Response to consultation on draft RTS on IRRBB standardised approach
Go back
Nevertheless article 6 sets out that cash flows arising from principal and fixed interest payments must be allocated at first repricing, and the risk of prepayment at fist repricing is not material, so its inclusion is not relevant for these cash flows. However, prepayment may affect the valuation of the spread component of long-term interest cash flows when discounted using risk free rates. The real importance is the valuation method for these products (a full revaluation approach versus first repricing).
The restrictions to be applied to passthrough rates and stable percentages are easy to implement for simple models (e. g. in models where customer rates are linear functions), but it is not straightforward for more complex models where customer rates depend on the past rates or other additional variables. More specific instructions would be welcome.
We understand that the proposed restriction for stability parameters and specially for the weighted average maturity may lead to very different results in markets with high interest rates and very low elasticity to the movements of the interest rates, such as Latin-America. In these countries, historical analysis shows that caps can be far from internal models and reality. These constrains leads to unrealistic metrics being not comparable with internal control and management.
Additionally, regarding the segmentation of NMD between “transactional” and “non-transactional” we think it fits better with a liquidity approach than with an interest rate perspective. Historical analysis shows that “transitional” does not necessarily mean “stable”. The drivers affecting stability usually are others, such as type of remuneration, product offered by competitors, economic cycles…
This segmentation (transactional / non- transactional / core / non-core) would lead to banks to maintain two different segmentations and model of NMD opening the breach between internal metrics and regulatory ones, therefore making difficult the comparability.
Consequently, regarding the restriction on the proportion of core deposits, 70% for retail non-transactional and 90% for transactional, we would suggest using a unique parameter, e.g., an 80% as the average between them in all segments to facilitate the calculations.
Additionally, for long term deposits slotting the cumulative amount of early redemptions in the ON bucket is not accurate, as those estimated outflows will be paid across the life of the deposit, and not immediately.
The time value (extrinsic) of embedded options linked to balance items accounted at amortised cost should be excluded from IRRBB calculations, as it does not have an effect in capital. The inclusion of their extrinsic value in the risk metrics may generate undesirable volatility derived from out-of-the-money options (e.g., 0% rate floors in environments with high interest rates), confusing the risk management decisions. Moreover, the market value of these options is not reflected in Equity, and they are not managed with the intention to be traded.
Additionally, we don’t understand the purpose of treating differently bought and sold options.
Regarding the volatility shock for explicit options and those embedded in fair valued instruments, we consider that implementing these criteria may not be operationally complicated for sophisticated banks, but it will introduce a great complexity when combined with interest rate shocks scenarios, especially for the small and medium size entities. Moreover, it should be clearly stated that volatility shocks only apply for explicit options or those embedded in fair value instruments. A clarification on the purpose of these calculations would be appreciated, as vega risk is not commonly managed in embedded options for IRRBB.
In any case, further detailed criteria about the relevant hypothesis would be appreciated as it would help on its implementation and harmonization among banks. The valuation of optionality is sensitive to models and parametrization, and the lack of common criteria would lead to non-comparable results. Variables like volatility type (implicit vs historical), absolute or relative shifts, and whether the valuation model follows a normal or log normal distribution should be specified.
Although it is reasonable to consider different levels of volatility in each of the interest rate scenarios and adapt this parameter depending on the scenario considered, increasing volatility in all scenarios does not seem appropriate. What can be seen when analysing the historical series of implied volatility is that as the model (e.g., Black-Scholes) acts as a translator between volatility and price, an increase in interest rates automatically implies a decrease in the implied volatility observed in the market and the opposite when rates fall, so when creating volatility scenarios, it is also necessary to take the “model” effect into account. Applying a 25% change in any scenario may be far from what is expected from a market point of view.
We understand that the shift (25%) should be considered as multiplicative (stressed volatility = base volatility x 1,25) to better adapt to each type of model, but it should be specified in the text.
Additionally, a 25% shock may not be representative of all currencies (and if not, more instructions about how to calibrate it for other currencies, an even maturity would be necessary).
Finally, it should be clearly stated in the RTS that volatility shocks do not apply for SOT calculations.
Moreover, operationally, the calculation of the fair value component introduces a higher complexity in the metric due to the overlapping between the NII and the FV changes during the projection period (e.g., 12M). The calculation of FV changes not only requires excluding the instruments maturing or repricing during the first year. It also requires estimating the future MtM impact at the end of the projection period (t=1 year) on the fair value instruments with longer maturities (>1Y) which will be computationally very demanding. It must be considered that the residual maturity of these instruments will be 1 year shorter at the end of the NII projection period.
Operationally, from our view the calculation of the net interest income add-on for basis risk is implementable but some additional clarification would be needed to consider such an add-on.
The basis risk is defined by taking into the reference terms: Overnight, 1 month, 3 months, 6 month or 12 months, and shocks are calculated comparing interest rates with the overnight reference.
