Response to consultation on Guidelines on technical aspects of the management of interest rate risk
Go back
Furthermore, it would be very useful if the definitions of Core/Transient Balances were detailed in the definitions section. The difference among core and unstable balances is not descripted in the document and a misinterpretation could result in a flawed regulatory report.
Additionally, we consider that the NPE's definition is not sufficiently clear in the Guidelines. We would appreciate if it could be confirmed that NPEs refers to the commonly used term NPLs (Non Performing Loans).
Furthermore, in paragraph (29), it is stated that “Economic Capital may be allocated back to Business Units and Products to ensure that the full costs of the underlying business units or products are properly understood by those responsible for managing them”. Except for certain optionality costs, we consider this request pointless, as there is no IRRBB risk for an individual transaction. IRRBB emerges from the mistmaching between assets and liabilities, and consequently it is centralised in the ALCO where it is managed.
While some banks prefer the exclusion of commercial margin from the EV metric, when discounting with risk free rate curve, other consider that extracting the commercial margins does not provide an accurate measure and introduces additional adverse effects undermining comparability. As commercial margins need to be estimated, its exclusion introduces subjectivity to the model as a result of the diverse margins estimations and extraction methods. Besides, it increases complexity to the data collection process and reduces the transparency of the EVE results.
The regulatory enforcement of any alternative should be assessed carefully due to the material additional resources that will have to be devoted. We think that a change of this nature should be discussed deeply before implementation, and final decision should be at least methodologically robust.
The adaptation cost will vary among banks, but undoubtedly, the exclusion of commercial margins increases complexity to the data collection process, and it can result very cumbersome and costly, raising doubts about the capacity of less sophisticated banking entities to fulfill supervisor’s expectations.
(PLEASE SEE ATTACHED PDF)
The minimum level reached (i.e maximum floor) across all tenors is summarized in the following tables:
(PLEASE SEE ATTACHED PDF)
The previous analysis evidences that a) -1.50% is extremely conservative, and b) a currency dependant floor may be considered.
In addition, over the last years, the use of a multi-curve framework has become an industry sound practice (e.g. OIS, LIBOR 1 month, LIBOR 3 months, LIBOR 6 months and LIBOR 12 months are usually associated to different curves). When considering the same regulatory floor for multiple yield curves in a single currency, the basis spreads between curves might become zero, which reveals a flaw in the proposed methodology. We believe that the current approach does not provide the right incentives to manage the basis risk when the floor takes effect. Hence, we suggest first applying the floor to the risk-free interest rate curve for each currency, and then constructing the rest of the curves for the same currency preserving the current basis spread.
We understand that the objective of the proposed currency risk aggregation methodology for the Outlier Test is to be simple and standardizable in order to maintain comparability among Banks. However, the methodology proposed is extremely conservative and methodologically wrong, as it fails to capture the advantages of diversification and it ignores the mitigation effects among different currencies.
The benefits of diversification are well-known and they can be observed on any historical analysis. While correlation among currencies is not perfect (100%), the aggregated losses of a currency diversified portfolio are far beneath the simple aggregation of the worst impacts in each currency (see Appendix 1).
Despite this, the proposal assumes perfect correlation among all currencies, independently of the degree of relationship among them. This is a fatal flaw of the Guidelines as it prejudice Entities with diversified portfolios and discourage Banks to prevent concentration risk.
Moreover, since the methodology does not take into account the correlation among currencies, it does not recognize any compensation effect. This assumption may have a severe impact on Banks operating in markets with multiple currencies promoting unintended consequences on their hedging strategies. Mitigation benefits will depend on the correlation (the higher the correlation, the greater benefit) and the exposure in each currency. Perfect correlations (positive or negative) are difficult to occur, unless IRs are linked by central policy. Likewise, strong negative correlations are uncommon to be observed in real markets. The majority of the currency correlations are positive and the recognition of a mitigation effect is expected.
The currency risk aggregation methodology may be determined by different approaches, but any of them should be based upon historical correlation and the current risk profile of each bank. In this way, the aggregation methodology should recognize diversification among losses in different currencies and also allow some mitigation benefit between losses and gains for each scenario.
