FW: Is Aid Oil?

Is Aid Oil?



An analysis of whether Africa can absorb more aid.

Paul Collier

June 16th, 2005

Centre for the Study of African Economies,

Department of Economics,

Oxford University





1. Introduction

The Report of the Commission for Africa, Our Common Interest, has proposed a

large

increase in aid to the region. The underlying motivation for this proposal

is an acute sense

of Africa's needs. Africa's neediness cannot be contested: on past trends it

will

increasingly become the epicenter of global poverty. The controversial

aspect of the

Commission's proposal is whether additional aid will be effective in

addressing these

evident needs. Criticisms of aid expansion can be grouped into two camps.

The more

radical critique sees aid as having been a cause of Africa's problems, which

expansion

would intensify rather than resolve. This view, which in academic circles

goes back to

Peter Bauer, is widely held in the US, and is also probably the dominant

position in

British public opinion.

1

An apparently softer, but potentially equally debilitating critique

sees aid as useful but subject to diminishing returns. On this view Africa

is already

around the point at which expansion of aid would be ineffective.

While extra aid is merely prospective, the current boom in the prices of

natural resources

is already delivering a massive increase in the resource transfer to Africa.

This does not

of itself make increased aid redundant: the resource boom is highly

selective, with only

around a third of Africa's population living in countries that are

substantial beneficiaries

of higher resource prices. Nevertheless, the two flows invite comparison.

They are the

two largest external resource flows to African governments and so

potentially generate

similar opportunities for investment and growth. The current surge in

resource rents at

least superficially provides a "natural experiment' for the consequences of

a large

expansion in aid.

If this is so, the latest growth data provide a somewhat discouraging

prognosis. During

2004 Africa's oil economies received an unprecedented bonanza in resource

rents. If big

unconditional resource inflows work, we should expect this bonanza to have

decisively

raised the growth rate of the non-oil part of their economies. In fact,

during 2004 the

growth rate of the non-oil part of the African oil exporters' economies was

identical to

the rest of Africa (IMF, 2005, Table SA2). The huge windfall to these

countries did not

confer any growth advantage whatsoever beyond the oil sector. The radical

critics of aid

would be unsurprised by this failure of an oil bonanza to induce growth, and

would

extrapolate this failure to aid. The key hypothesized causal mechanism is

that both are

"sovereign rents', generating dysfunctional rent-seeking behavior.

However, at least as regards the detail of how revenues are transferred, aid

and oil rents

differ. Resource rents are unrestricted finance for governments, allocated

on a near-

random basis. By contrast, aid is provided in four purposive ways: technical

assistance,

projects, packages linked to conditions on past or prospective government

behavior, and

debt relief. These modalities of transfer bring with them both expertise and

conditions in

varying degrees. Aid allocation is purposive both between countries and

within countries.

The expertise and conditions that come with aid, together with its purposive

allocation,

1

See for example, "Vast majority thinks Africa aid is wasted, poll shows',

The Daily Telegraph, June 4

th

,

2005.

________________________________________

Page 3

may be merely second-order qualifying footnotes to a basic equivalence of

aid to oil, or

they may give rise to first-order differences in consequences for

development

effectiveness.

While donors would like to think that their actions enhance the

effectiveness of aid above

that of natural resource rents, a priori either mode of transfer could be

associated with

superior development outcomes. The governments of developing countries

frequently

complain about the delivery modalities of aid. They criticize technical

assistance as a

waste of money. They criticize project aid for being uncoordinated, and

requiring

procedures that divert government attention. They criticize conditionality

for

undermining government autonomy and frustrating government priorities.

Recipient

governments would undoubtedly see resource rents as superior,

dollar-for-dollar, to aid.

The different allocations of the two revenue flows as between countries

might also

generate differences in average effectiveness. Again, however, the

presumption that

because aid allocations are purposive, aid should generate superior average

outcomes to

resource rents is not necessarily correct. Resource extraction companies

have tended to

avoid investment in the most politically disturbed environments. For

example, although

oil reserves in Chad and Sudan were discovered in the 1950s, they were not

exploited

until the twenty-first century. By contrast, aid has sometimes been targeted

on highly

problematic governments for strategic reasons, such as Western support for

Zaire during

the Cold War.

Although there are now literatures on the development effects of aid and

resource rents,

the two have not been explicitly compared. Section 2 attempts to make this

comparison.

Section 3 addresses the question as to why aid is so much more effective for

development

than oil revenues: evidently, and contrary to much popular perception, past

aid modalities

have been strikingly successful in adding value to the resource transfer.

Section 4

addresses the question as to whether, given this success, there is scope for

"scaling-up'

aid in the literal sense of simply increasingly it proportionately across

existing modalities.

I argue that unfortunately the evidence points to the scope being rather

limited. However,

in Section 5 I build on the evidence of previous sections to propose six

innovations that

may have the potential radically to increase aid absorption.

2. Simulating a big push? The development consequences of resource rents and

aid

compared.

The current boom in the prices of natural resources is greeted with

enthusiasm by the

governments of Africa's exporting countries. I have already noted that at

least in the short

term the consequences of this bonanza have been somewhat disappointing, at

least for

their non-oil economies. While these short term outcomes may merely reflect

lags in

economic responses, they conform to a global pattern. Indeed, global

responses to

resource rents are considerably more disturbing than indicated by this short

term response

in Africa. Far from extra time revealing benefits, it reveals costs. A new

study measures

resource rents annually since 1970 for each country, rents being the surplus

of revenues

over costs of production (Collier and Hoeffler, 2005). It finds that

resource rents

significantly reduce growth: the impact effect is negligible, but with a lag

of around four

________________________________________

Page 4

years the rents have a significantly negative effect. These new results are

entirely

consistent with the large existing literature on the adverse effects of

resource revenues.

In the 1970s the dominant explanation was Dutch disease. Recently, without

rejecting

Dutch disease as being part of the problem, more emphasis has been placed

upon the

adverse effects of resource rents upon governance. This focus on governance

is pertinent

for the African aid debate. Aid may potentially have similarly adverse

effects on

governance as resource rents. Further, around a third of Africa has

sufficiently large

resource rents for the governance problems generated by these rents to be

critical.

Collier and Hoeffler provide an account as to why natural resource rents

worsen

governance. They find that resource rents subvert and indeed reverse the

normally

beneficial economic effects of democracy. In the absence of resource rents

democracies

growth more rapidly than autocracies, but with large resource rents

autocracies

outperform democracies. The critical level of resource rents beyond which

democracies

under-perform is around 8% of GDP. They suggest that resource rents tend to

tip the

balance between two different strategies of electoral competition. In a

normal democracy

parties compete to spend public revenues effectively on public goods,

balancing the

benefits against the costs of taxation. However, an alternative way of

gaining votes is to

bribe opinion leaders through private patronage. Where voters have strong

ethnic

loyalties and limited objective information, as is common in Africa,

patronage politics is

likely to be more cost-effective. In these conditions the only inhibition on

the government

from adopting patronage politics is if public revenues are protected from

embezzlement

by checks and balances. Electoral competition in conditions of ethnic

loyalties and poor

information drives political parties towards patronage, and only strong

checks and

balances can prevent it.

