Response to consultation on draft RTS on IRRBB supervisory outlier tests
Go back
The EU framework for management of interest rate risk in the banking book has become very comprehensive and complex. Despite the legal basis for these documents being only two articles in the CRD (art 84 (5) and (6) and art 98 (5a)), the EBA is currently consulting on a total of 176 pages combined. In addition to this are the EBAs SREP guidelines, as well as guidelines and supervisory expectations from NCAs. In contrast, the Basel standard on which the framework is based is far less comprehensive and easier to understand. Also, the EU regulations apply to all banks whereas the Basel standards were initially developed for large internationally active institutions.
There is a general focus on ensuring proportionality in the prudential regulation. Although we acknowledge the need for sufficiently prudent management of interest rate risk amongst all EU/EEA banks, ESBG believes the current framework is too complex and challenging to implement for smaller institutions with non-complex operations and limited market risk exposure. Although there is a general possibility for institutions to, after a thorough and well-documented assessment, exclude certain risks if they can justify that those risks are not material, we believe there is a risk that supervisory practice will not be harmonized across the different jurisdictions. We hence believe that the guidelines and technical standards should provide more guidance on the application of the proportionality principle. This application should take into consideration the peculiarities of the national banking models and the interest risk inherent in national markets. In particular, small and non-complex institutions that are part of a group, subject to prudential requirement on consolidated level, should be excluded from the application of the strict thresholds for EVE.
Response to question 1:
ESBG does not see the prolongation of the floor to 50 years as being reasonable. Instead, we would propose to keep it at 20 years. Historically, 20 years as well as 30 years long-term rates were hardly ever negative in EUR particularly, for example. The 50-year rate was slightly negative only for 11 months. Moreover, many yield curves in the Central Easter European (CEE) region do not have yield curves until 50y. Keeping long-term rates negative for a significantly longer time as newly proposed has a multiplicative effect into discount factors, which does not have a historical background, and which will send adverse signals for otherwise stable balance sheet items, like customer deposits or floored client loans.
The suggested floor is considered too low taking into account what has been observed historically, especially inn the short term. A further lowering of the floor on the short end does not seem to be substantiated and negative consequences seem to be underestimated. Furthermore, it is our understanding that recalibrating the floor is mainly motivated by the observation of baseline bond yield points below the current floor, not by the relevant risk-free rates, which, aside from some minor exceptions, were respecting the floor in the past and current low-rate period. The EBA proposal could lead to a severe change in impact for EVE SOT and particularly for NII SOT, as the down shock more or less doubles for short tenors and increases with up to 90bps.
Option A/Metric 1:
We think that Option A is not an optimal choice for NII risk metric and undermines the complementary nature of the risk measures for NII and EVE. Option A uses capital (Tier 1) as a basis for the limiting measure. NII is a 1-year income risk measure, whereas Tier 1 is a stock quantity that should ensure long term stability of the institution. Tier 1 is already used in EVE, where it is more appropriate.
Option B/Metric 2:
Option B is more aligned with established internal interest rate risk management methodologies. Nevertheless, we think that the approach presented in the CP has material flaws:
o Unreasonably high NII utilization even in cases of low NII risk exposure
o Excessive sensitivity to input parameters
The inclusion of administrative expenses in metric 2 initiates unwanted consequences and makes the metric an arbitrary measure not fit for steering purposes. It will create high exposure to the threshold and high volatility of limit utilization which is not connected with the NII position but influenced by disconnected P/L contributions, the strategy of the institution and one-time effects. Technically speaking, metric 2 suffers from the small denominator problem. If the denominator in the formulae is close to (or even exactly) zero, the limit utilization is unbounded even when the NII risk is very small, extreme changes in NII limit utilization are created only by small changes in the underlying NII risk.
Realistic examples can be created easily for the small denominator problem. If Administrative Expenses are close to the Operating Income (and additionally NII base is stable around the last years NII hist) then the utilization becomes unbounded. Please find one below.
Example for High Sensitivity to input parameters (real example, anonymized by scaling of exposures)
NII Hist = 6.3 mn, Operating Income = 10 mn, NII Base = 5.1 mn, NII Shock = 4.2 mn, NII hist = 6.3 mn
NII risk = NII Shock – NII Base = -0.9 mln = 13.7% of NII hist
EBA metric Option A under small changes of inputs (10% increase)
Scenario EBA Option B EBA Limit
EBA Opt B -203% 35%
Op Inc +10% -102% 35%
NII Hist + 10% -2.378% 35%
Admin Exp +10% -2.378% 35%
NII Shock +10% -104% 35%
NII Base+10% -147% 35%
Although the institutions risk is only 13.7% compared to the historical NII, the EBA metric B results in 203%, which corresponds to 6 times above the proposed limit. As demonstrated in the example, small changes have a significant impact on the exposure between 104% and 2400%. In order to meet the EBA limit, the institution would need to reduce the NII risk to 160k or 2.5% of the historical NII. Additionally, administrative expenses on NII for currencies other than the home currency are usually not available.
