This paper argues that regional public goods in developing countries are under-funded despite their potentially high rates of return compared to traditional country-focused investments. Regional public goods only receive about 2.0-3.5 percent out of total ODA annually according to the definition used in this paper. The rate of return to regional investments is likely to be high, especially in Africa, where investments in regional infrastructure and institutional integration would reduce the high costs imposed by the region’s many small economies and many borders. There are several reasons for the under-funding of regional public goods. First, to produce regional infrastructure and manage multi-country institutions requires coordination among two or more developing country governments. The recent donor emphasis on countries’ “ownership” of their own priorities is more supportive of national programs. Second, bilateral donors prefer country-based transfers given their potential for providing geo-strategic and political benefits, and the multilateral banks have limited scope for lending for regional programs since their principal instrument is a loan to a single-country government that must guarantee its repayment. Countries could coordinate their borrowing, but that would require reaching agreement on attribution of the associated benefits to allow appropriate allocation of the burden of financing among two or more governments. Combined, these problems of coordination, (lack of) ownership, and attribution make financing of regional programs costlier for donors to arrange, and riskier in terms of their sustainability and benefits. In Africa the under-funding of regional public goods is primarily a political and institutional challenge to be met by the countries in this region. But the donor community ought to consider the opportunity cost – for development progress itself, in Africa and elsewhere – of its relative neglect, and explore changes in the aid architecture that would encourage more attention to regional goods.