The Association for Financial Markets in Europe (AFME) welcomes the opportunity to comment on the EBA’s Discussion Paper on the future changes to the EU-wide stress test (EBA/DP/2020/01) (referred to hereafter as the “Discussion Paper” or “DP”)

Our response starts with our overarching comments on the proposal discussed in the DP, which highlight our main concerns, and is followed with more detailed responses to the individual questions posed in the consultation.

I. Overarching comments

The future stress testing framework in Europe should build on the principle objective of the exercise, which is a microprudential exercise to identify risks in banks and assess their capital adequacy. The key strength of the framework is as a tool to facilitate a better understanding of these risks via supervisory dialogue, with disclosures to the market being an important by-product of the exercise, but not the principal objective. As we will elaborate on below, the disclosure of multiple perspectives on the outcome of the test risks entirely dominating the process for no added benefit. We therefore do not support the dual-leg approach being proposed and strongly recommend that the EBA retain a single leg approach with a single published outcome going forward.

This being said, we do think that there are important improvements which can be made to the framework to make it of greater relevance to all stakeholders involved, while maintaining a reasonable cost/benefit balance.

For instance, a key issue which needs to be dealt with is how the “severe but plausible” stress is achieved in practice. Addressing this matter will help improve the overall relevance and credibility of the exercise, while helping to reduce some of its complexity and associated costs. This is also one of the key issues flagged in the European Court of Auditors’ report 2019 report on the 2018 stress test exercises . In our view, the imposition of cumulative methodological constraints is not the best approach to achieving the appropriately severe outcome, but rather results in obscuring risk sensitivity and distorting business models and risk profiles which in turn reduces the relevance of the exercise for banks and supervisors. Rather than using methodological constraints to drive the level of the stress, it is the design of the scenario that should result in the appropriate shock on banks and, if this perceived as being sufficient severe, it will improve the credibility of the exercise in the eyes of the market.

The other key issue which would need to be addressed in the remaining lack of clarity on how the outcome of the stress test informs individual Pillar 2 Guidance and the link between P2G and other buffers more broadly. While this is also a complex issue, it is necessary that it be resolved to further increase the relevance of the exercise. There are likely to be important lessons learnt from the supervisory flexibility afforded to the industry in the context of the present Covid-19 crisis which can hopefully contribute to resolving this matter going forward.

II. Responses to the consultation questions

Proposed new framework

Question 1: What are your views on the proposed framework in general?

AFME’s general views on the proposed framework:

AFME agrees that the existing EU-wide stress testing approach would benefit from improvements to increase the utility of its outcome while reducing the costs involved for both the supervisors and the banks which carry it out.

The regulatory stress testing framework has evolved from initially being a crisis management tool, expanding its objectives over time to cover the multiple needs of supervisors, the market and banks themselves. We therefore welcome the recognition by the EBA, and others (such as the ECB), that the complexity of these competing demands has led to an outcome that does not fully satisfy any of these objectives, whilst placing significant demands on both the supervisory community and banks.

We also broadly agree with the issues identified in section 2.3 of the Discussion Paper describing concerns with the current framework and the rationale for change.
However, while we very much welcome the ongoing reflection on the future approach, AFME does not support the introduction of the dual-leg framework described in the Discussion Paper. While we think that improvements to the framework are required, a single disclosed outcome is a feature which should be preserved.

Our responses to the questions below will set out our underlying reasoning in further detail but, in general terms, we think that the proposed approach will increase the complexity and burden of the exercise for the firms required to take part. At the same time, there is no added value in having dual sets of outcomes being published as only the supervisory view will prevail in the eyes of the market. This has been confirmed to us in discussions with the analyst community. Additionally, without transparency on supervisory challenger models or top down methodologies and more clarity on the methodology used for determining P2G, banks will not be in a position to explain discrepancies in outcomes.

