Aid may become even more unpredictable, but there are ways to tackle the problem
LOW-INCOME countries face many sources of instability. Their economies are usually dependent on a single primary commodity, making them particularly vulnerable to climate- or trade-related shocks, and their political systems are prone to destabilizing regime changes. And even though low-income countries have few ties with international capital markets—which can be a source of instability in middle-income countries—they are still vulnerable to the consequences of volatile financial flows in the form of aid. Like private capital flows, fluctuations in aid can occur because of outside changes (for instance, shifts in donor sentiment) or in response to perceived domestic changes (for instance, in governance and economic management).
In the years ahead, the volatility of aid flows is likely to increase. Donors are planning to markedly increase aid and step up coordination and selectivity of aid recipients to help poor countries reach the UN Millennium Development Goals by 2015. In addition, donors are shifting away from project aid to program aid (given in the form of direct budget or sector support)—and countries will be seeking to underpin long-term recurrent spending (such as recruiting teachers and increasing the pay of nurses and doctors) with program aid. This shift will help reduce transaction costs and drains on limited capacity caused by the need to implement a large number of projects. But program aid flows tend to be more volatile than project aid, which is usually committed up front and disbursed on a multiyear basis.
Thus, the development community runs the risk of slipping into a low-level equilibrium—that is, countries that budget prudently over the medium term would discount pledges of assistance; donors would then see fewer funding gaps, in turn causing aid commitments to fall behind intended increases or even in absolute terms. Signs of this happening are already evident, with many low-income countries discounting aid commitments in their plans. To improve aid predictability, donors must lengthen funding horizons, and the annual review and programming cycle must be strengthened at the country level. However, even if progress is made on these fronts, four major challenges remain:
- How can countries deal with residual short-run volatility in disbursements?
- Can donors lengthen their commitment horizons without excessive risk of misallocating aid?
- How should levels and trends in performance influence the amounts allocated to project aid and budget support?
- What is the role for results-based aid allocations—as distinct from policy-based allocations—and how can results-based systems be improved?
We studied each of these questions to find ways to improve the predictability of aid, especially aid delivered in the form of budget support. We built on the existing literature on this topic, which tells us that aid is quite volatile (Alĕs Bulíř and Javier Hamann, and others estimate that variability is 30–60 percent of the mean). Volatility is higher for countries that depend heavily on aid, and program aid tends to be more volatile than project aid. Commitments are often statistically unhelpful in predicting disbursements—astonishing given the importance placed on commitments in medium-term fiscal programs—and despite efforts to improve predictability, there has not been much progress. A large body of evidence suggests that the costs of large macroeconomic shocks, including aid shocks, is high. And anecdotal evidence suggests that efficiency costs associated with unstable budgetary revenues are large, and that unpredictable cash limits on spending undermine agreed programs and so weaken ministries’ accountability for results.
Cushioning aid shocks
Given that aid volatility is here to stay, what can countries do to smooth the impact of short-run fluctuations in disbursements? Reserves represent the first line of defense for aid-receiving countries, and countries can adapt reserves and fiscal rules to cushion aid disbursement shocks. But could countries also develop a parallel mechanism, such as a stabilization fund—modeled, for instance, on Chile’s copper revenue stabilization fund—as further protection? To find out, we simulated a simple model, using a reserve buffer to keep unplanned deviations from aid-financed spending within 5 percent of target levels. When reserves are plentiful, the fund operates in “high mode,” protecting against downside shocks. When reserves are below target, the fund operates more cautiously. One could imagine more sophisticated mechanisms to manage volatility, but this simple instrument suits our purposes.
Our simulations suggested three main points. First, a reserve tranche of two to four months of import cover (less than the average level of five months for countries receiving Poverty Reduction Support Credits from the World Bank) can smooth expenditures quite effectively under a range of levels of aid instability. Second, while the simulated reserve fund does sometimes go “bankrupt,” this requires three to five years of large negative disbursement shocks, giving ample lead time for donors to organize an emergency response. Third, moderate improvements in the stability of flows, and a working process to offset negative shocks with higher flows can greatly ease reserve management strategy.
Lengthening commitment horizons
Multiyear aid commitments are controversial because they run the risk of over- or under-providing aid in the event of significant changes in the performance of the recipient country. But are efficiency losses from suboptimal aid allocations serious enough to discourage multiyear commitments? How can donors adjust their aid in response to changes in a country’s performance, while still operating within a multiyear commitment horizon?
We calibrated a simple model of aid allocation on the World Bank’s Country Policy and Institutional Assessment (CPIA) to assess the trade-off between optimal allocation and predictability within the framework of the International Development Association’s (IDA’s) performance-based allocation system. CPIAs are estimated annually for all World Bank clients, and the ratings are a major component of the allocation formula. They are currently available to the public only in the form of quintile rankings but will be released more fully in 2006. The model (described more fully in Eifert and Gelb, forthcoming) assumes that the marginal effectiveness of aid falls as aid increases, and that countries with high CPIA scores can absorb more aid productively than countries with low scores. Given countries’ scores during 1999–2003, how large would efficiency losses from aid misallocation have been if five-year donor programs had been implemented in 1999? On the upside, how much can commitment rules reduce aid volatility?
