Solving the Private Sector Imbroglio

Disagreement over how investments in the private sector are counting to aid is threatening to overwhelm the OECD Development Assistance Committee. There are no perfect solutions here; governments must find the least-bad compromise. We point the way forward.   

In 2014 the OECD Development Assistance Committee (DAC) agreed major changes to how concessional official-sector loans would be counted as aid.^[1] The basic idea was that donors can lend £100 on terms that are sufficiently generous that it amounts to giving away £20. Henceforth this “grant element” would be what gets counted as official aid. The DAC then turned its attention to aid invested in private sector enterprises, where it was expected the same principle could be applied without much difficulty. The February 2016 DAC high-level meeting communique specified that the measurement of “private sector instruments” (PSI) would be also based, where possible, on the grant equivalent method. All that remained was to iron out the technical details.

That proved easier said than done. The OECD is still pushing for an agreement in time for October’s high-level meeting, but those involved say that is unlikely. Aid to the private sector turned out to be a Pandora’s box of contradictions. Before getting to those, it’s worth reprising what is at stake.

Why these rules matter

If the way in which private sector investments are counted as aid is hard to justify, the credibility of aid statistics in the eyes of the public may be damaged. If no deal is reached, some countries may create their own aid statistics, which would also undermine public understanding. The DAC would then no longer be the sole arbiter of aid, and researchers would no longer have a single consistent source of aid data.

If the rules are too generous, then things that really should not be counted as aid will be, and donors will face an incentive to divert money away from health clinics or social protection programmes and into profitable private enterprises, to get more recognition for aid spending than they deserve. If checks and controls are too lax, aid could be used to subsidise domestic firms. There could be conflict with the OECD’s official export credits arrangement, and even a risk of governments being hauled in front of the WTO. Claims to have “mobilised” additional private finance could be hollow.

On the flip side, if ODA recognition does not reflect the true costs donors bear, they will be less inclined to put money into mobilizing the private sector, especially in the poorest countries, which is widely regarded as crucial to the success of the Sustainable Development Goals. If complying with safeguards is too time consuming, or project promoters and private investors are concerned about the confidentiality of financing terms, they may be unwilling to co-invest with development finance institutions (DFIs). In some cases, without the ability to retain confidentiality until deals are signed and sealed, DFIs fear that their ability to do business may unravel.

The opposing teams

One set of countries is more concerned about the former set of risks, another set the latter. As negotiations have dragged on, more countries have become increasingly impatient, creating pressure for a deal, raising the risk of a bad outcome. The EU has a new EUR3.6bn European Fund for Sustainable Development that needs to be ODA eligible and that it wants to start spending. The UK wants to know whether the capital injected into CDC will be eligible, or whether it needs to “change the law to allow us to use a better definition of development spending, while continuing to meet our 0.7 percent target,” as the Conservative manifesto put it.

To get things moving, many of the thorniest problems around competition safeguards have been kicked down the road into a “phase two” of talks (although there are still disagreements over what phase two should tackle). Negotiations are still stuck at square one: how to calculate the grant element of a loan. Although they have other objections too, there are three “holdouts” who think the proposed deal gives insufficient recognition for their private sector efforts: France, Germany, and Japan. On the other side is a fraying coalition of smaller countries and the United States, who are more concerned about overly generous ODA scoring and insufficient safeguards. But every country has things that matter more or less to them, so allegiances are shifting.

Instruments and institutions

But there is another twist in the pattern of alliances across countries. In 2016 two methods for reporting aid to the private sector were agreed: calculating the grant element of each transaction between a DFI and private enterprises (the “instrument” method); and scoring the entire face value of capital that governments put into (qualifying) DFIs (the “institutional” method). Furthermore, some countries do not even report their investments in the private sector as aid, perhaps because their DFI does not qualify, or because they regard it as more trouble than it is worth, or even because they believe it ought not count. 

