Panel Paper: Differential Impact of Literacy Boost Across Subgroups of Students – Experimental Evidence from Rwanda

Monday, April 10, 2017 : 12:00 AM
HUB 268 (University of California, Riverside)

*Names in bold indicate Presenter

Minahil Asim, University of California, Davis, Qiao Wen, Columbia University and Elliott Friedlander, Friedlander Research
According to UNESCO (2013), approximately 250 million children cannot read regardless of enrollment and attendance in schools. Research suggests that for a child to improve reading and writing skills, their environment, both in and out of school, should support their literacy development (Snow, Burns, & Griffin, 1998). In order to enhance school experiences with wider community activities, Save the Children adopted an educational approach they termed ‘Life-wide learning’ (Friedlander, Dowd, Borisova, & Guajardo, 2012). The Literacy Boost (LB) program in Rwanda implemented by Save the Children and independently evaluated by researchers at the Stanford Graduate School of Education in partnership with the Rwandan Education Board, adopted this approach. The program included community reading activities and reading awareness workshops to enhance the literacy environment for students. Friedlander, Zhou, Arshan & Goldenberg (2016) through a randomized control trial examined how students’ reading skills improve in the experimental group assigned to LB and estimated a positive and significant impact on students’ reading outcomes.

However, we do not know whether the program had differential effects among subgroups such as girls and students from low SES backgrounds. These unanswered questions about the project have tremendous implications for achieving parity in education, especially in a country like Rwanda where learners face diverse challenges in their learning development. In this paper, we use experimental data to estimate the differential impact of the LB program across sex and SES on children’s reading comprehension, reading fluency and oral comprehension in Kinyarwanda and English. Exposure to LB activities may impact students differently in the subgroups because of systematic differences at baseline in the home-literacy environment and academic achievement, or because of unobserved biases that may have influenced implementation. Our analysis offers insights to policy makers on how to address reading challenges for subgroups to achieve equity in education.

In particular, our study has three objectives: First, we estimate the intent-to-treat impact of being assigned to LB on subgroups including girls and low SES children. Second, we estimate a treatment-on-treated (ToT) effect of LB among those who actually received the LB intervention i.e. students who were assigned to a reading buddy, attended a reading club, or created materials to take home at the reading club. Finally, we estimate a “dosage” effect of LB among subgroups to determine if the intensity of exposure matters in improving reading outcomes. We use instrumental variables estimation (IVE) to estimate a ToT effect and obtain an asymptotically unbiased estimate of the causal impact of exposure to LB on reading outcomes (Murnane & Willett, 2011). 

Our preliminary findings suggest that the positive effect of LB accrued disproportionately to girls. There was a positive and significant effect size of 0.37 on reading fluency and 0.24 on reading comprehension for girls as opposed to no significant effect for boys. This suggests the effectiveness of this program in improving parity in educational outcomes across gender. In follow-up analysis we explore the ToT and dosage effects for girls and boys and extend our analysis to look at students across SES quintiles