The aim of the present study is to investigate how and why a country facing issues that needed to be tackled nationwide chose to decentralise political power and financial resources when it moved from military rule to redemocratisation. Furthermore, the study examines whether the decision to decentralise taken in Brazil in the period 1981-1991 has changed the allocation of public expenditure at sub-national level, especially to education. By analysing the decision to decentralise and its results at the sub-national level, the study embodies both an upstream and a downstream approach. The upstream approach encompasses the topics related to decentralisation in the Brazilian Constituent National Assembly that sat from 1987 to 1988. Research sources are based on the archives of the Constituent National Assembly and on interviews with key political leaders in Congress and practitioners. The decision to decentralise is analysed in three dimensions: the relationship between political parties and the State; intra- and inter-party competition; and regional cleavages. The downstream approach comprises three case studies: the state of Bahia, its capital, Salvador, and its most industrialised municipality, Camacari. The political analysis is based on (a) interviews with politicians in executive and in legislative positions, and officials and (b) newspaper material as a complementary source. The financial performance is based on the analysis of expenditure extracted from balance sheets. This study fills four gaps in political-science and public-administration works on contemporary Brazil First, it deepens the understanding of how and why Brazil became such a decentralised nation. Second, it links the analysis of political and financial resources. Third, it highlights differences between levels of government in their experiences with decentralisation. Fourth, it investigates the impact of decentralisation on political arrangements and on education expenditure. The results suggest that in Brazil there was a lack of social consensus on what was to be achieved by decentralisation. They suggest that decentralisation fosters democracy but its impacts on policy results have so far been limited. The evidence further implies that decentralisation and democratisation bring about a fragmentation of power without necessarily disintegrating previous political coalitions or changing the way public resources are spent. These findings indicate that various political and economic factors influence the outcomes of decentralisation, thus exposing the limits of decentralisation on policy results.