This has created the perception in some quarters that countries who will not be using the instrument method are wilfully causing trouble for those who will. Hopefully this perception is misplaced: the UK gains nothing worthwhile by France reporting somewhat lower ODA numbers, for example. The focus should be on what encourages the greatest development impact and produces a fair measure of donors’ costs. Nonetheless, this dynamic is present in the negotiations.

The institutional method was agreed at the instigation of the UK, not (so the story goes) to inflate its ODA numbers, but to give the Department for International Development a large number it could decide at the end of the fiscal year, to hit its 0.7 aid target bang on and avoid negative headlines in the Daily Mail for marginally overspending.

In theory, the institutional method works on a cash flow basis: positive ODA is scored as money goes into DFIs and negative as DFIs remit money back to the government.^[2] The grant element, if any, will emerge over time. But DFIs are unlikely to sell their assets and repay governments until the distant future, at a time when international development is no longer regarded as important. Little attention is then likely to be paid to negative ODA and the implications for aid data reported decades ago. In politics, a problem deferred is a problem solved. So, in practice, governments can put money into qualifying DFIs and score the lot as ODA today, without so much to constrain what they do with it. Some constraints should be imposed by the DAC requirement that donors who use the institutional method also calculate ODA by the instrument method, and if the results of the two methods are too far apart, do something about it. How well that will work, time will tell.

Some of the countries involved in negotiations regard the institutional method as an anathema that will inflate ODA figures and remove scrutiny at the transaction level. Other countries are drawn to its administrative attractions and (reportedly) seven countries have chosen to join the UK in using it. But the institutional method is part of the overall package, and donors will not be allowed to use it unless the whole deal is agreed. If that happens, the UK is threatening to break from the DAC and define its own aid statistics. “No deal is better than a bad deal” say those on the other side of the table, with some Brexit-related sniggering.

The bone of contention: discount rates

Now back to the impasse over loans. To calculate the true economic cost to governments of investing in private enterprises, one needs to pick a number. That number is called the discount rate, and it is the yardstick by which to measure the grant element of a loan. At this point it is hard to resist a lengthy digression into the conceptual foundations of discount rates, but I shall nobly forbear and concentrate on the dilemma currently occupying the DAC. A higher discount rate implies more ODA in every loan. For sovereign lending the DAC chose the IMF’s unified discount rate of 5 percent and added risk premia for countries based on their income category. On the basis that lending to the private sector is typically riskier than lending to a sovereign, if anything, discount rates for PSIs should be higher, all else equal (although there are some countries where the government is not creditworthy but one or two large enterprises might be). But even when using the discount rates agreed for sovereigns, some loans made to the private sector at market rates, and at rates classified as non-concessional by official export credit agencies (ECAs), would qualify as aid. Classing loans that a private lender would happily offer as aid is absurd, and one branch of government calling a loan aid that another branch of government would insist is not aid, is regarded as dangerous by the ECAs. Fixing this problem would require dropping discount rates beneath those used for sovereign loans, which is regarded as incoherent by some countries and might even raise the prospect of reopening those negotiations, which nobody wants. Language has been added to the agreement stating that whatever is agreed on PSIs has no implications for the agreement on public lending. But glaring inconsistencies may not go away so easily.

To an extent, these problems have always been there. Some sovereign loans may also count as aid even at rates seen in private markets. The DAC claims that private sector instruments are “non-concessional in nature” which on the face of it is incoherent—if you are not giving something away, it is not aid. But what the DAC means is that private loans are typically less concessional than loans to the official sector must be, to qualify as aid. The DAC suggests that a typical grant element for a private sector loan would be around 10 percent. A sovereign loan to an LDC needs to have a grant element of at least 45 percent. Some countries have argued that for private sector instruments, ODA should be thought of as measuring something called “donor effort,’’ which should not be interpreted as a grant equivalent and equated with concessionality. That is an argument made by civil servants looking for a way out of an impasse. The only intellectually coherent solution is to choose a sensible discount rate that only generates ODA content when you are doing something that the private sector would not, and which can be interpreted as a measure of concessionality. 

