Response to consultation on draft RTS on the assessment methodology under which competent authorities verify an institution’s compliance with the internal model approach
Go back
Hence, the Industry proposes to adjust the wording of Article 20 as follows:
(a) verify that the institution has regularly updated inventories, including:
(i) the difference between pricing functions or methods used in the internal-risk measurement model and the pricing functions or methods used to calculate the end-of-day value of the portfolio unless the institution demonstrates to the competent authorities that differences in pricing functions does not have a material impact on the results of the P&L attribution requirements as assessed in the requirement of paragraph (d);
Paragraphs (a)(iii) and (a)(iv) should then be adapted to refer to paragraph (a)(i). The industry would like to also remove paragraphs (a)(v) to (a)(vii) as the statements are unclear.
For example, there may be cases of desks managing their positions using similar hedging instruments (in agreement with their mandate). It should be clarified that using similar hedging instruments does not in itself mean that desks have similar positions or business objectives.
Foreign exchange and commodities in the banking book do not form actual desks and should be considered separately. Though their positions may be similar to those of desks in the IMA perimeter, banks should be given the choice to capitalise those exposures using the standardised approach.
The interest rates internal risk transfer (IR IRT) portfolios may contain positions that are similar to those of desks within the IMA perimeter. We believe that IR IRT portfolios should be granted an exception as their exposure is expected to be roughly flat. It would not justify the additional burden of having those portfolios under the IMA perimeter.
More generally, for sets of portfolios or desks which bear no material risks, the industry believes that an exception to the proposed rule should be granted.
There is also no consensus on the way these risks should be captured in market stress tests, therefore the Industry suggests removing any reference to environmental risk at this stage.
The industry is working on producing a conceptual framework for assessing climate risk within the trading book for the purposes of scenario analysis. This will be documented through the publication of a white paper due in early July 2023.
Further, as part of a paper that ISDA published during October 2022 on the challenges of Climate Risk Scenario Analysis for the Trading book , a survey was run to assess banks capabilities in this regard. Overall, banks qualified their climate risk scenario analysis capabilities as ‘basic’ or ‘evolving’.
As noted in the response to Question 3, ISDA is in the process of producing a conceptual framework for assessing climate risk in the trading book for the purposes of scenario analysis which will be published during July 2023. To help inform this paper we run another very comprehensive survey, and this still maintains that banks capabilities are either ‘basic’ or ‘evolving’.
As a consequence, the metric defined in article 21 may lead to various false positive or negative results when trying to detect outliers due to non-modellable risk factors.
Hence, we believe that this proposal would give unusable results and the MRF back-testing, as proposed and recommend that this should be removed.
Thus, the industry would suggest the deletion of this article 21 paragraph (a). Instead, the industry would advocate that the EBA provides flexibility to banks to choose their own additional internal model validation tests. They could, for instance, opt for one of the two approaches described thereafter to align the scope of VaR and P&L for a consistent back-testing. Such additional analysis may serve for the determination of the root cause of a regulatory back-testing overshooting: is it due to NMRF or could it be the result of a genuine internal model deficiency? There is no easy way to perform a consistent back-testing and each proposed approach have their pros and cons, but they still remain much preferable than the one of article 21(a).
(a) Option 1: To calculate an additional VaR that also includes non-modellable risk factors.
We note that the UK authorities (PRA) has proposed to allow the use of all risk factors within VaR for trading desk eligibility for IMA back-testing.
• Pros:
i. This back-tests a metric that can be used for risk management purposes.
ii. For some banks, the VaR inclusive of all risk factors is not overly burdensome.
• Cons:
i. For some banks computing a specific VaR, with the only objective to perform an additional back-testing, will be significantly costly in terms of IT (specific developments, CPU cost) and human resources (production and analysis of the new VaR figures).
ii. Use of time series with limited historical market data
(b) Option 2: To adjust HPL and APL by subtracting the P&L coming from NMRF.
