Impact Evaluation of an accelerated learning programme (the Luminos Liberia programme) for out-of-school children

The Luminos Liberia programme (also known as Second Chance, or SC) provides 10 months of intensive classroom instruction to out-of-school children (OOSC) to (re-)integrate students into mainstream public schools at the end of the program. The programme began in Ethiopia and expanded to Liberia in 2016 and is considered to be an accelerated learning programme, in which children learn basic literacy, numeracy and socio-emotional skills. IDinsight partnered with Luminos to conduct a randomized controlled trial (RCT) to assess the impact of the programme on children's learning outcomes. In this impact evaluation, 50 randomly assigned communities will participate in the programme, whereas 50 comparable communities will not. 20 children per community are tested on learning and numeracy skills (using respectively EGRA and EGMA) at the beginning and end of the programme, allowing for a rigorous estimation of the impact on learning outcomes. In addition, 5 children per community are tested on socio-emotional learning outcomes. In addition to the RCT, 5 children from public schools in every community are tested, to construct a scoring benchmark.

External Link(s)

Link to intervening organization

Registration Citation

Citation

Acero, Felipe et al. 2023. "Impact Evaluation of an accelerated learning programme (the Luminos Liberia programme) for out-of-school children." AEA RCT Registry. January 03. https://doi.org/10.1257/rct.10649-1.0

Sponsors & Partners

Partner

Name

Luminos Fund

Type

ngo

URL

https://luminosfund.org/

Name

IDinsight

Type

ngo

URL

https://www.idinsight.org/

Name

ROCH

Type

ngo

URL

https://rochliberia.page.tl/

Name

LIPACE

Type

ngo

URL

https://lipace.org/

Experimental Details

Interventions

Intervention(s)

The programme teaches children basic reading and numeracy skills and supports socio-emotional development. Luminos focuses primarily on providing the program to children who have never attended school, or have been out of school for at least the last two years. The SC programme uses a learning approach that, relative to the government school curriculum, is more adaptive to individual learning levels, offers diverse interaction and play-based learning, and is centred around intensive and semi-scripted lessons. For reading and comprehension, Luminos uses a phonics-based approach. In addition, both students and facilitators (community teachers) are intensively monitored and coached. Finally, the program and daily school lunches are consistently free.

Intervention Start Date

2022-09-19

Intervention End Date

2023-06-30

Primary Outcomes

Primary Outcomes (end points)

Our main outcomes for literacy and numeracy will be:
1. Composite scores from the Early Grade Reading Assessment (EGRA)
2. Composite scores from the Early Grade Math Assessment (EGMA)

Primary Outcomes (explanation)

EGRA and EGMA consist of 6 subtasks. The subtasks measure different literacy and numeracy skills and, as a result, the subtask scores will also be compared individually. Both EGRA and EGMA tools have procedures in place to not just measure raw scores (number of correct answers), but also compute percentage of successful attempts (disregarding questions that were not attempted), and scores per minute.
In addition to reporting subtask scores for EGRA and EGMA, we will also explore options for aggregating scores into a single score for each of the two assessments.

Secondary Outcomes

Secondary Outcomes (end points)

For socio-emotional learning (SEL), we are focusing on the following competencies: empathy, open-mindedness, self-awareness, and self-concept.

Secondary Outcomes (explanation)

To measure these, we will use several different tools. First of all, we will use the International Socio-Emotional Learning Assessment (ISELA) module on Self-concept to measure self-awareness and self-concept. In addition, we will measure scoring on the Empathy module from ISELA to measure empathy and social awareness. Finally, we will use the Dweck Growth Mindset scale to estimate scores on open-mindedness.

Experimental Design

The RCT consists of a 'treatment group' of 50 communities receiving the SC programme in the 2022-23 academic year and a 'control group' of 50 comparable communities not receiving the programme. For the sampling of OOSC, Luminos and its implementing partners compiled lists of roughly 35 children per expected SC class per scoped community (treatment and control). These were used as a sampling frame.

We aimed to sample 20 OOSC per community, stratified by gender and age (age was split into two groups with children above and children below 10 years old), with a target sample size of 2,000 OOSC (100 communities times 20 children).

Experimental Design Details

Randomization Method

Randomization has been implemented at the community level, given logistical concerns of within-community randomization and expected large spill-over effects within communities. The unit of analysis for the research questions will be the individual students. To balance the allocation to treatment and control across each IP and geographically, we stratified communities according to IP, anticipated number of classes per community, and county.

Randomization Unit

Randomization occured at the community level

Was the treatment clustered?

Yes

Experiment Characteristics

Sample size: planned number of clusters

100 communities

Sample size: planned number of observations

2500 children: 2000 out of school children (20 per community) and 500 public school children (5 per community) for score benchmarking

Sample size (or number of clusters) by treatment arms

50 communities for control, 50 for treatment.

Minimum detectable effect size for main outcomes (accounting for sample design and clustering)

Our estimates of intracluster correlation and autocorrelation of test scores come from a prior IDinsight RCT of a primary-school program, and we assume 80% power to detect significant effects at the 5% level. In order to detect effect sizes of 0.2 SD (moderate effects among education programs), our evaluation would require 47 treatment communities and 47 control communities if we conducted a baseline. Based on this, we planned to sample 50 communities across both treatment and control. In these calculations, we assume a sample of 20 OOSCs per community,

Supporting Documents and Materials

There is information in this trial unavailable to the public. Use the button below to request access.

Request Information

IRB