There are alternative methods that, keeping maths simple, solve the main drawbacks of the current EBA GL proposal by taking into account the relationship among currencies and the specific risk exposure of each Bank.
Thus, the aggregation methodology could be divided in the following steps:
For each scenario, aggregation of negative impacts (losses) among them, taking into account the diversification across currencies;
For each scenario, aggregation of positive impacts (gains) among, taking into account the diversification across currencies;
For each scenario, mitigation between the aggregated losses and aggregated gains, taking into account the diversification across currencies;
These steps are described in the following sections (see appendix 2):
Aggregated Loss by scenario
The aggregation of currencies with negative impact could be calculated using the following aggregation formula:
The magnitude of the impact on the consolidated EVE will strongly depend on the risk profile of each Entity. Indeed, those Banks whose exposure is spread over several currencies, and hence they are more resilient to a risk event, will be the most affected by the aggregation methodology.
We understand that the objective of the proposed currency risk aggregation methodology for the Outlier Test is to be simple and standardizable in order to maintain comparability among Banks. However, the methodology proposed is extremely conservative and methodologically wrong, as it fails to capture the advantages of diversification and it ignores the mitigation effects among different currencies.
The benefits of diversification are well-known and they can be observed on any historical analysis. While correlation among currencies is not perfect (100%), the aggregated losses of a currency diversified portfolio are far beneath the simple aggregation of the worst impacts in each currency (see Appendix 1).
Despite this, the proposal assumes perfect correlation among all currencies, independently of the degree of relationship among them. This is a fatal flaw of the Guidelines as it prejudice Entities with diversified portfolios and discourage Banks to prevent concentration risk.
Moreover, since the methodology does not take into account the correlation among currencies, it does not recognize any compensation effect. This assumption may have a severe impact on Banks operating in markets with multiple currencies promoting unintended consequences on their hedging strategies. Mitigation benefits will depend on the correlation (the higher the correlation, the greater benefit) and the exposure in each currency. Perfect correlations (positive or negative) are difficult to occur, unless IRs are linked by central policy. Likewise, strong negative correlations are uncommon to be observed in real markets. The majority of the currency correlations are positive and the recognition of a mitigation effect is expected.
The currency risk aggregation methodology may be determined by different approaches, but any of them should be based upon historical correlation and the current risk profile of each bank. In this way, the aggregation methodology should recognize diversification among losses in different currencies and also allow some mitigation benefit between losses and gains for each scenario.
There are alternative methods that, keeping maths simple, solve the main drawbacks of the current EBA GL proposal by taking into account the relationship among currencies and the specific risk exposure of each Bank.
Thus, the aggregation methodology could be divided in the following steps:
For each scenario, aggregation of negative impacts (losses) among them, taking into account the diversification across currencies;
For each scenario, aggregation of positive impacts (gains) among, taking into account the diversification across currencies;
For each scenario, mitigation between the aggregated losses and aggregated gains, taking into account the diversification across currencies;
These steps are described in the following sections (see appendix 2):
Aggregated Loss by scenario
The aggregation of currencies with negative impact could be calculated using the following aggregation formula:
(PLEASE SEE ATTACHED PDF)
Aggregated Gain by scenario
Similarly, the same aggregation formula could also be applied to currencies with positive impact:
(PLEASE SEE ATTACHED PDF)
Aggregated Impact by scenario
Finally, having calculated the aggregated positive changes and the aggregated negative changes, the final formula would mitigate losses against gains:
(PLEASE SEE ATTACHED PDF)
where Cmit represents the ‘mitigation coefficient’ between negative and positive impacts .
This is a simple approach that intends to be easy to implement and suitable for standardization, in order to keep comparability among Banks. Consequently, it should be used only for the “Supervisory Outlier Test”.
