Response to consultation on Regulatory Technical Standards on prudent valuation
If not, please describe the challenges you face with regard to a monthly calculation, and the monthly reporting of fair values and AVAs (e.g. with the COREP templates). Please make clear if those challenges arise in general or with regard to specific positions (e.g. type of instruments), whether they arise for positions assigned to the trading or non-trading book, and whether they arise for positions treated under the simplified or core approach. Please describe any simplifications and/or assumptions you would have to apply to determine fair values and AVAs on a monthly basis.
Calculating AVAs with a monthly frequency can be operationally challenging. The monthly calculation and reporting require significant system development (such as valuation process for banking book and automation of Variance-Ratio Test), prudent data contracting (such as market volume data currently provided on a less frequent basis for illiquid position for the calculation of CP AVA) as well as coordination of COREP submission process.The COREP reporting cannot be produced on a monthly basis and is not meant to.
While on trading books, fair value is calculated and reported monthly, with a limited setup of controls and certifications, thus, subject to a relevant industrialisation, a monthly measure of AVA should be theoretically possible (if banks can maintain this limited setup of controls and certifications). However, PVA core approach processes require the collection of a vast amount of risk and market data, numerous calibrations, while involve subsequent reviews and challenges as part of a robust governance, need proper commentaries and finally result in multiple reporting and escalation. The overall PVA process is time-consuming and therefore the cost/benefit of moving to a monthly calculation is not obvious.
Furthermore, AVAs are mostly calculated from valuation exposures and confidence intervals charges. On trading book positions the valuation exposures are calculated every day, yet the confidence interval charges cannot be easily updated on a monthly basis as this would require a full automation of market data gathering. The cost and complexity of such a project would be detrimental to the activity, with very little added value from supervisory or internal perspective. Indeed, the volatility of AVA over time remains reasonable except for specific cases (such as the Covid outbreak) but this can be tackled separately in exceptional circumstances without such strong and complex requirement. In addition, this cannot be done on the banking book positions where valuation for accounting and prudential requirements is only required on a quarterly basis. The specificity of the banking book, and in particular its own dynamic nature, is to be respected in the RTS as mentioned in the CRR under Article 105.1[1].
More generally due diligences related to instruments at fair value cannot be the same for banking book and trading book positions, due to the nature of the instruments, and the process implemented to monitor these two different books.
Besides, management awareness to valuation risk has been already achieved through various accounting fair value adjustments that exhibit similar dynamics to AVAs, such as XVAs and bid offer FVAs; various markets volatility indicators are also indicative of potential price dispersion and can provide real time information.
However, in case this requirement is maintained, the main criteria triggering a monthly calculation and/or COREP reporting should be more explicit, and the text should clarify whether shock calibration and VRT tests shall be also updated monthly. In addition, it would be also worth clarifying whether the monthly calculation implies to provide the reporting templates (C32.1-2-3-4) at the same frequency.
Finally, Level 1 text requires that reporting requirement shall be proportionate to the objective targeted by the requirement (see CRD 104.2).
Question 2. Do you have any comments on the amendments to Article 3 in general, and specifically with regard to the threshold of ten contributors set out in paragraph 2, point (d)? If you consider a different threshold should be applied, please describe how to set it, and provide a rationale and evidence supporting your proposal.
To begin with, Article 3.1 provides a list of data sources and requires that, “Institutions shall use at least the same range of market data as a basis for calculating the AVAs, as they use in the IPV process, provided that the market data meet the requirements set out in this Article”. Taking into account the recency requirement for AVA data per Article 3a, the “same range” requirement in Article 3.1 could be interpreted as when AVA is based on a 30-day history, IPV shall do the same in order to align the data range with AVA. This contradicts the intention of IPV, i.e. to use the best quality data source rather than a range with old data. To avoid such confusion, we kindly request the EBA to consider rewording the Article 3.1 into the following: “institutions shall use at least the same source of market data as a basis for calculating the AVAs, as they use in the IPV process.” Regarding the use of consensus, banks agree that they should be robustly back tested like any other pricing source. This should be part of a systematic pricing source accuracy, reliability, tradability and eventually observability assessment. The RTS should not be source specific as reliability assessment also applies to other sources, but it should rather be comparable analysis, evaluated pricing or broker quotes.
However, the classification of historical data that date back more than 1 month, as expert judgement will effectively increase the portion of AVA classified as expert judgement despite the existence of reliable data. As a consequence, banks are being forced to favour non-binding consensus rather than actual traded prices. Therefore, we propose to allow the use of transaction pricing data for up to 3 months in “Data Range” approaches with necessary adjustments if market conditions at the reporting date are materially different from those 3 months earlier.
The threshold of 10 contributors in consensus data seems excessive and is not justified. Moreover, if the number of participants in a consensus service was moving from 10 on one given reporting date to 9 on the next given reporting date, would the industry need to change the categorisation from range-based to expert based?
Therefore, it creates a wrong incentive for consensus providers to look for less active market participants to increase the number of contributors. The 10 contributors threshold would be also anti-competitive as it would favour established consensus services at the expense of emerging services: new entrants would be discouraged and deterred. Conversely, it might create a wrong incentive for consensus services to aggressively look for a 10th contributor which may just deteriorate the quality of the service. Technically, there are curve or surface services where some risk points receive more than 10 contributions, some other less than 10 points, which would make the implementation very complex. We observe that the 5 main dealers often account for 50% of market volumes /market shares [ESMA report on EU Derivatives Market 2023], which should be representative enough to calibrate an appropriate range of uncertainty. One may wonder if the split range based/expert judgment eventually just duplicate the fair value hierarchy assigned to valuation inputs.