Moreover, the supervisory leg set out in the DP involves an approach which can objectively be described as being more top-down (or mechanistic) than the current framework, even if this is not yet formalised outright as such. We do not think that capital guidance is best determined based on such a supervisory view as this will inevitably not reflect banks’ idiosyncrasies, including their business models or their specific abilities to mitigate certain stress situations. If firms are expected to integrate stress testing into their internal risk management frameworks, it is also key that the tool used for determining a regulatory outcome can be appropriately integrated into these internal risk assessments. This will not be the case with any approach largely based on a top-down approach. Finally, we are also concerned that the proposed approach will, to a large extent, remove important elements of supervisory dialogue whereby supervisors can question banks on their stressed outcomes and seek to better understand the specificities of their business models.
As such we do not think the proposed approach represents an overall improvement to the current set up and recommend that a single-leg approach be retained as it the case today. Nevertheless, the new framework could benefit from a number of refinements which we set out below.

AFME’s suggested way forward:

Rather than the proposed dual leg, our preferred approach would be to continue to rely on a bottom-up stress test, where the stress test outcome, including P2G setting, relies as far as possible on firms’ ICAAP stress testing approaches. This is because these are, by definition, designed to reflect the behaviour of a firm under stress together with idiosyncratic risks of its individual activities, markets or products and will inevitably be more precise when it comes to setting P2G capital outcomes than top down models developed by supervisors. They will therefore also provide the most useful information on banks’ risk profiles to the market.

For firms, using a single set of internal models for regulatory stress tests, ICAAPs and baseline challenge would present a significant simplification over the current piecemeal approach and would be consistent with supervisory recommendations that stress testing be embedded into their internal risk management processes.
We recognise there are concerns on the part of the supervisory community with respect to greater use of internal models, which have led to them preferring the dual-leg approach. We would like to address some of the key points raised by supervisors and reflected in the DP:

• Supervisory concerns on quality of firms’ internal stress testing capabilities: major EU supervisors such as the ECB are currently undertaking a multiyear review of firms’ ICAAPs. This can be leveraged to include assessments of stress testing capabilities for use in the context of regulatory stress testing. Only firms who would satisfy their supervisor of their capabilities should be allowed to use internal approaches (fully or to a greater extent than today), other firms could remain under the existing approach (or a more top-down, mechanistic approach) until they are able to do so. Strict model risk management frameworks would need to be implemented, similar to what has already been put in place elsewhere for example via the ECB’s TRIM exercise, and firms using internal capabilities would provide standard disclosures on the features of these models to ensure transparency and market discipline.
• Supervisors will retain the ability to quality assure firms’ results: including via benchmarking outcomes and to ultimately impose overrides in the event banks are unable to satisfactorily substantiate their results
• Workload for supervisors: while the review of ICAAP stress testing capabilities should not be underestimated and would take a number of years to complete, the workload involved in quality assurance for supervisors will be reduced as they progressively build up reliance on bank’s internal governance and model risk management processes under their ICAAP stress testing reviews.
• Comparability between firms’ outcomes: Comparability would be ensured via the continued use of common starting points, disclosure templates and a number of appropriate common methodological approaches, for instance to ensure that dynamic balance sheet approaches do not generate unrealistic assumptions or outcomes at the aggregate system level. Disclosure on banks’ internal approaches would facilitate the market discipline pillar. Finally, when considering comparability between outcomes, it is important to note that, while there is a need for a consistent approach to be taken in setting P2G levels, by definition P2G is required to be firm specific – something which cannot be achieved under a supervisory leg.
• Banks’ internal stress test projections can be fed into the supervisory leg: The DP makes a mention of this, but in the general context of the document, we do not see the intention of this being more than is currently the case. In fact, the DP highlights the extra workload involved in identifying areas where the supervisory leg result could be adjusted.
It is the industry’s clear preference to maintain a single-leg approach to the regulatory stress testing framework. From previous interactions with EBA staff and the many combinations and options presented in the Discussion Paper, it would seem that there may be general agreement between the supervisory community and the industry that there indeed needs to be greater alignment between the way an individual bank sees its stress test outcome (bottom-up view) and the supervisor’s view of the bank (top-down view). We think that rather than introducing multiple layers of tests and outcomes, the question becomes one of how to achieve this converged view most appropriately.

