Chapter

II Conceptual Framework and Its Application to Five Countries

Author(s):
Andrew Berg, Mumtaz Hussain, Shaun Roache, Amber Mahone, Tokhir Mirzoev, and Shekhar Aiyar
Published Date:
March 2007
Share
  • ShareShare
Show Summary Details
Author(s)
Andrew Berg, Shekhar Aiyar and Mumtaz Hussain 

This chapter develops an analytical framework for examining policy responses to a surge in aid inflows, summarizes the evidence from a sample of five low-income countries, and seeks out general lessons that may be of relevance to other countries expecting scaled-up aid. The complete country studies are contained in subsequent chapters. The sample of countries focuses on strong performers in terms of institutions and economic policies. This permits the drawing of lessons relevant for situations in which, broadly speaking, policymaking is not dominated by macroeconomic disarray, misgovernance, or post-conflict reconstruction. The goal is to learn how to help those countries that are well-positioned, institutionally and in terms of the policy framework, to absorb large quantities of aid. An important number of such countries have emerged in the past decade or so, including in Africa (World Bank and IMF, 2005). The selected low-income countries satisfy two criteria: first, each (except Ethiopia) ranks relatively high on the World Bank’s indicator of quality of economic institutions and policies—the Country Policy and Institutional Assessment (CPIA)—and second, each received large amounts of aid in the late 1990s and early 2000s, including a surge in aid inflows at some point over the period. The countries covered are Ethiopia, Ghana, Mozambique, Tanzania, and Uganda.1

Macroeconomic Framework for Analysis of Increases in Aid Inflows

The macroeconomic impact of aid depends critically on the policy response to aid. In particular, it is the interaction of fiscal policy with monetary and exchange rate policy that is important. In order to highlight this interaction, it is useful to introduce two related but distinct concepts: absorption and spending.

Absorption is defined in this study as the extent to which the non-aid current account deficit widens in response to an increase in aid inflows.2 This measure captures the quantity of net imports financed by an increment in aid, which represents the real transfer of resources enabled by aid. Absorption captures both the direct and indirect increase in imports financed by aid, i.e., direct purchases of imports by the government, as well as second-round increases in net imports resulting from aid-driven increases in government or private expenditures. Absorption reflects the aggregate impact of the macroeconomic policy response to higher aid inflows, encompassing monetary, exchange rate, and fiscal policies.

Absorption can be defined and understood in terms of the balance of payments identity:

Current account + Capital account = ΔReserves.

Breaking the current and capital accounts into their aid and non-aid components and rearranging items yields the following identity:

Aid inflows = ΔReserves – (Non-aid current account + Non-aid capital account).3

Thus, an increase in aid can serve some combination of three purposes: an increase in the rate of reserve accumulation; an increase in non-aid capital outflows; and an increase in the non-aid current account deficit. The rate of absorption of an increase in aid is then defined as the change in the non-aid current account deficit as a share of the change in aid inflows:4

For a given fiscal policy, absorption is controlled by the central bank, through its decision about how much of the foreign exchange associated with aid to sell, and through its interest rate policy, which influences the demand for private imports via aggregate demand.5 The mechanism will depend on the exchange rate regime, but under any regime, the monetary authority can choose to accumulate reserves or to make them available for importers. In the extreme case where the central bank uses the full increment in aid to bolster international reserves and does not increase net sales of foreign exchange, none of the extra aid will be absorbed.

Spending is defined as the widening in the government fiscal deficit net of aid that accompanies an increment in aid:6

Spending captures the extent to which the government uses aid to finance an increase in expenditures or a reduction in taxation. Even if the aid comes tied to particular expenditures, governments can choose whether or not to increase the overall fiscal deficit as aid increases. The aid-related increases in expenditures could be on imports or domestically-produced goods and services. Analyzing spending is important because of the natural focus on the budget as a policy variable, and also because of the importance of tensions between the fiscal policy response to aid and broader macroeconomic objectives with respect to the exchange rate and inflation. These definitions of absorption and spending take into account, by construction, the fungibility of aid.7

Absorption and spending are distinct, though related, concepts and policy choices.8 If aid comes in kind, or if the government spends aid dollars directly on imports, spending and absorption are equivalent, and there is no impact on macroeconomic variables like the exchange rate, price level, and interest rate.9 This paper, however, concentrates on the more difficult and empirically relevant case where aid dollars are given to the government, which immediately sells them to the central bank. Subsequently, the government decides how much of the local currency counterpart to spend on domestic projects, while the central bank decides how much of the aid-related foreign exchange to sell on the market; and, in general, spending differs from absorption.10

Different combinations of absorption and spending define the policy response to a surge in aid inflows. The four basic combinations of absorption and spending are described below, together with a discussion of the macroeconomic implications of each. Box 2.1 provides a numerical example showing how the central bank and fiscal accounting work in each of these four cases.

Aid Absorbed and Spent

This is the situation assumed (explicitly or implicitly) in most discussions of the macroeconomic implications of aid inflows (Bevan, 2005). The government spends the aid increment and foreign exchange is sold by the central bank and absorbed by the economy via a widening of the current account deficit. The fiscal deficit is larger, but financed by higher aid. Spending and absorption of aid allows an increase in government spending by redeploying resources that had been devoted to the traded goods sector. In terms of the familiar national income identity Y = C + I + G + (XM), for a given output, a fall in (XM) allows a rise in G. The aid dollars fill the foreign exchange gap that would otherwise result.

Of course, output may not be fixed. Government expenditures may well increase output, both in the short run through the effects of increased spending on aggregate demand and in the long run through the increase in the capital stock. To the extent that output can be increased through a fiscal expansion without leading to a deterioration in the current account, however, these increases in aggregate demand and investment could have been undertaken without the aid flows. Aid absorption refers to the use of aid to finance changes in the current account deficit associated with aid-related increases in aggregate demand, investment, and output in general.

Box 2.1.Absorption, Spending, and Central Bank and Fiscal Accounting

In this numerical example, the government sells the aid dollars to the central bank and receives a local currency deposit at the central bank in return. Net international reserves (NIR) increase by 100 and net domestic assets (NDA) of the central bank fall by 100 (because government deposits with the central bank are a negative NDA item). This places the economy in the lower-right box of the matrix. What happens next depends on whether the central bank sells the foreign exchange and on whether the government increases the deficit. Each case is discussed in the main text. The example below assumes a floating exchange rate regime. With a peg, the accounting story would be the same, but the numbers and details would be different.

Central Bank and Fiscal Accounts: Example With Aid Inflow of 100
SpendDon’t Spend
AbsorbCentral Bank Balance SheetCentral Bank Balance Sheet
NIR0RM0NIR0RM–100
NDA0NDA–100
Fiscal AccountsFiscal Accounts
External financing1+100Deficit+100External financing1+100Deficit0
Domestic financing20Domestic financing2–100
Don’t absorbCentral Bank Balance SheetCentral Bank Balance Sheet
NIR+100RM+100NIR+100RM0
NDA0NDA–100
Fiscal AccountsFiscal Accounts
External financing1+100Deficit+100External financing1+100Deficit0
Domestic financing20Domestic financing2–100
Note: NIR = net international reserves; NDA = net domestic assets; RM = reserve money.

Refers to external financing of the deficit.

Refers to domestic financing of the deficit.

Note: NIR = net international reserves; NDA = net domestic assets; RM = reserve money.

Refers to external financing of the deficit.

Refers to domestic financing of the deficit.

Some real exchange rate appreciation may be necessary and indeed appropriate in response to a sustained higher level of aid. This is because some combination of exchange rate appreciation and (if there is excess capacity) increased aggregate demand is necessary to generate the increased net imports that aid allows.11

The degree of exchange rate appreciation required to absorb the aid will depend, in general, on the structural response of the economy and the extent to which aid directly finances imports. For example, real appreciation would be higher to the extent that aid inflows finance expenditures on nontradable goods rather than directly financing imports.12 On the other hand, if higher incomes feed strongly into higher import demand, and if the supply of nontraded goods responds strongly to the increase in their relative price, the real appreciation would be limited. In economies with significant unemployment and the potential for a quick supply response, the additional demand for nontradable goods could induce additional employment and production, with little increase in the price level and limited real appreciation. In the longer run, investments that increase productivity in the nontradable sector could also reduce or even eliminate the real exchange rate appreciation.13

The mechanism for real appreciation would vary depending on the exchange rate regime. In a pure float, the central bank would sell the foreign exchange associated with the aid, causing a nominal (and real) exchange rate appreciation. In a peg, the real appreciation would take place through a period of inflation, with the increase in government expenditure accommodated by the central bank. The increase in aggregate demand and real appreciation would again increase net import demand, leading the central bank to sell foreign exchange in defense of the peg.

Aid Neither Absorbed Nor Spent

The authorities could choose to respond to the aid inflow by building international reserves, and neither increasing government expenditures nor lowering taxes. In this case there is no expansionary impact on aggregate demand, and no pressure on the exchange rate or prices.14

Not spending the aid may be infeasible over a longer time period, as donors need to account for how their assistance has been used. Of course, money is fungible, so that in principle not spending aid dollars is compatible with undertaking the projects favored by donors, while cutting back on other budgetary expenditures. In practice, the extent to which this is possible would depend on the room available—both fiscally and politically—to cut expenditures in other areas.

Aid Absorbed but Not Spent

Increased aid inflows can be used to reduce inflation in those countries that have not yet achieved stabilization. In such a case, the authorities can sell the foreign exchange associated with increased aid inflows to sterilize the monetary impact of domestically-financed fiscal deficits. The result would typically be slower monetary growth, a more appreciated real exchange rate, and lower inflation. Aggregate demand may increase as the inflation tax declines, with a corresponding increase in private consumption and investment. The deterioration of the trade balance that often accompanies such a stabilization program is financed by the aid inflow.15

In countries that have already achieved inflation stabilization but have large domestic public debt, the government could use the proceeds from aid to reduce the stock of local currency government bonds outstanding. This would tend to result in increased private consumption and investment, which would raise net imports through the indirect effect of higher private after-tax income on import demand. The extra foreign exchange sold by the central bank would finance this increased demand for net imports. Again, some real exchange rate appreciation is likely to be necessary to mediate the increase in net imports.

Whether a strategy of retiring public debt through absorbing but not spending aid is appropriate in a particular situation depends partly on whether lower interest rates would translate into higher domestic investment and/or consumption. If there are no good private investment opportunities, for example, an increase in credit to the private sector could result in private capital outflows or a buildup of excess commercial bank reserves at the central bank.16 In addition, as with the neither-absorb-nor-spend strategy, donors’ needs to account for the use of their assistance may make it difficult to sustain a no-spending approach.

Aid Spent but Not Absorbed

A fourth possibility is that the fiscal deficit, net of aid, increases with the jump in aid, but the authorities do not sell the foreign exchange required to finance additional net imports. The macroeconomic effects of this fiscal expansion are similar to increasing government expenditures in the absence of aid, except that international reserves are higher. The increased deficits inject money into the economy.