As a general concern, we consider that excluding available data from the scope of eligible information to support range-based approaches is by definition a wrong approach as a large scope of potentially eligible information is the best way to support over the long term a reliable use of market information. The implementation, the weight given to the different information, and the prudence level in the utilization of data should then be applied when the quality of information (including if relevant the number of contributors) is reduced.
However, if the EBA would like to keep this threshold, it should be estimated consensus service by consensus service depending on the product complexity, the existence of the “reducing risk market” and the parameter’s intrinsic variance (e.g. if the whole market has the same view a limited number of active market participants may be sufficient). Besides, more insight on what is considered a suitable consensus service data would be welcome.
In addition, regarding Article 3.2.(d), “consensus service data where the number of contributors is greater than or equal to 10 and the institution has performed a valuation back testing”, we kindly request the EBA to consider rewording it into “consensus service data where […] and a valuation back testing is performed” as the text should not be restricted to in-house backtesting.
Finally, it is noteworthy the fact that removing the indicative quotes from Article 3 paragraph 2 from the current framework may result in AVAs calculations that are based almost solely on the expert-based approach.
Question 3. Do you have any comments with regard to the requirements proposed in Article 3a? If you consider that some of those requirements should be adjusted, please describe how you would revise them in order to meet the policy objectives that the proposed amendments try to achieve, and provide the rationale supporting your proposal.
The one-month recency criterion is not manageable for banking book positions, they should therefore be excluded from this requirement. It is right that the valuation must be based on recent data reflecting the state of the market as of the reporting date. However,limiting historical data to one month is arbitrary and does not account for the difference in liquidity, potentially excluding good data for less liquid instruments. When measuring confidence interval for uncertainty, an historical approach is often necessary to build a relevant distribution, provided that the distribution obtained is a fair representation of the uncertainty as of the reporting data. Judgement in the choice of suitable data is required. Indeed, the length period should result from an historical analysis of the risk factor behaviour and the statistical consistency of the calibrated shocks. Exception may be permitted in order to eliminate the volatility of AVAs over time: for instance, dispersion observed on the most recent day of the year might not be representative (e.g. liquidity tightening at the end of year or quarter, market movement occurring on one single day) and might require some adjustment or the use of alternative set of data. Firms should have robust documentation as to why they depart from the “most recent data” guidance. Efforts to obtain reliable market data sources amenable to range base analysis should remain proportionate and not generate undue costs.
In addition, the proposed measure is likely to trigger a wide reclassification of risk factors from the range-approach scope to the expert-based approach, together with large solvency impacts. Moreover, this would be associated with a high degree of uncertainty as to the classification, from one month to another, and a large volatility of solvency measures.
Finally, we kindly request the EBA to consider rewording the current provision into the following one: “demonstrate that there are sufficient and reliable market data sources to calculate AVAs under range-based approach”, instead of “demonstrate that (a) there are no sufficient and reliable market data sources to calculate AVAs under range-based approaches”, because proving negative or non-existent data is logically impossible.
Question 4. Do you agree with the proposed amendment to capture valuation risks stemming from fair-valued back-to-back derivative transactions and SFTs? Do you agree that this would restore alignment with the treatment under the core approach? If not, please describe how you would suggest to revise the amendment providing any rationale and supporting evidence.
No reply.
Question 5. Do you agree with the proposed amendments to the calibration of the fall-back approach? If you consider that a different range of percentages should be considered, or that the AVAs under the fall-back approach should be calculated in a different manner, please suggest a range or a methodology, as applicable, and provide a rationale and evidence supporting your proposal.
The use of a percentage of notional for derivatives, especially the most complex ones that might be subject to the fall-back approach is not always a relevant measure of valuation uncertainty. Nothing can replace a sensitivity analysis with respect to the relevant risk factors. Some of these risk factors have boundary values (positive volatilities, correlations) that the notional based measure might not respect.
For non-derivatives, using a fixed percentage of Market To Market (MTM) would ignore nuances across different markets, instruments, and counterparties. This would not level the playing field, as x% could be aggressive for one institution, whereas conservative for another, depending on the composition of their portfolio.
In addition, a fundamental issue with an instrument-based approach, as opposed to a net risk approach, is that once a certain number of instruments are allocated to the fall-back approach, they are in theory no longer included in the core approach, so the vanilla risk exposure in the core approach becomes totally unbalanced as the fall-back deals are missing; the resulting Close-Out-Cost or Market Price Uncertainty becomes meaningless. Alternatively, allocating whole portfolios to the fall-back approach will inflate the total notional artificially. The result is that any instrument-based measure is bound to fail its objective. Imposing fall-back to exposure which is continuously risk managed in the trading book and for which models are approved in the context of the market risk framework would be inappropriate. Sharply disconnecting, as proposed, the market risk capital computation from the prudent valuation one, has no financial meaning.
For the industry, the point regarding the balance impact of removing a hedged structured product from AVA calculation while keeping its hedge is key. Therefore, the proposed fall-back approach is not applicable for derivatives without considering this issue.
Furthermore, banks would like clarifications on how to handle interaction between UCS/IFC and General Scope MPU/COC/MR. For instance, an issue in UCS credit LGD or credit spread related to a counterparty will impact all positions traded with this counterparty what is expected by the revised framework. Besides, they would like clarifications on whether for non-derivative cases “fair value” should include any FVA and IPV components as described in the example page 12.