They indeed find that checks and balances significantly and distinctively

raise growth in

the context of large natural resource rents. For example, press freedom

shows up

statistically as being critical in mitigating the adverse effects of

electoral competition.

Unfortunately, although resource-rich countries most need checks and

balances, they are

least likely to have them: over time, restraints are gradually eroded by

resource rents. The

most likely mechanism for this erosion is that large resource rents

radically reduce the

need for taxation. An indication of this comes from comparing Africa's oil

exporters with

its other economies. On average there is no difference in government

expenditure as a

share of GDP: the governments of oil economies do not spend more, they tax

less.

2

Scrutiny of government is a public good the supply of which is commonly

provoked by

the tax burden. The lack of scrutiny in countries with large resource rents

makes it easier

for public revenues to be diverted into patronage: not only are public

revenues larger, but

they are less well defended. Hence, on this thesis, resource rents subvert

democracy by

making patronage politics financially feasible. In such environments, trust

must be placed

not in the good faith of political parties, but in the efficacy of

appropriate checks and

balances that enforce accountability of politicians to citizens.

Where social conditions make patronage politics cost-effective, one

financial parameter

is critical in determining whether electoral competition takes this

destructive form or

2

See IMF, 2005, Table SA11.

________________________________________

Page 5

facilitates development through public goods provision. This is the ratio of

"sovereign

rents' to "scrutinized revenues'. Here, "sovereignty' refers not to the

rights of the nation

versus external actors, but to the rights of the government versus citizens.

Sovereign rents

are that part of public revenue over which there is no effective scrutiny.

Given political

pressures to raise finance for patronage, the government will be driven to

use sovereign

rents in this way. Political leaders who attempt to resist these pressures

risk being

replaced by those willing to do what is necessary to maximize electoral

support.

Conversely, "scrutinized revenues' are that part of public revenue over

which scrutiny is

able to enforce spending on public goods.

Since patronage politics is expensive, the lower are sovereign rents the

more difficult is

patronage politics. Conversely, the larger are scrutinized revenues the more

prominent is

public goods provision in political debate. Hence, as sovereign rents are

diminished and

scrutinized revenues increased, there comes a point at which, even in social

conditions of

ethnic loyalties and poor information, the winning electoral strategy

becomes effective

delivery of public goods rather than patronage. The ratio of scrutinized

revenues to

sovereign rents determines the incentive for effective government. The key

menace of

sovereign rents is not that they are "wasted' on patronage, but that they

can destroy the

normal incentive for effective government provided by citizen pressure.

I have described sovereign rents and scrutinized revenues as if there were

two distinct

flows, one subject to scrutiny and the other not. Sometimes, most notably in

respect of

aid, this may indeed be the case. However, more commonly, the ratio is

determined by a

generalized level of scrutiny across all public revenues: the level of

scrutiny determines

the proportion of public funds that can be diverted into patronage. The

ratio may differ

more across types of expenditure than across types of revenue: for example,

it is probably

easier to divert expenditures on capital projects into patronage than to

divert expenditures

intended for salaries. Differences in the intensity of scrutiny between

resource rents and

aid may potentially generate different consequences for growth. This is the

issue to which

I now turn.

Unlike resource rents, aid has generally been found to be effective in

raising growth.

Clemens et al. (2004) provide the most recent and systematic study. They

find that

economic aid (as opposed to aid for other purposes) has significantly and

substantially

raised growth in Africa. Bad as Africa's growth performance has been over

the past three

decades, it would have been markedly worse without aid. Aid evidently has

very different

effects from resource rents. Indeed, when aid is introduced alongside

resource rents in the

Collier-Hoeffler growth regressions described above, the hypothesis that

they have the

same effect can be decisively rejected. This suggests that the superior

average results of

aid are not simply due to better allocation among countries: within a given

county aid and

resource rents have distinctive effects. In turn, this tells us that the

in-country modalities

of aid have made an important difference.

Aid agencies are adding value to the transfers that they administer, and

indeed doing so to

a very considerable degree. The evidence of oil implies that aid agencies

face an intrinsic

problem: the baseline effect of resource transfers is negative and the

agencies have to

________________________________________

Page 6

offset this by purposive allocation and complementary inputs. Nevertheless,

such an

activity need not be forlorn: an analogy with the effects of hospitals might

help to clarify

the point. The baseline for hospital activity is also significantly

negative. By bringing

patients with a variety of illnesses together in a single building, a

hospital transmits

disease. Even in well-run hospitals, many people contract illnesses from

others while

there, and spread these illnesses when they leave. Nevertheless, societies

rightly see

hospitals as vital: the value-added of a well-run hospital far offsets this

negative baseline

effect. This seems to be the story with aid agencies. The radical critics of

aid are correct

in the sense that the effects they point to are adverse and important, as

demonstrated by

oil. But their overall assessment is as wrong as would be a proposal to

close hospitals.

Indeed, their critique would be far more usefully directed to reforming the

governance of

oil revenues: the task of making oil work more like aid is far more

promising than the

task of making aid work better.

Why does aid have such different growth consequences to resource rents?

Whereas

resource rents appear to undermine the political process, aid does not,

enabling the

additional resources to have a growth-enhancing effect through investment

and other

productive expenditures. This may be the net outcome of two opposing

effects. On the

one hand, we would expect aid to increase sovereign rents by reducing the

need for

taxation. However, offsetting this, aid comes with various donor-imposed

mechanisms of

scrutiny which may spill over onto other expenditures and so substitute for

reduced

pressure from citizens. To take this analysis further, I now consider the

four modalities of

aid delivery.

3. Why is aid more effective than oil?

Recall that aid is delivered through technical assistance, projects,

packages with

conditions, and debt relief. Each of these is so distinctive that their

effects on sovereign

rents and scrutinized revenues, and hence on the incentives for effective

government,

need to be considered separately.

3.1 Technical assistance

Technical assistance is not money in the hands of recipient governments.

Thus it does not

increase sovereign rents. Indeed, to the extent that it is effective, it is

likely to increase

scrutinized revenues and so increase the incentive for effective government,

precisely

contrary to the effect of resource rents.

