I have taken note with great interest of Euan Ritchie’s article “New DAC Rules on Debt Relief – A Poor Measure of Donor Effort”. He hits all the right spots with his criticism of OECD/DAC’s calculation of ODA regarding debt forgiven or otherwise reduced. However, I feel it would be good to complement his excellent analysis of how the new rules produce inflated figures with some additional commentary of a more general nature on the current approaches to statistics by the DAC.

What struck me most in the DAC agreement approved on 24 July 2020 (“Reporting on Debt Relief in the Grant Equivalent System”) was the “Narrative” presented in the Box on pages 2-3.

Right away, the use of the term “narrative” instead of the more traditional “rationale” suggests a deterioration of statistical culture at the DAC. It gives the impression of a marketing strategy needed because the proposed solution doesn’t withstand statistical scrutiny.

It also suggests a shift in the focus of the DAC’s audience. The old term “rationale” reflected an effort to show how new rules proposed or agreed were deemed consistent with underlying measurement concepts. As such, it was directed mainly at data reporters and users. In contrast, DAC’s narratives presenting the “new rules” are essentially sales pitches addressing non-expert “political” audiences, many of whom prefer flattering figures to sound figures.

In short, the use of “narrative” suggests a different underlying aim from DAC’s traditional rationales for statistical changes. DAC's main aim has shifted from helping informed readers understand the logic of the new scheme, to blurring the thinking of less expert audiences with plausible excuses for problematic decisions.

This impression that there has been a fundamental change in the DAC’s whole approach to statistics – turning them into a political and public relations instrument rather than trying to build an evidence base for policy-making – is further borne out by the bullet points in the narrative, as quoted and discussed below.

Bullet point 1:

It is in line with the 2014 HLM which called for "bearing in mind the past need to encourage debt relief initiatives such as HIPC and MDRI.

Already the first sentence confirms the shift towards understanding DAC Statistics as a political instrument rather than as a measurement system. The preceding introduction recalls that the DAC’s system now counts loans on a grant equivalent basis, so that they are valued upfront including their risk of default. It correctly recalls that the DAC undertook to change the debt relief rules so that this default risk would not be double-counted, and then asks the pertinent question “what does justify counting additional ODA for debt relief? But the first sentence of the first answer starts talking about the “ … need to encourage debt relief initiatives … “. This need is real enough, but it should never be satisfied with statistical measures: it needs to be addressed with political action. Statistics need to measure what is, whether the result is favourable or unfavourable, but should not have any role as an incentive or disincentive to policy action. The only role of statistical rules as an incentive is to encourage and facilitate measurement so that relevant facts become known and visible.

Bullet point 2:

Under the new grant equivalent system, ODA loans are required to conform with the IMF and World Bank debt policies. This strong requirement is expected to lead to more sustainable lending practices and less debt relief in future. However, it is also crucial to recognise that, under some circumstances, additional ODA effort may be necessary to make debt more sustainable. Loans and debt relief may be essential to bridge the financing gap to reach the SDGs; the 2030 Agenda recognizes that multiple sources of finance will be needed, not just grants. The DAC has duly recognised the key role of loans in development co-operation, and in mobilising resources to support the SDGs. The ODA modernisation aimed at better reflecting the development finance landscape, with the grant equivalent measure providing a more realistic comparison of loans and grants, and stronger incentives to use highly concessional loans.

“… ODA loans are required to conform with the IMF and World Bank debt policies. This strong requirement is expected to lead to more sustainable lending practices and less debt relief in future …
This again is put forward as a justification for counting extra ODA for debt relief even though the risk that relief would be required has already been factored into the ODA counted upfront on each loan. It impresses me as a typical “political placebo”, since there is nothing to convince me that the DAC Secretariat at the OECD will be in a position to check on its members’ compliance – and, by the way, there is no word on how this could be done in such an ultra-complex and constantly changing environment as the IMF and World Bank borrowing rules. In recent times, the Secretariat has been overwhelmed with elaborating and managing ever more complicated reporting schemes without being provided with sufficient resources to carry out comprehensive checks on conformity with rules. Not even in less complicated areas, where in-house data can be used for checks, has it been possible to carry out in-depth checks on members’ reporting. Now, I don’t blame Secretariat staff for this situation, they are doing a sterling job under the given circumstances. But I strongly disagree with the DAC’s new rules whose implementation is unfeasible both from a technical and a management perspective. Back in 2015, during the discussion on inclusion of private sector instruments (PSI) in ODA, Austria addressed this issue through official written comments, warning against agreeing a system that would no longer be manageable. But these valid concerns were not taken into account. The underlying phenomenon is a growing trend towards accommodating each and every political interest, no matter how much this harms the system, which actually no longer deserves the attribute “statistical”.

