Response to consultation on draft Guidelines on IRRBB and CSRBB
Go back
The exclusion from the modelling of wholesale NMDs from financial customers is not appropriate and in no way reflects the reality to be modelled. Moreover, this approach is not consistent with the Basel standards.
Should para 119 nevertheless be retained, we would ask for the following points to be taken into account:
It is totally understandable that basis, gap and option risk should have to be measured. But it should be made clear that separate measurement of these three risk sub-categories is not necessary and that a lack of separate measurement will not lead to categorisation as “non-satisfactory”.
In para 119(b) it is not clear what is meant by “calibrated” and “reviewed”. Who is supposed to be reviewing what: risk controlling, internal audit, …? It would also make more sense to replace “back testing” with the broader term “validation”.
We welcome the fact that the proportionality principle has been taken into account in the process of identifying inappropriate methods and procedures. However, some of the explicit requirements would require small institutions and institutions with non-complex portfolios to expend a disproportionate amount of time and effort, so we assume that the principle of proportionality will apply here too. Examples are the consideration of basis and option risk, the extensive, detailed requirements for reviewing and validating internal models, and the requirement in para 112(h) to isolate the impact of behavioural assumptions by using contractual terms instead. The realisation of contractual maturities is a strong behavioural assumption in itself and not suitable as a point of reference. This requirement should be dropped for all institutions. The requirement in para 109(b) concerning the analysis of elasticity in behavioural options is also unsuitable for small institutions since price effects from elasticities can only be mapped with the help of complex derivatives (a EUR 100 deposit, 70% of whose interest rate is linked to a current interest rate, is no longer worth EUR 100 in the event of an interest rate adjustment. This problem can be circumvented by using moving averages, for example).
We see a need to specify the procedure for handling institutions when a competent authority is considering classifying their modelling as non-satisfactory. First of all, the institution in question should be given a reasonable period of time in which to remedy identified shortcomings and thus avoid this classification. If an institution needs to switch to the (simplified) standardised methodology, orderly arrangements should be in place for doing so and there should be an adequate transitional period so that the institution can implement the methodology properly and make the necessary technical adjustments to its systems. Conversely, there should be a clear and straightforward road map for reverting to a (satisfactory) internal system in order to avoid undermining the incentive to use internal systems. This applies to both the standardised and simplified standardised methodologies (i. e. it should not be necessary to go via the standardised methodology). We assume that once identified shortcomings have been remedied, it will be possible for an institution to revert to its internal system without undue delay.
Moreover, given that many banks in Germany do not prepare their accounts in accordance with IFRS, the requirements should be formulated in such a way that they make sense for all institutions. The text should be adjusted accordingly.
We agree that IRRBB from an NII perspective is an important risk that needs to be measured and reported. However, given that NII is part of the normative perspective, the individual limitation of this specific risk is not in line with general normative assumptions and underlying management actions. Like other factors in the normative perspective, NII should only be assessed periodically under defined overarching scenarios and analysed in the context of the entire normative requirement. We would therefore ask the EBA to align the definition of limits in the normative context with the new ICAAP requirements to allow banks to streamline their efforts and to prevent conflicting steering implications. A lack of internal limits in one perspective should not lead to a bank’s internal methods being classified as non-satisfactory.
In addition, we recommend streamlining the text of some requirements to avoid duplication (e. g. paras 154 vs 155 and 146 vs 149).
The impact of this change is unclear; equally unclear is whether a material change is intended at all. In our view, the risk of a change in market price implies the existence of (fluctuating) market prices for products in the relevant business segment. Market prices are generally not available in the area of conventional customer lending. This area consequently falls outside the scope of the definition. German banks, for example, finance a large number of small and medium-sized businesses (such as “the baker on the corner”) for whose financing no market prices are available and thus no credit spreads can be determined. This also applies to residential real estate finance. If meaningful market information is available (e. g. for the corporate bond portfolio), these positions should be included in the measurement of CRSBB. Specifically, this therefore applies to transactions for which credit spreads are relevant, especially securities transactions. It should also be possible to exclude certain positions from the risk assessment to avoid double counting. This will be the case, for instance, if it can be demonstrated that the corresponding risks are already assessed and managed elsewhere (e. g. in combined credit risk and credit spread risk models).
