The link between residential and school segregation is widely recognised as a key to explaining urban inequalities. However, most studies have focused on countries of the Global North. This paper attempts to identify to what extent socio-economic residential segregation explains secondary school segregation in Buenos Aires (Argentina). Based on a linear programming method, the study proposes a hypothetical pupil allocation model that takes into account the capacity of schools and is used as an ideal typus to compare with the real socio-economic school composition. Using a decompose method' of segregation differences to analyse the differences in segregation indices and a local segregation analysis, this paper finds that in a residential context with low segregation but high social inequalities, school segregation is a social mechanism that allows maintaining spaces of differentiation and distancing between groups. In discussion with the idea of a vicious circle of segregation', this article argues for the potential of a multi-domain approach to segregation, to understand how different domains work in articulated and complex ways to reinforce urban segregation.