Given the complexities of the operational environment in the ASAL regions, the number of substantive

changes that took place with respect to the assignment of schools to interventions, as well as the duration

and timing of the interventions, UNICEF and FHI 360, in consultation with DFID, agreed to change the

methodology of this evaluation from a randomized control trial (RCT) to a quasi-experimental design

(QED).

The study focuses on the full spectrum of interventions implemented in the ASAL regions by UNICEF with

funding from DFID, including WASH, installation of solar lighting, CFS, and C4D, as well as the

combination of these interventions, with the goal of understanding the magnitude of interventions and their

expected effects, using a framework that accounts for the duration, “dosage” and enabling environment

surrounding interventions and their implementation. The framework also links interventions to their

intended impacts at the student, school, and community levels (see Annex D for specific outcomes by

intervention). Finally, we unpack the how and why of these linkages in an in-depth way through focus group

discussions. While the initial design presumed single-treatment arm interventions, it was revealed during

the baseline that most schools had already had more than one treatment in place or completed shortly before

the baseline. Therefore, the endline data are designed to capture the net effect of a combination of different

interventions completed at different times while our analysis controls for pre-baseline conditions and

background characteristics of the schools.

Selection bias cannot be fully ruled out from the study since the selection of schools into treatments was

not purely random and many schools had experienced prior interventions similar in nature to the ones being

evaluated which may have affected their participation in the interventions. However, the QED using the

full sample provides useful insights into what worked and what did not in the implementation of the

interventions participating schools. In addition, the study allows for an initial testing of the theories of

change that drove the design of the interventions. The results allow for a more nuanced understanding of

the complementary effects as well as the sequencing of different interventions, and an evidence base about

what changes can be expected given the nature of the operating environment and competing pressures

facing the schools and communities in the ASAL regions.

Key data sources The following instruments were used in the endline study. Following the baseline, it was determined that

additional instruments were needed to collect more in-depth data related to the interventions. The team

Page 12: ENDLINE EVALUATION · CBO Community-Based Organization CBA Cost-benefit analysis CEA Cost-effectiveness analysis ... consistent with the approach that is promoted by the intervention,

12

added a school observation protocol; classroom observation protocol; student survey; and expanded the

existing Head Teacher questionnaire and focus group protocols.

1. Head Teacher questionnaire

2. School observation tool, including a) a checklist of school facilities; and b) enrollment,

attendance and retention data tool capturing data from school registers

3. Focus group guides, including: a) student focus group and b) community/ parent focus

group;

4. Classroom observation tool

5. Rapid student survey on study behaviors

Each of these instruments were administered following an extensive series of training sessions carried out

by FHI 360 staff in consultation with a local data collection firm. Enumerators were recruited from a

selection of individuals that participated in the baseline, as well as recommendations from FHI 360 staff in

Nairobi. A series of training workshops were carried out during the last two weeks of June and into early

July, 2015, in Nairobi. The classroom observations were carried out using the Stallings methodology1

(Stallings, 1980) and the instrument was designed to capture differences in teaching practices and level of

resources across different types of schools. In each data collection team, at least one enumerator received

training on the Stallings methodology by FHI 360 staff, prior to the launch of endline data collection.

Training was carried out during a day-long workshop at the FHI 360 Nairobi country office, and follow-up

support was provided the following week during a more extensive week-long training on all of the data

collection instruments.

The data sources listed above provide information on the key immediate, lower-level outcomes that can be

plausibly expected in affected schools at the time of the endline. The initial data collection took place from

July 6 – 31, 2015. The follow-up and re-verification of data took place from September 21 – September 30,

2015. A teacher strike that was carried out during re-verification did not impact data collection, because

head teachers remained at public schools to administer examinations.

Constructing the comparison group Descriptive analysis The construction of comparison groups in this study is based upon the availability of data, the timing of

intervention durations, and the implementing organizations associated with the different interventions.

UNICEF vs Non-UNICEF interventions – We classify interventions by source of the implementing

partner at each school. During the baseline, we found that WASH, installation of solar lighting,

CFS, or C4D interventions were implemented in a large number of schools by non-UNICEF

partners.2 For example, the Government of Kenya is implementing a large-scale solar panel

intervention, and many schools in the sample received support from either/or, or both programs.

Including them in the comparison group without any additional markers may not be appropriate,

and therefore we separate schools into categories based upon whether they received the intervention

from UNICEF (or implementing partner organization), from a different organization, or not at all.

“Longer duration” vs, “shorter duration” – In accordance with our duration/dosage/ enabling

environment framework, we note the length/duration of the intervention as a way of predicting

impact on outcomes. As we note above, the baseline revealed that some schools had already begun

1 The Stallings Observation System was developed in the 1970s in the United States to evaluate time use in

classrooms, and has since been adapted to other country contexts by the World Bank and other organizations. 2 We identify an intervention as being implemented by UNICEF if the Head Teacher of a given school identified

UNICEF as one of the implementing partners of a given intervention.

Page 13: ENDLINE EVALUATION · CBO Community-Based Organization CBA Cost-benefit analysis CEA Cost-effectiveness analysis ... consistent with the approach that is promoted by the intervention,

13

implementation of their intervention prior to data collection. It was therefore determined that the

endline data collection instruments would capture whether the intervention had been implemented

by a UNICEF partner or a different organization, both prior to June 2014 and after June 2014. This

is particularly relevant for the WASH and CFS interventions, as no schools report a C4D

intervention and only 4 schools report that UNICEF installed solar lighting prior to June 2014.

In this report, the above categories are applied in descriptive analysis, baseline-endline changes, and

regression analysis, and comparisons to a constructed control group are done on the basis of propensity

score matching as outlined below.

Regression analysis To evaluate the impact of an intervention on key outcomes among schools receiving the intervention, we

estimate count and linear regression models, respectively. It is important to note that for the regression

models to isolate the impact of the intervention from potential confounders such as school size, number of

teachers, head teacher characteristics, and geographic location, we must compare schools that received the

intervention to similar schools that did not.3

Figure 2 below plots the propensity score distributions for schools that received the WASH intervention,

and schools that never received any treatments. In a randomized controlled trial, or in a quasi-experimental

design, schools receiving the treatment and schools in the control group would exhibit overlapping

propensity score distributions. However, we can see that the two groups of schools display rather distinct

distributions with little overlap in propensity scores providing evidence that the two groups are dissimilar

and may not serve as a viable point of comparison to evaluate the interventions.

Figure 2 - Propensity Score Distribution, by Treatment and Pre-Designated Comparison Groups

3 We compute propensity scores using logistic regressions of the probability of being assigned the treatment as a function of their

observable characteristics to ascertain the level of similarity between intervention and pre-designated control schools. The

propensity score measures the likelihood that a given school will be assigned the treatment.

Page 14: ENDLINE EVALUATION · CBO Community-Based Organization CBA Cost-benefit analysis CEA Cost-effectiveness analysis ... consistent with the approach that is promoted by the intervention,

14

As a result, we exclude schools from the designated control group that never received any treatments that

are dissimilar to the treatment schools. Therefore, for each treatment, we construct a control group that is

composed of schools that are somewhat similar to the treatment schools that did not receive any treatments

in the period prior to June 2014, and did not receive the corresponding treatment in the period following

June 2014. For example, the constructed control group for the WASH schools will have received no

interventions in the period before June 2014 and will receive no WASH intervention after June 2014. Figure

3 plots the propensity score distributions of each intervention and its corresponding comparison group.

Figure 3 - Propensity Score Distribution, by Treatment and Constructed Comparison Groups

Although not perfect, we observe a larger degree of similarity between the treated schools and the

‘constructed’ comparison groups, as evidenced by the larger degree of overlap in the propensity score

Page 15: ENDLINE EVALUATION · CBO Community-Based Organization CBA Cost-benefit analysis CEA Cost-effectiveness analysis ... consistent with the approach that is promoted by the intervention,

15

distributions of the treatment and comparison groups. As such, we will use these comparison groups in our

regression analyses that evaluate the impact of each intervention on outcomes of interest such as student

enrollment, attendance rates, and grade promotion rates.

We estimate a count (Poisson) regression model to ascertain the impact of each intervention on enrollment

growth between 2014 and 2015, and a linear regression model for measuring growth in attendance and

promotion rates over the same timeframe.4 Specifically, we regress 2015 enrollment in 5th and 6th grade on

WASH, installation of solar lighting, CFS, or C4D intervention status, 2014 enrollment for the same grades,

characteristics of the head teacher, number of teachers at the school, teacher experience distribution,

distance to the nearest source of water, and county fixed effects. Lastly, we assess the effectiveness of the

Solar intervention on an additional outcome, students’ reported frequency of study, using a logistic

regression model.5

Description of the study population The target population of the schools benefiting from UNICEF interventions consist of low cost primary

boarding schools (LCBPS) located in the ASAL region of Kenya. The schools received one of the five

interventions (i.e. WASH, installation of solar lighting, CFS, C4D, or combined), sponsored by UNICEF

Kenya in the counties of Garissa, Isiolo, Mandera6, Marsabit, Samburu, Tana River, Turkana, and Wajir.

These eight regions are chosen to achieve broad coverage of the ASAL region in Kenya, known for its low

enrollment and educational attainment rates. The unit of analysis is at the school level for the quantitative

data collection, and at the intervention cluster level for the qualitative focus groups. Cluster sampling by

intervention was done at the regional level and is not representative at the county level.

The baseline data were collected through a school head teacher questionnaire and complemented by a series

of focus groups with students, parents/community members, and teachers. The endline data collection

process included visits to the same schools to gather data using the instruments focused on the study

objectives listed above. The sample size for the study was 349 schools distributed across the counties.

During endline data collection, the teams were able to reach 321 schools, while 28 schools were unable to

be reached due to security reasons. Table 1 below highlights the number of schools reached by treatment

intervention. As the table shows, most of the schools that were unable to be reached were control schools.

Table 1 - Number of schools by type of intervention Treatment Total number of schools by

UNICEF intervention Number of schools reached in

endline survey

C4D 49 47

CFS 50 47

Solar 48 44

WASH 53 51

Combination 55 51

Control 94 81

Total 349 321

4 The Poisson model is appropriate when the dependent variable represents counts as in the case of total enrollment. 5 We estimate the impact of solar panels on the probability that a given student reports studying at least once a day. 6 Baseline data collection in Mandera County could not be completed due to security concern.

Page 16: ENDLINE EVALUATION · CBO Community-Based Organization CBA Cost-benefit analysis CEA Cost-effectiveness analysis ... consistent with the approach that is promoted by the intervention,

16

Characteristics of the sample As discussed in the methodology section, the original number of endline schools was 349 compared with a

sampling frame of 440 schools in the baseline7. For the endline, FHI 360 was able to reach approximately

28 percent more schools (321 compared to 250) than in the baseline, producing a larger dataset. In Isiolo,

Samburu and Tana River, the teams were able to reach all UNICEF supported schools. Mandera and Garissa

had the highest number of schools not visited due mainly to security issues that prevented the teams from

reaching those schools.

Figure 4 and Figure 5 provide some descriptive background statistics about the schools and counties that

were visited. 104 out of the 321 schools visited were in Turkana, while a small number were visited in Tana

River (11) and Isiolo (14). Error! Reference source not found. Figure 5 shows that pupil-teacher ratios

range across treatment arms from 71 in WASH schools to 98 in Solar schools.

Figure 4 - Number of schools reached by county

Figure 5 - Pupil-teacher ratio

RESULTS

Key Findings This section presents the overall summary findings for key outcomes associated with each intervention.

The data are disaggregated by intervention cluster and gender where appropriate. Since schools across all

treatment categories have received interventions from UNICEF and non-UNICEF sources, we unpack the

treatments by duration and source to get a better assessment of how UNICEF partner supported schools are

performing comparatively across the region.

WASH – We are able to divide schools that received the WASH intervention into those that received the

intervention prior to June 2014 (77 schools), after June 2014, and from non-UNICEF organizations.

Both the descriptive and regression analyses show that schools that implemented WASH the

longest (prior to June 2014) exhibited the largest gains in enrollment relative to all other schools.

Specifically, we estimate that enrollment among schools that implemented WASH the longest

7 Prior to the endline, UNICEF and FHI360 worked with the implementing partners to verify the schools where the interventions

had been implemented. This verification process produced the final list of 349 schools. We know from the baseline process that

many schools that were part of the 440 ceased to exist for different reasons.

104

52

51

35

32

22

14

11

0 50 100

(#)

Turkana

Marsabit

Samburu

Garissa

Wajir

Mandera

Isiolo

Tana River

98 9587 85

7771

02

04

06

08

01

00

(#)

Solar

Con

trol

CFS

Com

bo.

C4D

WASH

Page 17: ENDLINE EVALUATION · CBO Community-Based Organization CBA Cost-benefit analysis CEA Cost-effectiveness analysis ... consistent with the approach that is promoted by the intervention,

17

increased by approximately 20 percent relative to similar schools in the constructed comparison

group that never received the WASH intervention.

Attendance rates are also highest among WASH schools with the longest duration. Relative to

similar schools that never implemented WASH, we estimate an 11 percent increase in attendance

rates, although this estimate is not statistically significant at the conventional levels, due possibly

to the amount of variability in the attendance data given the sample size.

Head Teacher perceptions of WASH were more positive among those that had received WASH for

a longer duration than those that received WASH after June 2014.

Solar Power – There are only 4 schools that implemented the solar intervention prior to June 2014. As a

result, we consolidate all schools that received Solar power interventions from UNICEF and designate them

as UNICEF Solar schools.

A higher proportion of students in UNICEF Solar schools report studying for more hours per day

as compared to non-UNICEF Solar schools and schools that never received Solar.

All else being equal, the estimated likelihood that a given student reports studying at least once a

day is 13 percent higher among UNICEF Solar schools than similar schools that never received the

Solar intervention.8

Regression results show that enrollments are only modestly higher among UNICEF Solar schools

than similar schools in the constructed comparison group, by about 6 percent.

Head Teacher perceptions of the Solar intervention’s efficacy are more positive among UNICEF

Solar schools than non-UNICEF schools.

CFS – We are able to identify 18 schools that received the CFS intervention prior to June 2014.

Regression results suggest that enrollment growth among schools that implemented CFS prior to

June 2014 as well as those that implemented after June 2014 is 8.6 percent higher than the

constructed comparison group. However, non-UNICEF CFS schools exhibited the largest gains in

enrollment.

Changes in attendance and grade promotion rates among all UNICEF CFS schools are not

statistically different from those of the comparison group. However, descriptive analyses show that

changes in attendance rates among UNICEF schools that implemented CFS after June 2014 were

higher than schools that did not implement CFS. Changes in grade promotion rates among boys

were higher for all UNICEF CFS than non-CFS.

Classroom observations suggest that teachers in schools receiving the CFS intervention are more

likely to exhibit characteristics consistent with the approach that is promoted by the intervention,

e.g. ask probing questions such as how and why, ask students to explain in class discussions, and

ask students to apply understanding to relevant life experience. However, teachers in schools

receiving the CFS intervention are only slightly less likely to integrate gender sensitive material

into pedagogy.

C4D – The C4D intervention was largely implemented in conjunction with other interventions.

