In September 2011, the European Commission published the latest in a series of studies which estimate in quantitative terms what might be the benefits of measures to improve aid effectiveness at European level. In April 2012, ODIOverseas Development Institute (London) published a commentary, with a response by the original authors. Neither the original paper, by Bigsten, Platteau and Tengstam (henceforth BPT), nor the commentary, by Prizzon and Greenhill (henceforth PG), examined the practical implications. This note attempts to fill the gap. We need to begin, however, with a recap of the analysis. The key point out to watch out for is that the estimated benefits are notional, in the sense that they are not grounded in actual policy, but based e.g. on benefits that would accrue from a one standard deviation change in the variable being considered. Neither BPT nor PG translate the numbers into real world policy. Nevertheless, there are significant practical implications, summarised at the end of the paper. In brief they are:

1. The BPT figures are notional and should not be cited in any policy-related context.
2. The potential savings from reducing the number of countries are likely to be small in the short term, though larger in the medium term, as countries graduate. A significant benefit comes with the Division of Labour Agenda if aid administration costs by 27 Member States and all the EUEuropean Union institutions can be reduced as a result.
3. The costs of administering aid could increase substantially with a cut in budget support.
4. There may be some modest benefits in further work on untying, although on a small scale compared to the other benefits.
5. The biggest gains by far are to be found in allocating aid more efficiently as between countries.  This will increase efficiency, but also, if countries can graduate completely, save on aid management costs.

A recap of the analysis

BPT offer a careful analysis, with a comprehensive literature review, a thorough theoretical discussion, and a number of innovative empirical investigations. PG largely agree with this judgement, though with some important methodological qualifications. PG also point out that many aspects of aid effectiveness (captured in the Paris Declaration) have not been included.

BPT review five issues, covering the EUEuropean Union (EC and MS) and using data for 2009.

1.   The saving in transactions costs associated with:

a.   Reducing the number of countries covered by donors; and

b.   Moving from project to budget support.

2.   The benefits of reduced untying.

3.   Reducing the volatility and unpredictability of aid.

4.   The impact of (1) – (3) on growth.

5.   The benefit of reallocating aid so as to maximise the impact of EUEuropean Union aid on poverty reduction.

The overall findings are summarised in Table 1. Remember that the estimates are notional. Note especially, however, that the gains from reallocating aid dwarf those from other efficiency measures. On the efficiency side, growth effects are about equal to cost savings.

Table 1: Summary of the benefits of EUEuropean Union implementation of the Paris Agenda

With regard to savings on transactions costs:

1. BPT run a regression which estimates that reducing the number of countries receiving aid by 1 reduces the administrative overhead by 0.54%. They then assume a reduction in the number of aid-receiving countries by one standard deviation, 37%, which coincidentally is 37 countries – and this then gives a saving in admin costs for the EUEuropean Union of 20% or €461m. This is equivalent to €12.46m per country. PG endorse the methodology, though question some aspects of the definition of admin costs.
2. To work out the cost saving of moving from project to budget support, the authors use a literature review to underpin an assumption that budget support costs only 33.5% as much to administer as projects. They then assume an increase in the share of budget support in country programmable aid (CPA) from the present level of 43.7% to 66% (the Paris Declaration target). This saves €283m. PG agree, though caution that the saving is estimated on the basis of only one (Swedish) study of comparative admin costs for different types of aid.

With regard to the benefits of untying, the estimate of the benefits is based on a literature review, with the final assumption that tying increases the cost by 22.5%, so that the share of flows lost due to tying is 18.4%. This figure is then applied to the tied share of EUEuropean Union aid in 2009, to give a total cost of €800m. PG are happy with the estimate.

The benefit of reducing volatility is calculated by applying a Capital Asset Pricing Model. Looking only at country programmable aid, the dead weight loss associated with volatility is calculated as 15 cents per dollar. The total benefit is then calculated as €1.7 billion. PG endorse the model, but point out that it says little about predictability, which is not the same as volatility. BPT themselves offer a number of caveats: for example, volatility may reflect conditions in the donor country (‘bad’ volatility), or may reflect changes in need or objective conditions in recipient countries (which they qualify as ‘good’ volatility).