In this case, the aid does not serve to support the fiscal expansion. This point is central and deserves elaboration. A transfer of real resources to the recipient country occurs only if aid finances additional net imports. Aid also serves as a way for the government to finance its domestic expenditures, as an alternative to domestic tax revenue or borrowing, either from the public or from the central bank. It may seem, therefore, that the financing of domestic expenditures, such as the hiring of nurses, is an alternative use for aid, in addition to imports. But this approach to the function of aid is misleading; after all, the government could always simply borrow from the central bank (i.e., print money) to finance increased domestic expenditures. Rather, the purpose of the aid is to provide the foreign exchange required to satisfy the increased demand for foreign currency resulting from higher import demand.17

Consider a thought experiment in which, for a given level of aid, the government first decides on the appropriate level of government expenditure and its financing. This set of decisions, in principle, takes into account the scope for seigniorage, the supply response to increased fiscal expenditures, the productivity of the resulting public investment and the generation of higher exports that may result, and other such factors. Then, aid increases. What has changed is not that the government could now productively hire, say, more nurses to fight HIV/AIDS. It could have done that before. The difference is that, whereas before such additional expenditures would have caused too much inflation or a nonfinancable deterioration of the current account through second-round increases in import demand, now the incremental aid increases international reserves, which could be sold to pay for the higher imports. But this is the definition of aid absorption; aid that is not absorbed cannot fulfill this function.18

There are several monetary policy responses to a situation in which aid is being spent by the government but not absorbed in the economy. Absent foreign exchange sales to mop up the additional liquidity, the monetary policy options are the same as in the case of any domestically-financed fiscal expansion. One option is to allow the money supply to increase. This is essentially monetizing the fiscal expansion and would tend to be inflationary. Without foreign exchange sales, the nominal exchange rate will tend to depreciate as well, with a larger supply of domestic currency pushing up the price of foreign exchange. The resulting inflation tax helps contain absorption by transferring resources from the private sector. If, alternatively, the authorities resist nominal exchange rate depreciations, then the resulting inflation would generate a real appreciation, the demand for imports would eventually increase, and the supply of exports would decrease. Provided that foreign exchange was eventually sold to satisfy this demand for net imports, the foreign exchange sales would have a sterilizing impact and dampen inflation. With some delay, then, the aid would be used and absorbed. Although this strategy has the advantage that aid is eventually absorbed, the period of inflation may carry costs of its own, especially in countries with a history of high inflation.19

Another response is to sterilize the fiscally-driven monetary expansion through the issuance of treasury bills. This strategy would tend to raise interest rates and crowd out private investment. In effect, there is a switch from private investment to government consumption or investment.20 This strategy is likely to be particularly difficult and costly in countries with thin financial markets.21

There are opposing effects on the real exchange rate in the spend-but-don’t-absorb case. In a given situation, the net effect will depend on specific factors, including the strength of contrasting policy choices and other influences, such as the terms of trade. The fiscal expansion tends to raise demand for nontraded goods, causing an appreciation; on the other hand, it increases import demand and lowers export supply, pushing the exchange rate toward depreciation. The net effect depends, inter alia, on the price and income elasticity of the country’s export supply and import demand. In addition, the central bank’s resistance to absorption creates pressures for real depreciation. In a float, aid-related liquidity injections will tend to depreciate the nominal and, in the short run, the real exchange rate. Over time, higher inflation and the associated inflation tax will reduce private demand and lower the real exchange rate and absorption. Alternatively, sterilization through the sale of treasury bills will also depress private demand and hence the real exchange rate and absorption. In a peg, only the sterilization channel operates.

Choosing an Absorption and Spending Combination

Which of these combinations is best in the face of extra aid depends on many factors, including the level of official reserves, the existing debt burden, the current level of inflation, and the degree of aid volatility. For specific situations, some responses are more promising than others:

  • To absorb and spend the aid would appear to be the most appropriate response under “normal” circumstances. In this case there is a real resource transfer through an aid-financed increase in net imports, and a corresponding increase in public expenditures.
  • To neither absorb nor spend may be an appropriate short-run strategy where aid inflows are volatile or international reserves are precariously low.22 Accumulating international reserves while avoiding an injection of domestic liquidity through fiscal expansion could help smooth the path of the real exchange rate if aid inflows are temporarily high but expected to fall. However, it is not an appropriate response to a permanent increase in the level of aid, unless Dutch disease concerns fully outweigh the benefits from the absorption of aid inflows (Appendix 2.2).
  • To absorb but not spend the aid might be an appropriate response if inflation is too high (possibly owing to a very expansionary fiscal policy), resources are scarce for private investment, or the rate of return on public expenditure is relatively low. Sustained non-spending of aid may not be feasible, however, given donor objectives.
  • To spend and not absorb would appear to be the least attractive option. The use of aid to build reserves while financing the increased deficit domestically is generally unwise. Inflation can only finance a small amount of expenditure; attempts to go further tend to raise little finance while damaging the economy (Selassie and others, 2006). The use of domestic sterilization is also unlikely to be a sensible medium-run strategy, as it tends to shift resources from the private to the public sector and does not allow the country to benefit from a real transfer of resources financed by aid.

Findings from Country Studies

Pattern of Aid Inflows

Table 2.1 shows the pattern of aid inflows for all the countries in the sample. Gross aid inflows are the sum of grants and loans, including both program and project financing. Net aid inflows are gross inflows plus debt relief, net of amortization, interest payments on public debt, and arrears clearance.23 This is the headline measure of aid inflows, since it best captures the actual inflows of foreign exchange and hence the scale of the macroeconomic challenge. All the countries in the sample received debt relief over the period, which, in turn, permitted the clearance of external arrears in some cases and increased net aid inflows in others.

Table 2.1.Patterns of Aid Inflows(In percent of GDP)
199819992000200120022003
Ethiopia1
Net aid inflows4.76.08.816.115.0
Gross aid inflows11.78.824.318.117.5
Ghana
Net aid inflows3.22.8–0.310.62.67.1
Gross aid inflows8.77.58.814.96.19.5
Of which: program aid1.81.93.85.62.65.1
Mozambique
Net Aid Inflows11.611.420.415.416.415.0
Gross aid inflows13.413.420.016.718.517.4
Of which: program aid6.36.35.37.07.96.6
Tanzania1
Net aid inflows4.66.67.57.96.67.6
Gross aid inflows13.312.712.812.510.510.5
Of which: program aid2.01.82.32.73.85.1
Uganda1,2
Net aid inflows8.49.414.213.712.9
Gross aid inflows9.810.313.913.812.9
Of which: program aid3.03.56.88.38.2
Note: Figures in bold represent periods of aid surges.

Data are for the fiscal year. For example, July 1998 to June 1999 is reported as 1999.

Uganda data revised for net, gross, and program aid. Compiling a consistent series for recent aid inflows in Uganda is complicated by extensive revisions to data. For fiscal year 2000/01, the data in the table include about $80 million per annum of off-budget aid, which is not accounted for in previous years. Excluding this amount would somewhat reduce the size of the aid surge, without changing the analysis in any significant way.

Note: Figures in bold represent periods of aid surges.

Data are for the fiscal year. For example, July 1998 to June 1999 is reported as 1999.

Uganda data revised for net, gross, and program aid. Compiling a consistent series for recent aid inflows in Uganda is complicated by extensive revisions to data. For fiscal year 2000/01, the data in the table include about $80 million per annum of off-budget aid, which is not accounted for in previous years. Excluding this amount would somewhat reduce the size of the aid surge, without changing the analysis in any significant way.

All five countries experienced a surge in net aid during the study period. The net aid increment ranged from an average of 2 percent of GDP in Tanzania to an average of 8 percent of GDP in Ethiopia. The level of net aid was also high in all countries, ranging from 7 to 20 percent of GDP. In Ghana, there were two different episodes of surging aid inflows, with a sharp increase in 2001 followed by a slump the next year, followed by another surge in 2003. In all other countries, the surge in aid was persistent, in that after the initial jump, aid inflows remained substantially higher than in the pre-surge period.

Net Budgetary Aid

Net budgetary aid is the sum of budget grants and loans (including debt relief), net of public debt service and arrears clearance. Net budgetary aid usually differs from net aid inflows to the economy, for example, because some aid is channeled directly to the private sector and spent on projects outside the government budget. In this sample, however, the two aid measures behave similarly, in broad terms. On average, net budget aid has increased in recent years in all five countries. This is illustrated in Figure 2.1, where t = 0 denotes the country-specific year in which aid increased substantially. While the aid surge was gradual and steady in Tanzania, it was more volatile in the other cases.

Figure 2.1.Total Net Budget Aid

(In percent of GDP)

There is a clear shift from project aid to program assistance (Figure 2.2). Since the inception of the Poverty Reduction and Growth Facility (PRGF) approach in 1999, donors have been increasingly willing to channel their assistance to the recipient country’s general budget. This eases administrative and institutional constraints in recipient economies, and gives recipient countries more flexibility in spending the aid.24 However, there is no obvious shift from loans to grants, except in Ghana (Figure 2.2). This distinction is potentially important because loans add to debt service costs in the 0.future and therefore have implications for debt sustainability, while grants do not. On the other hand, there is some evidence that grants may have an adverse impact on the government’s revenue collection, while loans may have a positive impact (Gupta and others, 2003).

Figure 2.2.Changes in Composition of Budgetary Aid

(As percent of total gross aid)

Source: IMF staff reports.

Macroeconomic Context, Real Exchange Rate, and Dutch Disease

Growth was generally robust in most countries both before and during the aid-surge period, although exogenous shocks set growth back in some years (Table 2.2).25 Three of the sample countries—Ethiopia, Tanzania, and Uganda—kept a tight curb on inflation, both before and during the aid-surge period. In Mozambique, however, the aid surge coincided with a sharp increase in inflation. Ghana’s inflation was high and volatile before and during the aid-surge period. The private investment-to-GDP ratio was mostly stable in the sample. In Ethiopia and Tanzania, average private investment during the surge period declined slightly relative to the pre-surge average. In Uganda, it increased in the surge period. In most countries, the average public-investment-to-GDP ratio was higher during the aid-surge period.

Table 2.2.GDP Growth, Inflation, and Private Investment(In percent)
Pre-Aid Surge Average1Aid-Surge Average1Difference
Ethiopia
GDP growth5.71.8–3.9
Inflation4.72.6–2.1
Investment/GDP16.419.63.2
Private9.89.4–0.4
Public6.610.13.5
Ghana
GDP growth4.14.60.6
Inflation85.220.5–64.6
Investment/GDP23.623.2–0.4
Private14.113.8–0.3
Public9.59.4–0.1
Mozambique2
GDP growth9.77.3–2.4
Inflation1.812.811.1
Investment/GDP30.440.910.5
Tanzania
GDP growth2.85.42.6
Inflation9.94.9–4.9
Investment/GDP15.517.82.2
Private12.411.5–0.8
Public3.26.23.1
Uganda
GDP growth6.65.6–1.0
Inflation3.02.7–0.3
Investment/GDP19.621.01.4
Private11.213.92.8
Public8.57.1–1.3

For Ethiopia, Ghana, and Uganda, 1999–2000 is the pre-surge period and 2001–03 is the aid-surge period. For Mozambique, 1998–99 is the pre-aid-surge period and 2000–02 is the aid-surge period. For Tanzania, the corresponding periods are 1998–99 and 2000–04.

Mozambique lacks reliable data on private investment.

For Ethiopia, Ghana, and Uganda, 1999–2000 is the pre-surge period and 2001–03 is the aid-surge period. For Mozambique, 1998–99 is the pre-aid-surge period and 2000–02 is the aid-surge period. For Tanzania, the corresponding periods are 1998–99 and 2000–04.

Mozambique lacks reliable data on private investment.

There was no sign of Dutch disease in the sense that real exchange rates in the sample did not appreciate during the aid surges (Table 2.3).26 During the years in which aid inflows surged, there was typically a depreciation of the real effective exchange rate, ranging from 1.5 percent (Mozambique in 2000) to 6.5 percent (Uganda in 2001).27 Ghana, however, observed a small real appreciation over both episodes of surging aid inflows (Figure 2.3).

Table 2.3.Real Effective Exchange Rate(Percent change over previous year, unless otherwise specified)
Pre-Aid Surge Average1Aid-Surge Average1Difference
Ethiopia
REER (depreciation –)–2.0–2.1–0.1
NEER (depreciation –)–5.0–1.53.5
RER (bilateral with dollar)–5.7–1.93.8
Terms of trade–18.3–4.114.2
Exports–9.6–0.19.5
Nontraditional exports/exports (percent ratio)44.063.519.5
Ghana
REER–17.50.518.0
NEER–27.8–17.810.0
RER (bilateral with dollar)–12.0–6.65.4
Terms of trade–12.79.722.3
Exports–3.88.912.7
Nontraditional exports/exports (percent)30.233.02.8
Mozambique
REER–0.0–6.4–6.4
NEER1.6–14.1–15.7
RER (bilateral with dollar)–5.0–11.1–6.1
Terms of trade–8.81.210.0
Exports11.239.528.3
Tanzania
REER–2.3–9.8–7.5
NEER6.3–8.7–15.1
RER (bilateral with dollar)3.5–6.1–9.6
Terms of trade3.3–4.1–7.3
Exports–17.316.133.4
Nontraditional exports/exports (percent)41.374.733.4
Uganda
REER–6.6–6.30.3
NEER–8.6–5.82.8
RER (bilateral with dollar)–12.0–6.65.4
Terms of trade–14.0–3.610.4
Exports1.14.02.9
Nontraditional exports/exports (percent)51.478.927.5
Note: NEER = nominal effective exchange rate; REER = real effective exchange rate; RER = real exchange rate.

