46721 The World Bank PREMnotes S E P T E M B E R 2 0 0 8 N U M B E R 124 POVERTY Thinking about Aid Predictability Matthew Andrews and Vera Wilhelm Effectiveness and aid mitments are not good predictors of aid predictability disbursements and long-term commitments Researchers are giving more attention to aid by official lenders are unreliable. The studies predictability. In part, this is because of in- measuring the predictability of aid include creases in the number of aid agencies and aid work by Bulí and Hamann (2003, 2007); dollars and the growing complexity of the aid Celasun and Walliser (2008); Fielding and community (World Bank 2008). A growing Mavrotas (2005, 2008); Gemmel and McGil- body of research is examining key questions: livray (1998); Pallage and Robe (2001); SPA Is aid unpredictable? What causes unpredict- budget support surveys; and OECD-DAC ability? What can be done about it? This note surveys on monitoring the Paris declara- draws from a selection of recent literature to tion. For example, the results of the latest bring some clarity to the basic story emerg- OECD-DAC survey (OECD 2008) show that ing.1 We start by presenting evidence from on average only 45 percent of aid is delivered the literature on various problems with aid on schedule. Celasun and Walliser (2008) flows. Then we discuss how researchers use find disbursements are not always less than terms like volatility and unpredictability when commitments but that the "absolute devia- discussing aid predictability; we suggest that tion in per cent of GDP of committed and these concepts can be sharpened by intro- disbursed aid" in sub-Saharan Africa was ducing two new concepts--expectations and greater than three percent of GDP over the reliability. These new concepts are particularly 1990­2005 period. useful in conceptualizing the problems of Predictability varies by type of aid and unpredictable flows in government budget is associated with the recipient country's processes. This approach allows a basic analy- environment. Fielding and Mavrotas (2005) sis of how timing and different types of aid find significant deviation of aid flows from affect predictability, and the implications for trend and note that programmatic aid flows policy making. less smoothly than project aid. They also note various influences on aid stability, including Is aid unpredictable? the quality of the recipient country's institu- Some evidence tions. Celasun and Walliser (2008) find that Recent research finds that the predictability the predictability of aid increases with the of overall aid and of various types of aid is length of coverage under IMF programs, a significant and potentially costly problem a factor that is seen as a proxy for a stable in aid-dependent countries. Donor com- country environment. A recent World Bank study (World Bank 2007) confirms the gen- FROM THE POVERTY REDUCTION AND ECONOMIC MANAGEMENT NETWORK eral finding that aid flows deviate more from year, they can be predictable. In these cases trend than domestic revenues. The Bank the aid is based on expectations (related to study also notes that IDA flows have lower a hard commitment tied to a meaningful deviations than other aid--and domestic schedule or rule for disbursement, which revenues--presumably because these flows allows budgetary authorities to plan around are committed on the basis of improvements the expected revenues). Expectations are in the institutional quality of recipient gov- then met in a reliable fashion (where the aid ernments. disburses according to the schedule or the Poor predictability is costly. In his recent rules specified in the commitment, such that work, Kharas (2008) estimates the cost of budgeted revenues actually flow as planned). aid flow deviations at about 22 percent of This type of aid flow can be used to smooth the value of aid itself--a large number by business cycles, and it might reflect lumpy any account. In other work (Kharas 2007) investment requirements or come in the form he discusses how the "new reality of aid" of emergency or disaster relief (expected in creates further challenges and may imply times of disaster and reliably triggered by further costs. New players--including the such). Millennium Challenge Corporation (MCC), The joint concepts of expectation on the Chinese, the Global Fund, and private the part of the aid recipient and reliability foundations--are providing aid in large are important in better understanding the amounts that lack detailed planning, can aid predictability problem--especially how enter budgets quickly and unexpectedly, volatility relates to predictability--and in and lack transparency. Kharas attempts to thinking about appropriate policy responses estimate some of these new flows (especially to this problem. Figure 1 combines the two from middle-income bilaterals and private concepts in a basic matrix with four aid-flow funders) using "guesses," a reflection on the types in the quadrants. Quadrant I suggests challenge fiscal planners must have in recipi- the ultimate goal of