The target of this method is:
To be simple
To be suitable for standardization
To recognize the diversification among impacts on different currencies
To capture the mitigation effect among impacts on different currencies
To be sensitive to the correlation among currencies
To accommodate to the specific risk profile of each bank
For the purpose of comparability, we suggest the use of regulatory-specified correlations, as they depend on the selected curve instrument, time window and length of historical time series. The complexity of the process could be reduced by identifying clusters of currencies and then prescribing the correlations among them. We recommend, that the mitigation coefficient should also be prescribed by regulators, based on the correlations between currencies with negative and positive changes and the magnitude of their impacts.
In Appendix 3 we propose an alternative approach, that simplifies, even more, the implementation of the proposed aggregation methodology and ease the standardization.
In view of the foregoing, we deem that the EBA GL proposed approach is excessively simplistic and contradicts the spirit of the Guidelines, where a high level of accuracy in measuring IRRBB is expected, especially for the most sophisticated Banks. Although IRRBB by currency may be correctly measured, the final consolidated risk figure will be noticeably inaccurate as impacts are merely added without taking into account correlation effects.
Question 1: Are the definitions sufficiently clear? If not, please provide concrete suggestions and justify your answer.
The definition of the Credit Spread Risk in the Banking Book (CSRBB) is too broad (“any kind of spread risk of interest rate sensitive instruments that is not IRRBB or credit risk “) and does not relate to IRRBB. No only the definition, but also the scope of application is unclear. Consequently, we request to delete the reference to CSRBB from the IRRBB Guidelines.Furthermore, it would be very useful if the definitions of Core/Transient Balances were detailed in the definitions section. The difference among core and unstable balances is not descripted in the document and a misinterpretation could result in a flawed regulatory report.
Question 2: Are the guidelines in section 4.1. regarding the general provisions sufficiently clear? If not, please provide concrete suggestions.
NAQuestion 3: Do you agree that cash flows from non-performing exposures (NPEs) should be net of provisions and treated as general interest rate sensitive instruments whose modelling should reflect expected cash flows and their timing for the purpose of EV and earnings measures? If not, please provide concrete suggestions and justify your answer.
If NPE are treated as interest rate sensitive instruments and consequently included in the economic capital calculation, there can be some potential double counting on the capital calculation with credit risk (for the treatment of LGDs). We recommend to nuance the paragraph 17.g with the same comment that apply to pension obligations: “NPEs should be net of provisions and should reflect the expected cash flow associated to these assets, unless their interest rate risk is captured in another measure”.Additionally, we consider that the NPE's definition is not sufficiently clear in the Guidelines. We would appreciate if it could be confirmed that NPEs refers to the commonly used term NPLs (Non Performing Loans).
Question 4: Are the guidelines in section 4.2. regarding the capital identification, calculation, and allocation sufficiently clear? If not, please provide concrete suggestions and justify your answer.
We agree that IRRBB must be measured from a double approach (EVE and NII perspective), however, regarding the internal capital allocation, additional guidance on the consideration of earnings would be necessary. A capital charge should only be required when the bank is exposed to a risk of loss, so any capital charge due to a potential reduction of earnings should be excluded from the guidelines. As it is described on the guidelines, its application could generate harmful effects by duplicating the risk of Economic Value and Earnings. It would be necessary to clarify how Income and Economic Value risks should be blended, so that capital allocation appropriately takes into account their combination, avoiding double counting.Furthermore, in paragraph (29), it is stated that “Economic Capital may be allocated back to Business Units and Products to ensure that the full costs of the underlying business units or products are properly understood by those responsible for managing them”. Except for certain optionality costs, we consider this request pointless, as there is no IRRBB risk for an individual transaction. IRRBB emerges from the mistmaching between assets and liabilities, and consequently it is centralised in the ALCO where it is managed.
Question 5: Do you agree with the list of elements to be considered for the internal capital allocation in respect of IRRBB to earnings in paragraph 30? If not, please provide concrete suggestions and justify your answer.
NAQuestion 6: Are the guidelines in section 4.3. regarding the governance sufficiently clear? If not, please provide concrete suggestions and justify your answer.
