School closures and the compositional validity of Ideb in Brazilian municipal school networks

Abstract

School closures are a frequent administrative decision in Brazilian municipal school networks and account for a material share of the consolidation of school supply in the lower grades of primary education. This paper asks what network Ideb measures when that consolidation takes place, using a panel of 5,597 municipalities between 2007 and 2023 and combining difference-in-differences with a decomposition of the change in the indicator. The effect of closures on network Ideb is positive and concentrated in the rural segment (ATT of 0.14 points, against an estimate indistinguishable from zero for urban closures). Within that segment, however, the change in Ideb between 2007 and 2019 is almost entirely attributable to recomposition of the set of assessed schools: the within-school component accounts for 2% of the change, against 71% in the overall aggregate. At the same time, the more schools a municipality closes, the larger the dispersion of Ideb across remaining schools, the larger the urban-rural gap and the lower total municipal enrollment. Read on its own, network Ideb does not support inference about network quality where rural supply has been consolidated; using it for accountability requires reporting the decomposition between the within-school and the compositional components.