What sort of conditions should there be on aid?

Politicians, the media, bloggers and other armchair experts on development almost all agree that aid for developing countries should be conditional on reforms by recipient countries, and that aid should be tied to conditions about how the aid is used. But this approach is generally not supported by people who work in development (though some embrace it as an unfortunate political necessity.)

In this note, I will unpack the the reasons for this disagreement. I’ll look at the advantages and disadvantages of conditionality. My conclusion is that aid should take the form of long-term, predictable commitments, focused on countries that are pursuing policies that are likely to benefit the poor. I support aid “selectivity” linked to long-term outcomes, which is quite different from the policy conditionality that is currently conceived.

There are three distinct reasons why we might want to make aid conditional.

• Conditions on aid might increase incentives for policy reform by developing country governments.
• Allocating aid to countries with good policy environments might increase the impact of aid spending.
• Aid conditions might increase our ability to account for how the money was used and what effects it had.

Alongside these advantages, we should consider the possible disadvantages of aid conditionality.

• The conditions increase transactions costs, for both the donor and especially for the recipient.
• Conditions may reduce predictability, which in turn reduces the effectiveness with which aid is used.
• There is a possibility that some of the policy prescriptions are incorrect, either because they reflect donor interests or because some of the international experts have given poor advice.
• The conditions may undermine internal government systems for prioritising, allocating, managing and accounting for public spending.
• The imposition of external conditions may contribute to poor accountability of developing country governments to their own citizens.

The remainder of this note looks at the advantages and disadvantages in more detail.

Advantage 1. Encouraging policy reform (“incentives”)

It is a widespread view that donors should use aid to encourage recipients to improve their policies. There is a lot of sense in this: many problems in developing countries are partly or wholly the result of poor policy choices by governments. The hope is that by making aid conditional on policy reforms – such as restructuring state enterprises, liberalising the economy, or political reforms – donors might accelerate the changes which are likely to be essential for economic development. The policy changes which aid conditions are intended to produce may be as important as resource transfers, and possibly more so.

The problem with this view is that there is no evidence that policy conditionality is effective at promoting reform. For example, Tony Killick at ODI has marshaled a considerable body of evidence over the years, which concludes that aid conditions are only positively correlated with reform if there is a pre-existing intention on the part of the government to improve its policies. Other researchers have found the same.

This conclusion seems counter-intuitive, especially to those of us who believe that people respond to incentives. But decision-makers in developing country governments face an array of competing pressures. Experience and evidence show that while countries are often willing to agree to policy changes as a condition of receiving aid, the incentive is not sufficient to encourage them to maintain reforms if they become politically difficult. There are several possible reasons for this:

• Donors’ threats to withdraw aid are not perceived to be credible; decisions are seen as capricious; and donors have proven themselves reluctant to reduce aid if that impacts negatively on the poor, or involves canceling a project in which pat of the spending has already been sunk. Furthermore, donors are often unwilling to withhold aid from important allies and strategic partners even if conditions are not being met.
• Recipient governments do not reap sufficient political benefits from aid to compensate them for the political costs of reform. Aid is often (rightly) focused on marginalized and dispossessed communities with little political power; and donors choose to take much of the political credit for the benefits of aid. As a result, it may make little difference to the political position of developing country governments whether the aid arrives or not.
• Donors do not coordinate their approaches effectively, so that conditionality-related changes in aid from one donor is often compensated by offsetting changes by other donors.

Some of these barriers to creating effective incentives for reform could be overcome if donors were prepared to be tougher when aid conditions are not met, and to coordinate more effectively. But the last fifty years suggest little willingness on the part of donors to do this; and we have no evidence that, even if aid conditionality were more effectively applied, the incentives it creates would be sufficient to accelerate the needed policy reforms.

Advantage 2. Allocating aid where it will be most effective (“selectivity”)

The second possible benefit of aid conditionality, aside from whether it leads to changes in the behaviour of recipient governments, is that it might create a mechanism which induces donors to increase the flow of resources to good-performing governments while reducing the resources flowing to poor-performing governments.

The rationale for this approach is that aid spending in countries with a good policy environment produces greater benefits (e.g. in terms of long run economic growth) than the same amount of spending would produce in countries that have poor policy environments. So for a given aid budget, the total reduction in poverty will be increased if aid is targeted on those countries that can use the aid most effectively. Creating rules for allocating aid to countries that are likely to use it well is known in the jargon as “selectivity”.

The common-sense view is supported by a certain amount of empirical evidence. Collier and Dollar, building on the work of Burnside and Dollar, found that the effectiveness of aid in promoting growth depended in part on the policy regimes of recipient countries. According to these studies, an optimal aid allocation would favour countries with high levels of poverty, low per capita incomes and sound policy regimes.