The P&L from NMRF can be proxied as the difference between RTPL and RTPL_MRF (risk-theoretical changes in the institution’s portfolio’s value considering only changes to modellable risk factors). Hence, the back-testing can be performed on HPL_MRF and APL_MRF where HPL_MRF = HPL + RTPL_MRF – RTPL and APL_MRF = APL + RTPL_MRF – RTPL
• Pros:
i. This option is not overly burdensome to implement for some banks.
ii. It does not rely on NMRF data and risk calibration.
• Cons:
i. HPL_MRF and APL_MRF are proxies which may cast a doubt on the additional back-testing outcome, for instance if the risk pricers are simplified vs FO pricers
ii. A direct estimation of HPL_MRF and APL_MRF would be extremely difficult for some banks to implement because there is no full alignment between the risk factors used by FO in the APL and HPL and the model factors used in IMA. Even proxy estimates as proposed may be quite cumbersome since getting RTPL_MRF may be quite of a challenge.
Attempting to back-test one day returns instead also suffers from severe issues: At the present time, none of the methodologies available in scientific literature provide a robust way to directly back-test ES without significant assumptions on the P&L distribution, and all available methods still require a VaR back-test. Given that, any choice made would be inevitably challenged by both the internal model validation function as well as firms’ regulation authority as Article 21(b) states that the competent authority shall verify what drives the choice of the applied direct ES back-testing methodology as well as to analyze if the methodology is conceptually sound. It would also require extensive research to identify an approach that is most suitable on an individual bank basis.
Given the obstacles presented above, it is unclear what the benefit is of a direct back-test of the ES over more conventional tools such as VaR back-testing or full-distribution back-testing using quantile-quantile plots. Accurate computation of distribution quantiles ensures an accurate ES, and the tools available to back-test distribution quantiles are much more plentiful and accepted.
This approach is motivated by the fact that integrating all maturity mismatches within the core calculation of the DRC IMA would lead to significant economic inconsistency and even in a largely biased representation of the true underlying risk of default. Indeed, it is known that the liquidity issues in stressed conditions are much likely to happen on exotic credit or exotic equity positions than on stocks or listed equity derivatives that are highly liquid even in stressed market conditions. And yet with the “buy and hold until maturity” option, such mismatches between listed derivatives and stocks would be likely to generate the largest DRC impacts even though recent default of equities have demonstrated to have zero impact on an index future vs stock trading strategy (as the component where set to zero by the index provider with no liquidity impact).
Moreover, the “buy and hold until maturity” option could lead to very strong inconsistencies within the SA as the consideration of maturities is highly different (floor to 3 months). We also note that such positions would not generate any spot risk within the ES as the liquidity horizon is set to 10 days for large capitalization spot risk factors. Finally, even though those mismatches on most liquid positions are not a significant source of risk, some banks internal studies have shown that it could make the DRC IMA figures volatile due to the cyclicity of hedges rolling leading to difficult to explain variations of capital.
Then institutions are of opinion that the most consistent interpretation of the text is that the constant position assumption should be understood as a constant level of risk assumption and that the uncaptured maturity mismatch risk should be monitored through process that would take into account the liquidity of positions/instruments much like under the ES requirements.
The methodology used to establish a margin of conservatism (MoC) for internally derived PD could be very different from the PD MoC coming from the institution’s external sources, if that MoC is available at all (some external sources could provide simply a PD estimate with no expected range of estimation errors). This could make both options A and B hard to implement.
In the absence of an expected range of estimation errors, the institution could derive a crude estimation error from the externally sourced ladder of LGD by obligor grade (for example, by assuming that the LGD error at each ladder rung could be within the LGD average with the rung above and the rung below). Therefore, option C is preferred with the above mentioned potential additional requirement.
Q1. Do you agree with the provisions included in this chapter? Did you face challenges in complying with the governance chapter of the old RTS? If so, in which respect? Please elaborate.
