How should official development assistance be allocated across countries?

A market test for aid allocation and country graduation

South African Reserve Bank Governor Lesetja Kganyago speaks at IMFC joint press conference with International Monetary Fund (IMF) Managing Director Christine Lagarde during the IMF/World Bank spring meeting in Washington, U.S., April 21, 2018. REUTERS/Yuri Gripas - RC1983EAB0E0

Official development assistance, defined as grants or concessional loans provided for the purpose of development, is in scarce supply. Donors are naturally concerned with how to allocate aid resources most effectively. 

Economics has much to say about how to allocate scarce resources, but there is surprisingly little use of economics, as opposed to politics and international relationships, in practical aid allocation models. An early example is a paper by Collier and Dollar (1998) that uses a reduced form model to link aid with poverty reduction by estimating the effect of aid on growth. This allows them to identify a number of variables that should enter an efficient aid allocation formula: country income levels, headcount rates (or poverty gaps), and the strength of the policy environment. A more recent example is Lea and Dercon (2016) who look at aid’s impact on reducing poverty now and in the future.

Neither of these examples, however, sheds light on many practical questions that donor country officials face. How should one compare lending to a small island economy with a relatively high per capita income with a grant to a low-income country? Should being designated a fragile state or a least developed country imply that more aid assistance is warranted than the amount suggested by the country income level? If so, by how much? What if the fragility is a self-inflicted wound of poor governance? Should aid decline with recipient country income levels and, if so, at what rate? How should a donor balance using aid to mobilize additional private finance in a middle-income country against providing the same level of assistance to a low-income country?

This paper tries to shed light on the questions above by taking a new approach to aid allocation. Instead of trying to assess aid’s impact on poverty reduction, we use a classical welfare economics framework to measure the “consumer surplus” to the recipient country that accrues as a result of aid. In such a framework, aid is modeled as an inframarginal source of foreign exchange. The consumer surplus per unit of aid is determined as the difference between the marginal price of foreign exchange in the recipient country, proxied by the cost of foreign borrowing on commercial markets, and the price the country pays for aid (zero for a grant and below market for concessional loans). Such an approach is feasible today, because many developing countries (including a dozen low-income countries) do actually borrow and have publicly available credit ratings, whereas in the past low-income countries (LICs) had no market access and therefore the cost of commercial foreign borrowing was unobservable.

Our approach is as follows. We build on the general finding that the cost of foreign borrowing is closely linked to credit rating notations and outlook prepared by the major ratings agencies (Afonso et al. 2011). We construct and estimate a model to estimate the relationship between credit ratings and a number of economic variables, following the approach taken by Ratha et al. (2007, 2011). We take advantage of the fact that there are 92 developing countries, including 12 out of 31 low-income countries, with external long-term sovereign ratings from at least one of the big three rating agencies (Standard and Poor’s, Fitch, and Moody’s), so the data reflect the wide range of situations in developing countries.

The results of the model are used in two ways. We can generate predicted credit ratings for those that do not have them, thereby extending the analysis to almost all developing countries, including most low-income countries. With almost complete country coverage, it becomes possible to make judgments on the cross-country allocation of aid. We can also infer the appropriate weights to assign to variables like per capita income, small island status, governance scores, or debt levels, based on their empirical contribution to creditworthiness.

Our results confirm standard findings that, ceteris paribus, more aid should go to the poorest countries and those with good governance. They also encourage us to advocate for:

  • Expanding aid to small islands and least developed countries;
  • Expanding aid to countries facing economic volatility in GDP or exports;
  • Expanding aid to high debt countries (if inherited from previous regimes);
  • Expanding aid when blended with large amounts of non-concessional finance;
  • Eliminating thresholds for aid graduation in favor of smoother declines as incomes rise.

Naturally, expanding aid allocations for some countries implies less funding for others. Accordingly, we also advocate for more market-based loans from official bilateral and multilateral institutions to be extended to developing countries. The welfare benefit for the recipient may be lower than a pure grant, but is probably still positive. Hence, additional access to such non-concessional public resources could offset welfare losses from aid cut-backs in some countries.

The paper proceeds as follows: Section 1 lays out the current state of aid allocation and existing models surrounding it. Section 2 introduces the aid principles guiding our discussion and analysis. Section 3 lays out the data and methods used in our analysis, while Section 4 presents results and Section 5 lays out implications, while Section 6 concludes.


  • Footnotes
    1. Aid allocation models have found political factors, such as voting alignment at the United Nations, colonial history, and even membership on the Executive Board of Directors of multilateral funds, to be significant. (Dreher et al, 2008; Alesina and Dollar, 2000; Morrison, 2013; among many others)