Only six schools report receiving only C4D from UNICEF, therefore, we are unable to accurately

isolate the impacts of C4D from that of other interventions. Further, there are no schools that report

receiving a C4D intervention prior to June 2014, so we are unable to account for the duration of

treatment effect.

8 Study frequency of students is self-reported by the students in response to the student survey item “How often do

you study in a typical week?”

Page 18: ENDLINE EVALUATION · CBO Community-Based Organization CBA Cost-benefit analysis CEA Cost-effectiveness analysis ... consistent with the approach that is promoted by the intervention,

18

Overall, the descriptive analyses show a positive effect on enrollment and attendance, but not grade

promotion.

Additional descriptive evidence suggests that C4D schools are more likely to be involved in

community mobilization efforts around the importance of schooling and the promotion of education

for out of school children.

Combination Schools – Schools that receive two or more UNICEF interventions are more likely to show

positive effects on attendance and enrollment. We also analyze differences in the outcomes of interest

among specific pairwise intervention combinations.

Schools that received a combination of Solar power and C4D exhibited large gains in enrollment

relative to the constructed comparison group by 10.5 percent. The same combination of

interventions yielded a 46.5 percentage point increase in attendance rates and a 14.8 percent

increase in the likelihood of students studying more than once a day.

Schools implementing C4D in conjunction with CFS are estimated to have an attendance rate that

is 28 percentage points higher than that of the constructed comparison group.

Schools that implemented both CFS and WASH experienced approximately a 9.4 percent increase

in enrollment relative to the constructed comparison group. Although the estimate is positive and

certainly non-negligible, it is not statistically significant at the conventional levels.

Cost-Benefit Analysis – We compute the benefits of WASH, installation of solar lighting, and CFS

interventions as additional taxes paid over a lifetime as a result of increased educational attainment from

each intervention. Additionally, to compute the total benefits accrued to each intervention, we aggregate

across the total number of students who are estimated to have increased their educational attainment but

would not have if not for the intervention. Lastly, we compute the number of years required for each

intervention (investment) to yield a full return, i.e. each Kenyan shilling invested is returned in full.

We estimate that the total investment in WASH schools, where we are able to identify the

enrollment impact, may be returned in full between 11 and 21 years post-implementation,

depending on the discount rate. Lifetime benefits-to-cost ratios for WASH range between 1.19 and

3.85 per US Dollar invested.9 This means that every dollar invested in WASH is returned between

1.19 and 3.85 times 36 years after implementation.

Investments made for Solar power are estimated to be returned in full after 10-18 years, depending

on the discount rate. The lifetime benefits-to-cost ratios for the investment in Solar lighting range

between 1.27 and 4.17 per US Dollar invested.

Child friendly schools yielded the highest return on investment according to our simulation model.

Initial investments in CFS are returned in full after only 4-5 years and the lifetime benefits-to-cost

ratios for CFS range between .04 and .12 per USD invested.

Overview of changes in key outcomes This section presents trends in enrollment, attendance and retention over the intervention period. Schools

are organized according to the originally designated UNICEF intervention categories. Note that this is a

description of a trend, not accounting for the differences in the composition of the schools or duration of

their interventions. This contrasts with how interventions will be designated in the following sections,

9 Benefit-to-cost ratios are impervious to the currency used in the calculation. As such, we may use Kenyan

Shillings or US Dollars interchangeably in benefit-to-cost ratios.

Page 19: ENDLINE EVALUATION · CBO Community-Based Organization CBA Cost-benefit analysis CEA Cost-effectiveness analysis ... consistent with the approach that is promoted by the intervention,

19

where we seek to account for the duration of exposure to treatment and restrict analysis to comparable

schools.

Enrollment trends over the intervention period Table 2Error! Reference source not found. shows average enrollment per school across the initially

designated treatment groups. As the table shows, enrollments generally decreased across all schools in the

sample. The reason for the general trend is not known, although the focus group discussions provide some

indication – namely safety and security. However, some increases are seen, particularly for boys. Growth

in boys enrollment was observed in the Combination and Solar category schools (+14 percent for each

category), while enrollment in WASH schools has stayed relatively even. Enrollment also increased for

girls in Combination schools (+5 percent), but decreased across all other interventions. Also notable is the

change in Solar schools, which saw a large decrease in enrollment for girls (-17 percent) but also a large

increase for boys (+14 percent). These results should be interpreted as general trends, while the regression

results presented later are intended to serve as a means of comparison on enrollment trends between groups.

Table 2 - Enrollment changes over the intervention period10 Treatment Girls Boys % change

2013 2015 2013 2015 Girls Boys

C4D 441 459 345 382 4.1% 10.7% CFS 413 344 301 270 -16.7% -10.3% Combo. 488 422 351 327 -13.5% -6.8% Control 405 413 325 383 2.0% 17.8% Solar 369 343 317 358 -7.0% 12.9% WASH 322 351 290 299 9.0% 3.1%

Attendance rate trend over the intervention period In addition to enrollment, we measured attendance by taking information from class registers for the day of

enumeration, and dividing it by the corresponding enrollment figure for that grade. Figure 6 presents a

mixed picture of attendance changes from baseline to endline. While overall, attendance rates are relatively

high (generally around 90 percent, see Annex D), the positive changes one would like to see as a result of

the interventions are not seen across all categories. There are small but clear drops in attendance rates across

all interventions. Similar to enrollment, there are drops in attendance among control schools. Combination

schools are the only intervention type that see increases, most notably for boys (from 84 percent to 92

percent, see Annex D for overall numbers).

10 Note that intervention designations changed between baseline and endline, which is why enrollment numbers do not match

what is in baseline report.

Page 20: ENDLINE EVALUATION · CBO Community-Based Organization CBA Cost-benefit analysis CEA Cost-effectiveness analysis ... consistent with the approach that is promoted by the intervention,

20

Figure 6 – Attendance rate changes over the intervention period

Changes in retention A similarly mixed picture emerges when we look at grade-to-grade retention over the intervention period.

Error! Reference source not found. Figure 7 shows promotion rates between grades 5 and 6 for the two

year (Figure 7). Interestingly the combination schools see drops in promotion rates, while enrollment and

attendance remain high. There is also a drop in promotion for girls in CFS schools. However, on the whole,

promotion rates are relatively high and retention is not a major problem across the intervention schools (see

Annex D)11.

Figure 7 – Changes in retention over the intervention period

Taken together, the data on our outcomes of interest show that there is a substantial degree of fluctuation

on all key outcomes – enrollment, attendance, and retention (grade promotion). Girls’ attendance levels

were lower at the time of the endline survey, compared to the baseline. Promotion rate changes differed

across genders, and the measured changes in promotion rates are not consistent with our hypotheses (e.g.

WASH schools did not show growth in girls attendance). Overall, on face value there do not seem to be

clear differences in outcomes between intervention schools and schools in the designated control group.

However, as the next section will show, when duration of the intervention is taken into account and the

analysis sample is restricted to comparable schools the differences in outcomes are better identified.

11 It should be noted that there is automatic promotion between grades at the primary level, in Kenya. That said, many schools do

report repeaters, and the numbers presented are for a subset of schools for which the data are available.

-6.2-3.2

7.5

-3.8

-14.1-12.4

-20.0

0.0

20.0

% c

han

ge

C4D CFS Combo. Control Solar WASH

Girls Attendance Rate Change, Grades 5-6

-2.1

-11.7

8.0

-6.3

-14.3

-19.9

-20.0

0.0

20.0

% c

han

ge

C4D CFS Combo. Control Solar WASH

Boys Attendance Rate Change, Grade 5-6

-4.7

-16.6

-5.8

2.23.9

-11.8

-20

.0

0.0

20

.0

% c

ha

ng

e

C4D CFS Combo. Control Solar WASH

Girls Promotion Rate Change, Grades 5-6

-5.6-3.6

-9.0

1.1

-9.0

1.2

-20

.0

0.0

20

.0

% c

ha

ng

e

C4D CFS Combo. Control Solar WASH

Boys Promotion Rate Change, Grade 5-6

Page 21: ENDLINE EVALUATION · CBO Community-Based Organization CBA Cost-benefit analysis CEA Cost-effectiveness analysis ... consistent with the approach that is promoted by the intervention,

21

Indicative results of the WASH Interventions Status of WASH interventions Beginning in 2014, three international NGOs

(World Vision, CARE, and Food for the Hungry)

began implementing WASH interventions in

schools across the counties of Samburu, Turkana,

Tana River and Wajir. The objectives included

building latrines and hand washing facilities in

schools where none existed; connecting improved

water sources to schools for hand washing, drinking,

bathing and sanitation; building gender-separated

school latrines; conducting hygiene promotion

training and building the capacity of school

management committees to ensure the financial and

operational sustainability of services. In schools

where facilities already existed, the organizations

either refurbished existing facilities or

complemented the work that had been completed by

other organizations or the government.

How effective are the WASH interventions for outcomes of interest? The hypothesis regarding WASH interventions is

that improved water and sanitation systems will

increase school attendance, particularly among girls. As a result, the intervention will increase their overall

exposure to instruction leading to higher academic performance.

As noted earlier, our data allow us to distinguish schools that reported having received the WASH

interventions from UNICEF providers since prior to the program implementation period in question, from

those that had been receiving the intervention only since June 2014. Such a distinction is important for

determining the extent to which the duration of treatment may have had an effect on outcomes of interest.

As seen in Figure 8 presented below, enrollment increases were greater in schools that reported receiving

the WASH treatment from UNICEF since prior to the start of the program period in question (i.e. since

prior to June 2014), compared to a) schools that reported receiving the WASH intervention funded by

UNICEF for a shorter duration, and b) schools that did not receive a WASH intervention from UNICEF.

The enrollment increase in the “long duration UNICEF intervention” group was four times as large for

boys, and twice as large for girls as either of the comparison groups. Attendance rate changes should be

interpreted with caution, since overall level of observed attendance is relatively high (80%-95%);

nonetheless, when broken down by grade, small increases in girls’ attendance may be observed in “longer

duration” UNICEF WASH schools (Figure 8).

World Vision: World Vision (WV) implemented

WASH interventions in 56 schools across the four

counties. Their interventions consisted of building or

rehabilitating latrines; building or rehabilitating water

sources; provision of hygiene training to teachers; and

supporting the dissemination of school health policies.

Food for the Hungry: Food for the Hungry

implemented WASH in 25 schools across Marsabit and

Isiolo. FFH also GPS mapped the schools and

implemented similar interventions to WV.

CARE: CARE implemented WASH in 20 schools in

Garissa. Their focus was primarily on establishing a

stable water source (particularly piped water); and

building the capacity of staff to teach about WASH. In

six of these schools, there was no piped water so CARE

provided treatment chemicals to treat the local water

source, ensuring cleaner water for students and staff at

the schools. Similar to WV and FFH, CARE began

interventions in June 2014 with a rapid assessment and

initiated actual implementation in September 2014.

Implementation was completed in March 2015.

Page 22: ENDLINE EVALUATION · CBO Community-Based Organization CBA Cost-benefit analysis CEA Cost-effectiveness analysis ... consistent with the approach that is promoted by the intervention,

22

Figure 8 – WASH changes in key outcomes

The perceptions of head teachers in WASH schools were more unambiguously positive about the effects of

the interventions, particularly in schools that had received WASH interventions for a longer period of time.

As shown in Figure 9Error! Reference source not found., 94% of head teachers in schools that had

received UNICEF WASH interventions since prior to June 2014 agreed that it had positively impacted

attendance, compared to 83% in schools where WASH was implemented after this period. Similarly, head

teachers were more positive about effects on enrollment (84% to 76%) and dropout (75% compared to 66%)

in the former types of schools.

Figure 9 – Head teacher assessment of WASH outcomes

-23.3

76.7

-7.2

100

.0

0.0

-100

.0

# c

han

ge

Other WASHpre & post 6/14

WASHpost 6/14

Boys Mean Enrollment Change

-20.7-3.3

-30.9

100

.0

0.0

-100

.0

# c

han

ge

Other WASHpre & post 6/14

WASHpost 6/14

Girls Mean Enrollment Change

-4.9

8.2

-14.5

-20.0

0.0

20.0

% c

han

ge

Other WASHpre & post 6/14

WASHpost 6/14

Boys Attendance Rate Change, Grades 5-6

-8.0

11.0

-5.5

-20.0

0.0

20.0

% c

han

ge

Other WASHpre & post 6/14

WASHpost 6/14

Girls Attendance Rate Change, Grades 5-6

Page 23: ENDLINE EVALUATION · CBO Community-Based Organization CBA Cost-benefit analysis CEA Cost-effectiveness analysis ... consistent with the approach that is promoted by the intervention,

23

Analyzing the impact of the intervention in more depth, it is clear that there are moderate to large positive

differences between schools that have received the WASH intervention for a longer period and other types

of schools. For example, 56 percent of schools that report receiving the intervention since prior to June

2014 (hereinafter referred to as “longer duration” schools) have a water source on the school compound,

compared to 38-39 percent of shorter duration WASH schools and comparison schools with no WASH

intervention (see Annex D for a more detailed breakdown). Similarly, 29 percent of “longer duration

WASH” schools report a water source that is piped or from a well or borehole, compared to only 20-21

percent of comparison schools. Head teachers also considered handwashing and latrine facilities to be

somewhat safer in such schools, compared to comparison schools. Similarly, we see large differences

between types of schools on whether they have received training in hygiene promotion. 94 percent of longer

duration WASH schools have received such training, compared to 74 percent of WASH post-June 2014

only and only 56 percent of comparison schools (see Annex D).

One emphasis of the UNICEF WASH program is to provide separate latrine facilities for boys and girls,

with the goal of improving school attendance among girls. While most schools regardless of intervention

had separate latrines for girls and boys, a slightly higher (97 percent) percentage of longer duration WASH

schools reported separate latrines, and WASH longer duration schools reported a much greater number of

latrines compared to other types of schools (see Figure 10Error! Reference source not found.). The mean

number of latrines per school is significantly less for girls (5.0 compared to 7.6) and boys (5.2 compared to

8.2) in the baseline, compared to the endline. The pupil-toilet ratio has also dropped significantly from the

baseline. While the ratio decreased from 91 to 72 in non-UNICEF WASH schools, the ratio in UNICEF

WASH schools is roughly half what it was during the baseline.

Figure 10 – Changes in availability of latrines

Finally, a goal of the UNICEF WASH program is to establish clean handwashing facilities close to latrines.

While this remains a challenge across all types of schools, a greater percentage of WASH long schools (34

percent) report clean handwashing facilities close to latrines, compared to other types (32 percent in WASH

short and 24 percent in other types of schools).

WASH regression analysis We disaggregate the WASH treatment group into schools that received the intervention for at least two

periods (pre- and post-June, 2014) and those receiving the intervention only post-June 2014. Further, we

interact the group of schools who received the WASH treatment only in the post-June 2014 period with

their treatment status from the other interventions, i.e. for WASH, we identify the pairwise combinations

of schools that received WASH in conjunction with other interventions.