Alternative proposal:
In order to overcome the problem with Metric B we propose to abandon the term with administrative expenses and define the following metric:
o Option B alternative proposal: NII Shock/NII Base -1 < threshold where NII Base is the most recent 12 month rolling window
Additionally, the reference to the latest year-end NII can create unintended consequences for the metrics interpretation (e. g. from one-off effects). We suggest that a moving average over 12 months should be allowed instead.
“Narrow vs wide NII”: We regard the “narrow” NII as the better measure of NII for several reasons. Firstly, it is easy to compute, and secondly, it is easier to compare across banks, and since of the main purposes of the SOT is comparison across banks, we believe this consideration should be given weight. Secondly, the “wider NII” does have some conceptual aspects which makes interpretation and calculation difficult. For instance, in paragraph 18 it is even discussed whether other P&L lines such as fees and commission should be included. Many banks have governance structures set up where they distinguish between commission income and its relation to interest rate sensitivity, and NII and other “pure bank book market risk” activities. Secondly, the inclusion of such P&L lines will increase complexity and transparency and therefore comparability across banks, and perhaps also within banks. Finally, the discussion regarding inclusion of fair value items beyond the 1-year scope of NII-calculation complicates the interpretation and conceptual soundness of the definition of NII-Sensitivity, as it clouds the direct effects of the change in interest-rate levels of the banks’ earnings ability and capacity.
Calibration of threshold: Judging by Figure 2 (p.37) and Table 7, there are only 4 outliers in the SOT, two of which have positive NII. There remain therefore only 2 observations with a negative ∆NII/Tier1, one which indicates ∆NII/Tier1 of -2%, and the other of <-5%. Basing the -2.5% threshold on these two observations therefore seems to be imprudent. Especially considering that a standard deviation is 2.2% according to Table 2, and the 50th percentile is -0.6%, e.g. the threshold is calibrated to be a one-standard-deviation away from the median (and nearly 1-std from mean).
Respondents are also kindly requested to express whether they find an inclusion of market value changes in the calculation of the NII SOT clear enough.
General comments – Proportionality:The EU framework for management of interest rate risk in the banking book has become very comprehensive and complex. Despite the legal basis for these documents being only two articles in the CRD (art 84 (5) and (6) and art 98 (5a)), the EBA is currently consulting on a total of 176 pages combined. In addition to this are the EBAs SREP guidelines, as well as guidelines and supervisory expectations from NCAs. In contrast, the Basel standard on which the framework is based is far less comprehensive and easier to understand. Also, the EU regulations apply to all banks whereas the Basel standards were initially developed for large internationally active institutions.
There is a general focus on ensuring proportionality in the prudential regulation. Although we acknowledge the need for sufficiently prudent management of interest rate risk amongst all EU/EEA banks, ESBG believes the current framework is too complex and challenging to implement for smaller institutions with non-complex operations and limited market risk exposure. Although there is a general possibility for institutions to, after a thorough and well-documented assessment, exclude certain risks if they can justify that those risks are not material, we believe there is a risk that supervisory practice will not be harmonized across the different jurisdictions. We hence believe that the guidelines and technical standards should provide more guidance on the application of the proportionality principle. This application should take into consideration the peculiarities of the national banking models and the interest risk inherent in national markets. In particular, small and non-complex institutions that are part of a group, subject to prudential requirement on consolidated level, should be excluded from the application of the strict thresholds for EVE.
Response to question 1:
ESBG does not see the prolongation of the floor to 50 years as being reasonable. Instead, we would propose to keep it at 20 years. Historically, 20 years as well as 30 years long-term rates were hardly ever negative in EUR particularly, for example. The 50-year rate was slightly negative only for 11 months. Moreover, many yield curves in the Central Easter European (CEE) region do not have yield curves until 50y. Keeping long-term rates negative for a significantly longer time as newly proposed has a multiplicative effect into discount factors, which does not have a historical background, and which will send adverse signals for otherwise stable balance sheet items, like customer deposits or floored client loans.
The suggested floor is considered too low taking into account what has been observed historically, especially inn the short term. A further lowering of the floor on the short end does not seem to be substantiated and negative consequences seem to be underestimated. Furthermore, it is our understanding that recalibrating the floor is mainly motivated by the observation of baseline bond yield points below the current floor, not by the relevant risk-free rates, which, aside from some minor exceptions, were respecting the floor in the past and current low-rate period. The EBA proposal could lead to a severe change in impact for EVE SOT and particularly for NII SOT, as the down shock more or less doubles for short tenors and increases with up to 90bps.
Question 2: Do respondents have any comment related to these two metrics for the specification and the calibration of the test statistic for the large decline in Article 6 for the purpose of NII SOT? Specifically, do respondents find the inclusion of administrative expenses in metric 2 clear enough? Do respondents have any comment on the example on currency aggregation for metric 1 and metric 2?