However, there are several reservations to note about this approach:

• First, contrary to what is sometimes claimed, the evidence does not show that aid is ineffective in poor policy environments. It finds that aid is a bit more effective in good policy environments, but aid still has a large and statistically significant beneficial effect on medium term growth even in poor policy environments. The additional benefit of aid in good policy environments is small relative to the overall benefits of aid.
• Second, the statistical evidence for this finding is not robust. The existence of this effect is very sensitive to the specification of the statistical model. With modest changes to the structure of the model, the positive statistical relationship between aid and growth is confirmed, but the linkage to good policies disappears.
• Third, the very same studies that find a positive relationship with good policies also provide robust evidence that aid has the most effect if it is directed to countries which contain the most poor people. The most compelling conclusion to draw from the statistical evidence is not that we give too much aid to poor policy environments, but that aid is significantly skewed towards middle income countries rather than the poorest countries that could benefit most from it.

Overall, then, there is some empirical support for the view that aid is more effective if delivered to countries with good policy environments. But the size of the effect is modest, and the evidence is not statistically robust. The statistical evidence is not sufficient to justify the widely-held view that we should give aid only to countries with good policies.

Furthermore, the increase in the overall effectiveness that could be achieved by allocating aid on the basis of the number of poor people – that is, shifting aid to least developed countries – vastly exceeds the increase in effectiveness that can be obtained by channeling aid to countries with good policies. And yet those who advocate tying aid to policy performance very rarely argue with as much conviction the case for shifting aid towards the poorest countries, even though this is a much more statistically robust conclusion of the same analysis.

Advantage 3. Improving accountability to the taxpayers

Government aid is funded by taxpayers, whose money has been involuntarily taken from them to pay for it. Those taxpayers are entitled, at minimum, to know that their money has been used for the purposes intended and that it has produced the desired effects. Aid conditionality, which sets limits on how aid money can be used and demands information and evidence to show that this is what has happened, might contribute to the accountability to which taxpayers are entitled.

There is a significant danger, however, that conditionality gives an illusion of accountability which far exceeds anything that is possible in theory or in practice. Suppose a donor agrees to pay for a highway to be upgraded in a developing country. The donor can then show to its taxpayers that the money it provided was used for that purpose. But now the recipient government does not have to spend its own money on that same highway. The resources released can be used to increase rural health services, to buy military equipment, or to fund the lavish lifestyle of parliamentarians. Not only do donors not have any way to find out what has happened as a result of their aid, this is essentially unknowable, since nobody knows what would happen in a counterfactual world in which no aid is received. (This is known as the problem of “fungibility”). The donor can produce receipts to show that the money was spent on a road – known in some donor agencies as the “fig leaf” defence – but in reality there is no way to know what item of expenditure was increased as a result of the receipt of aid.

Another criticism of this approach is that it sets a much higher standard for accountability of aid spending than other forms of government spending. Government spending programs are designed on the basis of analysis and evidence that the spending will, on average, help the intended beneficiaries, that they are a cost-effective way to achieve the objectives, and that systems exist to ensure that the money is paid to people who are entitled to it under the rules. But recipients of agricultural subsidies or welfare payments are not are not required to account individually for how they used the money they have received. The taxpayers’ interest is in knowing that the overall approach is a cost effective way to achieve the results, not in having a detailed account of how the money was spent by the recipient. Perversely, to the extent that we insist on burdensome reporting of how aid is used, we may reduce the very cost-effectiveness that we say we want to ensure.

Accountability to taxpayers for how their money has been used is profoundly important, both because taxpayers have the right to know how their money has been used, and because continued and increased funding of aid programmes depend on taxpayers being willing to go on paying for it. But conditionality is not a good way to achieve this proper goal. Conditionality provides no information about the effect that aid has had on the recipient, and it adds to the bureaucratic costs of aid.

Disadvantage 1: Higher transactions costs

Aid conditions add significantly to the transactions costs of aid, for both recipient and donor.

Tanzania once estimated that it had to account for 20,000 different performance criteria a year to more than 50 different donors. Many African finance ministers reckon to spend 3 days a week or more talking to donors (or being talked to by them).

There are more than 100,000 foreign advisers resident in Africa , costing aid budgets a little under $10 billion. Some of these advisers provide valuable advice and expertise. This sort of knowledge-transfer is an important form of aid in its own right. But a substantial proportion of the work of these advisers is to design conditions to attach to aid, to monitor whether those conditions are being met, and to make judgements about what to do if they are not. Some of this work is valuable in its own right, since all governments need analysis of the programs they are implementing. But the bulk of it is jumping through hoops, which adds greatly to the overhead costs of aid, reducing the amount of taxpayers’ money actually being received by the people for whom it is intended.