Related to article 20:The industry recommends the requirements of Article 20(a) to be adjusted to avoid any overly burdensome efforts, as building an exhaustive inventory of all the pricing models used by the bank could be a significant challenge We suggest a review of the scope of this inventory to only include the internal-risk measurement model pricing functions that are different from the pricing functions used in the end-of-day valuation provided that the P&L Attribution Test shows high materiality impacts.Hence, the Industry proposes to adjust the wording of Article 20 as follows:
(a) verify that the institution has regularly updated inventories, including:
(i) the difference between pricing functions or methods used in the internal-risk measurement model and the pricing functions or methods used to calculate the end-of-day value of the portfolio unless the institution demonstrates to the competent authorities that differences in pricing functions does not have a material impact on the results of the P&L attribution requirements as assessed in the requirement of paragraph (d);
Paragraphs (a)(iii) and (a)(iv) should then be adapted to refer to paragraph (a)(i). The industry would like to also remove paragraphs (a)(v) to (a)(vii) as the statements are unclear.
Q2. What are your views in relation to the assessment methods relating to the requirements on the set-up of trading desks (see Article 7)? How do you plan to substantiate your choice of using internal models, in particular in the context of the requirement included in 325az(2), according to which the choice of not including a desk in the internal model shall not be motivated on the basis that the standardised approach requirements are lower compared to the internal model ones? Please elaborate.
With respect to article 7, paragraph (1)(f)(b): Trading desks are defined in agreement with their business mandate. As a result, cases of multiple desks managing similar positions should be rare. However, the occasional occurrence of two desks managing similar positions must be supported by a strong rationale and in some cases may justify a different approach in calculating own funds requirements. Hence, though in principle we agree with the EBA, we believe that supervisors should be granted the supervisory power to decide on the relevance of using different approaches for desks managing similar positions.For example, there may be cases of desks managing their positions using similar hedging instruments (in agreement with their mandate). It should be clarified that using similar hedging instruments does not in itself mean that desks have similar positions or business objectives.
Foreign exchange and commodities in the banking book do not form actual desks and should be considered separately. Though their positions may be similar to those of desks in the IMA perimeter, banks should be given the choice to capitalise those exposures using the standardised approach.
The interest rates internal risk transfer (IR IRT) portfolios may contain positions that are similar to those of desks within the IMA perimeter. We believe that IR IRT portfolios should be granted an exception as their exposure is expected to be roughly flat. It would not justify the additional burden of having those portfolios under the IMA perimeter.
More generally, for sets of portfolios or desks which bear no material risks, the industry believes that an exception to the proposed rule should be granted.
Q3. What are your views in relation to the requirement for credit institutions to specifically consider environmental risk as part of their stress-testing programme under the internal model approach? Do you agree that the assessment of that aspect should only apply from 2025? If not, by when EU credit institutions could be ready to be subject to this assessment? Please elaborate.
The Industry agrees on the fact that environmental risks should be carefully considered by banks as they could, for instance, deteriorate credit quality. However, there is no evidence on the fact that environmental risks (either transition or climate risks) could impact markets significantly in the short liquidity horizons usually seen in bank’s stress testing programme for traded market risk.There is also no consensus on the way these risks should be captured in market stress tests, therefore the Industry suggests removing any reference to environmental risk at this stage.
The industry is working on producing a conceptual framework for assessing climate risk within the trading book for the purposes of scenario analysis. This will be documented through the publication of a white paper due in early July 2023.
Q4. What is the status of credit institutions in relation to capturing environmental risks in their stress test for market risk under the internal model approach? Please elaborate.
As mentioned in the answer to question 3, the Industry is actively studying the integration of environmental risks in market risk. However, no consensus has been reached so far as banks continue to assess environmental risks within the trading business.Further, as part of a paper that ISDA published during October 2022 on the challenges of Climate Risk Scenario Analysis for the Trading book , a survey was run to assess banks capabilities in this regard. Overall, banks qualified their climate risk scenario analysis capabilities as ‘basic’ or ‘evolving’.
As noted in the response to Question 3, ISDA is in the process of producing a conceptual framework for assessing climate risk in the trading book for the purposes of scenario analysis which will be published during July 2023. To help inform this paper we run another very comprehensive survey, and this still maintains that banks capabilities are either ‘basic’ or ‘evolving’.
Q5. What is the status of credit institutions in relation to investigating whether environmental risk affects risk factor volatilities and/or the default risk? Are there credit institutions considering environmental (physical) risks as a form of event risk in their internal risk-measurement model? Please elaborate.
