This paper compares the increases in access rates to piped water in rural Brazil by localized water supply systems under two different management models. It finds that small-scale supply systems operated and maintained by user associations lead to significantly higher increases in access rates than comparable systems implemented and operated by local governments. Additional results point towards higher accountability as the reason for better performance. This paper is the first to evaluate the success of community-based water supply projects in rural areas by comparing them to non-participatory projects and is based on a valid econometric identification strategy. As service delivery is decentralized in Brazil, the results also contribute to the discussion of the merits and risks of decentralized water supply. In order to overcome the endogeneity problem, I use a difference-in-difference estimator in combination with a kernel matching approach. Treatment effects are robust to various specification changes; tests show no structural differences between treatment and control groups.