To calculate the add-on, it is necessary to obtain the sensitivity to each reference rate, but when calculating the add-on, we note that several cases could be applied:
1. Applying the highest calibrated shock to the sensitivity to that reference rate.
2. Applying the largest calibrated shock to the total sensitivity of the entire portfolio.
3. Applying to each sensitivity the shock calculated for its reference rate.
We understand that the first option is the one referred in the standard method, but further clarification on this would be needed in the text.
Another matter we would like to raise is how the add-on for basis risk should be treated in a multicurve environment. The methodology seems to be written only for the case of a single curve.
Question 1: What is the materiality of prepayments for floating rate instruments and what are the underlying factors? Would you prefer the inclusion of a requirement in Article 6 for institutions to estimate prepayments for these instruments?
Floating rate loans represent a significant part of the credit portfolio in some geographies in the balance sheets of Spanish banks, being the retail mortgages the most representative products. These types of portfolios present lower levels of prepayment, mostly driven by other factors more complex arising from macroeconomic and cultural situations in the market. This means that the normal techniques used to value options cannot be applied directly, and they must be combined with empirical statistical models that aim to capture customer behaviour.Nevertheless article 6 sets out that cash flows arising from principal and fixed interest payments must be allocated at first repricing, and the risk of prepayment at fist repricing is not material, so its inclusion is not relevant for these cash flows. However, prepayment may affect the valuation of the spread component of long-term interest cash flows when discounted using risk free rates. The real importance is the valuation method for these products (a full revaluation approach versus first repricing).
Question 2: Do respondents find that the required determination of stable/non-stable deposits, and core/non-core deposits as described in Article 7 is reflective of the risks and operationally implementable? In case of any unintended consequence or undesirable effect on certain business models or specific activities, please kindly provide concrete examples.
Unstable / stable are concepts usually used in internal models. Our view is that the subsequent split between core and non-core based on its interest rate sensitivity (not affected by passthrough rates), does not represent a realistic approach, given that internal prices and remuneration policies of the products are not based on these variables. Besides, the multipliers to be applied in upward/downward scenarios are fixed and calibrated independently of the interest rate levels and the product type, which may lead to a misrepresentation of the real behavioural of stable deposits under different scenarios.The restrictions to be applied to passthrough rates and stable percentages are easy to implement for simple models (e. g. in models where customer rates are linear functions), but it is not straightforward for more complex models where customer rates depend on the past rates or other additional variables. More specific instructions would be welcome.
We understand that the proposed restriction for stability parameters and specially for the weighted average maturity may lead to very different results in markets with high interest rates and very low elasticity to the movements of the interest rates, such as Latin-America. In these countries, historical analysis shows that caps can be far from internal models and reality. These constrains leads to unrealistic metrics being not comparable with internal control and management.
Additionally, regarding the segmentation of NMD between “transactional” and “non-transactional” we think it fits better with a liquidity approach than with an interest rate perspective. Historical analysis shows that “transitional” does not necessarily mean “stable”. The drivers affecting stability usually are others, such as type of remuneration, product offered by competitors, economic cycles…
This segmentation (transactional / non- transactional / core / non-core) would lead to banks to maintain two different segmentations and model of NMD opening the breach between internal metrics and regulatory ones, therefore making difficult the comparability.
Consequently, regarding the restriction on the proportion of core deposits, 70% for retail non-transactional and 90% for transactional, we would suggest using a unique parameter, e.g., an 80% as the average between them in all segments to facilitate the calculations.
Question 3: Do respondents find that the required determination and application of a conditional prepayment rate and term deposit redemption rate as described in Article 8 and 9 is reflective of the risks and operationally implementable? In case of any unintended consequence or undesirable effect on certain business models or specific activities, please kindly provide concrete examples.
We consider implementing these restrictions would not be complicated from an operational point of view. However, we would like to highlight that these parameters (multipliers) are not a rule that had been observed in historical analysis in all geographies/products. Therefore, as previously raised, this would lead to non-comparable metrics.Additionally, for long term deposits slotting the cumulative amount of early redemptions in the ON bucket is not accurate, as those estimated outflows will be paid across the life of the deposit, and not immediately.
Question 4: Is the treatment of fixed rate loan commitments to retail counterparties clear and are there other instruments with retail counterparties where a behavioural approach to optionality should be taken?
"Non-Applicable"Question 5: Do respondents find that the required determination of the impact of a 25% increase in implicit volatility as described in Article 12 is operationally implementable?
We consider that options embedded in products accounted at amortised cost should be evaluated differently from explicit options and those embedded in fair valued instruments.The time value (extrinsic) of embedded options linked to balance items accounted at amortised cost should be excluded from IRRBB calculations, as it does not have an effect in capital. The inclusion of their extrinsic value in the risk metrics may generate undesirable volatility derived from out-of-the-money options (e.g., 0% rate floors in environments with high interest rates), confusing the risk management decisions. Moreover, the market value of these options is not reflected in Equity, and they are not managed with the intention to be traded.