The multiple and vague conditions forecasted in the EBA proposal would likely lead to large scope of activities in the fall-back approaches, which would have material impacts on solvency measures, due to mechanisms mentioned in the previous paragraph. They are also vague so to be subjectively interpretated, leading to distortion between banks. As a general rule, a fallback approach should only apply in rare circumstances when no clear methodology can be applied.
In principle, the expansion of the scope and calibration of the charge result in various issues. For instance, the notional-related charges for derivatives are not reflective of the actual economic risk. Moreover, the framework lacks the concept of significance for risks that could not be independently verified/price-tested and thus removes bank’s ability to assess and capitalise the economic risk.
In addition, the range of charges (1-15%) in Article 7(4) is not consistent with the FRTB residual risk add-on (“RRAO”) [0.1%-1%] for derivatives.
Hence, we would suggest consideration be given to introducing significant considerations. If the risk not independently verified is immaterial, then there should be no fall-back requirement. If it is significant and a bank can demonstrate that its framework is adequate to conservatively quantify capital charge, there should be no fall-back requirement. If the afore-mentioned does not hold, the fall-back requirement on Net Notional could be linked to RRAO charges of 0.1%-1%.
The only goal of such fall-back is to act as an incentive for banks to invest in relevant uncertain measure. The desired outcome is not to use any fall-back. A regulation that proposes an increase of use of fall-back as a good way to achieve level playing field would have totally dismissed its purpose.
Whilst providing an apparent standardization, the fall-back scope is also unnecessarily subjective as it refers to “the usage of subjective pricing source”. Besides, according to the current proposal, the existence of immaterial, untested parameters could also trigger the use of the fall-back approach, which would result in an excessive degree of prudence.
Furthermore, the paragraph 3 (b) needs to be clarified, in order to eliminate potential ambiguity that may impact smaller institutions. These institutions conduct daily valuation directly through an independent unit, distinct from business trading units, with independent pricing sources. Although there isn’t an explicit IPV adjustment, their valuation positions are effectively IPV adjusted. The current proposal might lead them to believe that the fall-back approach must be applied on all of their valuation positions.
Question 6. Do you have any comments in relation to the positions proposed to be subject to the fall-back approach? If you consider a different treatment should be applied to these positions, please describe how you would treat them in order to meet the intended policy objectives, and provide the rationale and any evidence supporting your proposal.
We believe that the RTS should not be instrument-specific nor prescriptive, but rather principle-based. Notably, market transparency and information in relation to specific instruments could evolve over time and might depend on firms access to information.
More specifically, no explicit consideration should be made for unlisted equity. Various valuation techniques (e.g. comparables, discounted cash flow) already exist that are described in IPEV Valuation Guidelines and practically implemented (see IMF Valuation of Unlisted Equity). These techniques should enable the derivation of appropriate uncertainty range.
We kindly ask the EBA to clarify the scope and definition of “unlisted equities”, because there can be many unique cases around “unlisted”, especially in the trading book; for instance, temporary suspension of equity stocks resulting in no recent price quotes but historical data is available, a recent re-listing on another exchange where price information will become available for the new shares. Can the EBA clarify that the fall-back approach shall apply to the banking book valuation positions consisting of unlisted equities?
Moreover, current RTS state that PVA should be aligned with IPV process. The proposed fall back would contravene to this principle for the unlisted equity where a valuation technic is already used.
In addition, the imposed fall-back approach on unlisted share is risk insensitive and would not take into account the nature of the investments (some areas of Private Equity are riskier than others), the investment policy of the manager, and the way valuation is monitored in the vehicle. Imposing a fixed percentage is then an incentive to invest in the riskiest investments which would attract the best return on equity (since equity required is equal regardless of the aspects mentioned above).
Illiquid positions requiring valuation inputs that cannot be easily calibrated to market information are those for which a valuation uncertainty measure is the most useful and prudent value regulation goals are to achieve this. By resorting to an arbitrary fall-back approach in the case where a real uncertainty exists and requires judgment as to its quantification will dismiss the goal of such a regulation. This would be like saying that all credit risk capital or insurance risk capital (which are both not market based) should be measured with a forfeit like the old Cooke ratio. On the contrary, we think that risk sensitivity should be preserved as much as possible. A risk sensitivity-based calculation with more conservative defaulting shocks and a potential remove of diversification factor should be a better alternative.
For issues such as unavailability or non-reliability of IPV sources for some of the valuation inputs, a materiality assessment is necessary before triggering the fall-back approach since in this case an expert-based approach on this non-material valuation input will very likely be more realistic than a fall-back approach.
Any proposed fall-back should be envisioned only as last resort when the firm is not in a position to gauge an appropriate level of certainty, using any sensible inference or extrapolation technique. Regulators may alternatively use the fall-back approach as a challenger model.
Taking into consideration that it would further increase the uneven level playing field with other major jurisdictions, the proposal would be a total showstopper. Finally, this will contribute to the removal from the landscape of some dedicated areas of the banking industry the European regulated institution. This will also concern the private equity which will increase in the financing of the green economy and the offer to clients of complex products while some important players being in the EU. This uneven level playing field will be also emphasized between European players being subject to prudent valuation on the one hand and other players on the other hand.
