Summary
Key findings
- Reducing class size by about 16 pupils per stream (from roughly 52 to 36) raised first-term scores by 0.23 SD in English, 0.22 SD in Science and Elementary Technology and 0.20 SD in Social and Religious Studies; the 0.13 SD gain in Kinyarwanda was not statistically significant, and the effect on mathematics was close to zero.
- By the end of the school year (Term 3) the effects were smaller and no longer statistically significant in any subject, although the mathematics estimate rose to 0.06 SD.
- Gains were concentrated in Grades 1–3: first-term gains there included 0.35 SD in English, and gains in English, mathematics and social studies were still significant at year-end. In Grades 4–6 only English improved in the first term (0.19 SD), and by year-end the estimates were negative.
- Smaller classes helped in single-shift schools (for example 0.18 SD in English) but brought no significant gain in double-shift schools. Gains were largest where streams had about 40–45 pupils (0.31–0.42 SD in English, science and social studies); going below 40 added little.
Methodology & context
- Uses RwandaEQUIP administrative data (provided by NewGlobe) from 761 public primary schools in all five provinces of Rwanda for the 2023–2024 school year: first- and third-term exam scores in five subjects for Grades 1–6, aggregated to 4,471 school-grade observations.
- Uses a fuzzy regression discontinuity design based on the programme rule that a grade with more than 69 pupils be split into additional streams; crossing the threshold raised the probability of having two or more streams by about 49 percentage points. Estimated by two-stage least squares with baseline scores and other controls, using grades within 100 pupils of the threshold.
- Because enrolment takes few distinct values near the threshold, the estimates rely on a linear extrapolation and should be read with caution. They come from RwandaEQUIP schools, which already have structured lesson plans and coaching, so they may not carry over to other schools, and the evidence on mechanisms (more individual attention) is only suggestive.
Summary drafted with AI from the AFEP working paper and policy brief, and checked against them. See the full study for details.
Abstract
Developing countries have achieved substantial gains in school enrolment, yet improvements in learning outcomes have lagged behind. Overcrowded classrooms are a key barrier, as they tend to compromise the conditions for quality and effective instruction. Understanding the role of class size - one of the most debated and policy-relevant inputs in education - is therefore critical, particularly in resource-constrained systems where reductions are costly. This paper estimates the causal effect of class size on student learning outcomes using administrative data from the RwandaEQUIP programme across 761 public primary schools in Rwanda. Exploiting enrolment-based class size thresholds, we implement a fuzzy regression discontinuity design to identify exogenous variation in class size. The results show that reducing class size by approximately 16 pupils per stream - moving from roughly 52 to 36 pupils - improves short-run learning outcomes by 0.23 standard deviations (SD) in English, 0.13 SD in Kinyarwanda, 0.22 SD in Science, and 0.20 SD in Social Studies. Mathematics gains are negligible in the short run but grow to 0.06 SD by year-end, consistent with the cumulative nature of numeracy skill acquisition. Effects are largest in early primary grades and single-shift schools, and dissipates by Term 3, when estimates remain positive but statistically insignificant across all subjects. The findings highlight the importance of targeting class size reductions to early grades, where marginal returns are highest, and contribute to the limited causal evidence on class size and learning in Sub-Saharan Africa.