“… with the grant element measure providing a more realistic comparison of loans and grants …”
This claim for the new measure is untrue, given the current discount rates. These are used both to overstate “budgetary effort” through a generous risk margin applied upfront to generate loan's initial grant equivalents, and then again to generate additional "grant equivalents" up to the whole value of the loan when risk actually materialises. An outright grant in the same amount as a loan would be more “expensive” for the provider, yet as Euan’s analysis shows, a rescheduled or partially forgiven loan can now offer the same total ODA credit.

Bullet point 3:

The original grant equivalent is netted out; the method does not generate more ODA for a loan and subsequent debt relief than a standard grant would generate as a ceiling applies to avoid that situation. The ceiling will reduce the ODA amounts arising from forgiveness or rescheduling, in some cases substantially, including for the treatment of interests and late interests (for which there are no limits in the cash flow system).

This point concerns a “ceiling" that caps the reporting of relief on ODA loans so that total ODA reported on the loan before and after relief does not exceed the face value of the loan, which is the same as what would have been reported for an outright grant. The narrative’s claim that the ceiling means that the new system “does not generate more ODA for a loan and subsequent debt relief than a standard grant would generate …” is correct in itself, but diverts attention from the fact that relief on ODA loans is a small minority of total debt relief, and that the ceiling would only make a tiny difference to the amount reportable under the old rules. Thus the narrative deftly hides the fact that the ceiling will not be applied to debt relief of the much larger volume of non-ODA loans (former “other official” loans, export credits, and lending at market terms), where enormous ODA volumes are reaped from their forgiveness or rescheduling.

Bullet point 4:

The differentiated discount rates are a way to assess the degree of concessionality more precisely depending on the context, but the rates, applied ex-ante, do not cover for the “cost of default” ex-post, which entails additional donor effort. The risk of default cannot be assessed precisely at the time of commitment. The occurrence of debt relief means the context has changed in comparison with the original decision to extend the loan: ODA loans may have long maturities over which it is not possible to foresee all possibilities –conflicts or natural disasters may occur; other external factors such as volatility in resource prices or contingent liabilities unknown to the creditor may also lead to unforeseen difficulties for the borrower to repay its loan.

To me the wording of the narrative here is misleading, especially where it uses the term “differentiated discount rates” for the risk-adjusted discount rates decided by the DAC in 2014 to score the “grant equivalents" of ODA loans. The DAC rates have nothing in common with the well-known DDRs (Differentiated Discount Rates) used by the OECD Export Credit Community, which are differentiated by currency and tenor and updated annually, so that they reflect reality much more closely than the DAC’s discount rates. In discussions on this topic in the run-up to the DAC High Level Meeting in 2014, the DAC’s Working Party on Development Finance Statistics (WP-STAT) was very close to agreeing to recommend the use of the more realistic "real" DDRs as discount rates, but this proposal was stymied by political interests.

Bullet points 5 and 6:

The risk-adjusted discount rates agreed in 2014 will be regularly reviewed including to reflect evidence gathered on risk. See 2014 HLM Communiqué: “The discount rates and the grant element thresholds to be applied under the changes we are agreeing today will need to be regularly reviewed, reflecting changes in borrowing costs, emerging experience with risk (for example as reflected in default rates) and any need for further incentives for countries most in need.” Risk-adjusted grant equivalents are used on all ODA loans, and only a few loans will default incurring a real cost to the lender, hence the need for close monitoring of the implementation of the new method; the balance of the ODA amounts recorded ex-ante and ex-post can only be established based on real data.

In DAC statistics, the grant equivalent quantifies the concessionality of a loan (“distance” to the market). Through debt relief operations, donors modify the schedule of their loans by either forgiving or rescheduling the payments of principal and/or interests, thereby introducing additional concessionality in the loans. The method for accounting debt relief in a grant equivalent system aims at quantifying this additional concessionality by calculating the new grant equivalent of the loan, post treatment, and deducting the original grant equivalent.