Perhaps because of this, technical assistance tends to be unpopular with

recipient

governments. Sometimes the fact that around a quarter of aid is in the form

of technical

assistance is depicted as a scandal, or even as a deceitful way of inflating

the true aid

figures. Nevertheless, such aid can potentially both build capacity in the

public sector and

substitute for the lack of it. A key test of such capacity building and

substitution is

whether it assists policy change. Potentially, technical assistance might

induce reform,

persuading governments of the need for change. Alternatively, when

periodically,

________________________________________

Page 7

reformers happen to attain political power but lack the capacity to design

and implement

effective change, technical assistance might support reform.

Recent research suggests that technical assistance does not induce reform

but that it is

highly effective in supporting it. Provided at the appropriate moment,

technical assistance

substantially increases the chances of success during incipient turnarounds

from

situations in which policies and governance are very weak (Chauvet and

Collier, 2005).

3

Further, since these configurations of weak policy and governance are

otherwise highly

persistent, the payoff to effective assistance is very high. The highest

payoff is to large

technical assistance programs provided early during incipient reform. A

package worth

5% of GDP sustained for four years substantially raises the chance that the

incipient

reform will progress to major sustained change. On average, such a package

would cost

about $1bn per country and the payoff would be around 15 times its cost. At

present,

technical assistance early in reform is typically supplied at far below this

level. However,

technical assistance can also be ineffective. When it is provided prior to

any sign that

reform has already commenced, it appears to have no significant effect on

improving the

chance that a turnaround will occur. Donors can spend their money preaching,

but

governments are unreceptive.

3.2 Projects

Aid in the form of projects is commonly criticized as cumbersome and

uncoordinated.

One attraction of projects is the partly illusory notion of accountability

that it provides to

donors. They can reassure their legislatures that aid has been spent on an

identifiable set

of activities that can then be evaluated. This can be illusory because of

fungibility: many

donor projects would otherwise have been financed by the government, and so

donor

money actually releases government finance for other priorities. However,

where aid is

very large relative to government budgets as is often the case in Africa,

fungibility is

reduced at least at the margin. For example, once government funding of the

development budget has fallen to zero, as is the case in some countries,

there is simply no

further scope for fungibility.

Even in the absence of fungibility, some project aid can be embezzled and so

augment

sovereign rents. However, projects also augment public goods. Donor projects

must go

through due process. A reasonable indication of the counterfactual of

African government

expenditure in the absence of donor projects is provided by Nigeria. The

combination of

3

Statistical analysis of the payoff to technical assistance during

turnarounds must confront the problem of its

endogeneity: if donors get the allocation of assistance right, they will

direct it to the most promising turnaround

situations. The resulting association between assistance and successful

turnarounds can easily be misinterpreted:

whereas in fact likely success is causing an inflow of assistance, the

association may be seen as assistance causing

successful turnaround. Chauvet and Collier control for this problem by using

a standard set of instruments for aid.

While some changes in aid reflect changes in the circumstances of

recipients, other changes reflect changes

experienced by the donor. For example, since Cote d'Èvoire gets a lot of

French aid and Ethiopia a lot of Italian aid,

when the French aid budget goes up and the Italian aid budget goes down, aid

to Cote d'Èvoire is likely to increase

relative to that to Ethiopia. Further, since donors differ in the proportion

of their aid which is in the form of technical

assistance, it is possible to isolate changes in technical assistance that

are unrelated to circumstances in the recipient

country.

________________________________________

Page 8

oil, military regimes and a large population was sufficient to make project

aid utterly

marginal to Nigeria's development. In Nigeria, even the elementary

requirement of

putting out government projects to competitive tender was abandoned. Such a

requirement is standard to all donor projects. The demise of competitive

tendering

dramatically inflated the costs of projects in Nigeria, something that can

be quantified

thanks to its reintroduction as part of the current reforms. Competitive

tendering has

lowered the cost of projects by around 40% (Transparency International,

2005). Hence,

the scrutiny inherent in projects is likely to have been markedly higher

than that for the

own resources of governments.

In addition to due process, donor projects bring sector knowledge and

management

techniques. Even governments that are making a huge success of development,

such as

China, are willing to pay for such knowledge. An indication of this is that

China

continues to borrow from the World Bank. China does not need the money, its

reserves

dwarf the resources of the World Bank and in any case it could borrow more

cheaply on

the markets. Evidently, it values the knowledge that comes as part of the

lending

package.

3.3 Conditionality

Donors have made various attempts to link aid to the behavior of African

governments.

One rationale has been to provide an incentive for behavioral change: in

this case

potentially the main pay-off to aid is the behavioral change it induces

rather than the

actual use to which the resources are put. A different and less ambitious

rationale has

been to channel aid into those environments where behavior is already most

conducive to

aid effectiveness. The potential array of design choices for conditionality

is defined by

the time period and the indicator on which aid is conditioned.

The time period can be ex ante or ex post: that is forward looking or

backward looking.

Ex ante conditionality, where it is credible, provides the strongest

incentive effect

because the donor specifies precisely both the amount and timing of aid that

will be

provided, and the government performance that is required. Ex post

conditionality

provides a more reliable means of channeling aid into specific behavioral

environments,

because the behavior is already in place before the aid is agreed. It may

have some

incentive effects but they are likely to be weak because in general

governments do not

know until some time after implementing change whether it will result in

more aid.

The indicator on which aid is conditioned can be policies, outcomes, such as

reductions

in infant mortality, or governance. This two-by-three space gives six

potential designs for

conditionality (see Table 1).

Table 1: The Matrix of Design Options for Conditionality

Policies

Outcomes

Governance

________________________________________

Page 9

Ex ante World Bank in

1980s;

IMF currently

European Union

Currently

Ex post World Bank

currently (IDA);

DFID currently

US currently, (MCA)

Ex ante policy conditionality

In the 1980s the World Bank invented ex ante policy conditionality. The

motivation was

in part that it realized that the policy environments in which it was

operating was often

seriously deficient. Further, the emerging need for large scale defensive

lending to

prevent default could not be achieved through the slow and cumbersome

modality of

projects. It therefore developed an instrument aimed at improving policy

through

negotiated aid conditionality: aid was provided in return for the promise of

policy reform.

The resource transfer was termed structural adjustment lending. On the whole

this was

unsuccessful (Dollar and Svensson, 2000). The failure reflected two

fundamental

weaknesses. First, governments learned to game the system by reneging on

their

promises. Aid was committed on the basis of a promise, yet the limited

continuity in

Bank decision-taking and the strong incentives to disburse made enforcement

through

future aid commitments incredible. In the event, some governments were able

to sell the

same promise of reform to the Bank several times. This weakness of

conditionality is a

straightforward instance of a class of problem known in economics as time

inconsistency.

The second weakness was that the coercive nature of the Bank's promotion of

policy

reform deepened government resistance to policy change. This is also a

straightforward

instance of a class of problem known in the psychology literature as

"reactance' (Collier,

2001).