5.1 5.4 5.0

7.6

4.7

6.5

05

10

(#)

Other WASHpre & post 6/14

WASHpost 6/14

Mean No. of Girls Latrines

Baseline Endline

5.05.8

5.2

8.2

5.0

6.7

05

10

(#)

Other WASHpre & post 6/14

WASHpost 6/14

Mean No. of Boys Latrines

Baseline Endline

Page 24: ENDLINE EVALUATION · CBO Community-Based Organization CBA Cost-benefit analysis CEA Cost-effectiveness analysis ... consistent with the approach that is promoted by the intervention,

24

For ease of interpretation, we compute the marginal effects associated with the Poisson regression results

and are interpreted as the increase in the number of students enrolled from one year to the next for WASH

schools relative to similar comparison schools. The marginal effects and regression coefficients from the

linear regression model are one and the same. For attendance and promotion rates, we measure the

percentage point difference between WASH schools and similar schools in the comparison group.

Figure 11 – Margins plot of WASH outcomes Growth in Enrollment

Growth in Attendance Rates

Growth in Promotion Rates

Notes: Marginal effects associated with enrollment are expressed as changes in enrollment in response to the intervention, in counts.

Marginal effects associated with attendance and promotion rates are expressed as percentage point changes in response to the

intervention. Intervention statuses associated with fewer than 10 schools are suppressed. Intervention statuses associated with fewer

than 10 schools are suppressed.

Figure 11Error! Reference source not found. plots the marginal effects from the enrollment, attendance

rate, and promotion rate regression results along with 90 percent confidence intervals for each estimate.

Schools that received the WASH intervention in both the pre- and post-June 2014 periods experience a 23

student increase (effect size of 22.4 percent of a standard deviation) relative to similar schools that never

received the WASH intervention and did not receive any other interventions prior to June 2014. We also

see a statistically significant 15.4 student increase (effect size of 10 percent of a standard deviation) in

enrollment among non-UNICEF WASH schools relative to similar schools in the comparison group. We

find almost no effect on WASH schools that have had the intervention only recently, suggesting that the

intervention may require additional time for its effects to materialize. Schools that implemented WASH

along with CFS exhibit a positive increase in enrollment. However, this estimate is not statistically

significant.

We find a positive, though statistically insignificant, effect of WASH on attendance rates among schools

that received the treatment in two periods by 11.1 percentage points. WASH schools that received the

treatment in the most recent period and in conjunction with CFS exhibit no differences in attendance with

comparable non-WASH schools. The estimated treatment effect among non-UNICEF WASH schools is

also not statistically different from zero. Lastly, when examining the impact of the WASH intervention on

grade promotion rates, we find modest positive effects across all types WASH schools by between 1.5

percentage points and 4.3 percentage points. However, these estimates are not significantly different from

zero.

Challenges in implementation and monitoring of WASH interventions. Across the literature, we know that there are a number of constraints in sustaining WASH interventions. To

understand the constraints that UNICEF-supported currently face in maintaining WASH facilities, we asked

school Head Teachers a series of questions related to these challenges. Error! Reference source not found.

Figure 12 shows the percentage of head teachers reporting challenges in initial installation or construction,

maintenance or upkeep of facilities, and usage of facilities by students, teachers and staff, by type of school.

Page 25: ENDLINE EVALUATION · CBO Community-Based Organization CBA Cost-benefit analysis CEA Cost-effectiveness analysis ... consistent with the approach that is promoted by the intervention,

25

As can be seen, schools working with non-UNICEF WASH implementers report slightly more challenges,

particularly in ongoing maintenance of facilities. We can also get some idea of how head teachers view the

causes of these challenges. Error! Reference source not found.2 also shows the percentage of schools

citing a lack of funding, no training in how to use the facilities, and poor quality of the service provider as

primary causes of challenges in implementation.

Figure 12 - WASH challenges in implementation, as reported by Head Teachers

Findings from focus groups As discussed at the beginning of this section, increasing the equitable access to and use of safe water and

basic sanitation services improves health and education outcomes. Numerous reports confirm the

importance of the link between WASH and education (UNICEF, 2006; Gleaton, 2012; Jasper, Le, and

Bartram, 2012). These studies also indicate that issues of corruption in the water committees and delayed

repair of damaged infrastructure often result in erratic service and the lack of financial support, and

community engagement, were major limitations to sustaining and improving WASH interventions.

Access to clean water and hygiene facilities was of high importance to all of the parents, community

members and students who participated in the focus group discussions. These different groups noted that

having water at the schools allowed students to bathe, cook, and wash their clothes. Access to water made

it easier for to justify coming to school; and many parents noted the particular importance that the hygiene

training has had on reducing diarrhea, vomiting, and other water borne diseases. Several parents across all

counties noted:

“There are many pupils that are boarding in the school so this water is of great help because they use it

for cooking their food, bathing and washing their clothes. It is important for them to have water when at

the school” Parent FGD, Garissa.

“We have handwashing facilities which have been place next to the toilets and also if you look at

this basins, you see this one is filtered so you put water there and it’s safe for the children to drink

water.” Community Member FGD, Marsabit

Parents often noted that they were made aware of the WASH practices through meetings at the school and

that organizations such as UNICEF not only held workshops, but handed our materials to teach them about

the importance of hand washing and sanitation. One parent noted, “there was a meeting in this school where

all the schools from around here attended and we were taught how to wash our hands and were also given

leaflets with this information. The information was also written on buildings - wash hands after using the

toilets. As the pictures below show, many of the schools we visited had colorful walls painted with

instructions on good hygiene for students to follow.

63.663.3

74.8

85.2

48.6

59.0

05

01

00

(%)

Initial Maintenance Use

Percent Reporting Challenges in Implementation

UNICEF Non-UNICEF

50.9 54.1

29.537.7

20.524.6

05

01

00

(%)

Funding No Training Service

Percent Reporting Causes of Challenges

UNICEF Non-UNICEF

Page 26: ENDLINE EVALUATION · CBO Community-Based Organization CBA Cost-benefit analysis CEA Cost-effectiveness analysis ... consistent with the approach that is promoted by the intervention,

26

One additional issue that was raised in the focus groups centered around the way many of the WASH

facilities had been modified to ensure that physically challenged students could also use the facilities. One

of the areas that UNICEF specifically supported was making modifications to make WASH facilities more

child friendly – particularly for children with disabilities. Therefore, hearing the issue discussed among

parents in Marsabit reinforced the fact that the interventions were seen as important by the communities.

Finally, parents and students also mentioned that having the water and hand-washing facilities on the school

grounds reduced the amount of time that students had to take away from classes or studying, thus, increasing

the students time to focus on learning.

While much of the discussion was focused on the positive aspects of the WASH interventions, the

moderators were asked to also discuss the types of challenges that are faced in the school, vis a vis the

WASH intervention. As expected, parents noted that this year there have been more challenges related to

the access to clean water than in previous years. The drought was mentioned by nearly 30 percent of focus

group participants. They indicated that bore holes had been drying up at the schools or that there was no

longer access to clean water for hygiene purposes as a result of the drought. The following two quotes from

focus group discussions highlight the importance communities do place on the provision of clean water:

“Our school had a bore hole but it got dry, now students do not have water”.

Another participant notes that in some cases, schools end up relying on rainwater to fill the tanks because

the volume of children drinking the water exceeds the tank size. He noted, “…the water that we have….the

toilets are okay, but the water tank is too small to store enough water for all the children, the water will

not be enough because the children are many, so the water does not last long when they open school….water

does not last even a month, then we must depend on rainwater.”

The size and upkeep of the infrastructure posed a particular problem for many communities. Parents and

students noted, “….the toilets are okay but the water tank is small to store enough water for all the students,

the water will not be enough because the children are many. The water does not last long when they open

school the water does not last even a month….the, students go home or have to look for other sources of

water”.

The types of challenges that parents and students mentioned mirror both the literature and the perceptions

surveys conducted during this evaluation. The triangulated data shows that schools lack the resources to

maintain the infrastructure and that having sufficient water is often a key issue. Parents also indicated that

they continue to have problems with pipes bursting at the schools, which leads to water wastage and

subsequently, insufficient water for students. It was noted in several focus groups that,” …The toilets should

be kept clean and there is no water and soap….[communities need to support, but they often don’t]”.

While parents indicated that things have gotten better as a result of UNICEF WASH interventions and both

groups noted that having access to clean water does make a difference in attendance, this is a challenge that

UNICEF should consider supporting in the future. Parents in a focus group in Marsabit noted that, “If we

are provided with water, our children given food and the school is fenced we will keep them here. Nowadays

we are enlightened not like there before, we want our children to be educated – water and other things will

help keep them in school.”

Page 27: ENDLINE EVALUATION · CBO Community-Based Organization CBA Cost-benefit analysis CEA Cost-effectiveness analysis ... consistent with the approach that is promoted by the intervention,

27

Indicative Results of Solar-Power Interventions The provision of solar panels for lighting in schools was intended as a way of strengthening the schools

ability to provide safe study spaces after dark,

reduce the costs of power for schools, and

ultimately improve learning opportunities for

students. As part of this intervention, UNICEF

worked with Power Point Systems East Africa to

install solar lighting in two dormitories (boys and

girls), in ablution facilities, and in two upper

primary classrooms, in each school. To

understand the extent to which schools in this

sample have lighting, we collected data from head teachers, teachers and students at the school and

conducted school observations to verify the status of the infrastructure.

Status of Solar interventions For the solar intervention section, we look at UNICEF solar intervention schools and non-UNICEF solar

schools (most of which are supported by the GoK program). Non-solar schools mostly have regular

electricity. Approximately 198 of the 322 schools visited during the endline data collection process have

solar lighting. The solar lighting was provided to schools by many different actors including UNICEF, local

governments, and other international organizations. The solar interventions provided to schools by

UNICEF are located in dormitories, grade 7 and 8 classrooms, or the highest grade levels in the school,

latrines, and in some cases in the library. The objective of providing the solar panels is to increase the

amount of time students have available to study as well as provide additional security to the schools at

night. Error! Reference source not found.Figure 1312 shows the percentage of all schools by their source

of power, while Figure 1413 shows the number and proportion of solar schools by their implementing

organization. The largest implementer of solar power in ASAL schools in the study was reported to be the

government, followed by UNICEF.

Figure 13 – Solar: type of power

Figure 14 – Solar: UNICEF and Non-UNICEF

How effective are the solar interventions for outcomes of interest? Increasing the amount of study time for students is one of the key intended outcomes of the UNICEF has

supported solar power interventions (see Annex D). The belief is that if schools have power in places like

the dormitories and classrooms, that students – particularly those that board – will have more time to study

12 These percentages do not add to 100 because of some missing data on this response. 13 Although some schools have both UNICEF and government as implementers of solar intervention, a school is considered

UNICEF if it has both.

77.6

15.66.5

05

01

00

(%)

Type of Power

Solar Electricity None

78

120

UNICEF Other

Power Point Systems East Africa: In June 2014,

UNICEF contracted with Power Point to install

solar panels in approximately 100 schools across

all 8 counties in the study. At the time of

installation, two school staff members then

received training on use and maintenance of the

system.

Page 28: ENDLINE EVALUATION · CBO Community-Based Organization CBA Cost-benefit analysis CEA Cost-effectiveness analysis ... consistent with the approach that is promoted by the intervention,

28

since they can now study later into the evening and in the mornings when it is dark. Annex D shows the

head teachers’ evaluation of reported sufficient light in key areas of the school. UNICEF schools that are

implementing the solar intervention reported sufficient lighting in their classrooms at a higher rate than

other types of schools (63 percent of head teachers report sufficient light in the classroom in schools that

installed solar lighting compared to 39 percent where the intervention is not present at all).

Despite this, students report studying after dark in only about 50 percent of UNICEF solar schools,

compared with 52 percent of non-UNICEF solar schools and almost 62 percent of non-solar schools (Annex

D). Error! Reference source not found. Figure 15 shows the mean percentage of students who report

studying in the evening, per school. Both UNICEF and non-UNICEF solar schools appear at a disadvantage

compared to non-solar schools in facilitating the ability of students to study in the evening. Further, lighting

does appear to be a key impediment in the ability of schools to promote learning. While most of the schools

in the “no solar” category have electricity, Figure 15 shows the percentage of students who report being

unable to study at night14 due to poor lighting, across the same category of schools.

Figure 15 – Solar: differences in evening study

While it appears that solar schools overall struggle to provide opportunities for night study, it does appear

that a greater percentage of students in solar schools study for a longer period of time, per day. Error!

Reference source not found. Figure 16 shows the percentage of students by the number of hours they

report studying per day, by solar and non-solar schools. There is a trade-off between students studying less

than 2 and 2-4 hours per day, so that students in non-solar schools are more likely to study less than two

hours per day while students in solar schools are more likely to study 2-4 hours per day. A similar

percentage of pupils study more than 4 hours a day, regardless of intervention or implementing

organization.

Figure 16 – Hours of study

14 Students who indicate they typically study during the day (and not also during the evening) are asked why they do not study at

night.

24.4

18.2 18.9

01

02

03

0

(%)

Mean % of Students Studying in Evening per School

No Solar UNICEF Non-UNICEF

8.6

17.9

13.3

05

10

15

20

(%)

% Schools With Lighting Impediments to Night Study

No Solar UNICEF Non-UNICEF

36.529.8

22.7

49.6

59.163.9

13.911.113.3

03

57

0

(%)

Less Than Two Two to Four More Than Four

Mean % Students by Hours of Study per School

No Solar UNICEF Non-UNICEF

Page 29: ENDLINE EVALUATION · CBO Community-Based Organization CBA Cost-benefit analysis CEA Cost-effectiveness analysis ... consistent with the approach that is promoted by the intervention,

29

Similar to other interventions, head teachers were slightly more positive about the interventions impact on

students. Figure 17 shows the percentage of head teachers who agree or strongly agree that the named

intervention positively influenced a specific outcome. Schools where the solar intervention was

implemented by UNICEF saw slightly higher rates of agreement than schools where solar was implemented

by a different organization.

Figure 17 – Solar: head teacher assessment

Solar regression analysis We disaggregate schools’ treatment status to identify schools that received the Solar from UNICEF and

from a non-UNICEF organization. We do not distinguish between Solar schools that implemented the

intervention in the pre- and post-June 2014 periods since there were only four schools that did so. As such,

we consolidate all UNICEF schools as a single group. We still interact Solar treatment status with receiving

other treatments at the same time in the post-June 2014 period. We present graphically the marginal effects

of receiving the Solar intervention on student enrollment growth, growth in attendance rates, and growth in

grade promotion rates relative to comparable schools that never received the Solar intervention and did not

receive any other treatments prior to June 2014. Figure 18, below, plots the marginal effects associated with

various Solar treatment statuses along with their corresponding 90 percent confidence interval.

Figure 18 - Margins plot of Solar outcomes Growth in Enrollment

Growth in Attendance Rates

Growth in Promotion Rates

Notes: Marginal effects associated with enrollment are expressed as changes in enrollment in response to the intervention, in counts.

Marginal effects associated with attendance and promotion rates are expressed as percentage point changes in response to the

intervention. Intervention statuses associated with fewer than 10 schools are suppressed.

