Please consider also the uploaded document file, which includes additional important comments.
The definitions are only clear in parts:
Interest rate risk arising from non-trading book activities: We would like to propose removing the words “current or prospective” from the definition. In particular, the prudential requirement for measuring interest rate risk in the non-trading book on the basis of an institution’s economic value allows no more than an initial estimate of the current interest rate risk. The meaningfulness of simulations of future trends must be clearly called into question in the context of measuring EV. However, the present definition permits such a conclusion.
We would also like to propose deleting the words “or have an impact on IRRBB”. We believe there is a significant danger here that risks will be intermingled.
Option risk: Exercise scenarios that depend not on the market rate of interest, but on the personal situation of the customer, are not described here. If these are not classified under interest rate sensitive instruments with an impact on IRRBB, they should be added here. These options are not interest rate sensitive, but still impact interest rate risk.
Credit spread risk: Credit spread risk from non-trading book activities (CSRBB) is not a part of IRRBB but a standalone risk class. Therefore, it should not be regulated in the scope of the IRRBB guidelines. There are various ways to correctly model CSRBB, so institutions may analyse it separately from IRRBB or may model the two risk types together.
Moreover – as already stated in comments on the corresponding Basel consultation paper (BCBS 319/368) – defining CSRBB as a residual is not appropriate.
Overall, we believe that the newly added definition of CSRBB in the IRRBB guidelines goes too far. Even if the requirements for CSRBB are still very vague in the present document, they may well be spelled out in more detail in future versions. It should therefore be ensured that this definition is appropriate. We see a risk in the more sweeping definition that e.g. full fair value measurements of loans and the own credit spread risk will have to be included in EVE, or that margin risk from loans will fall under the definition of CSRBB for NII. In terms of substance, it would, for example, be appropriate to include CSRBB in EVE only if it results from exposures (especially securities) that are measured at fair value. The definition, however, should exclude the own credit spread risk. Spread effects from securities should only be relevant (if at all) in NII in the context of IRRBB if the securities were acquired for purposes of interest rate management (i.e. normally securities held by treasury).
As noted above, we wish to stress that CSRBB is not a part of IRRBB, but rather represents an independent risk class that, by definition, should not be mixed up with interest rate risk. This should follow clearly from the text.
As a general rule, provisions should be reflected in risk measurement and management. However, there are various methods for including provisions. One example is to include them in interest cash flow, as described in the Draft Guidelines. However, because expected credit losses have of course not yet materialised and there is still a possibility that outstanding instalments will still be received, it must also be possible to reflect expected credit losses in credit risk. Therefore, there should be no fixed requirements about when and where provisions should be included; rather, they must be included to reflect the individual institution’s management. Additionally, the procedure adopted should be reported to the supervisor so that the latter can perform an appropriate assessment of the reported figures for the calculated change in EVE following an interest rate shock.
The situation is different if the credit loss has already occurred and there is no longer any contractual claim for payment of interest. In this case, the entire defaulted interest-related transaction is no longer part of the total cash flow. The default risk of troubled interest rate products is measured and managed in counterparty risk.
Regardless of when and where provisions are included, there is also no hard-and-fast rule for the amount to be recognised. What is important is that there must not be any one-to-one recognition of accounting carrying amounts in order to avoid mixing business management and accounting rules. The cash flows expected by the bank must be recognised, regardless of the amounts actually written down. These may be both lower and higher than the write-downs (e.g. pull-to-par effect if a fair value is recognised, ECB provisioning rules, and similar). It is important to understand that write-downs must be appropriately reflected. In this respect, the qualification “net of provisions” should be moderated in respect of the amounts to be recognised and modified to read “including provisions on a reasonable basis”.
Moreover, clarification is requested as to whether the Guidelines refer to specific provisions only, or whether general provisions are being considered as well. General provisions would be difficult to allocate at a single exposure level within the IRRBB system. We would prefer an approach where the impact of provisions on cash flows can be calculated on an aggregated (e.g. portfolio) level.
There should also be a clarification that the requirement applies solely to the credit portfolio, and that IFRS provisioning or unwinding effects do not have to be included in the analysis.
Finally, for operational reasons we would like to propose that modelling of cash flows from NPLs should only be required when material and according to the proportionality principle. We would therefore appreciate a reference to proportionality in this paragraph.
