Improving Primary Education in the Democratic Republic of the Congo: End-Line Results of a Cluster-Randomized Wait-List Controlled Trial of Learning in a Healing Classroom

Abstract

We used a cluster-randomized, wait-list controlled trial to examine impacts of a school-based social-emotional learning intervention on Congolese students and teachers. Seventy-six school clusters in two groups (A and B) were randomized to treatment or control. The groups differed in geographic location, accessibility, exposure to violence, and external donor investment. We estimated causal impacts in Group A, tested whether those impacts were replicated in Group B, and conducted sensitivity analyses on the pooled sample. Pooled analyses had higher statistical power and were therefore more likely to represent the true average impacts of the program. Improvements in students’ perceptions of school predictability and in addition and subtraction, geometry, and reading performance were specific to Group B. Only the effect on addition and subtraction remained significant in the pooled analysis. Improvements in teachers’ sense of accomplishment were found in Group A and remained significant in the pooled analysis. We detected no impacts on other outcomes. School-based interventions embedding social-emotional learning principles into the academic curricula are a promising but not yet proven approach to improving children’s outcomes in low-income countries affected by war.

Keywords:

• social-emotional well-being
• reading
• mathematics
• Democratic Republic of Congo

Acknowledgments

Additional support for scientific analysis and write-up of data was provided to authors Aber, Tubbs, and Wolf by the New York University Abu Dhabi Research Institute and to Aber, Tubbs, Starkey, and Johnston by the U.K. Economic and Social Research Council.

The authors wish to thank the schools (directors, teachers, children) and education officials in the DRC whose participation made this study possible.

Disclosures

Coauthors Jeannie Annan and Anjuli Shivshanker are employees of the International Rescue Committee and Nina Weisenhorn was an employee of the International Rescue Committee at the time of the evaluation. They are (or were) part of the Research, Evaluation, and Learning division and were committed to and helped guarantee a truly independent evaluation.

Notes

1 We use the World Bank’s classifications of low- and middle-income countries.

2 Other kinds of interventions invest in infrastructure and provide materials and incentives for teachers and students, such as performance-based rewards and report cards.

3 Learning circles are regular meetings where educators share information and support each other in learning and implementing new pedagogical techniques.

4 North Kivu Province was included in the original design. However, security threats from M23 rebels and other armed groups led to the suspension of 2013 data collection, and thus school clusters from North Kivu are not included in this study.

5 We assumed an alpha of .05 and power of .80 for all estimations. We also assumed that covariates would explain 25% to 50% of the variance in the outcomes; this was reasonable because we assumed we would be able to follow individual students over time. Because the study was designed to detect impacts on outcomes with different intra-class correlations (ICCs), we also assumed that ICCs would range from .10 to .30 and estimated power under the two scenarios.

6 d_WT represents a standardized mean difference between intervention and control clusters. This was calculated with the following equation: $$d_{WT}= \frac{b}{\sqrt{{\widehat{\sigma }}_{BS}^2+ {\widehat{\sigma }}_{BC}^2+ {\widehat{\sigma }}_{WC}^2 }}$$ where b represents the unstandardized regression coefficient with covariate adjustment, and the three terms of the denominator represent variances at the cluster, school, and student levels, respectively, controlling only for the intervention effect. The rationale behind covariate adjustment for the intervention effect, but not the variances, was to obtain a more precise intervention effect (i.e., adjusted), but standardized based on typical (i.e., unadjusted) variances at each level (L. V. Hedges, personal communication, November 3, 2014). This same approach was used to estimate d_WT for this and other main effects presently reported.

Additional information

Funding

Data collection for this evaluation was funded by the United States Agency for International Development under cooperative agreement #623-A-10-00023 and the NoVo Foundation.