UNICEF and non-UNICEF implementing schools both show positive gains in enrollment as a result of the

Solar intervention by 6.99 and 5.74 students, respectively. The equivalent effect size is 7 percent of a

91.085.8

79.5

69.2

87.2 85.8

05

01

00

(%)

Attendance Dropout Enrollment

Agree or Strongly Agree

UNICEF Non-UNICEF

Page 30: ENDLINE EVALUATION · CBO Community-Based Organization CBA Cost-benefit analysis CEA Cost-effectiveness analysis ... consistent with the approach that is promoted by the intervention,

30

standard deviation for UNICEF Solar schools and 5.3 percent of a standard deviation for non-UNICEF

Solar schools. The estimated effects of the Solar intervention, however, are associated with relatively wide

confidence intervals. We find no effect among schools that implemented both the Solar and CFS

interventions in the post-June 2014 periods. However, we find a statistically significant 21.98 student

increase in enrollment among schools that implemented both the Solar and C4D interventions. We estimate

that the Solar intervention had little to no effect on attendance rates among schools that received the

intervention from UNICEF as well as from non-UNICEF organizations. Similar to the effects seen in

enrollment, schools that implemented both the Solar and C4D interventions exhibit a 46.5 percentage point

increase in attendance rates relative to similar schools in the comparison group. Grade promotion rates are

largely unaffected by the Solar intervention with estimated effects ranging between -0.81 and .72

percentage points, these estimates are not statistically significant at the conventional levels.

Lastly, Figure 19 displays the marginal effects of the Solar intervention on student reported study frequency.

In this case, we estimate the change in the likelihood of students who report studying at least once a day

relative to similar students in control schools who did not receive the Solar intervention. We find that

students attending UNICEF Solar schools are 13.2 percent more likely to report studying at least once a

day than similar students in non-Solar schools. Students in schools that implemented both Solar and WASH,

and Solar and C4D are 8.8 percent and 14.8 percent more likely to report studying at least once a day than

their counterparts. These estimates are marginally significant at the 80 percent confidence level.

Figure 19 - Margins plot of Solar effect on frequency of studying

Challenges in implementation and function of solar power As with WASH facilities, maintaining the solar infrastructure presents a challenge to the sustainability of

the intervention. It is not sufficient to implement the panels. Schools and communities need to monitor and

maintain the infrastructure to ensure long term use. When the head teachers were asked about the challenges

in monitoring and maintaining the solar panel, three main challenges emerged in the UNICEF supported

schools: Insufficient funding (47 percent); No training on the maintenance of the panels (41 percent); and

service quality (25 percent) (see Figure 20). The results were similar in the schools supported by other

organizations, with higher percentages of head teachers reporting these three areas as a challenge.

Page 31: ENDLINE EVALUATION · CBO Community-Based Organization CBA Cost-benefit analysis CEA Cost-effectiveness analysis ... consistent with the approach that is promoted by the intervention,

31

Figure 20 – Solar challenges in implementation

On the whole, challenges in implementation and monitoring appear to be correlated with outcomes of

interest (percent of students who study at night) to a slightly higher degree than with the WASH challenges,

particularly in training in use and quality of service. This would seem to indicate that challenges in

maintenance of the solar interventions represent a serious challenge, and that UNICEF may consider more

regular support to these schools in improving the delivery of education.

Findings from focus groups The implementation of solar power was intended to help improve student’s safety and their ability to study

for longer periods of time. As the data has shown, the intervention is both cost-beneficial in the long term

and educators perceive that it does support improved learning. The results of the focus groups can be broken

down into three main themes: more time for teaching and learning; increased school safety; and indirect

benefits.

Increased time for teaching and learning The importance of having focused time to study is well documented in the literature (Bruns et al, 2014;

Destefano et al, 2011). In the ASAL schools, the contribution of solar power to increased time on teaching

and learning was mentioned approximately 90 percent of the time across the 23 focus group discussions.

Parents and community members noted that students are now able to study in the morning and in the evening

- “It [solar power] has helped because nowadays children can go to school early in the morning to do

morning studies but before the installation of solar it was impossible because of the dark. So it has helped

so much.” Parent FGD, Marsabit

They also noted that teachers have more time to do their prep work as well as work with students who board

at the school in the evening. Parents and community members noted that, “These days we have night preps

and also morning preps. The teachers who are residing in the school are also able to teach at night for

those children who are around or those who are in the boarding.”

Students in the FGDs agreed that having solar power increased their ability to study for extended periods

of time. Students also extensively discussed the ability to study at night or early in the morning; read text

more clearly at night as a result of having lights; and being able to study in various locations. When asked

how many additional hours of study time they believed having solar power provided, at least six students

indicated that they had about 2-3 hours of extra study time now available that previously did not exist. The

following quote captures the collective sentiment of students related to the issues of learning: “We see well

when reading our books so we are able to read in the evening even when it’s dark for a longer time than

before.” Student FDG, Tana River.

38.5 38.7

70.578.3

39.749.2

05

01

00

(%)

Initial Maintenance Use

Percent Reporting Challenges in Implementation

UNICEF Non-UNICEF

47.4

55.0

41.046.7

25.6 24.2

02

04

06

0

(%)

InsufficientFunding

No Training Service Quality

Solar: Challenges in Monitoring and Implementation

UNICEF Non-UNICEF

Page 32: ENDLINE EVALUATION · CBO Community-Based Organization CBA Cost-benefit analysis CEA Cost-effectiveness analysis ... consistent with the approach that is promoted by the intervention,

32

Finally, while the FGD do not provide concrete evidence that increased lighting leads to improved reading

results, several parents/community members did feel that having more lighting was directly linked to the

learning outcomes of the school. One of the parents from Marsabit noted, “It [solar power] has improved

student learning because last year’s result of Kenya Certificate Exams this school did better than all other

public schools - it was number 1 in Moyale sub-county, Marsabit County.”

School Safety Multiple counties in the ASAL region are currently facing issues related to school safety. Schools are

occasionally attacked by members of armed groups and communities often feel unsafe sending their

children to school. The issue of school safety was brought up as a theme throughout the 23 focus groups.

Parents, community members and students indicated that having solar power allowed the schools to ensure

there was sufficient lighting in the school compound, which they felt, worked as a deterrent to crime and

made students feel safer. One student in Samburu stated that, “Lighting has improved security as criminals

do not come in school.” Students also noted that having lighting kept students from getting hurt at school.

“At night when going to the dormitory, we are able to see our way clearly and avoid dangers or injury –

for example, there could be a snake passing, a thorn on the ground, or something else that could hurt us.”

Finally, students in the FGDs also noted that having lighting at the school contributed to better hygiene

practices. Several students commented that before the school had lighting, many students would just “go

around the corner” to relieve themselves at night. But now, having the lights around the latrine facilities

enable students to use the latrines at night and is helping to keep the school cleaner.

Indirect Benefits The focus group discussions re-emphasized the elements of adding solar power that one would expect –

increased time for teaching and learning as well as school safety. But there were two additional indirect

benefits that parents and community members mentioned during the discussions. First, a parent in Garissa

County noted that while their school was only a primary school, that in the evenings students from the

nearby secondary school would come over to study in the evenings. While the indirect benefit of having

students from other schools use the lighting was only mentioned in Garissa, it is an important extended

benefit that should be examined in a future study.

The second indirect benefit is linked to teaching and learning, but relates specifically to class 7 and 8

students. Solar lighting was added to class 7 and 8 classrooms. Parents, community members and students

all mentioned that having lighting in these classrooms specifically helped students in these grades study for

their exams by giving them a more isolated area to study.

Challenges of Solar power While the FGDs mainly focused on the benefits that communities feel have accrued to their students, there

were a few important challenges that parents raised during the discussions. First, having the resources to

sustain the solar panels was a challenge to communities. Several focus group participants mentioned the

lack of resources to fix and maintain the solar panels. This challenge was also re-enforced in the head

teacher survey. Second, parents in the Garissa FGD mentioned that one of the buildings with solar power

had burned down and had to be rebuilt, which the community was able to eventually finance and complete.

However, it was not clear from the discussion whether the fire was related to the solar infrastructure, or

some other reason. Finally, parents and community members mentioned that in the schools where there are

no security fences, the solar panels were often damaged, or destroyed by “criminals”. While this issue was

only mentioned two or three times, it is another challenge that UNICEF should consider and further explore

with communities.

Page 33: ENDLINE EVALUATION · CBO Community-Based Organization CBA Cost-benefit analysis CEA Cost-effectiveness analysis ... consistent with the approach that is promoted by the intervention,

33

Indicative Results of Child-Friendly Schools Interventions UNICEF characterizes a child friendly school as one that is inclusive, healthy and protective for all children,

and positively engaged with families and communities with the aim of promoting safe and healthy

educational delivery at the school level. Planned project interventions involve capacity development

training aimed at enhancing teacher skills in child-centered pedagogy that focus on notions of inclusiveness,

including gender sensitivity and sensitivity towards students with disabilities. The CFS approach includes

the following activities: (1) how to develop teaching and learning materials; and (2) how to improve school

management through the use of tools such as a CFS monitoring toolkit, action research, and a mobile phone-

based school data monitoring and evaluation system. This intervention attempted to establish children’s

governance groups to enhance student leadership, participation, and empowerment in school decision

making.

How effective are the CFS interventions for outcomes of interest? The UNICEF intervention is intended to boost attention and retention in CFS schools, and to improve the

application of student-centered instruction and time use in the classrooms. In addition to the regular head

teacher, student and school instruments, the CFS intervention was assessed using a classroom observation

instrument following the Stallings methodology. Similar to the WASH intervention, we are able to separate

schools that have been implementing the intervention for a longer period (prior to June 2014), compared to

schools that have been implementing for a shorter period and schools that have not implemented the

intervention. As is clear from Figure 21, the change in girls’ outcomes does not appear positive. Although

overall attendance and promotion rates are relatively high (see Annex D), the changes are only slightly

positive on attendance in schools implementing CFS for a shorter period, and there does not appear to be a

difference between schools that have been implementing the intervention for a longer period. As can be

seen in Figure 21, however, the impact on boys is mixed, with a clear positive effect for retention (for both

longer and shorter reported duration) and a negative effect with attendance rates.

Figure 21 – CFS: Changes in key outcomes

-13.3 -13.1

8.0

-20.0

0.0

20.0

% c

han

ge

Other CFSpre & post 6/14

CFSpost 6/14

Girls Attendance Rate Change, Grades 5-6

-9.2-11.2

-0.4

-20.0

0.0

20.0

% c

han

ge

Other CFSpre & post 6/14

CFSpost 6/14

Boys Attendance Rate Change, Grades 5-6

-1.8

-7.6-10.6

-20

.0

0.0

20

.0

% c

ha

ng

e

No CFS Prior Post

Girls Promotion Rate Change, Grades 5-6

-0.3

14.5

8.8

-20

.0

0.0

20

.0

% c

ha

ng

e

No CFS Prior Post

Boys Promotion Rate Change, Grades 5-6

Page 34: ENDLINE EVALUATION · CBO Community-Based Organization CBA Cost-benefit analysis CEA Cost-effectiveness analysis ... consistent with the approach that is promoted by the intervention,

34

Classroom Observation - Background To capture differences in student-centered instruction and time use at the classroom level, we implemented

a classroom observation tool following the Stallings methodology. The CFS intervention is designed to

promote changes in the style of teaching practice, with teachers trained in child-centered pedagogy, and

urged to move away from teaching through rote memorization. Classroom observations were conducted in

54 (17 percent) of the 321 schools reached in the endline (Figure 22). 13 of these schools received the CFS

intervention and 41 did not.

Figure 22 – Number of schools in which classroom observations were conducted15

Academic and non-academic activities The Stallings tool differentiates class time into that devoted to academic and non-academic activities. Figure

23 shows teacher engagement in different academic activities. Academic activities include copying,

debating, desk work, exposition and demonstration, practice and drilling, and reading out loud. Note that

these numbers do not add to 100 percent because not all of class time is devoted to academic activities. The

graph shows the percentage of class time (average) devoted to these different activities for CFS and non-

CFS schools. As can be seen, there are similar percentages across both types of schools in terms of the

percentage of class time devoted to each activity. Most class time is devoted to exposition and

demonstration (18-22 percent), debate (12-13 percent), followed by practice and drilling (6-8 percent), and

reading aloud (5-8 percent).

Exposition and demonstration includes lecturing and showing students how to do something new16. While

this activity is less likely to be done in a collaborative manner, debate is identified as an activity that

involves the collaborative exchange of ideas and discussion between teachers and students. While there are

not large differences between CFS and non-CFS schools in terms of class time devoted to academic

activities, both classes include a mix of these types of activities. Of particular interest, it appears that in

both CFS and non-CFS schools the proportion of class time devoted to copying is zero. This is important,

as an objective of the CFS intervention is to move teachers away from using rote memorization, and

stimulate active participation in learning among students.

15 The number of schools per county is provided in parentheses, for reference 16 The CFS training manual that was used can be made available upon request

16

12

8

6

4

3

3

2 (11)

(14)

(51)

(32)

(52)

(35)

(22)

(104)

0 10 20#

Turkana

Mandera

Garissa

Marsabit

Wajir

Samburu

Isiolo

Tana River

Page 35: ENDLINE EVALUATION · CBO Community-Based Organization CBA Cost-benefit analysis CEA Cost-effectiveness analysis ... consistent with the approach that is promoted by the intervention,

35

Figure 23 – Percentage of class time devoted to academic activities

Figure 24 – Percentage of class time spent on nonacademic activities17

Figure 24 shows the percentage of class time devoted to non-academic activities. Note that the total class

time between the two graphs does not add up to 100% because of a small number of missing observations

in one school. Classroom management is defined as activities that include passing out papers, cleaning the

board or putting away materials. A significant proportion of non-academic class time is devoted to

classroom management, although note that classroom management tends to be more collaborative in CFS

schools.

At the end of the classroom observation, enumerators were asked to make a series of evaluative judgments

about the frequency of CFS-related content delivered by teachers in the classroom. Table 3 shows the

percentage of class time that the teacher devoted to specific tasks. CFS schools tended to have a greater

percentage of class time devoted to activities that would be consistent with participation in a CFS training.

Teachers were more likely to ask probing and open-ended questions, to ask students to expand and explain

the meaning of concepts during class discussion, and to apply their understanding of topics to relevant life

experiences. The exception is with gender and disadvantaged students, where CFS schools were as likely

as non-CFS schools to discuss this content. Of note is that across all 54 schools, enumerators observed both

CFS and non-CFS classrooms did not discuss gender or other biases against disadvantaged students.

Table 3 - Highlights of the classroom observation Status Proportion of

Class Time Teacher…

...asks probing and open-ended questions such as how and why

...asks students to expand and explain in class discussion

...asks students to apply understanding to relevant life experiences

...talks about gender or other biases against disadvantaged students

...integrates gender-sensitive material into pedagogy

Non-CFS 1 or 2 times

29% 20% 24% - 12%

CFS 46% 38% 38% - 8%

Non-CFS About half 24% 20% 2% - 2%

17 Reading out loud does mean aloud and not loudly. Classroom management refers to time spent by the teacher

ensuring that that lessons are being delivered as planned and managing student behavior.