The impact of changes in the aid relationship on growth are calculated by regressing aid modalities on growth, with the variables being (a) aid fragmentation, (b) the share of General Budget Support, and (c) the share of tied aid. The number of countries included is not given. Only the share of budget support is found to be significant. Increasing the share of budget support by 11% (one standard deviation) increases the GDPGross Domestic Product of recipients in the following year by €1.8bn. PG are most sceptical about this finding, mainly for econometric reasons, and say that the ‘findings should be approached with a great deal of caution’. BPT do not disagree.

Finally, the benefit of reallocating aid so as to maximise poverty reduction. This is calculated by using a modification of the Collier and Dollar approach. The aid saturation point for Country Programmable Aid is assumed to be 10% of PPP GDPGross Domestic Product (derived from Clemens and Radelet’s figure for all aid and nominal GDPGross Domestic Product of 30%). Major reallocations are required, from aid darlings to aid orphans. Major losers (in absolute terms) include Afghanistan, Turkey, India, Morocco, China and Indonesia. Major gainers include Bangladesh, Ethiopia, Nepal, Kenya and Uganda. In total, up to €19bn is to be reallocated, from total CPA of €27bn. The poverty reduction effect of a dollar in the ‘darling’ countries is only 15.2% of the effect the same dollar would have in the ‘orphan’ countries. Hence, if there were no differences in governance (or effectiveness of aid), then the benefit of reallocation (i.e. the amount available for poverty reduction) would be €19bn – (0.15*€19bn)  = €16.4bn. However, adjusting for differences in governance  between (some) darlings and (some) orphans, gives a net benefit of €7.8bn. NB This is aid available for poverty reduction, not an estimate of GDPGross Domestic Product impacts or income for the poor. PG approve the methodology, but point out that reallocation on this scale is politically unlikely; they wonder about the impact of more modest reallocation. BPT agree that such analysis would indeed be useful.

Practical implications

What, then, might be the practical implications. We need to take this a step at a time.

First, it would be quite wrong for policy-makers to cite the aggregate numbers in the summary table. It is easy to see the temptation to claim €3.2bn or even €7.8 bn as the benefit of full coordination, but, as noted, these are notional numbers which do not reflect actual, feasible or desirable policy decisions.

Reducing the number of countries

Second, then, with regard to reducing the number of countries, BPT suggest that the benefit to the EUEuropean Union of donors collectively reducing the number of countries to which they provide aid by one is €12.46m. They say the current number of recipients is 100. This compares to the total number of aid-eligible recipients listed by the DAC of 148. Of these, 48 are least developed, 6 other low income, 40 lower middle income, and 54 upper middle income. Some of the countries are large (China, India) and some are small (Palau, St Vincent): it is unlikely that it costs precisely € 12.46m to provide aid to each, but that is a statistical artefact we have to live with.

Say that the donors were to provide aid only to least developed countries and other low income countries. That would mean that 46 countries could be cut from the list of aid recipients, giving a total saving to the EUEuropean Union of €573.2m.

Say that the donors were to provide aid to LDCs, OLICs and LMICs. That would cut 6 countries from the list, and save the EUEuropean Union €74.76m.

Those numbers probably provide upper and lower boundaries of possible savings. Note that the Commission is proposing to cut bilateral aid to 19 countries, which could theoretically generate savings of €236.74m – except (a) the Member States would have to follow suit, and (b) the 19 countries will still be eligible for aid under thematic programmes, for example from the DCI, which will need to be administered.

A range of savings from €75-573m may not be very helpful for policy purposes, but it seems evident that the potential savings are likely to be small in the short term, though larger in the medium term, as countries graduate.

It is important to note what this calculation does and not do. It estimates the savings to the EUEuropean Union if all donors pull out of a country. It does not estimate the benefits if some actors pull out of some countries. This, of course, is the principle behind Division of Labour, and also what many countries are doing in practice: the UK, for example, has reduced the number of recipients to only 28. It would be interesting to apply the BPT methodology in that disaggregated way.

