In this paper, we focus on the determinants of the relationship between aid and corruption. We propose a static principal-agent model where a donor faces the problem of giving aid to a recipient country in which the phenomenon of corruption is widely spread. We distinguish among two different types of corruption: one, that we call endemic, that depends on the political and institutional environment of the recipient; the other, that we call aid related, is the consequence of moral hazard arising from the ability of corrupt burocracies to divert resources from their intended use. Through the design of appropriate contracts, donors can act only on the second type of corruption, contributing to reduce the entity of the phenomenon. We use the restrictions implied by our theoretical framework to test a model of aid allocation. For the majority of the donors (Germany, Italy, the Netherlands, Norway, Spain and the UK), we find some indication that efficiency considerations are taken into account in allocating aid. For some of them (Germany, Italy and the Netherlands), however, strategic/economic considerations are important, while the UK is also motivated by purely altruistic concerns. According to our model, Denmark and Japan are mainly driven by recipient needs, while the USA, and to a lesser extent France, allocate aid mainly on the basis of strategic/economic interests.