Delivering Remote Learning in Developing Countries using a Low-tech Solution – Bangladesh

The intervention will test whether IVR will improve the grade-level learning of participating
students in Bangladesh. By offering content of different proficiency levels to the
secondary school students, the intervention will in effect offer a menu of choices so that
students can continue to learn more if they have current grade level proficiency or catch up
with the learning gap if they are behind their grade level. The study will conduct a
randomized control trial (RCT). RCT allows to avoid endogeneity issues and identify the
causal impact of the intervention. We will randomly assign 208 secondary schools into three
groups of equal size:
T1: Self-help group – At the beginning of the program, participants will be briefed by field
staff on how to access IVR-based lessons. Also, the field staff will explain how this program
could lead to better educational outcomes.
T2: Assisted group – Students in this group will get the same treatment as T1. In addition,
they will get a bi-weekly call from a tutor. These tutors will be hired from those within the local
community who have the required teaching skills. Students of T1 may face two constraints.
The first is lack of willpower: students may lack the motivation to use the lessons regularly.
The second is incomprehension: students may fail to understand some of the contents of the
one-way audio lessons. To counteract these two possible shortcomings, this additional bi-
weekly phone calls should increase participants’ motivation and resolve problems of not
understanding the lessons. These discussions will cover the content to be covered in our
modules.
T3: Control – No intervention will be provided.

2023-01-15

2023-06-15

(1) Student Learning
(2) Student’s Aspirations
(3) Student’s Hopes
(4) Student time spent on exam preparation

a) Student Learning outcomes
Standardized test score of English exams at the endline
Standardized test score of Mathematics exams at the endline
Notes on the construction of the test score outcomes:
Students’ learning will be measured using a standard assessment test based on the national curriculum of Bangladesh. There will be two tests, one each for English and Mathematics. To examine the effect of this treatment (IVR lessons and direct assistance) on learning outcomes, we will construct the outcome variable by standardizing the raw test score with respect to the mean and standard deviation of the control group. We will also standardize the baseline test scores with respect to the baseline control mean and standard deviation.
Unit of observation: Students
Data source(s): Assessment tests (Mathematics and English Test Scores)
b) Student Aspirations: Using a method like the one employed by Beaman et al. (2012)
and Bernard et al. (2014), we will estimate aspirations.
Unit of observation: Students
Data source(s): Baseline and endline surveys
c) Student hope: We will measure the hope of the participants by using the Hope Scale,
a 6-item self-report measure of students’ perceptions that their goals can be met (Snyder,
1994, Snyder, 2002). There are two categories of hope, with three specific items within each
category. The two categories are agency and pathways. We will construct three outcome
variables, agency, pathways and the overall hope score.
Unit of observation: Students
Data source(s): Baseline and endline surveys
d) Student time spent on exam preparation: Students' time investment preparing for the
exam will be measured by asking the following questions:
(a) How much time (hours) in various academic activities (i.e., reading, practising exam
questions, taking lessons from private tutors etc.) do you spend each week? Total hours
spent in various academic activities will be used as an outcome variable for student time
investment in academic activities.
(b) To what extent do you spend time on academic activities (5-point Likert-scale response;
‘none’ to ‘a great deal’)
Unit of observation: Minutes spent on exam preparation by each student
Data source(s): Baseline and endline surveys

This study uses an RCT to assess whether there is a causal link between the program and changes in outcomes. The evaluation of this employs a three-arm clustered RCT design (two treatment groups and one control group).
Randomization is done in 2 stages: We first pick 208 schools from a pre-existing list provided by our partner organizations. We then randomly distribute these 208 schools into the three study arms: T1 (self-help), T2 (assisted), and C (control).
T1 (self-help) and T2 (assisted) will have 69 schools each and Control will have 70 schools. From each school, we will randomly select 15 students on average to participate in the program. We will ensure that about half of the students are female when we randomly select the students to be treated from each treatment school.
The intervention contains three study arms:
a) T1: Self-help group (69 schools, 1058 students selected to receive treatment) –will only be provided with information on how to access IVR-based lessons and how this program could result in improved educational outcomes (such as higher marks on the SSC exams). Participants will receive regular text messages from the IVR system.
b) T2: Assisted group (69 schools, 1022 students selected to receive treatment) – Will receive the same information as students in T1. Additional treatment will be provided in the form of biweekly phone calls from a tutor. Participants will also receive regular text messages from the IVR system.
c) T3: Control (70 schools, 3201 students) - This group of students will not be receiving any intervention.

Randomization done in office by a computer using Stata/BE 17.0

A two-stage randomization procedure will be employed. In particular, we first randomly
assign 208 schools to two treatment groups and one control group, and then within each
school we randomly select 15 students on average (in a way that ensures almost equal numbers of boys
and girls) from each of the 208 schools (for treatment in the treatment schools, and for data
collection in the control schools).

Yes

208 Schools

9352 Students (which includes untreated students in treated schools)

T1: Self-help group: 69 schools (units of randomization), 1058 students selected for treatment (units of observation), 2260 students not selected for treatment (but data will be collected from these students)
T2: Assisted group: 69 schools (units of randomization), 1022 students selected for treatment (units of observation), 1811 students not selected for treatment (but data will be collected from these students)
Control: 70 schools (units of randomization), 3201 students (units of observation)

We consider a 0.24 SD effect size, alpha levels of 0.05, and a 0.2 intra-cluster correlation (ICC) to achieve 80% statistical power.