Our recommendations for achieving this convergence are to let firms use internal stress testing approaches for regulatory stress tests, subject to the following:

1. Ensure that model risk management principles, similar to those which already exist for capital requirement model purposes in the ECB’s internal model guide are extended to internal stress testing models and that a review of these principles is prioritised in the ongoing assessments of ICAAPs. These principles include:
• Banks have an established definition of a model and maintain a model inventory
• Banks have implemented an effective governance framework, policies, procedures and controls to manage their model risk
• Banks have implemented a robust model development and implementation process and ensure appropriate use of models
• Banks undertake appropriate model validation and independent review activities to ensure sound model performance and greater understanding of model uncertainties

2. Banks should produce information according to documentation guidelines developed by the EBA setting out common standards of communication on their stress testing models for the purpose of supervisory review (which should be prioritised) and also including “transparency sections” which could be disclosed to the market together with the outcomes of the stress test, providing analysts with key information on how the results are obtained.

3. Quality assurance processes could include benchmarking portfolio tests to assess the effectiveness of level playing field and ensure the exercise is fit for determining supervisory outcomes. For example, if an individual bank continues to display material, different results compared to a peer group when it applies its own models to a common benchmarking portfolio, this would be an indication that supervisory conservatism may be required.

Beyond the above, it is recognised that the process of obtaining supervisory confidence in internal stress testing capabilities will take time. We therefore suggest that the removal of constraints be progressive and aligned with the above supervisory review of ICAAPs according to a roadmap. We provide our suggestions below on which constraints should be removed as a matter of priority as they impede the realism of the stress test to the greatest extent.
Moreover, banks which would not satisfy supervisory requirements for using internal models would continue to use the current approach.

Status quo: advantages of the current approach over the proposed approach:
Given the significant issues we see with the proposed approach, and the lack of cost savings it represents, if there is no willingness to pursue and invest in a true bottom-up approach to regulatory stress testing, AFME would recommend retaining the existing framework. While the need for improvement of the existing approach is acknowledged, we nevertheless consider it is of higher utility than the proposed approach as it places greater reliance on supervisory dialogue and supervisory understanding of an individual firms’ idiosyncrasies under stress which are essential for setting capital guidance.

This being said, while efforts to simplify and accelerate the current exercise have already been noted in practice (e.g. for the 2020 exercise) and are welcome in principle, we are concerned that the reduced number of submission cycles, particularly when accompanied by an accelerated “comply approach”, is effectively already introducing greater supervisory overrides with the potential to mask underlying idiosyncrasies further. This does not, in our view, represent a benefit in terms of realism.

As alluded to in the question, the DP envisages several options for the supervisory leg in the dual approach; however, it would clearly be an approach where there is less supervisory dialogue and a greater use of supervisory overrides of banks’ inputs compared to the exiting approach. We view this is another form of top-down approach.

Supervisory overrides are based on peer benchmarks and challenger models which are not transparent to banks or the market. The opacity surrounding the conception of benchmarking criteria may prevent the EBA from duly considering the specific risks that each institution could be exposed to, therefore potentially contradicting the institution-specific nature of the stress test exercise, or its microprudential objective and crucial use in determining individual P2G levels. At a minimum, information on the design and calibration of these should be disclosed. The same comment applies if explicit top-down approaches are introduced for certain portfolios/risk categories.

In cases where the supervisory leg would consider some bank leg components, the relevance of the exercise might be further improved because such an approach would take into account idiosyncrasies of the bank that are captured only in the bank leg. However, it would require more interactions between competent authorities and banks to identify areas where the supervisory leg result could be adjusted.

All in all, the supervisory leg is at this stage rather undefined – the QA process involved is not clear, to what extent and for which portfolios or risk categories mechanistic approaches/overrides and use of top-down models would be used is not clear, neither is the intended relaxation of constraints or the statist balance sheet assumption. The workload potentially involved under this approach is therefore difficult to assess.

Finally, we are concerned that the absence of transparency on the calculation methodologies for the supervisory leg would make it impossible for institutions to anticipate, or understand, capital ratio depletions resulting from this leg. Explanations to the market of differences between the supervisory and the bank legs cannot be done without an adequate understanding of the supervisory calculations.

In addition, we believe that supervisors have the duty to motivate their decisions in terms of final outcomes of the stress tests as these are communicated to the public and strongly influence financial markets.

As noted above, we do not support the introduction of two legs to the stress testing framework but are in favour of a maintaining a single approach which relies as much as is feasible on ICAAPs, and with better designed methodology constraints than today’s framework.

