SUMMARY
This paper discusses the implementation of national and municipal assessment systems in Brazil, and how it has influenced the educational debate and newly shaped policies. The empirical analyses focus on public municipal schools in Rio de Janeiro and test the plausibility of the school composition effect theory. The approach combines different datasets provided by the Rio de Janeiro Educational Department: a) Prova Rio – longitudinal assessment for all public schools; b) Pupil Profile – longitudinal with information on all pupils, including family profile and school placement. Evidence from different models suggests the plausibility of school-mix effect theory. Data corroborates the claim that at least part of a purported “school effect” could be explained by a particular mix of pupils.