In addition to the response to our answer to question 4, some participants have studied the impact of environmental crisis / natural disaster on markets within a short-term trading horizon. It appears that the impact was minimal over a short-term horizon (short term horizons usually used to calibrate shocks for trading book stress tests), although clearly this is very much dependent on individual bank’s trading book composition and material exposure to vulnerable sectors, regions, or counterparties.Q6. What are your views on the provisions included in Article 21(a)? In particular, do you think that monitoring APL_MRF, and HPL_MRF is relevant for identifying potential deficiencies in the model? Please elaborate.
Due to the weaknesses of the APL_MRF and HPL_MRF (as has been highlighted in the answer to question 7), the Industry does not see any advantage emerging from the use those metrics to identify deficiencies in the model.Q7. Do you think that the scaling proposed in APL_MRF and HPL_MRF in Article 21(a) could lead to frequent numerical issues (e.g. due to a denominator, i.e. RTPL, that is close to zero)? Please elaborate.
The industry believes that the proposed approach of using the ratio of RTPL_MRF and RTPL to scale APL and HPL, respectively, to assess whether the outliers are due to NMRF is not conceptually sound. In fact, it assumes that APL (HPL) in relation to APL_MRF (HPL_MRF) behaves in the same way as RTPL to RTPL_MRF which cannot be expected.As a consequence, the metric defined in article 21 may lead to various false positive or negative results when trying to detect outliers due to non-modellable risk factors.
Hence, we believe that this proposal would give unusable results and the MRF back-testing, as proposed and recommend that this should be removed.
Q8. What could be alternative definitions of APL_MRF and HPL_MRF in Article 21(a) that could provide an estimate of the contribution of modellable risk factors? Please elaborate.
The EBA proposal requires significant additional IT developments, which would not lead to sensible results. Together with other costly additional requirements that the EBA would like to impose on IMA banks (ex. ES back-testing), it questions the sustainability of the internal model approach and might generate level-playing field issues, as banks from other jurisdictions might not be subject to such expectations.Thus, the industry would suggest the deletion of this article 21 paragraph (a). Instead, the industry would advocate that the EBA provides flexibility to banks to choose their own additional internal model validation tests. They could, for instance, opt for one of the two approaches described thereafter to align the scope of VaR and P&L for a consistent back-testing. Such additional analysis may serve for the determination of the root cause of a regulatory back-testing overshooting: is it due to NMRF or could it be the result of a genuine internal model deficiency? There is no easy way to perform a consistent back-testing and each proposed approach have their pros and cons, but they still remain much preferable than the one of article 21(a).
(a) Option 1: To calculate an additional VaR that also includes non-modellable risk factors.
We note that the UK authorities (PRA) has proposed to allow the use of all risk factors within VaR for trading desk eligibility for IMA back-testing.
• Pros:
i. This back-tests a metric that can be used for risk management purposes.
ii. For some banks, the VaR inclusive of all risk factors is not overly burdensome.
• Cons:
i. For some banks computing a specific VaR, with the only objective to perform an additional back-testing, will be significantly costly in terms of IT (specific developments, CPU cost) and human resources (production and analysis of the new VaR figures).
ii. Use of time series with limited historical market data
(b) Option 2: To adjust HPL and APL by subtracting the P&L coming from NMRF.
The P&L from NMRF can be proxied as the difference between RTPL and RTPL_MRF (risk-theoretical changes in the institution’s portfolio’s value considering only changes to modellable risk factors). Hence, the back-testing can be performed on HPL_MRF and APL_MRF where HPL_MRF = HPL + RTPL_MRF – RTPL and APL_MRF = APL + RTPL_MRF – RTPL
• Pros:
i. This option is not overly burdensome to implement for some banks.
ii. It does not rely on NMRF data and risk calibration.
