In many low-income and 'fragile' states, citizens' committees are elected to co-manage basic social services. However, the effects of such committees on service delivery, and the way they are influenced by local contexts, remain understudied. This thesis seeks to fill these gaps by examining the case of the health facility committees in Burundi and South Kivu between 2011 and 2014. It relies on original health facility and committee surveys, household surveys, nested interviews and focus groups, and interviews with key informants. The thesis firstly explores how the committees came about. It then looks at the questions, What makes them get involved in decisions at their health facility? and, How do measures designed to improve committee functioning lead to changes in service delivery, if at all? Mixed-methods work finds that chief nurses largely dominate the health facilities, and the committees appear to be both the product of recent political and administrative changes and a façade of community governance. The work's randomised controlled trial tests the idea that this inefficiency arises from an 'institutional knowledge gap': the committee members and nurses do not know the committee's (official) functioning. An information session has strengthened the committees and led to changes in health facility management in South Kivu, but not in Burundi. This difference seems to come from dissimilar management structures and people's relationships to service providers. The intervention has had no effect on service provision. The remaining chapters report on additional interventions in Burundi, which theory and qualitative research suggest might improve the effects of the knowledge intervention: trust-building between nurses and committee, information about health facility performance, and increased interaction between local leaders and committees. These are either ineffective or have unintended consequences. Overall, the thesis nuances the promises of social accountability mechanisms and stresses the importance of power relationships within basic social services.