Regarding the breakdown into MoC A, B and C, this process will be particularly burdensome for large banks given the high number of portfolios and rating models to be covered and consequently the number of calculations this will require for the parameters under different approaches (with conservatism and without it, removing a different level of conservatism under each scenario). Currently, the MOC contributions to the PD/LGD are not stored separately but integrated in the calibration targets. Collecting and storing these MoC contributions will therefore require significant IT adjustments, especially for large banks with a high number of rating models, requiring a full review of IT processes for doing so.
Furthermore, the EBA specifies the collection of MoC included in risk weights where fallback internal credit ratings are assigned. The EBA identifies that fall back internal credit ratings are assigned when information is outdated, missing or incomplete. Our understanding is that the segregation of components of risk weights can be difficult given the exponential relationship between internal credit ratings, PDs and the resultant risk weight. Collection of this information will therefore also involve changes to systems and/or processes. Consequently, we propose that the EBA considers both a delay in the submission of this data point and further clarity on the aggregation of this information.
Indeed, we recommend the implementation timing is reviewed holistically in light of the Progress Report on the IRB Roadmap published the 9 July 2019. The collection of information for the 2022 Benchmarking exercise requires the rating systems adjusted under the new IRB Program regulation (in particular, the MoC framework defined in the GLs on PD and LGD) to be implemented by 31 December 2021. However, according to paragraph 18 of the IRB Roadmap, the final deadline for making changes to the rating systems has been postponed until the end of 2021. Therefore, depending on institutions’ implementation plans, the 2022 benchmarking exercise could be biased as some institutions’ figures may be based on simulations instead of based on adjusted models. For instance, some regulators have prescribed timelines for review of models under the EBA’s IRB Repair programme that extend beyond 2022, while for other institutions they will submit changes for almost all models during the second half of 2021 and await an ECB decision. Therefore, it is very likely that by 31 December 2021 a significant number of model changes will still be awaiting regulatory approval and not yet implemented.
Given this situation, and to ensure the comparability and consistency of the figures collected from institutions, we consider it would be more appropriate to postpone the new collection of data to the 2023 benchmarking exercise. At a minimum if the EBA does decide to proceed on this, we suggest that in the next benchmarking exercise the EBA focus on collecting MoC data at aggregate level with an option for banks to provide MoC A, B and C. Further benchmarking can then review this collection.
Gathering information on both margin of conservatism and supervisory multipliers/floors should be done at the highest level of aggregation in the template as they are applied to the rating system model (for each IRB parameter PD/LGD). In particular, for the PD we suggest this should only be collected at the highest performing level of aggregation. We would also emphasise the challenge in disaggregating multipliers.
As regard the add-on imposed by supervisors due to deficiencies in the IRB model (column 690 of template 02.00 of Annex I to implementing regulation (EU 680/2014), this could be gathered qualitatively on the single IRB model affected.
As raised in question 1.1, we recommend the implementation timing is reviewed as per the Progress Report on the IRB Roadmap. Although any rating system which includes in its scope any corporate exposures other than financial institutions treated as corporates and large corporates will be subject to the deadline of the end of 2021 (application date January 2022, as per paragraph 18), as stated in paragraph 19 of this report, the deadline for the adjustment of LGD and CCF models considering the MoC framework defined in the GLs on PD and LGD has been postponed until the end of 2023 (application date January 2024, applying the same criteria as for HDP) where institutions have stand-alone rating systems for exposures to institutions, financial institutions treated as corporates or large corporates as defined under the final Basel III framework. Therefore the comparison of data collected from institutions until this date could be biased given that for some institutions the figures could be real but for other institutions the figures would be based on simulations.
Indeed, those institutions which have planned their LDP model developments to take account of the postponement until end of 2023, may not be able to report real or simulation data. In this respect, a bias would come about both during the voluntary and mandatory (from 31.12.2023) collection data period, due to the lack of available data for those institutions which postponed the model development, which could lead to an unlevel playing field (even if they are complying with the regulatory obligations/timeframe). Hence, we consider the postponement of the new collection of data for LDP to the 2025 Benchmarking exercise is needed to ensure a level playing field as well as the comparability and consistency of institutions’ data. If the EBA do decide to proceed with this data collection on a voluntary basis, then models approved under the previous regime should be out of scope and it should only benchmark models based on new regulation or guidance that will result in changes in the rating system to ensure comparability of models.
As commented above, the new requirement of data is burdensome for large banks given the high number of portfolios and rating models that will need to be covered, as it implies the calculation of parameters under different approaches (with conservatism and without it, removing a different level of conservatism under each scenario).
Secondly, depending on the implementation plan of the credit institutions, the 2022 Benchmarking exercise could be biased, as for some institutions the figures could be based on simulations instead of considering the actual figures based on adjusted models.
We note this will mainly affect non-retail portfolios. Materiality will be dependent on the exposure size, the granularity of the credit rating system and the fallbacks applied. The above will introduce inconsistency of application which defeats the benchmarking exercise. This is also linked to the segregation of MoC point above which requires IT updates.