There is currently no industry standard for measuring CSRBB: on the contrary, approaches differ widely. No industry standard should be “enforced” in this area by the EBA. Requirements should be “modelling neutral”, which is not the case at present. A typical example of the use of trading book VaR models adapted for banking book purposes is the integrated measurement of IRRBB and CSRBB including idiosyncratic risk (with or without changes in individual creditworthiness). Alternatively, credit risk models can be extended to include components for CSRBB measurement which enable an integrated risk measurement on the basis of creditworthiness and market-induced spread changes. When such integrated models are used, there is a particular danger of the proposed procedure quickly leading to double counting, which must be avoided at all costs. Existing models are all well understood and have proved their worth; it should continue to be possible to use them.
According to para 157, however, it will only be possible to include idiosyncratic credit spread components in CSRBB measurement if it can be demonstrated that the results will be conservative. This alleged relief is nothing of the sort and the obligation to produce evidence should be deleted since, depending on the observed reporting date, inclusion may lead to either an increase or a reduction in risk as a result of possible diversification effects that the inclusion may generate. Since the consideration of additional elements of the overall credit spread risk tends to increase unavoidable double counting owing to overlaps with migration and default risk, results should be considered conservative per se and permitted without any requirement for verification. By contrast, a precise measurement of the individual components of credit spread risk (market spread, liquidity spread, separate consideration of idiosyncratic spread) may be desirable in theory but cannot be accurately carried out in practice. The explanatory box on page 46 includes a qualification to this effect. Even if separate consideration of the spread components were possible, the question arises as to whether this would offer any added value and make a meaningful contribution to a bank’s risk management practices. For example, the default premium might already be included in the credit risk and the liquidity premium in the liquidity risk of the institution. Anticipated rating changes, i. e. expected migration to other rating classes, may in turn have an impact on the default premium and possibly the liquidity premium.
We believe greater account should be taken of an institution’s individual circumstances (portfolio composition, design of internal systems, availability of reference data, etc.) when assessing credit spread risk. It should be up to the bank itself to decide on the specific design of its system, including whether to opt for integrated or isolated measurement. Analysing currency-specific dimensions, for example, as proposed in para 123, will not make good sense for every bank. For all larger portfolios, further clustering (such as by rating class, sector, region, product, or possibly also more granular, issuer-specific mapping) is required for internal risk management purposes and has thus become market standard. It should consequently be up to institutions to choose which type of clustering is most appropriate. This is the only way, in our view, to achieve meaningful results from which sensible risk management practices can be derived. On top of that, a generic and arbitrarily constructed curve will lead to inappropriate mapping of the credit spread risk of certain portfolios and thus to undesirable risk management incentives. Institutions should therefore be allowed to use bank individual appropriate credit spread curves with which observable market prices are also set for traded instruments.
Backtesting (realisation vs forecast) does not always provide meaningful results and is sometimes not possible at all. It is not possible, for example, to observe on the market the “actual result” of a loan to a client such as the baker on the corner (see, in particular, para 147). The question therefore arises as to what value should be backtested in such cases and what meaningful implications (in terms of risk management takeaways) should be derived from this theoretical exercise.
According to our understanding of the definition of CSRBB (cf. our reply to question 3), positions should, in addition, only be included in the scope of CSRBB measurement if meaningful market information is available about them. Reference prices are not sufficient in this respect: position-specific prices must be available. This means instruments must be eligible for trading on a stock exchange and have a certain minimum market liquidity. Prices from secondary markets, by contrast, are not a suitable means of measuring CSRBB (registered bonds traded on secondary markets with only issuer spreads would also have to be excluded, for instance). In summary, therefore, only liquid securities positions measured at fair value in the banking book should be included in the measurement. As we see it, this is the only way to achieve appropriate risk management incentives.
In particular, we strongly oppose the requirement in para 124 for institutions not to exclude certain positions ex ante but to provide detailed documented evidence of an “absence of sensitivity to credit spread risk” for each exclusion of a group of similar positions. This will not be feasible to implement in practice since, for items such as customer loans or demand deposits, evidence of this kind is virtually impossible to document in a way which is audit compliant. It would make more sense to assume that CSRBB generally plays no material role for certain types of products (with negative lists at least at national level). If neither idiosyncratic risk nor migration and default risk are taken into account, this assumption is likely to be largely correct. Even if credit spread changes were observable for these product types, they would play no role when it comes to conditions for either new or existing business (in a classic buy-and-hold business model). The changes would have no influence on P&L nor, in consequence, on equity and would ultimately have no effect on the bank’s solvency. Their measurement would therefore be irrelevant.