For Ethiopia, Ghana, and Uganda, 1999–2000 is the pre-surge period and 2001–03 is the aid-surge period. For Mozambique, 1998–99 is the pre-aid-surge period and 2000–02 is the aid-surge period. For Tanzania, the corresponding periods are 1998–99 and 2000–04.

Note: NEER = nominal effective exchange rate; REER = real effective exchange rate; RER = real exchange rate.

For Ethiopia, Ghana, and Uganda, 1999–2000 is the pre-surge period and 2001–03 is the aid-surge period. For Mozambique, 1998–99 is the pre-aid-surge period and 2000–02 is the aid-surge period. For Tanzania, the corresponding periods are 1998–99 and 2000–04.

Figure 2.3.Real Effective Exchange Rates and Aid Inflows

Source: IMF and country authorities.

Note: All indices are 100 at t = 0.

A real depreciation in the face of surging aid inflows may indicate (1) structural features of the economy such as a rapid supply response to aid expenditures or high import propensities, though this would tend to mitigate the appreciation rather than cause a depreciation; (2) a fiscal and monetary policy stance that leans against real appreciation; or (3) other exogenous events, notably a negative terms-of-trade shock. Subsequent sections consider the first two explanations. With respect to the latter, two countries in the sample, Ethiopia and Uganda, were hit by significant negative terms-of-trade shocks during the aid-surge period. However, as shown in Box 2.2, even in these cases the incremental aid flows were much larger than the scale of the terms-of-trade shocks. There is also no case where a significant change in private inflows counteracts the pattern of aid inflows.

Consistent with real depreciation, export performance was strong in most of the sample, especially in Mozambique and Tanzania. In Ghana as well, export performance was strong despite a stable real exchange rate. In both countries that were affected by the decline in coffee prices, real depreciation helped export performance. In particular, nontraditional exports grew strongly, and increased as a proportion of total exports, enabling robust export growth in Ethiopia and moderating the decline in exports in Uganda.

Was Incremental Aid Absorbed?

In general, country authorities were reluctant to fully absorb the incremental aid. Following the framework introduced earlier in this chapter, Table 2.4 decomposes the increment in aid in each country into the change in the non-aid current account, the change in the rate of reserve accumulation, and the change in the non-aid capital account.

Table 2.4.Balance of Payments Identity(Annual averages in percent of GDP)
Pre-Aid Surge Average1Aid-Surge Average1DifferenceIncremental Aid Absorbed?2
Ethiopia
Net aid inflows5.313.38.0
Non-aid current account balance–9.2–10.8–1.6Partly absorbed
Non-aid capital account balance2.01.3–0.720%
Change in reserves (increase –)1.9–3.8–5.7
Ghana
Net aid inflows1.36.85.5
Non-aid current account balance–13.4–3.410.0Not absorbed
Non-aid capital account balance9.92.1–7.80%
Change in reserves (increase –)2.2–5.4–7.6
Mozambique
Net aid inflows11.517.45.9
Non-aid current account balance–19.7–23.6–3.9Mostly absorbed
Non-aid capital account balance8.78.3–0.466%
Change in reserves (increase –)–0.5–2.1–1.7
Tanzania
Net aid inflows5.67.82.2
Non-aid curent account balance–9.2–6.82.3Not absorbed
Non-aid capital account balance4.11.7–2.40%
Change in reserves (increase –)–0.6–2.7–2.2
Uganda
Net aid inflows8.913.64.7
Non-aid current account balance–10.1–11.4–1.3Partly absorbed
Non-aid capital account balance1.6–1.1–2.827%
Change in reserves (increase –)–0.4–1.1–0.7
Source: IMF staff reports.Note: Errors and omissions are included in the capital account.

For Ethiopia, Ghana, and Uganda, 1999–2000 is the pre-surge period and 2001–03 is the aid-surge period. For Mozambique, 1998–99 is the pre-aid-surge period and 2000–02 is the aid-surge period. For Tanzania, the corresponding periods are 1998–99 and 2000–04.

Non-aid current account deterioration as a percent of incremental aid inflow is truncated at 0 and 100.

Source: IMF staff reports.Note: Errors and omissions are included in the capital account.

For Ethiopia, Ghana, and Uganda, 1999–2000 is the pre-surge period and 2001–03 is the aid-surge period. For Mozambique, 1998–99 is the pre-aid-surge period and 2000–02 is the aid-surge period. For Tanzania, the corresponding periods are 1998–99 and 2000–04.

Non-aid current account deterioration as a percent of incremental aid inflow is truncated at 0 and 100.

In three countries, the aid led to some increase of the non-aid current account deficit, but this increase was typically modest, except in Mozambique, which absorbed over half the incremental aid inflow. In Tanzania and Ghana, the non-aid current account deficit actually shrunk by 2 and 10 percentage points of GDP, respectively, implying that the incremental aid was not absorbed. In all countries, the aid surge increased the rate of reserve accumulation. This pattern is consistent with the failure of the real exchange rate to appreciate in line with the surge in aid inflows.

In all countries, part of the aid increment was lost through reductions in the rate of capital inflow. In Ghana, the deterioration in the non-aid capital account exceeded the entire increment in the aid inflow. In Tanzania and Uganda, the reduction in the rate of non-aid capital inflows was comparable to the aid surge. Some short-run movements in the non-aid capital account could reflect lags between foreign exchange being made available for absorption and the actual increase in imports that comprises absorption.28 However, this would not seem to be an adequate explanation for the more sustained changes observed in the sample.29Box 2.3 explores possible links between private capital flows and aid.

Was Incremental Aid Spent?

The governments in Mozambique, Tanzania, and Uganda spent most of the additional foreign assistance (Table 2.5). A variety of factors encouraged these countries to spend the incremental budgetary aid. Because these countries had attained macroeconomic stability in the mid-to-late 1990s before the aid surge, reducing domestic financing of the budget deficit was not a major goal. Neither was retiring domestic public debt a key objective, as these countries had rather low domestic financing of the deficit as well as domestic debt and debt service prior to the aid surge (Table 2.6). Furthermore, they had strengthened their expenditure management systems, partly because of the Heavily Indebted Poor Countries (HIPC) Initiative, which helped them spend most of the incremental aid that they received as program assistance.30 To the extent that these countries spent the aid increments, the additional spending was concentrated on capital and poverty-reducing expenditures.

Table 2.5.Allocation of Incremental Net Budgetary Aid: Spent or Saved(In percent of GDP)
Pre-Aid Surge Average1Aid-Surge Average1DifferenceIncremental Aid Spent or Not?2
Ghana
Net fiscal aid inflows1.37.36.0
Revenue (excluding grants)17.119.01.9
Expenditure (excluding external interest)26.929.32.3Not spent
Overall fiscal balance before aid–9.9–10.3–0.47%
Ethiopia
Net fiscal aid inflows5.311.25.9
Revenue (excluding grants)17.919.41.5
Expenditure (excluding external interest)31.832.50.7Not spent
Overall fiscal balance before aid–13.8–13.00.80%
Mozambique
Net fiscal aid inflows12.917.95.0
Revenue (excluding grants)12.613.91.3
Expenditure (excluding external interest)26.032.76.7Spent
Overall fiscal balance before aid–13.0–18.5–5.5100%
Tanzania
Net fiscal aid inflows4.78.63.9
Revenue (excluding grants)12.112.50.4
Expenditure (excluding external interest)16.720.74.0Spent
Overall fiscal balance before aid–4.8–8.3–3.591%
Uganda
Net fiscal aid inflows9.312.53.2
Revenue (excluding grants)12.612.80.1
Expenditure (excluding external interest)22.224.72.5Mostly spent
Overall fiscal balance before aid–9.6–12.0–2.474%

For Ethiopia, Ghana, and Uganda, 1999–2000 is the pre-surge period and 2001–03 is the aid-surge period. For Mozambique, 1998–99 is the pre-aid-surge period and 2000–02 is the aid-surge period. For Tanzania, the corresponding periods are 1998–99 and 2000–04.

Non-aid fiscal balance deterioration as a percent of incremental aid inflow is truncated at 0 and 100.

For Ethiopia, Ghana, and Uganda, 1999–2000 is the pre-surge period and 2001–03 is the aid-surge period. For Mozambique, 1998–99 is the pre-aid-surge period and 2000–02 is the aid-surge period. For Tanzania, the corresponding periods are 1998–99 and 2000–04.

Non-aid fiscal balance deterioration as a percent of incremental aid inflow is truncated at 0 and 100.

Table 2.6.Domestic Debt and Debt Service Indicators(In percent)
Pre-Aid Surge Average1Aid-Surge Average1Difference
Ethiopia
Domestic debt237.839.11.3
Interest payments37.45.6–1.8
Nominal interest rates on treasury bills43.41.6–1.8
Real interest rates in treasury bills4–0.8–2.1–1.3
Ghana
Domestic debt223.023.10.2
Interest payments328.027.5–0.5
Nominal interest rates on treasury bills438.126.5–11.6
Real interest rates in treasury bills419.31.7–17.6
Mozambique5
Domestic debt0.32.62.2
Interest payments30.23.83.6
Nominal interest rates on treasury bills411.824.012.2
Real interest rates in treasury bills49.011.12.2
Tanzania
Domestic debt210.19.5–0.6
Interest payments36.97.40.6
Nominal interest rates on treasury bills411.68.0–3.6
Real interest rates in treasury bills41.33.01.7
Uganda
Domestic debt23.48.14.7
Interest payments32.66.94.3
Nominal interest rates on treasury bills48.210.32.2
Real interest rates in treasury bills43.67.23.6

For Ethiopia, Ghana, and Uganda, 1999–2000 is the pre-surge period and 2001–03 is the aid-surge period. For Mozambique, 1998–99 is the pre-aid-surge period and 2000–02 is the aid-surge period. For Tanzania, the corresponding periods are 1998–99 and 2000–04.

Domestic public debt as percent of GDP.

Interest payments on domestic public debt in percent of government revenue.

Average interest rates on treasury bills (in percent).

In Mozambique, values for domestic public debt and treasury bill rates are for 1999 only.

For Ethiopia, Ghana, and Uganda, 1999–2000 is the pre-surge period and 2001–03 is the aid-surge period. For Mozambique, 1998–99 is the pre-aid-surge period and 2000–02 is the aid-surge period. For Tanzania, the corresponding periods are 1998–99 and 2000–04.

Domestic public debt as percent of GDP.

Interest payments on domestic public debt in percent of government revenue.

Average interest rates on treasury bills (in percent).

In Mozambique, values for domestic public debt and treasury bill rates are for 1999 only.

The governments in Ghana and Ethiopia, however, spent very little of the incremental aid. These countries had a relatively weak record of macroeconomic stability, and a low level of international reserves before the aid surge, which limited their ability to spend additional aid. As these countries also had relatively high domestic debt and domestic financing of the budget prior to the aid surge, reducing domestic public debt (and hence domestic debt service) was also a consideration for not spending the additional aid. In Ghana, which experienced highly volatile aid inflows, aid volatility appears to have been a major factor in saving incremental aid in 2003. In Ethiopia, limited administrative capacity and weak institutions following the conflict with Eritrea may have been additional factors.

The government can also spend aid indirectly by lowering taxes and transferring aid to the private sector. Contrary to theoretical considerations of moral hazard and some evidence from other countries,31 the revenue-to-GDP ratio either improved (in Ethiopia, Ghana, and Mozambique) or remained largely unchanged (in Tanzania and Uganda). For the latter cases, additional fiscal aid might have reduced the incentives for aid recipients to strengthen revenue efforts, as total revenues stagnated below 15 percent of GDP.32

Box 2.2.Terms-of-Trade Shocks and Aid Inflows

The table below attempts to disentangle the terms-of-trade effect from the aid-inflows effect for the two countries in the sample that were affected by a significant terms-of-trade shock during the aid-surge period: Ethiopia and Uganda. In both countries, the main export commodity is coffee. A sharp and prolonged decline in world coffee prices caused a deterioration in the terms of trade for both countries, and in each case this deterioration coincided with surging aid inflows.

The table contains estimates of the loss in dollar inflows through net exports resulting from the terms-of-trade shock, and compares it with the increase in dollar inflows due to the surge in aid. In this calculation, year t prices of exports and imports are fixed at the level of year t – 1. This yields a counterfactual series for exports and imports; the difference between this series and the actual data on exports and imports is taken as the terms-of-trade effect.