We found, not surprisingly, that risks are larger if countries’ performance scores are volatile. Roughly half of the countries remained in the same CPIA quintile, with one quarter moving up and the other down. Most movements were across one quintile, but some countries slipped more (for example, Côte d’Ivoire and Zimbabwe). We considered three types of programs. The first, a pure performance-based system, delivered the optimal quantity of aid to each country in each year. The second, “pure precommitment,” held each country’s aid level over 2000–2003 at its (optimal) 1999 level. This locks in allocations to countries with deteriorating performance, thus also preventing the reallocation of aid to improving countries. The third, “flexible precommitment,” adjusted aid levels if a country’s CPIA score drifted one-third of a point above or below its 1999 level. This corresponds to about a 90 percent confidence interval given the likely standard error of the CPIA score (Gelb, Ngo, and Ye, 2004). This rule would, therefore, not adjust aid flows unless a country experienced a clearly observable performance change.
Under the pure performance-based system, allocations have an average standard deviation of 17 percent of 1999 levels, which is far lower than estimates of historical volatility. Flexible precommitment is successful in further reducing volatility, except for the low-rated countries—simulations show that it can halve variability for countries in the top four performance quintiles (see Chart 1). For those countries that stay on track throughout the program, it reduces variability all the way to zero. Of course, where programs go rapidly off track (as in several of the worst managed countries), the flexible rule has little stabilizing effect.
Chart 1Commit, but with safeguards
Source: Authors’ calculations, based on data from World Development Indicators and the Country Policy and Institutional Assessment (CPIA) database, World Bank.
Note: CPIA quintiles as ranked in 1999. Q1 represents first quintile (best-managed) countries, Q5 the fifth quintile (worst-managed) countries.
Pure precommitment over five years stabilizes flows completely, but is risky. Efficiency losses (relative to annual optimal allocation) under this option represent 10.7 percent of aid. But under the flexible rule, average efficiency losses drop to only 2.3 percent of aid because levels respond to large changes in performance. Losses are concentrated more heavily in the more poorly managed countries where absorptive capacity constraints bind more rapidly. Countries that perform consistently benefit the most from flexible precommitment.
The results of this exercise argue for using performance-based aid systems that allow latitude for small changes in ratings—the “typical” change of +/- 0.1 in a country’s annual CPIA score is within the range of measurement error. Small changes in scores do not usually foreshadow subsequent drift in the same direction. Indeed, minor changes in the survey instrument itself can cause small changes in ratings. There is little to gain from continually fine-tuning aid levels; identifying and responding to large changes is more important.
Balancing budget and project support
How much of the total aid given to a country should be channeled through budget rather than project support? Country circumstances will shape the answer, but some common principles may apply. The World Bank usually restricts budget support to stronger performing countries, a practice that has a better chance of providing stable financing in such cases. How would selectivity be implemented? It could be formula-based, where countries become eligible above a certain CPIA performance cutoff, and where the maximum share rises with performance. A precommitment formula could be adopted to maximize predictability—subject to adequate performance. If recipient countries prefer budget support, this provides them with an incentive to bolster performance. For very high-capacity clients, donors could validate budget support simply by “certifying” country systems.
However, there may also be a role for budget support in countries whose budget and financial management systems are still fragile, but where donors are willing to make an investment as a means to help strengthen them. This concept of budget support as an investment in country systems suggests that criteria for budget support should reflect both levels and trends in performance. How, then, to weight them? Too small a response to trends might not provide adequate safeguards or incentives to improve. But too strong a response reduces predictability and undermines the value of budget support itself. There is no simple answer, especially as it is not easy to distinguish small, observed trends in performance from measurement errors.
One approach would be to set a multiyear base level of budget support and supplement it by incentive payments of up to 10 percent based on interim “light” assessments of performance. These incentives would anticipate future performance–driven changes in the aid mix, and be applied to the following year’s support in order to improve predictability. Every three years—long enough to expect to identify changes—there would be a “deep” systematic review of progress in country systems, supported by independent assessment and comprehensive output and outcome measurement, including through surveys. This would feed back into the CPIA and help shape the decision on aid levels and how much to channel through budget support. Again, major performance changes in the interim should trigger a comprehensive review.
Improving results-based aid
In recent years, the development community has begun to shift away from emphasizing only policy prescriptions and actions as a basis for support to a focus on results, thus creating more room for aid recipients to develop their own policies. The European Union’s budget support programs represent the most ambitious move in this direction. They combine a fixed tranche with a variable tranche that disburses at a level based on the recipient country’s success in meeting a set of mutually agreed targets for service delivery (such as immunizations or primary enrollments) and public financial management. The European Commission recently concluded that this approach has been quite successful in combining a reasonable degree of predictability with performance-based incentives. However, it emphasized that the global community still lacks an analytic framework to guide the setting of targets.