Regarding Article 7.3 (b), we kindly ask the EBA to clarify this sentence and ensure that different types of IPV are allowed (e.g. embedded IPV, stand-alone/ add-on IPV, expert-based IPV), as well as IPV adjustments of 0, by rewording this article into the following: “Valuation positions for which institutions are not able to estimate and record an adjustment for any types of independent price verification in their accounting, due to the fact that independent pricing sources are not available or pricing sources are more subjective.”
Question 7. Are the requirements included in Article 8 clear? If you consider them to be not clear or to be particularly challenging to meet in specific circumstances, please describe the issue you encounter and how you would address it in order to meet the intended policy objectives, and provide the rationale and any evidence supporting your proposal.
A major existing shortcoming of the core approach is the partial double counting of risks with respect to Day One Profit Deferrals under IFRS. Indeed, article 8 still does not address the issue of offset between Day One Profit Deferral (DOPD) and AVA. We understand that the offset prohibition stated by Q&A 2019_4458 still stands but this is not justified given that DOPD and AVA can account for the same uncertainty due to unobservable valuation inputs.
For this reason, we urge the EBA to rectify this issue by allowing a partial offset subject to the following safeguards:
- Banks need to map the deferred Day One Profit reserves directly to the corresponding AVA for the respective valuation uncertainty drivers ;
- The amount that can be offset is generally capped at the AVA.
For this purpose, the words “Fair value” should be replaced by “accounting value” in lines 2 and 5 of paragraph 3, to avoid a double counting in the way the CET1 is computed. You will find in Appendix a proposed methodology by European banks to offset accounting adjustment from deferred Day One Profit to AVA.
Besides, the end of paragraph 3 is meant to impose a floor cap on the use of accounting adjustments that can be offset with AVA so that the same adjustments is not offset several times. This requirement is expressed in a complex and unclear way. Indeed, banks would need more precision and ideally an example or to be explicitly written in the Annex.
Section 8 also states that “In addition, considering that accounting fair value adjustments are often determined at portfolio level, while AVAs are determined at the level of a valuation position, it should be ensured that institutions do not use one and the same accounting fair value adjustment to reduce more than one individual AVA. Therefore, this Regulation should specify a cap for the amount of accounting fair value adjustments that institutions can actually use to reduce AVAs” and is equally confusing.
Likewise, in paragraph 6 banks are wondering whether this includes model pricing used for accounting valuation. Clarification would be welcome.
Paragraph 7 does not have its place in this article. It is a justification requirement, specific to methodologies detailed in article 9 and 10 (sensitivity-based adjustments), that should better be stated in these articles themselves where the methodology is described.
In addition, this exercise is burdensome and the additional charge is disproportionate compared to the materiality of this additional requirement. The transversal ECB on-site missions issued recommendation to justify the utilization first order sensitivities and this exercise has allowed justifying the utilization of first order sensitivities.
In particular, regarding Article 8.5 “For the determination of the AVAs, institutions shall use the same pricing models that they use to determine the changes in the value of positions reported in their financial statements. If sensitivities are used in the determination of AVAs, they shall be based on those models”, we would like to point out the cases where end-of-day valuation is based on third-party pricing model whereas sensitivities that are used for market risk and AVA calculations are produced by internal models. For example, end-of-day valuation for the multi-callable bonds is based on Bloomberg prices, while credit risk sensitivities are not provided by the same vendor but generated from validated internal models. Therefore, we kindly suggest considering the rewording of the last sentence into the following: “If sensitivities are used in the determination of AVAs, they shall be based on those models, or on validated sensitivity models for end-of-day risk calculations.”
Moreover, the requirement of Article 8.7 of inclusion of convexity and cross-order effects is challenging and requires significant effort of implementation and introduces a significant layer of computational burden with limited expected effect from a mostly plain vanilla portfolio. Gamma/x-gamma risks require large moves to become material, which is more relevant for VaR / stressed VaR calculations rather than valuation uncertainty. Note that even in a full revaluation AVA approach not all cross-order effects are covered, not forgetting that a full revaluation solution might potentially be inconsistent with the Variance Ratio Test in case of dimensionality reduction, which somehow assumes sensitivities are used. Next to that, not all residual risks are supposed to be closed out with MPU/COC, given the requirements of Future Administrative Costs AVA, referring to Article 15.1 “institutions may consider the AVA for future administrative costs for a valuation exposure to be zero, where all of the following conditions are met: (a) the individual market price uncertainty, close-out costs, and concentrated positions AVAs implying fully exiting the exposure”. When FAC AVA is calculated for the total exposure, it already addresses any residual risks missed out from MPU/COC. Given no diversification benefit is allowed for FAC AVA, FAC AVA shall not overlap with MPU/COC. All in all, Therefore, this requirement should be dropped.
The requirement to calibrate End of Day models quarterly means an unnecessary operational burden, as institutions already use these models to demonstrate marking to market (Fair Value). We propose to drop this requirement.
Question 8. Do you have any comments with regard to the amendments to Article 9, 10 and 11? If you do not agree with the amendments, please describe how you would adjust or design the requirements to meet the policy objectives that the amendments try to achieve. When giving your answer, please provide the rationale and relevant evidence supporting your proposal.
The modifications on Articles 9 and 10 are much more complex to follow, inevitably leading to differences in interpretation among banks. From a presentational point of view, we consider that adding additional requirements with regards to the alpha factor applicability as part of the Annex is not appropriate and any such significant requirements should be part of the articles themselves.
The articles allow calculation of MPU/COC with available and reliable data sources based on a reduced set of parameters. However, the article should also clarify what constitutes “an unreduced set of parameters”. We therefore request EBA to add a clarification specifying that End-of-Day (hereinafter:EoD) risk representation buckets where risk sensitivities are generated by EoD valuation models can be considered as the unreduced set of parameters.