Considering “all elements and expectations” in the present Draft EBA Guidelines goes well beyond the bounds of feasibility. Many requirements are – taken by themselves – generally understandable, but only the calculations that are actually material for each institution should need to be performed. This is not sufficiently reflected in the qualification “implement […] commensurate with […] exposure to IRRBB”.
We support the requirement for institutions to use their own methodologies to suit their own situation. However, this paragraph must be viewed critically especially from the perspective of smaller institutions because it exemplifies the large number of calculations that these Guidelines are now demanding from the institutions. We would therefore appreciate a reference to the principle of proportionality in this paragraph. As a general point, the added value of the many analyses must be called into question.
What is important is that the nature and number of the analyses must be guided by their ultimate relevance for management purposes.
There should be no requirement to capture basis risk for any institutions for which basis risk is insignificant due to its very low level compared with overall interest rate risk.
Paragraph 26(e) and (f):
The meaning of subparagraphs (e) and (f) is not clear.
We reject the amendments made in paragraph 30 for a number of reasons.
The statements (e.g. on dividend policy) refer to ICAAP in general and are not specific to an individual risk type. We rather believe that the link between IRRBB and the dividend policy should be avoided. The revision of the dividend policy cannot be considered as losses. Payment of dividends is not directly correlated with the level of net interest income.
The detailed requirements suggest that losses from movements in interest rates are particularly toxic compared with losses from other risk types. We cannot agree with this. Additionally, we regard the requirements as largely redundant. The precise difference between points a, b and d, e is not evident from practical experience. A particular problem in this context is that competent authorities (e.g. the ECB) and auditors generally require the institutions to demonstrate for each paragraph in the Guidelines how the requirements are implemented internally at the banks. This can be illustrated using the example of paragraph 30: we believe that it will be impossible to provide the required evidence for the new points cited in paragraph 30, in particular for NII risk in isolation. We strongly recommend reducing the text to its essence i.e. that institutions themselves are responsible for the allocation of Tier 2 capital taking account of a) the relative importance of interest income for their business model and b) the overall design of their ICAAP.
We think it is difficult to make a bank’s management body directly responsible for decisions that can often only be made in day-to-day business operations. The management body can only define the overall framework. A good example is paragraph 41(c), because major hedging initiatives are developed and decided in the course of day-to-day business operations. It is neither practicable to define a suitable framework for this in advance, nor is it possible to involve the management body in these decisions in a timely manner. We would like to propose not giving the management body any direct responsibility for business operations. Paragraph 41(c) should be deleted.
This paragraph is a good example of how the accounting and risk perspectives are mixed together. We would like to propose deleting the last sentence: “Institutions should also take into account the impact on profit and loss accounts of hedging interest rate derivatives whose effectiveness may be hampered under certain interest rate changes”. When a hedging strategy is developed, there is an assumption that it will continue to be effective in the future. It does not make any sense to simulate its future effectiveness for all interest rate scenarios because the future ineffectiveness of any hedge accounting relationship results from potential larger differences in the actual vs. the assumed interest rates, and making reasonable assumptions in this respect is not possible.
Section 4.3.4 c. (IT system and data quality):
As a general remark, the requirements for IT systems should not be identical for all institutions, but should only cover the relevant material risks for the institutions in question. For example, basis risk is usually not relevant at small German banks. For this reason, it is not necessary for the corresponding IT systems to be capable of modelling this risk. This should be clearly formulated in the Guidelines. This is another example for the inadequate consideration of proportionality in the Draft Guidelines.
Section 4.3.4 d. (Internal reporting):
We believe that the in-depth requirements for internal reporting are, overall, too detailed and too far-reaching. At a fundamental level, we wish to refer to BCBS 239, where the reporting obligations are already addressed. The principle underlying BCBS 239 is that the level of detail decreases as hierarchy increases. This concept has been completely lost in paragraph 67, for example. On the contrary: the requirements mean that the reports become bloated and hence lose their informative value (because of the necessity to report certain matters regularly to the management body although they cannot be used to generate any management triggers). A dramatic loss of acceptance of internal reporting to the management would be inevitable. Thus, the risk of mismanagement would be rather increased. Because of the complexity of the overall topic of IRRBB and the different business models, the institutions should have more freedom here to decide the form in which a breakdown of risks for the management body makes sense. The existing requirements in the Guidelines governing internal reporting were entirely adequate.
Some examples are listed in the following.
We believe that the requirement to break down information by currency is not appropriate.