We also recommend that pension plans (assets and liabilities) be explicitly excluded from the scope. The pension plans of banks should be excluded from the scope of CSRBB because of the way they are uniquely governed and managed relative to the bank’s commercial activities. Pension plans, specifically defined benefit plans, contain credit-risky investments along with equities and other investments in order to hedge the credit-risky liabilities of the plan. These investments are managed by an investment manager, typically a third-party asset manager, and governed in many cases by a board of trustees who are either independent or semi-independent of the bank. In other words, the bank may not be able to legally apply these guidelines to those plans, especially considering the granularity of requirements envisioned by the revised CSRBB rules. Even where the bank is able to influence investment strategy, it takes a portfolio approach that looks at the overall risks of the plan across assets and liabilities and establishes broad investment guidelines for the investment manager to adhere to. We do not think the rules were intended for pension plans and ask that they be explicitly excluded from the scope of CSRBB.
Furthermore, pragmatic procedures should be established to enable individual institutions to exclude other non-material positions from the scope if necessary.
In special cases and depending on the bank’s business model, the com
Question 1: In the context of the measurement of the impact of IRRBB under internal systems, paragraph 111 envisages a five year cap repricing maturity for retail and non-financial wholesale deposits without a specified maturity. Would you foresee any unintended consequence or undesirable effect from this behavioural assumption in particular on certain business models or specific activities? If this is the case, please kindly provide concrete examples of it.
Provided that the cap is defined as a weighted average – as envisaged in para 111 – we see only limited potential for unintended side effects. We see no danger of results being significantly distorted. The reference to the weighted average should therefore definitely be retained. On several occasions in the paper (such as in this question) the wording is inconsistent and mentions a cap on the maturity itself rather than the average. The wording should therefore be adjusted accordingly.The exclusion from the modelling of wholesale NMDs from financial customers is not appropriate and in no way reflects the reality to be modelled. Moreover, this approach is not consistent with the Basel standards.
Question 2: Do respondents find that the criteria to identify non-satisfactory IRRBB internal models provide the minimum elements for supervisors’ assessment?
We support the idea of considering an internal measurement system (IMS) non-satisfactory if a case-by-case analysis concludes that the system does not materially comply with the guidelines. We see no need for further detailed assessment criteria and believe para 119 should therefore be deleted.Should para 119 nevertheless be retained, we would ask for the following points to be taken into account:
It is totally understandable that basis, gap and option risk should have to be measured. But it should be made clear that separate measurement of these three risk sub-categories is not necessary and that a lack of separate measurement will not lead to categorisation as “non-satisfactory”.
In para 119(b) it is not clear what is meant by “calibrated” and “reviewed”. Who is supposed to be reviewing what: risk controlling, internal audit, …? It would also make more sense to replace “back testing” with the broader term “validation”.
We welcome the fact that the proportionality principle has been taken into account in the process of identifying inappropriate methods and procedures. However, some of the explicit requirements would require small institutions and institutions with non-complex portfolios to expend a disproportionate amount of time and effort, so we assume that the principle of proportionality will apply here too. Examples are the consideration of basis and option risk, the extensive, detailed requirements for reviewing and validating internal models, and the requirement in para 112(h) to isolate the impact of behavioural assumptions by using contractual terms instead. The realisation of contractual maturities is a strong behavioural assumption in itself and not suitable as a point of reference. This requirement should be dropped for all institutions. The requirement in para 109(b) concerning the analysis of elasticity in behavioural options is also unsuitable for small institutions since price effects from elasticities can only be mapped with the help of complex derivatives (a EUR 100 deposit, 70% of whose interest rate is linked to a current interest rate, is no longer worth EUR 100 in the event of an interest rate adjustment. This problem can be circumvented by using moving averages, for example).
We see a need to specify the procedure for handling institutions when a competent authority is considering classifying their modelling as non-satisfactory. First of all, the institution in question should be given a reasonable period of time in which to remedy identified shortcomings and thus avoid this classification. If an institution needs to switch to the (simplified) standardised methodology, orderly arrangements should be in place for doing so and there should be an adequate transitional period so that the institution can implement the methodology properly and make the necessary technical adjustments to its systems. Conversely, there should be a clear and straightforward road map for reverting to a (satisfactory) internal system in order to avoid undermining the incentive to use internal systems. This applies to both the standardised and simplified standardised methodologies (i. e. it should not be necessary to go via the standardised methodology). We assume that once identified shortcomings have been remedied, it will be possible for an institution to revert to its internal system without undue delay.
