It is UK Finance’s view that there are a number of definitions that require further clarity and a few terms where it would be helpful if a formal definition was given. The key concerns are:
It is not clear what is meant by the term automatic option. For example, are they explicit option contracts/products where they are indeed exercised automatically, or is this term inherited from the Basel papers that assumed any non-retail implicit option (e.g. a loan commitment) that would be exercised automatically?
Constant balance sheet/constant margins
The definition of constant balance sheet does not reflect best practice and should be amended. The current definition states:
“A balance sheet in which the total balance sheet size and composition are maintained by assuming like-for-like replacement of assets and liabilities as they run off.”
This definition is very rigid, and more flexibility should be given for banks to ensure the balance sheet remains constant in a more realistic fashion. For example, banks may desire to replicate the current term structure of new business rather than the term structure of the back book. An example would be if a bank is now issuing 80% of its fixed rate mortgages in the 2-year tenor it may want to model 80% of maturing 5-year mortgages into that tenor.
There should also be more clarity about whether product margins should be held constant or reflect current bank pricing. If products are assumed to be replaced at the same margin this is likely to lead to a disconnect between the results in the stress prescribed by the regulator compared to banks actual measurement of interest rate risk for management purposes and could be one of the main drivers of the outcomes.
Conditional/unconditional cash flow modelling
Conditional and unconditional cash flow modelling needs to be clarified. Is this referring to product caps and floors and taking into account a number of rate paths?
Short & medium term
It is not clear what time period is meant by short and medium term. This is particularly important for earnings metrics which are particularly sensitive to the time horizon that they are measured over. Further clarity, if not strict definition, of what is meant by short/medium/long term (paragraph 14) would be helpful. For instance, is short term considered up to one or two years? Is medium term considered up to three or five years?
Credit spread risk from non-trading book activities (CSRBB)
It is UK Finance’s view that the definition of CSRBB requires further clarification both in terms of what it is measuring and what parts of the banking book it should be applied to.
The current definition of CSRBB as “Any kind of spread risk of interest rate sensitive instruments that is not IRRBB or credit risk” is very broad and captures items that it is likely not intended to, such as product margins. UK Finance of the opinion that a positive definition of what CSRBB should capture would be less ambiguous and avoid confusion.
UK Finance would also question why CSRBB is being captured as a component of an interest rate risk in the banking book when the definition itself indicates that this is not a type of IRRBB. If there is a regulatory desire to capture CSRBB would it be more appropriate to do so in an individual standard rather than as a subset of an unrelated risk as is happening here.
In addition, any impact and overlap from other parts of the prudential framework need to be considered and understood, including IFRS9 overlaps.
Our recommendation is that the EBA should conceptualize CSRBB as per the definition presented in Annex 1, section 1.3, figure 1 of the BCBS revised standards (see Appendix to this response), whereby a market credit spread is inherent to traded instruments (i.e. bonds accounted for at Fair Value to P&L and Fair Value to Other Comprehensive Income, OCI) whereas the credit component for items at amortised cost is defined in terms of credit margin.
BCBS’s definition addresses price risk arising from credit spreads and its impact on capital ratios resulting from the Mark-to-Market of those instruments. Conversely credit losses from items at amortised cost are already captured by provisions and the Pillar 1 credit risk framework. Changing the scope defined by BCBS would result in double counting the impact of the risk.
It would be beneficial if the definition were to cover the specific instruments or possible sources of CSRBB such as:
1. The sensitivity of bonds or other instruments held for liquidity purposes which maybe fair valued using a specific curve which leads to value changes as a result of movement in the credit spread element of that curve.
2. The commercial margin of customer products which can be driven in part by underlying credit spread and the resulting sensitivity of that to changes in any spread, which may be applied for the purposes of calculating economic value sensitivity, for example.
3. The sensitivity of funding sources or other instruments which may be exposed to value sensitivity as a result of changes to an institution’s own credit spread.
We are of the opinion that CSRBB should extend to item 1 above only. A clear definition of this and ideally guidance on methodology would minimise the risk of double counting with other risk types such as Credit Risk or CVA Framework.
Basis risk is a broad term that can capture the variability in a number of different types of basis. For example, it can refer to interest rate indices, as the guidance identifies in some cases, whilst it could also be referring to the performance of a derivative as a hedge in comparison to the underlying risk that is being hedged. In our view, the definition of basis risk requires further clarity.