ESBG would have some considerations on the proposed potential metrics for the definition of the large decline. In our view, Option A, referring to a capital related metric, is not an optimal choice as it undermines the complementary nature of the risk measures for NII and EVE and since capital is not directly related to NII generation. Instead, Option B referring to a cost related metric looks more in line with established internal interest rate risk management methodologies. Yet, the addition of the administrative expenses term makes the metric excessively volatile and unreasonably complex and constitutes a fatal flaw of this option. We therefore believe that option B should be favoured, but only if it is adjusted as per our proposal outlined below.Option A/Metric 1:
We think that Option A is not an optimal choice for NII risk metric and undermines the complementary nature of the risk measures for NII and EVE. Option A uses capital (Tier 1) as a basis for the limiting measure. NII is a 1-year income risk measure, whereas Tier 1 is a stock quantity that should ensure long term stability of the institution. Tier 1 is already used in EVE, where it is more appropriate.
Option B/Metric 2:
Option B is more aligned with established internal interest rate risk management methodologies. Nevertheless, we think that the approach presented in the CP has material flaws:
o Unreasonably high NII utilization even in cases of low NII risk exposure
o Excessive sensitivity to input parameters
The inclusion of administrative expenses in metric 2 initiates unwanted consequences and makes the metric an arbitrary measure not fit for steering purposes. It will create high exposure to the threshold and high volatility of limit utilization which is not connected with the NII position but influenced by disconnected P/L contributions, the strategy of the institution and one-time effects. Technically speaking, metric 2 suffers from the small denominator problem. If the denominator in the formulae is close to (or even exactly) zero, the limit utilization is unbounded even when the NII risk is very small, extreme changes in NII limit utilization are created only by small changes in the underlying NII risk.
Realistic examples can be created easily for the small denominator problem. If Administrative Expenses are close to the Operating Income (and additionally NII base is stable around the last years NII hist) then the utilization becomes unbounded. Please find one below.
Example for High Sensitivity to input parameters (real example, anonymized by scaling of exposures)
NII Hist = 6.3 mn, Operating Income = 10 mn, NII Base = 5.1 mn, NII Shock = 4.2 mn, NII hist = 6.3 mn
NII risk = NII Shock – NII Base = -0.9 mln = 13.7% of NII hist
EBA metric Option A under small changes of inputs (10% increase)
Scenario EBA Option B EBA Limit
EBA Opt B -203% 35%
Op Inc +10% -102% 35%
NII Hist + 10% -2.378% 35%
Admin Exp +10% -2.378% 35%
NII Shock +10% -104% 35%
NII Base+10% -147% 35%
Although the institutions risk is only 13.7% compared to the historical NII, the EBA metric B results in 203%, which corresponds to 6 times above the proposed limit. As demonstrated in the example, small changes have a significant impact on the exposure between 104% and 2400%. In order to meet the EBA limit, the institution would need to reduce the NII risk to 160k or 2.5% of the historical NII. Additionally, administrative expenses on NII for currencies other than the home currency are usually not available.
Alternative proposal:
In order to overcome the problem with Metric B we propose to abandon the term with administrative expenses and define the following metric:
o Option B alternative proposal: NII Shock/NII Base -1 < threshold where NII Base is the most recent 12 month rolling window
Additionally, the reference to the latest year-end NII can create unintended consequences for the metrics interpretation (e. g. from one-off effects). We suggest that a moving average over 12 months should be allowed instead.
“Narrow vs wide NII”: We regard the “narrow” NII as the better measure of NII for several reasons. Firstly, it is easy to compute, and secondly, it is easier to compare across banks, and since of the main purposes of the SOT is comparison across banks, we believe this consideration should be given weight. Secondly, the “wider NII” does have some conceptual aspects which makes interpretation and calculation difficult. For instance, in paragraph 18 it is even discussed whether other P&L lines such as fees and commission should be included. Many banks have governance structures set up where they distinguish between commission income and its relation to interest rate sensitivity, and NII and other “pure bank book market risk” activities. Secondly, the inclusion of such P&L lines will increase complexity and transparency and therefore comparability across banks, and perhaps also within banks. Finally, the discussion regarding inclusion of fair value items beyond the 1-year scope of NII-calculation complicates the interpretation and conceptual soundness of the definition of NII-Sensitivity, as it clouds the direct effects of the change in interest-rate levels of the banks’ earnings ability and capacity.
Calibration of threshold: Judging by Figure 2 (p.37) and Table 7, there are only 4 outliers in the SOT, two of which have positive NII. There remain therefore only 2 observations with a negative ∆NII/Tier1, one which indicates ∆NII/Tier1 of -2%, and the other of <-5%. Basing the -2.5% threshold on these two observations therefore seems to be imprudent. Especially considering that a standard deviation is 2.2% according to Table 2, and the 50th percentile is -0.6%, e.g. the threshold is calibrated to be a one-standard-deviation away from the median (and nearly 1-std from mean).