I was once on a multi-donor mission to one of the poorest countries on earth, during which one of the large donors deliberately linked future aid to conditions which it knew had already been met. This bizarre agreement enabled the donor show that aid was linked to performance, satisfying domestic constituencies, while avoiding for both donor and recipient any risk that the conditions would not be met. Though the conditions has been met, this convenient fiction required auditors, statisticians and other experts to be brought in from abroad to gather information to prove conclusive that the conditions had been achieved, adding hugely to the transactions costs of the aid program for no benefit.

Disadvantage 2: Lack of predictability

The single biggest constraint on the effectiveness of aid is not the corruption or incompetence of the recipients, but lack of predictability by donors. It is shaming (or should be) that aid is the single biggest cause of fiscal volatility in least developed countries (more volatile than commodity prices, economic growth or tax collection). It is shaming that past aid levels are a better predictor of future aid flows than the formal projections and promises made by donors.

This lack of predictability in aid matters for developing country governments for the same reasons that macroeconomic stability matters in affluent countries: there are huge economic costs associated with variability in demand for both public and private services. Many developing countries face challenges that require reform with short-term costs and enormous long-term benefits. An example is public sector reform. If a country is to pay off redundant staff with inappropriate skills while hiring new staff to replace them, it requires a considerable investment over a number of years, albeit with huge returns in the long run. Other examples include liberalisation of state monopolies, or reductions in import tariffs. All such reforms with high costs in the next few years, but significant long term benefits, are unthinkable for developing country governments with little ability to borrow. Only if they have reliable access to long term aid flows can they begin to contemplate the sorts of reforms they need to make to make these essential changes.

Aid conditionality adds to lack of predictability. Sometimes aid is withheld because the policies or actions of the recipient have deteriorated significantly. But more often, aid is withheld for more technical reasons: accounts have not been submitted, or one of the conditions has yet been met in full. Sometimes the problem is that the donor agency does not have the capacity to process the information that it originally asked for. As a result, aid programs stop and start, with little link to the underlying state of policies and commitment of the recipient country or the effectiveness with which they are able to use the aid they receive.

Disadvantage 3: The wrong conditions might be imposed

Some NGOs who oppose aid conditionality dislike the policies that donors insist upon, such as increasing private involvement in the provision of public services, or reducing public sector deficits. That is not my view. The policies advocated by agencies such as the World Bank and the IMF, and bilateral aid agencies such as DFID, are generally sensible. For example, Africa is better off as a result of overcoming the macroeconomic imbalances of the 1970s and 1980s which led to huge deficits, run-away inflation and macroeconomic instability. I believe in the combination of macroeconomic stability, openness to trade, reform and liberalisation of state enterprises, and creating an improved climate for business, which are advocated as part of the “Washington Consensus”.

But these policies are complex to design and implement in developing country contexts, and there is huge variation in the competence of the technical advisers sent to draw up detailed policy prescriptions. I have personally seen governments being given advice, for example on sterilizing exchange rate inflows, or on the design of systems for public spending management, which is technically incompetent, out of date or downright ill-conceived.

Furthermore, some policy advice is given with the interests of the donor, not the recipient, foremost on the donor’s agenda. The UK provided aid to Malaysia to build the Pergau Dam, even though the UK Government had concluded that there were more cost-effective ways to generate the power it was intended to produce, because the consortium with the contract to build the dam was led by a British construction firm. (Aid is frequently conditional on buying goods and services from the donor, adding hugely to the cost and reducing its effectiveness.) It is very difficult for recipient governments to distinguish advice given impartially and in good faith from conditions designed to promote the commercial interests of the donors.

Disadvantage 4: Undermining internal government systems

I worked for many years in the UK Treasury, latterly on public expenditure management and control. The Treasury occupies a powerful position in the UK government. Spending departments have to negotiate performance agreements with the Treasury, on the basis of which their spending allocations are determined by ministers. Other departments have a strong incentive to work closely with the Treasury, as it is the only source of discretionary spending increases to enable them to respond to political priorities.

Suppose that a wealthy donor from Mars arrived in Britain and offered spending departments an alternative source of funding for all the additional projects and activities that they wanted to implement. Even if the Treasury were providing substantial core funding for the department’s spending, its power would evaporate overnight. Marginal spending decisions would be made bilaterally between departments and the men from Mars, with the Treasury left to try to keep track of what was going on while trying to influence decisions from the outside. Cabinet decision-making, coordinated prioritisation and accountability of ministers to the centre of government would disappear. The result would be a rash of uncoordinated initiatives operated by independent barons with substantial patronage of their own and little or no accountability to anyone except the martians.