• Cons:
i. HPL_MRF and APL_MRF are proxies which may cast a doubt on the additional back-testing outcome, for instance if the risk pricers are simplified vs FO pricers
ii. A direct estimation of HPL_MRF and APL_MRF would be extremely difficult for some banks to implement because there is no full alignment between the risk factors used by FO in the APL and HPL and the model factors used in IMA. Even proxy estimates as proposed may be quite cumbersome since getting RTPL_MRF may be quite of a challenge.
Q9. What are your views in relation to the assessment method to verify that the internal validation process includes a direct back-testing of the expected shortfall, as per Article 21(b)? Do you expect this requirement to put significant burden on institutions? Which of the methods available in the literature do you expect credit institutions to use to back-test their expected shortfall? Please elaborate.
The industry believes that direct back-testing of ES would create a significant additional operational burden for participating institutions with no straightforward benefit beyond what is already provided by VaR back-testing using one day returns. In fact, IMCC is not back-testable in its fully aggregated form, as it is an aggregation of distinct metrics calculated on multiple liquidity horizons and asset classes. Even if the diversified ES with a 10-day liquidity horizon was considered in isolation, one would require a corresponding buy & hold P&L on a 10-day horizon which would be challenging from a calculation and operational point of view . We also note that this is not required for the current capital calculation requirements for 10-day VaR calculations.Attempting to back-test one day returns instead also suffers from severe issues: At the present time, none of the methodologies available in scientific literature provide a robust way to directly back-test ES without significant assumptions on the P&L distribution, and all available methods still require a VaR back-test. Given that, any choice made would be inevitably challenged by both the internal model validation function as well as firms’ regulation authority as Article 21(b) states that the competent authority shall verify what drives the choice of the applied direct ES back-testing methodology as well as to analyze if the methodology is conceptually sound. It would also require extensive research to identify an approach that is most suitable on an individual bank basis.
Given the obstacles presented above, it is unclear what the benefit is of a direct back-test of the ES over more conventional tools such as VaR back-testing or full-distribution back-testing using quantile-quantile plots. Accurate computation of distribution quantiles ensures an accurate ES, and the tools available to back-test distribution quantiles are much more plentiful and accepted.
Q10. What are your views in relation to the requirement included in this Article (i.e. Article 23) on sufficient risk coverage? Do you agree that institutions should monitor the impact of the exclusion of some risk factors from the internal model? Please elaborate.
Following Article 325bh(1)(a), validating the perimeter of risk factors should be strongly related to the results of P&L attribution tests which already constitutes a significant challenge. The consultation paper seems to ignore this fundamental axis, advocating in favor of an exhaustive list of risk factors. Our proposal is to adjust this article following the spirit of the CRR by conditioning on PLAT materiality and not monitoring non-material risk factors from a PLAT perspective.Q11. Do you agree with the provisions included in Article 24 and the relevant assessment techniques to verify that interest rate risk is properly captured? Do you think there are additional aspects that should be covered and/or assessed? Please elaborate.
In addition to the overall recommendation in section 3.1.1.2, the industry would like to highlight that Article 24 suggests providing a full inventory of sensitivities at market parameter level for general interest rates with all related risk factors which represents a strong operational burden. This approach gold-plates the wording of the CRR2 (Article 325bh(1)(c)) which mentions, using a principle-based approach, that all sensitive positions should be mapped to a relevant set of risk factors. Besides, point (c) of article 24 does not include any materiality constrain, potentially implying that all bases, material or not, should be captured by the model. We suggest introducing the materiality constrain in this article.Q12. Do you agree with the provisions included in Article 25 and the relevant assessment techniques to verify that equity risk is properly captured? Do you think there are additional aspects that should be covered and/or assessed? Please elaborate.
In addition to the overall recommendation in section 3.1.1.2, the industry would like to highlight that Article 25 suggests providing a full inventory of sensitivities at market parameter level for equity positions with all related risk factors which represents a strong operational burden. This approach gold-plates the wording of the CRR2 (Article 325bh(1)(e)) which mentions, using a principle-based approach, the strong link between the materiality constrain (significant positions + materiality) and sophistication of the modelling approach. This logic is not retained in article 25 which opts in favor of an inventory approach disconnected from materiality constrains. We suggest introducing the materiality constrain in this article.Q13. Do you agree with the provisions included in Article 26 and the relevant assessment techniques to verify that credit spread risk is properly captured? Do you think there are additional aspects that should be covered and/or assessed? Please elaborate.