In both cases, in the first year the incremental aid inflow dominated the negative effect from the terms-of-trade shock. This is also true of the average over the aid-surge period. Nonetheless, in both cases there was a nominal and real depreciation.

Terms-of-Trade Shocks(In millions of U.S. dollars, unless otherwise specified)
19992000200120022003
Ethiopia
1. Terms-of-trade effect on net exports1–54–484–13–36–59
2. Change in aid inflows–152423541721
3. Net effect (1 + 2)–69–460222381–38
NEER (percent change)–8.4–1.65.9–1.6–9.0
REER (percent change)–5.11.1–3.5–4.92.2
Uganda
1. Terms-of-trade effect on net exports1–53–106–521160.1
2. Change in aid inflows–8254246–210
3. Net effect (1 + 2)–135–52194970
NEER (percent change)–14.0–3.2–6.92.3–12.7
REER (percent change)–13.0–0.2–6.5–1.7–10.6
Note: Figures in bold represent periods of aid surges. NEER = nominal effective exchange rate; REER = real effective exchange rate.

Calculated as the difference between actual net exports and net exports, keeping unit export and import prices unchanged.

Note: Figures in bold represent periods of aid surges. NEER = nominal effective exchange rate; REER = real effective exchange rate.

Calculated as the difference between actual net exports and net exports, keeping unit export and import prices unchanged.

In neither Ghana nor Ethiopia were program targets in IMF-supported programs generally responsible for the underspending of the aid increments. In most cases, programs supported by the PRGF allowed aid-recipient countries to spend the aid inflows. These programs accommodated expected increases in aid flows with a comparable expansion in targets for the fiscal deficit excluding aid (Figure 2.4).

Figure 2.4.Programmed vs. Actual Levels of Fiscal Deficit (Excluding Aid) and Net Budget Aid

(In percent of GDP)
(In percent of GDP)

Source: IMF staff reports.

Note: PRGF = Poverty Reduction and Growth Facility. Shaded areas represent the aid-surge period.

However, in a few exceptional cases, the IMF-supported programs envisaged only a partial spending of additional aid. One significant case is Ethiopia in the 2001/02 fiscal year, when a budgetary aid surge of about 5 percent of GDP was envisaged but the PRGF-supported program did not accommodate this expected increase in aid with a commensurate increase in public spending. In this instance, the program was aiming to reduce inflation driven by an excessive deficit and associated monetary financing during the war period of 1997–2000.33 In Ghana, the IMF-supported programs anticipated only partly spending the aid increments because the goal was to use part of the incremental aid to reduce domestic financing and thus lower domestic interest rates and the government’s interest bill. In the event, probably in part because of the high aid volatility, even these targets for spending were not binding, particularly in 2003.

Aid volatility also contributed to some governments’ choice of not spending the aid increments. There are two broad fiscal responses to aid volatility: adjust expenditures in line with the aid fluctuations, or smooth spending by building up reserves (and government deposits) when aid is up and drawing down these savings when aid shortfalls occur. For example, in Ghana, overall expenditures were relatively unresponsive to aid volatility, and past aid savings as well as additional domestic financing were used to offset shortfalls in budgetary aid. To some extent, the preference to build up official reserves in Ethiopia and Ghana appears to have contributed to saving the incremental aid, particularly in the second aid surge in Ghana in 2003. In general, the IMF-supported programs allowed for measures to mitigate the impact of volatile aid on public expenditures (Box 2.4).

Box 2.3.Aid, Absorption, and Capital Flows

Were the reductions in capital inflows in Ghana, Tanzania, and Uganda a result of the aid surge itself? If so, the aid surges did not serve their intended purpose of promoting absorption. Aid inflows that trigger capital outflows weaken the link between central bank sales of foreign exchange and absorption, because a (possibly volatile and unpredictable) part of the foreign exchange may leave the country through private capital account transactions.

The question arises as to how to relate the reduction in capital inflows in many sample countries to the policy response to aid. It is important to identify the direction of causation between the changes in aid and the changes in capital flows, and also to identify the appropriate counterfactual, which is what would have happened in the absence of aid.

Assume that the reductions in capital flows are exogenous to the aid flows. Then, there are two possible counterfactuals. First, the central bank could have sold dollars to accommodate the outflow even in the absence of the aid. Given this counterfactual, the aid allowed the authorities to accumulate reserves and pursue a don’t-absorb strategy. Second, in the absence of the aid surge the central bank could have resisted the capital outflow by not making dollars available in the foreign exchange market. This would have caused the domestic currency to depreciate, thus increasing net exports through the price effect. In this case, the aid dollars allowed the authorities to accommodate the capital outflow, increasing absorption relative to the counterfactual. But associated with this capital outflow would have been a reduction in demand for domestic money or assets. Thus, the use of reserves for this purpose would not have enabled financing of domestic spending. Therefore, whichever the counterfactual, it remains the case that the same aid dollar cannot be used twice: if a dollar is used to accommodate an exogenous capital outflow, it cannot simultaneously be used to finance government spending.

In Ghana, for example, the reduction in capital inflows seems to have been associated not with the aid surge but with macroeconomic disarray. Following a negative terms-of-trade shock and with reserves almost depleted, non-aid capital inflows fell sharply in 2000 and 2001. In 2000, the exchange rate weakened sharply and inflation shot up. With an aid surge in 2001, the authorities were able to avoid devaluing the exchange rate. In this case, the aid inflows likely kept absorption higher than it would have been.

Whether to treat changes in capital flows as exogenous to increased aid is, however, an open question. It is possible that an aid surge could encourage private capital inflows. For example, Buffie and others (2004) argue that increased aid, by reducing seigniorage and hence expected inflation, may lead to domestic agents substituting foreign currency assets with domestic currency assets, which would appear in the balance of payments as a capital inflow.

Similarly, an aid surge could trigger a capital outflow in certain circumstances. For example, when the authorities attempt to absorb but not spend aid, channeling incremental aid to the private sector through the financial system by reducing the stock of domestic bonds outstanding, lack of investment opportunities at home as well as portfolio diversification goals could encourage private investors to invest abroad—which will appear as a deterioration of the capital account. However, none of the countries pursued a policy of channeling aid to the private sector through the financial system. It would thus appear unlikely that such a policy resulted in the reduction in capital inflows observed during the aid-surge period.

An aid surge could also cause a capital outflow if it leads the authorities to pursue an excessively loose monetary policy. Aid-related fiscal spending tends to increase the money supply. If the authorities allow this to lead to excessively low interest rates and excess liquidity in the banking system, capital outflows could result. As discussed later in this chapter, aid inflows to Tanzania were associated with periods of relatively loose monetary policy, and this may have contributed to the slowdown in capital inflows. Direct evidence is scarce, however. Identifying the conditions under which aid would lead to a capital outflow or inflow is, therefore, a nontrivial undertaking that for now is left to future work.

Table 2.7 presents a summary of the absorption and fiscal responses discussed above. Surprisingly, a full absorb-and-spend response is not observed in any of the sample countries. Two countries (Ethiopia and Ghana) adopted a neither-absorb-nor-spend response to the aid surge. Both entered the aid-surge period with a precariously low level of reserves and used the additional aid to build those reserves. In Ethiopia, reserves were accumulated to bolster the de facto exchange rate peg against the dollar. In Ghana, a buffer against extremely volatile aid inflows was built. The remaining sample countries (Mozambique, Tanzania, and Uganda) spent the incremental aid without fully absorbing it. In all three countries, concerns about the negative impact of a real appreciation on competitiveness dictated the pattern of aid absorption. In general, aid was much more likely to be spent than to be absorbed.

Table 2.7.Policy Response to Aid Surge(In percent)
Not Spent1Partly SpentMostly SpentFully Spent
Not Absorbed2Ghana (0, 7)Tanzania (0, 91)
Partly AbsorbedEthiopia (20, 0)Uganda (27, 74)
Mostly AbsorbedMozambique (66, 100)
Fully Absorbed

“Spent” variable = non-aid fiscal balance deterioration as percent of incremental aid inflow. Truncated at 0 and 100. This variable is the second entry within brackets for each country.

“Absorbed” variable = non-aid current account deterioration as percent of incremental aid inflow. Truncated at 0 and 100. This variable is the first entry within brackets for each country.

“Spent” variable = non-aid fiscal balance deterioration as percent of incremental aid inflow. Truncated at 0 and 100. This variable is the second entry within brackets for each country.

“Absorbed” variable = non-aid current account deterioration as percent of incremental aid inflow. Truncated at 0 and 100. This variable is the first entry within brackets for each country.

Monetary Impact of Aid and Policy Response

On the basis of the policy response to aid, two main groups of countries can be discerned: (1) countries where the aid impact was limited because only a small part of it (if any) was either absorbed or spent (Ethiopia, Ghana); and (2) countries where expenditure exceeded absorption, resulting in an injection of domestic liquidity and creating upward pressure on prices (Mozambique, Tanzania, and Uganda).34 These patterns of absorption and spending defined the challenge for monetary policy in response to the aid surges.

Cases with Limited Aid Impact

In neither Ethiopia nor Ghana was the aid surge accompanied by an increase in domestic spending in excess of revenue generation. Therefore, over the period as a whole, the aid did not lead to a substantial injection of domestic liquidity.35

The monetary policy responses of these two countries were not restricted by the PRGF-supported program (Figure 2.5). In general, Ethiopia overperformed on monetary and reserve targets under the PRGF-supported program during the aid-surge period.36 This implies that the authorities’ chief concern—accumulating reserves to keep the nominal peg against the dollar—diverged to some extent from the PRGF target path, which envisaged more spending (supported by a higher target for net domestic assets—NDA) coupled with greater sales of foreign exchange (reflected in a lower reserves target).37

Figure 2.5.Ethiopia and Ghana: Limited Aid Impact

(NDA and NFR in millions of local currency, NIR in millions of U.S. dollars, and RM growth in percent)

Source: IMF staff reports.

Note: NDA = net domestic assets; NFR = net foreign reserves; RM = reserve money; NIR = net international reserves. Target data are occasionally unavailable. Wherever possible, adjusted program targets are used.

In Ghana, the strategy of the PRGF-supported program was to absorb and partly spend the expected aid increments. During the first aid surge in 2001, the government focused on reducing high inflation from the previous period, including sterilization through the sale of foreign exchange. While NDA and money growth exceeded the program targets, inflation came down over this period. Aid collapsed in 2002. When aid surged again in 2003, the authorities were much more cautious and saved most of the incremental aid inflow through the accumulation of government deposits at the central bank. Consequently, in this period, Ghana overperformed on both monetary and reserve targets under the PRGF-supported program; better adherence to the program path in 2003 would have resulted in more spending of aid combined with sterilization through foreign exchange sales. Presumably influenced by the negative consequences of the aid collapse in the previous period, the authorities opted for the more risk-averse strategy of reserve accumulation.

Cases Where Spending Exceeded Absorption

Given that the fiscal authorities in these countries (Mozambique, Tanzania, and Uganda) spent the aid, but central banks were unwilling to allow full aid absorption, the question arises as to how they handled the increase in money supply associated with this spend-but-don’t-absorb response.

Each of the three countries followed a combination of options: either allowing the money supply to increase or sterilizing monetary expansion through treasury bills. In all three countries, concerns about the negative impact of a real exchange rate appreciation on competitiveness led the central banks to contain net foreign exchange sales to a level consistent with a depreciating nominal exchange rate.38 Because the governments in these countries simultaneously increased domestic expenditures, this injection of liquidity led to inflationary pressures and various episodes of attempted sterilization through treasury bill sales.39

In Tanzania and Uganda, the authorities were largely successful in keeping inflation in check, with underlying consumer price index (CPI) inflation never exceeding 10 percent during the aid-surge period. However, in both cases this was achieved at the expense of squeezing private sector investment through the sale of government paper during some periods. In Tanzania, the treasury bill rate rose from 2.6 percent in September 2002 to 7.6 percent by end-2003, while in Uganda two episodes of treasury bill sterilization pushed rates to over 20 percent in early 2001 and end-2003.40

Box 2.4.Aid Volatility and Programs Supported by the Poverty Reduction and Growth Facility

Aid inflows remain generally volatile and unpredictable, and volatility has potentially serious adverse consequences for the recipients of those aid inflows.1 For example, Celasun and Walliser (2005) find that periods of excess aid and tax revenue are not used to accelerate domestically-financed investment spending to compensate for previous shortfalls. This implies that the lack of predictability of aid may have permanent costs in terms of lost output. In addition, volatility may complicate systemic liquidity management by injecting large and unpredictable amounts of money into a thinly monetized economy.