If two sets of capital depletion information are published, the market will in practice only take that of the supervisor into account, disregarding information published by the bank, even if the bank tries to produce a reconciliation between the two approaches. Not only is the reconciliation expected to be burdensome and time-consuming for banks, it will not be of value to the market which is principally interested in understanding the link with the stress test outcome and the determination of any resulting supervisory decisions including P2G setting. As this is the outcome of the supervisory leg there would be no practical benefit in conducting or publishing the outcome of a bank leg.

Moreover, as already mentioned previously, banks are unlikely to be able to fully explain or reconcile completely the differences between their legs and the supervisory leg if there is no transparency on top-down or challenger models used under the supervisory leg. As long as these remain undisclosed, reconciliation efforts are likely to be seen by the market as a relatively fruitless exercise.

If ever the EBA were to pursue an approach with two distinct legs, it will therefore be essential to adopt the alternative proposal discussed in the DP whereby a single view of CET 1 depletion is published.

As noted above, AFME considers that there should be a single leg with a single outcome. This would de facto involve taking banks’ views of their capacities to react to the stress scenario into account when determining the final outcome of the exercise.

We see the proposed framework introducing substantial costs without necessarily producing benefits for the market or for banks. The principal sources of increased costs arise from:
• Two submissions and processes for banks (and supervisors) to manage, in addition to what banks will do separately for internal risk management purposes
• Significant efforts to explain differences in approaches (unless banks decide that the bank leg is effectively the supervisory leg), without the expectation that these efforts will be recognised by the market, which will principally be interested in understanding the stress outcome and associated consequences or supervisory actions

We recognise that the proposed approach will result in less supervisory dialogue or interactions with banks compared to today, which may be perceived as a cost saving for supervisory authorities. We would argue however that this ignores the benefits of the additional information supervisory authorities can gain into bank-specific situations via such interactions. Moreover, supervisory dialogue will likely be made more efficient if sufficient and appropriate transparency on challenger models is provided.

We consider the following constraints to be those which should be removed as they obscure the realism of the exercise, have the greatest effect on underlying economics and the potential to distort business models.

Credit risk:
• Banks should be allowed to recognise more dynamism with respect to short term exposures
• The imposed assumption that stage 3 assets cannot cure is overly conservative and should be reviewed (creating more consistency with LGD modelling approaches), taking into account the impacts of the regulatory prudential provisioning backstop and supervisory expectations regarding NPL ratios.

Market risk:
• The approach to market risk should be entirely reviewed to achieve a better relevance vs cost balance, particularly for banks with significant global market activities.
• The industry has pointed out on several occasions that the prescribed floors on trading losses can override a bank’s full revaluation results (particularly around tail risk hedges). The full revaluation exercise is particularly resource consuming for both banks and supervisors but does not yield any benefits as the final market risk impact is driven by the floored approach. When combined with other constraints in this risk area (such as client revenue floors and net trading income caps) the overall approach is unduly conservative.
• The blanket exclusion of one-off adjustments to net trading income (e.g. due to exits from an entire business line) also obscures underlying economics. Economic risk management dictates that market risk losses in such cases will be zero.
• In the current methodology, at least 25% of client contributions is lost given the floor mechanism. This should be amended and brought in line with other revenues such as fees and commissions (i.e. 10% for modelling banks). Banks could use past data (e.g. from the 2020 stressed environment) to demonstrate that client contributions can be maintained even in a volatile environment. Client contributions should also be stressed separately from trading P&L (the client contribution prescribed formula refers to historical NTI). Trading P&L is already stressed separately via the Full Reval template.
• The scaling factor applied to the end of year VaR is justified by EBA as a means to correct possible end of year window dressing. However, banks that can demonstrate the absence of window-dressing should not be subject to this factor.
• The current methodology for market reserves should be improved. The behavior of such reserves during the present crisis should provide an opportunity to improve both modelling in banks and draw lessonss which could be reflected in the EBA’s methodology.
• It would be more appropriate to arrive at severe outcomes via input scenarios rather than methodological constrains and we suggest that market risk losses could be based on an imposed level of severity which is aligned with the adverse macroeconomic scenario and the correspondingly severe internal stress scenarios used in market risk models (which are approved and reviewed by supervisors).
• Note that the floor mechanism referred to above also creates a similar situation with respect to the economic reality of CVA risk which should also be reviewed. In particular, gains arising from CVA hedges should be eligible for offsetting CVA impacts.
• Exposures are to be reported net of stressed collateral. However, no collateral that is to be called beyond what is held at the reference date may be assumed. Not allowing the inclusion of future margin calls in stress calculations is inconsistent with regulatory requirements and the EBA’s methodology itself – as assuming no collateral is exchanged is equivalent to considering the default of all market counterparties – which is not part of the given assumptions of the stress test.