The changes with regard to the Variance Ratio Test disincentivise hedging because they penalise hedging strategies not matching exactly End-Of-Day risk buckets. This increases dispersion of capital across banks, subject to each bank's risk bucket definition and choice of hedges. Moreover, it is inconsistent with other capital requirement measures (e.g. FRTB NMRF), where the prescribed bucket reduction does not penalise aggregation. Therefore, we suggest removing opposition between VRT and the alpha factor usage requirement and instead introducing targeted restrictions on the VRT application:
- Requiring liquidity of selected buckets and replicating instruments to be supported by market data.
- Introducing prudency when buckets with sufficient traded activity/observability and buckets without sufficient trading activity/observability are netted together, by assigning the maximum shift between the two buckets.
The proposal to make Fair Value reserves subject to supervisors horizontal reviews expand prudential supervisory discretion to accounting Fair Value although proper accounting practice is a responsibility of the bank and its auditors enforced by ESMA. Banks are penalised if they do not attribute their Fair Value based on the ECB’s view. Reducing alpha for the entire AVA category to zero seems too punitive, given the fact that the problem above will most likely affect a small part of the portfolio. Moreover, the framework/principles required to ensure no adjustment to alpha factor are unclear and, thus, banks will have to assume the worst case in pricing, increasing competitiveness gap versus banks from third countries. We propose to drop this requirement or at least consider scaling the alpha reduction according to some estimate of the extent of the actual problem.
Banks are forced to book all IPV variances to avoid punitive AVA outcome, resulting in:
- Removing the concept of risk appetite for valuation uncertainty
- Removing the 1st Line / 2nd Line segregation of duties for Fair Value Recognition in P&L of adjustments linked to subjective or uncertain pricing sources, potentially reducing the quality of reported earnings,
In order to avoid these unintended consequences, we suggest :
- maintaining the 0.5 aggregation factor ;
- capitalising for Group level net aggressive unadjusted IPV variance for reliable independent pricing sources ; and
- identifying via a systematic assessment the unreliable subset where the asymmetric approach is to be followed for subjective independent pricing sources.
Article 9 required firms to “calculate individual market price uncertainty AVAs on valuation exposures related to each valuation input that the institutions used in the relevant valuation model to determine that exit price. Each of the valuation inputs shall be treated separately”.
Particular attention should be paid in order to avoid any double counting as the calibration errors observed on one single input may capture other input uncertainties as well.
These articles combined with the Annex make the use of the reduction of the number of parameters of a valuation input through the variance ratio test (VRT) conditional to the loss of the diversification benefit (alpha = 0). It seems that the distinct concepts of risk offsetting and diversification have been mixed by the authors of the revised framework. We remind that the former assumes high correlation between closely related risk factors (in particular within a valuation input matrix) and the latter assumes low correlation between unconnected uncertainties. The two effects coexist, so it is not logical at all to require having to choose one or the other effect. It is important to highlight that diversification factor is a way to handle the decorrelation effect between two uncertainties of different asset types which will no longer be taken into account with the new proposal.
The proposal put forward with the Consultation Paper fundamentally contradicts the meaning and the spirit of the diversification benefit, as previously elaborated by the EBA in its discussion paper on Prudent Valuation:
“The principle of allowing for a diversification benefit is based on the theory that an institution with many small valuation uncertainties may face a very different total valuation uncertainty when compared to an institution with one large valuation uncertainty”.
“Any diversification benefit should in any case only apply to certain AVAs as some relate to uncertainty around the fair value of individual positions which could be positive or negative (and which would not all be expected to be at the bottom of the range of plausible values at the same time) while others relate to reserves that are required for particular reasons that are only negative.”
Besides, the revised framework no longer allows securities with quoted prices to be treated as a curve (like a government bond often managed together with interest rate swaps) and specifies that each point of a curve/surface must be treated as autonomous valuation inputs. This would not allow measuring the uncertainty of the marking of a curve/surface as a whole by analysing dispersions of quotes of whole curves/surfaces that are provided by consensus services for the purpose of measuring MPU. Indeed, curve/surface cannot be distorted in erratic way by moving each point in the adverse direction of the exposure (long or short). This would result in AVAs unfounded and beyond reasonableness.
Banks consider it important to keep the possibility of defining a global curve or surface as a global valuation input especially when it is issued from a validated calibrated data model and for which curve or surface scenarios will be used to model plausible 90th percentile uncertainties.
Current revised framework, lead to an inconsistency of the valuation input definition between:
- Art. 9.1: valuation input = “every parameter in the matrix”, and
- Art. 9.4(b): “where the value of a parameter is extrapolated from parameters of the same valuation input”. In this case we understand that the matrix is the valuation input.
- Art 9.9: “Any weakness in the methodology applied shall be addressed by including a margin of conservatism in the determination of the estimate”. Some details and illustrative examples on the definition of such margin of conservatism would be also welcome.
Finally, under the revised framework, it is mentioned under article 9 paragraph 11 that institutions that apply the expert-based approach should report the largest individual MPU AVAs determined based on that approach. Is this reporting compatible with a sensitivity-based calculation as this would imply calculating the AVAs at valuation exposure level instead of position level?