In the sample considered here, aid was hard to predict even one year ahead (Figure 2.4). Part of the aid volatility is a welcome response to exogenous shocks—for example, aid inflows to Mozambique increased sharply in response to floods in early 2000, and in Ethiopia in response to drought in 2002. Because low-income countries are disproportionately prone to exogenous shocks such as terms-of-trade declines or natural disasters, aid inflows should ideally be able to cushion at least part of the adverse impact of these shocks (IMF, 2003a; Guillaumont and Chauvet, 2001). Volatility may also reflect aid conditionality and thus is, to some extent, endogenous to the recipient government’s actions. It may also reflect the donors’ budget cycles.

While prediction errors were common, the International Monetary Fund was not systematically overcautious in projecting net aid inflows for the budgets. In none of the five sample countries considered here was there a consistent pattern of under- (or over-) prediction, except in Mozambique.

In most cases, programs supported by the Poverty Reduction and Growth Facility (PRGF) allowed aid recipient countries adequate fiscal space to spend anticipated aid inflows. The PRGF-supported programs generally dealt with aid surprises more cautiously than with expected changes in aid. As shown in the table, in the case of positive aid surprises, PRGF-supported programs limited the spending of aid in excess of projections by reducing the ceiling on net domestic financing of the government in three of the five countries. Such an adjustment in the ceiling sets an implicit limit on the fiscal deficit.2 When actual program aid exceeded the level projected under the program, net domestic financing of government budget was reduced by the full amount of the excess aid.

In the case of negative aid surprises, the PRGF-supported programs implied a limited degree of fiscal adjustment by allowing some increase in net domestic financing of the government and lowering international reserve targets. At one extreme, the PRGF-supported program in Tanzania allowed increases in the net domestic financing target by 100 percent of any aid shortfall; on the other extreme, the program in Mozambique did not allow for any adjustment in net domestic financing. In other cases, the programs allowed for only partial upward adjustment in net domestic financing in response to a negative aid shock.

During the event, only in Ghana were these adjusters binding. The downward adjustment in the ceiling on domestic financing of the government in Ghana limited spending of incremental aid in 2001. In view of the subsequent unexpected collapse in aid, this turned out to be welcome. In other cases there were no significant positive surprises during the period of study or there was no adjuster (Mozambique), allowing surprises to be spent (and requiring a contraction in the case of shortfalls).

Adjustments in the Net Domestic Financing Target under IMF-Supported Programs in Response to Aid Shocks
Excess Aid (actual aid is higher than programmed)Shortfall in Aid (actual aid is lower than programmed)
Ethiopia1Reduced net domestic financing by 100 percent of excess aidRaised net domestic financing by 50 percent of the shortfall in aid
Ghana2Reduced net domestic financing by 100 percent of excess aidRaised net domestic financing by some proportion of the shortfall in aid
Mozambique3No adjuster: no effect on netdomestic financingNo explicit adjuster: no change in net domestic financing of the budget deficit
Tanzania (2001–04)4No effect on domestic financing of the government budgetRaised net domestic financing by 100 percent of the shortfall in aid; in 2001 adjustment was capped at $60 million
Tanzania (1996–2000)5,6Reduced net domestic financing by 100 percent of excess aidRaised net domestic financing by 60 percent of the shortfall in aid
Uganda7Reduced net domestic financing by 100 percent of excess aidRaised net domestic financing by 100 percent of the shortfall in aid

A shortfall (or excess) in foreign program assistance, defined as the cumulative sum of nonproject external funding, excluding assistance under the Heavily Indebted Poor Countries (HIPC) Initiative, from the programmed levels triggers adjustments.

The adjuster allowing for a higher ceiling to account for unexpected aid shortfalls changed from 50 percent of the shortfall to a fixed $50 million cap in June 2001. This cap was raised to $75 million in March 2002.

Domestic primary deficit was the target in Mozambique, which implicitly implied a target for net domestic financing of the government. However, an increase in net domestic assets in response to an unexpected shortfall in program aid implicitly allowed additional domestic financing.

A shortfall in foreign program assistance, which is defined as the cumulative sum of program grants and loans, from the programmed levels, triggers adjustments. For 2001, the trigger was net foreign financing, which was defined as the cumulative sum of program grants and loans minus external debt service paid.

A shortfall (or excess) in net foreign financing, defined as the cumulative sum of program grants and loans minus external debt service paid, from the programmed levels triggers adjustments.

In 1996–97, performance criteria on net credit from the Bank of Tanzania to the government (ceiling) were also in place, adjusted upward by 60 percent (downward by 100 percent) for any shortfall (excess) in net foreign financing.

Net credit to the government (ceiling) adjusted up (down) for any shortfall (excess) in import support (basically, program aid), including debt relief.

A shortfall (or excess) in foreign program assistance, defined as the cumulative sum of nonproject external funding, excluding assistance under the Heavily Indebted Poor Countries (HIPC) Initiative, from the programmed levels triggers adjustments.

The adjuster allowing for a higher ceiling to account for unexpected aid shortfalls changed from 50 percent of the shortfall to a fixed $50 million cap in June 2001. This cap was raised to $75 million in March 2002.

Domestic primary deficit was the target in Mozambique, which implicitly implied a target for net domestic financing of the government. However, an increase in net domestic assets in response to an unexpected shortfall in program aid implicitly allowed additional domestic financing.

A shortfall in foreign program assistance, which is defined as the cumulative sum of program grants and loans, from the programmed levels, triggers adjustments. For 2001, the trigger was net foreign financing, which was defined as the cumulative sum of program grants and loans minus external debt service paid.

A shortfall (or excess) in net foreign financing, defined as the cumulative sum of program grants and loans minus external debt service paid, from the programmed levels triggers adjustments.

In 1996–97, performance criteria on net credit from the Bank of Tanzania to the government (ceiling) were also in place, adjusted upward by 60 percent (downward by 100 percent) for any shortfall (excess) in net foreign financing.

Net credit to the government (ceiling) adjusted up (down) for any shortfall (excess) in import support (basically, program aid), including debt relief.

1 Bulíř and Hamann (2005) find that aid volatility—measured as the ratio of the variance of aid to the variance of revenue—increased over 2000–03 compared with the level over 1995–98 for all five countries in the sample. Simulations from a calibrated real business cycle model in Arellano and others (2005) suggest that aid volatility may cost poor countries as much as 3 to 4 percent of GDP.2 If the aim is to fix spending regardless of aid, then net domestic financing would be adjusted downward by the full amount of unprogrammed aid. If the aim is to allow spending to vary fully with aid, there would be no adjuster on net domestic financing of the government.

There were considerable year-to-year policy variations in these countries. For example, in Tanzania, an early response focused on sterilizing through treasury bill sales was largely abandoned subsequently in favor of allowing the money supply to increase. This latter policy created excess liquidity in the banking system and eventually led to an increase in inflation. Thereafter, the authorities turned toward selling foreign exchange to sterilize the monetary injection associated with aid-related spending, resulting in a sharp rise in aid absorption and some real exchange rate appreciation. In effect, the authorities moved toward a delayed spend-and-absorb strategy.41

In Mozambique, despite more aid absorption than the other countries, the large fiscal expansion was also accompanied by more monetary loosening than in the other countries. Reserve money growth shot up from about 7 percent per annum before the aid surge to 53 percent in 2001. Inflation followed suit, peaking at well over 20 percent in early 2002. From 2002 onward, however, the authorities undertook more sterilization through foreign exchange and treasury bill sales, bringing down reserve money growth. In addition, rapid GDP growth led to an increase in the demand for money. Consequently, inflation was brought under 10 percent by 2003.

Better adherence to the IMF-supported program paths (lower reserves accumulation and greater sales of foreign exchange by the central bank) would have reduced the need for sterilization through treasury bills, helping to avoid crowding out private sector investment in Tanzania and Uganda, and reducing inflation in Mozambique (Figure 2.6). This would have been a more suitable response to the surge in aid inflows because, unlike in Ethiopia and Ghana, the level of import coverage afforded by gross reserves was quite high in all three countries.

Figure 2.6.Mozambique, Tanzania and Uganda: Spending Exceeds Absorption

(NDA in millions of local currency, NIR in millions of U.S. dollars, and RM growth in percent)

Source: IMF staff reports.

Notes: NDA = net domestic assets; NIR = net foreign reserves; RM = reserve money. Target data are occasionally unavailable. Wherever possible, adjusted program targets are used. In Uganda, the 2000 target for NDA is for the banking system.

Conclusions and Policy Implications

Summary of Findings

Were there significant macroeconomic constraints on aid absorption? Yes, in the sense that in most countries absorption fell significantly short of the increment in aid inflows. Although there is considerable variation from year to year, on a cumulative basis no country entirely absorbed the increased level of aid.42 Absorption ranged from two-thirds for Mozambique to zero for Ghana and Tanzania. Despite the conjecture that to absorb and spend is generally the best use of aid, no country in the sample systematically pursued this strategy.

Was Dutch disease a problem? There is no evidence in the sample of significant real exchange rate appreciation as a result of a surge in aid. This is consistent with the pattern of aid absorption noted above; if aid is accumulated in reserves, then there is no need for a real exchange rate appreciation to mediate a fall in net exports and thereby absorb the aid. Hence the conclusion that Dutch disease was not a problem ex post for any of the countries studied. Of course, part of the reason that real appreciation (and consequently, Dutch disease) was not observed is precisely because authorities were concerned with competitiveness and restricted absorption accordingly.

Why did some countries save the increase in aid, neither spending nor absorbing? Two of the countries studied—Ethiopia and Ghana—neither absorbed nor spent a significant part of incremental aid. A number of related but distinct factors seemed to underlie this response. First, both countries went into the aid-surge period with a precariously low level of international reserves and a need to establish macroeconomic stability. Hence, building sufficient import coverage was one motive for accumulating the aid. Second, aid volatility probably played a role in determining absorption behavior in some years. For example, in 2003, when aid surged again in Ghana after collapsing the previous year, part of the motivation for the don’t-absorb-don’t-spend strategy was to protect against excessive fiscal tightening in the event of a future reduction in aid inflows. Third, nonabsorption may have been governed by a desire to avoid appreciation and preserve international competitiveness and, when coupled with a policy of not increasing government expenditures, to curb inflation (in Ghana’s case) or to keep inflation in check (in Ethiopia’s case).

Why did some countries fail to use aid to increase net imports at the same time as they increased fiscal spending? That is, why did they spend but not absorb? In Mozambique, Tanzania, and Uganda, and in Ethiopia over 2002–03, government spending out of incremental aid exceeded the amount of aid absorbed. The governments of these countries increased domestic expenditures, but there was a much smaller increase in net imports. This is potentially the most problematic response to aid surges. It can create high inflation, as it did in Mozambique, or require substantial treasury bill sterilization, and hence high interest rates and increases in domestic debt, to keep inflation in check.

What were the main reasons for this policy response? In Ethiopia over 2002–03, absorption was only partial even though the aid increase was largely spent, and the monetary authorities took no obvious actions to reduce absorption, such as sterilizing the associated monetary expansion. However, the monetary authorities’ sales of foreign exchange of the aid inflows were dictated by the exchange rate peg to the U.S. dollar. It would have taken further action by the authorities to absorb more of the aid.

In the other countries, the dominant factor behind this response appears to have been a desire to preserve international competitiveness, manifested in an unwillingness to see the nominal or real exchange rate appreciate. Thus, the central banks of each country accumulated international reserves throughout the aid-surge period, despite a relatively comfortable level of import coverage.

This still begs the question of why, in the face of the desire to avoid appreciation, the aid was spent at all? One explanation may be that political pressures make it difficult to resist spending aid money. For example, donors may object to having the aid they provide simply accumulated in government deposits at the central bank. Indeed, loans for projects might require a certain amount of domestic expenditures every year, at the risk of being stopped altogether.

