Official development assistance (ODA), the most commonly used measure of aid, is intended to assess donor effort. This is clearly valuable: an accurate measure of donor effort allows us to hold donors to their commitments, encourage them to do more for developing countries, and assess their priorities. However, the incentives this measurement creates also matter, as recognised by the OECD’s Development Assistance Committee (DAC), which sets the rules for measuring ODA: do the rules provide an incentive to direct resources to poorer countries who need them the most? If so, it could be argued that slightly less accuracy in measuring donor effort may be justifiable.
However, this blog argues that for rules on measuring ODA loans, the answer is no: if they have any incentive effect at all, the rules are more likely to incentivise lending to middle-income countries at near market rates, rather than truly concessional finance to countries that need it most. The DAC should consider whether alternative ways of measuring loans would strike a better balance between accurately measuring donor effort and incentivising lending to poorer countries, or if the latter is an appropriate objective for statistical rules.
Towards a measure of donor effort
The DAC’s “ODA modernisation” project aimed to make ODA a better measure of donor effort. As part of this move, the way loans are recorded has changed. Previously, loans that were sufficiently concessional were recorded at face value when disbursed, and repayments were subsequently subtracted (broadly speaking). Now, the “grant element” of a loan is recorded when it is disbursed (a measure of the subsidy on the loan) and subsequent payments are not subtracted (their present value is “built-in” to the grant element calculation).
The old method overstated donor effort in the first year, when loans counted the same as a grant, and understated it in subsequent years, when repayments reduced the value of ODA. Furthermore, interest was ignored, so a loan charging 3 percent scored the same amount of ODA as an otherwise identical loan charging no interest, despite the greater degree of sacrifice in the latter case. Along with a very low bar for concessionality, these features of ODA measurement were heavily criticised by researchers and civil society. The switch to the grant equivalent system addressed these criticisms. So higher interest loans now count for less aid than low interest loans, and yearly figures are more comparable to grants.
This change, however, also increased the stakes when choosing the “discount rate,” the benchmark rate used to assess concessionality. If the interest rate on a loan is lower than the discount rate, the loan is at least partly concessional, and the bigger the difference, the higher ODA is scored. While under the old system the discount rate was more generous—at 10 percent, compared to 9 percent, 7 percent, and 6 percent for less-developed countries (LDCs), lower-middle income countries (LMICs), and upper-middle income countries (UMICs) respectively—it was only used to assess whether a loan qualified as ODA. Once a loan was designated ODA, the discount rate had no impact on the actual value of ODA recorded. Now, since the grant equivalent is itself recorded, the discount rates have a direct bearing on ODA recorded, giving them an importance they didn’t have under the old system.
For the grant equivalent to accurately measure donor effort, the choice of discount rates needs to accurately capture the risk of lending. It makes sense to use higher discount rates when lending to riskier countries, given that there is a higher chance that future payments will not be received, so the present value of those future payments is lower. If donors lend at 3 percent to a country that can only borrow from private lenders at 10 percent, this is more concessional than if private lenders were willing to charge only 5 percent. But this risk is not well approximated by having one rate per income group (lumping together Venezuela and China for example). For discount rates to accurately account for risk (and so provide an accurate measure of donor effort) they would need to take into account more information, such as currency, duration of the loan, and previous borrowing behaviour, among other factors.
Not just about risk
But the risk of lending is not the only consideration. The DAC has emphasized the importance incentivising lending to LDCs. In the 2014 communique announcing the changes, and in the narrative to the new rules on counting debt relief as ODA, the discount rates are said to be “expected to incentivise lending on highly concessional terms to LDCs and LICs,” and were to be regularly reviewed, in case of “any need for further incentives for countries most in need.”
Incentivising lending to LDCs is a laudable aim: These countries sorely need additional resources to fight poverty, and loans are clearly an important part of the toolbox (debt sustainability issues aside). But aside from whether the rules regarding ODA measurement are the right tools for this job, do the rules actually produce these incentives?
Income and substitution effects
The DAC’s argument is that by allowing more ODA to be scored for lending to LDCs, donors will be enticed into lending more to those countries. This is analogous to the “substitution effect” in basic economics: when someone’s hourly rate increases, they may be inclined to work more hours because the payoff to doing so is higher. However, there may also be an effect analogous to the “income effect”: that same person may also choose to work fewer hours because the greater pay means they can afford to enjoy more leisure. When both effects are present, the overall incentive is ambiguous and depends on how much value the person places on each extra unit of pay and leisure.