In addition to the overall recommendation in section 3.1.1.2, the industry would like to highlight that No specific sub-articles were developed in Article 325bh(1) for credit spread risk factors thus this article should be linked to general sub-article Article 325bh(1)(a) and should be strongly related to any impact on P&L attribution requirements. Instead, the same inventory logic is retained and is decorrelated from the logic developed in CRR2.Q14. Do you agree with the provisions included in Article 27 and the relevant assessment techniques to verify that foreign-exchange risk is properly captured? Do you think there are additional aspects that should be covered and/or assessed? Please elaborate.
In addition to the overall recommendation in section 3.1.1.2, the industry would like to highlight that Article 27 suggests providing a full inventory of sensitivities at market parameter level for foreign-exchange positions and is decorrelated from P&L attribution requirements. Besides, the specific risk of unpegging events is added with no connection to Article 325bh(1)(d). Unpegging events are usually not market driven and should not be included in the market risk framework.Q15. Do you agree with the provisions included in Article 28 and the relevant assessment techniques to verify that commodity risk is properly captured? Do you think there are additional aspects that should be covered and/or assessed? Please elaborate.
The industry would like to highlight the recommendation in section 3.1.1.2.Q16. What are your views on assessment techniques laid down in Article 29 and 30? Do you see alternative or additional techniques that could be introduced to assess whether the modelling of curves and surfaces is accurate? Please elaborate.
The industry would like to highlight the recommendation in section 3.1.1.2.Q17. Do you agree with the provisions included in Article 31 relating to the inclusion of implied correlation risk factors? Please elaborate.
In addition to the overall recommendation in section 3.1.1.2, the industry would like to highlight that If the implied correlation parameters do not impact the P&L attribution requirements, then the materiality constrain should prevent from forcing the institution to model this risk factor.Q18: Do you agree with the assessment techniques included in Article 32? Please elaborate.
We have not provided a response for this question.Q19: How do you expect institutions/third parties to determine that the volume of a transaction or a quote is non-negligible, and that the bid-offer spread does not substantially deviate from current market conditions? How do you expect institutions to determine that there is a close relationship between the verifiable price and the risk factor to which this price is mapped? Please elaborate.
We have not provided a response for this question.Q20. Do you agree with the provisions included in Article 33 aiming at assessing a sound implementation of the requirements relating to the risk factor mapping to the appropriate liquidity horizons? Please elaborate.
We have not provided a response for this question.Q21. Do you think that institutions would face challenges in providing the details referred to in paragraph 1 at risk-factor level? In particular, as regards data inputs to mark the risk factor (see paragraph 1(c)), do you think institutions would face challenges in providing a high-level/simple description of data inputs in order to verify that the RF is mapped to the appropriate (sub-)category as per Article 1 of RTS on LH? Please elaborate.
We have not provided a response for this question.Q22. Do you think that institutions will make use of the derogation referred to in Article 325bd(3) of Regulation (EU) No 575/2013 regarding the use of longer liquidity horizons? Please elaborate.
We have not provided a response for this question.Q23. Do you agree with the assessment techniques proposed in Article 34? What could be alternative techniques for assessing whether a proxy is conservative and keeps track of the actual position held? Please elaborate.
We have not provided a response for this question.Q24. What could be reasons why the institution decides to use a proxy approach to determine the stress scenario risk measure for a non-modellable risk factor despite N being above 12? Please elaborate.
We have not provided a response for this question.Q25. What are your views on the provisions and techniques included in Article 35? Do you consider the indicators included therein adequate? What could be alternative or additional indicators? Please elaborate.
We have not provided a response for this question.Q26. Do you agree with the provisions and the assessment techniques included in Article 36 dealing with back-testing? Please elaborate.
We have not provided a response for this question.Q27. What is your view in the relation to the analysis that institutions are to perform for their top-of-the-house overshootings in accordance with Article 37?