What is an appropriate three-year target for raising primary enrollment or vaccination rates? How rapidly can literacy test scores improve or child mortality decline? Norms for target-setting can be derived in a number of ways. Clemens (2004) estimates long-run functions to derive norms for increasing primary enrollment. We use quantile regressions to relate annualized changes in infant mortality and mortality for children under five years old to their initial levels. Where mortality rates are already low, the scope for further gains are limited and changes tend to be small, but at higher mortality levels country performance diverges. Some countries have achieved rapid declines in mortality, whereas rates have stagnated or even risen further in others, generally because high mortality is a symptom of persistent problems (conflict, poor governance) or because new challenges are emerging (HIV/AIDS).
Quantile regressions allow us to study this relationship over a range of percentiles of country performance. With strong efforts, countries with infant mortality rates above 100 per 1,000 (as is the case in most of Africa) can reduce mortality at a rate of 2.5–3.5 per 1,000 a year (this would represent the 75th percentile of experience). For countries with weak systems and difficult circumstances, median performance may be a more appropriate target; at the median, infant mortality rates improve by 1.7–2.3 per 1,000 a year. These estimates might be used to target rates of improvement for forward-looking programs. Chart 2 illustrates the estimated “paths” from high to low infant mortality rates at different percentiles. Over 20 years, the 75th percentile country path would bring a country from an infant mortality rate of 150 to 100 per thousand. The 90th percentile country would drop the rate to 80 per thousand.
Chart 2Targeting improvement
Source: Authors’ calculations, based on data from World Development Indicators and the Country Policy and Institutional Assessment (CPIA) database, World Bank.
Changes in the underlying model of aid away from conditionality and fragmented projects and toward country leadership supported by more coordinated, harmonized, and selective donor flows require careful rethinking of how to design the mechanisms for providing support. Budget support is becoming an important mechanism, especially for countries with a stronger, and more consistent, performance record. But it will be important to ensure that the shift in aid modalities does not replace the problem of uncoordinated flows with that of coordinated, yet even less stable, support. Our findings suggest possible approaches.
First, fiscal and reserve management rules can be adapted to cushion short-term disbursement shocks that are not directly performance-related. Such a system will work best when there is a clear performance framework and with mechanisms to convene donors to respond to persistent deviations of disbursements from commitments. Reserve management and fiscal programming in low-income countries should thus take into account the objective of stabilizing spending.
Second, aid can be performance-based with far greater predictability than in the past. While unconditional multi-year commitment can be risky, flexible precommitment—where flows are committed several years ahead and revised only when performance deteriorates or improves to a substantial degree—appears to be a more attractive option. Relative to a system of continuously “optimal” allocation, efficiency losses from this option are modest, and predictability is improved except for lower-performing countries where aid levels will need to change in response to large swings in performance. Even without precommitment, however, flows calibrated by an IDA-type CPIA system are less volatile than historical aid flows.
Third, some emphasis on performance trends is appropriate, because budget support can be seen as an investment in country systems of budget management and service delivery, but the weighting on these trends cannot be too great as this would destabilize flows and nullify the gains from support. The weighting suggested here allows performance trends to be treated as signals of likely future changes in aid levels, advancing potential gains in the form of incentive payments without inducing excessive volatility.
Finally, there has been considerable debate on the merits of service delivery, output, or outcome-based indicators as alternatives to policy-based indicators, especially for budget support. We do not wish to take sides in this debate, seeing the two approaches as more complementary than competitive. But if output-type indicators are to be used, it will be important to have a reference framework for judging progress. If this is not done, countries setting more ambitious goals will be penalized relative to those aiming for more modest improvements. Little research exists so far on this question, but historical progress can be used in a comparative framework to construct performance-based norms.
Alan Gelb is Director for Development Policy in the World Bank’s Development Economics Department where Benn Eifert worked previously as a Junior Professional Associate.
Bulíř, Alĕs and JavierHamann, 2003, “Aid Volatility: an Empirical Assessment,”IMF Staff Papers, Vol. 50 (April), pp. 64–89.
Bulíř, Alĕs and JavierHamann, 2005, “Volatility of Development Aid: From the Frying Pan into the Fire?” paper delivered at Maputo seminar on aid, available athttp://www.imf.org/external/np/seminars/eng/2005/famm/pdf/hamann.pdf
Clemens, Michael A., 2004, “The Long Walk to School: International Education Goals in Historical Perspective,”Center for Global Development Working Paper 37 (Washington).
Eifert, Benn, and AlanGelb, 2005, “Improving the Dynamics of Aid: Toward More Predictable Budget Support, World Bank Policy Research Working Paper (forthcoming) (Washington).
European Commission, 2005, “EC Budget Support: An Innovative Approach to Conditionality and Development,”http://www.spapsa.org/index.jsp?sid=1&id=1100&pid=1137
Gelb, Alan, BrianNgo, and XiaoYe, 2004, “Implementing Performance-Based Aid in Africa,”World Bank Africa Region Working Paper 77 (Washington).
OECD/DAC, 2005, Paris Declaration on Aid Effectiveness (February) http://www.oecd.org/dataoecd/11/41/34428351.pdf