While not abandoning conditionality, donors recognized these failures and

and by the late

1990s had substantially changed its design. The only agency left operating

the ex ante

policy conditionality mode was the IMF. Because of its core mission of

responding to

crises, the Fund did not have the option of change: of necessity it

continued to direct its

finance into situations in which the key problem was that policies needed

rapid revision.

However, the Fund took note of the ownership issue by reining back the scope

of its

conditionality to those policy changes that were demonstrably critical to

the success of

the program.

Ex post policy conditionality

The World Bank gradually shifted to ex post policy conditionality. That is,

aid was

allocated not on the basis of promises of policy change, but on the basis of

attained levels

of policy. The instruments for this were the Poverty Reduction Strategy

Papers (PRSPs)

which were to be drawn up by each country, and the Country Policy and

Institutional

Assessment, an annual rating of attained levels of policies which was used

to guide

allocations of the Bank's concessional lending, IDA. The term for such

lending is budget

________________________________________

Page 10

support. This reform substantially addressed both reactance and time

inconsistency.

Governments were now designing their own policies: governments were "in the

driver's

seat', and "owned' their policies. Any promises they chose to make were to

their citizens

not to donors. Some major bilaterals, most notably DFID, followed broadly

the same

allocation principles.

Ex ante outcome conditionality

The European Union retained ex ante conditionality, but switched from

policies to

outcomes: governments promised to achieve certain outcomes such as

reductions in

infant mortality in return for aid. This also substantially overcame the

problem of

ownership: governments were free to attain outcomes with whatever policies

they wanted

as long as they delivered the donor-desired outcomes. However, the

time-consistency

problem remained and was indeed probably accentuated. To see this, consider

what

happens if a government breaches its commitments. It is possible to detect

that a policy

commitment has not been kept well before a failure in an outcome commitment

can be

detected, and even once such a breach has been detected the government can

resort to the

argument that the missed outcome was due to circumstances beyond its

control.

Ex post governance conditionality

The US government switched to ex post governance conditionality, through its

new

vehicle, the Millennium Challenge Account (MCA). In this allocation

mechanism the

dominant consideration was the attained level of governance processes.

Underlying the

shift to governance conditionality were considerations both of legitimacy

and efficacy. In

terms of legitimacy, while policy conditionality was recognized as an

intrusion upon

national sovereignty, governance conditionality was concerned with

strengthening the

accountability of governments to their own citizens. So conceived,

governance

conditionality was intended to accelerate the process that elsewhere had

taken centuries

whereby governments came to share sovereignty with their citizens. In terms

of efficacy,

weak policies were diagnosed as generally reflecting underlying weaknesses

in

governance.

Mutual conditionality

Meanwhile, African governments themselves recognized that ex ante policy

conditionality had not merely been an affront to their sovereignty, but had

painted them

into a deeply damaging corner in which they were seen as invariably

resistant to reform.

The resulting perception had probably contributed to the severe erosion in

aggregate aid

flows during the 1990s. The governments of Nigeria and South Africa,

Africa's two

major powers and both newly democratic, led an internal movement for

improvements in

governance. Both through the power of their example, and through the peer

pressure

formalized in the review mechanism of NEPAD, they offered prospects of

improved

governance across the region. This was the basis for the Monterrey

Consensus, in which

African governments committed to improve governance and donors committed to

increase aid. The purported spirit of these commitments was that of mutual

trust rather

________________________________________

Page 11

than of formal conditionality: donor distrust had been replaced by donor

trust. However,

the deal can equally be understood as the passage from donor mistrust to

mutual mistrust.

Because neither side trusted the other, each side committed to gradual

processes of

improvement each of which could be monitored. African governments committed

to

regular self-scrutiny of political and economic governance through the peer

review

mechanism of NEPAD, and donors committed to move towards the UN target of

aid at

0.7% of GNP. Both parties committed to periodic monitoring at G8 meetings

and at

specially convened conferences. The recognition of mutual distrust has thus

initiated a

gradual shuffle towards the desired deal of improved governance matched by

increased

aid. The problem of reactance is addressed because there is no policy

conditionality, and

indeed no explicit conditionality at all. The problem of time inconsistency

œ this time for

both parties - is addressed by the two processes being in tandem: each party

must be able

to demonstrate to the other credible evidence of progress at each review

meeting.

However, the Monterrey Consensus introduces a new problem into

conditionality: each

side in this deal has very little control over its members and so is subject

to a potentially

severe free-rider problem. The limits of NEPAD's power to influence

recalcitrant

governments have been cruelly exposed over Zimbabwe. The limits of OECD

coordination have been similarly exposed in a highly variable pattern of

donor

commitments. Each side has tried to assist the other to coordinate. The MCA,

which was

announced at Monterrey, can be seen as an attempt to reinforce the NEPAD

incentives

for improvements in governance. Similarly, the recent creation of an

independent league

table of donor country performance can be seen as a way of increasing

pressure on

recalcitrant donors.

3.4 Debt relief

Debt relief is primarily driven by the ethical confusion and bureaucratic

duplication

involved in providing aid while at the same time demanding debt repayment.

However,

the case for debt relief would be more compelling were oil revenues more

effective than

aid: debt relief is the aid modality that, unless carefully managed, comes

closest to

turning aid into oil.

Once debts are cancelled a government has no incentive to abide by any

continuing

conditions. HIPC arrangements have attempted to overcome this problem by

having

donors pay debt service into an account which only liquidates the debt at

some future

date, the liquidation being conditional upon adherence to certain

conditions. However, it

remains to be seen whether this will be a credible threat. Fundamentally,

time-

consistency problems can only be overcome by conditioning behavior upon

flows rather

than stocks. Not only does debt relief face a potentially severe time

consistency problem,

the criteria of allocation tend to favor precisely the bad policy and

governance

environments where resources are least likely to be effective. This is

because large debts

to GDP or exports tend to arise where the latter have been depressed due to

the policy

and governance environment.

________________________________________

Page 12

The case for debt relief is, however, strengthened by four other

considerations:

unsustainable debt, odious debt, debt overhangs, and signals of

international approval.

Not all debt relief is a resource transfer. Indeed, were "debt

sustainability' to be taken

literally as the criterion for debt relief, then it would not generate any

resource transfer.

The only debt that would be written off would be that which was

unsustainable œ it could

not have been repaid. Such debt relief is strongly to be encouraged. Indeed,

were official

creditors subject to the same regulations as commercial banks, such debt

write-off would

be a legal requirement. However, to reiterate, by definition it would not

produce a

resource transfer to debtor countries. In practice, however, most debt

relief does produce

a resource transfer: the criterion of "sustainability' is a political

fig-leaf rather than an

economic assessment.