predictability--aid that ent nations. He states that, "The new reality is expected and reliable. In this case, fiscal of aid is one of enormous fragmentation and planners working in budgetary authorities volatility, increasing costs and potentially in developing countries have the ability to decreasing effectiveness." predict (based on schedule and rule-based disbursement information provided with Volatility and predictability commitments) and can be confident in the Volatility is a prominent word in the literature prediction (disbursements actually follow the on aid predictability. Frequently it refers to schedules and/or rules outlined in commit- the standard deviation of the change in aid ments). Aid in this quadrant could be volatile with a specific time horizon and therefore but is predictably so, thus not forcing any ill- does not imply direction (World Bank 2008a). advised or policy-defeating fiscal adjustments More volatile aid is likely to increase or de- on governments. crease in value more than less volatile aid. Quadrant II shows aid that budgeters OECD (2005) specifies that "aid is volatile expect (and program) but do not receive in when fluctuations in aid flows are large, rela- a reliable fashion. In simple terms the ex- tive to the volume involved." pectation upon which budgeters base their Volatility in aid flows does not always predictions proves weak, evidenced in a gap equate with unpredictable flows, however. between commitments and disbursements. Aid is predictable when recipient countries Celasun and Walliser (2008) measure exactly can be confident about the amounts and the this issue, finding significant absolute devia- timing of aid disbursements (OECD 2005, p. tions between commitments and disburse- 20; World Bank 2008a, p. 1). In other words, ments. This creates a hole in the budget even where aid amounts fluctuate year by in some periods (where commitments are 2 PREMNOTE SEPTEMBER 2008 Figure 1. Thinking about Aid Volatility and Predictability Reliable aid Unreliable aid Aid is expected and reliable Aid is expected but unreliable Expected aid I II Aid is unexpected but reliable Aid is unexpected and unreliable Unexpected aid III IV under-resourced) but can also create revenue Quadrant IV is where aid is both unex- spikes (where more money is provided than pected and unreliable, meaning there is no committed). In both cases the lack of reli- basis for predicting the aid in budgets. This ability can lead to volatility and undermine includes aid that is provided off budget and the allocative efficiency of budgets in devel- without any coordination with budgetary oping countries. Celasun and Walliser show, authorities as well as aid that enters into for example, that unreliable flows can lead to budgets unexpectedly and in a haphazard cuts in investment spending during times of fashion. Examples include certain kinds of aid deficiency and increases in payments on food aid that are provided without any kind domestic debt and/or government consump- of preexisting commitment and with no set tion in times of aid excess.2 schedule for provision. Quadrants III and IV relate to aid that is Timing is important in thinking about unexpected by budgetary authorities, either volatility and predictability. This is reflected when being off budget altogether (and not in recent work by the Strategic Partnership coordinated with budgetary authorities) or for Africa (SPA 2007), which finds that 92.5 entering into the budget without being ex- percent of budget support committed in plicitly programmed into budgets. Aid that 2006 was actually disbursed in-year--the is not programmed is, by definition, a source budget support being expected and reli- of volatility (causing flows that budgeters able.3 In contrast, SPA finds the same budget did not forecast). The volatility may not be support programs less predictable over the negative, however, in that the unexpected aid medium term, with only 69 percent of these might relate to some kind of countercyclical programs even providing commitment de- flow. Disaster relief is an example, as is food tails (the basis of expectation) for 2008 and aid. Apart from being countercyclical, these only 35 percent for 2009. In terms of figure unexpected flows can also be reliable--mean- 1, this budget support fits into quadrant I in ing that they sometimes flow according to the short run in a good-performing country, established commitments and rules, fitting accommodating annual budgeting. However, into quadrant III. Disaster relief is sometimes the same budget support is a quadrant III or provided on the basis of agreements that IV candidate for the medium term, because trigger flows on certain conditions: when an it does not allow clear expectations of when earthquake exceeds 6 on the Richter scale, for money will enter medium-term fiscal frame- example, certain amounts of aid are provided works (the research does not show whether in specific time periods. The relief may be the funds flowed according to any reliable unexpected by budgeters (as is