“The risk-adjusted discount rates agreed in 2014 will be regularly reviewed including to reflect evidence gathered on risk.… the balance of the ODA amounts recorded ex-ante and ex-post can only be established based on real data.”
The narrative here tries to create the impression that the “real loss” on defaulting loans is likely to be greater than allowed for upfront by scoring all ODA loans in their risk-adjusted grant equivalents, or at least that there is still mystery about whether default risk has been fully priced in upfront. However, if you look at the entire portfolio of an ODA loans provider, it is more likely that the overall picture is already an over-estimation of ODA, since the DAC’s discount rates are generally far above market levels. In any case, data on the performance and default rates on ODA loans have been available for decades, and the Secretariat should have been tasked with an assessment, before agreeing these rules, of what percentage of ODA loans actually goes into arrears and finally receives debt relief treatment in donors’ ODA loan portfolios as a whole. Actually, a member of the OECD Export Credit Group, in a WP-STAT meeting several years ago, reported that the percentage of export credits actually going bad was “about 10%” versus about 90% that perform well. The percentage could be somewhat different for ODA loans, but this could easily have been verified by a Secretariat calculation before the DAC’s decision on the “new rules”. The result could have been very informative figures about the magnitude of overall losses or gains of ODA through the new measurement approach. 

The DAC’s promise of an ex-post check on the balance between grant equivalent ODA upfront and “real losses” ex post is eloquently silent about any downward adjustment of ODA that turns out to have been overstated upfront. The intention seems to be an ex-post correction only for undercounting and never for over-counting by the grant equivalent system.

In any case, none of the points in these two bullets addresses the real issue of double-counting risk by scoring it upfront on every loan, but then scoring new ODA whenever the risk materialises and an individual loan requires relief. At best, the points suggest that the discount rates may turn out to be too high or too low, given actual observed risk – but they provide no justification whatsoever for counting that risk twice.

Bullet point 7:

Not all loans default and need debt treatments. Original grant equivalents of loans that do not default still reflect the concessionality of these loans, as assessed against the agreed benchmarks (differentiated discount rates) at the time of their commitment. While the market will adjust interest rates downwards for a country originally assessed as highly risky but which eventually never defaults, the differentiated discount rates will remain unchanged, thus possibly overstating the concessionality of future ODA loans. Differentiated discount rates are approximate and represent averages for developing countries in the same income group, they may overstate concessionality in some cases, but also underestimate it in other cases. It will be important to regularly review them, as stated in the 2014 HLM Communiqué, to maintain their overall relevance. The Secretariat plans a first review in 2023 (five years after implementation of the grant equivalent system).

“It will be important to regularly review them [the discount rates], as stated in the 2014 HLM Communiqué, to maintain their overall relevance. The Secretariat plans a first review in 2023 (five years after implementation of the grant equivalent system)”.
While the real DDRs (used by the export credit community) are recalculated annually from actually observed rates and thus stick close to reality, the Secretariat plans a first review 9 (in words: nine!) years after having fixed the discount rates. Interestingly, the narrative speaks of five years, counting from the year of introduction of the grant equivalent system. But the DAC rates were determined in 2014 based on 2014 conditions and, moreover, grant equivalents have been calculated and reported as from 2015 anyway, they just have not entered the ODA/GNI calculation. This is another manipulative twist of the narrative, trying to make the revision interval not look all that bad. Announcing revisions of the DAC discount rates is another political placebo, while no real intention to approximate them to reality is recognisable. The DAC does not even promise to take any notice of the outcome of the Secretariat’s planned review! There is reason for doubt that the DAC actually realises how central the discount rate is to any grant equivalent calculation. They keep treating grant equivalents as valid and accurate measures of real effort, without acknowledging that you can get any number you like out of a grant equivalent calculation by just sliding the discount rate up or down. Credible grant equivalent accounts of ODA require that discount rates are kept under regular frequent review that ties them closely to market rates. Of this there is no sign.

Bullet point 8:

The Secretariat will monitor closely the implementation of the grant equivalent system. Separately, it will monitor the impact of the debt relief methodology including trends in ODA loans versus grants, with close attention to the risk of circumventing the rules (instead of debt relief, providers to give grants for borrowing countries to repay their loans).

Based on the experience with what didn’t work that well with monitoring ODA loan reporting in the past, there is little credibility in the promise made here about future monitoring and its implied impact on the accuracy and integrity of ODA statistics. Also, there is no hint as to the monitoring methodology. One can virtually guarantee that such an unspecific political promise will, when the point in time comes around, fail to deliver any serious results or impact. 

*

There is also another brief Narrative in a box on page 1. This concerns the treatment of non-ODA claims:

“As no ODA grant equivalent will have been recorded for the original claims, ….”
This statement is untrue, because in Associated Financing Packages, where specific export credits or other types of loans at market terms are softened by a grant paid to export credit/loan extending agencies to reduce their interest or charges, this will have been reported both as an ODA commitment and disbursement (see DAC Statistical Reporting Directives for reporting on Associated Financing). This has definitely been the case for Austria, and in our case the type of lending involved is financially supported export credits. But also other DAC members report ODA against the category of Associated Financing, so this is not a unique issue just for Austria. By ignoring this fact in the “new rules”, the DAC has introduced a second form of double-counting (in addition to double-counting risk). Since much the larger share in ODA volume can be expected from debt relief of non-ODA claims, this is an absolutely relevant issue.

