A link to the pdf of the full paper is here. Note this is a working paper.

Abstract: School closures during COVID-19 caused learning losses worldwide, often compounding existing deficits. Brazil, with one of the longest closures, is a particularly striking case. This paper examines how the city of Recife responded after reopening schools in 2022 and 2023. With education largely decentralized and managed at the municipal level in Brazil, Recife’s education department rolled out various system-wide measures—adjusting pedagogy, offering opportunities to group students by learning levels, and sharing learning outcome data with schools to inform teachers. Given the multifaceted nature of the problem, heterogeneity in implementation, and the evolving nature of the intervention, a quasi-experimental design was neither feasible nor appropriate. Instead, a simple application of an adaptive evaluation approach—involving systems diagnostics and process tracing—was employed for a structured assessment of policy implementation, school-level variation, and shifts in learning outcomes. While the core guiding principles remained stable, 2023 saw a shift toward increased school-level discretion, reduced testing burdens, and more timely feedback. Process tracing of schools showed that guidance was generally transmitted reasonably effectively, but uptake varied, revealing limitations in translating data into classroom practice. Although statistical causal attribution was not possible, average learning outcomes in Portuguese and Math recovered to or surpassed pre-pandemic levels, with substantial variation across schools. Qualitative analysis of high performing “positive deviant” schools highlighted the importance of data use, internal collaboration, and flexible regrouping. These findings support reform strategies that combine central coordination with local autonomy and show how an adaptive evaluation can illuminate the workings of complex education systems, where traditional causal inference is not possible.