At times, the guidelines seem to suggest that banks will take positions simply in the expectation of future interest rate moves – e.g. paragraph 26 (b) and 42 (b). This gives the wrong message to both banks and supervisors. In practise mismatches in the banking book leading to open IRRBB positions arise for a number of reasons.
For example, they can arise because of operational time lags, holding positions in the reasonable expectation of offsetting customer flow, achieving best market execution and more structural positions that cannot readily be hedged except, possibly, by means of macro hedges. We consider that these are fully legitimate reasons for mismatches that do occur in the banking book whereas the guidelines should be amended to reflect this in order to send an unambiguous message to banks and regulators about the purpose and risk exposures in the banking book.
Paragraph 14 requires running a stress on instruments that are fair valued through OCI. In our opinion an EV type stress (for instance VaR) would appear more appropriate for this stress.
We agree banks should also consider the behaviour of assets once an impairment has been recognised. Banks will have different approaches to the treatment of such net exposures, some may treat them as a fixed, rather than variable rate assets. Either approach should be acceptable, subject to its challenge and justification in the Pillar 2 process.
Paragraph 26 suggests that capital adequacy assessments should incorporate a variety of factors. We feel the Guidelines would benefit from a more explicit or detailed definition for these factors to ensure they are understood and unambiguously interpreted. Particular points of discussion are:
26(b) - Further clarity on what is meant by the ‘effectiveness’ of a position would be beneficial. It is clear that the ‘cost’, or risk exposure, of an open position is something which should form part of an institution’s capital assessment. However, making an assessment of the effectiveness of an open position to take advantage of an expectation of rate levels i.e. was the open position successful, may well be something banks will do for internal performance measurement but is not something that would be expected to be included for a capital assessment, as long as the risk the position generated is captured.
26(f) - Embedded losses requires elaboration and definition. This should also cover embedded gains.
Embedded losses can arise within banking books as a result of different activities and in some instances, may only be realised as a result of change in management strategy or customer behaviour. Conversely, the same can also be said of embedded gains within the banking book. We would like to seek confirmation that any realised embedded losses under stress, expected to be included in capital assessment, should be net of any embedded gains that may also be realised under stress.
Some further points in section 4.2 that could be further clarified are:
1) Para 23 states that there is no expectation to double count internal capital for EV and earnings measures. However, in Para 24, both measures are expected to be taken into account;
2) Para 26(c) requires banks to factor in the sensitivity of the imperfect modelling assumptions. While we agree with the point, the wording could be improved to remove ambiguity. For example, it could simply state that banks should measure the sensitivity to key modelling assumptions.
3) Para 30 raises a number of issues that are wider than IRRBB and are potentially more relevant to stress testing. Additionally, certain expectations in this section are contradictory to the rest of the document. For example, measuring changes in NII resulting from liquidating positions, dividend policy and decrease in business operations contradicts the principle that earnings calculations are expected to be done on constant balance sheet with new positions raised on like for like basis.
Section 4.2 on capital is generally welcomed because they provide explicit confirmation that banks are expected to model their capital requirement using their own internal models and that an earnings, as well as an economic value approach should be used. There is also recognition that the risk of assumptions being incorrect is an important aspect to consider. Finally, the use throughout the section of the term economic value as opposed to EVE acknowledges the diversity of business models and hedging practices across the region.
However, it would be useful if there some guidance on the issue of on how a bank should combine the results of their EV and earnings approaches to arrive at an appropriate internal capital allocation. Depending on the business model of the bank these different measures can return very different results and regulatory guidance on addressing this discrepancy would ensure greater consistency of capital allocation across the industry.
Paragraph 30 and 31 state that capital should be held to protect against variability of earnings / reduced earnings. Our view is that any capital requirement for IRRBB should only be applicable where there is a risk of loss, not variability of P&L.
It would be helpful if there was an articulation of the purpose of holding additional capital in respect of an EV calculation. EV metrics capture the change in the market value of the portfolio under the specified rate shocks. In business-as-usual conditions this change in market value will never be fully realised as the position is normally held to maturity and pulls to par (i.e. has no P&L at maturity). It is only when this EV shock is accompanied by a change in behavioural assumptions, such as an increase in prepayment behaviour, or the bank fails, will this variability in income ever be realised. As the EV metric does not incorporate the existing mark-to-market of the portfolio it does not follow that an EV loss will result in a loss to the bank, it may only be a reduction in earnings. Accordingly, it does not seem appropriate to fully capitalise this metric.
Much is made of the need for the capital computation to consider limits and appetite as opposed to point in time data. However, there is an issue that the risk appetite might not be expressed and defined in the same way as the capital calculation and the two are not directly comparable.