We have not provided a response for this question.Q28. Do you agree with the provisions and the assessment techniques included in Article 38 dealing with the profit and loss attribution requirements? Please elaborate.
We have not provided a response for this question.Q29: do you agree with the assessment methods included in Article 39 for verifying a sound implementation of the requirements on the calculation of own funds requirements for FX and commodity risk? If not, what would be alternative proposals?
We have not provided a response for this question.Q30: in accordance with the assessment methods referred to in paragraph 3 and paragraph 5 (of Article 39), institutions are expected to perform reconciliation exercises. Do you think that institutions may face difficulties in performing those reconciliations? If so, which? What would be alternative methods to assess the corresponding requirements?
We have not provided a response for this question.Q31: do you consider the term “valuation inputs” appropriate to define inputs that form an accounting value? If not, what would be an alternative terminology on which the assessment method referred to in paragraph 7 (of Article 39) should be built?
We have not provided a response for this question.Q32. Do you agree with the assessment techniques relating to non-linearities? Please elaborate.
The Industry would like to highlight that the ability for the model to capture non-linearities is already measured in the P&L attribution test. We believe that in most of the cases, using first order sensitivities is sufficient. Consequently, we suggest removing any reference to “second-order terms of Taylor series approximations” in Article 40. The use of second-order sensitivities could be significantly burdensome as it would require the calculation of those sensitivities even if they are not material.Q33. What are your views in relation to the assessment techniques included in Articles 41 and 42 dealing with aspects relating to Article 325bb and 325bc CRR? Please elaborate.
We have not provided a response for this question.Q34. Do you have comments on the analysis included in the background section and in Annex I relating to ES and VaR estimators? Please elaborate.
We have not provided a response for this question.Q35. Do you agree that the RTS should require competent authorities to verify that the estimators used are concordant? Please elaborate.
We have not provided a response for this question.Q36. How competent authorities should verify that the estimators used are reasonably accurate in measuring risks? Do you agree in adding a list of estimators ranked by conservativeness as that provided in Annex I? Please elaborate. Please elaborate.
We have not provided a response for this question.Q37. Do you have comments on the examples of concordant ES and VaR estimators proposed that can be used by institutions in their calculations? Please elaborate.
We have not provided a response for this question.Q38. Do you agree with the provisions included in this Article (Article 43)? Are banks fully relying on an historical approach able to fulfil the requirement in Article 12(d) of these RTS, i.e. how in practice they could test alternative correlation patterns than those historically observed (please note that this provision was already included in the ‘old’ draft RTS on internal model for market risk)?
We have not provided a response for this question.Q39. Do you agree with the assessment methods included in this Article (i.e. Article 44)?
We have not provided a response for this question.Q40. Paragraph 5 of this Article (i.e. Article 44) implies that an institution is able to retrieve data used for calibrating a shock in the ES for a risk factor that was modellable. This provision has been built on the requirement in Article 3(2) [RTS on SSRM]. Hence, banks are expected to keep those data in their systems. Do you agree with the requirement, or do you think that institutions may face difficulties in making this operational? Please elaborate.
We have not provided a response for this question.Please elaborate.
We have not provided a response for this question.Q42. When institutions face pricing failures, or when they are to identify the stress period, institutions are allowed to use sensitivity-based P&Ls. Do you consider that the assessment methods proposed to assess the accuracy of sensitivity-based calculations are appropriate? Or do you think that additional/alternative checks should be envisaged? Please elaborate.
We have not provided a response for this question.Q43. What are your views in relation to the requirements included in this chapter (i.e. chapter 4) dealing with the internal default risk model?
We have not provided a response for this question.Q44. Institutions are required to use two systematic risk factors to model issuers' defaults, as per Article 325bp(1) CRR. Do you agree that this requirement implies that single issuers cannot be modelled directly (by calibrating the correlation structure without the use of a factor model)? Do you agree that this requirement allows factor models based on principal components analysis to be used? Please elaborate.
We have not provided a response for this question.Q45. How do you think EU credit institutions will model PDs relating to equities? Are there specific requirements you think should be included in these RTS in relation to equity issuers' PDs? Please elaborate.
