Undue credit: Are France, Germany, and Japan subverting the definition of aid?

Last April, former Development Assistance Committee chair Richard Manning penned a scorching alarum in the Financial Times. Unbeknownst to the public, donors were inflating their aid totals by including loans that would profit the donors if paid in full. And the DAC, whose job it is to validate and publish these tallies, had proved feckless in defending the principle that for a loan to count as aid, it should profit the recipient not the donor.

The OECD must put in place a definition of concessionality that reflects the real cost of capital and requires real fiscal effort. It is shocking that the OECD should publish official statistics that allow “different practices” on such a key issue and which make a mockery of its own requirement that loans are concessional in character. It is encouraging OECD finance ministries to get away with murder as they seek to massage reported aid upwards at minimum cost. If the OECD cannot do a professional job on this, the UN should take over the reporting for international aid flows.

Richard’s causticity would impress you even more if you, like me, had observed his conduct as chair at last October’s board meeting of 3ie, which was a masterpiece of understatement and diplomacy. I assume his jibe about professionalism was aimed not at the staff who support the DAC, who are quite sharp, but at the Committee per se, whose membership consists of donor representatives and whose collective positions tend toward least-common-denominator consensus and inertia.

So what’s the issue? Ever since the aid world began, foreign assistance has come in two main forms: grants and loans. 1 The DAC counts all grants as aid, provided they’re meant to support development. But it only counts loans if they are sufficiently generous or—to use the term of art—“concessional.” When Japan lends to Myanmar at 0.01% with a 10-year grace period and 30-year repayment period, that’s concessional enough to count uncontroversially as aid. But what if Japan lent at 5% or 10% or 20%? In recent years, a few donors, notably France and Germany, have pushed along that slippery slope, increasing their aid totals and riling their peers.

To judge whether a loan is concessional enough to be Official Development Assistance (ODA), the DAC tries to impose two criteria, one mathematical, one fuzzy. The deep problem behind the controversy over when a loan is aid is that the mathematical criterion is broken, forcing those concerned to rely on the fuzzy one. The mathematical one is based on the assumption that rich-country governments can borrow long-term capital at about 10% interest per annum. Maybe that was reasonable in 1972 when the rule was adopted but it is laughable now. Today, donors can borrow super-cheap then on-lend the money at a rate high enough to earn a profit but low enough to meet the mathematical definition of aid.

The result today is a stand-off within the DAC over whether such loans meet the fuzzy criterion, which is that the loans be “concessional in character.” As I’ll discuss below, France and Germany argue that their loans that look profitable on paper are not profitable once you factor in the chance that borrowers will default. In December 2012, the Committee asked its staff to look into the matter and report back. The goal is to forge a new consensus in 2015.

In my discussion below, I concede that the French and German argument for factoring in default risk has merit, but find it inconsistent with countries’ historical insistence that there loans be repaid in full.

But that issue should not be allowed to obscure the deeper problem of which it is a symptom. The set benchmark of 10% is simplistic and, as a result, out of date. Not even France and Germany defend it. Moreover, it is easily fixed; for just across the hall from the DAC, metaphorically speaking, is an office that has honed a formula that is much more sensible and has already won approval from all 29 DAC members save Iceland. The OECD Export Credit Arrangement is a gentlemen’s agreement aimed at limiting subsidy in the loans governments make to finance purchases of their nations’ exports, such as power plants and aircraft. Where the DAC uses 10%, the Arrangement introduces Differentiated Discount Rates (DDRs), which vary across currencies, maturities, and time.

I built a database (requires SQL Server; tables and graphs in this Excel file) to simulate the substitution of DDRs for 10%. Since the 2008 financial crisis (2009–12), DAC donors have disbursed $32.3 billion in “aid” loans—loans whose interest rates look low compared to the DAC’s 10% threshold—that yet fail the concessionality test when using the lower and more-realistic DDRs. Japan accounts for $17.3 billion of that sum; France, $9.1 billion; and Germany $4.9 billion. France and Germany’s pre-crisis figures are much lower, suggesting that the official aid tallies have recently gone out of kilter.