the disaster) rules once actually released). but once it is triggered, its quality should be The basic observation is that aid predict- judged by whether it flows in accordance with ability is time sensitive because expectations the commitment. are time sensitive. In contrast to the budget SEPTEMBER 2008 PREMNOTE 3 support example, some aid flows (like disaster however. Budget departments thus have less relief) may be negotiated over the medium certain expectations to build on when devel- term and expected to flow when disaster oping multiyear budget frameworks, making hits (actually programmed into a four-year, budget support a quadrant III or IV type aid medium-term budget in some countries). in the medium term. The longer the plan- However, there is never an explicit short- ning horizon the less certain are expectations term expectation for such aid flows written, and the more problematic the aid. for example, in annual budgets. Thus, when Budget support is scheduled with cer- disaster hits aid may be unexpected in the tainty only on an annual basis, when donors short run (not included in the budget) even can assess political conditions and those con- though it is reliably following the rules (es- ditions do not change dramatically (that is, pecially those triggering the aid) negotiated the country is considered a stable performer). for the medium term. In the medium term, Budgeters can program current-year com- therefore, disaster aid may fall into quadrant mitments with a fair degree of confidence8 as I while in the short run (within a budget year) a result, and research shows that annualized it might fall into quadrant III (unexpected to commitments are well serviced. Budget sup- the budgeters but reliably provided, accord- port, at least in a good-performing country, ing to trigger clauses).4 falls into quadrant I in the short term. In a fragile environment, however, large shares Aid type also matters of budget support can be unreliable even in This brief discussion alludes further to the the short term, and would therefore fall into importance of aid type in the discussion of quadrant II. predictability. The examples show that gener- Even in a stable environment, however, al budget support might have different short- the primary concern is that the lack of bud- term and medium-term predictability effects get support predictability over a multiyear compared to disaster aid, largely because the period minimizes governments' ability to two have very different commitment modali- allocate funds for investment spending (gen- ties5 (on which expectations are based). Such erally multiyear in nature). This increases differences in aid effectiveness and volatility the likelihood that budget support will be are commonly found in the literature, with used to fill short-term holes in consumption regard especially to project aid and budget and to pay off domestic debt, as suggested support.6 Although different studies use dif- in Celasun and Walliser (2008). ferent terms for the two types of aid, they all Project aid: Project aid is provided over commonly find budget support more volatile multiple years, according to specified plans than project-based aid (see, for example, and schedules. The multiyear plans allow Bulí and Hamann 2003, 2007; Fielding for clear expectations, both annually and and Mavrotas 2005, 2008). This should not over the medium term. Project supervision be surprising because different aid types are requirements, monitoring and evaluation argued to have different effects in general: timetables, and other donor-driven rules for The different aid predictability effects of dif- these projects generally result in fairly reli- ferent aid types emerge, it appears, because able disbursement over their lifetime.9 In the of differences in the processes underlying medium term, therefore, these projects often the aid types. These aid types are discussed fall into quadrant I in figure 1.10 below and summarized in table 1. Nevertheless, there are many reasons Budget support: In the medium term, do- why annual project aid may flow in an unreli- nors program budget support in the context able fashion, such as aspects of project com- of their own internal planning processes.7 mitment and design. Disbursements commit- These programs result in multiyear com- ted for the first year of a project might not mitments that are no more than indicative, materialize because the conditions for such 4 PREMNOTE SEPTEMBER 2008 disbursement are not met--either because into a budgeted environment without signal donors do not follow their own administra- and does not have a schedule of its own (thus tive procedures and rules (a quadrant II potentially being unexpected and unreli- scenario) or because there are government able). These programs tend to settle down delays due to capacity constraints and lag- over the medium term, with most principal ging speed in implementation (a quadrant recipients managing disbursements better in I scenario if disbursement delay is not the the second and third years. donor's