Where debts have been incurred by unaccountable governments that are

manifestly not

likely to use the money for the benefit of their populations, the lender

lacks moral

authority: the debt is odious. A significant variant of this is where the

borrowing

government has borrowed advantageously thanks to penalty clauses which it

has then

chosen to breach, leaving successor governments with the legacy of the

penalties. If

lenders introduce penalties knowing that the borrowing government lacks the

integrity to

avoid triggering them, the resulting liabilities are odious.

Debt relief locks donors into a resource transfer. It thus tackles the time

consistency

problem as perceived by recipient countries and this extra certainty can

have favorable

economic effects. Government debt reflects future tax liabilities œ the

overhang effect -

that can discourage private investment, since the investments will become

the future tax

base for the economy. Firms that invest in highly indebted economies are

thus

volunteering to be milked. By removing the uncertainty over how these debts

will be

repaid, debt relief thus encourages private investment.

Finally, whereas aid is a flow and so not particularly newsworthy, debt

relief is an event.

It therefore has the potential to provide a signal from the international

organizations that

the recipient government has already implemented changes that they judge to

be

important. Because the underlying behavioral changes are likely to have been

incremental, they may also have failed to attract media attention. The key

audience for

such a signal is usually international investors, but it may on occasion be

the domestic

population which has lacked objective information on which to judge

government

performance. Evidently, such a signal effect depends upon debt relief being

conditioned

upon attained performance.

4. So should aid be "scaled up'?

Were resource rents effective in development it would be a simple matter to

raise aid

effectiveness. There would be no difficulty in turning aid into oil:

technically this could

be done through either unconditional budget support or unconditional debt

relief.

Politically such a change would be popular with recipients and many NGOs and

indeed

characterizes the broad thrust of the current aid discourse. However, given

the

________________________________________

Page 13

considerably superior performance of past aid relative to resource rents,

such a

transformation of the modalities of aid would most probably reduce rather

than enhance

its effectiveness. The radical critics of aid are correct in arguing that

large sovereign rents

have been detrimental to development, but wrong in their casual conflation

of resource

rents and aid. So, have the existing modalities of aid created conditions

whereby it can

productively be scaled up?

4.1 What will be achieved on current modalities: evidence from past

experience?

Given that aid as channeled through existing modalities is effective in the

growth process,

the evidence might appear to support a substantial scaling up. However, like

much of the

literature, Clemens et al. find that aid is subject to diminishing returns.

They estimate that

economic aid hits the point at which it ceases to contribute to growth when

it is around

8% of GDP. Although some African counties receive less than this in economic

aid, on

average on this figure there would seem to be little scope for expansion. A

few of the

countries with aid inflows well below 8% are indeed "aid orphans,' that is

countries that

have only weak historical ties with the important bilateral donors: here

scaling aid up

would be appropriate without qualification. However, other apparently

under-aided

countries either have serious governance problems such as Zimbabwe, or large

resource

rents, and in such cases donors may well be appropriately taking into

account factors that

are inadequately considered in the econometric models.

The evidence for diminishing returns is not overwhelming, but it the best

assessment of

the dispassionate empirical literature, and it has some plausibility. It

suggests that aid to

Africa has more-or-less reached its appropriate scale. In Section 5 I

challenge this

position head-on. First, however, I consider a more conventional

justification for

enhanced aid: times have changed.

4.2 Have recent changes invalidated past experience?

Improved policies and governance in Africa

There is reasonable evidence that policies and governance are improving in

Africa. This

is certainly indicated by the Country Policy and Institutional Assessment of

the World

Bank, an annual country-by-country rating system. The average CPIA score for

Africa

has risen considerably. While this might possibly reflect "grade inflation'

on the part of

World Bank staff, objective performance as indicated by higher growth and

lower

inflation is also at a historic high. African governments might also point

to the

establishment of NEPAD and the AU as being indicative of greater concern

with

governance and as vehicles for its improvement. However, the NEPAD process

of peer

review is still in its infancy, and the earliest sensible evaluation would

be around 2007.

Although loosely modeled upon the OECD process of peer review, it is in some

respects

noticeably weaker: indeed, unlike the OECD, the mechanism is predominantly

one of

self-evaluation, rather than peer evaluation.

________________________________________

Page 14

Whether improvements in policies and governance actually increase the

absorptive

capacity for aid is surprisingly controversial: the econometric evidence is

mixed and

hotly disputed. This may be due to a conflation of two distinct processes.

As discussed

above, there is evidence that aid can sometimes be highly effective in

promoting policy

turnarounds in the weakest policy and governance environments. It can also

cushion

adverse shocks in the most vulnerable countries. These may be muddying an

underlying

but highly likely economic relationship that better policies and governance

raise the

return to public spending. My own view is that to the extent that the

improvement in

African policies and governance is genuine it will surely have increased

absorptive

capacity for aid. Clearly, improvement is not region-wide. However, the aid

allocation

rules of IDA and many other agencies already incorporate appropriate

selectivity. Hence,

an improvement in the average CPIA for Africa warrants an increase in

overall resources,

which existing procedures can then be left to allocate.

Improved donor practices

The second justification for expanded aid is that donor practices have

improved:

administering and managing aid has becoming less costly to its recipients,

and donors are

allocating aid more appropriately. Underlying this is a change in the

motivation for aid.

Until the 1990s much aid was provided for political reasons associated with

the Cold War

with relatively little engagement from OECD electorates. The focus on the

Millennium

Development Goals both reflects and promotes a greater concern among both

donors and

electorates that aid should be effective in reducing poverty. The extent to

which practices

have improved and its effect upon aid absorption is as yet not quantified.

Nevertheless,

my expectation is that this has received more emphasis than is warranted.

Most debate on

aid absorption is donor-dominated and the discussion may reflect a

self-absorbed focus

on the minutiae of their operations.

5. Potential innovations in aid modalities: complements to scaling up

Thus, so far, the more moderate critics of aid expansion appear to be

correct: aid is useful

but the scope for proportionate "scaling-up' is limited. The limits to

scaling up should not

surprise us. If a mouse were "scaled up' proportionately to the size on an

elephant it

would collapse under its own weight: increased size usually requires radical

change in

design. The present composition of aid may well be appropriate for its

present scale, but

not for a substantially larger scale. I now turn to six innovations that

have the scope to

yield quantum changes in aid absorption. The first four are new "funds'.

Unlike the new

Global Funds they are not specific to some narrowly defined activity such as

primary

education, nor would they require new bureaucracies. Rather, they would be

additional

money that donors would allocate to specific countries in particular

circumstances: they

would reflect changes in country allocation criteria. The remaining two

innovations are

complementary policies that both donors and African governments could

feasibly adopt

to address particular constraints on aid absorption.

5.1 Fund 1: Finance for big pushes

________________________________________

Page 15

While diminishing returns are plausible in theory and appear on average to

hold in

practice, there are circumstances in which, at least over a range, aid can

be expected to

yield increasing returns. The theory underlying this is sometimes referred

to as the theory

of the big push: it depends on the existence of complementarities and

thresholds.