Paragraph 42(b) - There is now a clear requirement for a separation between those functions monitoring and controlling IRRBB and what the document terms the ‘risk takers’. While we are fully supportive of this principle, the language is borrowed from the trading book and should be reworded. In most banks IRRBB is originated in the customer facing business units and then gets passed to the Treasury/ALM function. These risks are managed, as required. The Treasury/ALM functions do not generally ‘take’ risk actively.
Paragraph 44(e) - We believe that the regulatory guidance should not prescribe whether to include sub-limits on basis risk or optionality risk as long as these risks are measured, analysed and capitalised (if necessary). The Pillar 2 framework should continue to allow organisations to exercise some flexibility in the way they set their limit structure.
Paragraph 62 - The requirement to ensure that data (and, by implication, assumptions) used to measure earnings risk is consistent with that used for financial planning could be problematic, as the purposes of risk management and finance are different. For example, it is common risk management practice to use an interpretation of business re-pricing assumptions, rather than the actual business assumptions, to model IRRBB through the cycle, to better reflect the structural nature of the risk.
Core vs. transitory
Paragraph 106(a) suggests that the identification of ‘core’ transactional balances (i.e. current accounts) should be based primarily on analysis of the behaviour of individual customer accounts. However, consistent with a going concern approach, many banks model the aggregate balance including replenishment. The stability of balances on individual customers’ accounts is not particularly relevant – what matters is the total aggregate balance in the system, then, whether a particular bank can retain its current market share. An individual account based approach will underestimate the core balance and leave banks exposed to interest rate risk.
The draft EBA Guidelines are generally reflecting the importance of assumptions in the measurement of IRRBB and the greater understanding that IRRBB is not purely driven by rate changes. However, it is overly simplistic to draw significant conclusions from the impact on risk measurements from changing one of the modelling assumptions, such as the term of NMDs.
The EBA should also outline the context in which it envisages that all NMDs would reprice overnight and how this information should be used by institutions. Without such clarification the comparison of EV metrics calculated with and without behavioural modelling assumptions for NMDs might be causing more confusion about the regulators’ view of IRRBB. Using the example of modelling NMDs as overnight, this is not necessarily conservative as suggested in paragraph 105(c).
Although the regulators seem to be alluding to margins of conservatism, it should be emphasised that there is no risk neutral position in the banking book and assuming a very short repricing date of NMD would in fact increase NII volatility. This would not reflect real behaviour and create misalignment between the measurement and management of IRRBB.
Paragraph 105(c) - The EBA should appreciate that ‘conservatism’ cannot always be achieved. For investment of equity and NMDs, for example, there is no truly risk neutral position as there are dangers inherent both in investing too long and in investing too short.
Similarly, with regard to paragraph 106(d), there is no assumption that is necessarily prudent in all circumstances; For example, assuming all NMDs could re-price or be withdrawn overnight is clearly prudent in a liquidity context, but, not necessarily in an IRRBB context, as falling rates could damage a bank that had ‘prudently’ assumed a short re-pricing maturity.
Paragraph 108 should be expanded upon. Hedging of equity should be risk reducing from an earnings perspective, unless there is a real possibility that equity itself may decrease either through insolvency of the bank or a material reduction in the scale of its operations. If this is what the concern is, it should be clarified and other parts of the prudential framework would need to be taken into account, such as recovery and resolution.
We suggest that an assessment of the impact of the standard outlier test across the industry should be performed before it is implemented and the shocks set at a level calibrated by local regulator, rather than at a single level across all jurisdictions.
The draft Guidelines emphasise that the purpose of the standard outlier test is to aid supervisors in comparing banks and to identify possible outliers. However, it is not clear how the regulators are going to deal with exceptions to the 15% threshold. The EBA mentions that there should be no automatic supervisory measures from a breach of the threshold, but it is not clear if this is an interim approach that will be replaced with a list of potential actions or if this is a permanent approach.
The standard outlier test should remain a confidential supervisory tool as part of the ICAAP discussion and not be subject to Public Disclosure, which effectively is the result of the Basel standard. The inclusion of a specified standard result in the ICAAP appears similar to the inclusion of IRRBB within the Pillar 1 rather than Pillar 2 framework, as it raises the concern that the results will influence the setting of Pillar 2 capital. This is not necessarily in the spirit of the Pillar 2 framework which understands the significant differences between institutions’ balance sheets and therefore gives an element of freedom to determine required capital, subject to supervisory approval.