Moving from project to budget support

Thirdly, the benefit of moving from project to budget support is estimated on the basis that 44% of CPA is currently distributed in the form of budget support, with the figure moving to the Paris target of 66%. The best that can be said about this is that it seems extremely unlikely, even if ‘budget support’ is defined generously to include sector budget support and some other aspects of programme-like support – there are major definitional problems, discussed in a piece I wrote on budget support in 2011, drawing on research by ODI’s Centre on Aid and Public Expenditure.

In fact, many donors are moving in the opposite direction. DFID, for example, will be cutting budget support by over 40%, although with some relabeling as results-based contracts. Recent parliamentary enquiries have endorsed that position. The EUEuropean Union Commissioner also appears to have resiled on a commitment to spend 50% of aid as budget support. ACommission Communication was published in 2011, and Council Conclusions are awaited.

It seems likely that the costs of administering aid might well increase if these new policies are followed, or perhaps better should increase, if budgetary constraints allow. Assuming symmetry in the costs and benefits if budget support rises or falls, a reduction in the budget support share in CPA to, say, 30%, should logically result in an increase in administrative costs of €177m!

The benefits of reduced untying

With regard to untying, BPT make their estimate on the basis that tying would be eliminated altogether, falling from a current level of about 9% (excluding technical assistance) to zero (for all aid, including technical assistance). It is a moot point how likely is this. Some EUEuropean Union countries tie very little or not at all (the UK is an example), others very substantially: from BPT’s figures, Austria, Greece, Italy and Portugal all have shares in excess of 40%. All are small donors, however. The EUEuropean Union average (for DACDevelopment Assistance Committee (of the OECD) members only) is well below that for the DACDevelopment Assistance Committee (of the OECD) as a whole (9% compared to 15%) – and remember that the EUEuropean Union provides about 70% of all DACDevelopment Assistance Committee (of the OECD) aid, so the figure for non-EU DACDevelopment Assistance Committee (of the OECD) members will be much higher.

Say that half of all EUEuropean Union tying could be eliminated, which might be a realistic ambition over a number of years. That would generate savings of perhaps €400m.

Reducing volatility

The gains from reducing volatility are estimated at €1.7bn for the Commission and Member States together, which dwarfs the savings in transactions costs from improved aid management. This is not straightforward, however. First, the figure assumes complete elimination of all volatility in country programmable aid (NB not humanitarian aid or other insurance-type payments). Secondly, both BPT and PG list problems with the methodology, including a rather key question about whether it is really applicable to aid flows. In addition, as noted, some volatility may be desirable. For example, BPT suggest that variable flows from the EUEuropean Union institutions may play a role in smoothing volatility which arises from action by Member States. PG make the important point that volatility is not the same as unpredictability, and that many of the supposed costs of volatility (enumerated in the literature review by BPT) may actually be the result of unpredictability.

Unfortunately, neither BPT nor PG provide enough information to judge whether the estimated gains from reducing volatility are completely convincing or completely spurious, both unlikely, or where the pendulum might settle between the two extremes. Nor is it easy to see what level of reduction in volatility might be desirable or achievable.

The policy conclusion might be that predictability and ‘good’ volatility need to be protected and expanded. The way to do this is probably to introduce greater transparency and accountability into aid planning, using explicit results frameworks, formal review processes and arbitration or conciliation procedures. Most donors are moving in this direction, at least as far as results frameworks and review processes are concerned. DFIDDepartment for International Development partnership agreements provide examples. The EUEuropean Union has gone further, for example with respect to the mutual accountability and arbitration provisions of the Cotonou agreement.

It is very hard to quantify the benefits globally, but might be interesting to examine through country case studies.