On the question of when loans are aid, the OECD’s development loan and export credit regimes diverged around 1987, when DDRs were invented. According to William Hynes and Simon Scott’s excellent history of the ODA concept, many have since suggested that the DAC catch up with the Arrangement and use benchmark interest rate that adjust to the times. Nothing has come of those suggestions. But today’s extremely low rates and the resulting spat over French and German loans have created an opportunity to overhaul DAC’s obsolete rule and restore an increment of credibility to this system for marshaling resources for the global public good.

More details

To explain my calculations and reasoning more fully, I’ll start by clarifying the DAC’s 10% rule. For a loan to be aid, its “grant element” must be at least 25% when using a benchmark interest rate of 10%/year. What does that mean? The DAC formulas (Annex 14) take a loan’s actual payment terms—grace period, interest rate, pattern of principal payments, years to repay—and compute how large a loan balance those payments could pay off if the loan actually charged 10% interest. The difference between that and the loan’s face value is the “grant element,” and it must be at least 25% of the face value. The 25% requirement assures that loans that have tiny grant elements don’t slip into the aid tallies, where they would be counted in full. As Richard put it at ODI in November, it puts “blue water” between loans that are aid and those that are not.

For example, suppose you were buying a house in the United States in 1990, when 30-year mortgage rates indeed were just about 10%. A mortgage broker of today, shell-shocked but still in business, wormholes back to 1990 to offer you a $100,000 mortgage at a typical 2014 rate : 4.5%. It works out that the payments on this 4.5% loan would, if poured instead into a 10% mortgage, suffice to pay off a balance of only $60,000. So: you’d get a $100,000 loan disbursement check and the repayment burden equivalent, in 1990, to only $60,000. That’s a $40,000 gift, for a grant element of 40%. So DAC would call the mortgage “aid.”

The formulas chugged here are Finance 101. But the numbers plugged into them—10% and 25%—are arbitrary. They are now the center of controversy, especially the 10%, since it is stationary in a world of oscillating interest rates.

To be fair, the DAC system has usually been reasonable, even conservative. In 1990, for example, the U.S. government had to pay 8.6%/year on money borrowed for 30 years. If it on-lent to Brazil at, say, 9.6%, giving it a 1% spread to cover administrative costs, it would have brushed too close the DAC’s 10% benchmark. Only by charging a lower rate and taking a loss could the U.S. have reported such a loan aid. But the financial crisis has flipped the picture by lowering interest rates. In 2011, France could borrow $111 million long-term at 3.5%, on-lend it to Colombia at 5.5%, keep the 2% spread, and call it aid. And it did. 2

The reporting of potentially profitable loans as aid sparked the controversy within DAC that Richard’s Financial Times piece brought into public view. France—along with Canada, Germany, and Spain—defend the practice not by citing the obsolete 10% rule, but by contending  that a judgment about whether a loan meets the “concessional in character” test should factor in the riskiness of loans like that to Colombia. NYU economist Aswath Damodaran estimates that private creditors charge Colombia an extra 2%/year (200 basis points) to compensate for the perceived risk of default. So the market is saying that, factoring in the possibility of default, France can expect to earn not 5.5% a year on that loan, but 3.5%, which happens to be exactly what it pays for borrowed funds. Subtract overhead—the cost of running l’Agence Francaise de Developpment (AFD)—and France can expect to lose money. In this view, Colombia is getting a good deal, and the loan is aid.

This argument is relevant not only for whether France’s loans satisfy the fuzzy “concessional in character” criterion, but also for how the mathematical criterion should be updated. A new formula that factored in default risk would count more loans as aid than one that did not.

The argument also has merit. The IMF’s Balance of Payments Manual, which instructs governments on tracking international financial flows, emphasizes that financial transactions should be valued in comparison to market prices (paragraph 93)—and market prices include risk premiums.

Having pondered the case for putting default risk in the definition of aid, I am inclined against it. Default certainly does raise the cost of lending for the lender. But the way the international aid system works today, default also raises the cost of lending for the borrower. And that is precisely why it is usually considered a bad idea to make high-interest loans to the poorest, riskiest countries, even though—factoring in the probability of default—they might make money on the deals. I am wary of any formula that would valorize high-rate loans to indigent nations.