These cases illustrate the risk that policies on spending aid inflows may be inconsistent with policies on exchange rate and monetary management of these same inflows. This may be partly because the link between these two sets of issues may not be fully understood by all the relevant policymakers. It may also be because institutional responsibilities for these two sets of issues are separated. Fiscal authorities and donors will find it entirely appropriate that aid inflows to the budget be spent. Central bank officials, on the other hand, may be more concerned about implications for the real exchange rate and the export sector. A spend-but-don’t-absorb response may be an unfortunate outcome that neither party fully desires.43

Were aid inflows inflationary? Whether the aid surges were inflationary or stabilizing depended on the macroeconomic policy reaction:

  • In Mozambique, Tanzania, and Uganda, aid inflows did create inflationary pressures, because the excess of spending over absorption implied an injection of domestic liquidity. In Mozambique, the gap was particularly large, and hence the country experienced high inflation. In Tanzania and Uganda, the inflationary pressures led to various episodes of sterilization through treasury bill sales. Although inflation in these countries was consequently contained below 10 percent, this came at the cost of rising interest rates, and in Uganda’s case, a rapidly rising level of domestic debt.
  • In Ethiopia and Ghana, the don’t-absorb-don’t-spend response implied that, over the aid-surge period as a whole, incremental aid did not contribute to inflationary pressures.
  • With a different policy response, aid can help reduce inflation. The strategy of absorbing but not spending was executed successfully in some years, notably in Ethiopia and Ghana in 2001. In each case it was part of a successful macroeconomic stabilization program. It was planned but not executed more generally in Ghana during 2001–03, in part perhaps because of concerns about exchange rate appreciation but also as a reaction to the volatility of aid.44

PRGF-Supported Program Design Issues

Strategies for Reacting to Aid Surges

In general, monetary and net international reserve (NIR) targets under the PRGF-supported programs were consistent with a textbook absorb-and-spend response to aid inflows. Where macroeconomic stability had not been established or domestic debt was too high, the programs appropriately envisaged absorbing but not spending.

In practice, these goals typically diverged from the authorities’ concerns about international competitiveness and their desire to avoid nominal appreciation. Hence, in those countries where spending exceeded absorption—Mozambique, Tanzania, and Uganda—NIR remained consistently above the programmed path (because of a reluctance to sell foreign exchange to the extent envisaged by the program), while NDA remained below the programmed path (because of attempted treasury bill sterilization to dampen the consequent inflationary pressures). The countries therefore did not benefit from the full extent of aid, inflationary pressures were created, and treasury bill sterilization led to an increasingly high debt service burden on domestic debt.

The consistency of monetary and exchange rate policy with fiscal policy needs greater attention in cases where the authorities spend but do not absorb the aid. Typically, the IMF recommends exchange rate flexibility and foreign exchange sales when a spend-and-don’t-absorb outcome emerges. However, where countries are unwilling to follow this advice—perhaps in order to guard competitiveness—more care needs to be taken that an appropriate outcome is achieved.

One option would be to limit spending as well, following the pattern of Ghana and, for part of the time, Ethiopia. This would have the merit of avoiding macroeconomic difficulties while saving the aid for later, and would make sense if Dutch disease concerns outweigh the benefits from the absorption of aid inflows. Even here, the best response would be to work to improve the quality of public expenditures and the quality of aid. However, where aid inflows are volatile, international reserves are too low, or good projects cannot be implemented, reserve accumulation may be the most reasonable short-run response.

Another option is to allow the monetary expansion necessary to accommodate increased expenditures, and accept inflation as the mechanism through which real appreciation occurs. This policy may work well in the short run, especially if a rapid supply response tempers inflation. In the medium term, the incremental domestic expenditures should increase the demand for imports, and the saved foreign exchange should then allow the authorities to accommodate this increased import demand without running low on international reserves. In essence, this is a spend-and-eventually-absorb scenario, of the kind followed by Tanzania after 2003. This requires some tolerance for a period of higher inflation to achieve the required real appreciation and, critically, a willingness on the part of the central bank to defend the nominal exchange rate against depreciation. In other words, the central bank must tolerate the real appreciation and eventually sell foreign exchange as the current account deficit increases.45

Most of the countries in the sample—Mozambique, Uganda, and initially, Tanzania—attempted to combat inflationary pressures through treasury bill sterilization. This approach is likely to be suboptimal under many circumstances. Again, aid is not absorbed, while domestic public expenditures in conjunction with higher interest rates potentially crowd out private investment. In addition, where monetary expansion has not yet led to higher inflation and the nominal exchange rate remains stable, programs must exercise care before prescribing treasury bill sterilization that could prevent delayed absorption.

However, sterilization would generally be preferable to inflation as a source of domestic financing (IMF, 2005b). Thus, in a country with a given spend-and-don’t-absorb policy, perhaps politically dictated, where inflation is high or rising, there would be a case for sterilization through treasury bills.

Tanzania’s experience illustrates how the different policy options work. The aid increases were largely spent. Initially, the monetary authorities attempted to control money growth through sterilization, raising interest rates, and depressing private investment. A subsequent relaxation of monetary policy led to rapid money growth but still no inflation, real exchange rate appreciation, or absorption. Finally, more aggressive sterilization through foreign exchange sales led to an increase in absorption and some real exchange rate appreciation.

Dealing with Aid Volatility

Aid volatility raises a special set of issues. In general, whenever a significant increase in aid was anticipated, program ceilings on the government’s primary balance were raised, thereby allowing greater government expenditure. Similarly, reserve targets were also consistent with absorbing expected increases in aid flows.

When there was a deviation from the anticipated aid path, however, programs were more cautious about spending the “excess” aid in most cases. This is because the programs often included adjusters that would fully limit the spending (and absorption) of surprise aid.46 A symmetric response to aid shortfalls would imply allowing a full increase in domestically-financed spending when aid shortfalls emerged. However, in most cases the PRGF-supported programs limited additional domestic financing of the budget in cases of shortfalls.47 In practice, however, these adjusters were not a constraint because the budgetary aid outcomes were broadly in line with projections (except in Ghana in 2001).

Further consideration should be given in individual cases to whether such an asymmetric response to aid shocks is appropriate. In general, aid volatility should presumably be smoothed by saving some of the aid in high-aid years (both fiscally and in terms of reserve accumulation that is neither spending nor absorbing) and by dissaving (that is, running down reserves and increasing the deficit after aid) when aid shortfalls emerge. Against this logic must be balanced the temptation to treat positive shocks as permanent and negative ones as temporary, in which case such a policy would lead to a rundown of reserves. Clearly, the level of reserves may itself be an important factor in deciding whether to have symmetric or asymmetric adjusters—countries with lower reserves have less scope to smooth negative aid shocks.

Critically, effective smoothing requires coordinated fiscal and monetary policy responses. Taking the opportunity of a temporary aid surge to build reserves may make sense, but generally only if accompanied by increased budgetary savings as well. Otherwise, the aid surge leads to an increase in the budget deficit that is effectively domestically financed.

Final Considerations

The key long-run strategic choice is whether to use the aid—by absorbing and spending—or not, in which case the aid should be neither absorbed nor spent. The latter choice is equivalent to refusing aid altogether. Other responses should be at most short-run. To absorb and not spend can be very helpful in stabilizing the economy, because it permits a reduction in domestic debt and/or inflation without the fiscal contraction that this would normally require. It cannot be a feasible longrun strategy once stability has been achieved, however, because domestic debt stocks cannot fall indefinitely, and also because donors typically would want their aid spent. To spend and not absorb also is not a desirable long-run strategy because it implies that aid is serving not to finance additional net imports but to trigger additional domestically-financed spending.

From this perspective, the typical discussion of the monetary management of aid inflows, which centers on the question of how to manage the consequences of aid-related spending, risks missing the point. The common logic proceeds in two steps: (1) the government necessarily spends the aid receipts; and (2) the central bank chooses the appropriate instrument for sterilization—that is, whether to sell foreign exchange causing some real appreciation, sell government bonds inducing an interest rate hike, or employing a combination of both. This line of thought leads to the view that central banks need to balance the risks of excessive exchange rate appreciation against excessive interest rate increases and choose some reasonable middle ground.

This perspective fails to treat the analysis of monetary and fiscal policy jointly. The first consideration should be whether extra public spending at some potential cost to export competitiveness is desirable. If not, the question should immediately arise as to whether the aid should be spent in the first place. After all, additional domestically-financed spending can take place in the absence of aid. Thus, the question becomes whether to spend and absorb or neither spend nor absorb.

In considering this question, it may be helpful to review the nature of macroeconomic absorptive capacity constraints. Frequently, macroeconomic concepts such as inflation and the real exchange rate misalignment are posed as constraints that limit the ability of a country to spend aid on building schools or roads or fighting HIV/AIDS.

However, the real issue is not a trade-off between allowing more inflation and hiring more nurses. The real question is how the productive resources in a country—its workers, natural resources, and physical capital—are to be deployed. Aid raises this question because it allows more domestic resources to be devoted to building physical and human capital at home or to satisfying consumption needs, because fewer domestic resources are needed for producing exports or import substitutes. To the extent that there are domestic projects carrying a high rate of return, or that have an ameliorating impact on poverty, this reallocation of resources is beneficial. The real exchange rate appreciation that may result is not, in itself, a macroeconomic cost of using aid, but rather part of the mechanism whereby aid is useful, because it may be required to draw resources out of the traded goods sector.

When aid comes in the form of grants, the opportunity cost for the recipient of absorbing and spending aid is the foregone use of domestic resources, particularly in the traded goods sector. The traded goods sector—especially nontraditional exports—may play a special role in generating productivity growth. If so, the aid may lower growth by slowing the growth in that sector. It may also have adverse consequences for poverty by squeezing margins in traditional exports. A related but distinct potential cost has to do with the unpredictability of aid. For example, if aid inflows are high now but fall in the future, then in the presence of transaction costs and imperfect capital markets it may be difficult to resuscitate export firms crowded out by the aid.48

The critical strategic question for aid recipients is how to balance these costs and benefits. The question of when the aid-financed investments cease to become sufficiently productive is closely related to the concept of absorptive capacity.49 The rate of return to investments will likely decline as the rate of investment rises—the best projects should be the first undertaken. Moreover, resources will be drawn out of other uses, and this will likely become progressively costlier—the least-well-used resources should be the first to be drawn away. The implication is that there may be a level of investment beyond which the rate of return will be lower than that achieved in alternative uses. Absorptive capacity has been reached when aid-financed investments do not yield enough to justify the resources used to produce them.

If the judgment is made in a particular situation that the costs of absorbing and spending outweigh the benefits, what is to be done?

One possible solution is to recognize that absorptive capacity cannot be taken as given. It is critical to make expenditures more effective, including by improving project choice and expenditure management, and more broadly the overall policy environment (Bevan, 2005). Attention should be focused on how, and how fast, to scale up aid so as to minimize competitiveness problems, such as by focusing on ways to use aid to increase productivity and critical imports.

For example, a carefully designed and scaled-up investment program may raise the rate of return while minimizing the cost for the traded goods sector. An investment program aimed at improving productivity in the medium term may result in a traded goods sector even larger than it would have been without the aid, as the productivity gains resulting from better roads, education, or health may outweigh the effects of the real exchange rate appreciation (Bourguignon and others, 2005).

Similarly, aid spent directly on noncompetitive imports may create fewer tensions with an export-led growth strategy. For example, using aid to import factors of production used in the export sector (e.g., chemical fertilizer) would not tend to create pressure on the real exchange rate.

However, raising absorptive capacity is easier said than done, and the stakes are high: if the aid-related spending turns out not to be effective, not only is the aid wasted, but scarce domestic resources are misallocated and the traded goods sector is shrunk.

Another possibility is to save the aid until it can be effectively used. This would involve accumulating reserves and avoiding an increase in the budget deficit net of aid. This solution has two major problems. First, it is politically hard to limit government spending of the local currency counterpart to the aid inflows—after all, the aid has been given and the needs are great. This would lead to the spend-and-don’t-absorb policy denigrated previously. Second, a country that saves aid inflows in this fashion would alienate donors, who might understandably decide to reallocate the aid to a more eager recipient. Saving a large part of an aid surge is particularly appropriate when there is a concern that the aid boom will prove temporary; however, “treating the aid flow as temporary may well make it so” (Adam, 2001, p. 11).

Donors can help resolve some of these tensions by improving the quality of aid. Aid that is less tied and for which the administration uses up fewer scarce management resources in the recipient country would be more useful. Equally important, aid that is more predictable, and in particular that can be relied on over the medium term, can be spent and absorbed more effectively without the otherwise valid concern that its disappearance will leave the recipient with unsustainable expenditures and an overvalued real exchange rate. Aid that buffers temporary negative shocks may also be more readily used, insofar as stabilization of the exchange rate and aggregate demand in the face of a temporary contraction would not tend to raise competitiveness concerns.