Disentangling MoC in the RWA is difficult to do. If the EBA decides to include it in the benchmarking exercise, such data would only be possible to collect on a qualitative basis at this stage.
This process will be burdensome for banks, as it will require collecting information step by step originated from the final rating calculation and complementary analysis for the information collected via a survey (for example, general level of outstanding ratings and the time periods these ratings are already overdue, as it is mentioned in the consultation paper). In respect of deriving updated information from the final rating calculation it’s not possible to retrieve accurate data step by step (for example conservatism related to missing data, or outdated information from a general process which includes other adjustments to risk parameters not linked to deficiencies where you cannot disentangle each component).
It is challenging to reconcile FINREP and COREP, banks will make best efforts to align
The addition of new sub classes and the introduction of categorisation of corporate exposures by size might require multiple changes to existing reporting systems which will require effort from the teams to ensure that information is collected and agreed appropriately. Specifically, annual turnover is typically stored in credit approval systems whilst RWA information is included within Financial and Regulatory Reporting systems within large banks. The addition of this data point together with changes to corporate exposures requires enhancements to IT systems. This enhancement when considered with regulatory changes within the next 24 months namely IRB Repair and the Implementation of Basel III will put additional strain on resources. For corporates that fall close to the 500m EUR threshold it is already very complex for banks to assess and then aggregate, let alone integrate this within financial reporting systems.
Large banks typically use actual LGD values for internal management reporting and statutory reporting under IFRS 9. Consequently, the calculation of hypothetical IFRS 9 LGD values is likely to be a one-off exercise for the purposes of submission of data to the EBA. This calculation will have to be performed outside IFRS 9 databases, for each low default portfolio and will require an exercise to identify the sub-set of senior unsecured instruments that have no negative pledge. The calculation will therefore be operationally intensive, with the information related to the “LGD IFRS9 unsecured 12m hypothetical” not necessarily easy to extract. In addition to these operational complexities, we believe that hypothetical IRB LGD values provide better risk insights for low default portfolios and propose that only hypothetical LGD values for LDPs for IRB are retained.
As highlighted in previous questions, it will be particularly burdensome for large banks given the high number of portfolios and rating models to be covered. The calculation process will be operationally complex and for some banks it will be an IT challenge to extract the data.
We do not believe this is necessary as we believe the weights will be constant per time horizon. As a priori, weights used in economic scenarios for Large Exposure models that some banks have in force are constant throughout the projection and therefore this is not a relevant issue.
Template C 120.01 requires firms to submit Implied Volatility for Vega. Many banks are computing sensitivity via a proportional shift of the volatility surface, therefore obtaining directly a sort of ‘vega-weighted’ (i.e. vega x vol) quantity.
The option to enter the “vega-weighted” quantity directly should be retained; use of ISDA SA-CRIF format might resolve any potential inconsistencies in banks submission.
We confirm the EBA will be granted a free license by ISDA to use the ISDA SA-CRIF standards for the purpose of collecting risk data from supervised firms.
Relying on the existing ISDA SA-CRIF standards will have the benefit of easing the burden on supervised firms since most IMA supervised firms currently use the ISDA SA-CRIF standards for ASA and hence will not need to develop any new risk data reporting format.
The Industry would like to inquire about the planned availability of final ITS for the envisaged 2021/2022 EBA benchmark exercise changes also in relation to report 106.01, should it end up becoming a mandatory delivery in early October 2021.
It is not the content of the report, rather the delivery infrastructure at the heart of this inquiry.
In order to specify the newly added reporting templates (106.01, 120) for statutory reporting, the final ITS will be required. Considering that this year the implementation of new benchmark reports will be competing for resources with several other statutory reporting changes going live and a stress testing exercise the Industry would like to have a better idea by when is the final ITS expected to be delivered to Competent Authorities (‘CA’)?
Given for the initial market valuation the amount of data points collected per portfolio increases from one value per portfolio to potentially dozens of values per portfolio the proposed time span between the booking data and the IMV remittance data of one week seems challenging. This period will also overlap with the FRTB-SA reporting go-live which will add further strain on the firm’s resources.
The Industry proposes to extend this period to at least three to four weeks.
The industry would like to raise awareness on implications regarding risk sensitivities and capital calculations in a currency other than the firm’s reporting currency. Under FRTB context firms have the optionality to either calculate capital on reporting currency or on a base currency and then translate it to the reporting currency. However, both approaches are associated with infrastructure build and computational cost as new sensitivities are required either for FX delta or for FX curvature. Once a firm chooses a preferred approach and currency as part of their implementation then is not easy to switch. EBA benchmarking requires banks to be able to calculate capital not only on reporting currency but on the portfolio’s currency. That comes along with an associated cost to build the infrastructure and/or produce the new sensitives that they would not have to do otherwise for firms capital calculation.
We need to highlight that the calculation of FX delta and FX curvature is not trivial and requires more guidance from the regulators.