To drill into this paradox, imagine an extreme case. Suppose France could choose between giving Senegal a grant of €50 million or a one-month loan of €100 million at 0% interest, which Senegal had a one-in-two chance of repaying. Going by the French/German/Canadian/Spanish argument, these two options would have equal economic cost on average for France and equal economic value to Senegal. In reality, if France made the loan and Senegal defaulted, the French government would turn from a beneficent donor to an insistent loan collector. One indicator of France’s stance is its practice of carrying loans made to developing countries on its books at face value, structurally denying the likely costs of default. As loan collector, France might halt further aid. Or it might enter the quagmire of “defensive lending,” haphazardly making new loans with which Senegal could repay old, raising fiscal uncertainty in Senegal. Any debt forgiveness would take years and, France would insist, be negotiated at the Paris Club, where all the rich-country creditors unite on one side of the table to face poor debtors appearing singly. Cue nightmare flashbacks of human chains of debt protesters around G-8 summits (chains=slavery=debt), books with titles like A Fate Worse than Debt, and haltingly constructed, avowedly one-time debt relief programs such as HIPC. (OK, also nice monographs with titles like Still Waiting for the Jubilee.)

If France’s AFD and Germany’s KfW more smoothly accepted default as part of the risky business of development—as former AFD chief economist Pierre Jacquet has proposed—then I would find their rationale for factoring default risk into the value of credit more persuasive. Default would become more truly a zero-sum game, with losses for the lender matched by benefits for the borrower. As things stand, France and Germany insist for domestic accounting purposes that loan losses do not incur, while insisting for international accounting purposes that they do.

Not that I endorse the 10-and-25 status quo. Not that anyone does. Today’s low interest rates have made obvious that a simple, static benchmark is ill-fit to a world in which the only economic constant is change.

How should the 10% rule be overhauled? It could be done a million ways, this being finance. Fortunately, and remarkably, every member of the DAC except Iceland has endorsed a reasonable alternative in a closely related context. It is the system of Differentiated Discount Rates (DDRs).

As I mentioned at the outset, the OECD distinguishes between credits for development and credits for exports. (It’s not clear that China does. Within the OECD, the conceptual split emerged in 1969.) The big problem in the world of export credits has been the arms race: to promote Caterpillar’s sales to a Malaysian construction firm, the U.S. government offers to finance the purchase at 5% interest; Japan swoops in with a 4% deal on behalf of Mitsubishi; the U.S. counter-offers, etc. Government budgets are strained and trade is distorted as sales go to the biggest subsidizers, not the most efficient producers. To slow the arms race, the majority of OECD members have incrementally fashioned that gentlemen’s agreement, the Arrangement on Officially Supported Export Credits. Among other things, the Export Credit Arrangement seeks to cap subsidies embedded in low interest rates and to clearly distinguish export credits from aid loans. Along the way, it sets forth a more sophisticated and sensible system for measuring grant elements, using the DDRs. DDRs are “differentiated” by currency as well as the term of the loan, with longer-term DDRs being slightly higher. They are computed as a government’s borrowing cost plus 1% for overhead, and updated every January. 3 They do not factor in the risk of default.

DDRs are available back to 1999, and I back-calculated them to 1984 using underlying data. In this graph, you can see they fluctuate with economic events. Meanwhile, 10% is still…10%.

Differentiated discount rates, 1984-2012, UK, US, Eurozone, Japan

Determining the implications of a switch from 10% to the generally lower and tougher DDR standard took some work and some assumptions, which I’ll need to document are documented here. For now, let’s look at results. The next graph shows totals disbursements of loans that have been counted as ODA, but would not be if the DAC benchmarked would-be aid loans against DDRs:

ODA loan disbursements not aid when benchmarking against lender's cost of funds

I see two stories in this graph. First, it turns out that Japan has been reporting billions of dollars in loans as aid since 1996 that would not be considered aid against the more-realistic DDRs. This perhaps should be unsurprising since Japan is the biggest aid lender in the group and has the lowest domestic interest rates, thus the toughest concessionality test. I was surprised, though, because French and German loans, not Japanese, have figured in the recent controversy.

Click for a map of where these loans go.
Click for a map of where these loans go.