Additionally, we don’t understand the purpose of treating differently bought and sold options.
Regarding the volatility shock for explicit options and those embedded in fair valued instruments, we consider that implementing these criteria may not be operationally complicated for sophisticated banks, but it will introduce a great complexity when combined with interest rate shocks scenarios, especially for the small and medium size entities. Moreover, it should be clearly stated that volatility shocks only apply for explicit options or those embedded in fair value instruments. A clarification on the purpose of these calculations would be appreciated, as vega risk is not commonly managed in embedded options for IRRBB.
In any case, further detailed criteria about the relevant hypothesis would be appreciated as it would help on its implementation and harmonization among banks. The valuation of optionality is sensitive to models and parametrization, and the lack of common criteria would lead to non-comparable results. Variables like volatility type (implicit vs historical), absolute or relative shifts, and whether the valuation model follows a normal or log normal distribution should be specified.
Although it is reasonable to consider different levels of volatility in each of the interest rate scenarios and adapt this parameter depending on the scenario considered, increasing volatility in all scenarios does not seem appropriate. What can be seen when analysing the historical series of implied volatility is that as the model (e.g., Black-Scholes) acts as a translator between volatility and price, an increase in interest rates automatically implies a decrease in the implied volatility observed in the market and the opposite when rates fall, so when creating volatility scenarios, it is also necessary to take the “model” effect into account. Applying a 25% change in any scenario may be far from what is expected from a market point of view.
We understand that the shift (25%) should be considered as multiplicative (stressed volatility = base volatility x 1,25) to better adapt to each type of model, but it should be specified in the text.
Additionally, a 25% shock may not be representative of all currencies (and if not, more instructions about how to calibrate it for other currencies, an even maturity would be necessary).
Finally, it should be clearly stated in the RTS that volatility shocks do not apply for SOT calculations.
Question 6: Do respondents find that the required slotting of repricing cash flows in accordance with the second dimension of original maturity/reference term as described in Article 13 is operationally implementable?
For fixed-rate instruments, the implementation of the slotting according to the original maturity at transaction level may be operationally burdensome for certain entities. We recommend applying a simplified approach by assigning the average original maturity at product level to reduce the operational workload.Question 7: Do respondents find it practical how the determination of several components of the NII calculation, with in particular the fair value component of Article 20 and the fair value component of automatic options of Article 15, is generally based on the processes used for the EVE calculation (in particular Article 16 and Article 12)?
From our view these positions should not be included, since they do not impact in margin following the accounting rules, in line with other variables like fees or commissions. This risk is captured with additional metrics but not with NII.Moreover, operationally, the calculation of the fair value component introduces a higher complexity in the metric due to the overlapping between the NII and the FV changes during the projection period (e.g., 12M). The calculation of FV changes not only requires excluding the instruments maturing or repricing during the first year. It also requires estimating the future MtM impact at the end of the projection period (t=1 year) on the fair value instruments with longer maturities (>1Y) which will be computationally very demanding. It must be considered that the residual maturity of these instruments will be 1 year shorter at the end of the NII projection period.
Question 8: Do respondents find that the calculation of the net interest income add-on for basis risk is reflective of the risk and operationally implementable
Conceptually, the basis risk calculation, as it is defined, includes the variations not only among different interest rate references, but also among different tenors of the same yield curve. These changes in the slope are directly related to the interest rate scenario evaluated and they should not be calibrated and evaluated separately. We consider that shocks should be consistent with the IRs scenario evaluated.Operationally, from our view the calculation of the net interest income add-on for basis risk is implementable but some additional clarification would be needed to consider such an add-on.
The basis risk is defined by taking into the reference terms: Overnight, 1 month, 3 months, 6 month or 12 months, and shocks are calculated comparing interest rates with the overnight reference.
To calculate the add-on, it is necessary to obtain the sensitivity to each reference rate, but when calculating the add-on, we note that several cases could be applied:
1. Applying the highest calibrated shock to the sensitivity to that reference rate.
2. Applying the largest calibrated shock to the total sensitivity of the entire portfolio.
3. Applying to each sensitivity the shock calculated for its reference rate.
We understand that the first option is the one referred in the standard method, but further clarification on this would be needed in the text.
Another matter we would like to raise is how the add-on for basis risk should be treated in a multicurve environment. The methodology seems to be written only for the case of a single curve.
Question 9: Do respondents find that the adjustments in the Simplified Standardised Approach as set out in Article 23 and 24 are operationally implementable, and do they find that any other simplification would be appropriate?
"Non-Applicable"Question 10: Do respondents find that all the necessary aspects are covered and the steps and assumptions for the evaluation of EVE and NII as laid out in the standardised approach and simplified standardised approach clear enough and operationally implementable?
"Non-Applicable"Upload files
AEB CP on IRRBB SA (EBA-CP-2021-38) + EX SUM.pdf
(141.84 KB)