Today, impact evaluation of development projects is dominated by quantitative methods, both experimental and quasi experimental. Yet, they are difficult to apply to agricultural development projects, and do not allow for quantification of the impact over the useful life of the project realisations. This thesis proposes a new method, the systemic impact evaluation. It enables us to understand both the process through which impact materializes and to rigorously quantify the impact of agricultural development projects on the producers' income. The analysis is notably based on the comprehension of the agrarian dynamics and the farmers' strategies, and permits the quantification of ex-post impact but also to devise a model of ex- ante evolution for the following years. The systemic impact evaluation is applied to two major types of development projects in the forest region of Guinea: inland valley development projects and the oil-palm and rubber tree plantation project. These two case studies notably prove the importance of going beyond a mean impact on households: the existence of highly differentiated impacts by farming system is one of the main conclusions of this thesis.