Net Interest Income:
• Without allowing a greater use of internal models, the Net Interest Income (NII) component of the exercise simply fails to be risk sensitive. Firms should be allowed to rely on Pillar2/ICAAP approaches as much as possible to reflect their underlying businesses and risks.
• This would apply in particular for interest bearing sight deposits (i.e. non-regulated, other floating rate sight deposits) where there should not be a prescribed formula.
• The quality assurance process of banks’ NII projections could be facilitated via dedicated, jurisdiction-specific workshops to ensure comparability while reflecting the specifics features, market practices and legal requirements in place in the various countries of the EU.

FX risk:
• In order to ensure that international businesses are not unduly penalised over domestic ones, a welcome simplification would be to remove FX risk, particularly as the added benefit in a capital depletion exercise is not clear.
• At present, banks with business in countries with weaker currencies are particularly affected given the mismatch between the treatment of NII versus the cost base and associated credit RWAs.

Conduct losses:
• The so-called QA floor should be reviewed as past losses are not reflective of future losses.
• At the very least, the constraint could be adapted and refer e.g. to a percentage of a more relevant “business” metric such as revenues, etc.

Non-interest income & expenses:
• Banks which have a proven track record in modelling non-interest income and expenses should be allowed to do so.
• Incentives should be given to banks to model operating expenses in the same way as is done for fees, commissions and client contributions. This could be achieved through the reduction of the haircut to 10% when banks have valid internal models, and a higher haircut when they do not.
• The share of profit of investments in subsidiaries: the 75% cap for banks with an internal model is overly conservative, affecting in particular financial conglomerates with a consolidated insurance business. This constraint is distortive of business model choices.

We prefer a single approach, with a limited number of constraints designed to ensure comparability and conservatism where necessary.

In addition to there being no impact of supervisory overrides/top-down models, other differences will be the result of having a dynamic balance sheet and banks recognising mitigating actions.

Such guidance could be useful but only in the context of a single leg approach to ensure necessary comparability and transparency of banks’ outcomes.

As already noted, we do not support two legs and consider that there should be a single leg, with a single set of disclosures as is the case today.

The most useful information for the market is understanding the maximum CET1 depletion and the key drivers behind this, including RWA and line by line P&L changes. Beyond this, very granular, 3-year projections at specific portfolio levels are not necessarily of use to the market, which tends to rely more on the EBA’s Transparency exercise for granular, yet comparable, information as this does not reply on relatively long term projections. Projections are only useful at a higher level of aggregation.

To ensure that the regulatory stress test exercise can be better integrated into firms’ internal decision-making processes, the type and level of information provided to the market would benefit from being consistent with internal management views to the greatest extent possible, while of course recognising the need for comparability across the industry. In our view, this implies that FINREP/COREP structures should be used as a starting point for ST disclosures, to ensure standardisation of the information presented, but that these should remain at the level of key info lines (i.e. risk and P&L areas), without drilling down into individual portfolio level information.

Finally, we wish to stress that we have particular reservations with the suggestion in the DP that banks should disclose their results under the baseline scenario at the level of granularity prescribed by the current transparency templates as this would effectively amount to disclosure of business plans for the next years.

See above. The particular challenge is that banks are unlikely to be able to entirely explain differences due to the limited disclosure of supervisory models.

See above. The particular challenge is that banks are unlikely to be able to entirely explain differences due to the limited disclosure of supervisory models.

The extent to which the stress test outcome impacts the determination of P2G, together with a clear understanding of this mechanism, is critical to ensuring that banks hold the appropriate amount of capital to be able to deal with a stress situation, as well as for the exercise to be market-relevant. AFME therefore considers that the link between the stress test and P2G is paramount.

