Response to consultation on draft RTS on the materiality of extensions and changes to the use of FRTB IMA and changes to the subset of MRF
Q1. What are your views on requiring additional information for specific extensions and changes?
The industry believes that the suggested qualitative criteria for categorizing model extensions and changes as material for the below cases are unnecessarily restrictive and make the IMA overall less attractive due to the rigidity of the draft proposals:
a) changes in the organization or structure of trading desks, and
b) ‘reverse extension’ when an institution aims to apply the Standardized Approach (i.e. a simpler but more punitive method)
Furthermore, the suggested criteria may also result in incorrectly classifying a normal organizational change as material for the purposes of the IMA.
We believe that the RTS should explicitly mention that any change in the structure of the desks that are the product of:
- splitting an existing IMA desk into two or more IMA desks should be placed under the “non-material extension and changes requiring additional information” category. This is on the basis that there is no material change on the overall products and pricing models of these desks post re-structure.
- merging two IMA or more desks into one new IMA desk should be placed under the “non-material extension and changes requiring additional information” category. This is on the basis that there is no material change in the overall products and pricing models of the new desk post merging.
Separately, for an institution that maintains an overall IMA own funds requirement (OFR) over total market risk OFR ratio that is greater than 10% as per Article 325 (3), any desk move from IMA to SA should be categorized under the “non-material extension and changes requiring additional information” category. This is to ensure that there is no additional regulatory burden stemming from a move to the simpler and more conservative SA, which acts as a fallback to the IMA in any case.
For any other types of changes related to desk structures (i.e. move of an SA desk to IMA), the quantitative conditions should first be assessed to determine whether the change is categorized in the material or non-material bucket. In case the quantitative assessment results in categorizing the change in the non-material bucket, then institutions should provide the required additional information of the change but still be able to proceed with an ex-ante notification to the competent authorities.
Q2. Do you agree with the proposed quantitative conditions for determining the materiality of extensions and changes?
We believe that where a risk measure does not represent more than 5% of the sum of all IMA risk measures at day 1 of the assessment (e.g. DRC), institutions should completely exclude this metric from the 15-day testing. This will avoid excessive operational burden on institutions.
The draft RTS mandates that quantitative assessments are made on the IMA perimeter of desks that meet the back-testing requirements and the profit and loss attribution test. Indeed, article 2 paragraph 4 in the RTS stipulates that the assessment shall be made on the “portfolio of all positions assigned to the trading desks for which the institution has been granted permission as referred to in Article 325az(2)”. Consequently, any changes to the models that affect primarily desks currently ineligible for IMA will have by construction a low quantitative impact. Some changes may focus on remediating model deficiencies observed on one or several desks that fail the eligibility tests. We support the EBA proposal: model change impact assessments should be performed on the current actual IMA perimeter, since this is a true reflection of the own funds requirements impact. For a desk currently capitalised in SA, a model change may have an impact only if in the future the desk is included within the IMA perimeter. In this case, the impact would be driven by the difference of OFR when in IMA versus SA.
In some instances, when a change that targets model deficiencies of a desk failing the IMA eligibility tests, the bank may provide as part of their model change submission package, an impact assessment on some risk monitoring metrics covering desks within the IMA intended perimeter (i.e. the perimeter for which the bank has received approval from its Competent Authority for the use of internal models conditional to meeting the back-testing and P&L attribution test requirements) but currently capitalised in the SA (because either back testing or P&L attribution test has failed). For instance, a bank monitoring the internal model firm-wide VaR (i.e. the VaR of all desks intended to be capitalised under IMA) may provide, as part of the documentation package of RTS article 9, the model change firm-wide VaR impact.
Q3. What are your views on the selected threshold levels?
We believe the suggested thresholds for both material model changes or extensions and for changes of the subset of modellable risk factors require further review. We recommend two options to amend the thresholds and materiality assessment.
Option A
As an alternative we recommend aligning the upper and lower thresholds of the material model changes and extensions presented in this RTS, to the maximum of the two i.e. +/- 15% for the IMA OFR and +/- 20% change in at least one of Expected Shortfall (ES), Stressed Expected Shortfall (SES) and Default Risk Charge (DRC). Similar thresholds should also apply when assessing changes of the subset of modellable risk factors. Specifically, our recommendation is to align the upper and lower thresholds of the material changes and extensions in the subset of modellable risk factors to +/- 15% for the IMA OFR and +/- 20% change for ES.
<Detailed on the proposed thresholds can be found in the attachment >
Option B
We recommend a two-step approach that links the materiality thresholds with the impact of the model change or extension to the total OFR (i.e. IMA + SA OFR) of the institution. By assessing the impact of a change or extension to the total OFR (not just the IMA OFR), we believe it will provide a better view to the competent authorities of the true magnitude of the change or extension has on an institution’s total OFR.
