This paper analyzes optimal foreign aid policy in a neoclassical framework with a conflict of interest between the donor and the recipient government. Aid conditionality is modelled as a limited enforceable contract. We define conditional aid policy to be self-enforcing if, at any point in time, the conditions imposed on aid funds are supportable by the threat of a permanent aid cutoff from then onward. Quantitative results show that the effectiveness of unconditional aid is low while self-enforcing conditional aid strongly stimulates the economy. However, increasing the welfare of the poor comes at high cost: to ensure aid effectiveness, less democratic political regimes receive permanently larger aid funds.