Reshaping the International Aid Architecture
by Mark McGillivray
Growth in developing countries would be lower in the absence of aid. This is clear from the results of recent research on the macroeconomic impact of aid, evidence from micro studies, and field experience. One can reasonably infer that by implication poverty would be higher without aid. Aid works, on balance, therefore. Yet it is also true that aid has many failings.
Reform to the international aid architecture should not simply be reform for the sake of having reform. It should aim at improving the developmental effectiveness of aid by inter alia addressing these failings. This note addresses three perceived failings: (i) insufficient consideration of absorptive capacity constraints; (ii) under- and over- aiding; and (iii) year-on-year aid instability. It looks at why they are failing, and floats some proposals that aim to fix them. These proposals involve greater inter-donor coordination of aid allocation decisions than currently appears to exist. Ideally, this coordination should result in the joint determination of country level aid allocations, involving all donors, bilateral and multilateral. The case for this becomes stronger the greater the level of global aid, especially if it reaches the levels called for to achieve the Millennium Development Goals.
II. Aid Effectiveness Issues
Absorptive Capacity Constraints
While aid is positively associated with growth, there can be too much of a good thing, with aid being subject to diminishing returns. This is based on the findings of a number of studies that tested for non- linearity in the aid-growth relation- ship, with aid being positively related to per capita income growth up to a certain level of aid relative to recipient GDP and negatively related thereafter. This is a seemingly highly robust finding, with almost all studies testing for such a relationship finding evidence of its existence. These studies find that negative returns set in when the aid inflow reaches anywhere between 15 and 45 percent of GDP. This has been interpreted as indicating limited aid absorptive capacities, with recipient governments being limited in the amounts of aid they can use efficiently. Even if this constraint was removed, there will still be an upper limit to the amounts of aid that an economy can efficiently absorb. This issue is especially relevant to fragile states (those with critically poor policy regimes and weak institutions) as they will inevitably have very limited absorptive capacities, as conventionally defined. Increased aid to these countries, in line with proposals to achieve the MDGs, could actually increase poverty. Yet diminishing returns will in principle always be an issue in all countries; an economy can only efficiently absorb aid up to a particular level no matter how good its policies and institutions might be.
Absorptive capacity constraints need to be addressed urgently. Clarity is required over the precise levels at which aid impact becomes negative. The wide range of levels currently reported in the research literature is of little operational use to policy makers. Moreover, these aid levels, expressed as ratios of GDP, have typically been obtained from panel data for diverse samples of countries. One size will not fit all: country- or category-specific levels are required. It must also be recognized that empirical evidence of diminishing returns is based almost entirely on aid flows that have gone to recipient country governments. There are other potential, non-government channels through which aid might be efficiently provided. There will presumably be optimal levels of aid that each channel can absorb, and information is needed on these levels. The different forms in which aid is provided will require individual consideration. Information is also needed on how to increase the absorptive capacities of the various channels. Both the research and donor communities should collaborate to work towards the provision of this information. The donor community should provide aid amounts up to but not beyond these optimal levels, working directly with government and non-government sectors in developing countries.
Under- and Over-aiding
In an ideal world the donor community would provide aid to all recipient countries up to the point at which the incremental impact on poverty falls to zero in each.(1) Moreover, in such a world the incremental efficiency of aid should be equal in all recipients, so that reallocating a given amount of aid from one country to another results in no change in global poverty. The second of these criteria is perhaps an impossible dream. But the first is not, and it should be a guiding principle for donors.
The current evidence suggests that fragile states as a group are substantially under-aided, having received far less aid than their populations, per capita GDPs and policy and institutional performance levels would predict.(2) But there appears to be tremendous diversity within this group, with some countries being over-aided and others being under-aided, in this sense. The under-aided countries are typically those of little global political or strategic significance and supported by relatively few donors. While this information doesn t quite tell us how donors are performing in terms of the first above-mentioned criterion, it certainly rings alarm bells. Nor does it tell us anything about less fragile states; under- and over-aiding is relevant to all aid recipients.
What would appear to be needed is the better absorptive capacity information, and coordination among donors to minimize deviations from the maximum levels of aid consistent with these capacities. While this ideally requires agreement among donors regarding aid allocation criteria and allocative processes, it certainly requires much greater coordination among them.
Year-on-year instability in government revenues is bad for growth and poverty reduction. Yet research shows that aid inflows are more volatile than other sources of developing country government revenue. Fragile states are particu- larly subject to this phenomenon, as aid to these countries appears to be roughly twice as volatile as aid to other low-income countries. Possibly worse still, the volatility of flows to the under-aided countries is roughly twice that of the aid to the over-aided countries. This would on the surface be a relatively straightforward matter to address. Donors need to be more conscious of aid allocations over time, seeking to stabilize year-on-year variations, possibly through increased use of longer-term partnership agreements. But it is also a matter of greater coordination among donors, as a lack thereof could inter alia simply make smaller year-on-year variations much larger.
III. What's Next?
The absorptive capacity, under- and over-aiding and instability issues discussed above all converge. The second and third of these issues arise if there is deviation, spatial and inter-temporal, from aid levels corresponding to full capacity or optimal levels. Thus there is a clear message for the research community, which is to do more work and provide more operational clarity on absorptive capacities. It is up to donors to both evaluate and use where appropriate this information. Yet this is only part of the story. Donors need to decide how to use it, on what to do. This includes decisions on appropriate channels within recipient countries, so as to alleviate absorptive capacity constraints. Beyond this, what donors will hopefully do is better coordinate among themselves to ensure that variations from optimal, poverty reducing country level allocations are minimized.
One way of achieving the preceding outcome is, of course, some sort of global fund that administers all aid. But that will not happen. Less utopian is the joint determination of recipient country level allocations. This would involve all donors, bilateral and multilateral, or if not at very least the largest agencies, coming together to finally decide how much aid will be allocated in the following year or years and then acting in accordance with these decisions. If the DAC, in conjunction with major multilateral agencies can organize senior level fora on selected policy issues, why cannot it organize a similar annual event specifically to coordinate inter-recipient aid allocation?
Recognizing that politics imposes constraints on bilateral agencies, selected multilateral agencies could play a 'balancing role in this process, ironing out any residuals between actual and optimal allocations to the full extent possible. This is broadly consistent with the 'Multilateral Balancing Fund recently proposed by DFID. If joint determination is not possible across the board, for all country-level aid allocation decisions, then it at least should be possible for certain country groups. The fragile state group, arguably that with the greatest per capita need for aid, is a good starting point.
(1) This reasonably assumes an inverted-U relationship between aid and poverty, consistent with the absorptive capacity issue just discussed. Ideally this relationship would be observed directly, using data on aid and poverty. But given the lack of poverty data one would need to use GDP data instead, drawing appropriate inferences for poverty reduction. The best approximation of this outcome is the level of aid at which the incremental impact on per capita GDP growth reaches zero.
(2) The fragile state for the current purposes is defined as one belonging to the bottom two quintiles of the World Bank's Country Policy and Institutional Assessment (CPIA).
This paper was originally prepared for the OECD Development Centre Workshop 'The International Aid System: What's Next?', held on 4 February 2005 at the OECD Headquarters in Paris.
Mark McGillivray is the Director of WIDER research projects 'Measuring Human Well-being', 'Development Aid: a Fresh Look' and 'Millennium Development Goals: Assessing and Forecasting Progress'.