responsibility). Multiyear commitments from entities These delays can cause project-related like the Global Fund also allow solid expec- aid flows to falter in the short-run--meaning tations for fund flows, and these are gener- that annual budgets routinely might include ally reliable over the medium term (most project allocations that do not materialize in programs disburse above 90 percent of the that year. Projects that fall behind schedule committed amounts in the medium term). in one year might catch up in others, creat- These interventions thus might be quadrant I ing bunching. These predictability issues are funding sources (predictable) over a four- or short term in nature and make it difficult for five-year period.12 fiscal planners to set budgets with any kind of Private giving: Private giving has not certainty; where funds deviate from expected been captured in studies on predictability amounts they result in the misallocation of and is still subject to "guesses" (Kharas 2007). resources (especially skilled personnel) across It is provided to many parties including government.11 quasigovernment entities and nonprofits Vertical funds: Vertical funds are pro- (like vertical aid) and is generally off budget. vided either by single donors or by combi- This type of aid is increasing in size and has nations of donors, to flow to specific policy implications for public resource allocation areas. Prominent examples include funding (especially people, who can be displaced from for HIV/AIDS (PEPFAR and the Global Fund, budgeted government programs to off-bud- for example) and the environment. These get NGO-type interventions, for example). funds are committed and disbursed in many Thus, the flow of private giving is important ways, so it is difficult to provide a standard in any discussion on predictability. discussion. However, a basic characteristic As with vertical funds, one expects private of many vertical funds is rapid disburse- giving to disburse quickly after commitment. ment following commitment. This is seen in The commitment of such funds is often not Global Fund support in particular, which is very transparent or reflected in government committed after the approval of proposals fiscal timetables--and is thus not conducive (a process that can take years) and disbursed to short- or medium-term budget develop- shortly after commitment. ment. Private giving thus generally classifies The final commitment and initial deci- as unexpected aid and falls in quadrants III sion points often fall in the middle of budget and IV of figure 1.13 The essential issue is years, implying an inability to plan for such difficulty in predicting how, when, and in funds in the annual government budget. what amount these funds will enter into any Principal recipients in many countries strug- development setting, which makes it dif- gle to allocate any of the disbursed funds in ficult in the short and long run to allocate the first year (and even longer in many cases) resources. Private funds might unexpectedly of the program because of poor up-front result in a school being built, for example, planning (weak execution, procurement, and with no current or future plans to staff the supply management plans, for example). In school (or even to fund an inspection of the a short-run perspective, therefore, such aid school). Private funds might support the can create quadrant III and even quadrant purchase of drugs that cannot be dispensed IV problems, where unexpected money flows because of lack of infrastructure. SEPTEMBER 2008 PREMNOTE 5 Concluding thoughts: Clear depending on slower-moving conditions). thinking and clear policy? A country having problems with short-term Increases in aid and changes in aid architec- predictability of vertical funds and project ture, including ramped-up private giving, funds might seek to establish budgetary suggest that problems of aid predictability "buffers" (as recommended by Eifert and will continue and worsen. Unpredictability Gelb 2006). The influence of significant potentially undermines the effectiveness of off-budget funds from private foundations aid, as quality of spending is threatened and and other donors could be mitigated by the true value of aid quantities is reduced. coordination forums, or even rules whereby Studies reviewed here suggest unpredict- all donors have to register their intended ability is already significant. However, one interventions prior to action. Crucial for all needs to separate findings about volatility countries, however, is balancing different from those about predictability to really un- types of aid in the aid portfolio, which is a derstand the issues. A simple attempt at sepa- significant step toward smooth aid flows and rating findings identifies aid flow problems enhance overall predictability--in the short arising because of two dimensions: expecta- and medium term (IMF 2007a). tion and reliability. Where flows are expected and reliable, they may be volatile but are This PREM note was prepared by Matthew An- still predictable and do not undermine fiscal drews, Kennedy School of Government and Vera planning and allocation. Where aid is unex- Wilhelm (PRMPR). The authors are grateful pected and/or unreliable, aid predictability for comments received from Louise Cord, Punam is threatened and one can expect costs in Chuham-Pole and Jan Walliser. the form of forced resource adjustment. The problems and costs differ for different kinds References of aid and in the short and medium term, as Bulí, A., and A.J. Hamann. 