This paragraph requires the results of model reviews and audits to be included in the (regular) reports. This should also be possible, for example, in the form of separate annual reports on model validation. We believe that the requirement to compare historical stress analyses against the current performance is (at least for earnings risk) unnecessarily time-consuming and not expedient. In contrast to economic value risk, earnings are only stressed over a certain period. In order to obtain meaningful backtesting, both the historical scenario and the corresponding balance sheet change would thus have to be perfectly modelled.
The calculations and reporting obligations imposed here are far too complex and comprehensive. Not only that many of the required analyses are not helpful for management purposes – the mass of information means that the management triggers from the suitable indicators will actually be lost. The examples show clearly that in the case of such detailed minimum requirements, there can be no mention of applying the proportionality principle. We would therefore like to propose rewording the passage “they should include at least the following” to “accordingly, they may include the following”.
We believe that the obligation to always report the proportion of interest rate derivatives and Level 3 instruments in the regular reports to the management body (regardless of the importance of these positions or how stable their proportion is) is not appropriate. Any separate treatment of derivatives in the banking book seems to misunderstand the impact of their role: hedging the IRRBB exposure. Institutions using derivatives to steer their day-to-day IRRBB exposure will possess a huge number of internal derivatives for steering purposes – the analysis of only one exposure type in the banking book does not show any risk in the banking book. The treatment of derivatives should therefore be aligned with other exposures in the banking book, such as mortgage loans. We do not understand the separate treatment of these hedging instruments for reporting purposes and would therefore like to propose deleting paragraph 67(e) in particular.
From a purely practical perspective, the requirement for reports to be understood easily contradicts the requirements of paragraphs 66 and 67, which automatically result in complex reporting.
It is important to understand the models used, their strengths and weaknesses, their consistency with other methodologies used and the assumptions and their consequences. However, adequate consideration should be given to the principle of proportionality especially with regards to the understanding of the analytics, including by permitting external certification.
The requirements for the components to be modelled are excessive as a whole. They are almost impossible to meet, even at large institutions with appropriate specialised departments. In addition to reducing the requirements imposed for all institutions, the principle of proportionality must be taken into account to a greater extent especially in this section.
We do not understand the reasons behind the reference to derivatives. The use of derivatives does not result in any interest rate risk that has to be measured separately. Interest rate risk is analysed as a whole. Derivatives are used to manage this risk. We would therefore like to propose deleting the clause “(ii) the IRRBB of their banking book interest rate derivatives”.
We presume that a “transparent methodology for the identification…” means a description and justification of the approach and would like to propose wording this accordingly.
We believe that institutions should use their own assumptions and methods that are suitable for the specific situation of the institution concerned. This is the only way to calculate a realistic risk – something that is not possible with a one-size-fits-all approach. We do not understand in this context how the SOTs, which cannot accurately capture the risk of the institutions, can be integrated as complementary measures in risk measurement. The use of two additional measures that are not meaningful to estimate the interest rate risk exposure does not constitute any added value for the institutions, but merely increases calculation effort and the complexity of decision-making (see also Question 4) and may lead to adverse management incentives hampering the appropriate internal modelling of interest rate risk.
Own measures that are appropriate for the risk to be measured but are not listed in Annex I should be allowed to replace the methods described in Annex I. Please therefore add this to the paragraph.
The descriptions of automatic and behavioural options are incorrect: we would like to propose limiting automatic options to caps and floors. Swaptions or prepayment options embedded in wholesale assets are very often determined by behavioural factors and should therefore not be mentioned in the context of automatic options.
Behavioural options are not merely limited to smaller retail customers. There are also wholesale customers who base their decisions on other factors than the mere interest rate change. This needs to be addressed in order to make good management decisions. The definition of behavioural options therefore needs to be widened.
Basis risk is not relevant for all institutions (see also Question 1). We are requesting the corresponding qualification of this requirement.
We believe that the old wording (“may supplement […], for instance:”) was sufficient. We would therefore like to propose restoring it in order to allow application of the proportionality principle.
The calculation of any second round effects is not capable of implementation. The wording should at least be qualified (e.g. “major second round effects”).
The future hedge accounting ineffectiveness is by definition difficult to model as only the expectation of an effective hedge relationship allows for future hedge accounting. The effectiveness of hedge accounting should be left outside the simulation. For this reason, we would like to propose deleting paragraph 102(f).
It is not practicable to model all the parameters cited. Rather, it should be possible to select one or more parameters with a significant influence that will be used in the modelling. The reflection of all behavioural options is not practicable and not target-aimed. The obligation should be restricted to essential options.
