Does bilateral aid work better through multilateral channels? Evidence from the GODAD database

Abstract

Abstract How does the choice of aid modality influence development effectiveness? While scholars have extensively studied the drivers of aid delegation, the relative effectiveness of different modalities remains under-theorized. I develop a theory of aid effectiveness rooted in the interplay of donor motives, recipient incentives, and principal−agent dynamics. I argue that earmarked aid to international organizations (IOs) will succeed in achieving specific sectoral outcomes by leveraging IO capacity to constrain recipient slack, though its rigid mandates and high transaction costs will limit its broader development impact. Conversely, bilateral aid—which often serves donor particularistic interests and grants recipients more autonomy—is generally ineffective, except when political competition disciplines recipients to prioritize long-term gains. To test these propositions, I utilize the GODAD database to analyze geolocated aid projects from 19 Western donors across 1,753 subnational regions from 1990 to 2024. Using two-step counterfactual difference-in-differences estimators, I examine the impact of different funding types on household wealth, child mortality, and governance. The findings reveal that earmarked aid is effective for specific goals, such as health aid reducing child mortality, whereas the success of governance aid depends on low baseline development levels. In contrast, bilateral aid is frequently ineffective or counterproductive, correlating with increased mortality and decreased state capacity. Sectoral targeting has limited effects, positively affecting only household wealth creation. These results highlight how the institutional design of aid delivery mediates the tension between donor control and recipient agency, offering critical implications for the study of multilateral delegation and aid effectiveness.

The authors' abstract, as published at the source. The Review of International Organizations, 2026 · DOI ↗