NAQuestion 7: Are the guidelines in section 4.4. regarding the measurement sufficiently clear? If not, please provide concrete suggestions and justify your answer.
Paragraph 106c asks to consider potential constraints on the repricing of retail deposits in low or negative interest rate environments (for instance, embedded floors in the customer deposits rate). We agree with this statement, but we belive that it should be extended to considered the assets side embedded options (potential restrictions on the customer loans repricing rates).Question 8: Do you consider the comparison between EV metrics calculated using contractual terms for NMDs with the EV metrics calculated with behavioural modelled assumptions sensible and practical? Please justify your answer.
The comparison of the EV metrics considering internal models with those resulting from their contractual characteristics does not seem to offer valuable conclusions about the Model Risk of the Institution. Moreover, this comparison could generate misleading results if it is used to make a peers analysis, as it is an inappropriate metric to compare Model Risk among Banks. We would consider more useful to run an analysis about the sensitivity to the assumptions, with comparable shocks to the hypothesis parameters.Question 9: Are the guidelines in section 4.5. regarding the supervisory outlier test sufficiently clear? If not, please provide concrete suggestions and justify your answer.
NAQuestion 10: Is the proportionality adequately reflected in the guidelines, in particular in relation to the transitional period for SREP category 3 and 4 institutions and the frequency of calculation for the additional outlier test under paragraph 112?
NAQuestion 11: If relevant, do you manage interest rate risk arising from pension obligations and pension plans assets within the IRRBB framework or do you cover it within another risk category (e.g. within market risk separately from IRRBB, etc.)?
NAQuestion 12: Which treatment of commercial margins cash flows do you consider conceptually most correct in EV metric, when discounting with risk free rate curve: a) including commercial margins cash flows or b) excluding commercial margins cash flows? Please justify your answer.
We believe that EVE calculation has to be coherent with banks’ internal risk frameworks and businesses, which may differ among entities. The decision about the inclusion or not of commercial margins in cash flows and/or discount factors should be anchored in the basis of methodological robustness and comparability.While some banks prefer the exclusion of commercial margin from the EV metric, when discounting with risk free rate curve, other consider that extracting the commercial margins does not provide an accurate measure and introduces additional adverse effects undermining comparability. As commercial margins need to be estimated, its exclusion introduces subjectivity to the model as a result of the diverse margins estimations and extraction methods. Besides, it increases complexity to the data collection process and reduces the transparency of the EVE results.
The regulatory enforcement of any alternative should be assessed carefully due to the material additional resources that will have to be devoted. We think that a change of this nature should be discussed deeply before implementation, and final decision should be at least methodologically robust.
Question 13: Are your internal systems flexible enough to exclude margins for the purpose of calculating EV measures for the supervisory outlier test? If not, what would be the cost to adapt your systems (high, medium, low)? Please elaborate your answer.
Not all the institutions’ internal systems are flexible enough to exclude margins, as its development depends on the internal management framework.The adaptation cost will vary among banks, but undoubtedly, the exclusion of commercial margins increases complexity to the data collection process, and it can result very cumbersome and costly, raising doubts about the capacity of less sophisticated banking entities to fulfill supervisor’s expectations.
Question 14: Do you consider the level of the proposed linear lower bound as described in paragraph 113 (k) appropriate? If not, please provide concrete suggestions and justify your answer.
We find the level of the floor (-150bps) overly conservative. To illustrate this opinion an analysis of the maximum expected movement in base of the historical volatility has been performed. The analysis shows that even for the current very low levels, it is highly improbable for rates to go below -100bps. The analysis has been performed on EUR and GBP, using the 6 month EURIBOR curve for the earlier and 3 LIBOR month curve for the later. Over a 5 year series of returns (computed quarterly and semi-annually) the 1%-ile of the return distribution (representing the most adverse down movements in the curves) was computed. For each tenor of the curve, this extreme down movement was subtracted from the current level of the rate:(PLEASE SEE ATTACHED PDF)
The minimum level reached (i.e maximum floor) across all tenors is summarized in the following tables:
(PLEASE SEE ATTACHED PDF)
The previous analysis evidences that a) -1.50% is extremely conservative, and b) a currency dependant floor may be considered.
