Response to consultation on RTS on Assessment Methodology for Market Risk Internal Models
Go back
A consistent treatment should be applied for back-testing purposes.
This seems to be the most appropriate choice, for the sake of consistency:
- between the IMA and the standardized approaches (since an institution’s own creditworthiness is beyond the scope of the standardized charge)
- between the numerator and the denominator of the capital ratio (since the effects of an institution’s own creditworthiness are excluded from own funds).
As a general matter, the EBA should promote the adoption of consistent treatment for an institution’s own creditworthiness throughout the Basel capital framework, at least in those parts that are currently being redrafted (such as the CVA risk).
Since the Basel Committee only considers application at consolidated level, a profile that needs specific attention by EBA relates to the definition of “own”. In ABI’s opinion, it should be clarified that the scope of the definition of “own” includes all entities within the prudential consolidation, or entities that are fully consolidated and fall under the same jurisdiction.
Anyway, appropriate criteria for defining “complexity” should be identified in the light of the specific measures to which the definition applies. In that regard, ABI argues that, in the draft RTS, the application of the different measures to the three categories is quite vague.
Moreover, ABI deems that the introduction of a definition of “simple” vs “complex” (instruments or models) deserves careful consideration since, once established, this definition could be referred to in further regulations.
The request that a model has to remain unchanged for one year appears particularly penalizing. Moreover, the period between application and authorization can be used in order to collect the observations needed prior to approval.
In ABI’s opinion, the observation period before application could be reduced from one year to 6 months.
ABI acknowledges that the methodology for excluding certain risk factors from the P&L may represent a regulatory issue, as the EBA might not be willing to leave banks excessive room for discretionary approaches. To this end, the industry can work together with the EBA to define appropriate solutions.
Anyway, ABI does not in principle criticize the use of P&L including all risk factors for reporting to senior management or to the supervisor. ABI rejects its use for back-testing purposes, determining automatic effects on the capital charge.
ABI rejects the assumption that gaps between the outcome of the IMA (that can exclude some risk factors) and the P&L including all risk factors are manifestations of model flaws.
Given that the outcomes of back-testing affect calculation of the own funds requirements, the inclusion of risk factors outside the scope of IMA could give rise to a double counting issue.
If the inclusion of risk factors outside the scope determined a higher multiplier for the IMA, these risk factors would be covered twice: by the increase of the multiplier in the IMA and by the standardized capital charge (that the bank applies since these risk factors are outside the scope of IMA).
It seems that the CRR does not explicitly ask EBA for setting out a definition for net interest income. ABI suggests that EBA deletes Article 40(5)(e) of the draft RTS and activates dialogue with banks in order to identify a definition to be introduced upon transposition of the FRTB into EU law.
An alternative methodology, for the intra-day risk monitoring, would be to produce ad hoc reporting for the only intra-day component in terms of statistics (of the historical series), in order to isolate the effect and demonstrate its role compared to the overall profitability.
Making the use of stochastic correlations mandatory would introduce an unnecessary, burdensome task for banks, with a short time horizon.
Furthermore, the use of stochastic correlations is questionable from the theoretical point of view and poses further modelling issues (formulation of a consistent stochastic process for the correlations among several risk factors), implementation issues (calibration of the parameters of these stochastic processes) and practice (numerical simulation could be required, Montecarlo square-like computational issues).
It should also be considered that the FRTB addresses the issue of “correlation risk”, adopting a different solution (constraints on the effects of correlations).
VaR tests using (even historical) correlation scenarios require important operational efforts. We suggest to identify historical scenarios of high and low correlations on the basis of a restricted set of risk factors, suitably selected for their relevance using risk factor sensitivities.
In addition, with reference to Article 54(3) of the draft RTS, ABI would appreciate clarification about the meaning of “materiality” of the proxy in risk measure, given that it cannot be assessed against the “true” value (which is not available if a bank needs to use a proxy).
In the light of the above, ABI thinks that no modifications of IRC models should be requested in these RTS, before implementation of the FRTB. In that respect, the FRTB says that “PDs implied from market prices are not acceptable unless they are corrected to obtain an objective probability of default” (1).
ABI also notes that, in addition to theoretical reasoning, technical issues have to be taken into account. In fact, internal models do not always provide transition matrices and, above all, AIRB models usually do not cover the entire scope of the issuers of trading book instruments.
(1) Basel Committee on Banking Supervision, Minimum capital requirements for market risk, January 2016, page 61.
Q1: What are stakeholders’ views regarding the two proposed interpretations for the capture or exclusion of an institution’s own creditworthiness as a risk factor in internal models (non-default only), and consistent treatment for back-testing purposes?