Moreover, given that many banks in Germany do not prepare their accounts in accordance with IFRS, the requirements should be formulated in such a way that they make sense for all institutions. The text should be adjusted accordingly.
We agree that IRRBB from an NII perspective is an important risk that needs to be measured and reported. However, given that NII is part of the normative perspective, the individual limitation of this specific risk is not in line with general normative assumptions and underlying management actions. Like other factors in the normative perspective, NII should only be assessed periodically under defined overarching scenarios and analysed in the context of the entire normative requirement. We would therefore ask the EBA to align the definition of limits in the normative context with the new ICAAP requirements to allow banks to streamline their efforts and to prevent conflicting steering implications. A lack of internal limits in one perspective should not lead to a bank’s internal methods being classified as non-satisfactory.
In addition, we recommend streamlining the text of some requirements to avoid duplication (e. g. paras 154 vs 155 and 146 vs 149).
Question 3: Is there any specific element in the definition of CSRBB that is not clear enough for the required assessment and monitoring of CSRBB by institutions?
The draft guidelines propose to change the definition of CSRBB set out in the old 2018 guidelines. While para 7 of the old guidelines refers to the “risk driven by changes in the market perception about the price of credit risk, liquidity premium and potentially other components of credit-risky instruments … which is not explained by IRRBB or by expected credit/(jump-to-)default risk”, the new para 7 defines CSRBB as the “risk driven by changes of the market price for credit risk, for liquidity and for potentially other characteristics of credit-risky instruments, which is not captured by IRRBB or by expected credit/(jump-to-) default risk.”The impact of this change is unclear; equally unclear is whether a material change is intended at all. In our view, the risk of a change in market price implies the existence of (fluctuating) market prices for products in the relevant business segment. Market prices are generally not available in the area of conventional customer lending. This area consequently falls outside the scope of the definition. German banks, for example, finance a large number of small and medium-sized businesses (such as “the baker on the corner”) for whose financing no market prices are available and thus no credit spreads can be determined. This also applies to residential real estate finance. If meaningful market information is available (e. g. for the corporate bond portfolio), these positions should be included in the measurement of CRSBB. Specifically, this therefore applies to transactions for which credit spreads are relevant, especially securities transactions. It should also be possible to exclude certain positions from the risk assessment to avoid double counting. This will be the case, for instance, if it can be demonstrated that the corresponding risks are already assessed and managed elsewhere (e. g. in combined credit risk and credit spread risk models).
There is currently no industry standard for measuring CSRBB: on the contrary, approaches differ widely. No industry standard should be “enforced” in this area by the EBA. Requirements should be “modelling neutral”, which is not the case at present. A typical example of the use of trading book VaR models adapted for banking book purposes is the integrated measurement of IRRBB and CSRBB including idiosyncratic risk (with or without changes in individual creditworthiness). Alternatively, credit risk models can be extended to include components for CSRBB measurement which enable an integrated risk measurement on the basis of creditworthiness and market-induced spread changes. When such integrated models are used, there is a particular danger of the proposed procedure quickly leading to double counting, which must be avoided at all costs. Existing models are all well understood and have proved their worth; it should continue to be possible to use them.
According to para 157, however, it will only be possible to include idiosyncratic credit spread components in CSRBB measurement if it can be demonstrated that the results will be conservative. This alleged relief is nothing of the sort and the obligation to produce evidence should be deleted since, depending on the observed reporting date, inclusion may lead to either an increase or a reduction in risk as a result of possible diversification effects that the inclusion may generate. Since the consideration of additional elements of the overall credit spread risk tends to increase unavoidable double counting owing to overlaps with migration and default risk, results should be considered conservative per se and permitted without any requirement for verification. By contrast, a precise measurement of the individual components of credit spread risk (market spread, liquidity spread, separate consideration of idiosyncratic spread) may be desirable in theory but cannot be accurately carried out in practice. The explanatory box on page 46 includes a qualification to this effect. Even if separate consideration of the spread components were possible, the question arises as to whether this would offer any added value and make a meaningful contribution to a bank’s risk management practices. For example, the default premium might already be included in the credit risk and the liquidity premium in the liquidity risk of the institution. Anticipated rating changes, i. e. expected migration to other rating classes, may in turn have an impact on the default premium and possibly the liquidity premium.