What is on the table

But the DAC is not starting with a blank slate. It is stuck between the precedents set by the sovereign lending rules and the regulated rates of the ECAs. The further it moves in one direction, the worse it gets in another. The current deal on the table—and it should be stressed that negotiations are continuing and the deal may change rapidly—takes the IMF base rate of 5 percent, adds country risk premia of 4 percent, 2 percent, and 1 percent for low-, lower-middle, and upper-middle income countries respectively and then adds a “private surcharge” based on the tenor of the loan (maturity above or below 10 years) and the riskiness of the country. For the purposes of this surcharge, riskiness is sorted by ECA country risk categories, whereas the country risk premium is based on country income category, which seems like an unsatisfactory muddle. For short-term less-risky loans, this surcharge subtracts 2 percent from the discount rate (and adds 2 percent to long-term riskier loans). One set of countries see this as bringing things more in line with market benchmarks. The three holdouts find it absolutely unacceptable.

Table 1. The deal

Tenor

The DAC reckons the average loan to the private sector is around 10 years whereas the average official sector loan is closer to 20 years. The holdouts argue that the grant element calculation already accounts for tenor, but that is only true in the sense that the timing of repayments is part of a net present value calculation. One might also adjust the risk-free element of the discount rate with reference to the yield curve (i.e., the rates on US Treasury bonds of different maturities). For administrative purposes, it might be acceptable to approximate that with a step-change in discount rates at a 10-year maturity threshold, but the proposed 2 percent step looks too large because the yield curve is currently quite flat: there is less than a 0.5 percentage point difference between 10 and 20 year bonds (although that could change). It is not a very satisfactory solution however; a discontinuity in the quantity of recorded aid as a loan moves from 9 to 10 years is likely to produce a clustering of loans at 10 years.

Thresholds

There is also a minimum concessionality threshold of 5 percent for loans (and 1 percent for guarantees, zero for equity). The holdouts do not like that either. With the right discount rate and necessary safeguards to avoid competition issues, there would be no need to set concessionality thresholds for ODA eligibility. If a loan is on close-to-market terms, then it would simply represent very little ODA. We want DFIs to minimise concessionality, to avoid wasting public money and creating rents for project promoters and other private investors. Setting a minimum concessionality threshold would create incentives to needlessly grant subsidies to gain ODA recognition. But if the discount rate is too high, then a minimum concessionality threshold is needed to avoid truly profitable lending counting as aid, and to avoid potential competition issues. Those countries pushing for a minimum concessionality threshold also think it will help create some clear blue water between what the ECAs do, and what gets counted as aid. 

Incentives

The holdout countries worry that without sufficient ODA recognition, DFIs with mandates to balance financial and development returns will not put enough money into the riskiest countries, where it is most needed. It is hard for outsiders to parse these arguments. If mobilizing private finance is so crucial to the success of the SDGs, what additional incentives do donors and DFIs need, other than the mandates that they already have? It is up to DFIs’ shareholders to give management and staff the right incentives to fulfil their missions. We don’t score spending on maternal health as ODA at 150 percent of cost to encourage donors to do it. DFIs are not motivated by ODA recognition; their shareholders are. Are we to believe that without the “right incentives” governments will not direct DFIs to invest in riskier projects in poorer countries? The bottom line is that ODA recognition rules should accurately reflect the true economic cost of investing in private enterprises, and talk of incentives does not change that.

Perhaps talk of incentives is motivated by the following: suppose the selected discount rate for a country is 8 percent, but some projects in that country are relatively low-risk and merit a rate of 5 percent, whereas others are riskier and should be discounted at 10 percent. In which case, one could argue an 8 percent discount rate does not give DFIs the right incentive to invest in high-risk projects with potentially greater development impact (and where their intervention is less likely to be merely crowding-out the private sector). This argument reveals that a single country-level discount rate cannot confer the right incentives at a project level. The solution is not to choose a 10 percent rate, which would make those low-risk projects even more appealing, but a sensible average. DFIs must be given the right incentives, by their shareholders, to manage a portfolio of investments that mixes risk such that in aggregate the discount rate is accurate. 