2003. "Aid summarized in table 1. Volatility: An Empirical Assessment." IMF Note that table 1 is highly simplified. Staff Papers 50(1): 64­89. The impacts of timing and aid type on pre- ------. 2007. "Volatility of Development dictability are largely contingent on other Aid: An Update." IMF Staff Papers 54(4): contextual factors. For example, countries 727­39. enjoying strong institutions, transparent Celasun, O., and J. Walliser. 2008. "Predict- processes, and well-coordinated government ability of Aid: Do Fickle Donors Under- structures are likely to see more predict- mine Aid Effectiveness?" Economic Policy able aid in all forms. The basic message for (July): 545­94. policy makers is, however, that predictability Eifert, B., and A. Gelb. 2006. "Improving the problems differ from volatility problems, and Dynamics of Aid: Towards More Predict- that not all predictability problems look the able Budget Support." In S. Koeberle, Z. same--some are about unexpected aid and Stravreski, and J. Walliser, eds., Budget some about unreliable aid. The problem can- Support as More Effective Aid? Recent Expe- not be understood through general analyses riences and Emerging Lessons. Washington, and solved through general solutions, but will DC: World Bank. always be best understood in context of the Fielding, D., and G. Mavrotas. 2005. "The types of aid in the aid portfolio of specific Volatility of Aid." Economics Discussion countries and related problems. Papers No.0508. University of Otago, A country having problems with me- Dunedin, New Zealand. dium-term predictability of budget support ------. 2008. "Aid Volatility and Donor-Re- might pressure donors to strengthen the cipient Characteristics in `Difficult Part- medium-term commitment of donors (for nership Countries'." Economica 75(299): example, by locking in funds for three years, 481­94. 6 PREMNOTE SEPTEMBER 2008 Table 1. Different Types of Predictability Problems Aid flow Budget support Project aid Vertical funds Private giving characteristic Predictable aid In the short In the short and In the medium Potentially in run for good medium term, term, because of the medium performers, because of defined multiyear term, if funding because of defined multiyear commitments and is provided QI: Expected clear annual commitments and program length transparently and reliable commitments and project length limitations and within uncomplicated limitations general allocation disbursement discussions, and mechanisms according to a clear schedule Unpredictable aid In the short In the short run, n.a. Potentially in the run for fragile because donors medium term if states and poor do not follow "due funding is not Q2: Expected performers process," causing provided within and unreliable delays general allocation discussions, and according to a clear schedule In the medium n.a. In the short Presumably in term (potentially) run, because the short run and because of programs are medium term, unclear multiyear developed because resource commitments; outside of commitments are Q3 + Q4: and because resource not scheduled Unexpected and annual reviews allocation within the reliable/unreliable allow policy processes, and government and political are often poorly allocation adjustments planned in early process, thus periods creating perpetual "surprises" n.a. Not applicable. Gemmell, N., and M. McGillivray. 1998. ------. 2007b. "Aid Inflows--The Role of "Aid and Tax Instability and Government the Fund and Operational Issues for Budget Constraints in Developing Coun- Program Design." Paper prepared by tries." CREDIT Working Paper No.98/1. Fiscal Affairs Department, June. IMF, University of Nottingham Center for Washington, DC. Research in Economic Development and Kharas, H. 2007. "The New Reality of Aid." International Trade (CREDIT), Notting- Brookings Blum Roundtable 2007. ham, England. Wolfensohn Center for Development, International Monetary Fund (IMF). 2007a. The Brookings Institution, Washington, "Fiscal Policy Response to Scaled-Up Aid." DC. Available at: http://www.brookings. Paper prepared by Fiscal Affairs Depart- edu/~/media/Files/rc/papers/2007/08aid_ ment, June. IMF, Washington, DC. kharas/08aid_kharas.pdf. SEPTEMBER 2008 PREMNOTE 7 ------. 2008. "The Volatility of Aid." Draft 3. Note that budget support is also being pro- Working Paper, July. Wolfensohn Center vided earlier in the budget year than previously, for Development, The Brookings Institu- which gives budgetary flexibility in using the tion, Washington, DC. money (as opposed to receiving money in the final Mavrotas, G., and P. Nunnenkamp. 