The other story is the emergence since the global financial crisis of Spain, Germany, and France  as major dispensers of loans of the same type. For them, this is a break with the past.

The next graph is the same except that it nods to the argument that the risk of default should be factored in. The method for estimating the average cost of default again comes from the Export Credit Arrangement. Before applying the 25% test, I adjust grant elements upward based on how risky a borrower is perceived to be. 4 Again, I’ll skip the complexities and focus on results:

ODA loan disbursements not aid when benchmarking against borrower's cost of funds

Under this more lenient test, most of the Japanese loans that were rejected as aid above now pass, suggesting that a lot of their lending only barely fails the DDR test. But the majority of the French and German loans still do not make the cut. So the French and German standard, applied using formulas they have endorsed, rejects much of their loan portfolios as aid.

But, as I say, I favor the DDR standard, as in the first bar graph above.

One natural question is how the switch from 10% to DDRs would affect headline aid totals, which are Net Official Development Assistance as a share of GDP. As the “net” signals, DAC subtracts repayments on aid loans from aid totals. So if we banish certain loans from the kingdom of aid, we need to  adjust net ODA in two ways: remove disbursements on the excluded loans and add back repayments on them (if they were never aid loans, repayments on them should never have been removed from net ODA). For a given donor in a given year, the combined effect of these two adjustments can be positive, zero, or negative. Here’s what actually happens to the 2012 numbers:

Net ODA-GDP by donor, 2012, with and without DDR filtering

Most countries don’t do aid loans, and they are unaffected. The big loser is France, whose aid/GDP ratio falls from 0.45% to 0.38%. Japan gains under the change, going from 0.17% to 0.21% of GDP.

Here are France and Japan’s net ODA/GDP over time, with and without the DDR filter:

France--Net ODA-GDP with and without DDR filter

Japan--Net ODA-GDP with and without DDR filter

For France, there is little difference until the financial crisis. The tighter aid definition hurts Japan during the big-lending 1990s but now that those loans are coming due and the country’s lending has shrunk, the DDR sword cuts the other way.

Does the minimal, or even positive, impact of the DDR filter on net ODA mean that the loan concessionality debate is overblown? Not exactly. We need to distinguish between the statistical impact and the incentive impact of putting loans to the DDR test. The statistical impact is: the exclusion of disbursements of certain loans + the exclusion of repayments on certain loans. But only the first of those, disbursements, is easily influenced by current policy (unless donors want to give debt relief more routinely). That is why I think it is proper to focus on the impacts on disbursements, as in the bar graphs above. They are the domain of the incentive effect of a move to DDRs.

I hope in a future post to say more about how best to measure the quantity of aid. [Update: done.] Years ago, for instance, I tweaked net ODA to remove interest payments (I think they should be treated just like principal repayments) along with cancellation of old non-aid loans (that were were probably not going to be repaid much anyway). The DAC’s sharp 25% cut-off is arguably problematic since a loan with a grant element of 24.99% does not count as aid while a nearly identical one at 25.01% does–even though the loans are essentially identical. At least since 1998, economists have suggested that only the grant element of loans be counted. And the IMF has a new interest rate system that offers an alternative to DDRs.

But the DAC is governed by donors, so it is a political organization; and politics is the art of the possible. I think the best and most realistic way forward is, at a minimum, to retire 10% after 42 years of service and substitute the Differentiated Discount Rates that DAC donors have already endorsed.



  1. Perhaps a third form should be added: in-kind donations such as of grain. For this post, I’ll call them grants.  (back)
  2. Some information about the Colombia loan is in row 5 of the “Loan effort” tab here. More details are in the 2011 Creditor Reporting System data file.  (back)
  3. DDRs are, however, not differentiated within the EU. So there is one rate for euro loans even though Germany and Greece borrow at substantially different rates.  (back)
  4. The Arrangement defines borrower-specific Minimum Premium Rates (MPRs) which are expressed as up-front charges as a percentage of loan amount. Failure to charge less than the MPR is considered a subsidy. To factor in an MPR, I first compute the net present value of the debt servicing stream using the relevant DDR. Then I multiply it by 1–MPR and subtract it from the loan amount to get the grant element. Annex VIII of the Arrangement states the MPR formula.  (back)