Safeguards and hidden motives

Some of the smaller countries, and the export credit agencies of the some of the larger countries, believe that the true motivation of some of the largest PSI providers is not a desire to report as much ODA as possible: it is a desire to provide de facto state aid to domestic exporters. The desire to remove concessionality thresholds and objections to ex-ante reporting requirements could be seen in that light. In public the OECD promotes the importance of protecting competitive markets from state interference, but, so some countries believe, the DAC is poised to agree rules that will open the door to state aid via DFIs. These suspicions will be hotly denied, but it is not hard to see how countries with large construction or energy firms that want to win business in competition with state-subsidised Chinese firms might want to level the playing field.

Other countries find concerns about competition frustrating—their DFIs rarely make investments that are used to fund capital goods imports that could blunder onto the regulated territory currently occupied by export credit agencies. They do not want their DFIs jumping through unnecessary hoops.

These issues have been deferred until phase two. Some countries want them removed altogether. On the table for phase two is the question of how to demonstrate additionality. There is a risk of conflating input measurement with impact assessment here. Traditional aid may also fail to be additional: perhaps the local government would have built that health clinic if donors had not. The DAC does not require that donors demonstrate additionality before they can score grants for building health clinics as aid. Nonetheless, some donor governments are more worried about additionality in this context, because crowding-out private investors is a greater risk than occasionally substituting for the local government. The right way to approach this question is not to ask how can we measure additionality, but rather to ask under what circumstances are we prepared to accept an investment is additional, knowing that we will sometimes be wrong. The DAC may agree eligibility guidelines based on geography and sector, but these will need to be drafted carefully. For example, an investment into a telecoms company in a middle-income country may look like it does not need public money, but perhaps the purpose of the investment is to fund expansion into countries or regions that the firm would not otherwise have undertaken. Some countries want to see ex-ante notification amongst participating institutions (i.e., not public), as practiced by ECAs, so that others can inspect the deal and potentially challenge its rationale. DFIs strongly object to such proposals, arguing their investments are much less standardised than those made by ECAs and they are not prepared to expose themselves to the risk of losing a deal after many months of work. But it may be possible to differentiate between investments, whether debt or equity, which may plausibly finance imports from the donor country, and investments where those concerns are not relevant, and apply different rules accordingly.

Equities: an ODA pump

The proposals for scoring equity investors as ODA must also be revised during the second phase of the negotiations. The current proposal to cap negative ODA when reporting under the instrument method must be scrapped. This cap applies when using an ex-post method, which is a modification of the existing system—there is also an ex-ante method under consideration, which we will get to later.

The existing system for equity works on a cash flow basis: positive ODA is recorded when an investment is made (cash out), and negative ODA when it is sold (cash in). The problem with this method is that successful investments create negative net ODA, yet we want DFIs to create successful businesses, so the incentives are all wrong. The proposed solution, however, replaces that incentives problem with a new one

The current proposal has two new elements: it will introduce discounting, and eliminate negative net ODA by capping it at zero. The introduction of discounting makes sense, and would mean ODA is calculated on a grant equivalent basis, in retrospect. As with loans, if the discount rate is too high, ODA will be inflated and profitable investments will count.^[3] Some countries argue the proposed rates are too high and will again raise competition concerns with the ECAs, if equity investments are used to finance imports.

The introduction of a cap on negative ODA is more serious; even with an appropriate discount rate it would allow DFIs to run profitable portfolios whilst reporting large amounts of ODA. Here is a highly simplified example, for a single cohort of transactions, purchased in 2010 and sold in 2018.

Table 2. The ODA pump

When some investments make a lot of money and other lose it, a cap would turn equity investing into an ODA pump. Without the cap, reported ODA for the first two profitable investments would be -13 and -17, instead of -10. This portfolio of investments has made a 10 percent nominal return over 8 years and a 10 percent discounted loss. Many DFIs make higher returns than that. The ex-post grant element would be 4 without the cap, but is 14 with the cap.