2007. quarter and being forced to make adjustments in "Foreign Aid Heterogeneity: Issues budgetary plans to that point). and Agenda." Review of World Economics 4. It logically follows that differences in the 143(4): 585­95. quality of short- and medium-term commit- OECD (Organisation for Economic Co-oper- ments undermine fiscal planners' abilities to ation and Development). 2005 Harmoniz- connect annual budgets to medium-term plans. ing Donor Practices for Effective Aid Delivery. The 2007 SPA report identifies this issue more Vol. 2. DAC Guidelines and Reference Se- effectively than other studies, in which it is dif- ries. Paris: OECD. Available at: http://www. ficult to see whether measured deviations are oecd.org/dataoecd/53/7/34583142.pdf. between disbursements and short-term (annual) ------. 2008. "2008 Survey on Monitoring commitments or disbursements and medium-term the Paris Declaration: Effective Aid by commitments. The difference matters, however, 2010? What It Will Take." OECD, Paris. in a practical sense, as it potentially distorts dif- Pallage, S., and M. Robe. 2001 "Foreign Aid ferent strategic processes--annual budgeting in and the Business Cycle." Review of Interna- the short-term sense and medium-term planning tional Economics 9 (November): 636­68. otherwise. SPA (Strategic Partnership with Africa). 5. The term modality refers to the organiza- 2007. Survey of the Alignment of Budget tional rules and processes governing the different Support and Balance of Payments Support kinds of aid flows. with National PRS Processes. Brussels and 6. There are also studies on other aid types, London: SPA. including food aid. World Bank. 2007. "The Role of IDA in the 7. The World Bank Country Assistance Strategy Global Aid Architecture: Supporting the is an example, but bilaterals and other multilater- Country-Based Development Model." als have equivalents. Report DC2007-0016, June. Prepared 8. See the 2007 SPA report. Kharas (2007, p. 5) by World Bank staff for the International cites the GAO finding that only 59 percent of Development Association, Resource committed Millennium Corporation Funds were Mobilization Department (FRM). World disbursed to nine compact countries by 2007 (with Bank, Washington, DC. the other 41 percent still uncommitted and thus World Bank. 2008a. "Drivers of Aid Predict- unscheduled). ability." Concept Note, p. 1. World Bank, 9. Adjustments to the time frame of projects Washington, DC. are often made, but these too are generally ------. 2008b. Global Monitoring Report 2008. scheduled, transparent, and thus facilitate clear Washington, DC: World Bank. expectations. 10. Studies finding these flows more reliable Endnotes than budget support seem to take a medium-term 1. The core papers reviewed for this note were perspective--analyzing whether medium-term presented at a recent workshop: "The Delivery of disbursements match medium-term commitments Scaled Up Aid: Does Predictability Matter?" For in both types. a workshop report, go to: http://go.worldbank. 11. If project funds do not disburse, then org/OI5S2XXNU0. Other papers are referenced project employees are underutilized. If project at the end of this note. funds disburse more quickly than expected, then 2. See also Mavrotas and Nunnenkamp (2007), employees are overutilized, and either project who discuss government responses to various aid quality suffers or personnel is drawn from other issues. areas. 8 PREMNOTE SEPTEMBER 2008 12. However, concerns about these funds go arising from continuous new donor commitments beyond predictability. Concerns include the (by different donors), could undermine long-term degree to which funds are off budget (in many institutional and policy actions of government. cases) and whether they draw resources from 13. This may not be the case if private foun- other policy areas of the government budget dations coordinate their flows with budgetary (for example, when the funds are earmarked for authorities. A foundation building a school, for specific donor-defined policy areas). These con- example, could inform government budgeters of cerns are heightened in countries like Rwanda its intentions and allow the budgeters to modify where earmarked funds for HIV/AIDS exceed the their own programs accordingly. government health sector budget by multiples. In these situations continual short-term disruptions, SEPTEMBER 2008 PREMNOTE 9 This note series is intended to summarize good practices and key policy findings on PREM-related topics. The views expressed in the notes are those of the authors and do not necessarily reflect those of the World Bank. PREMnotes are widely distributed to Bank staff and are also available on the PREM Web site (http://www.worldbank.org/prem). If you are interested in writing a PREMnote, email your idea to Madjiguene Seck at mseck@worldbank.org. For additional copies of this PREMnote please contact the PREM Advisory Service at x87736. PREMnotes are edited and laid out by Grammarians, Inc. Prepared for World Bank staff