Solid Foundations and the Transition to Post-Primary School

The NULP was developed by Mango Tree Educational Enterprises Uganda, in collaboration with teachers, government officials, and the local Language Board. Because of the complex challenges in many African classrooms, the model involves a carefully-designed bundle of inputs focused on providing quality local-language classroom materials, training and supporting teachers in literacy instruction, and building the capacity of local government officials to train, supervise and support teachers.

The Full-Cost Program: The model focuses on mother tongue literacy and training and supporting teachers in literacy instruction. The first training module involves a five-day residential workshop on the Leblango orthography. In P1-P3, teachers undergo three intensive residential trainings in a school year on literacy methods (both whole language and phonics approaches) and participate in six Saturday in-service training workshops. Mango Tree staff visit classrooms to observe and provide feedback to teachers once per month. Government tutors make similar visits twice per term (six visits per year). The Mango Tree model introduces content slowly, providing time for repetition and practice; it takes two years to introduce the Leblango letter sounds, in contrast to other models which introduce all the phonemes in one year. Every teacher is provided with teachers’ guides that provide a script for each literacy lesson. In P4, the program did not deliver the same type of intervention, rather, Mango Tree supported teachers through a mentorship program.

The Reduced-Cost Program: Mango Tree’s goal was to create the highest-quality, most-effective literacy program possible. However, the materials, one-on-one support, and residential trainings are costly. Not including the initial costs of curriculum and materials development and community engagement activities, the program costs $19.88 per student. Mango Tree therefore created a modified, more easily scalable, reduced-cost version of the NULP.

There are two main differences between the full- and reduced-cost versions. The first is the use of a cascade model of training and support, rather than working directly with teachers. This approach involves Mango Tree staff directly training government tutors, who conduct teacher trainings and support visits themselves. The second difference between the full- and reduced-cost versions is that schools in the reduced-cost version received fewer support visits: two visits per term (from tutors only) instead of five (two from tutors and three from Mango Tree staff). In both program versions, tutors were given financial resources to make school visits and hold training sessions.

2013-01-01

2017-12-01

• Learning (through math and English literacy in reading and writing, formal and functional) and school progression (enrolment)
• Life skills, socio-emotional skills
• Life transitions and activities including sexual behaviour & marriage, labor market participation,
• Attitudes and expectations about the future

The two different versions of the program (full vs. reduced cost) and two different Cohorts treated either once, or continuously through early grades 1-4, allow us to compare different types of strategies to support basic skill formation on post-secondary outcomes. We will estimate the ITT impact of the NULP on student outcomes among Cohort 2 students with the following model (1), below:

Yis = β_0+β_1*FullCost_s+β_2*ReducedCost_s+ L'_s*γ+ δRead_is^base+ϵ_ij (1)

where i indexes students and s indexes schools. Yis is a student’s outcome at each of the two surveys. We control for a vector of indicator variables for initial randomization stratification cells L′s (Bruhn and McKenzie, 2009). In some specifications we will control for basic demographics such as gender, age, and parents’ education. We also control for each student’s baseline reading score in first grade, prior to exposure to the intervention Readbase, as specified in our pre-analysis plan. Among the full-cost and control schools, we will also estimate (2), below:

Yis = β_0+ β_1*FullCost_s+β_2*Cohort2_s+ β_3 FullCost*Cohort2_s+L'_s γ+ δRead_is^base+ϵ_ij

In equations (1) and (2) we will explore heterogeneous treatment effects by students’ gender, baseline test-scores, and socio-economic status. This will help us understand if some characteristics are particularly important for affecting students’ school progression and foundational skills. Following Bitler, Gelbach, and Hoynes, (2006), Djebbari and Smith (2008), and Koenker & Bassett (1978) we will estimate quantile treatment effects on test scores. Using data on enrollment at the student-year level (or student-term level if there is enough variation) and achievement scores for each year of available data, we will estimate the determinants of progression and learning over the study period using similar methods as in Ardington, Lam and Leibrandt (2011).

Public lottery

Cluster randomization at the school level

Yes

107 Schools

2,674 Pupils

• 38 schools in half-treatment (all in cohort S2),
• 34 schools in control (in all schools S2 will be studied, and in 7 of these schools, pupils in cohort S1 will also be studied),
• 35 schools in full treatment (in all schools S2 will be studied and in 8 of these schools, S1 will also be studied).