AbstractThe paper assesses the effectiveness of two institutional strategies, one centralized and the other decentralized, for sustaining infrastructure investments in the Third World and the organizations that have developed from them. It argues that each has been flawed. Institutional over‐centralization, involving a centrally directed and hierarchical bureaucracy is widely found to have been ineffective in undertaking complex tasks of rural development under conditions of uncertainty and severely limited resources. It has failed to secure beneficiary participation, adequate maintenance and utilization of infrastructure, administrative co‐ordination, responsiveness to local needs, as well as flexibility and adaptability. But decentralization has not been a widely successful solution. Decentralized planning, attempts to increase the role of beneficiaries in development projects, and localized training and organizational developments efforts have all produced disappointing results. There is a lack of theoretical understanding in this area. Human behaviour and relationships within organizations have been neglected. An alternative strategy, using the individual as the unit of analysis, is suggested. It requires ‘institutional analysis’ of the incentives and disincentives involved in getting people to perform specific tasks relating to specifi