We have not provided a response for this question.Please elaborate.
We have not provided a response for this question.Q47. How institutions are going to implement the constant position assumption? Are there cases where the model set-up does not allow to easily capture the risk deriving from maturity mismatches, and if so, how institutions in those cases monitor such a risk? Please elaborate.
The industry feels that competent authorities will expect banks to have implemented the constant position assumption as a constant level of risk assumption much like under Basel 2.5 IRC. Then the authority should expect that the uncaptured maturity mismatch risk is monitored by integrating an additional specific criterion related to the liquidity of instruments.This approach is motivated by the fact that integrating all maturity mismatches within the core calculation of the DRC IMA would lead to significant economic inconsistency and even in a largely biased representation of the true underlying risk of default. Indeed, it is known that the liquidity issues in stressed conditions are much likely to happen on exotic credit or exotic equity positions than on stocks or listed equity derivatives that are highly liquid even in stressed market conditions. And yet with the “buy and hold until maturity” option, such mismatches between listed derivatives and stocks would be likely to generate the largest DRC impacts even though recent default of equities have demonstrated to have zero impact on an index future vs stock trading strategy (as the component where set to zero by the index provider with no liquidity impact).
Moreover, the “buy and hold until maturity” option could lead to very strong inconsistencies within the SA as the consideration of maturities is highly different (floor to 3 months). We also note that such positions would not generate any spot risk within the ES as the liquidity horizon is set to 10 days for large capitalization spot risk factors. Finally, even though those mismatches on most liquid positions are not a significant source of risk, some banks internal studies have shown that it could make the DRC IMA figures volatile due to the cyclicity of hedges rolling leading to difficult to explain variations of capital.
Then institutions are of opinion that the most consistent interpretation of the text is that the constant position assumption should be understood as a constant level of risk assumption and that the uncaptured maturity mismatch risk should be monitored through process that would take into account the liquidity of positions/instruments much like under the ES requirements.
Q48. Do you agree that the requirement in Article 325bp(2) CRR should be read as applicable to LGD only? How it could be argued that a IRB-LGD is already reflecting the economic cycle, and it already dependent on the systematic risk factors? Please elaborate.
The statement in Article 325bp(2) could be interpreted in our view as: both PD and LGD can reflect the economic cycle. Whether or not PD should reflect the economic cycle could be left for the financial institutions to decide. In the current implementation of the IMA DRC model, if the economy is in a downturn, i.e., when systematic risk factors are lower, issuers are more likely to default since their ability to pay, which depends positively on the systematic risk factors, would be lower and more likely to fall below the default threshold. Additionally, in case of default, the LGD distribution (whose unconditional average is the IRB-LGD) depends negatively on the systematic risk factors, which implies that lower systematic risk factors would lead to greater losses under default.Q49. How institutions, in their on-going implementation of model, plan to meet the requirement in Article 325bp(2) CRR? Please elaborate.
The current IMA DRC model, with its 2 stochastic processes for ability to pay process and LGD process, already reflects the requirements in its current state.Q50. What is your favourite option among options A, B, C as presented in Article 48(2)(k)(iii)? What are the challenges that an institution would face in implementing option A? and option B? Please elaborate.
The preferred option for the industry is option C. It is the preferred option with the potential additional requirement of conducting a sensitivity analysis similar to that required under the ‘fallback’ PD methodology, where institutions are to conduct a sensitivity analysis and scenario analysis to assess the qualitative and quantitative reasonableness of the ‘fallback’ PD approach, when neither external sources nor an IRB-like internal methodology are available for producing PD estimates.The methodology used to establish a margin of conservatism (MoC) for internally derived PD could be very different from the PD MoC coming from the institution’s external sources, if that MoC is available at all (some external sources could provide simply a PD estimate with no expected range of estimation errors). This could make both options A and B hard to implement.
In the absence of an expected range of estimation errors, the institution could derive a crude estimation error from the externally sourced ladder of LGD by obligor grade (for example, by assuming that the LGD error at each ladder rung could be within the LGD average with the rung above and the rung below). Therefore, option C is preferred with the above mentioned potential additional requirement.