Potentially complementarities and thresholds could apply to a variety of

development

goals. However, for Africa the critical problem has been a lack of economic

growth and

this is the objective on which I will focus.

Complementarities in the growth process imply that if several things are

done together

the returns to all of them are higher. Improved ports may only be useful if

the rural road

network is improved so that produce can be transported, and the extra

produce may not

be forthcoming unless credit facilities are enhanced. If each of these

improvements can

be made incrementally, then the economy can shuffle forwards: slightly

better ports

induce slightly better roads, which then feed back onto the returns to

roads. However, the

process of creeping improvement might be arrested altogether if there are

also important

thresholds.

Thresholds in the growth process imply that only once some scale of change

is reached

does investment become effective. One threshold that may be very important

for some

African countries is the point at which the country becomes competitive in

global

markets for manufactures. There is now strong evidence that there are

powerful

economies of agglomeration in manufacturing. Given that Asia already has

such

agglomerations, this creates a threshold that currently shuts Africa out of

global markets

in which it might well be competitive if only it could reach the necessary

scale. Note that

when Asia broke into export manufacturing it did not face competition from

established

low-wage producers and so did not encounter such a threshold problem. A key

attraction

of manufacturing exports as a growth strategy is that because African

countries are so

small relative to the global market, once they were over the threshold of

competitiveness

they could in effect grow without limit. EPZs could be used as focal points

for

infrastructure and business services that lower the costs of exporting.

The conjunction of complementarities and thresholds requires coordinated

leaps: no

investment looks to be worthwhile in isolation. Aid can provide both the

resources and

the coordination to make these leaps. Although the potential is largely

speculative, the

effect of such aid would be remarkably easy to monitor. By benchmarking the

costs in an

activity against the global competition it would be clear by how much costs

would need

to be lowered, and by monitoring these costs it would be apparent whether

aid

investments had achieved the objective.

A different type of threshold is that once growth rates become exceptional,

they get

noticed. Africa has not been devoid of economic success stories, but its

successes,

notably Botswana and Mauritius, have been very small countries. Were a few

larger

countries to grow rapidly, especially if the growth took the economy well

beyond any

previous point so that it could not be belittled as a mere recovery, the

pioneers would

become role models. In economic terms, beyond some growth threshold success

________________________________________

Page 16

generates learning externalities. It is not possible to quantify such

benefits, but the

process has obviously been important in both Asia and Latin America.

The big push postulates a range of increasing returns. This has a radical

implication for

aid allocation: aid should be concentrated in a few countries at a time. It

would obviously

be sensible to concentrate aid in those countries whose governments had

developed

ambitious but credible growth strategies. For example, if the key objective

is to break into

global manufacturing markets it would be sensible to concentrate on

countries that are

coastal, and that do not have large resource rents with attendant Dutch

disease effects.

Further, it would be sensible to concentrate on those countries that

currently have

manufacturing costs that are closest to competitive global levels. Such a

strategy of

concentrating aid would raise some political difficulties. However, for any

one country it

would essentially be temporary: aid would be big until the country was over

the

threshold, and then donors would move on down the queue of potential

recipients. If the

theory of the big push is correct, such a sequenced selectivity would be in

the interests of

all recipients.

There have been no recent experiments with such big push aid and so it is

not possible to

verify empirically whether such a strategy would work unless it is tried on

an

experimental basis in some countries. Within a chosen country donors would

simply need

to align their support around a growth strategy which could be part of a

PRSP. The more

difficult part is a significant increase in country selectivity. The most

difficult part is that

choices as to which countries were being favored would need to be common

across

donors.

5.2 Fund 2: Venture capital for turnarounds in failing states

As we have seen, technical assistance is effective in supporting

turnarounds. Situations of

incipient reform could have been supported more speedily and more

substantially with

technical assistance, and subsequently with more finance. This requires

changes in donor

allocation procedures and evaluation methods.

In terms of procedures, donors need to be able to respond swiftly to

turnarounds with

large increases in technical assistance. In effect, the provision of

technical assistance

should be organized more akin to emergency relief than to normal projects.

Without

changes in aid evaluation, aid agencies are unlikely to put more money into

turnarounds.

Like most bureaucracies, aid agencies are risk averse: staff avoid

situations in which their

decisions lead to visible failure. Much of the time, even with appropriately

supportive

aid, turnarounds will continue to abort. The rationale for support is not

that it has a high

probability of success, but that where it does make a difference the pay-off

is massive.

Such uses of aid need to be evaluated in ways analogous to a venture capital

fund. That

is, staff decisions are not assessed investment-by-investment, but rather on

the overall

return on a portfolio of decisions. The current attention to "results

orientation' in aid

agencies may have the inadvertent consequence of increasing risk aversion,

and so aid for

turnarounds probably needs a distinct fund which would be evaluated only as

an

aggregate.

________________________________________

Page 17

Post-conflict situations, of which Africa now has several, could also

productively absorb

considerably more aid that has typically been provided. Such situations are

distinctive not

only in their manifest need for reconstruction, but in their need for policy

reform.

Typically, countries emerge from conflict with a legacy of very poor

policies. However,

whereas poor policies are usually highly persistent, in post-conflict

situations they are

fluid. Starting from the same level, substantial turnaround is around four

times more

likely if the situation is post-conflict (Chauvet and Collier, 2005).

Indeed, by the end of

the first post-conflict decade, as long as there is no relapse into

conflict, policies (as

measured by the CPIA) are typically above the developing country average

(Collier and

Hoeffler, 2004). Aid is found to be more productive in the growth process

during this

decade than in other periods. Although some post-conflict situations, such

as Bosnia,

have attracted large aid inflows, the more typical situation is that there

is an initial spurt

of aid for the first couple of years after which it tapers out. Collier and

Hoeffler find that

this is not the pattern most conducive to growth: the peak phase for aid to

accelerate

development appears to be during the middle of the decade rather than at the

beginning.

Perhaps, in the early post-conflict period, although needs are great, the

capacity to spend

money effectively is limited and so large aid is more useful after

institutions have been

rebuilt. Overall, they find that aid flows to post-conflict situations could

productively be

substantially increased relative to historical levels.

Just as the big push approach is attractive because of the externalities

that follow from

having some strong successes in the region, so turnarounds are particularly

attractive

because those states that are currently failing generate large costs for

their neighbors.

Indeed, funding for turnarounds is the most promising way of using aid to

address the

problem of failing states, about which there is widespread global concern.

