Raising schooling quality in low-income countries is a pressing challenge. Substantial research has considered the impact of cutting class sizes on skills acquisition. Considerably less attention has been given to the extent to which peer effects, which refer to class composition, also may affect outcomes. This study uses new microdata from East Africa, incorporating test score data for over 250,000 children, to compare the likely efficacy of these two types of interventions. Endogeneity bias is addressed via fixed effects and instrumental variables techniques. Although these may not fully mitigate bias from omitted variables, the preferred IV results indicate considerable negative effects due to larger class sizes and larger numbers of overage-for-grade peers. The latter, driven by the highly prevalent practices of grade repetition and academic redshirting, should be considered an important target for policy interventions.