The IMF can provide only supportive guidance regarding the strategic question of how to best absorb and spend aid in the long run. However, the long run famously never arrives. In the meantime, the authorities in aid-recipient countries must balance a complex set of objectives involving fiscal policy and exchange rate and reserve management. But one message is simple: a given aid dollar can be used to build reserves or to increase the fiscal deficit, but not both. The cases reviewed in this paper suggest that trying to do so may make aid less effective.

Appendix 2.1. Methodology for Sample Selection

For each country in the sample, this paper examines the pattern of net aid inflows, which are defined as gross inflows (including debt relief) less amortization, interest payments, and arrears clearance.50 In particular, the paper identifies a year/period in which net aid inflows increased substantially, and a detailed analysis is conducted of the policy responses to increased inflows and macroeconomic outcomes for that year/period and succeeding years. While gross aid flows may be a better aggregate for analyzing donors’ leverage in policy dialogue, net aid inflows are a more appropriate measure for assessing the macroeconomic effects of a scaling up of aid. The focus in this paper is on countries characterized by (1) a relatively sound policy record, (2) a large rise in net aid inflows, and (3) net aid inflows that comprise a significant percentage of GDP.

First, the avoidance of the worst performers in terms of institutions and economic policies was driven by a desire to draw lessons of relevance for situations in which, broadly speaking, policymaking is not denominated by macroeconomic disarray, misgovernance, or post-conflict reconstruction. The focus is on how to help those countries that are best positioned, institutionally and in terms of the policy framework, to absorb large quantities of aid. The criteria used were the World Bank’s indicators of quality of economic institutions and policies (the CPIA). Only those countries with a CPIA ranking in the third quartile and above were considered.

Second, a country must have experienced a significant rise in net aid inflows as a percentage of GDP. The quantitative screen was for a rise in the net aid/GDP proportion by at least 1 percentage point compared with a period identified by the data as the pre-surge period.

The period at which aid may have begun to rise varies across countries. To divide the 1996–2003 period into two subperiods of pre-surge and ongoing-surge, the “breakpoint” was found that minimizes the sum of squared deviations between annual aid levels and within-period average aid levels. In some cases, aid flows rose linearly, so there was no obvious breakpoint. In these cases, the breakpoint was subjectively and uniformly identified as 2000.

Third, to focus on cases where aid is macroeconomically central, net aid inflows should be large relative to the economy receiving them. The screen applied to capture this feature was that only countries for which net aid was at least 10 percent of GDP over the post/ongoing-surge period were considered.

Excluding small island economies and countries that were emerging from either a major civil conflict or that were not PRGF-eligible, this left seven countries, all of them African (see Table 2.A1; Table 2.A2 presents aid and CPIA data for PRGF-eligible countries). Of these, Mauritania was excluded because of a history of significant misreporting. Although Malawi emerged from the quantitative screens, its aid “surge” was in 1998 and reflected a temporary dip in aid flows from the previous year. In contrast to the other countries in the sample, there was no significant and sustained increase in aid, relative to previous periods. As a result, Ethiopia, Ghana, Mozambique, Tanzania, and Uganda were selected.

Table 2.A1.Aid Flows: Changes and Levels(As a percent of GDP)
Aid-Surge Breakpoint Year1Net Aid: Change Over the Surge Period2Net Aid: Level Following Initial Surge2
Ethiopia200111.422.1
Ghana20003.011.5
Malawi19987.824.7
Mauritania20016.628.2
Mozambique20026.433.0
Tanzania20021.413.7
Uganda20002.513.5

Where this year came after 2001, 2000 was selected as the breakpoint for the calculations.

Using net aid flow data from the Development Assistance Committee of the Organization for Economic Cooperation and Development.

Where this year came after 2001, 2000 was selected as the breakpoint for the calculations.

Using net aid flow data from the Development Assistance Committee of the Organization for Economic Cooperation and Development.

Table 2.A2.Net Official Development Assistance (ODA)(In percent of GDP)
CPIA Ranking Ranking Quintiles1ODA Average 1996–99ODA Average 2000–03
AlbaniaThird9.86.7
ArmeniaFirst12.510.5
AzerbaijanSecond3.54.1
BangladeshSecond2.72.3
BeninSecond10.29.6
BoliviaSecond8.68.8
Burkina FasoSecond14.413.1
BurundiFifth9.122.8
CambodiaFourth10.311.6
CameroonThird5.15.8
Central African RepublicFifth11.96.2
ChadFourth13.810.3
Congo, Dem. Rep. ofFourth2.631.7
Congo, Republic ofFourth9.51.9
Cote d’IvoireFourth5.74.0
EritreaFourth21.036.2
EthiopiaThird10.718.0
GambiaFourth9.014.1
GeorgiaThird7.87.4
GhanaSecond8.511.6
GrenadaFirst2.93.0
GuineaFourth8.77.1
Guinea-BissauFifth45.839.4
GuyanaThird20.012.3
HaitiFifth9.95.5
HondurasFirst9.17.6
KenyaThird4.33.9
Kyrgyz RepublicThird15.912.8
LaosFifth19.615.1
LesothoThird7.67.3
MadagascarSecond13.78.7
MalawiThird20.924.7
MaliSecond15.813.6
MauritaniaFirst21.326.5
MoldovaThird4.38.1
MongoliaThird23.420.5
MozambiqueThird26.633.0
NepalSecond8.07.2
NicaraguaFirst32.317.7
NigerFourth13.513.8
PakistanSecond1.32.4
Papua New GuineaFifth7.67.1
RwandaSecond20.819.1
SenegalFirst11.18.6
Sierra LeoneFourth15.139.3
Sri LankaFirst2.52.3
SudanFifth2.02.2
TajikistanFourth10.613.5
TanzaniaFirst11.614.6
TogoFifth8.33.9
TongaFourth16.214.7
UgandaFirst11.013.6
VietnamFirst4.24.5
YemenSecond5.43.9
ZambiaThird16.216.1
ZimbabweFifth4.32.2
Sources: Organization for Economic Cooperation and Development; International Monetary Fund, World Economic Outlook; and World Bank and IMF (2005).

The Country Policy and Institutional Assessment (CPIA) is a measure used by the World Bank. The first quintile is the best ranking.

Sources: Organization for Economic Cooperation and Development; International Monetary Fund, World Economic Outlook; and World Bank and IMF (2005).

The Country Policy and Institutional Assessment (CPIA) is a measure used by the World Bank. The first quintile is the best ranking.

Appendix 2.2. Dutch Disease: Theory and Evidence

Dutch disease is related to the idea that productivity growth is particularly high when resources are devoted to exports, particularly of nontraditional products, because of learning-by-doing or other dynamic externalities in these relatively competitive and technologically-advanced industries. The decline of the export sector, mediated by an increase in the demand for and price of nontradables, may lower the attainable growth path of the economy. For this argument to hold, dynamic externalities in the export sector would have to outweigh the benefits of capital accumulation associated with aid-financed investment (as well as any related productivity growth). A slightly different argument is premised on imperfect capital markets and hysteresis: if aid is temporarily high and crowds out export firms through real appreciation, it may not be possible to resuscitate these firms once aid falls and the real exchange rate depreciates.

The theoretical case for Dutch disease is ambiguous. For example, when learning-by-doing externalities can also take place in the nontradable sector, the long-run adverse impact will be limited, even if the real exchange rate appreciates in the short term (Torvik, 2001). In the longer run, the investments in physical and human capital, both in the government and in the private sector, begin to bear fruit and productivity increases not only in the tradable sector but also in the nontradable sector, potentially offsetting the initial loss of competitiveness.51

The effects of Dutch disease can be enhanced if the aid-recipient economy has weak financial markets. For example, in thin foreign exchange markets, volatile and lumpy aid disbursements can cause overshooting in the exchange rate or interest rate. Similarly, in the short term, when the real exchange appreciation due to excess demand for nontradables is not yet compensated for by the increase in productivity, firms may be forced out of business if they do not have access to adequate credit to smooth out the shock. Temporary overshooting of the actual real exchange rate after an increase in aid may therefore be more damaging than the longer-term shift in the equilibrium real exchange rate.

Despite a substantial body of theoretical literature on the implications of Dutch disease from aid inflows, empirical work is limited—particularly in low-income countries. Recent cross-country studies find some evidence for the real appreciation effect. For example, Elbadawi (1999) finds that a 10 percent increase in the aid-to-GDP ratio appreciates the real exchange rate by about 1 percent. Individual country studies, however, offer mixed results. Some (e.g., Malawi and Sri Lanka) find that aid inflows cause real appreciation, but others (e.g., Ghana, Nigeria, and Tanzania) find that aid flows are related to real depreciations.52

In a related literature, some papers find evidence of a significant detrimental impact of real appreciation on exports, particularly nontraditional exports (Sekkat and Varoudakis, 2000; Elbadawi, 2002). Empirical evidence also suggests that real appreciation contributed to the widening trade deficits in four African economies (Adenauer and Vagassky, 1998).

A recent approach is to look directly at the impact of aid on exports without attempting to trace through the real exchange rate channel. Rajan and Subramanian (2005a) examine the effects of aid in a sample of 33 countries in the 1980s and 15 countries in the 1990s. They find that export and labor-intensive manufacturing industries grew significantly slower in those countries that received the most aid, and that a 1 percentage point increase in the ratio of aid to GDP is roughly equivalent to a 4 percentage point overvaluation of the exchange rate. Arellano and others (2005) find that aid significantly depresses the export sector in a sample of developing countries.

The risks of Dutch disease need to be balanced against the potential benefits from the investment that aid can finance. Here, the evidence is also mixed. The benefits of public investment are not clearly established empirically, as a general matter (Leite and Tsangarides, forthcoming). Of course, the rate of return will depend on the particular investment and a variety of country-specific circumstances. A strong case can nonetheless be made for a higher level in poor countries (United Nations Millennium Project Report, 2005). While the systematic evidence for a positive growth impact of private investment is stronger, it is less clear that aid can be effectively channeled into higher private investment.

More broadly, a huge literature asks directly whether aid affects growth, with somewhat mixed conclusions.53 However, there is substantial cross-country evidence that exchange rate overvaluations are one of the few policy variables that matter for growth after controlling for institutions. There is also substantial micro-based evidence on the benefits of trade and, to some extent, learning-by-doing associated with exports. Weak exchange rates may also help predict the incidence of episodes of growth acceleration.54 Case studies that examine the entire chain from aid through export performance to final outcomes include Nkusu (2004b), who finds little sign of Dutch disease in Uganda.

On balance, the evidence on Dutch disease is mixed. Presumably, the seriousness of the problem and the benefits of aid-financed investments depend on the particular circumstances of each country. A country with strong dynamic externalities in the tradable goods sector may want to carefully consider the level of aid it can absorb without triggering too much real appreciation. It may also wish to seek aid in forms that are less likely to induce real appreciation.55 It can safely be concluded, however, that the risk of Dutch disease raises the stakes: if aid-financed investments have poor rates of return, not only is the aid wasted, but there is a risk that overall growth may be impaired.

