The originality of the REDD proposal is its incentives-based mechanism designed to reward the governments of developing countries for their performance in reducing deforestation as measured against a baseline. This mechanism is founded on the hypothesis that developing countries 'pay' an opportunity cost to conserve their forests and would prefer other choices and convert their wooden lands to other uses. The basic idea is, therefore, to pay rents to these countries to compensate for the anticipated foregone revenues. The reference to the theory of incentives (in its principal-agent version) is implicit but clear. In this REDD-related framework, the Government is taken as any economic agent who behaves rationally i.e. taking decisions after comparing the relative prices associated to various alternatives, then deciding to take action and implementing effective measures to tackle deforestation and shift the nation-wide development path. Such an approach ignores the political economy of the state, especially when dealing with "fragile" or even "failing" states facing severe but chronicle institutional crises, which are often ruled by "governments with private agendas" fuelling corruption. Two assumptions underlying the REDD proposal are particularly critical: (i) the idea that the government of such a state is in a position to make a decision to shift its development pathway on the basis of a cost-benefit analysis that anticipates financial rewards, and (ii) the idea that, once such a decision has been made, the "fragile" state is capable, thanks to the financial rewards, to implement and enforce the appropriate policies and measures which could translate into deforestation reduction. The first sections of the article discuss the pertinence of applying such a REDD version of the theory of incentives to Governments, and particularly to Governments in fragile states, with respect to the historical patterns and the practical way those states work. The last sections discu