Para 113 (k) – The idea of a moving floor is conceptually understood, however, many institutions will find this difficult, if not impossible to implement. A large portion of institutions use third party vendor software such as QRM to model IRRBB. QRM currently only allows a user to attach a constant floor assumption to a yield curve rather than the period specific one as prescribed here. While this new sliding floor feature will be built in, it will require a version which institutions may have internal restrictions against upgrading to. As such, it is recommended to remove this sliding floor and to simply prescribe a single floor across the curve.
Para 113(m) – This would artificially increase the EVE sensitivity for any institutions with significant multi-currency balance sheets. Diversification benefits across currencies should be allowed as per institutions’ internal approaches subject to robust governance and documentation, otherwise comparability across institutions is lost.
Para 113(o) - We do not believe it is possible to define an appropriate maximum average duration for these positions given the different individual customer characteristics and behaviour in different institutions within the same jurisdiction, let alone across different countries. We believe institutions should use the same assumptions as used for the management of IRRBB internally and if customer behaviour and / or product characteristics result in a longer duration then that profile should be used.
In a resource constrained environment, supervisors should focus their attention on banks, and activities undertaken by banks, that could impact financial stability or put the deposit guarantee system at risk. Smaller banks are generally well capitalised, not highly leveraged and hold significant amounts of high quality liquid assets. So we support the inclusion of proportionality in the Guidelines and recognition that smaller banks may calculate and report SOT results less frequently than larger ones.
Further guidance on what constitutes ‘small trading book business’ (para 113 b, would be appreciated.
Our assumption is that the Guidelines apply at consolidated level and not at solo and sub-consolidated levels. Confirmation of this would be welcomed.
In our view, the EV metric should be seen as a tool for measuring cash-flow mismatches in the current balance sheet position; both alternatives are valid depending on the risk profile and framework of the institution.
Some institutions may require the analysis of EV with dynamic assumptions in order to make the most of this metric, with their risk frameworks, measurement techniques and infrastructure reflecting this accordingly. In those cases, excluding commercial margins may prove complex and result in the loss of valuable information required to understand their risk profile. This appears to be an area where one size does not fit all, giving banks the ability to apply either treatment appears a sensible approach.
In relation to the treatment of commercial margins specifically when discounting with a risk-free rate curve, conceptually, we believe it to be more appropriate to exclude commercial margins from the EV metric. At least in theory, this would provide a better alignment between the respective components of the numerator and denominator of the calculation. In practice it can be very difficult for institutions to achieve such purity in their approach given the subjectivity of defining exactly what constitutes a commercial margin, as well as the data and technological dependencies required to achieve this.
UK Finance member firms’ view is that generally there will be some level of cost for all institutions. This will vary by circumstance and scale of the institution, but there will be common areas of development required:
- Larger institutions may likely have more sophisticated risk management tools or systems, which may have the embedded capacity to model commercial margins separate to other cash-flows. However, given their scale, they may face challenges in terms of underlying data sourcing and architecture development. This would likely come with an associated cost which may become high considering the scale of balance sheet and systems which may need to be covered.
- Smaller institutions may have simpler data architecture and therefore possibly lower development costs. However, they may also not have the risk management tools or systems in place with the capacity to model commercial margin separately to other cash-flows, hence leading to development cost.
It is difficult to categorise the size of the cost across institutions, as what is high for one may not be for another; however, it is reasonable to surmise that for some institutions it may prove material to implement.
Apart from the fact that interest rate scenario floor specification entails some degree of subjectivity, we agree that regulators should be positioned to determine the most appropriate lower bound for interest rates in their jurisdictions. Our concern, however, is that global banks will be subject to different prescribed lower bounds from different regulators; IRRBB disclosures may end up losing comparability across geographies and would require devoting scarce IT resources to run additional regulatory scenarios.
We would like to propose the possibility of regulators authorising different floors for currencies out of their jurisdictions, when the bound is defined by a foreign regulator. We think that this measure will increase comparability across the industry and efficiency in reporting processes.
Our second concern is the implementation of a non-parallel floor. Although we acknowledge that there is some logic to phasing out the lower bound over a given horizon, its implementation may be subject to IT constraints depending on the functionality offered by different ALM software solutions. Given the proposed relatively ‘shallow’ gradient of the lower bound, we believe this introduces additional complexity to the calculation for an unknown (but possibly immaterial) impact to EVE results for many institutions, viz a viz a parallel floor.
As such we would like to propose the possibility of setting a parallel floor where this approach does not lead to material differences in the EVE results.