As a rule, articles 9 and 10 have become more complex and many aspects will require further assistance on their expected application. This complexity will most likely generate discrepancies among peers regarding the implementation of the norm, thus obtaining results contrary to the objective of the RTS. For this reason, the authorities should aim to incorporate clear-cut definitions, accompanied by multiple examples for different cases, which illustrate how the guidelines should be interpreted.
Question 9. Do you have any comments with regard to the amendments to Article 12? If you do not agree with the amendments, please describe how you would adjust or design the requirements to meet the policy objectives that the amendments try to achieve. When giving your answer, please provide the rationale and relevant evidence supporting your proposal.
In particular, the CVA scope extension to Fair Value (FV) Securities Financing Transactions (SFTs) is disproportionate. FV CVA is not relevant when trading SFTs and no accounting adjustments for CVA are applicable. Moreover, the scope extension is inconsistent with other capital measures: FRTB CVA allows for the exemption of FV SFTs subject to materiality considerations. US NPR exempts all SFTs and UK PRA significantly limits the scope. Therefore, the CVA scope should not be extended.
Also, the alignment of the MPOR to be used with the regulatory one is inappropriate and will lead to unjustified additional conservatism. There is no evidence corroborating such an alignment. Moreover, it is inconsistent with the definition of the prudent value:
„[…] the value at which institutions are 90% confident that they will exit a position based on the applicable market conditions at the time of the assessment[…]”
Using a floored MPOR in in contradiction to applicable market conditions and adds a further layer of conservatism that is also not justified from a conceptual point of view in our opinion.
Another driver of unjustified operational burden is the requirement to take into account CVA correlations due to the prescribed granularity of application, without a necessarily meaningful change in outcomes. This requirement should be dropped as well.
In article 12 paragraph 4, the requirement to set the margin period of risk to be equal or greater than that employed for the purposes of determining the own funds requirements for CVA risk might create a double count of capital requirement. If there is a default, there is indeed a value slippage risk leading to insufficient collateral posted leading to a real counterparty risk. However, at the current date when default has not been materialised, valuation is about the purchasing price offered by a willing market participant. Default risk is rarely considered in the interbank market for fully collateralised derivatives or Securities Financing Transactions.
As correlations between risk factors used for the risk factor diffusion are usually not hedgeable and often calibrated historically, clarifications on what kind of AVA is expected for such risk factor (MPU, COC, MR, CP) would be welcome by the industry.
Question 10. Do you have any comments with regard to the amendments to Article 14 and 15? If you do not agree with the amendments, please describe how you would adjust or design the requirements to meet the policy objectives that the amendments try to achieve. When giving your answer, please provide the rationale and relevant evidence supporting your proposal.
In article 14, the use of the liquidity horizon set by Article 325 bd of CRR below which no concentrated position AVA is required is a welcome amendment that compensate for increased market risk charge when liquidity horizon is longer for less liquid risk factors and avoid double counts.
This amendment is consistent with the 2013 EBA CP on prudent valuation that stated “Concentrated positions AVA is only required where the prudent exit period exceeds the time horizon for the market risk measure used to calculate own funds requirements as defined in Article 365 of Regulation (EU) 575/2013.”
Referring to Article 14.1.(c) “where the prudent exit period exceeds the liquidity horizons that are assigned to the valuation inputs of the position in accordance with Article 325bd of Regulation (EU) No 575/2013”, Article 325bd of CRR requires banks to map each risk factor assigned to the trading desk for which they have been granted permission as referred to in Article 325az(2), i.e. FRTB IMA models, and prescribes the liquidity horizons for the risk factors. We would gladly welcome from the EBA a clarification whether these liquidity horizons shall be considered for CP AVA by an institution adopting FRTB SA only?
The reference to the trading desk liquidity horizon seems to suggest that the prudent exit period is only relevant to trading book positions. We would like to ask the EBA to clarify whether the requirements in this Article are applicable to fair value positions in the banking book, such as the fair value illiquid loan and equity positions in the banking book which are not capitalised under FRTB market risk framework?
Moreover, taking into account that cut-off levels of prudent exit period are aligned with the liquidity horizon prescribed by IMA VaR/Expected Shortfall models for market risk capital, does EBA interpret CP AVA charge as an additional charge beyond these liquidation periods, which are not covered by VaR or Expected Shortfall when the prudent exit periods are longer? In that case, should the main driver of the charge be the market price movement of instruments/valuation inputs as in the VaR or Expected Shortfall model and not bid-offer spreads as required in Article 14.1.(c) “where the prudent exit period exceeds …, they shall estimate an AVA taking into account the volatility of the valuation input, the volatility of the bid-offer spread and the impact of the hypothetical exit strategy on market prices”?
While Article 14 refers to the identification and AVA calculation for concentrated valuation positions, amended Article 15.1 condition (a) requires future administrative costs (FAC) AVA for a valuation exposure where the individual MPU, COC and Concentrated Positions AVAs do not imply fully exiting the exposure. We request EBA to confirm that the identification of concentrated positions shall be aligned with the exit scenarios, where, in case of derivatives, hedges corresponding to net valuation exposures and needed to close out the relevant valuation positions shall be taken into account.
In article 15, the precisions added in paragraph 1 are useful but:
- sub-paragraph 1(b) does not provide a definition of tradable instruments
- sub-paragraph 1(c) can be potentially misleading as all option positions require some form of dynamic re-hedging even when these options are actively traded and all risks could be exited with a measurable CoC, MPU and CP. There should be a precision that dynamic re-hedging to maturity of the position is the only possibility as there is no competitive market price for exit of the risk.