Indirect effects

BPT move onto fairly speculative territory when they examine the growth effects of reinvesting the money saved by more effective aid in growth, as they themselves acknowledge, and as PG emphasise. Remember that only increasing budget support is significant in their calculation. Increasing the share of budget support by 11% (one standard deviation) increases the GDPGross Domestic Product of recipients in the following year by €1.8bn. As it happens, this is only half the increase used in the calculation of the saving in admin costs of moving to budget support – but also we have argued that the likelihood of moving to budget support is remote, so the savings are also, and therefore the growth effect.

More seriously, there is an apples and pears problem. Can the growth effect legitimately be added to the cost savings? Would a better comparison not be with the growth effect of reinvesting the savings from lower transactions costs and lower volatility? This issue is discussed in the report on Pg 121, which makes the point that ‘this money cannot directly be compared with the yearly cost savings discussed elsewhere’. In that case, it seems perverse to include it, even in brackets, in the summary table.

Reallocating aid

Finally, BPT look at the question of aid allocation. The make a point frequently made in aid debates, that poverty impact could be increased by allocating aid more efficiently as between countries. The potential gains are substantial, dwarfing the others considered. However, BPT observe that

‘one must then ask why the actual allocation is so far from the “optimal” allocation. Obviously EUEuropean Union donors have other aims apart from the maximization of global poverty reduction. They want to be present in a broader range of countries for economic and political reasons, which mean that there are political constraints on the reallocation our analysis suggests. Coordination could be organised in the form of Commission aid or by coordination of bilaterals, but irrespective of what form the coordination would take, there remains strong political restrictions on what can be achieved.’

PG agree with these qualifications, especially with regard to complete reallocation. They wonder what would be the impact of ‘more moderate, partial’ reallocation. Given the scale of the potential gains, this is obviously a sensible question to ask.

In fact, the aid allocation issue is at the heart of many current debates about EUEuropean Union aid, especially with regard to the European Institutions. It underpins Commissioner Piebalgs’  – as it turns out, controversial – proposals on differentiation, embodied in Agenda for Change, and in the proposed revision of the Development Cooperation Instrument. TheInternational Development Select Committee of the UK House of Commons has recently looked at the question, with particular reference to the European neighbourhood. For example, it has asked tough questions about aid to some of the countries listed as ‘losers’ in the BPT analysis, especially Turkey, but also China (and India). The Committee has also examined the question of whether DACDevelopment Assistance Committee (of the OECD) criteria for oda eligibility should be revisited.

Say that half the benefit of reallocation estimated by BPT could be achieved. That would amount to €3.9bn, easily dwarfing the other gains. Even a 25% success rate would ‘liberate’ nearly €2bn, a very large sum

Conclusions and further research

There are some important areas where it would be useful to carry out further work. PG make a number of suggestions regarding methodological development. In addition, we have suggested or might suggest:

1. More disaggregated work on the benefits of reducing country coverage by individual donors;
2. Better data on budget support and other programme-like approaches, along with careful tracking of donor trends;
3. Perhaps an analysis of the scope for further untying of aid in the EU;
4. Country level case studies of the benefits of greater predictability and lower ‘bad’ volatility; and
5. More work to understand aid allocation decisions.

Practically speaking, what conclusions can be drawn from the analysis? Look again at Table 1, reproduced below for ease of reference.

Table 1:Summary of the benefits of EUEuropean Union implementation of the Paris Agenda

First, to repeat, no-one should cite the aggregate figures in any policy-related context. They are not designed for that purpose.

Second, I would disregard the ‘indirect effects’ which are apples, so to speak, not pears, and of a different character to other gains. That does not mean there are not growth benefits from more efficient use of aid, but they follow the immediate cost savings and are different.

Third, the biggest gains by far are to be found in pursing the inter-country allocation issue, which will increase efficiency, but also, if countries graduate completely, save on aid management costs (in other words, a combination of the savings in first and last lines of the table).

Fourth, there should also be significant benefit in pursuing the Division of Labour Agenda if aid administration costs by 27 Member States and all the EUEuropean Union institutions can be reduced as a result.

Fifth, there may be some modest benefits in further work on untying, although on a small scale compared to the other benefits.