At present however, it neither clear to banks nor to the market how the stress test outcome feeds into P2G, with analysts using basic rules of thumb such as a 100bp P2G applying across the board. This is counter to the objective of P2G being bank-specific.

It is our view that before any form of P2G disclosure can be considered, it will be necessary to ensure that the market has a full understanding that P2G is not a requirement and that its use does not result in a trigger of any MDA restrictions. Moreover, this issue will also have to be considered across the multiple metrics and triggers of the prudential framework.

Supervisory communications confirming that banks can use P2G buffers in the present crisis should be leveraged to ensure that the market comes to having a complete understanding of the role of this feature before any form of disclosure of P2G levels is considered.

The following additional aspects also need to be clarified prior to considering P2G disclosures:
• How does P2G setting interact with levels of excess capital banks have? For instance, if a bank has a relatively large excess CET1, would it have a lower P2G than a bank with a smaller excess, all else being equal (same maximum stress impact, etc.)?
• If the maximum stress impact does not lead to a breach of the TSCR, there should be no P2G (no “floor to P2G”). What would the conceptual justification be for a P2G floor?
• How are potential overlaps between P2R and maximum stress impacts determined, and can they be addressed via the determination of P2G?
• How will CAs and NDAs coordinate to determine the amount of the CCyB which may be consumed in stress (if any)?
• How will “credible management actions” be dealt with in P2G determination in the future framework? -
Once there is more clarity on the above, as well as on the future framework more broadly, we would very much welcome the opportunity to engage further with the EBA and Competent Authorities on possible disclosure approaches.

We understand this question is posed in the context of the bank leg but as explained above, we do not think that there should be two separately published outcomes. Moreover, until there is greater clarity on the link between P2G and stress test outcomes, as well as on the future design of the framework more broadly, we would suggest that management actions, capital plans and dividend policies are better communicated based on real events and performance rather than on the basis of a hypothetical stress test exercise. We would however welcome the opportunity to further engage with the EBA and Competent Authorities on these communication to market matters once the framework is more settled.

As already noted in response to previous questions, although we agree with the EBA’s assessment of the existing framework, we do not consider the proposed approach to be preferable against any of the assessed criteria. From the point of view of banks, the approach would clearly be less relevant, transparent and more expensive. We think this is also true when considering the perspective of all stakeholders more generally:
• Relevance would be reduced as the capital depletion would be solely based on the supervisory leg which would more constrained and less reflective of banks’ individual risk profiles.
• In terms of transparency, the “double disclose” would likely add a layer of confusion and complexity, with the market ultimately only relying on the supervisory leg but lacking the means to fully understand how it has been calculated.
• From a cost effectiveness perspective, the proposed approach may be less expensive for supervisors if there is less QA to perform. However, this will come at the cost of reduced relevance of the outcome and a lesser understanding by the supervisor of the bank’s specificities and own views of its risks. Banks on the other hand would have to run multiple variations of stress testing (supervisory leg, bank leg, ICAAP, etc.) which will not all be relevant for their internal business purposes and will be faced with the (near impossible) challenge of having to explain discrepancies in outcomes to the market. As such, the costs of the proposed approach for banks are very high, with no added benefit being visible.

Our comments relate to the overall exercise:

Relevance: Likely to decrease over time with the increase in top-down, mechanistic elements. Relevance will strongly depend on the quality of the exchange between banks and supervisors

Comparability: Unchanged compared to today – it is challenging to measure the comparability criteria in practice

Transparency: Minimal as markets are really interested in the elements that drive the P2G

Cost-efficiency: Potentially more cost-efficient for ECB and EBA, but more expensive for banks

With regards to scenarios, we think that the focus should first be on achieving the appropriate consistency within the context of the single scenario. This applies in particular to the consistency across geographical areas and the scenario should not be the simple sum of different “crises” in different geographies.

Adding further scenarios when their design is subject to complex governance questions and when the framework itself is in flux appears to be an unnecessary source of complexity and cost at this stage. We would nevertheless be keen to discuss this matter further in due course with the EBA and macro-prudential authorities.