In addition, over the last years, the use of a multi-curve framework has become an industry sound practice (e.g. OIS, LIBOR 1 month, LIBOR 3 months, LIBOR 6 months and LIBOR 12 months are usually associated to different curves). When considering the same regulatory floor for multiple yield curves in a single currency, the basis spreads between curves might become zero, which reveals a flaw in the proposed methodology. We believe that the current approach does not provide the right incentives to manage the basis risk when the floor takes effect. Hence, we suggest first applying the floor to the risk-free interest rate curve for each currency, and then constructing the rest of the curves for the same currency preserving the current basis spread.
Question 15: Do you consider the minimum threshold for material currencies included into the supervisory outlier test (5% for individual currency and minimum 90% of the total non-trading book assets or liabilities) sufficient to measure IRRBB in term of EVE? If not, please provide concrete suggestions and justify your answer.
We consider that a minimum materiality threshold of 5% for individual currencies and 90% for the total is adequate and aligned with the common practices. However, we deem that an exception should be made in the case that there is an aggregated exposure over 10% which is very fragmented among many currencies with very low materiality (i.e. <2%).Question 16: When aggregating changes to EVE in the supervisory outlier test, does the disregarding of positive changes to EVE have a material impact on the calculation of the supervisory outlier test?
The magnitude of the impact on the consolidated EVE will strongly depend on the risk profile of each Entity. Indeed, those Banks whose exposure is spread over several currencies, and hence they are more resilient to a risk event, will be the most affected by the aggregation methodology.We understand that the objective of the proposed currency risk aggregation methodology for the Outlier Test is to be simple and standardizable in order to maintain comparability among Banks. However, the methodology proposed is extremely conservative and methodologically wrong, as it fails to capture the advantages of diversification and it ignores the mitigation effects among different currencies.
The benefits of diversification are well-known and they can be observed on any historical analysis. While correlation among currencies is not perfect (100%), the aggregated losses of a currency diversified portfolio are far beneath the simple aggregation of the worst impacts in each currency (see Appendix 1).
Despite this, the proposal assumes perfect correlation among all currencies, independently of the degree of relationship among them. This is a fatal flaw of the Guidelines as it prejudice Entities with diversified portfolios and discourage Banks to prevent concentration risk.
Moreover, since the methodology does not take into account the correlation among currencies, it does not recognize any compensation effect. This assumption may have a severe impact on Banks operating in markets with multiple currencies promoting unintended consequences on their hedging strategies. Mitigation benefits will depend on the correlation (the higher the correlation, the greater benefit) and the exposure in each currency. Perfect correlations (positive or negative) are difficult to occur, unless IRs are linked by central policy. Likewise, strong negative correlations are uncommon to be observed in real markets. The majority of the currency correlations are positive and the recognition of a mitigation effect is expected.
The currency risk aggregation methodology may be determined by different approaches, but any of them should be based upon historical correlation and the current risk profile of each bank. In this way, the aggregation methodology should recognize diversification among losses in different currencies and also allow some mitigation benefit between losses and gains for each scenario.
There are alternative methods that, keeping maths simple, solve the main drawbacks of the current EBA GL proposal by taking into account the relationship among currencies and the specific risk exposure of each Bank.
Thus, the aggregation methodology could be divided in the following steps:
For each scenario, aggregation of negative impacts (losses) among them, taking into account the diversification across currencies;
For each scenario, aggregation of positive impacts (gains) among, taking into account the diversification across currencies;
For each scenario, mitigation between the aggregated losses and aggregated gains, taking into account the diversification across currencies;
These steps are described in the following sections (see appendix 2):
Aggregated Loss by scenario
The aggregation of currencies with negative impact could be calculated using the following aggregation formula:
The magnitude of the impact on the consolidated EVE will strongly depend on the risk profile of each Entity. Indeed, those Banks whose exposure is spread over several currencies, and hence they are more resilient to a risk event, will be the most affected by the aggregation methodology.