1Appendix 2.1 discusses sample selection in more detail.
2This usage of absorption should not be confused with the related concept of “absorptive capacity,” which, in addition, involves questions about the rate of return on investments financed by aid.
3The non-aid current account balance is the current account balance excluding official grants and interest on external public debt, while the non-aid capital account balance is the capital account net of aid-related capital flows, such as loan disbursements and amortization.
4With this definition, aid that finances capital outflows is not absorbed. This makes sense insofar as aid that flows back out of the country does not transfer real resources to the country.
5Aid that is directly used to finance imports by the government (e.g., a grant in kind, a grant of foreign exchange that the government immediately uses to purchase imports, or aid that goes directly to nongovernmental organizations to finance imports) effectively bypasses the central bank and would lead directly to absorption. Technical assistance is also directly absorbed.
6The deficit net of aid is equal to total expenditures (G) less domestic revenue (T), and is financed by a combination of net aid (including both external loans and grants) and domestic financing: GT = Fiscal deficit net of aid = Net aid + Domestic financing.
7For example, if the government allocates a new grant to financing a domestic project that was earlier financed from different sources, this does not constitute an increase in spending, because the non-aid fiscal deficit remains unchanged. Similarly, if the foreign exchange associated with a particular grant is sold by the central bank, but overall net sales of foreign exchange do not increase, this does not constitute an increase in absorption, because no extra foreign exchange is available to finance an increase in net imports.
8The distinction between absorption and spending, in the terminology used in this paper, is one of the central issues associated with the “transfer problem” discussed in Keynes (1929). Keynes was concerned with Germany’s problems in generating current account surpluses to pay reparations after World War I. He argued that for the fiscal authorities to accumulate the local currency counterpart to the required transfers was only part of the transfer problem—the other part being generating the net exports and therefore the required foreign exchange. See Milesi-Ferretti and Lane (2004) for a recent general discussion of the transfer problem and the real exchange rate.
9Strictly speaking, this is true only if the gifted or directly imported good is one for which there was no existing effective demand. If the good transferred was already demanded domestically, then increasing the good’s supply would depress the price of tradables relative to nontradables, leading to real appreciation.
10Prati and Tressel (2005) find that monetary policy can control the timing of absorption. Aid could also go to the private sector directly. Here, too, if the private sector uses the dollars to directly finance imports, there is unlikely to be much macroeconomic impact. Where the private sector sells the dollars to the central bank and uses the local currency proceeds to finance domestic expenditures, similar issues will arise as in the case of government spending.
11The real exchange rate is generally understood in this paper to refer to the relative price of nontraded to traded goods, as a conceptual matter. When it comes to measurement, however, data availability usually dictates the choice of real exchange rate concept. The case studies here follow the common practice of measuring the real exchange rate as a function of the nominal exchange rate and changes in consumer price indices. It turns out for the cases under consideration that this is unlikely to make a major difference, but further work on the correct measurement of the real exchange rate would appear justified.
12One category of nontradable goods that might be important in this process is skilled labor; if aid raises the wages of skilled professionals, this could translate into real appreciation.
13Many of these channels are explored in Chapter VIII.
14There may be second-order effects, e.g., expectations may change as a result of the central bank’s higher international reserve position.
15This is the case emphasized by Buffie and others (2004).
16The IMF Independent Evaluation Office (2004) argues that program assumptions of the Poverty Reduction and Growth Facility that crowding-in will ensue from an increase in availability of credit to the private sector are often left unexamined and also often do not turn out to be correct.
17Related to this point is an accounting issue: “domestic financing” as usually defined in the budgetary accounts is misleading as an indicator of aid usage. It may be useful to consider the following example. Suppose aid is saved entirely in the form of gross international reserves, the government builds up deposits at the central bank, and the fiscal deficit excluding aid remains unchanged. By construction, the fiscal accounts will show a shift in financing from domestic financing (which will fall due to a reduction in net central bank credit to the government) to external financing. But the aid has no macroeconomic effects in this don’t-absorb-and-don’t-spend case—the money supply, fiscal stance, and interest rates are unaffected (except insofar as interest earnings of the central bank are higher). More generally, aid that is not absorbed does not contribute to financing of the government deficit in an economic sense. Thus, it would be misleading to conclude from a perusal of below-the-line financing items in the budget that aid inflows were actually financing the deficit to a greater extent than before.
18It is possible that optimal domestically-financed spending would be greater with higher reserves, perhaps because the higher reserves lower the risk premium the country must pay. This is likely to be a small effect, however. Moreover, borrowing domestically while building reserves would imply losses on the spread between the cost of borrowing and the return on reserves.
19Inflation can dampen private sector confidence and hinder financial deepening, in addition to disproportionately hurting the poor.
20Private investment and government expenditure could have different import intensities, which would modify the details of the argument but not alter the main point. Similarly, the fiscal expansion may increase aggregate output, so it is not the case that there need be a one-for-one trade-off between government spending and private investment. But such an aggregate output expansion could have been engineered without the aid.
21Depending on how open capital markets are, such a strategy can also be self-defeating. That is, sales of treasury bills, by raising the domestic interest rate, may attract international capital flows and appreciate the currency.
22Recent cross-country evidence (e.g., Bulí and Hamann, forthcoming) indicates that aid continues to be volatile, that aid commitments consistently exceed disbursements, and that aid disbursements are generally procyclical—thereby increasing volatility of public expenditures rather than lowering it. Prati and Tressel (2005) construct a theoretical model to consider the optimal pattern of absorption. Implicitly, they compare absorbing and spending to neither absorbing nor spending, in the terminology used here.
23Net aid inflows = Gross aid inflows + Debt relief (including relief under the Heavily Indebted Poor Countries Initiative) − Debt service + Arrears accumulation; with a clearance of arrears taking a negative sign. This paper uses aid data from IMF country reports. This is generally based on recipient country reporting and may differ significantly from donor data, such as those reported by the Organization for Economic Cooperation and Development’s Development Assistance Committee (OECD/DAC), for a number of reasons, including the timing of aid delivery, and donor reporting of technical assistance delivered outside the recipient country as aid.
24For example, in 2001, more than 1,200 donor-funded projects were being implemented in Tanzania. Managing and coordinating such a large number of projects was a challenge for the authorities.
25Devastating floods reduced Mozambique’s growth rate in 2000, a drought reduced Tanzania’s growth rate in 1999, and a severe drought caused a two-year contraction in Ethiopia.
26See Appendix 2.2 for a discussion of the theoretical and empirical literature on Dutch disease.
27These real effective exchange rate (REER) indices are based on nominal exchange rates and consumer price index (CPI) inflation in the target country and its trade partners. Lack of data prevents supplementing these indices with the REER measured by unit labor costs, or the REER measured as the price ratio between nontradables and tradables.
28For example, consider a case in which government expenditures raise wages for a set of workers. This increases their demand for imports. However, when they purchase dollars from the central bank, they do not immediately spend them on imports, but in the first instance, deposit them in dollar accounts held with domestic commercial banks. This would count as a deterioration in the non-aid capital account (due to an increase in commercial banks’ net foreign assets). Subsequently, when they spend the dollars on imports, there would be a corresponding improvement in the non-aid capital account.
29In some countries, large errors and omissions in the balance of payments accounts could be partly responsible for measured fluctuations in the capital account.
30Mozambique, Tanzania, and Uganda reached their decision point under the HIPC Initiative before mid-2000. Improving expenditure management and tracking was part of the fiscal conditionality in all three countries.
31Gupta and others (2003) find evidence that aid lowers revenue effort in a large sample of developing countries, though the effect is modest except in countries with relatively high levels of corruption. McGillivray and Morrissey (2001) reported a significant negative incremental impact of aid on domestic revenue for Pakistan and Côte d’Ivoire.
32Of course, in the absence of a counterfactual, it is difficult to gauge whether revenue effort was harmed by aid. One indicator, however, is provided by PRGF revenue targets. In general these were met by most countries in the sample for most periods. The exceptions were Ethiopia and Mozambique, which missed several quarterly targets on government revenue. However, these episodes did not appear to be the result of moral hazard arising from increased aid inflows. In Ethiopia, the 2004 ex-post assessment argues that the revenue targets were overly ambitious given the pace of structural adjustment in the country. In Mozambique, these episodes were more attributable to a drop in excise and import taxes arising from an increase in world oil prices.
33Of course, project aid was expected to be spent; the program targeted overall expenditures and thus implied a reduction in non-aid-financed spending.
34This broad grouping of sample countries is based on comparing period averages for pre-aid-surge and aid-surge periods. Since the periods under consideration range from two to four years, the averages mask considerable year-to-year policy variations in many of these countries. The countries’ policy responses are discussed further in Chapters III through VII.
35Although both countries avoided substantial real appreciation, this was achieved through different combinations of inflation and nominal exchange rate movements. In Ethiopia, the birr was pegged to the dollar, which necessitated a low rate of inflation to keep the real effective exchange rate on a downward path. Consistent with this objective, reserve money growth remained low during both preaid-surge and aid-surge periods. Ghana, on the other hand, experienced a combination of high inflation and nominal depreciation, the net effect of which was a fairly stable REER.
36During most of the aid-surge period, net domestic assets remained below their targeted path, while reserves remained above the target path.
37It appears that the program may have been too restrictive in its reserve money path. Despite base money exceeding its programmed path during most of the aid-surge period, inflation was contained at a low level. Implicitly, it seems that the increase in money demand accompanying post-war monetization was underestimated. This was inconsequential in practice because the reserve money target was only indicative.
38In Uganda, this concern was magnified by the simultaneous terms-of-trade shock. An appreciating real exchange rate would have further reduced earnings in the coffee sector, with possible adverse consequences for poverty, and may have hindered the expansion of noncoffee exports. In addition, the central bank was concerned that the commercial banks’ appetite for foreign exchange would be limited, putting a limit on sterilization through foreign exchange sales.
39See Chapters III to VII for more details on varying monetary policy responses to the aid surges.
40In the absence of a counterfactual, it is difficult to ascertain by how much the private sector was squeezed. In Tanzania, private sector investment fell from an average of 12.4 percent of GDP in the pre-surge period to 11.6 percent in the aid-surge period. In Uganda, the private investment ratio improved from 11.2 to 13.9 percent, but given the substantial increase in interest rates during the sterilization episodes, it is possible that more improvement may have occurred in the absence of treasury bill sterilization.
41Ethiopia in the latter part of its aid-surge period also saw an increase in absorption.
42Heller and Gupta (2002) discuss a number of challenges for aid-recipient countries at both the micro and macro level that could limit aid absorption.
43This issue raises the possibility that there is a potential cost to central bank independence in the context of aid-dependent, low-income countries, as mentioned in Heller (2005). Chapter VIII discusses this issue further.
44This is broadly consistent with the conclusions of Buffie and others (2004). However, they assume aid is fully absorbed. Their empirical conclusion that aid is on average 80 percent spent in sub-Saharan Africa also would seem to reflect a wide variety of country experiences, some with full spending and some with none. The inflationary impact evidently depends on the interaction of spending and absorption.
45If the authorities fail to defend the nominal exchange rate, inflation will increase without causing real appreciation. This is an inferior response: the aid is not absorbed, while the highly inefficient and regressive inflation tax pays for the spending increase.
46In three of five sample countries, spending of excess aid was limited to zero through adjusters on net domestic financing.
47For example, in Ethiopia and Ghana, additional domestic financing was restricted to 50 percent of the aid shortfall. Since both countries completely restricted the spending of excess aid, the adjusters in each case were asymmetric.
48The same considerations apply when aid arrives in the form of concessional debt, except that the opportunity cost of aid includes the need to repay the debt. As a result, the risks are higher: if aid is poorly used or Dutch disease effects are strong, the debt burden will be hard to bear.
49Adler (1965), Guillaumont (1971), and Berg (1983) are early treatments of this issue.
50Gross inflows include all the resources a low-income country receives in grants (including HIPC debt relief) and loans.
51Nkusu (2004a) discusses the theoretical determinants of Dutch disease and emphasizes the mitigating role of excess domestic capacity. Adams and Bevan (2003) describe a nonmonetary theoretical model and calibrate it for Uganda.
52See White and Wignaraja (1992) for Sri Lanka; Ogun (1995) for Nigeria; Nyoni (1998) for Tanzania; Sackey (2001) for Ghana; Ouattara and Strobl (2003) for the CFA (African Financial Community) countries; and Fanizza (2001) for Malawi.
53See Clemens, Radelet, and Bhavnani (2004) for results showing a positive correlation between aid and growth, as well as Rajan and Subramanian (2005b) and Easterly, Levin, and Roodman (2003) for more skeptical views.
54Acemoglu and others (2003) present important evidence on overvaluation and growth, while Easterly, Levin, and Roodman (2003) summarize the literature. Hausmann, Pritchett, and Rodrik (2004) discuss the role of depreciated real exchange rates in sparking growth accelerations. Berg and Krueger (2003) summarize some of the literature on learning-by-doing and exports.
55This is harder than it seems. It is sometimes argued that aid in kind has no impact on the real exchange rate. This is true, however, only if the transferred good is one for which there was no existing effective demand. If the good transferred was already demanded domestically, then increasing the good’s supply would depress the price of tradables relative to nontradables, leading to real appreciation. On the other hand, the transfer of a good for which there is no preexisting demand is clearly of limited utility in general (although not always—for example, one could imagine aid taking the form of expensive drugs or treatments for which there is no effective preexisting demand).

    Other Resources Citing This Publication