*

Some general remarks on the underlying approaches to the “new rules”:

There are many other things I consider wrong with the new rules. For a start, they are incomplete, as they fail to address forms of debt relief other than forgiveness or reorganisation (e.g. debt swaps). Numerous flaws and inconsistencies in the case example calculations need examination, though this requires going into technical details that some audiences may not be willing or able to follow. Euan Ritchie has already embarked on this exercise in his article, more of which would be most welcome.

But what matters even more than filling the gaps or trying to set the calculations right is to highlight the wrong fundamental approaches in the new rules, which should not be accepted by the development community, and especially by those who monitor and analyse aid flows.

When I say wrong approaches I am not thinking so much about the conceptual flaws in the new scheme, though these are substantial, but more about its impact on the practical activity of statistical reporting. The new rules introduce a system the very nature of which means that DAC statistical reporters will no longer be able to check effectively on compliance or data plausibility. The rules are so complicated that they are thoroughly understood only by financial experts in this field, and calculations require data that are held exclusively by Finance Ministries and/or Export Credit Agencies and may not be passed on to third parties for confidentiality reasons. This means that ODA calculations in this area will depend on Finance Ministries and/or Export Credit Agencies. They will send their results to statistical reporters, most likely not disclosing any of the underlying data. Statistical reporters will no longer be able to fulfil their role of trying to verify data before reporting them to the DAC. In practice, we will see “any old numbers” being reported under the umbrella of debt relief. I am saying this as someone who has long-standing experience in Finance-Ministry-dependent practical reporting on contentious items such as ODA grants in Associated Financing packages, debt relief reporting of non-ODA claims and the like.

Debt relief reporting of non-ODA claims involves another feature that has not been addressed sufficiently so far and that leads to a vast overstatement of ODA in that full ODA credit is given for the balance-of-payments value of a debt relief operation that may not involve any budgetary effort at all. This striking anomaly in DAC’s ODA measurement was present in the “old system” and was expected to be corrected in new rules. For non-ODA lending, especially for export credits that are guaranteed by the official sector and for other officially guaranteed lending at market terms, it is common practice that the official sector collects (usually market-based) insurance premia from the private sector in compensation for taking over the risk of default. In case of default and eventual forgiveness, the accumulated receipts of premia are used to cover any indemnification claims by extending agencies, and, as I understand, the respective accounts are overflowing with resources, since it is only a minor share of credits that actually go bad. An Austrian case example of this was the 2005 Iraq debt forgiveness operation (through the Paris Club) that generated huge amounts of ODA credit, stretched out over several years, but that did not involve any budgetary effort. All losses were in effect covered by premia paid by the private sector, no transfers from official budget accounts for indemnification of the Austrian export credit agency were traced, in contrast to debt forgiveness cases of ODA claims where such indemnification payments were reflected in official budget accounts. I am convinced that this example is not an unusual case, but that it applies to the entire creditor community. Even now the 2005 to 2007 spikes in the curve showing all DAC ODA need to be explained as resulting from just two debt relief operations: Iraq and Nigeria. And the major share of this ODA is likely to have been gained without payment of a single euro/dollar from official budgets, since default was covered by private premia money. I remember discussions on this way back at the time of 2005 to 2007 statistical reporting, when I, as Austria’s Statistical Reporter, was incredulous that the DAC Statistical Reporting Directives should actually allow for this, and the DAC Secretariat replied that this was an anomaly that called for correction in a future overhaul of the Directives. The DAC, a body of political delegates with the exclusive power of defining statistical rules, has decided not to tackle the issue.

In conclusion, I feel sad that what I used to consider a “cultural heritage” – DAC’s statistical data on development finance – as I said in my farewell address as WP-STAT Chair at the end of 2013 – is deteriorating so quickly from an evidence base into a source for discretionary political story-telling.

Weiterführende Links

In der Rubrik "Kommentar der Anderen" bietet die AG Globale Verantwortung ExpertInnen die Möglichkeit, aktuelle und relevante entwicklungspolitische Themen zu kommentieren sowie ihre Meinung zu präsentieren. Das Ziel ist, Debatten über Entwicklungspolitik zu ermöglichen, den demokratischen Diskurs zu fördern und die Bedeutung der Umsetzung der Agenda 2030 hervorzuheben. Die inhaltliche Verantwortung für den Text liegt ausschließlich bei den AutorInnen. Die AG Globale Verantwortung teilt nicht notwendigerweise die vorgetragenen Ansichten.