We believe greater account should be taken of an institution’s individual circumstances (portfolio composition, design of internal systems, availability of reference data, etc.) when assessing credit spread risk. It should be up to the bank itself to decide on the specific design of its system, including whether to opt for integrated or isolated measurement. Analysing currency-specific dimensions, for example, as proposed in para 123, will not make good sense for every bank. For all larger portfolios, further clustering (such as by rating class, sector, region, product, or possibly also more granular, issuer-specific mapping) is required for internal risk management purposes and has thus become market standard. It should consequently be up to institutions to choose which type of clustering is most appropriate. This is the only way, in our view, to achieve meaningful results from which sensible risk management practices can be derived. On top of that, a generic and arbitrarily constructed curve will lead to inappropriate mapping of the credit spread risk of certain portfolios and thus to undesirable risk management incentives. Institutions should therefore be allowed to use bank individual appropriate credit spread curves with which observable market prices are also set for traded instruments.
Backtesting (realisation vs forecast) does not always provide meaningful results and is sometimes not possible at all. It is not possible, for example, to observe on the market the “actual result” of a loan to a client such as the baker on the corner (see, in particular, para 147). The question therefore arises as to what value should be backtested in such cases and what meaningful implications (in terms of risk management takeaways) should be derived from this theoretical exercise.
Question 4: As to the suggested perimeter of items exposed to CSRBB, would you consider any specific conceptual or operational challenge to implement it?
We recommend a narrower interpretation of the CSRBB perimeter, which we believe would also be consistent with the Basel standards. Given their sensitivity to credit spread risk, fair value positions would generally be included. Positions where changes in the balance sheet value are not observable, such as those accounted for at amortised cost, would generally be excluded from the CSRBB perimeter as they have no (material) influence on credit spread risk and their inclusion would be give rise to extremely costly calculations.According to our understanding of the definition of CSRBB (cf. our reply to question 3), positions should, in addition, only be included in the scope of CSRBB measurement if meaningful market information is available about them. Reference prices are not sufficient in this respect: position-specific prices must be available. This means instruments must be eligible for trading on a stock exchange and have a certain minimum market liquidity. Prices from secondary markets, by contrast, are not a suitable means of measuring CSRBB (registered bonds traded on secondary markets with only issuer spreads would also have to be excluded, for instance). In summary, therefore, only liquid securities positions measured at fair value in the banking book should be included in the measurement. As we see it, this is the only way to achieve appropriate risk management incentives.
In particular, we strongly oppose the requirement in para 124 for institutions not to exclude certain positions ex ante but to provide detailed documented evidence of an “absence of sensitivity to credit spread risk” for each exclusion of a group of similar positions. This will not be feasible to implement in practice since, for items such as customer loans or demand deposits, evidence of this kind is virtually impossible to document in a way which is audit compliant. It would make more sense to assume that CSRBB generally plays no material role for certain types of products (with negative lists at least at national level). If neither idiosyncratic risk nor migration and default risk are taken into account, this assumption is likely to be largely correct. Even if credit spread changes were observable for these product types, they would play no role when it comes to conditions for either new or existing business (in a classic buy-and-hold business model). The changes would have no influence on P&L nor, in consequence, on equity and would ultimately have no effect on the bank’s solvency. Their measurement would therefore be irrelevant.
We also recommend that pension plans (assets and liabilities) be explicitly excluded from the scope. The pension plans of banks should be excluded from the scope of CSRBB because of the way they are uniquely governed and managed relative to the bank’s commercial activities. Pension plans, specifically defined benefit plans, contain credit-risky investments along with equities and other investments in order to hedge the credit-risky liabilities of the plan. These investments are managed by an investment manager, typically a third-party asset manager, and governed in many cases by a board of trustees who are either independent or semi-independent of the bank. In other words, the bank may not be able to legally apply these guidelines to those plans, especially considering the granularity of requirements envisioned by the revised CSRBB rules. Even where the bank is able to influence investment strategy, it takes a portfolio approach that looks at the overall risks of the plan across assets and liabilities and establishes broad investment guidelines for the investment manager to adhere to. We do not think the rules were intended for pension plans and ask that they be explicitly excluded from the scope of CSRBB.
Furthermore, pragmatic procedures should be established to enable individual institutions to exclude other non-material positions from the scope if necessary.
In special cases and depending on the bank’s business model, the com