Furthermore, it is not clear from the current RTS what “future administrative costs” refer to. To avoid misinterpretation of the article, we kindly suggest EBA to consider rewording in Article 15 paragraph 3 that “Future administrative costs refer to all incremental staffing cost, incremental fixed costs and future hedging costs that are likely to be incurred in managing the portfolio but a reduction in these costs may be assumed as the size of portfolio reduces.”
Besides, in article 15, all non-linear positions are impacted, including liquid vanilla options. Additional charges will occur, even if the back-testing of Prudent Valuation charges is successful. In addition to that, the discounting rate will be disconnected from the cost of capital, we suggest removing this requirement if Prudent Valuation is successfully back-tested and to limit the application to portfolios with products that are in scope of the RRAO. Moreover, the discount rate should be linked to the cost of equity of the institution.
Question 11. Do you agree with the requirements set out in Article 19a and Article 19b? If you do not agree, please describe how you would suggest to revise those Articles and address the mandate on extraordinary circumstances outlined in Article 34 CRR. When giving your answer, please provide the rationale and any relevant evidence supporting your proposal.
The proposed relief measure will only be applicable in the event of a major market stress but there is no provision for market dislocations on significant, short-term market stress events (e.g. CS/SVB, Russia/Ukraine war). Therefore, we propose the introduction of a relief measure for significant, but not yet major market stress events with an alpha factor set at 0.58%.
Taking into account extraordinary circumstances with a mechanism to avoid a pro- cyclical effect of AVA is a welcomed addition. The use of the aggregation factor alpha is simple.
Article 19a provides examples of market factors to take into account to trigger the state of “extraordinary circumstances”. We agree that there should not be an automatic trigger from quantitative factors as no pre-set criteria will ever account for all possible cases and an element of human judgement is required. However, the regulation does not tell who makes the ultimate decision of declaring the state of “extraordinary circumstances”. It is written “EBA shall assess”. Does EBA act as an advisory body or as a decision body? It is unlikely that EBA has an executive decision and enforcement power given its regulation setting power and the principle of separation of powers. We understand that the ultimate decision lies in the European Commission. This regulation should precise this.
Question 12. Which of the two options presented do you consider more appropriate for the purposes of addressing concentration of UCS AVAs? When giving your answer, please provide the rationale and any relevant evidence supporting your proposal.
For the purpose of valuation positions to be deemed concentrated on certain counterparties, option 2 appears quite simple to implement but it seems arbitrary, as being part of the 5 largest UCS AVA has no relationship with concentration. The 5 counterparties may simply have a slightly larger AVA than others. Option 1 is even worse as it disallows the usage of diversification benefit for all UCS AVA calculations if a single position is determined to be concentrated (>10%).
The industry does not understand the rationale of detecting a CVA concentration issue only by comparing AVAs between different counterparties without taking into account any volume information or CVA levels relatively to the MtMs. Therefore, banks believe that both options cannot answer the raised concentration issue.
More fundamentally, both option 1 and 2 subjectively measure “inner” concentration and are irrelevant from a valuation perspective, be accounting, regulatory or in a resolution context.
Market participants/Valuers would not factor in these metrics when pricing individual or aggregated XVA exposures not to mention that 10% threshold / 5 most significant counterparts are arbitrary and without justification.
Alternatively, the question of “outer” concentration (i.e. the extension of article 14 to article 12) might be raised despite comments in footnote 5.
In conclusion, the intention of this amendment seems not entirely clear for the banking industry.
Therefore, we propose an alternative as the only viable approach in the form of:
- A requirement for institutions to perform systematic portfolio level analysis to identify concentrated positions (which may give consideration to the identification of individual exposures which exceed a threshold of 10% applied to Fair Value CVA).
- Only consider the removal of the aggregation factor provided the institution has not taken alternative, bespoke measures to address the specific risk in the UCS AVA, to the satisfaction of the competent authority.
- Apply the removal of the aggregation factor only to the counterparty being seen as concentrated and not to the whole UCS AVA.
Question 13. Do you have any comments with regard to the amendments introduced in the Annex? If you do not agree with the amendments, please describe how you would adjust or design the requirements to meet the policy objectives that the amendments try to achieve. When giving your answer, please provide the rationale and relevant evidence supporting your proposal.
The rationale leading to the removal of the method 2 of aggregation does not find us entirely agreed. Furthermore to method 1, method 2 leads to a CET1 level independent from the choice of Fair value methodology and level of conservativeness embedded in Fair value. Method 2 relies on a precise definition of the Expected Value (EV) that is insufficiently defined in the current framework in particular for CoC (expected exit cost for Prudent value vs principal market exit cost for Fair value). However, for MPU and Model Risk, Expected value is clearly defined as the value obtained with valuation models without conservative layer (valuation theory defines the value as the expectation of future cashflows).
With the use of method 1, the effect well explained in figure 1 page 9 of the consultation paper (impact of absence of fair value adjustment) leads the revised framework to propose a form of penalty (alpha = 0) when fair value adjustments are deemed insufficient (defined as not commensurate with the adjustment other market participants would consider when determining the reference fair value). This may lead to inconclusive challenges by the supervisory authority as to whether fair value adjustments are insufficient as there are no objective criteria for this. Setting an appropriate level of EV in method 2 with more regulatory guidance would lead to better EU level playing field in the determination of CET1 without the need to challenge each firm’s accounting practice with subjective judgement. Indeed, the accounting fair value in IFRS 13 does not require a precise confidence interval (even if a 50% confidence and above is deemed acceptable by accounting firms) and EBA is overstepping its mission when making an assessment of insufficiency. Such competency of judgement lay down to the audit firms signing on the accounts of institutions and having such conflict of competency is not seen as an appropriate one.