5.3 Fund 3: Aid in the context of weak governance: an alternative to social

funds

Where African governments are even moderately accountable to their citizens,

donor

conditionality is inappropriate: donors should be supporting credible

government-

conceived strategies. Unfortunately, in some African countries governments

are far from

being accountable to their citizens. At present, such countries receive

little aid and given

current aid modalities this is probably appropriate. The main aid modality

designed for

these environments is "social funds'. These finance projects channeled

directly to

communities rather than through the government. In effect, they pretend that

the

government was not there. The issue is whether donors might devise a more

effective

alternative.

Donors should not attempt to play the role of substitute citizens. However,

there may be

scope for redesigning conditionality so as to change the incentives in the

weakest

governance environments towards greater accountability to citizens. Recall

that I have

suggested that based on the evidence from resource rents, the roots of weak

governance

are the circumstances in which patronage politics trumps the delivery of

public goods.

Potentially, aid can be used to address this problem rather than to compound

it. To do this

aid would need to reduce the finance available for patronage and increase

expenditures

________________________________________

Page 18

on public goods. This can be thought of as ex ante governance

conditionality. If feasible

it offers the enormous advantage of channeling aid resources into the

countries with the

greatest needs for public goods, while at the same time addressing the

problems of weak

governance.

As implied by Table 1, the option of ex ante governance conditionality has

not yet been

tried to any extent. This is surprising because it is far more legitimate

than other forms of

conditionality. The essence of governance conditionality is that aid is

being used to

increase accountability not to donors, but to citizens. The sovereignty of

the government

is being limited, but not the sovereignty of the country. Such struggles to

limit the

sovereignty of governments occurred throughout the now-developed world.

International

pressures are generally judged to have played a crucial role in these

struggles: external

military threats forced governments to raise tax revenues, and this in turn

empowered

citizens to demand that governments create checks on their own power.

Clearly, this is

neither a feasible nor a desirable process in modern Africa, but other

external pressures

may be needed as a substitute.

The arrangements put in place for the management of revenues from the

Chad-Cameroon

pipeline provide a rare illustration of how ex ante governance

conditionality might work.

The arrangements actually addressed how resource rents should be used, but

the principle

could equally well be applied to aid flows. Indeed whereas the circumstances

in which oil

revenues came to be restrained are unique, the application to aid could

readily be

generalized to many other situations. Chad is a pre-reform environment in

which

resources under the sole control of the government are likely to be badly

used. This was

the rationale for the Revenue Management Law which required that 80% of the

resource

rents were to go into a special fund. Money from this fund could only be

spent upon a

specified set of social priorities, and this money must be incremental to

government

spending. To ensure that spending was incremental, the law required the

government to

allocate at least the same proportion of its non-oil revenue on these

priority sectors as it

had done in 2002, prior to the oil, namely 42%. The money in the special

fund could only

be released from the fund on the authority of a college including

representatives from

civil society, who would verify that all expenditures were indeed for the

approved

purposes. The money into the fund was released from an escrow account on a

quarterly

basis. This arrangement was time-consistent in the key sense that because

verification

was continuous the government did not have stronger incentives to make

promises than it

had to fulfill them.

These arrangements provide a possible model of ex ante governance

conditionality in

environments where governance is weak but needs for public goods are great.

As in that

model, aid would come as budget support, but would be limited to particular

parts of the

budget. As in that model, aid would be matched by the government's own

resources,

channeled into the same system month-by-month, on which draw-downs from the

aid-

funded resource flow would be conditioned. At the core of this system would

be

independent scrutiny such as that provided by the college, which would

verify and

authorize each item of expenditure from the resources provided. A key

advantage of such

an arrangement is that it would expand resources devoted to public goods

provision, at

________________________________________

Page 19

the same time as it squeezed sovereign rents of the government, and hence

the finance of

patronage. At some point the winning political strategy would switch from

patronage to

public goods provision. Once this switch occurred, the aid could gradually

evolve into

unrestricted budget support.

Not all governments where governance is weak would opt for such

arrangements,

although on the evidence of Chad some clearly would. However, the different

choices

would reveal where the improvement of governance was more and less

tractable. One

lesson to be learnt from the past failures of conditionality is to recognize

that the

instrument is indeed likely to have limits to its effectiveness. Unlike past

experience, the

choices of governments would reveal the likely limits before resources were

committed.

5.4 Fund 4: Aid to cushion terms of trade shocks

A final potential use of aid is to cushion adverse terms of trade shocks.

Whereas most

non-African developing countries have diversified their exports away from a

narrow

range of primary commodities, several African countries remain highly

vulnerable.

Uganda, Ghana and Madagascar, each with governments implementing economic

reforms, have been hit by large adverse terms of trade effects at

politically vulnerable

moments. There have been some experiments with using aid to provide direct

insurance

for African exporters. However, such an approach is administratively complex

and

misses the point that large shocks generate macroeconomic effects. The

depreciation of

the real exchange rate in response to a shock cushions exporters and

transfers the income

loss to other agents in the economy: it is thus like an automatic domestic

insurance

arrangement. What is most needed is therefore not a scheme targeted on

individual

exporters, but rather one that cushions the entire economy. Much the easiest

way of doing

this is for aid flows to governments to be countercyclical to the terms of

trade. The

essence of such a scheme is that it should be swift, and for this it needs

to be as near to

automatic as possible. Past aid schemes for cushioning shocks such as Stabex

fail this

test. One study finds that adverse terms of trade shocks have large adverse

repercussions

for growth but that timely aid can offset these effects, implying

exceptionally high returns

to aid (Collier and Dehn, 2001).

5.5 Complementary policies 1: offsetting Dutch disease

One likely reason for diminishing returns to aid is Dutch disease. Dutch

disease occurs

when aid reduces the competitiveness of the tradable sector. Many analysts

of the growth

process stress the importance of a depreciated real exchange rate,

(Hausmann, Pritchett

and Rodrik, 2003), and the Dutch disease effect of aid is a recurrent

concern of the IMF

(Rajan and Subramanian, 2005). There are four means of avoiding this

problem.

First, incremental aid can be targeted to expenditures that reduce some of

the costs that

are faced by important parts of the tradable sector, such as transport costs

and power

costs.

________________________________________

Page 20

Secondly, the import content of aid can be increased. Both technical

assistance and

infrastructure projects have much higher import content than expenditure on

government

salaries and so have a smaller impact on the real exchange rate,

dollar-for-dollar. It is

notable that the social priorities favored by donors during the past decade

neither reduced

costs in the tradable sector nor directly raised the demand for imports.

Third, recipient governments can offset the increase in the supply of

imports implied by

aid, with an increase in the demand for imports. The instrument for

achieving such an

offsetting increase in demand is trade liberalization. Where Dutch disease

is a real

concern, donors may need to confine large increases in aid to those

governments which

coordinate aid increases with trade liberalizations. Note that the purpose

of the

liberalization would not be to open markets to firms in donor countries, but

rather to

prevent the exchange rate effects of aid from contradicting its objectives.