We presume that 104(b) does not mean that there is a requirement to use elasticities, but rather that the term “elasticity” is being used to refer to the speed of the adjustment of product rates. The weaknesses of elasticities (e.g. insufficient tradability) are known, which is why many institutions use another model here. The use of different models in an institution leads to inconsistencies in management. A single model for quantification is sufficient. What matters is the quality of the model. We therefore suggest retaining the original wording (“speed/elasticity”) or replacing the word “elasticity” by “speed”, which is valid across models.
It should be clarified here that it may also happen that a customer may pay at the contractual date, but e.g. only part of the outstanding amount (e.g. 30%). In this case, too, the transaction would be managed as a behavioural transaction.
Overall, the requirements far exceed the requirements of BCBS 368. In particular interest rate risk is mixed together with liquidity risk. We are requesting that this paragraph be modified accordingly.
It is not evident why inflows and outflows to and from individual accounts should have anything to do with the institution’s interest rate policy for a product. Given an overall constant product volume and a constant rate of interest, in an extreme case 100% of the balance could be withdrawn daily by customers and reinvested by other customers without having any effect on the interest rate risk of the product in question.
See Question 8.
We are requesting clarification that only realistic stress scenarios for interest rate risk, and not for example – as required in paragraph 106(g) – an overnight outflow of demand deposits have to be modelled. That would be neither realistic, nor would it primarily be interest rate risk.
No, because there is no connection between the contractual terms and the behavioural assumptions for NMDs.
Whereas contractual terms are an unrealistic case of a right that theoretically applies to all customers, the behavioural assumptions primarily involve modelling repricing using a trade executed on the money and capital market as precisely as possible. On the one hand, the margin remains a variable that is as constant as possible. On the other, interest rate risk from this product is optimally eliminated.
A comparison of both variables (contractual term and model) is thus not necessary, nor does it lead to any understanding in terms of interest rate risk. In light of the overall expected additional effort, all unnecessary calculations should be avoided.
The SOT systematically disadvantages institutions that have created a high level of assets in their interest rate book thanks to a stable investment policy. As a rule, these assets do not lead to a 1:1 increase in capital. The higher the hidden reserves generated by the investments, the greater the interest rate coefficient of the SOT will be. This means that particularly stable banks (with a high net asset value) are especially impacted. As a result, the outlier criterion cannot serve as an objective measure for assessing IRRBB in a cross-country comparison. This indicator does not become more meaningful if the numerator is based on six scenarios that all represent extreme stress cases and the denominator is still an accounting measure (whether own funds or Tier 1 capital). This does not change the effect described. If these indicators continue to be used to classify interest rate risk, no automatic mechanism can be based on them from which an own funds requirement can be derived.
In addition, we wish to point out that requiring two SOTs anticipates the European legislative process. The EBA does not have a mandate to do this, in particular by means of a guideline. We are therefore not in favour of the coexistence of two prudential indicators.
We wish to point out again that this results in the absurd situation that, in addition to the required basically reasonable institution-specific modelling exercises, two SOTs also have to be calculated from which – as explained above – no meaningful management triggers can be derived for the institution due to their design.
Specifically we are opposed to the following requirements:
We believe that the requirement for behavioural assumptions that depend on the interest rate scenario is not appropriate – after all, behavioural assumptions are independent of interest rates. As an example, we wish to point out that, especially for the overnight shock modelled for the SOT, there cannot be any change in behaviour. Even if there were a slight dependence on interest rates, a direct change in behaviour would be extremely unrealistic. In addition, this requirement blatantly contradicts the general demand for comparability.
What is meant by “repricing” here? Principal is automatically remeasured when the interest rate risk cash flow is remeasured at overall bank level following the simulated interest rate shock. After all, this is precisely the idea behind determining a change in EVE.
We would welcome a situation where not only discounting using a risk-free yield curve per currency were to be permitted, but also discounting using several yield curves per currency. We believe that it is appropriate for institutions – as before – to be able to use the yield curves they use for their internal management for the SOT as well. We wish to draw attention to the future disclosure requirements at this point. It will be difficult to explain the differences between internal and external calculations resulting from this requirement.
The five-year cap for NMDs required for the EU represents clear gold plating compared with the Basel requirements. We therefore continue to forcefully reject this requirement. The stiffer requirement added to this Draft excluding financial institutions from behavioural deposits is based on the assumption that financial institutions always act as perfect economic agents. It has been repeatedly shown in practice that this is not the case. Insurers and pension funds in particular do not always make their investment decisions merely on the basis of interest rate changes. We therefore clearly reject this stiffer requirement.
