Community-based and -driven development projects have become an important form of development assistance, with the World Bank's portfolio alone approximating $7 billion. A review of their conceptual foundations and evidence on their effectiveness shows that projects that rely on community participation have not been particularly effective at targeting the poor. There is some evidence that such projects create effective community infrastructure, but not a single study establishes a causal relationship between any outcome and participatory elements of a community-based development project. Most such projects are dominated by elites, and both targeting and project quality tend to be markedly worse in more unequal communities. A distinction between potentially “benevolent” forms of elite domination and more pernicious types of capture is likely to be important for understanding project dynamics and outcomes. Several qualitative studies indicate that the sustainability of community-based initiatives depends crucially on an enabling institutional environment, which requires government commitment, and on accountability of leaders to their community to avoid “supply-driven demand-driven” development. External agents strongly influence project success, but facilitators are often poorly trained, particularly in rapidly scaled-up programs. The naive application of complex contextual concepts like participation, social capital, and empowerment is endemic among project implementers and contributes to poor design and implementation. The evidence suggests that community-based and -driven development projects are best undertaken in a context-specific manner, with a long time horizon and with careful and well-designed monitoring and evaluation systems.