This could, for

example, be achieved by a trade liberalization in favor of other developing

regions:

Africa currently has higher trade barriers against other developing regions

that against the

OECD. Since virtually all recipient governments currently impose fairly high

trade

restrictions there is plenty of room for increases in import demand.

Fourth, developed countries could reduce their current trade barriers to

developing

countries. This would raise the price received by exporters in aid-receiving

countries and

so would offset the disincentive to exports created by an appreciation in

the exchange

rate.

5.6 Complementary policies 2: donor coordination without harmonization

To the extent that improved donor coordination matters, the present approach

of

"harmonization' is on the past record unlikely to deliver much change. After

all, with

largely the same players the European Commission failed to achieve

harmonization of

product standards despite many years of effort. A simple approach to

reducing the

demands placed upon recipient administrations is for donors to adopt a

mutual

recognition of each others procedures. Mutual recognition was the key

procedural

advance that enabled the European Union to develop an integrated market for

goods: if a

product was accepted as meeting conditions for sale in the market of any

member country

it was accepted in all of them. The strategy of mutual recognition replaced

harmonization, which, though attractive in theory, proved impossible to

achieve in

practice. The history of aid coordination to date has in effect been that

the unattainable

holy grail of harmonization has impeded more practical alternatives. With

mutual

recognition, a recipient government would be free to adopt whichever donor

system was

most compatible with its own procedures. Other donors would then adopt that

system for

that country. Donors would find themselves running different systems in

different

countries, but this administrative burden is probably within their

competence. The

problem at present is that it is the weak administrations of recipient

countries that are

confronted with the burden of multiple systems. However, if donors found the

burden of

multiple systems intolerable, this would be the only incentive for

harmonization that is

likely to prove effective.

________________________________________

Page 21

Beyond coordination and mutual recognition, much the most effective way of

coordinating aid is to pool resources financially, letting a single agent

decide how to use

them within certain agreed limits. To a degree this is done already. Donors

pool funds

into IDA, leaving the World Bank to implement. They also pool resources into

budget

support, leaving the government to implement. However, for political reasons

bilateral

donors retain most of their funds for their own projects. One political

reason is that

projects provide at least a semblance of accountability. For example, when

the

government of Malawi was criticized in the international media for

purchasing a fleet of

Mercedes, DFID was able to claim that this had not been financed by British

aid because

all aid to Malawi was in the form of projects. Were budget support to be

redesigned in

such a way as to provide much greater accountability, effectively precluding

such uses by

recipient governments, it might be possible for donors to devote more

resources to it.

6. Conclusion

If aid is like oil the economic case for a major expansion of aid to Africa

is disturbingly

weak. The evidence strongly indicates that on average aid has been much more

effective

at promoting development than oil. Since both are transfers to governments,

the profound

difference in their consequences must lie in the details of how aid

transfers are made.

Even though aid is not like oil, the scope for substantial expansion may

nevertheless be

limited by diminishing returns. The available evidence suggests that if aid

is simply

scaled-up proportionately, the incremental aid might indeed be much less

effective,

dollar-for-dollar, than existing aid.

There is, however, scope for innovations. Additional aid could support four

strategies that

have to date been neglected: a big push through concentrating large

resources in the most

promising environments; assistance for turnarounds; in the weakest

environments finance

for public goods channeled through required improvements in governance; and

finally aid

to cushion terms of trade shocks. Each of these would generate externalities

across the

region. Investing aid in accentuating success would provide African role

models for the

rest of Africa to follow. Investing aid in the weakest environments would

reduce the

crisis situations that spill over onto neighbors. There is also scope for

complementary

policy changes that would mitigate Dutch disease and reduce the transactions

costs

associated with aid projects.

The evidence for these innovations varies. The most speculative is the big

push.

However, fortuitously, it lends itself to being an evaluated experiment. The

country-

sequenced strategy implied by the logic of scale economies simulates a pilot

experiment,

and since success involves being a role model for the region, performance is

inherently

observable to donors. There is thus a good case for trying the approach,

recognizing that

it may have to be abandoned. The venture capital model of financing

incipient

turnarounds is, by contrast, the most difficult to evaluate, but there is

reasonably robust

statistical evidence that it would have a very high payoff. The attempt to

use ex ante

governance conditionality to build new institutions for channeling money

into the

weakest and most needy environments is new territory but not inherently

risky. The

________________________________________

Page 22

money would only flow where the new institutions of governance ensured it

would be

genuinely accounted for, and provided some prospect of increased

accountability to

citizens. The provision of aid for macroeconomic insurance against shocks

would begin

to address the evident problem that the poorest countries face high levels

of risk and so

need some contingent financing. While the evidence for a high pay-off is

currently

limited, the lack of contingent financing instruments can only be rectified

by a phase of

experimentation. The evidence for the complementary policies is analytic

rather than

empirical and essentially straightforward.

Something surely needs to be done decisively to raise African growth rates:

the continued

marginalization of Africa in the world economy would have incalculable

consequences

that it is in our common interest to avert. The new strategies look

sufficiently promising

to be worth trying. New strategies do not come with guarantees, but in the

face of

problems that have proved intractable there is no alternative to such

strategic

experimentation.

Finally, I note that in this article I have used the gulf between the

development

consequences of aid and oil to consider aid. However, the more important

implication

from the aid-oil contrast is for oil. If oil could be turned into aid the

development

consequences would be enormous. Oil and other resource rents are in

aggregate much

larger flows to developing countries than aid, and their effects to date

have been on

average significantly negative. Raising the development effectiveness of oil

revenues and

other natural resource rents must, however, be the subject of a separate

paper.

References

Chauvet, L. and P. Collier, 2005, Policy Turnarounds in Failing States,

Department of

Economics, Oxford University.

Clemens, M., S. Radelet and R. Bhavnani, 2004. Counting Chickens when they

Hatch,

WP40, Center for Global Development, Washington DC.

Collier, P. 2001. —Consensus Building, Knowledge, and Conditionality," in

Annual Bank

Conference on Development Economics 2000, B. Pleskovic and N. Stern, eds.,

Oxford

University Press.

Collier and Dehn, 2001. Aid, Shocks and Growth, WP2688, World Bank,

Washington

DC.

Collier, P. and A. Hoeffler, 2004, Aid, Policies and Growth in Post-Conflict

Societies,

European Economic Review, 48, 1125-1145.

_____________________, 2005, Democracy and Natural Resource Rents,

Department of

Economics, Oxford University.

________________________________________

Page 23

Dollar, D. and J. Svensson, 2000. What Explains the Success or Failure of

Structural

Adjustment Programs? Economic Journal, 110, 894-917.

IMF, 2005. Regional Economic Outlook: Sub-Saharan Africa, Washington DC.

…