26

21

1513

7

11

78

2

6

01

02

03

0%

Exposition& Demonstration

Debate Copying Practice& Drilling

DeskWork

Non-CFS CFS

15

10

15

119

7

4

12

01

02

03

0%

ClassroomManagement

Alone

Teacher Outof Class

ReadingOut Loud

ClassroomManagement

Non-CFS CFS

Page 36: ENDLINE EVALUATION · CBO Community-Based Organization CBA Cost-benefit analysis CEA Cost-effectiveness analysis ... consistent with the approach that is promoted by the intervention,

36

CFS 8% 15% 8% - 0%

Non-CFS Entire

5% 5% 7% - 5%

CFS 15% 0% 0% - 8%

Non-CFS Did not observe

41% 56% 66% 100% 80%

CFS 31% 46% 54% 100% 85%

CFS regression analysis Following the same structure for the regression analysis, we assess the impact of the CFS intervention on

growth in student enrollment, attendance rates, and grade promotion rates. Similar to the analysis of WASH

schools, we are able to estimate the impact of the CFS intervention among schools that implemented the

intervention prior to and after June 2014, schools that only implemented the intervention after June 2014,

and schools that received the intervention from organizations other than UNICEF (some head teachers

reported receiving the CFS intervention from KEPSHA and/or the Kenyan Ministry of Education and not

through UNICEF). We further disaggregate CFS schools that implemented only in the post-June 2014

period to identify those that also implemented WASH, Solar, and C4D over the same period. Error!

Reference source not found. Figure 25 presents the marginal intervention effect across different

intervention statuses on growth in student enrollment, attendance rates, and grade promotion rates relative

to similar schools that never implemented CFS and did not receive any other intervention prior to June

2014.

Figure 25 – Margins plot of CFS outcomes Growth in Enrollment

Growth in Attendance Rates

Growth in Promotion Rates

Notes: Marginal effects associated with enrollment are expressed as changes in enrollment in response to the intervention, in counts.

Marginal effects associated with attendance and promotion rates are expressed as percentage point changes in response to the

intervention. Intervention statuses associated with fewer than 10 schools are suppressed.

Schools that implemented CFS in both the pre- and post-June 2014 periods display an increase in their

enrollment, relative to similar schools in the comparison group, by an average of 9.2 students (effect size

of 15.3 percent of a standard deviation). However, this estimate is not statistically significant. Schools that

implemented the intervention only after June 2014 exhibit a statistically significant 11.4 student increase

in student enrollment relative to similar non-CFS schools, equivalent to an effect size of 6.9 percent of a

standard deviation. Schools that implemented CFS through non-UNICEF organizations exhibited the

largest increase in student enrollment by 16.1 students, relative to comparable non-CFS schools. CFS

schools that also implemented other interventions displayed modest positive effects in some cases and

negative among schools that also implemented C4D. The CFS interaction effects are all statistically

insignificant.

Page 37: ENDLINE EVALUATION · CBO Community-Based Organization CBA Cost-benefit analysis CEA Cost-effectiveness analysis ... consistent with the approach that is promoted by the intervention,

37

We find mixed and insignificant results of the effect of CFS on attendance rates among all CFS schools

except those schools that implemented CFS and C4D where we see a 25 percentage point increase relative

to similar non-CFS schools. Lastly, we find small positive effects of CFS on grade promotion rates between

0.6 and 3.7 percentage points across all types of CFS schools, although these estimates are not statistically

significant.

Findings from Focus Groups One of the goals of CFS is to ensure that all children are included in the teaching and learning process and

feel that they have an opportunity to learn. For children with disabilities, the challenges of learning in

developing contexts can be even more challenging as they are often excluded from schools – or the physical

environment does not facilitate their school attendance. CFS is trying to change that. Parents, community

members and students discussed how their schools are working to integrate disabled students into the

learning environment. Several parent focus groups mentioned that UNICEF had improved WASH facilities

to facilitate their use by physically disabled students. Parents also noted that, “..there are students who

have been appointed to support the students with disabilities and they also help others appreciating them.

Like last week there was a function where this group of student volunteers was introduced to the

missionaries as care givers of the pupils with disability.

Parents also noted that, “Teachers also go out of their way to help the physically impaired children. They

even go to the extent of hiring motorbike using their money to take these students home. This has also

contributed to children coming to school.” Teachers also teach students to. “…accept them [disabled

students] and respect them the way they are. They should not laugh at them or abuse them. They now

understand them and consider them as part of the community.”

Indicative Results of C4D Interventions Communication for Development (C4D) has played an integral part in international development efforts

over the past five decades. For UNICEF, C4D is considered one of the core cross-cutting strategies essential

to advancing its program goals, particularly to ensure that the rights of children are realized. C4D aims to

build community capacity to meaningfully engage and participate in analyzing challenges and barriers and

identify priority solutions for education access and quality issues in schools. The approach to C4D

encompasses a range of interventions from advocacy to mobilization of communities to changes at both the

individual and community levels facilitated through many diverse communication channels and processes.

The intervention uses local media channels to communicate and promote understanding of the legal

framework of education governance. Community members were inducted in CFS principles and disaster

risk reduction (DRR), and engaged parents, teachers, and pupils together in school-based change.

Integrating C4D principles and approaches is a way to create an enabling environment for the empowerment

and engagement of individuals and communities.

How effective are the C4D interventions for outcomes of interest? The C4D intervention is designed to improve enrollment, attendance, and retention. In particular, by

targeting out of school children and promoting the importance of schooling through media and other forms

of communication to the wider community. The hope is that education can be seen as a more attractive

investment for families. As the C4D intervention implemented by UNICEF may be similar to community

participation interventions implemented by other organizations, we sought to compare interventions

implemented by UNICEF to similar ones implemented by other organizations. Table 4 – Changes in

enrollment, C4D status shows enrollment changes over the intervention period, and we construct

comparison groups based on whether a C4D intervention was implemented by UNICEF partner, a different

organization, or whether the school did not receive a C4D intervention. While schools without a C4D

intervention saw enrollment decreases, those implementing a C4D intervention saw enrollment increases

Page 38: ENDLINE EVALUATION · CBO Community-Based Organization CBA Cost-benefit analysis CEA Cost-effectiveness analysis ... consistent with the approach that is promoted by the intervention,

38

for boys. The enrollment level for girls in UNICEF C4D schools stayed almost exactly the same over the

period, while for boys it increased by 14%.

Table 4 – Changes in enrollment, C4D status C4D Status Girls Boys % change

2013 2015 2013 2015 Girls Boys

No C4D 327 286 343 322 -13% -6%

UNICEF 477 472 654 742 -1% 14%

Non-UNICEF 282 242 345 358 -14% 4%

UNICEF supported C4D schools also saw increases in attendance over the implementation period. Figure

26, on the following page, shows attendance rate changes for girls, which increased 14.1 percent points

over the period. Both non-UNICEF C4D interventions and schools without C4D interventions saw

decreases. Figure 26 illustrates a similar result where attendance rates increased by just under 10 percent

while the other two categories saw decreases of 9-10 percent for C4D interventions.

Figure 26 – Changes in attendance, C4D status

Retention does not show quite as positive a change across the intervention categories. Figure 27 shows

promotion rates for girls, which decreased over the intervention period in all three types of schools. The

promotion rate for boys actually decreased to the largest extent in UNICEF schools (-9.8%) compared to

1-4% drops in non-UNICEF C4D schools.

Figure 27 – Changes in promotion, C4D status

Figure 28 shows the percentage of schools that receive training in advocacy, community outreach and social

mobilization across different types of schools. UNICEF schools implementing C4D receive substantially

more training of this type compared to schools that have not received the C4D intervention.

-9.6

14.1

-1.2

-20

.0

0.0

20

.0

% c

ha

ng

e

No C4D UNICEF Non-UNICEF

Girls Attendance Rate Change, Grades 5-6

-9.1

9.8

-10.4

-20

.0

0.0

20

.0

% c

ha

ng

e

No C4D UNICEF Non-UNICEF

Boys Attendance Rate Change, Grades 5-6

-2.0

-11.9

-16.0

-20

.0

0.0

20

.0

% c

ha

ng

e

No C4D UNICEF Non-UNICEF

Girls Promotion Rate Change, Grades 5-6

-3.6

-9.8

1.1

-20

.0

0.0

20

.0

% c

ha

ng

e

No C4D UNICEF Non-UNICEF

Boys Promotion Rate Change, Grade 5-6

Page 39: ENDLINE EVALUATION · CBO Community-Based Organization CBA Cost-benefit analysis CEA Cost-effectiveness analysis ... consistent with the approach that is promoted by the intervention,

39

Figure 28 – Has received C4D training

Figure 29 shows schools that report a communication plan to attract out of school children. C4D schools

report these at a similar rate to other types of schools. Figure 29 shows the reported components of out of

school children communication plans. Posted messages at school and the community are the least (7.9%)

common forms, compared to media campaigns (13.2%), targeted conversations with out of school children

in the community (50.0%) and community meetings (68.4%).

Figure 29 – OOSC communication plan

Figure 30 shows the percentage of schools that reported engaging in advocacy and social mobilization

around school participation in the past year. As can be seen, UNICEF schools implementing C4D are the

highest at 87 percent, followed by non-UNICEF schools and those not reporting any C4D intervention.

18.9

71.2

42.6

05

01

00

(%)

Has Received Advocacy and Outreach Training

No C4D UNICEF Non-UNICEF

68.676.4 77.5

05

01

00

(%)

Has OOSC Communication Plan

No C4D UNICEF Non-UNICEF

7.9 7.913.2

68.4

50.0

05

01

00

(%)

UNICEF

OOSC Plan Components

School Community

Media Meetings

Conversations

Page 40: ENDLINE EVALUATION · CBO Community-Based Organization CBA Cost-benefit analysis CEA Cost-effectiveness analysis ... consistent with the approach that is promoted by the intervention,

40

Figure 30 – C4D: Engaged in advocacy or social mobilization

Figure 31 shows the percentage of C4D schools that believe C4D activities have had an impact on the

school. 94.6 percent and 83.8 percent believe that communication activities at the school have been

successful in improving stakeholder perceptions of the value of education, and have been successful in

increasing the intake of out of school children back into the school, respectively. Still, many believe that

the barriers to schooling are too great to be overcome with communication activities, with a higher

percentage (56.8 percent) in UNICEF schools compared to non-UNICEF ones.

Figure 31 – C4D: Head teachers’ assessment

C4D regression analysis The regression results for the C4D intervention are presented below in Figure 32. We differentiate between

schools that received the intervention UNICEF and from non-UNICEF partners. Similar to the analysis of

the Solar intervention, we do not observe enough schools who implemented the intervention prior to June

2014 to distinguish UNICEF C4D schools. We do, however, stratify C4D recipients among those schools

that received a combination of C4D and CFS.18 However, since there were a relatively small number of

schools with sufficient data that received the C4D intervention alone, we are unable to confidently estimate

the intervention effect for these schools. Further, the majority (22 out of 32 UNICEF C4D schools) received

C4D in conjunction with CFS, and more than half of all C4D schools received the intervention from non-

UNICEF organizations, which are the only two types of C4D schools that we are able to identify the

intervention effect for.

18 There were fewer than 10 schools implementing C4D and WASH and/or C4D and Solar, the results of which are suppressed

due to lack of statistical power.

59.6

87.0

75.0

05

01

00

(%)

Engaged in Advocacy or Social Mobilization

No C4D UNICEF Non-UNICEF

83.892.3

56.847.5

94.6 94.9

05

01

00

(%)

Attendance Barriers Perceptions

C4D: Agree or Strongly Agree

UNICEF Non-UNICEF

Page 41: ENDLINE EVALUATION · CBO Community-Based Organization CBA Cost-benefit analysis CEA Cost-effectiveness analysis ... consistent with the approach that is promoted by the intervention,

41

Figure 32 – Margins plots of C4D outcomes Growth in Enrollment

Growth in Attendance Rates

Growth in Promotion Rates

Notes: Marginal effects associated with enrollment are expressed as changes in enrollment in response to the intervention, in counts.

Marginal effects associated with attendance and promotion rates are expressed as percentage point changes in response to the

intervention. Intervention statuses associated with fewer than 10 schools are suppressed.

We find a positive and significant effect of C4D among non-UNICEF implementers on growth in student

enrollment, by about 16.6 students (effect size of 24.7 percent of a standard deviation), while UNICEF C4D

schools that also implemented CFS exhibit a negative but insignificant effect relative to similar comparison

schools. On the other hand, we find a statistically significant 27.6 percentage point increase in attendance

rates among C4D schools that also implemented CFS relative to comparison schools, while non-UNICEF

C4D schools do not show any statistically significant improvements. Finally, we estimate small and

statistically insignificant increases in grade promotion rates among all C4D schools relative to the

comparison group.

Findings from Focus Groups During the FGDs, parents, community members, and students were asked to discuss strategies, training,

and activities related to helping get out of school children back in school. The findings from these FGD

varied greatly from one school to the next and should be taken with caution. The main themes arising in the

community and parent focus groups related to C4D include: the role of the school committees and student

government at the school; and the extent to which UNICEF was recognized for the training and activities

taking place in the schools.

Community activities Parents and community members who participated in the FGDs clearly understood the importance of having

students attend school. However, the actions taken by communities and their understanding of best practices

for recruiting and retaining children in school varied widely across communities. Parents talked about how

the school committees travel out to remote areas and talk to parents about the importance of education.

They also mentioned that the school director and members of their BoMs had participated in meetings that

focused on strategies for recruiting out of school students. A parent in the Isiolo focus group captured the

role parents understand that they should be playing through the following quote, “… the school committee

sometimes goes round the village telling women that the government want every child to go to school. Also

as a committee member you are asked to pass this information to all parents in the village and that is what

we usually do.”

They also noted that, “we [the SMC] help by reporting all the children that have dropped to the chief. We

also follow-up with the parents to know why their children are not going to school and insist that they

should take them back to school.”

Page 42: ENDLINE EVALUATION · CBO Community-Based Organization CBA Cost-benefit analysis CEA Cost-effectiveness analysis ... consistent with the approach that is promoted by the intervention,

42

Parents also indicated that they have been instructed (i.e. trained by either the chief or someone else) – and

often meet with the teachers to understand how students are performing in school and help the school

improve attendance by visiting the homes of students who fail to come to school. These activities are all

positive and in line with the types of activities that the C4D training hopes to instill in communities.

It was clear from the discussions that the Chief plays a significant role in organizing and ensuring that

children go to school. Parents and community members tended to work with and through this person, which

aligns with the culture and practices of the nomadic tribes. However, it was less clear whether the techniques

used by the communities were also in line with what UNICEF and other organizations teach in terms of

recruiting new students. One disconcerting theme that arose in several focus groups was the idea that parents

are arrested and prosecuted for not sending their children to school. A parent in Isiolo noted, “We were told

that if a parent does not take her children to school it’s an offence and [they] can be arrested and

prosecuted. This has made the number of children in this school increase very much and this is due to the

fear of being arrested and also because parents have now know that there is a law that protect the children.”

The context of the discussion centered around parents discussing the rights of children, but in many cases

– and across several communities, the issue seems to have been taken to the extreme. UNICEF and other

organizations working on awareness and mobilization need to understand the extent to which these type of

fear tactics are used and focus their training to ensure communities use more positive recruiting methods

moving forward.