In reality, DFIs are continually buying and selling equity. In this example, if the 44 raised from selling equity is reinvested in the same year, ODA reported in 2018 would be 44-26=18. That is a lot of ODA from a growing fund recycling 44 of assets in a year, even if the discounted returns are negative (meaning the profits are not sufficient to compensate for the risk-adjusted cost of capital).

A government using the institutional method would report 40 for the first year of operation and then negative ODA at some point in the distant future when the DFI is wound up (perhaps governments will decide to cap negative net ODA at zero under the institutional method too). Donors must report flows under both the institutional and instrument methods, so that the DAC can check they are reasonably consistent. Some countries regard the institutional method as a scheme for inflating aid, but with this cap in place, the inconsistency is more likely to be the instrument method producing bigger numbers.

But the DAC is also considering an alternative ex-ante grant equivalent calculation, for the instrument method. This has the potential to be the best system. When a DFI puts money into a private equity fund, for example, the targeted rate of return could be used as the basis for transforming the investment into an expect flow of income and discounting it back, similarly to loans. A formula for creating a synthetic flow of income ex-ante could also be concocted for direct equity investments, based on expected returns. Some governments do not like this idea, because they do not want to divulge expected returns, even behind closed doors.

As with all such things, the devil is in the detail. But the attraction of an estimated ex-ante grant element calculation is this: if there is no ex-post correction once realized returns are known, then the incentives to make good investments would be preserved. There would be no ODA penalty for doing better than expected or bonus for doing worse. If DFIs get expected returns roughly right, then errors would wash out on average, so there would be no need to correct individual investments ex-post. Current thinking seems to be, however, that ex-ante methods always need ex-post correction. The DAC should reconsider how that is done. Without an ex-post correction there is a risk of donors systematically low-balling ex-ante expected returns to inflate ODA, but there may be alternatives to applying corrections to individual transactions. An ex-ante grant equivalent method would not introduce the asymmetry of capping profits but not losses, for ODA reporting purposes, so it would not turn DFIs into ODA pumps. 

What next?

There are no perfect solutions here. But hopes of achieving the SDGs rest on donors mobilising private finance—they need to get on with it.

DFIs themselves argue the idea of basing discount rates on country income categories makes little sense: there is far more variation in risk within countries, depending on the nature of the project and the sector, than there is across countries. This applies to official sector lending too—Cambodia is an LDC, but its state-owned electricity companies do not need to be borrowing on a 45 percent grant element. It is widely acknowledged that in some countries lending to a strong private enterprise can be less risky than lending to the public sector, so the idea that discount rates on private loans would sometimes be lower than those for official sector loans should not cause outrage. The holdouts should accept it, in the right circumstances.

If we were starting with a blank slate, we could think of more differentiated discount rates, based on sector and tenor. Or perhaps the best solution would be for the IMF to adjust its baseline discount rate more frequently, to reflect market conditions. Many of these problems would go away if the IMF rate was a couple of percentage points lower.

Negotiations are hard for outsiders to read, and perhaps the holdouts will fold. But the threat of no deal seems credible. So, let’s assume some movement on discount rates will be needed to reach an agreement. What is the best way of getting there?

One option could be to use the agreement for official sector loans, but consider something that allows certain deviations, to adjust discount rates down in those safer-than-the-sovereign sectors (either with or without some adjustment for tenor). This could replace that 2 percent reduction on shorter tenor loans that some countries find unacceptable. There is a precedent: the “sector understandings” used by the ECAs, when lending against aircraft or renewable energy projects for example, play an analogous role (although they do not change minimum premiums, but other terms). If necessary, a risky sector adjustment could be introduced too (where what is considered risky would change from place to place). Rather than try to pin down lists of high-risk country/sector pairs, DFIs could make a case for high-risk status and be policed by their peers.   

