The World of Work - A study of the transition from secondary school in Tanzania - PRE-ANALYSIS PLAN FOR THE ENDLINE SURVEY

Future Yangu Podcasts is A Work Readiness Training Program Career Guidance and Counselling program preparing secondary school students for the World of Work (WoW). Future Yangu is designed specifically for this research project in close collaboration between HIP edutainment, Ruka creatives and the research team. It is an extracurricular learning material, delivered by teachers to a collective of students who form listening groups.

The program is delivered in a new innovative digital format, using audio podcasts. This is new to the education sector in Tanzania. The aim is to support the access of new digital learning materials. The program consists of 12 episodes which are accessible (to the treated schools) through a mobile phone App.

The program is designed to address one of the big challenges in Tanzania – youth unemployment and the lack of job readiness among secondary school leavers. The aim is to assist Form 4 students navigate the world of work so that they have a smoother school/work transition and can find education pathways and make inspiring career choices while enhancing their knowledge of work readiness and labor market skills. The program is particularly useful for those who will proceed to vocational and technical training and do not aim to go to university. They may or may not have received their Form Four Certificate - the ‘ticket to education pathways’. Common to all is that they must navigate their way into the Tanzanian labor market.

2023-02-01

2023-07-01

The first main objective of this survey is to assess whether and how the podcast intervention affected the students’ occupation a year after they finished lower-secondary school. They may be studying, working in a paid occupation, or doing unpaid work.
We will therefore test the treatment effects on the following outcomes (denoted by O, for occupation):
• O1: a binary variable indicating if their main occupation is a paid job (self-employed or salaried)
• O2: the income from their main occupation
• O3: a binary variable indicating if they are enrolled in any educational programme
• O4: a binary variable indicating if they are enrolled in an alternative educational programme (VETA school, professional training, vocational colleges…)
The second main objective is to test whether the intervention had a sustained impact on the students’ hope and aspirations.
To measure hope, we use an index based on Lybbert & Wydick (2017) and, originally, Snyder (1994). The Hope Index is a combination of three indices (aspirations, agency and pathways). We construct each index based on the students’ level of agreement with five sentences (from 10-totally agree to 0-totally disagree), following the method described in Kling et al. (2007) and Anderson (2008). We also use alternative data-driven scoring techniques.
We will test the treatment effects on the following outcomes (denoted by H, for hope):
• H1 The Hope Index,
and its sub-indices:
• H2 The Aspiration Index
• H3 The Agency Index
• H4 The Pathways Index

We are planning further analysis in order to document how the treatment affected our main outcomes.
One potentially important explanation, is that the treatment may have pushed the students to develop B-plans. In order to test this possibility, we will estimate the treatment effects on:
• A1 A binary variable for having listed any alternative education choices on the “Selform”, the official document indicating their preferences in terms of further education.
Furthermore, now that the uncertainty related to the exam is lifted, we will test the treatment effect on how realistic student’s expectations are:
• A2 A binary variable for expecting to graduate from university
• A3 A binary variable for expecting to get a governmental job
We will also test whether the treatment affected their exam results:
• A4 A binary variable for obtaining Division IV or 0
• A5 Their overall grade
Using recall questions, we will test whether the students behaved immediately differently after they finished the final lower-secondary school exams, and before they received the exam results:
• A6 A binary variable for being employed, or self-employed,
• A7 A binary variable for actively looking for a job, following a training, or being actively looking for one.
Finally, we will test the program effect on stress and mental well-being:
• A8 PSS index
• A9 Cantril index

Experimental Design
Cluster randomized trial. Randomization is at the school level. The target population of participants are the students of Form IV.
Experimental Design Details
We selected two study regions in Tanzania: Dodoma and Tanga. Dodoma is very central and inland, while Tanga is a (large) costal region. These two regions offer a good mix of the different geographical and cultural particularities of the country, in particular difference between coastal and inland areas.
Within these two regions, we established the pool of all schools which hosted a Fema club from our partner organization. This criterion is meant to facilitate communication with the schools, and the logistics later on. This left us with 211 Schools, 131 in Tanga, and 81 in Dodoma.
We then also imposed the restriction that the share of students obtaining division 4 or 0 in the CSEE exam to be bigger than 60%. This is to ensure that we are treating the students most in need of the program, as highly performing students don’t really need information and exposure on alternative pathways.

Finally, we restricted the sample to schools that have between 40 and 80 students. This is for logistics reasons: We wanted to avoid bigger schools because it would have been too costly to treat every student, and to avoid smaller schools because these tend to be more remote and difficult to access.

We ended up with 60 schools, from which we randomly selected 50 in our final sample. We interviewed at baseline 2863 students. We had three types of school, depending on their size:
- A) Every students in the cohort were surveyed (34)
- B) We surveyed the students for the weaker stream (12)
- C) When streams were not organized by academic levels, we tossed a coin to know which stream to survey (4)

Using STATA, 25 schools were randomized into the treatment group, 25 into the control group. Schools were stratified by their region and type. Two teachers from each treated school were trained to deliver the Podcast material to form IV students, for 12 sessions. Each school followed their own calendar, but all had completed the treatment before the summer break.

Not available

randomization done in office by a computer.

school

Yes

50 schools

2,800 students

25 schools per arm.

given the ICC observed in pilot data and while preparing the project, we expect to be able to detect an effect of 0.1 to 0.2 sd depending on the outcome (with 5% level of significance and 80% power).