Additionally, we presume that the cap will be calculated as a volume-weighted average for all liabilities. We would ask you to clarify this.
We would like to emphasise that the information contained in the numerator of the SOTs can only be comparable if there are no standardised requirements for modelling NMDs. Overly restrictive requirements produce meaningless results, especially in the case of interest rate risk (diversity of business models, heterogeneous customer behaviour in Europe due to cultural and legal factors).
We would like to declare a very clear “No” at this point. The reason is that the prudential proportionality principle means not only marginal relief due to reduced regulatory reporting obligations, but also qualitative and quantitative relief overall. This is also especially the case if the indicators to be captured do not provide any objective added value for supervisory monitoring. A single meaningful economic indicator such as value at risk is much better and also adequate compared with various indicators or measures that – already because of their mere quantity – do not have any relevance for management.
It is generally correct that pension obligations generate future cash flows (expenditures). The question that needs to be asked, however, is whether the present value (the current price of the obligation) is materially affected by a change in interest rates. This will certainly be the case if the amount and term of the pension obligation are fixed in the same way as a fixed-rate transaction. For example, if the pension agreement has a term of 10 years and the amount is already fixed today, the pension payment will thus neither be continued after the end of those 10 years without a renegotiation, nor will the payment be terminated early due to the death of the beneficiary. Such agreements are uncommon in Germany.
Rather, the pension obligation is normally linked to general wage trends/inflation and/or other economic variables. Additionally, the longevity of the recipient or their spouse or partner is another factor.
This shows clearly that the primary factor determining the value is not only the interest rate, but that other factors have a much stronger impact on the economic value and the risk associated with it.
From a management perspective, it seems questionable whether a fixed-rate investment based on the estimated pensionable age of the beneficiaries really makes sense. If this investment were to be made at today’s low rates, and inflation were to emerge over time due to pay increases, a risk would have arisen for the bank.
It would make more sense to secure the pension entitlements to match the bank’s long-term investment strategy, i.e. by making the investment not only in interest-bearing securities, but also in real estate (funds) and equities.
As a consequence, from a management perspective, risk stemming from pension obligations should be included in the IRRBB-framework at most with its interest-driven part or should be considered as a whole in a separate position.
The current practice varies. Both methods are used by German institutions - in line with the respective internal management system. Both approaches are consistent and thus most correct". EBA should, in accordance with the requirements of BCBS368, allow the equal application of both procedures. From our point of view, this would be the only way to counteract the ever-widening split of reporting and internal management."
This question cannot be answered unequivocally for the German banking market. Most significant institutions have already completed this separation in their internal systems. However, the systems of the majority of less significant or even medium-sized institutions are not able to exhibit this flexibility. The IT costs for those institutions would be extremely high because the corresponding adaptations to the upstream systems would have to be made. There is also no unequivocal answer to the question of whether the calculation is better with or without margins from a management perspective. We are therefore very clearly against the proposal to exclude margins from the prudential calculation mandatorily for all banks.
We understand the reason for specifying an interest rate floor, but wish to draw attention to one aspect. The drawback of applying a post-shock interest rate floor can be illustrated by the following example:
If the interest rate before the shock is lower than the floor and the scenario shifts it lower, the described logic means that first the shock and then the floor would be used. In this example, that means that there is a de facto upward shift. In order to prevent misunderstandings and to ensure a proper and timely implementation we are therefore asking for clarification of this requirement.
As the floor, we propose using the lower of the interest rate existing at the observation date and the rate following from the definition above. For example, if the interest rate is –100 bp and the floor is -90 bp, it will not be shifted any lower by the downward shock, but also not any higher. This approach is already applied in practice by the German supervisor.
It is our understanding that only currencies that exceed a materiality threshold of 5% are included. However, the reference variable for the materiality threshold is unclear. Potential values here include the nominal value or the market value, for example. However, with respect to derivative positions, the nominal would be inadequate and the market value preferable.
This approach is not appropriate because the interactions between currencies and interest rates are not adequately taken into consideration. Merely showing negative EVE changes is arbitrary and extremely conservative. It does not reflect the bank’s risk position and, for example, punishes banks whose FX funding is fully hedged by FX derivatives. The objective should be a scenario-specific analysis. We are therefore requesting the modification of the requirements.