In ABI’s opinion an institution’s own creditworthiness should not in principle be captured as a risk factor in internal models.A consistent treatment should be applied for back-testing purposes.
This seems to be the most appropriate choice, for the sake of consistency:
- between the IMA and the standardized approaches (since an institution’s own creditworthiness is beyond the scope of the standardized charge)
- between the numerator and the denominator of the capital ratio (since the effects of an institution’s own creditworthiness are excluded from own funds).
As a general matter, the EBA should promote the adoption of consistent treatment for an institution’s own creditworthiness throughout the Basel capital framework, at least in those parts that are currently being redrafted (such as the CVA risk).
Since the Basel Committee only considers application at consolidated level, a profile that needs specific attention by EBA relates to the definition of “own”. In ABI’s opinion, it should be clarified that the scope of the definition of “own” includes all entities within the prudential consolidation, or entities that are fully consolidated and fall under the same jurisdiction.
Q2: What is industry current practice in this regard for VaR, SVaR and IRC?
The institution’s own creditworthiness is currently included in the scope of internal models (for IRC, only with respect to migration risk).Q3: What are the main operational challenges?
The operational effort required to exclude an institution’s own creditworthiness seems reasonable. No major challenges have been identified.Q4: Do stakeholders agree with the General-Specific model application hierarchy introduced by the RTS?
ABI agrees with the proposed hierarchy, which reflects current practices. Additionally, ABI observes that strict criteria should be provided for identifying the risk factors pertaining respectively to “General” and to “Specific” risk, since some cases may still be ambiguous (in particular with respect to equity).Q5: Do Stakeholders consider that the categories of instruments listed above provide an appropriate guide to assess the complexity of an internal model?
In principle, ABI deems that the complexity of an internal model cannot be properly assessed just by taking into account the complexity of the instruments’ payoff. The features of the risk factors modeled (liquidity and market risk factors involved in pricing, at least) should also be considered.Anyway, appropriate criteria for defining “complexity” should be identified in the light of the specific measures to which the definition applies. In that regard, ABI argues that, in the draft RTS, the application of the different measures to the three categories is quite vague.
Moreover, ABI deems that the introduction of a definition of “simple” vs “complex” (instruments or models) deserves careful consideration since, once established, this definition could be referred to in further regulations.
Q6: Do stakeholders agree with the use of two differentiated approaches for general and specific risk to assess the significance of positions included in the scope of the model?
ABI acknowledges that the EBA has to define criteria in order to fulfill the CRR mandate, even though it is likely that they will only be applied in a very few cases before the FRTB enters into force (introducing a different approach). The proposed solutions seem reasonable, provided that the EBA sets out clear definitions of “long” and “short” net positions for the purpose of Article 12.Q8: Do stakeholders agree with the two metrics required to assess regularly the relevance of positions excluded from the scope of the internal model?
ABI considers the proposed approach to be quite effective.Q9: What are stakeholders views regarding the proposed requirements on the internal committee structure?
ABI agrees with the EBA proposal, reflecting the current practice of Italian banks.Q10: Do stakeholders agree that the internal validation requirements are relevant and capture all material risks?
ABI agrees.Q11: Are there any missing elements that should be incorporated or current elements that may be too burdensome?
ABI has not identified any elements that are missing or too burdensome with respect to the internal validation requirements.Q12: Do stakeholders agree that the proposed requirements on limit structure, regular limit update and limit breach approval processes are appropriate?
ABI deems the proposed requirements appropriate.Q13: Do stakeholders agree with the rationale to provide some flexibility for the introduction of new products?
ABI agrees and welcomes the EBA approach.Q14: What are stakeholders’ views regarding the specific limitations introduced in the RTS regarding the delegation of authority to the new product committee?
ABI agrees.Q15: Do stakeholders agree that the model should have been working in a stable way during a minimum period of 250 days prior to application for permission to use the model?
In ABI’s opinion, requiring a one-year history before applying would result in excessive delay.The request that a model has to remain unchanged for one year appears particularly penalizing. Moreover, the period between application and authorization can be used in order to collect the observations needed prior to approval.
In ABI’s opinion, the observation period before application could be reduced from one year to 6 months.
Q16: Do stakeholders agree that the results obtained for the portfolios published by the EBA during this period are useful for validation purposes?
ABI agrees, provided that only portfolios with products actually traded by a bank are taken into account.Q17: Do stakeholders agree with the requirements related to the model accuracy track record and Stress Testing programme?
