An aid loan is not just a throw of the dice

After two years of communiqués, consultations, and commentary, the Development Assistance Committee has gotten down to brass tacks in fixing the definition of Official Development Assistance (ODA). I hear that a negotiating group composed of key donor representatives named for its chair, the UK’s Mark Lowcock, has reached a tentative deal. The Lowcock group’s role being unofficial, DAC chair Erik Solheim has turned its consensus into a proposal for discussion with the entire DAC membership.

I have not seen the document, so I cannot report definitively on it. It appears purely focussed on the issue that forced the discussions in the first place, which is when and how much to count loans as aid. I gather that the principle elements of the compromise are:

  • Counting only the grant element of loans, not the face value, toward ODA. (This CGD post of mine explains the concept.)
  • Dropping the fixed 10% discount rate in favor of an established benchmark system that will adjust as world interest rates rise and fall. I’ve heard cited both the “unified” rate of the IMF, currently at 5%, and the Differentiated Discount Rates of the OECD. The IMF rate appears favored.
  • Adjusting the new benchmark for risk, country by country. One formula floated would follow the World Bank’s tripartite division of aid recipients into low-income, lower-middle-income, and upper-middle-income. It would add 4% to the benchmark for LICs, 2% for LMICS, and 1% for UMICs, on the idea that wealthier countries are less likely to default and thus cheaper to lend to. (You know, like UMIC Argentina…)

I appreciate how this compromise would balance financial realism with simplicity. The IMF’s 5% + 1%/2%/4% is elementary as financial formulas go. Despite the arbitrariness in such a simple formula, in hitching the benchmark to actual interest rates and only counting the grant element, the compromise offers a big step forward in accuracy. And I understand that whatever deal is reached is just that: a deal, negotiated by a complex community of donors motivated by blends of principles and politics. It won’t be perfect in anyone’s eyes.

On the specifics, I say:

Thumbs up for just counting the grant element. With no originality, I advocated this change. This way, if two loans are for the same amount, the one charging less interest will count as more aid.

Likewise for dropping the hoary 10% benchmark rate. Few defend that either. However, I argue that the alternative the DAC now seems drawn to, the IMF unified rate, is inferior to the OECD’s own DDRs. It is not fit for purpose. The IMF uses its discount rate to project whether  a country’s current borrowing threatens financial stability in the long run. Since future economic conditions–in particular, future interest rates–are hard to predict, it makes sense to discount against a rate that changes slowly. Given the great uncertainties, it is not credible if a country’s IMF-perceived financial solidity jumps up or down every time rich-country bond markets move. This is why the IMF rate is based on the average U.S. medium-term interest rate over the last decade.

But to assess a donor’s financial “effort” in making a loan, all that matters is the donor’s current borrowing rate. If Japan makes a 40-year loan to Myanmar at 1% today, the net cost to the Japanese government depends on what it pays to borrow on the same terms today, not 10 years ago. This is why I favor the DDRs, which are averages over just 6 months.

It is probably more than coincidence that the DAC members prefer the higher of the two contending benchmark systems. Higher discount rates make for larger grant elements being counted as ODA. At the moment, DDRs are 2.9–3.5% for the dollar, euro, and pound, and 2.2–2.7% for the yen—all well below the IMF’s 5%. When one switches from a DDR of 3.5% to the IMF’s 5%, the grant element of a 30-year €10 million loan charging 2% interest rises from €1.6 million to €2.9 million.

But to the extent that donors favor the IMF rate system because it would produce bigger ODA totals today, they may be misguided. For when world interest rates rise, as they inevitably will, the inertia of the IMF formula will cause it to lag by many years. In contrast, DDRs will promptly adjust. Then the DDRs, not the IMF rate, will make the donors look better. Advocates for ODA increases might welcome such a turnabout, with the DAC imposing such a tough standard on itself.

But a formula cannot be ideal that systematically overstates donors’ efforts in some periods and systematically understates it in others—thus hardly ever getting it right. Better to use something like the DDRs, which more accurately reflect donors’ borrowing costs all the time.