C4D training workshops It was clear from the focus group discussions that communities – particularly the head teacher and chief

had participated in meetings and training workshops that improved their knowledge and awareness of the

importance of education. However, it was less clear the extent to which any of the activities can be traced

back to UNICEF support verses other organizations. During the focus group discussions, parents and

community members were asked who provided the community mobilization training and whether they were

familiar with the Elimu Yetu Coalition EYC). No participants recognized EYC as having provided

training,19 nor did they attribute the training and activities to UNICEF. Rather, the communities referenced

groups such as Youth trust, TUSOME and the government as the organizations most linked to improving

their knowledge and awareness around the importance of education. Additionally, the results from the focus

groups around the C4D topic were fairly generic. Parents and communities rarely specified activities

beyond going to homes and talking to parents about getting their children in school. There was no mention

of radio messages, community advertising/marketing or additional local training workshops – even when

participants were specifically asked to elaborate on other types of activities the community engaged in

throughout the year.

Raising community awareness does take time and parents noted that, “… we have just started [the process]

last year - you remember we are nomads and it is very hard for us to take the messages at first, so it’s better

if we are given time to catch up because there’s a lot of illiteracy in our community so it not the way you

expect the results very fast, they still need time.” So while the results of this evaluation appear to be less

promising for C4D in this first year, the types of activities and the length of time and intensity of training

required for strong uptake likely affect our ability to see any results. The FGDs seem to indicate that training

was not consistent, short in duration and that there was no on-going support to the communities for

conducting different types of activities in support of education. As with some of the previous interventions,

UNICEF needs to look carefully at the training approach used to improve community awareness and

perhaps strengthen both the delivery in communities and on-going support for implementation. The issue

of duration and dosage are further elaborated in the discussion section.

19 The primary target of C4D training was the head teacher.

Page 43: ENDLINE EVALUATION · CBO Community-Based Organization CBA Cost-benefit analysis CEA Cost-effectiveness analysis ... consistent with the approach that is promoted by the intervention,

43

COST-BENEFIT ANALYSIS Cost-benefit, cost effectiveness and other economic analyses are often used to guide policymakers and

others interested in understanding the types of interventions that have the greatest impact on specified

outcomes (Flay et al., 2005; Spoth, Greenberg, & Turrisi, 2008). Cost-benefit analysis allows

policymakers to determine if their investments are sounds; and provides a basis for comparing the results

to other alternatives to determine which of the investments provides the greatest returns. In this study we

present the cost-benefit analysis of three UNICEF interventions: WASH, Solar power, and CFS – each of

which [together and individually] seek to help retain students in school. We examine whether the

interventions, when implemented in low cost boarding schools in the ASAL region of Kenya, are a good

investment of donor dollars.

Interventions such as WASH, solar power and CFS attempt to address issues of enrollment, retention and

completion in schools. As this report has shown, the UNICEF interventions have had different levels of

success when compared to schools that had similar interventions by other organizations – or no

interventions. While this evaluation looks at results after only one year, over time, it is expected that

UNICEF will see greater improvements in enrollment and retention. The following analysis will show the

length of time that it will take UNICEF to attain returns on their investment. It is important to note that the

cost-benefit model (CBA) that is presented in this section serves as a means of modeling the potential long-

term benefits that may accrue to society based on the potential for better future employment. Caveats and

limitations of these findings are discussed at the end of the section.

Assumptions To build this investment ‘simulation’ model, we made the following simplifying assumptions.

1. Intervention costs are provided to us by each implementing organization. The benefit-cost analysis

is conducted separately for WASH, Solar power, and child friendly schools to determine the

monetary returns to each intervention.

2. We do not include the costs of running the schools since these are provided by the government and

essentially equivalent across the ASAL schools.

3. To determine the number of students and schools affected by each intervention, we rely on our

enrollment regression results for the WASH, Solar power, and CFS interventions. The analysis

examined 31 WASH only schools, 23 schools that received solar power, and 48 schools that

implemented CFS. This analysis does not include C4D schools due to the lack of positive results

observed in those schools, i.e. we found no measurable benefits to enrollment as a result of C4D.

Table 5 – Estimated enrollment benefits for WASH, Solar, and CFS Cell size Marginal effect Standard error

WASH 31 +22.99 (10.36)

Solar 23 +6.99 (16.32)

Child friendly schools 48 +10.66

66

(7.97)

Note: Cell size refers to the number of treated schools included in the regression analysis to identify

the marginal effects.

4. Benefits to society are calculated based on the net present value of increased salaries and tax

revenue paid to society over time in 2015 US Dollars (USD) using a standard CBA approach (Levin

and McEwan, 2001). Specifically, the benefits are the net present value of those taxes paid due to

increased earnings as a result of increased educational attainment over and above the taxes they

would have paid from not benefitting from the interventions. We group students into the following

Page 44: ENDLINE EVALUATION · CBO Community-Based Organization CBA Cost-benefit analysis CEA Cost-effectiveness analysis ... consistent with the approach that is promoted by the intervention,

44

categories: No education (i.e. dropped out prior to grade 5), attained grades 6 through 8, and

completed grade 8.20

5. The mean wages used to calculate the tax gain to society are drawn from the World Bank STEP

surveys. We use these wage data to construct the age-earnings profiles of individuals with different

levels of educational attainment. We assume that students begin to earn wages at age 18 (3 years

after the intervention year)21. The age-earnings profile data enable us to compute total lifetime

earnings for different types of individuals in Kenya. As Figure 33 below shows, the returns to

education to those students completing grades 9-12 is approximately 69 percent. For those students

who complete grades 9-12 their annual earnings start at nearly USD 2,007 at age 18 and continue

to increase over the next 36 years. For students who attain grades 6-8, earnings typically start at

USD 978, reach a peak of USD 1,634 when the student is in their mid to late 30s and then declines

over time. For students with no education, the line remains relatively flat over their life time.

Figure 33 – Age-earnings profile by educational attainment – Kenya, 201322

Source: The STEP Skills Measurement Program 2013, the World Bank.

6. Enrollment data for students in intervention schools are not currently available for each individual

grade level. We measure the benefits for the current cohort of 5th and 6th grade students over their

labor market life-cycle. Further, we use the following grade level estimates to calculate the number

of students that move on to the higher earning brackets.

Table 6 – Grade promotion, dropout, and repetition rates – Kenya, 2014 Promotion rate Dropout rate Repetition rate

Grade 6 89.9% 5.6% 4.5% Grade 7 78.7% 7.8% 13.6% Grade 8 75.2% 2.0% 23.1% Grade 9 100.7% 0.6% -1.3%

20 Refer to the Annex for a detailed description of the net present value calculations. 21 We acknowledge that students who drop out of the education system are likely to work in the informal sector, or as herdsmen,

helping their parents, but the income is minimal [or in lieu of parental income] and therefore not included in the analysis.

22 The Kenyan Shilling to US Dollar exchange rate used is 102.17 KES to the dollar.

Page 45: ENDLINE EVALUATION · CBO Community-Based Organization CBA Cost-benefit analysis CEA Cost-effectiveness analysis ... consistent with the approach that is promoted by the intervention,

45

Grade 10 99.8% 1.2% -1.0% Grade 11 90.1% 2.3% 7.5% Grade 12 -- 3.2% -- Source: Calculations made by FHI 360 Education Policy and Data Center, using data from the

Demographic and Health Surveys in Kenya, 2009.

Based on the mean enrollment in the WASH and Solar schools and the promotion rates in Table 7, we

estimate the following number of students to have benefitted from the interventions directly. We show that

713 students attain at least grade 6 as a result of WASH, 161 students in Solar, and 512 students in CFS

who would not have otherwise.

Table 7 – Increased enrollment and grade attainment, by intervention WASH Solar CFS

Number of schools 31 23 48

Mean enrollment 114.9 116.9 124.1

Total enrollment 3,562 2,689 5,957

Total students affected +713 +161 +512

Students progressed through grade 8 +334 +75 +240

Students completed grade 8 +379 +86 +272

Results of the benefit-cost analysis Below, we plot the cumulative lifetime benefits associated with the WASH and Solar interventions under

different discount rate assumptions. We incorporate different discount rates in increments of two percentage

points to provide a range of possible net present values of the cumulative lifetime benefits of the

interventions. The benefits are presented in the form of the future tax payments made as a result of the

interventions over and above the tax payments from a scenario where the intervention never took place.

Table 8 displays the total and per pupil cost of each intervention in US Dollars.

Table 8 – Costs per school, per student, and total, by intervention

WASH Solar CFS

Total intervention cost $1,646,486 $315,582 $231,424

Total cost for schools in the analysis $505,357 $111,667 $112,206

Cost per school $16,302 $4,855 $2,338

Cost per student $142 $42 $19

We plot the cumulative benefits of WASH, Solar, and CFS separately, over the 36 years post-

implementation in Figure 34. The time horizon for the cost-benefits simulation represents the time needed

for the current student beneficiaries to be of employment age as well as the time covered by their labor

market lifecycle, i.e. from ages 18 through 50 years.

Page 46: ENDLINE EVALUATION · CBO Community-Based Organization CBA Cost-benefit analysis CEA Cost-effectiveness analysis ... consistent with the approach that is promoted by the intervention,

46

Figure 34 – Cost-Benefit simulation results – NPV of cumulative taxes paid over lifecycle WASH

Solar

Child Friendly Schools

Note: The horizontal red line represents the total intervention costs for the schools present in the analysis, i.e. 31 WASH schools,

23 Solar schools, and 48 child friendly schools.

Based on this simulation, it is clear that the cost-benefit simulation with the lowest discount rates yields the

highest benefits, and vice versa. However, we can see that over an entire lifecycle the WASH intervention

may generate between USD 602,000 and USD 1,944,000 in benefits, the solar intervention may yield

between USD 136,000 and USD 440,000 in benefits, and the CFS intervention is expected to generate

between USD 432,000 and USD 1,395,000 in benefits.

Using the results shown in Figure 34Error! Reference source not found., we compute the number of years

required for each investment to be returned in full, i.e. the number of years required for the benefit-cost

ratio to reach 1. For WASH, the initial investments will be completely returned by between 11 years and

21 years, depending on the discount rate used. The Solar investments will be returned in full after 11 to 20

years following implementation of the intervention. Finally, the simulations show that the CFS investments

yield the highest returns, relative to their initial investments. Depending on the discount rate, CFS will

return all of its initial investment within 5 years of implementation.

Over an individual’s labor market lifecycle, the WASH intervention yields a benefit to cost ratio of 1.19 to

3.85 per US Dollar invested, the lowest benefit to cost ratio denotes the highest discount rate of 10 percent

while the highest ratio denotes the lowest discount rate of 2 percent used in the simulation model. The

investments in Solar power produce a benefits to cost ratio ranging from 1.27 to 4.17 per US Dollar invested

over a lifetime. Lastly, investments made in the CFS intervention yield the highest benefit to cost ratios

over a lifetime between 4.0 and 12.5 per US Dollar invested.

Table 9 – Lifetime benefit-to-cost ratios per intervention, by discount rate23 Discount rate r = 2% r = 4% r = 6% r= 8% r = 10%

WASH 3.85 2.70 2.00 1.52 1.19

Solar 4.17 2.86 2.13 1.61 1.27

Child friendly schools 12.50 9.09 6.67 5.26 4.00 Note: the benefit-to-cost ratios are calculated as the ratio of cumulative benefits after 36 years

over the initial cost of the intervention for the schools included in the simulation model.

Limitations of the Benefit-cost model The benefit-cost analysis has four main limitations that restrict inferences about economic benefits. First,

the analysis assumes that the infrastructure remains intact and that schools have the required resources to

repair and refurbish the WASH and solar power infrastructure as needed over time. If schools do not invest

23 For the full list of benefit-cost ratios over time and for each intervention under different discount rate scenarios, please refer to

the Annex.

Page 47: ENDLINE EVALUATION · CBO Community-Based Organization CBA Cost-benefit analysis CEA Cost-effectiveness analysis ... consistent with the approach that is promoted by the intervention,

47

in the upkeep of the facilities, it is likely that they will depreciate and would likely be an associated decline

in enrollment and retention.

A second limitation is that these estimates are based on projected versus actual benefits as well as drawn

from national surveys rather than earnings data specific to the ASAL region and the counties included in

this study. We estimate lifetime beginning at age 18 and assuming that the students work continuously over

the next 36 years. Although this is a standard approach, actual earnings from communities in these counties

may provide a more accurate accounting of economic status and increased tax revenues.

The third study limitation is that there are likely to be unmeasured economic benefits. For example, fringe

benefits were not estimated as part of lifetime earnings (i.e. health insurance). Benefits for parents, other

family members, or the community more broadly were not estimated.

Finally, the estimations assume a fixed number of students that move through the system with different

levels of education. The estimates do not account for future cohorts of students or additional positive

changes in enrollment and retention that would impact the overall benefits and reduce the time it takes for

UNICEF to gain back its initial investments. This final assumption implies that the results of our simulation

model are conservative and may serve as a lower bound given our model assumptions.

Concluding remarks for cost-benefit analyses

The cost-benefit analyses show that, over the long-run and regardless of the discount rate used, investing

in child friendly schools yields the highest returns in the form of additional taxes paid over a lifetime. Every

shilling invested in CFS is expected to yield a return between 4.0 and 12.5 after 36 years. Further, CFS

investments are returned in full faster than investments in WASH or Solar power. Our simulation model

predicts that the initial investments made in CFS will be returned in full after 4-5 years since implementing

the intervention. As for WASH and Solar power interventions, the lifetime return yields are between 1.19

and 4.17 for each shilling invested. It is important to note that initial investments are returned in full between

10 and 21 years post-implementation.

Although it appears that the initial investments in WASH, Solar power, and CFS are returned after a

substantial period of time, it is important to note that this simulation model includes the benefits from only

one cohort of 5th and 6th graders that we were able to observe. Ideally, with more data, we would observe

multiple cohorts affected by the same initial investment before the need to reinvest to update and maintain

current physical capital (WASH facilities, solar panels, etc.) as well as human capital (professional

development and training for new teachers). As such, this simulation model provides UNICEF with an

informed notion of the relative cost-efficiency of each investment under current data constraints.

DISCUSSION This section will draw on the results to discuss the operational hypothesis that higher duration and dosage

of interventions (or specific combinations of interventions) have a greater impact on outcomes (i.e.

enrollment, attendance, study time, outreach). The section begins with an elaboration of our dosage,

duration and enabling environment framework. It is followed by a (1) discussion of whether the duration

and dosage of training was sufficient to begin to affect changes in behavior among communities, parents

and teachers; (2) the extent to which the enabling environment contributed to differences or similarities in

findings; and (3) whether and how different combinations of interventions impacted the outcome variables.

Understanding Dosage, Duration and Enabling Environment When scaling up education reforms, implementing an intervention in thousands, even millions of

classrooms where teachers may lack the proper skills is a different kind of challenge (Thompson and

William, 2007). The challenges of successfully implementing education reforms and seeing an impact on

Page 48: ENDLINE EVALUATION · CBO Community-Based Organization CBA Cost-benefit analysis CEA Cost-effectiveness analysis ... consistent with the approach that is promoted by the intervention,

48

educational outcomes include the sheer number of classrooms, the complex system in which these

classrooms often function (i.e. different environments even within the same country), the separateness of

the classrooms, the private nature of teaching, and the type of pedagogical support teachers receive on a

regular basis (Thompson and Wiliam, 2007). For an intervention to show impact, it means that every

teacher, every head teacher, and every parent and community member has to understand and commit to the

intervention, on their own. Projects can have great interventions and theories of action – but the enabling

environment to move an action plan forward can reduce the impact of interventions and slow progress. The

projects can be well designed, but if the teachers cannot deliver the interventions with a high level of quality

with the appropriate intensity, then the implementation effort is wasted (Thompson and Wiliam, 2007).