The way we give aid to developing countries can have exactly this effect. Aid conditionality, in which aid is linked to particular programs and projects, prevents the government from making decisions about which programs are a priority. Performance agreements are made with donors, not with Finance Ministries or Cabinet. Separate project implementation units, managers, and auditors are hired by the project to ensure that the program remains on track with its conditionality. So government systems are hollowed out: resource allocation, performance management, auditing and accountability is moved from collective government decisions under the control of the Finance Ministry to individual ministers and officials in consultation with their donor patrons.

This is more than a theoretical curiosity. One common feature of many developing country governments is that systems for government resource allocation and accountability are inadequate; and this both causes and is caused by, poor economic performance. It is difficult to overstate the long-term importance of resolving this.

Disadvantage 5: Undermine government accountability

It is sometimes claimed as a benefit of aid conditionality that it enables governments to take difficult decisions, because Ministers can tell the affected constituencies that the decision was forced upon them by donors.

This is a double-edged sword, however. To the extent that domestic stakeholders accept that the government has little discretion because of conditions attached to aid, this reduces the accountability of governments to their own citizens. Governments can, and frequently do, blame donors for having imposed conditions upon them, and assert – often truthfully – that the eventual outcomes were beyond their control.

In the worst cases, the donors, acting individually or collectively, effectively take the place of the opposition that should hold the government to account. Admittedly, the capacity of some legislatures and political systems to carry out this function is weak, which leads donors to step into the role; but in doing so, they crowd out the role that should be filled by local stakeholders.

This is important because lack of government accountability is a common feature in many developing countries. The long-term solutions lie in strengthening legislatures, civil society and the media to enable them to hold the government to account, and so enable them to press for more rapid reform in the interest of the broader population. To the extent that aid conditionality obstructs increasing accountability of governments to their citizens, the system delays rather than accelerates reform.

Assessment: the need for long term partnership

The arguments for aid conditionality look unconvincing when considered in detail. There is compelling evidence that conditionality does not create meaningful incentives for reform, which is the main rationale, explicitly or implicitly, for those who advocate it.

There is some empirical basis for saying that aid should be targeted on countries with good policy environments, though the effect is commonly exaggerated and only partially supported by the evidence. Moreover, if this is our prime concern there are many, much more significant ways in which we could improve the allocation of aid budgets to increase the impact on poor people.

Set against the advantages of conditionality, there are significant disadvantages, both in overhead costs, but also in undermining the systems of decision-making, accountability and control that need to be strengthened if developing countries are going to prosper in future.

The proliferation of multiple conditions, foreign advisers, and competing commercial interests impose huge overhead costs on donors and recipients alike, and the lack of predictability of aid flows makes genuine underlying reform almost impossible.

There is nothing new in the need for reform of the aid system. This was spelled out in detail in the Berg report as long ago as 1981, and in many studies since. Donors should be more willing to engage in long term partnerships with developing countries, providing reliable, long-term financial support as well as expertise, linked to the country’s long-term progress towards economic growth and poverty reduction. It is high time that we moved to long term partnerships in place of the cacophony of conditions, projects and competing systems created in the name of aid effectiveness.

For once, this is not a case in which producer interests oppose reform to protect their jobs. Most aid workers would prefer to bring and end to the proliferation of aid conditions, and move instead to long term partnership, even though that would mean fewer jobs in the design and appraisal of aid projects. It is not them, but the media, politicians and pundits that have created and sustain this industry, for entirely honourable reasons. If we want to increase the effectiveness of aid, we would be better off without the immense burden of conditionality.

Owen Barder
December 2005

Killick, T., (1998), Aid and the Political Economy of Policy Change, Routledge, London and New York.

For example, Collier, Paul. 1997. “The Failure of Conditionality.” In Catherine Gwin and Joan Nelson, eds., Perspectives on Aid and Development. Policy Essay 22. Washington, D.C.: Overseas Development Council.

Devarajan, Shantayanan, David Dollar, and Torgny Holmgren. 2001. Aid and Reform in Africa. Washington, D.C.: World Bank.

Nelson, J.M., (1996), “Promoting Policy Reforms: The Twilight of Conditionality?”, World Development, Volume 25 Number 9, September.

Collier, P. and D. Dollar (1999), Aid Allocation and Poverty Reduction, World Bank Policy Research Working Paper No. 2041, World Bank, Washington.

Collier, P. and D. Dollar (2002), Aid Allocation and Poverty Reduction, European Economic Review, Vol. 26, pp. 1475-1500.

Burnside, C. and D. Dollar (1997), Aid, Policies and Growth, Policy Research Working Paper 1777, World Bank, Washington.

Burnside, C. and D. Dollar (2000), “Aid, Policies and Growth,” American Economic Review, Vol.90, No.4, pp.847-868.