While in principle sharing the EBA objectives, ABI argues that the Stress Testing programme seems oversized, since the capital charge includes the SVaR as a countercyclical component.Q18: Do Stakeholders have any additional comments or concerns regarding the requirements outlined in the governance section?
ABI does not have any additional comments.Q19: What are stakeholders' views on the proposed requirements for the computation of VaR and PandL at consolidated level?
ABI does not have any comments.Q20: Do stakeholders’ agree with the distinction between ‘global’ and ‘local’ price risk factors?
ABI agrees, provided that “local” prices are recognized for accounting purposes.Q21: What are stakeholders’ views on the burden a more frequent update than monthly creates?
In current practice by Italian banks, data sets are updated daily.Q22: For “partial use” IMA, do you agree with the use of a hypothetical PandL calculated from mark to market PandL including all pricing factors of the portfolio´s positions?
In order for “hypothetical” back-testing to represent a genuine benchmark for assessing the adequacy of a model, it should consider the P&L stemming only from the risk factors included in the scope of IMA.ABI acknowledges that the methodology for excluding certain risk factors from the P&L may represent a regulatory issue, as the EBA might not be willing to leave banks excessive room for discretionary approaches. To this end, the industry can work together with the EBA to define appropriate solutions.
Q23: If your answer to Q22 is no, what impact does this have on the PandL used for back-testing purposes and how do you monitor the appropriateness of the model? Are there alternatives to ensure a proper reporting to senior management?
In banking practice, several analyses already reported to senior management can be considered suitable for this goal.Anyway, ABI does not in principle criticize the use of P&L including all risk factors for reporting to senior management or to the supervisor. ABI rejects its use for back-testing purposes, determining automatic effects on the capital charge.
ABI rejects the assumption that gaps between the outcome of the IMA (that can exclude some risk factors) and the P&L including all risk factors are manifestations of model flaws.
Q24: What are stakeholders’ views regarding the relative merits of the inclusion of all risk factors for the actual PandL computation?
Even for “actual” back-testing purposes, the performance of the model should be assessed against the P&L stemming from risk factors already included in the scope of IMA.Given that the outcomes of back-testing affect calculation of the own funds requirements, the inclusion of risk factors outside the scope of IMA could give rise to a double counting issue.
If the inclusion of risk factors outside the scope determined a higher multiplier for the IMA, these risk factors would be covered twice: by the increase of the multiplier in the IMA and by the standardized capital charge (that the bank applies since these risk factors are outside the scope of IMA).
Q25: What are stakeholders’ views regarding the proposed definition of ‘Net interest income’?
ABI highlights that setting a definition for net interest income is not trivial and that a modification of current methodologies – approved by the supervisor – would be burdensome for banks.It seems that the CRR does not explicitly ask EBA for setting out a definition for net interest income. ABI suggests that EBA deletes Article 40(5)(e) of the draft RTS and activates dialogue with banks in order to identify a definition to be introduced upon transposition of the FRTB into EU law.
Q26: What are stakeholders’ views regarding the requirement to assess the importance of intraday and new trades to determine the VaR and SVaR multipliers?
In ABI’s opinion, the VaR methodology is not suitable for addressing all the peculiarities of intra-day trading and, hence, it is not proper to address the related risks using the VaR and SVaR multipliers.Q27: What alternative methodology, if any, might be appropriate to capture this intra-day risk?
ABI considers that, to the extent possible, the risk is captured through the consideration of “actual” P&L, including the results of intra-day trading.An alternative methodology, for the intra-day risk monitoring, would be to produce ad hoc reporting for the only intra-day component in terms of statistics (of the historical series), in order to isolate the effect and demonstrate its role compared to the overall profitability.
Q28: What are stakeholder’s practices regarding adjustments computed less regularly than daily?
Different practices are in place. Anyway, the approach proposed in Article 40(12)(a) of the draft RTS seems reasonable.Q29: What are stakeholders’ views regarding the treatment of Theta in VaR and as a component of PandL?
ABI thinks it important, in principle, that the treatment of Theta be consistent between the VaR and the P&L (included in or excluded from both, according to modeling assumptions). In practice, Italian banks calculate VaR assuming instantaneous shocks – in accordance with the Basel rule – and therefore support the exclusion of Theta.Q30: Taking into account the CRR requirement to capture ‘correlation risk’ do you consider that the use of stochastic correlations should be required?
ABI thinks that the use of stochastic correlations should not be required.Making the use of stochastic correlations mandatory would introduce an unnecessary, burdensome task for banks, with a short time horizon.