Last, there is the question of whether the DAC should adjust discount rates upward for risk, the way financial markets do. (Ghana pays more to borrow than China does.)

I’ve written that the case for discounting for risk is intelligent and serious—and yet have sided against it. Why? Because an aid loan is not just a throw of the dice, not just a financial transaction that may or may not result in full repayment. Historically, when a borrowing government has fallen behind in servicing an aid loan, perhaps for some clearly identifiable reason such as a global recession or plunge in the price of coffee, donor governments have not accepted such default with aplomb. They have not gamely taken the losses and moved on the way investors in the stock market must.

Rather, generally as a matter of law, donors have become insistent loan collectors. In the 1980s and 1990s at least, they acceded to the evident need for permanent debt relief only slowly, a tardiness which harmed borrowers by casting a pall over them of uncertainty and permanent crisis. In the meantime, donors charged overdue penalties, and interest on the unpaid penalties, and penalties on the unpaid interest, etc…and then wrote off the unpaid sums as aid, even though not a penny of them ever reached developing countries. A loan with a 10% chance of repayment has radically different implications for development than a grant 10% the size. Yet in spirit the Lowcock Group’s compromise would equate them.

In sum, an aid loan is not just a throw of the dice that picks the winner and the loser. More like: overlending poses a risk to both sides. And once one considers the cost for the borrower as well as the lender, it begins to look strange to tilt the reward for lending in the direction of the the riskiest, most vulnerable countries. Multilateral banks charge the poorest countries less, not more. Discounting for risk sends the opposite signal: the poorer and less secure the country, the more interest it can be charged while still achieving a given ODA level.

If the newly suggested discounting formula had been in force in the 1980s and 1990s,the IMF base rate would have exceeded 11%, I estimate, to which 4% would have been added for the poorest countries, for a total of at least 15%. Compared to the DAC discounting regime then in force, 10%, this would have substantially increased the reward for increasing the debt of future Heavily Indebted Poor Countries (HIPCs).

So in my view, it is appropriate to discount aid loans for risk if the implied metaphor of a gamble is made apt. As I wrote,

One intriguing option is for the DAC to only discount for default risk for loans that eschew stiff penalties for default and contain automatic risk sharing mechanisms. For example, Guillaumont et al. (2007) and Cohen, Jacquet, and Reisen (2006, 2007) propose that instead of lending on soft terms to low-income countries, donors lend to them on hard terms and set aside funds to cover debt service when the borrowers experience certain price shocks. Any funds left over after full repayment might be disbursed as grants. The approach could be extended to capital account shocks and exchange rate shocks. The idea is to share risk while minimizing moral hazard by tying the triggers to external events rather than factors under the borrower’s control.

At the least, if loans are to be valued for risk ex ante then no debt relief ex post should count as ODA.

A smart and friendly French official argued with me this summer that it is not the job of the DAC to worry about overlending. That’s the IMF’s responsibility, which it discharges by regularly conducting debt sustainability analyses—at least for the poorest countries—and giving cooperating lenders red, yellow, or green lights. The DAC’s job, the argument goes, is to assess  donor effort only. Relevant is the risk that an aid loan poses to the donor, not the recipient.

This argument, like that for discounting for risk, is intelligent and serious. But it is not ironclad. If effort is all that matters, why not count the trillion-dollar American effort to accelerate democratic development in Iraq? In truth, the definition of ODA incorporates (consensus-based) value judgements about instruments, not just effort. These affect whether and how much various “efforts” count toward ODA. Not only does the U.S.-led invasion of Iraq not count, but only 6% of U.N.-led peacekeeping worldwide counts. It is therefore fully within the logic and tradition of ODA to value efforts in light of consensu views about potential consequences for recipients.

It might be argued more generally that donors have safeguards in place to prevent too much lending to poor countries. This may be so. But can we assume it will be so for all donors for the next several decades? The current discounting regime has been in place since 1972. The new regime could be in place just as long.

As they finalize their agreement, I hope the donors will maintain historical perspective. Opting for political expedients such as the presently higher IMF benchmark could constitute a failure to learn from the history that led to the current breakdown: conceptually incoherent choices can come back to bite you.