We understand that the objective of the proposed currency risk aggregation methodology for the Outlier Test is to be simple and standardizable in order to maintain comparability among Banks. However, the methodology proposed is extremely conservative and methodologically wrong, as it fails to capture the advantages of diversification and it ignores the mitigation effects among different currencies.
The benefits of diversification are well-known and they can be observed on any historical analysis. While correlation among currencies is not perfect (100%), the aggregated losses of a currency diversified portfolio are far beneath the simple aggregation of the worst impacts in each currency (see Appendix 1).
Despite this, the proposal assumes perfect correlation among all currencies, independently of the degree of relationship among them. This is a fatal flaw of the Guidelines as it prejudice Entities with diversified portfolios and discourage Banks to prevent concentration risk.
Moreover, since the methodology does not take into account the correlation among currencies, it does not recognize any compensation effect. This assumption may have a severe impact on Banks operating in markets with multiple currencies promoting unintended consequences on their hedging strategies. Mitigation benefits will depend on the correlation (the higher the correlation, the greater benefit) and the exposure in each currency. Perfect correlations (positive or negative) are difficult to occur, unless IRs are linked by central policy. Likewise, strong negative correlations are uncommon to be observed in real markets. The majority of the currency correlations are positive and the recognition of a mitigation effect is expected.
The currency risk aggregation methodology may be determined by different approaches, but any of them should be based upon historical correlation and the current risk profile of each bank. In this way, the aggregation methodology should recognize diversification among losses in different currencies and also allow some mitigation benefit between losses and gains for each scenario.
There are alternative methods that, keeping maths simple, solve the main drawbacks of the current EBA GL proposal by taking into account the relationship among currencies and the specific risk exposure of each Bank.
Thus, the aggregation methodology could be divided in the following steps:
For each scenario, aggregation of negative impacts (losses) among them, taking into account the diversification across currencies;
For each scenario, aggregation of positive impacts (gains) among, taking into account the diversification across currencies;
For each scenario, mitigation between the aggregated losses and aggregated gains, taking into account the diversification across currencies;
These steps are described in the following sections (see appendix 2):
Aggregated Loss by scenario
The aggregation of currencies with negative impact could be calculated using the following aggregation formula:
(PLEASE SEE ATTACHED PDF)
Aggregated Gain by scenario
Similarly, the same aggregation formula could also be applied to currencies with positive impact:
(PLEASE SEE ATTACHED PDF)
Aggregated Impact by scenario
Finally, having calculated the aggregated positive changes and the aggregated negative changes, the final formula would mitigate losses against gains:
(PLEASE SEE ATTACHED PDF)
where Cmit represents the ‘mitigation coefficient’ between negative and positive impacts .
This is a simple approach that intends to be easy to implement and suitable for standardization, in order to keep comparability among Banks. Consequently, it should be used only for the “Supervisory Outlier Test”.
The target of this method is:
To be simple
To be suitable for standardization
To recognize the diversification among impacts on different currencies
To capture the mitigation effect among impacts on different currencies
To be sensitive to the correlation among currencies
To accommodate to the specific risk profile of each bank
For the purpose of comparability, we suggest the use of regulatory-specified correlations, as they depend on the selected curve instrument, time window and length of historical time series. The complexity of the process could be reduced by identifying clusters of currencies and then prescribing the correlations among them. We recommend, that the mitigation coefficient should also be prescribed by regulators, based on the correlations between currencies with negative and positive changes and the magnitude of their impacts.
In Appendix 3 we propose an alternative approach, that simplifies, even more, the implementation of the proposed aggregation methodology and ease the standardization.
In view of the foregoing, we deem that the EBA GL proposed approach is excessively simplistic and contradicts the spirit of the Guidelines, where a high level of accuracy in measuring IRRBB is expected, especially for the most sophisticated Banks. Although IRRBB by currency may be correctly measured, the final consolidated risk figure will be noticeably inaccurate as impacts are merely added without taking into account correlation effects.