A quarterly magazine of the IMF

March 2007, Volume 44, Number 1

Straight Talk

Aid Can Work

Aid effectiveness is getting better, even though it's tough to prove

François Bourguignon and Mark Sundberg

The international call for increased aid to fight HIV/AIDS, cut extreme poverty, and expand access to clean water presumes that aid promotes development. This presumption lies behind most of the $106 billion dollars that international agencies and bilateral donors spent in 2005 on official aid to developing countries. But the effectiveness of aid remains a highly controversial issue for economists and development practitioners, who have debated it ever since aid first appeared as an expenditure line in national budgets. The resurgence in interest coincides with the push to scale up transfers to poor countries to help them achieve the Millennium Development Goals.

From both sides of the political spectrum, attacks on aid effectiveness have become a growth industry. Some critics point to bad projects funded by aid, others to aid granted to bad governments, usually inferring this is the rule and not the exception. Several economists have used econometric models on cross-country data to test whether aid leads to economic growth, but the results have been largely ambiguous.

This approach is a bit like setting up a straw man only to knock it down. The aid industry unquestionably provides ample fodder for critics: many cases exist of aid funding poorly conceived, badly executed, unsustainable projects (for example, cement factories built far from sources of gypsum and sand). And some badly managed countries have, indeed, received millions, especially during the Cold War, when aid was extended for geopolitical objectives. At times, aid agencies followed fads that later proved misguided (recall the popular integrated rural development projects of the 1970s). This does not prove that all aid has been, or is, ineffective.

It is entirely unsurprising that many economists have found the relationship between aggregate aid and growth to be weak. Evidence suggests a high level of heterogeneity in the effects of aid, which comes on top of the typical statistical problems that arise in cross-country analysis. Multiple markers for development success—income growth, poverty reduction, literacy, access to sanitation, and inoculations—further complicate empirical analysis. Case studies do not solve this problem because of the difficulty of establishing a counterfactual: some argue that aid has not prevented growing numbers of poor in Africa; others argue that the situation would be far worse without aid.

Although these findings may make aid seem indefensible, much of the criticism is misguided. This isn't to say the impact of aid is easily known or that we can fine-tune aid to improve results. Even though it will be difficult for some time to come up with adequate evidence, there are strong grounds for believing that aid fosters development.

What a dollar can buy

The basic problem with assigning blame or credit to aid for outcomes is one of attribution. Donors would love to know what development outcome they are buying for each dollar of aid given. But it is not really possible to say, for example, how many children are inoculated for a dollar of aid. Additional children being vaccinated may be due to an aid-supported immunization program, but could also be attributable to extra health system funding, or possibly an unrelated improvement in health care delivery. And, of course, money is fungible, even with earmarking.

Between the disbursement of foreign aid and eventual development outcomes, there is a long causality chain. This chain has three main links that shed light on what we do know about aid effectiveness. First is the link between country policies (macroeconomic stabilization, regulation, trade, public finance) and final outcomes. Although not perfect, a stock of knowledge exists on the development impact of such policies. Economic research and evaluation can generate this knowledge through ex ante and ex post analysis of national experiences and impact evaluations of specific interventions.

The second link is the ability of policymakers to make appropriate policy choices given existing knowledge—in other words, the quality of governance: bureaucratic capability, institutional capacity, checks and balances mechanisms, and so on.

The third link—or, in effect, the beginning point—is between external donors or aid agencies and policymakers or even policies. Agencies provide funds and technical assistance, both of which influence the policy debate. But they often also seek to impose conditionality even though they operate with imperfect knowledge and little control over implementation.

Within this chain, aid effectiveness depends on the behavior of donors toward national policymakers, the governance environment, and the knowledge available on the development outcome of a large range of policies in alternative contexts.

Good governance counts

Each link can be monitored, albeit imperfectly. Even if some skepticism is in order and country specificity is not always easy to factor in, a lot is known about the impact of several policies on development outcomes. For example, macro stability ensures a good investment climate, and selective trade openness influences growth. This is also true at the project level, with impact evaluations considerably expanding our knowledge of what works and what does not at the household, firm, and community levels.

There's also a growing understanding of governance systems and how they lead to better policy formulation and, ultimately, development outcomes. Although causality is difficult to untangle, evidence shows that some dimensions of governance have a favorable impact on development. This logic explains the emphasis placed on governance in performance-based aid allocation systems used by development institutions such as the World Bank.

Thus, aid should be effective in countries that meet specific governance conditions and where agencies and governments have access to the right knowledge in terms of policies and programs. Some countries have, indeed, done well while relying to varying degrees on aid. But many examples exist of donors who have been unable to convince policymakers to adopt an effective strategy. Why? Political and historical drivers of aid allocation include postcolonial relations, commercial interests, the Cold War, and, more recently the "war on terror," with donors having only marginal interest in how the aid is actually used. Donors have often used aid to support what they think are the right development policies—such as public enterprise privatization or financial liberalization—often with little understanding of local parameters that may make such measures risky or even counterproductive. The difficulty of implementing policy conditionality seems obvious in hindsight. Supply problems that undermine aid effectiveness are also fairly well understood nowadays: donor fragmentation, aid volatility, lack of predictability, and "compliance costs" can deeply erode the value of aid to recipients.

Toward a new aid model

Increasing awareness of these difficulties should lead to better aid effectiveness. A new partnership model seems to be emerging that relies on the preceding principles and that some donors in a selected number of countries are applying.

Two features of official development assistance have emerged over the past 10 years: country ownership and aid allocation based on country performance. The first requires that donors align their support with the strategy picked by recipients, thus moving away from strict policy conditionality. This is easier said than done, but the Paris Declaration agenda, agreed to by more than 90 countries, represents a major shift from past aid practices and appears to be slowly having an impact. Indeed, in 2006, indicators of donor alignment were collected and monitored for the first time. The second feature bases the level and modalities of aid on what can be inferred from the development achievements recipient countries are expected to attain. These include the strength of their governance and policies and some intermediate indicators of future results. It is already clear that aid selectivity based on this type of performance indicator is on the rise.

The aid framework that is emerging increasingly resembles the familiar principal-agent model: principals (donors) recognize they cannot directly determine agents' (recipients') development effort for lack of relevant information and the relative ineffectiveness of strict policy conditionality. But principals can observe the nature of policies being implemented and infer the likely outcomes, such as rising literacy, immunization coverage, and falling poverty. The objective is to move aid commitments toward a contractual basis built on achieving development outcomes.

A remaining challenge is balancing aid between rewarding and encouraging good performance and addressing needs. The good performance models risk focusing aid on a few countries that already perform well. A need-based approach risks putting resources in poorly managed, inefficient, and possibly ineffective environments, and may undermine incentives for better performers. How should aid be given to failed states or countries with very weak governance—often where the need is greatest? These "fragile states" clearly need aid and require a different approach (see F&D, December 2006, p. 37). The global community is still struggling to improve practices in this area. We hope these troubled nations will take lessons from good performers and be able to replicate successful aid contracts, so long as donors and agencies keep attuned to local exigencies.