In the case of ODA loans, this “income effect” is a real risk. Donors can record more ODA per transaction, so if they see benefit in recording large ODA figures, then the “payoff” to these transactions is higher. But this also means that the donors can afford to maintain ODA levels while doing less, which may be preferable. This outcome is particularly likely if donors have publicly committed to targets (like 0.7 percent), which creates discontinuities. Reaching the target could be politically important, but there may not be any additional political benefit to overshooting the target, and it could even be damaging—like in the case of the UK, where policymakers were loath to spend more ODA than the legally mandated floor of 0.7. These targets therefore increase the importance of the income effect, and so allowing lenders to score more ODA per transaction could mean fewer transactions as donors meet their targets more easily and reallocate scarce resources elsewhere.
This argument doesn’t apply just to ODA as a whole, but ODA to LDCs in particular, given the lesser-known Istanbul declaration which urges donors to allocate 0.15-0.2 percent of gross national income (GNI) to LDCs. Few countries achieve this percentage, but new rules allowing them to so while spending less may have the opposite of the intended effect. In addition, some donors are concerned with recipients’ capacity to absorb aid, and monitor recipients’ aid to GNI ratio. Recording more aid per loan clearly gives the impression that recipients are nearer to their absorption capacity than they really are. Allowing more ODA to be scored by way of a higher discount rate only incentivises greater lending if you ignore these effects.
What matters is not the discount rate itself, but its level relative to risk
Even ignoring these effects, what really matters for the incentives donors face is the level of the discount rates relative to risk. Whether or not the discount rates are supposed to accurately capture the risk in lending to different income groups, that risk is nevertheless different across these groups. If the discount rates are a fair reflection of risk, then there is no additional incentive to lend to LDCs, since donors may be able to score more ODA, but they’ll also lose more money.
What matters is the discount rate relative to actual risk. Looking at past data, it appears that the current discount rates are highly unlikely to incentivise more lending to LDCs. In constructing the discount rates for each borrower income group, the DAC starts with a base rate (5 percent) and then adds an “adjustment factor” for each income group (table 1) to account for the greater risk of lending to lower-income countries (paras 6, 14). Comparing these adjustment factors to the average percentage of principal that has been forgiven since 2003 (“actual risk” in table 1) shows that LDCs/LICs is the only group for which the adjustment factor is lower than the realised level of risk. This analysis suggests the new rules likely provide a disincentive for lending to LDCs/LICs.
|Table 1: DAC Risk adjustment factor, and actual risk*
|DAC Risk adjustment factor
*Actual risk measured as the percentage of outstanding loans forgiven, average between 2003 and 2018
Note: In reality, taking the risk margin as the full estimate of risk would imply that the base rate of 5 percent is the “risk-free” rate, which it is clearly not. Therefore, the actual risk implied by discount rates is higher.
Source: OECD Creditor Reporting System, author’s analysis
This same effect can be seen by examining the (commercial) market for China’s bonds. Based on their yield, every Chinese government bond with a maturity above two years could qualify as ODA on the concessionality test. What’s more, several ODA loans have higher interest rates than the yields on bonds with comparable maturities. So, far from being concessional, ODA loans are earning a better return than bonds on the market. There is no grant element on these loans in any real sense, but they score about two-thirds as much ODA as identical loans to Kenya or the Democratic Republic of Congo (depending on terms).
More generally, because the differentiation between countries is crude—only the income group is taken into account—the discount rates give the incentive to give all aid to the safest countries in that group. Bangladesh is set to do well, with a 10-year government bond yield well below its discount rate, suggesting that lending on commercial terms could count as ODA, but Angola isn’t. According to my colleagues examining models of global aid allocation, Bangladesh is already “over-aided” and Angola “under-aided,” suggesting this incentive is exactly the wrong way around.
Less accurate, worse incentives
ODA strives to be an objective measure of donor effort, with grants and loans measured on a comparable basis. This is important: the DAC peer reviews rely heavily on the value of ODA when assessing how well member countries are performing, researchers frequently make cross-donor comparisons with ODA figures, and cross-border ODA is used in assessments of what finance is available to developing countries. ODA will only measure donor effort accurately if the discount rates used in calculating the grant equivalent of loans do a good job of capturing risk, and this may or not coincide with discount rates that create the incentives to lend to LDCs. But the current discount rates don’t do either.
Instead, there is a risk that these discount rates incentivise donors to lend to middle-income countries at market rates, allowing them to meet aid targets while costing them little (or maybe even nothing). It would be useful for the DAC to publish any evidence to support the idea that the effects go in one direction: this would perhaps go a long way in reassuring those who are skeptical about the new rules. Otherwise, alternatives to the current arrangement have been suggested (such as using the “Differentiated Discount Rates” to measure concessionality instead) and these could be considered when the new rules come under review.
The author is grateful for valuable comments from Ian Mitchell and Ranil Dissanayake but remains responsible for any errors.