Assistance to refugees is shifting from a humanitarian model, which focuses on protection, emergency relief, and shelter, to a development model promoting refugee self-reliance through income-generating activities, market development, and cash transfers. Evidence on the effects of this paradigm shift is limited. Exploiting a regression discontinuity design, this paper tests whether the adoption of a development approach to refugee assistance in a new settlement in Kenya has a positive impact. We find that refugees benefiting from the new approach have better diets and perceive themselves as happier and more independent from humanitarian aid. We find no effect on assets and employment. These effects appear to be driven by the switch from food rations to cash transfers and by the wider promotion of small-scale agriculture. Our findings argue in favor of the development approach to refugee assistance, which is cheaper and leads to better outcomes.