This is the challenge we face in closing the learning gap.

In 2014, the Basic Education Coalition Working Group for Monitoring and Evaluation wanted to

understand whether different combinations of interventions had more or less of an impact on improved

reading outcomes. Using data from a multitude of USAID-funded programs, the working group devised a

framework for analyzing the effect sizes of these interventions with the purpose of being able to inform

expectations of what it really takes to improve reading outcomes, particularly at scale. This framework,

comprised of three components, is used to frame the discussion around the findings of this evaluation.

Duration Duration is defined as the time period of implementation, or how long the intervention is applied. Similar

to the concept of dosage (below), duration refers to the exposure of an intervention over time – or, the

length of time that interventions are provided to participants. Through our analysis, we were able to separate

out schools that had longer periods of the interventions (pre June 2014) and compare the outcomes at those

schools with the results from schools that had only post 2014 interventions. The results for the WASH and

solar schools are discussed below. The results for CFS and C4D are discussed in the section on dosage.

Schools that received the WASH intervention in both the pre- and post-June, 2014 periods experienced an

average increase in student enrollment of 23 relative to similar schools that had never received WASH and

did not receive any interventions prior to June, 2014. We also saw an increase of 15 students, on average,

among non-UNICEF WASH schools relative to similar schools in the comparison group. This finding tells

us that WASH does have some relationship to increases in student enrollment. Interestingly, for the schools

that have had WASH implemented only recently, there was almost no effect on enrollment, suggesting that

duration is an important factor in affecting enrollment and attendance outcomes.

As a comparison, the introduction of a latrine/Wash interventions in India increased schools’ eighth-grade

student enrollment by only 1.5 students, with a standard error of 0.755 (Aduki, 2014). A review of eleven

articles by Joshi and Amandi (2013) also showed that the child’s gender, age, and access to water, sanitation

and hygiene training over an extended period of time were positively associated with education outcomes.

The positive impact of WASH interventions on school enrollment in 42 primary schools in Kenya after a

year of interventions is also supported by Patel et al. (2012).

In terms of the solar power interventions, it is important to note that we had only three schools that

implemented solar power both pre and post 2014, so the results cannot be extrapolated to other schools.

However, in the three schools that had the longest duration of the intervention, the schools exhibited an

increase of about 40 students compared to schools that had not received solar power.24 We also note that

schools that implemented both Solar and C4D exhibited significant growth in enrollment by 22 students.

24 In the final regression analyses of Solar power interventions, we consolidate schools that received Solar power from UNICEF

regardless of duration to maintain statistical power.

Page 49: ENDLINE EVALUATION · CBO Community-Based Organization CBA Cost-benefit analysis CEA Cost-effectiveness analysis ... consistent with the approach that is promoted by the intervention,

49

Dosage Dosage is the frequency and time that is spent implementing technical interventions. Implementation

dosage refers to the intensity with which interventions are carried out with fidelity. This intensity includes

the amount of training that participants receive in preparation to deliver interventions; the amount of time

that coaches spend working with teachers on an intervention; or the amount of time teachers receive training

on the use of new materials (OPRE, 2013), Understanding what dosage of an intervention is needed is

important because it affects the outcomes as well as the costs, staffing, replication and scale-up efforts that

follow.

The existing research on dosage is clear that one dose of an intervention is usually not enough (Boller et

al., 2004; Winton, 2008; Joyce & Showers, 1980). However, more is not always better (OPRE, 2013). As

shown by Boller et al. (2004); Joyce & Showers (2002); Raikes, et al. (2006); and Winton & McCollum

(2008). One-day workshops for teachers do not provide the necessary dosage to affect teacher learning or

change and improve their delivery of classroom practices in the long term. Professional development

interventions specifically, need to be delivered more intensely, usually with longer duration or frequency

to make a difference (Halle, Zaslow, Tout, Starr, Wessel, & McSwiggan, 2010). However, there is little

research that sheds light on the optimal amount.

In Kenya, the duration of training workshops and its effect on CFS uptake seems to be linked, but not

necessarily in the way we would have expected. As previously discussed in this report, regression results

show that schools that implemented CFS in both the pre- and post-June 2014 periods (i.e. a longer duration)

display an increase in their enrollment by an average of nine students. Schools with lower duration of

implementation (i.e. post June 2014 only), exhibit a statistically significant 11 student increase in student

enrollment relative to similar non-CFS schools. When we look at the dosage of training, the teachers in the

UNICEF schools received a higher dosage of general professional development training (nine days)

compared to the non-UNICEF schools (seven days) and the schools with no CFS interventions (eight days).

Similarly, when we looked at the reported number of days of training on CFS specific content, teachers at

the UNICEF-supported schools reported receiving eight days of training compared to five days in the non-

CFS schools and seven days in schools that received CFS training from another organization. The standard

errors in both cases ranged between 0.4 and 0.6 – similar to what the literature has found in other training

cases.

So does the dosage and duration of training matter? In the case of CFS in Kenya, it is difficult to say after

only one year. It is clear, however, that teachers received more than one day of training, and that seems to

have had a positive effect on enrollment and attendance. The schools that had CFS training also had higher

levels of enrollment and attendance when compared to schools that had no CFS training. Additionally,

FGD, student surveys, classroom observations and head teacher surveys further suggest that the schools

where CFS has been implemented (whether by UNICEF or someone else) are more welcoming and there

is a positive perception about the environment (learning, security, and inclusion) for all children. These

items all suggest that the dosage and duration of the interventions is perhaps sufficient, though we would

have expected the schools that had received two doses of training (pre and post) to have had the highest

enrollment and attendance results. This result likely indicates that other factors such as the enabling

environment, teacher migration and attendance or other factors may also be influencing the results within

the schools that have received CFS training.

In terms of the C4D interventions, anecdotal information obtained through FGDs and surveys suggest that

there is a lot of inconsistency in how the C4D interventions were implemented. While the C4D schools that

received the intervention from UNICEF in conjunction with implementing CFS had higher attendance rates

than similar non-C4D schools by approximately 27 percentage points. We further found that the growth in

enrollment over the course of the year was only statistically significant in schools that received the

intervention from non-UNICEF organizations (i.e. the length of training data is not available to FHI360).

Page 50: ENDLINE EVALUATION · CBO Community-Based Organization CBA Cost-benefit analysis CEA Cost-effectiveness analysis ... consistent with the approach that is promoted by the intervention,

50

What we learned through the FGDs lead us to believe that the dosage and duration of C4D was not sufficient

to see impacts on student outcomes. The training, done through a cascade model, helped communities

understand the importance of education, but it is less clear that the communities learned the different skills

needed to properly mobilize and help retain students in school (e.g. use of fear tactics, no radio or public

messaging mentioned in FGD). It is likely that interventions such as C4D need more time and support in

the implementation phase to ensure that communities are able to implement the interventions according to

the design (fidelity to implementation) and use different types of activities depending on the contexts to

influence community perceptions on education.

Enabling environment An enabling environment is defined as a set of interrelated conditions (i.e. political, institutional, technical,

and cultural) that impact the capacity of actors (i.e. teachers, parents, communities, donors, and

governments) to engage in development processes in a sustained and effective manner. The challenge often

faced in implementing education reform and taking the reform to scale is that the existing situation in a

country is not an accident. There are often well-entrenched interest groups who want to ensure that

institutional arrangements do not change since they often directly benefit from the arrangement (Crouch

and DeStefano, 2006; Moe 2003; Hess, 2004). Often proposed school level reforms or pilot programs are

seen as only affecting a small part of the system, so these groups do not feel threatened. However, when

these changes (albeit policy or scale up of interventions) begin to change the resource allocations, interest

groups who may not benefit from the change respond – and often forcefully (Crouch and DeStefano, 2006).

In the Kenya ASAL region there are several “environmental” issues that are also at play. While we are not

able to statistically measure those at this time, the issues are worth mentioning because they likely affected

the outcomes of this evaluation in different ways.

Security and Politics In several of the ASAL counties (i.e. Garissa, Mandera, and Wajir) security issues have plagued both the

implementation of UNICEF interventions as well as data collection for this impact evaluation. Certain

armed groups are known in Kenya for attacks on various public areas, including schools and universities.

The activities and threats from these groups delayed implementation of several UNICEF interventions

(reducing the duration of the interventions) and prevented the data collection teams from reaching

approximately 34 schools in Mandera, Garissa and Wajir.

In the best of circumstances, education reform efforts need stable environments and support from the

political (government), institutional (teachers, head teachers, inspectors) and technical (donors and

technical assistance) agents to succeed. Through discussions with parents and community members as well

as meetings with the implementing organizations and data collectors, we know that security issues and fear

of attack have affected the ability of students to attend school. Fear of attacks have increased parents

wariness to send their children to school; the lack of fencing – or even long distances students and teachers

may have to walk have affected their desire and ability to attend the school; and the danger associated with

the unrest made it difficult for data collectors to reach the schools to even see if the interventions had made

a difference. These security issues greatly impact and work against creating an enabling environment for

learning and improvement – and thus, must be taken into consideration when understanding the results of

this evaluation.

The Drought In many cases, understanding the enabling environment means that we look at different political,

institutional and technical aspects of interventions and try to understand who the winners and losers are –

and how each group may have affected the results. But, sometimes, natural disasters also affect the

environment and can impact the results we see in schools. Throughout the focus group interviews, parents

and community members mentioned the challenges the region has faced as a result of the drought. While

Page 51: ENDLINE EVALUATION · CBO Community-Based Organization CBA Cost-benefit analysis CEA Cost-effectiveness analysis ... consistent with the approach that is promoted by the intervention,

51

the communities recognize that organizations such as UNICEF have worked hard to improve or establish

water sources at the schools, many communities noted that bore holes have dried up, or that the tanks at the

schools did not have sufficient water (i.e. community use vs. student use) as a result of the drought. While

the evidence is anecdotal, the lack of water may have also reduced the impact of the interventions during

this year. Families have lost their herds and need children to work in the informal sector; there is no water

at the school so students stay home; and in schools where there is piped water, communities often share the

water source possibly reducing the availability of water for students.

Community Commitment In closing, it is important to note that many of these counties and schools do have a positive enabling

environment when it comes to education. Throughout the FGDs we saw an incredible commitment of the

parents and community members to their school. They recognized the importance of education. They were

trying to use strategies they had been taught to bring more children into the schools; and the chiefs were

committed to ensuring all children received education. This enabling environment creates an opportunity

for UNICEF to continue to work with and through these communities to implement hygiene, CFS and C4D

training. There is a base of institutional capacity (i.e. people power) that has been built and in many counties

a great deal of political will (i.e. support from the chiefs, elders, parents) to improve education. More

training and direct support over a longer period of time is needed to ensure that communities implement

CFS and C4D properly, but the foundation has been laid and UNICEF can build on that enabling

environment.

Page 52: ENDLINE EVALUATION · CBO Community-Based Organization CBA Cost-benefit analysis CEA Cost-effectiveness analysis ... consistent with the approach that is promoted by the intervention,

52

REFERENCES Adukia, A. (2014). Sanitation and Education. Harvard University

Ashish Joshi and Chioma Amadi (2013). Impact of Water, Sanitation, and Hygiene Interventions on

Improving Health Outcomes among School Children. Center for Global Health and Development, College

of Public Health, UNMC, Omaha, USA

Birdsall, N., Levine, R. & Ibrahim, A. (2005). Towards Universal Primary Education: Investments,

Incentives, and Institutions. Sterling, Va. Stylus Publishing.

Chapman, D.W. and Quijada, J. J., (2008). An analysis of U.S assistance to basic education in the

developing world, 1990-2005, International Journal of Educational Development: 29(3), 268-280.

Crouch, L. & DeStefano, J. (2015). A Practical Approach to In-Country Systems Research. This paper was

prepared for the Research on Improving Systems of Education (RISE) Program Launch Conference,

Washington, DC, June 18–19, 2015.

Education for All Global Monitoring Report (2014). Teaching and Learning: Achieving Quality for All.

UNESCO Publishing.

Flay BR, Biglan A, Boruch RF, Castro FG, Gottfredson D, Kellam S. Standards of evidence: Criteria for

efficacy, effectiveness and dissemination. Prevention Science (2005); 6:151–175.

Hanushek and Woessman (2012). Do better schools lead to more growth? Cognitive skills, economic

outcomes, and causation. Journal for Economic Growth. 17:267–321

Levin, H. and Lockheed, M. (1991). Effective Schools in Developing Countries. International Bank for

Reconstruction and Development, The World Bank.

Lockheed, M. & Verspoor, A. (1992). Improving Primary Education in Developing Countries. Oxford

University Press.

Ministry of Education, Science and Technology. (2014). Basic Education Statistical Booklet. Retrieved

from http://www.mygov.go.ke/.

Patel, M., Julie R. Harris, Patricia Juliao, Benjamin Nygren, Vincent Were, Steve Kola, Ibrahim

Sadumah, Sitnah Hamidah Faith, Ronald Otieno, Alfredo Obure, Robert M. Hoekstra, and Robert Quick.

(2012). Impact of a Hygiene Curriculum and the Installation of Simple Handwashing and Drinking Water

Stations in Rural Kenyan Primary Schools on Student Health and Hygiene Practices. American Journal of

Tropical Medicine and Hygiene. 87(4): 594–601.

Spoth R, Greenberg M, Turrisi R. (1980) Preventive interventions addressing underage drinking: state of

the evidence and steps toward public health impact. Pediatrics 121(Suppl 4): S311–336.

Stallings, J. (1980). Allocated Academic Learning Time Revisited, or Beyond Time on Task.”

Educational Researcher: 9 (11): 11–16

Thompson, M., Wiliam, D. (2007). Tight but Loose: A Conceptual Framework for Scaling Up School

Reforms. Paper presented at the annual meeting of the American Educational Research Association (AERA)

April 9, 2007 - April 13, 2007; Chicago, IL.