Specifically, our recommendation is that an additional pre-condition needs to be met for a model change or extension to be considered material. In the first instance, before proceeding with an assessment of the impact on the IMA OFR and individual metrics (i.e. ES, SES, DRC). The suggested precondition is that only changes or extensions with an impact of greater than 5% in absolute terms, on the Total OFR (i.e. IMA + SA OFR) should be considered material. If the pre-condition is not met the change will be categorized as non-material. If the precondition is met the institution will proceed in the next step of assessing the impact on the IMA OFR and individual metrics based on the thresholds described in this RTS. Similar logic should also apply when assessing whether changes of the subset of modellable risk factors are considered material or not.
Q4. Do you agree with the proposal to introduce a condition for assessing the relevance of the risk measures to be checked for the materiality assessment of model extensions and changes?
Yes, but we suggest that if on day one a risk measure does not represent more than 5% of the sum of all IMA risk measures it should be considered irrelevant straight away.
Q5. Do you agree with the proposal to always assess the quantitative conditions on a 15 consecutive business days period?
Assessing the impact for 15 consecutive days is very burdensome in contrast with the Basel 2.5 framework. The FRTB IMA framework is more complicated and computationally intensive by requiring more runs and analysis (i.e. the need to perform three main runs for ES (RS (Scaled), FC (Full Set), RC (Reduced Set)) plus one run for Non-Modellable Risk Factors (NMRF) and one run for IMA DRC). In addition, based on the outcomes that we see in the current parallel runs for IMA, the impact for both IMCC and NMRF remain reasonably within the proposed thresholds of this RTS and therefore the requirement to perform testing for 15 days seems excessive.
Further, we want to understand how a 15-day assessment will be performed if it covers a quarter end when the IMA scope changes. In extreme cases, post quarter end, if all affected desks fail the IMA eligibility tests and are capitalised in SA, the impact will be zero (see also comment to Q2).
We note that in the current framework, the assessment period is shorter. The 15-day assessment period could be stopped whenever the materiality threshold has been exceeded. Although in practice we rarely see such a situation, this does however show how the regulation attempts to reduce the operational burden. We would like this RTS to help limit the operational burden to the maximum possible way.
We therefore suggest that the assessment period should be reduced to our recommended 5-day period, which should give Competent Authorities sufficient comfort into the materiality assessment.
Q6. Do you agree with the proposal to require non-material extensions and changes to be notified four weeks before their planned implementation?
We note that the existing RTS requires two weeks notification before the planned change which we believe is sufficient and does not need to change. In addition, under the Basel 2.5 framework the ex-ante notification period did not apply to all changes that exceeded the day-1 1% threshold (for instances that do not meet the qualitative criteria).
Under the FRTB framework when the day-1 assessment of any risk number exceeds the 1% threshold, a 15-day quantitative assessment period is mandated and alongside the extension or change will be considered of some significance no matter the outcome of the 15-day materiality assessment, and hence will have to be notified 4 weeks ahead of implementation. In total, the full process will take at least 7 weeks to complete.
This very lengthy process will jeopardise remediating observed model weaknesses in time before the following end of quarter. Since remediation cannot be implemented speedily, it may lead in some instances for a desk to fail the next quarters eligibility test. This would have severe consequences in a reduction of the IMA scope, for potentially a long period, since when a desk is in the red zone, a desk must return to the green zone to allow capitalization in IMA which is something that may prove challenging.
We suggest that the ex-ante notification period should be shortened to the maximum extent possible and in any case be no longer than the current 2-week period.
Q7. Do you agree with the proposed conditions for determining the materiality of changes to the institution's choice of the subset of the modellable risk factors?
Satisfying all three conditions imposes additional burden in an already complex framework. We suggest to either keep the Partial Expected Shortfall (PES) ratio or the conditions on the impact of ES and IMA OFR.
Q8. Which one of the options do you think is more appropriate for a quantitative condition based on the ratio PEStRC/PEStFC (option 1 or option 2)? Please provide the rationale for the chosen option.
We prefer Option 1 as it is simpler and easier. Option 2 penalizes institutions that already have a high PESRC / PESFC (e.g. a ratio close to 100%).
Q9. What are your views on the proposed clarification to the content of the reports of the institutions' independent review or validation?
We agree with the suggested RTS.
Q10. Do you agree with the proposed qualitative criteria to identify material extensions and changes, as set out in Part I of the Annex?
No, as mentioned in the summary and on Question 1, changes related to the structure of the desks or changes related to the business strategy of a desk should be placed under the category “Non-material extensions and changes which require additional information”.