Furthermore, the use of stochastic correlations is questionable from the theoretical point of view and poses further modelling issues (formulation of a consistent stochastic process for the correlations among several risk factors), implementation issues (calibration of the parameters of these stochastic processes) and practice (numerical simulation could be required, Montecarlo square-like computational issues).
It should also be considered that the FRTB addresses the issue of “correlation risk”, adopting a different solution (constraints on the effects of correlations).
Q31: Do stakeholders agree with the additional requirements introduced for banks using empirical correlations?
We agree with the requirement to review empirical correlations at a minimum monthly: in the current practice of Italian banks, data sets and consequently (empirical) correlations between risk factors are updated daily.VaR tests using (even historical) correlation scenarios require important operational efforts. We suggest to identify historical scenarios of high and low correlations on the basis of a restricted set of risk factors, suitably selected for their relevance using risk factor sensitivities.
Q34: Do you agree that the SVaR multiplier should always be the same or higher than the one used for VaR purposes?
ABI does not agree. It cannot be taken for granted that weaknesses in the VaR models (that give rise to a higher multiplier) are reflected in the SVaR, as some flaws might not affect the stressed calculation.Q35: Do Stakeholders have any additional comments or concerns regarding the requirements outlined in the VaR section?
ABI does not have any additional comments.Q36: Do stakeholders consider that any proxy validated for VaR should be acceptable for SVaR purposes?
Yes, in ABI’s opinion, if a proxy is considered suitable for VaR purposes, it should be considered suitable for SVaR purposes as well.Q37: Do Stakeholders have any additional comments or concerns regarding the rest of requirements outlined in the Stressed VaR sub-section?
As for the back-testing add-on to ms, an overshooting of VaR that does not cause an overshooting of SVaR should not be taken into account for SVaR addend purposes.In addition, with reference to Article 54(3) of the draft RTS, ABI would appreciate clarification about the meaning of “materiality” of the proxy in risk measure, given that it cannot be assessed against the “true” value (which is not available if a bank needs to use a proxy).
Q38: Do stakeholders agree with the EBA interpretation regarding the treatment of event risk for credit positions after the implementation of IRC?
ABI agrees insofar as credit positions are concerned.Q39: What are stakeholders’ views regarding the capture of the FX position stemming from Banking Book activities and the treatment proposed in the RTS?
ABI agrees with the proposed approach. Anyway, ABI observes that making reference to data based on observations over the previous year can lead to misleading results. Since banks provide quarterly and monthly reporting for accounting and prudential purposes, ABI suggests that reference should be made to more updated values (i.e. single observations or average figures).Q40: Do Stakeholders consider appropriate the requirements established in this Article regarding the constant level of risk and constant position assumptions?
ABI considers the requirements to be appropriate.Q41: Do stakeholders agree that internally-derived ratings shall be prioritised for IRC?
ABI does not agree. We believe external ratings, provided by rating agencies, would be preferable.Q42: Do you consider that PDs derived from spreads or external ratings are more appropriate for IRC modelling than those internally-derived?
ABI remarks that the Basel framework is ambiguous about whether “market” PDs or “internal” PDs are to be preferred, and hopes for the adoption of a consistent solution between the market risk and the CVA risk frameworks.In the light of the above, ABI thinks that no modifications of IRC models should be requested in these RTS, before implementation of the FRTB. In that respect, the FRTB says that “PDs implied from market prices are not acceptable unless they are corrected to obtain an objective probability of default” (1).
ABI also notes that, in addition to theoretical reasoning, technical issues have to be taken into account. In fact, internal models do not always provide transition matrices and, above all, AIRB models usually do not cover the entire scope of the issuers of trading book instruments.
(1) Basel Committee on Banking Supervision, Minimum capital requirements for market risk, January 2016, page 61.
Q43: Do stakeholders agree with the exclusion of zero PDs for IRC?
ABI acknowledges that this measure is in line with the FRTB. Anyway, in principle, ABI does not agree with the imposition of a floor. A generic floor can introduce moral hazard in the trading strategy (i.e. same PD for different issuer may determine appetite for higher yield). As a matter of fact, several papers show valid methodologies suitable for a sound PD estimation even when no defaults are observed in the historical sample.Q44: Do stakeholders consider that losses due to default should be based on the market value or the instrument’s principal?
In ABI’s opinion they should be based on market value.Q45: Do Stakeholders have any additional comments or concerns regarding the requirements outlined in the IRC section?
ABI has no additional comments.Q46: Do Stakeholders have comments or concerns regarding the requirements outlined in the correlation trading section?
ABI has no observations with respect to the correlation trading section.Upload files
EBA RTS on IMA assessment_ABI PP_def.pdf
(203.58 KB)