Table 3. Adjust for “safe sectors” (and tenor)

Another option could be to remove the currently proposed combination of a “country risk premium” based on country income, and a “private surcharge” based on ECA country risk classifications split into two groupings, 2-5 and 6-7, and replace these with a single private sector risk premium differentiated by ECA country risk classification (and possibly tenor). This would look like a distinct private sector regime, and remove any read-across to the rules for sovereign loans. For example, the current proposed discount rate for a short-tenor LMIC loan is 5 percent, comprising 5 percent base plus 2 percent country risk premium, minus 2 percent private surcharge for countries in risk categories 2-5. This could be replaced by the base rate plus a single differentiated private sector risk premium. The point of allowing for more fine-grained differentiation is to introduce scope to meet the holdout countries halfway—rather than completely remove the objectionable deduction. An adjustment for tenor could be layered on top, but the current differential of 2 percent looks excessive.  

Table 4. Adjust by country risk category (and tenor)

In theory, there is also scope to move the institutional method to an estimated ex-ante grant equivalent basis (although that boat might have sailed). The idea under consideration for computing an ex-ante grant element for equity investments under the instrument method could also be applied here. An estimated grant element for capital that governments put into a DFI could be calculated with reference to that institution’s targeted financial returns and a discount rate based on a weighted average of its targeted country allocations. If DFI runs different funds with different targeted returns, money put into those funds gets reported separately. These parameters can be chosen at the point of capitalisation and thus still allow for accurate budgeting. This should also ensure greater correspondence between the institutional and instrument methods. However, choosing an appropriate methodology would be difficult, and would require some arbitrary assumptions. Negotiations over that could take a long time.

Once a deal on discount rates is struck, we will know that some loans will be reported as ODA that really do not deserve to be, and that in the current low-interest rates environment, ODA figures will probably be somewhat inflated. But that is inevitable in a simplified administrative system that cannot calculate grant elements using parameters fine-tuned for each individual investment. Central banks are now talking about raising rates, so the problem of inflated aid may not be around for long. There may also be some investment in the private sector where the true cost is probably understated. If the overstatement is not enough to influence donors’ allocation decisions, there is not much harm in it—the aid sausage has more worrying unpalatable ingredients. And if some countries feel the rules allow private sector instruments to count as aid when they are really subsidies to domestic exporters, the next phase of negotiations should focus on qualitative safeguards to ensure that when a country is claiming ODA on finance which is plausibly funding imports from domestic firms, there is a defensible development rationale and the transaction plainly would not have been financed by the market.  

The DAC should also introduce more oversight of governments’ private sector operations, holding them to account against the blended finance principles that the DAC is hoping to announce at the October high-level meeting. This could include naming and shaming countries that are stacking their portfolios with low-risk investments that do not deserve to be discounted at the agreed rates.

Data and transparency

A compromise is desirable on the presumption that the magnitude of investments being counted as aid that ought not be will be relatively small, so that most reported PSI will be genuine aid, even if somewhat inflated. Most participants in these negotiations seem to believe that, although there does not appear to be any shared quantitative analysis. That presumption could be wrong. And as things stand, we will never know because a decision has been taken to keep some transaction-level data for PSIs confidential.

That may be a mistake. If the DAC members want to keep civil society on their side, and have a constructive conversation about how to make blended finance more effective rather than face accusations this is all a sham to disguise state aid, they should not act like they have something to hide. Data can be aggregated to a level that preserves commercial confidentiality, but it should still be possible to disclose useful information about what investments are being made in what places on what terms, on average, and discounted at what rates, by grouping transactions into appropriate “bins.” We want to observe how much ODA is being claimed for investments on close-to-market terms. Presumably the fear is that without sufficient understanding of the context, civil society would interpret some investments as not aid, whereas in fact the risks were greater than appreciated and they genuinely achieved something additional with development impact. But greater openness is probably the better response than secrecy. DFIs should publish the ex-ante investment case for each investment they make, and refer doubters to these. Those who want to see a system of ex-ante notification amongst DFIs and ECAs argue it would build trust that the integrity of ODA is being preserved. The public need to trust that too.