Inequalities in children’s learning are widely recognized to arise from variations in both household and school-related factors. While few studies have considered the role of sorting between schools and households, even fewer have quantified how much sorting contributes to educational inequalities in low- and middle-income countries. We fill this gap using data on over 1 million children from three East African countries. Applying a novel variance decomposition procedure, our results indicate that sorting of pupils across schools accounts for at least 8 percent of the total test-score variance, equivalent to half a year of schooling or more. This contribution tends to be largest for children from families at the ends of the socio-economic spectrum. Empirical simulations of steady-state educational inequalities reveal that policies to mitigate the consequences of sorting could substantially reduce inequalities in education.