Page 53: ENDLINE EVALUATION · CBO Community-Based Organization CBA Cost-benefit analysis CEA Cost-effectiveness analysis ... consistent with the approach that is promoted by the intervention,

53

UNESCO. (1990). Meeting basic learning needs: A vision for the 1990s. Background Document: World

Conference on Education for All. Retrieved from http://www.unesco.org/education/pdf/11_92.pdf

UNICEF. (2014). Annual Report 2014 – Kenya.

http://www.unicef.org/about/annualreport/files/Kenya_Annual_Report_2014.pdf

United Nations (2015). Sustainable Development Goals.

http://www.un.org/sustainabledevelopment/sustainable-development-goals/

Wasik, B. A., Mattera, S. K., Lloyd, C. M., & Boller, K. (2013). Intervention dosage in early childhood

care and education: It’s complicated (OPRE Research Brief OPRE 2013-15). Washington, DC: Office of

Planning, Research and Evaluation, Administration for Children and Families, U.S. Department of Health

and Human Services. doi: 10.4269/ajtmh.2012.11-0494 PMCID: PMC3516306

Page 54: ENDLINE EVALUATION · CBO Community-Based Organization CBA Cost-benefit analysis CEA Cost-effectiveness analysis ... consistent with the approach that is promoted by the intervention,

54

ANNEXES

Annex A: Supplemental Methodology Information

Annex table 1 – Number of schools by reported treatment status Treatment Status N

C4D Not Reported 271

C4D UNICEF 38

C4D Other Org 40

CFS Not Reported 201

CFS UNICEF 117

CFS Other Org 31

Solar Not Reported 151

Solar UNICEF 78

Solar Other Org 120

WASH Not Reported 176

WASH UNICEF 112

WASH Other Org 61

Page 55: ENDLINE EVALUATION · CBO Community-Based Organization CBA Cost-benefit analysis CEA Cost-effectiveness analysis ... consistent with the approach that is promoted by the intervention,

55

Instruments and Key Data Sources

Outcome indicator

Definition Source Sample size Timing

Duration, dosage,

enabling environment

Amount and intensity of

training interventions;

presence of other (third

party) contributing

interventions

Head teacher

questionnaire

All schools Endline

Attendance (boys and

girls)

Percentage of students

from two consecutive

grades present on the day

of data collection,

disaggregated by gender

Primary data collection:

school attendance

registers at Day &

Boarding Schools

All schools Baseline

and

endline

Retention Number of students

promoted to the next

grade as a proportion of

students enrolled in the

previous grade the

previous year

Primary data collection –

data on pupils, repeaters

and new entrants by year

by grade

All participating

schools with data

availability

Endline

only

Solar power effect on

studying habits

Percentage of students

reporting studying on

school premises after dark

Student survey

Student focus groups

All participating

schools

Endline

only

Ongoing activities

supporting enrollment

drives

Percentage of schools

where enrollment

promotion activities took

place

Head Teacher survey

Teacher focus group

Student focus group

Endline

only

Community

perceptions of the

value of interventions

and their impact

Parent and community

perceptions on the value

of interventions and

persistent barriers to

school participation

Primary qualitative data

collection: focus group

discussions and key

informant interviews in

communities around

selected schools;

All participating

schools

Baseline

and

endline

Student perceptions Student reflections on the

value of interventions and

persistent barriers

Focus group discussion

for std. 6 pupils in

selected schools

Student rapid survey

All participating

schools

Endline

Effective use of

student-centered

instruction in class

Percentage of time that is

used for the application of

student-centered methods

promoted by CFS training

Primary data collection

(classroom observation)

At least 1

classroom per

school

Endline

Cost-effectiveness

Cost per student of each

intervention, in constant

2014 currency, relative

the change observed in

attributable lower-level

outcomes

Implementing org. will

provide their school-

level budget information

for each intervention.

Will also request per

student costs per district.

All

implementing

organizations, by

county

Endline

only

Page 56: ENDLINE EVALUATION · CBO Community-Based Organization CBA Cost-benefit analysis CEA Cost-effectiveness analysis ... consistent with the approach that is promoted by the intervention,

56

Hypotheses to be analyzed in the observational study:

The research hypotheses for this evaluation are based on the UNICEF and DFID Theory of Change which

states:

Interventions that improve water and sanitation systems will increase school attendance, particularly among

girls. As a result, the intervention will increase their overall exposure to instruction leading to higher

academic performance. The lower-level outcome for this intervention will be measured by changes in both

enrollment and attendance, particularly among girls.

Installation of solar-powered electricity will expand the effective time available for instruction, including

homework, leading to higher academic performance. The lower-level outcome measure is evidence of

students using school facilities for study during evening hours (after dark).

Community involvement interventions will raise the public awareness of the value of education and

enhance the positive perception of schooling for boys and girls. The lower-level outcome for this

hypothesis is evidence of the community involvement activities taking place and communities reporting

changes in attitudes and behaviors with respect to school enrollment. The outcome measure related to this

intervention are attendance and retention in the affected schools.

Teacher training for CFS will increase the effectiveness of instruction, ensure child friendly and inclusive

classrooms; safe schools; schools that promote health, nutrition and equitable education; and enhanced

community linkages and partnerships. The hope is that by bringing together these elements, it will

contribute to increasing the pupils’ academic performance. While the ultimate outcome – academic

performance – will not be measured for this observational study, a lower-level outcome will involve the

extent to which the student-centered instruction is applied at the classroom level. .

The null hypotheses to be tested related to the causal chain include the following:

i. H0: No effect of receiving WASH interventions on attendance and retention

ii. H0: No effect of receiving Solar Power on student studying patterns

iii. H0: No effect of receiving CFS training on application of student-centered instruction and time

use.

iv. H0: No effect of receiving CFS teacher training on attendance and retention

v. H0: No effect of receiving training on community engagement on attendance, retention and

sustainability of enrollment promotion

vi. H0: No effect of the combined interventions on attendance, retention and learning

environments.

Each of these null hypotheses is plausible for this evaluation. First, the duration of the interventions may

not be sufficiently long or intensive for the effects in enrollment and retention to manifest themselves.

Where expected effects are particularly modest, measurement error may sometimes crowd out the effect –

which is especially plausible when outcomes are self-reported as is the case with observational tools and

self-reported practices. Further, enrollment, attendance, and retention rates are vulnerable to demographic

shifts (for example, due to security concerns, climate or other external factors) that may affect different

counties differently.

Finally, and crucially, most of the schools already had some interventions taking place at the time of the

baseline, including the control schools. Nearly all schools reported having more than one intervention in

the two years prior to the baseline. Due to the high level of contamination and the lack of a clear start date

Page 57: ENDLINE EVALUATION · CBO Community-Based Organization CBA Cost-benefit analysis CEA Cost-effectiveness analysis ... consistent with the approach that is promoted by the intervention,

57

for the interventions, the estimates of the effect between baseline and endline in this study may

underestimate the true causal effect of interventions.

For these reasons, the observational study approach, with the focus on lower-level outcomes and the

implementation of interventions using the duration-dosage-enabling environment, is the appropriate design

for the endline study. In addition, an in-depth look at the “how” and the “why” of the interventions and

their effects on lower-level outcomes will be made possible through focus groups. Finally, the focus groups

will also provide a gauge of the remaining barriers for school participation and learning, and allow for better

understanding of the appropriateness of interventions to the real needs and barriers facing communities in

the ASAL regions.

Variables and analyses Qualitative analysis In terms of qualitative research, we conducted focus groups with students and teachers at each school;

community members; school and classroom observations; as well as interviews with key informants from

the implementing organizations. These various qualitative techniques allowed us to understand the

challenges and contributions of the interventions from various perspectives and document how the

interventions were implemented. The classroom observations – newly added for the endline – allowed us

to determine the extent to which training content on CFS and C4D was implemented at the school level.

We explain both the instruments and methodology for collecting and analyzing the data under each

instrument.

1. Focus Group Protocols: The endline data collection team used slightly revised and refined focus

group protocols to gather data on the effect of the interventions on the experience of students,

teachers and the communities. One focus group with each stakeholder group will be conducted

during the school visit.

2. Classroom Observation Protocol: The Stallings classroom observation instrument was used to

measure time on task and document whether the techniques taught to teachers during CFS and C4D

training workshops is translating to the classroom.

We conducted 25 focus groups across schools and communities. Classroom observations were conducted

in 54 classrooms across 54 schools, primarily those with CFS interventions. The focus groups were recorded

using the tablets, transcribed and upload to NVIVO for coding and analysis. Once analyzed, we estimated

the level of agreement using Cohen’s Kappa. We found that the coefficient was high, and therefore, we did

not feel that collecting additional data would be useful

Changes to the data collection and analysis process from baseline to endline Based on a review of the processes, procedures, and outcomes of the baseline, FHI360 made the following

changes to the data collection process to ensure quality control and meaningful outcomes for the overall

study.

1. Reviewed the instruments from the baseline and based on conversations with UNICEF Kenya,

made revisions in the protocols to ensure consistency, minimize missing data, and provide more

in-depth data and responses related to the interventions.

2. Refocused existing instruments and analysis towards the intended lower-level outcomes of the

interventions.

Page 58: ENDLINE EVALUATION · CBO Community-Based Organization CBA Cost-benefit analysis CEA Cost-effectiveness analysis ... consistent with the approach that is promoted by the intervention,

58

3. Changed the methodology to an observational study with the understanding that effects from the

interventions may not always be solely attributable to interventions, but must serve as initial gauges

of likely future outcomes.

4. Provided extended training to enumerators on the use of the newly revised instruments.

5. Worked with the field team to provide additional training on data cleaning so improve the quality

of the data in the system.

6. Strengthened the management of data collection on the ground.

7. Attempted to gather missing data from the baseline to better account for the enabling environment

and possible effects of third party interventions.

8. Added a classroom observation component in a sub-sample of schools to gather data on whether

and how teachers demonstrate skills received during CFS and C4D training.

9. Added a short student survey to collect data on student study habits.

10. Used tablets to collect data. The use of tablets increased the accuracy with which data was collected;

allowed FHI360 staff to spot check data daily and identify any problems with data quality; and

conduct the analysis more quickly.

Page 59: ENDLINE EVALUATION · CBO Community-Based Organization CBA Cost-benefit analysis CEA Cost-effectiveness analysis ... consistent with the approach that is promoted by the intervention,

59

Annex B: Computation of Net Present Value in Cost-Benefit Analysis The net benefit from WASH, Solar, and CFS is calculated as the (net) present value of taxes paid over the

life-cycle for students progressing to at least grade 6 over and above the taxes they would have paid had

they halted their educational progression in grade 5, less the initial cost (investment) of the intervention.25

Formally, we compute the net present value of the benefits accrued to each individual benefitting from the

WASH or Solar intervention as follows:

𝑁𝑃𝑉𝑡𝑏 =

∑ (𝑌𝑡𝑏 − 𝑌𝑡

0)𝑇𝑡=1

(1 + 𝑟)𝑡− 𝐶0

𝑁𝑃𝑉𝑡𝑏 is the net present value of the cumulative taxes paid by individuals with education attainment b less

the taxes paid had they not benefitted from the program, at time T. r is the discount rate for the value of

money in the future and 𝐶0 is the initial cost/investment of the WASH, Solar, or CFS intervention. For the

simulation model, we present the cost-benefit analyses under a range of possible discount rates. Finally, the

total net benefit of the interventions is the NPV associated with educational attainment b multiplied by the

number of students with education attainment level b.

Annex table 2 – Benefits-to-cost ratios for WASH, Solar, and CFS

r = 2% r = 4% r = 6% r= 8% r = 10%

Years post-implementation WASH

5 Years 0.39 0.37 0.35 0.33 0.31 10 Years 0.90 0.81 0.72 0.65 0.59 15 Years 1.49 1.27 1.09 0.93 0.81 20 Years 2.13 1.69 1.39 1.16 0.97 25 Years 2.70 2.08 1.64 1.32 1.08 30 Years 3.23 2.44 1.82 1.43 1.15 36 Years 3.85 2.70 2.00 1.52 1.19

Solar

5 Years 0.42 0.40 0.37 0.35 0.33 10 Years 0.96 0.86 0.77 0.69 0.63 15 Years 1.59 1.35 1.15 1.00 0.87 20 Years 2.27 1.82 1.49 1.23 1.04 25 Years 2.86 2.22 1.75 1.41 1.15 30 Years 3.45 2.56 1.96 1.52 1.22 36 Years 4.17 2.86 2.13 1.61 1.27

Child Friendly Schools

5 Years 1.33 1.25 1.19 1.12 1.06 10 Years 3.03 2.70 2.44 2.22 2.00 15 Years 5.00 4.35 3.70 3.13 2.78 20 Years 7.14 5.88 4.76 4.00 3.33 25 Years 9.09 7.14 5.56 4.55 3.70 30 Years 11.11 8.33 6.25 4.76 3.85 36 Years 12.50 9.09 6.67 5.26 4.00

25 We cap our benefits at attaining the final year of secondary school in order to produce a conservative estimate that could serve

as a lower bound for the benefits associated with each intervention.

Page 60: ENDLINE EVALUATION · CBO Community-Based Organization CBA Cost-benefit analysis CEA Cost-effectiveness analysis ... consistent with the approach that is promoted by the intervention,

60

Annex C: Additional Focus Group Results In this section, we highlight additional findings from the focus group discussions with parents, students,

and community stakeholders. These FGD findings highlight challenges still facing parents and students that

donors and UNICEF can provide support for in the schools.

School fees are an issue among the schools participating in the study as costs of uniforms and other materials

for school are expensive. Some parents indicate that the community may provide financial support to help

a family buy uniforms to get their children in school, however fees are still brought forward as a non-

negligible issue. It was mentioned across many of the FGD by both parents and students. Several parents

mentioned that providing scholarships, especially for orphaned children, may prove helpful in encouraging

enrollment.

“If the parents of that child cannot afford to buy for this child school uniform or books they will assist him

by contributing for him money so that he can buy for him.”

“one day when Iwas walking around the town came across a girl who had never enrolled to school and

when I asked the parents why they had not taken the child to school they said that they did not have money

to buy for her uniform and books. So, I went to the shop and bought for her uniform, shoes and books and

gave them to the parents and asked her to take the child immediately to the school and now the girl is in

school.”

“So most of the people are now in the town and if we come across any children not in school we find out

what is wrong and if it is because they do not have uniform or books we go to the UN office and ask them

for assistance, which they often provide and the books are provided by the government. We have done this

several time and almost all the children are in school although these uniforms are now torn and the

organization is no longer helping because there is peace in this region. The UN helps the region when there

is insecurity but when security improves they move to another region.”

“When our students do not have uniform and we are not able to provide for them we go knocking to all

NGOs doors in this region asking them to help our kids”.

School feeding programs and the need to ensure children have food at school was the topic most mentioned

in the focus groups when parents were asked what more could be done for the schools. They indicate that

the drought has severely impacted the households both in terms of access to water and food. Parents also

mention fairly frequently that when children get meals at the school, they are more likely to attend and stay.

Below, are some quotes from the FGD, where provision of meals at school was mentioned in every focus

group (all 23).

“children report to school at 6am and they don’t take breakfast so if schools would provide milk at break

time it would be very good. Also as we said, the school is not provided with enough food so the children

are served with very little food and they are not satisfied, so the government or others should add enough

food so that the children can comfortably stay in school.”

“I think the diet should be improved, apart from the one which is given by the school feeding program or

school meals program, at least even supplements should be provided because if you look at the diet which

is given by the school meals, there are no vitamins only beans, maize and oil, vegetables oil, but some fruits

should be provided for the students.”

Page 61: ENDLINE EVALUATION · CBO Community-Based Organization CBA Cost-benefit analysis CEA Cost-effectiveness analysis ... consistent with the approach that is promoted by the intervention,

61

“Children in boarding school should be given good food and those going back home should also be given

food – they cannot study when they are hungry. Water could also be of great help to the community and

build wash facility for the children to use after going to the toilet.”

Parents in Garissa and Wajir especially mentioned the need for security fences around the perimeter of the

schools. Here is what some of the participants said related to this issue:

“The school is not properly fenced and because of insecurity when the teacher sees the school is not

completely fenced they feel not secured and so they don’t come to school. The fence should go up so that

even the wild animals cannot pass through.”

“On the issue of security, they [the organizations] should also build fence around the school because of

security issues. Since the school is not fenced they [the students] are not safe.”