• Firms should calculate the ASA amounts using existing systems and translate at spot as required for reporting benchmarking results
• As firms will be reporting detailed information at the risk class / risk factor / currency level (in the SA-CRIF) EBA will be able determine the intrinsic element of FX risk
In addition, the Industry would like to propose to change the term ‘base currency’ used in the instruments and portfolios documentation with e.g. ‘instrument currency’ as the term ‘base currency’ is used in the FRTB-SA and therefor might create a confusion when reporting FRTB-SA risk measures.
Instrument #4 – Peugeot futures
The merger of Peugeot and Fiat under Stellantis was completed in January 2021. Whilst the Bloomberg Ticker (UG FP) remains valid for the Peugeot SA underlying the traded future is now Stellantis with a Euronext code UG6.
For all futures the exchange code of the future could be quoted rather than the Bloomberg ticker of the underlying.
Instruments #9 - #16
The descriptions for the equity options include the comment (1 contract = 100 shares) implying these are exchange traded options. As for futures, the intended exchange should be specified together with the exchange code of the option.
e.g. Bayer single stock future on Eurex (Instrument #3) has exchange code BAYG and the Bayer equity option (instrument #9) has Eurex exchange code BAY
Auto-callable Equity product
The autocall level had changed for the 2021 exercise and the same definition is kept in for the 2022 exercise:
Autocall level (‘Initial value’): End of day Booking date - 1 year
This used to be (pre-2021 IMA benchmarking exercise):
Autocall level (‘Initial value’): End of day Booking date + 1 month
The ‘End of day Booking date – 1 year’ is not a standard market practice.
Revert to 2020 definition:
Autocall level (‘Initial value’): End of day Booking date + 1 month
Instruments #24 and #30
Both instruments (one long, one short) are defined with same ISIN and same notional. These instruments are part of Portfolio #18. Is that intended?
5-year Mark to Market (MtM) Cross Currency EUR/USD SWAP
It can be operationally problematic to exclude the cash balance from the benchmarking results.
The instructions should be changed to include the cash balance while maintaining the notional exchange at inception and termination. Note, this eliminates most of the FX risk in the instrument which reflects the way cross currency swaps are typically managed.
If the EBA intention is to retain the relatively large FX risk then the Industry requests the EBA specify an instrument that does not require subsequent manipulation to include/exclude the cash balance, for example specifying notional exchange at maturity only
Short position in spread hedged Super Senior tranche of iTraxx Europe index
The iTraxx Europe standard tranches are 0-3, 3-6, 6-9, 9-12, 12-22, 22-100. The attachment point of 25% creates a bespoke tranche. Is that intended? Also, instrument #80 is in fact two instruments, the tranche and the CDS index hedge; it may be clearer to describe these two instruments separately, particularly to specify whether the institution is long or short protection in the tranche.
The Industry welcomes the portfolio structure simplification resulting from amending the instrument definitions to include the quantity and notes than in section 3. Individual Portfolios there is therefore no need to mention “1 instrument” each time.
The answer to this question is likely to be very firm specific as it depends on the respective business model.
The Industry recommends that the EBA have those discussions bilaterally with individual firms.
As noted earlier, if there is a consideration for a broader ASA benchmarking exercise then it is worth ensuring that the set of portfolios as a whole covers at least all SA risk classes and respective risk positions. However, given the internal nature of IMA this may be considered less relevant for an IMA benchmarking exercise.
ISO 20022 has been expanded to cover OTC derivatives and ESMA is indeed requiring it use for MIFID trade reporting. However, ISO 20022 does not provide the level of detail that FpML provides and many trades would not be able to be represented in ISO 20022. Firms are familiar with ISO 20022 messaging in the payments space (because of Swift) but in the derivatives spaces it has been limited and only to send specific info externally as required by ESMA.
EBA should move towards using FpML for instrument representation
Instrument specifications in a markup language such as FpML are useful to align on booking details of trades. However it needs to be stressed that these are currently used in a pilot mode and firms may not be able to consume them systematically. Hence such specification can enhance but should not replace a careful instrument definition.
EBA should continue with the Annex 5 instrument description supplemented with FpML.
An option might be to publish term sheets of trade FpML on EBA’s web page at the beginning of the exercise and run an informal public Q&A process on these portfolios during the exercise. This may prove more efficient than formal requests for clarification and formal requests to amend bookings for certain trades as part of the IMV validation process.
The industry would find it helpful for such values to be added into the instrument specification (i.e. moving away from defining instruments being “at the money on the booking date” and rather give an explicit strike ) and this would remove some remaining ambiguity on instrument definitions. However this might lead to a situation where instruments are far away from the money on the booking date due to the significant time span between the publication of the ITS and the booking date.
In order to avoid a situation where instruments are far away from the money on the booking date the Industry proposes that the EBA publish a supplement to the ITS just prior to the booking date. The supplement, which could be in the form of an update to the FpML described in the response to Q.17 above, would set the instrument parameters subject to risk factor levels (e.g. strike prices, fixed rates, etc.) to give the required moneyness as at a more recent date, say 1st September 2021.
Equity, FX and Commodity prices are publicly available on the appropriate exchanges. Fixed rates for swaps should be taken from the relevant benchmark.