The revised framework is too subjective with a potential high impact on AVA amounts that each bank cannot anticipate since it will depend on other bank’s FVA amounts and policies. It would be better to link this point to internal and/or external review conclusions regarding the consistency of the FVA amounts in regard with the bank’s positions and valuation policies.
The diversification benefit has been designed to account for the structural de-correlation between uncertainties across all sources and when measured at a conservative confidence level. There is no theoretical rationale for adjusting it to offset various flaws or bias in the FVA or PVA framework.
EBA could alternatively expand the hypothetical portfolio exercise for IMA benchmarking to assess consistency in FVA and PVA calibration across firms.
The use of tenor dimension reduction with the Variance ration test (VRT) and loss of diversification benefit (alpha = 0) is discussed with question 8.
Regarding the need to reintegrate unadjusted IPV difference into the CET1 measure through the AVA, we can understand the logic, yet there are two elements to consider:
- If IPV is not adjusted because the independent source is not proven to be more reliable than the trader marking, there is no more reason to adjust CET1 than Fair value apart from recognising that this is a source of uncertainty that should benefit from diversification benefit. For very liquid valuation inputs such as standard yield curve, equity spot and fx spot, an IPV with accounting P&L impact may be considered as not necessary as long as an independent control ensuring that FO mark is consistent with mid-market values or when marks are automatically fed from a validated source. For such cases banks would like to still benefit from the current AVA formula.
- Only considering the negative unadjusted IPV differences may lead to a large negative adjustment as even if these differences are a form of valuation noise, the cumulative negative noise may be important when the total noise is not. The cumulative negative IPV would be also dependent on the granularity at which the IPV is performed /recorded. At the end, it is fairer to simply record all IPV differences in Fair value than the proposed AVA treatment. The mathematical formula to be used in order to reintegrate the unadjusted IPV amount must be clearly defined with a mathematical formula in the Annex in order to avoid any misunderstanding. Based on the example p.12 we understand that the proposed formula is just a way to reintegrate “exactly” in the AVA amount the unadjusted IPV. In this case all the IPV shall be reintegrated and not only the negative part.
A more cosmetic point to highlight is that the Annex should precise that the alpha values of 0.5 or 0 should only apply in normal circumstances when the condition of article 19a and 19b are not triggered (extraordinary circumstances).
Finally, on the format of the annex, we think that all these conditions for setting alpha = 0, if they are adopted, should be set in the RTS articles themselves and the annex should only contain the aggregation formula as it is the case in the current framework.
Question 14. Do you have any other comments on this consultation paper? If you do not agree with any of the proposed requirements, please describe how you would adjust or design them in order to meet the policy objectives that the proposals try to achieve. When giving your answer, please provide the rationale and relevant evidence supporting your proposal.
Operational risk AVA:
We would like to make a point on Article 17 OpRisk AVAs, even though there is no respective question included in the consultation.
In particular, we think that the AVA size is not a good indicator of Operational Risk. As the AVA is applicable to all products, regardless of their complexity, it penalises conservative business models such as market making.
Additionally, from a conceptual point of view, AVA size is not a good indicator of Operational Risk as it disincentives institutions from implementing conservative practices due to the subsequent increases in Operational Risk AVA, given it is a function of the overall MPU & CoC. Besides, this is a double counting of the operational risk RWA on valuation, as the full AMA risk is replaced by the Standard Approach of Operational risk. Adding another dedicated Operational risk charge will duplicate what was earlier in AMA and now in the Standard Approach.
We would therefore kindly request Operational Risk AVA to disappear from the new RTS.
General comments:
The revision of the Prudent valuation RTS was not mandated by CRR apart from the definition and impact of the extraordinary circumstances to avoid pro-cyclical effect of capital measure (as mandated by CRR3 article 34). It appears that the highest motivation for this proposed revision is to create a better EU level playing field as there seems to be a diversity of application of the current framework.
The revised framework is indeed more prescriptive, while it is at the same time so complex and open to interpretation, hence it is unlikely to lead to a greater level of harmonisation of application. Furthermore, the cost in terms of capital requirements and operational complexity is considered prohibitive by members. Adding that this will not be implemented in the US and also that the PRA has not expressed any will to follow the same road for the UK, it is foreseen that this proposal will be a driver of supremacy of the non-European players on the European market.
If, for the purpose of EU harmonisation, a fall-back approach is meant to be used whenever an economic method would require subjective judgements, this is probably not the right way to motivate a financial institution to analyse their valuation uncertainty with a critical mind. A method that could be challenged by supervisory authorities is always better than an arbitrary forfeit unless harmonisation and simplicity are the only goal.
Prudent value regulation is only applicable in the European Union. It means that US banks that are the main competitors of the largest European banks are not affected by it. In addition, the revised framework will also not apply to UK banks given that EBA regulation has no longer effect in the UK, apart from existing ones integrated in UK law. We are not aware that PRA intends to copy this revised framework in the UK regulation. This is contrary to the level playing field intention at international level.
As the result of all these comments, we believe that the revised framework was not requested, therefore it is inappropriate and it should be discarded. The RTS should rather focus on the mandate given by CRR through Article 34.
Appendix : Proposed methodology (uploaded) to offset accounting adjustment from deferred Day One Profit to AVA.