We welcome the EBA’s initiative to carry out a fundamental review of the methodology of the current European stress testing exercise. However, it should be remembered that a fundamental review will not necessarily result in a fundamental change of the methodology. In fact, we reject the introduction of the proposed 2-legged approach and regard the status quo as significantly better. A change of the methodology would only be acceptable if the key stakeholders, especially the relevant banks and supervisors as well as the relevant public (including analysts and journalists), regarded the envisaged proposals as an improvement of the status quo, and the changes would not be implemented at the expense of one or more parties. This is not to be expected in this case as the alleged advantages of the proposed changes will at best and unilaterally benefit the supervisory authorities. For the institutions and the public, the disadvantages will prevail:
Relevance: Under the new framework, there would be two outcomes for the same adverse scenario; one from the bank leg (ostensibly under the banks’ responsibility) and one from the supervisory leg (sole responsibility of the supervisor, cf. text no. 54). Market participants would thus receive two answers to the question of the bank’s resilience in the given scenario. We believe that, overall, the bank and supervisory leg framework in the proposed model leads to an almost uncommunicable message, which ultimately reduces the relevance of the exercise for third parties. This is particularly true since neither the bank leg nor the supervisory leg can reasonably reflect an individual institution’s specific risk situation in the proposed form. Furthermore, in connection with the envisaged partial employment of top-down elements in the supervisory leg, the question arises why the supervisory models - e.g. in the credit risk area - should be superior to the banks’ models. It is also entirely unclear whether the supervisory models will be subject to similarly rigorous validation as the banks’ own internal models. Insofar as different outcomes are also affected by the use of undisclosed supervisory models, market participants cannot even be provided with a meaningful reconciliation to help them with the necessary interpretation.
Comparability: As before, the supervisory leg continues to provide only limited scope for the inclusion of the specificities of the business model in the outcome figures. The result is systematic discrimination against certain business models. The proposed bank leg represents only marginal progress. Although its results are standardised, they are not comparable in the sense of a (fair) level playing field. Instead of a comparable relevance of outcomes, the suggested approach delivers apparent comparability at the level of the calculation formulae. All in all, by adhering to methodological guidelines and constraints within which the scenario effects are allowed to occur, the exercise fails to provide any comparable information.
Transparency: In our view, there is a danger that the top-down adjustments of the supervisory leg will be implemented without sufficient quality assurance and discussion with the banks, and will be adopted without change as the basis for the P2G setting in the SREP. Such approach would be accompanied by a significant reduction of the transparency of the projections. The outcomes of the supervisory models are not plausible and cannot be verified (black box), affecting negatively the plausibility and transparency as well as the control relevance for the banks, and ultimately also the relevance of the exercise. However, precisely the dialogue with the supervisor in the quality assurance context is a decisive element in order to arrive at an appropriate assessment within the framework of the microprudential assessment of institutions’ capital adequacy and resilience. Moreover, transparency would decline in practice since market participants would now have to deal with three different risk assessments in order to obtain a complete picture: the bank’s internal assessments, the supervisory leg and a bank leg which, as the proposal stands, will by no means represent the bank's view.
Even if the respective outcomes were used for separate purposes (bank leg: transparency and market discipline through detailed publication by the market, supervisory leg: subsequent setting of the P2G), and if the outcomes varied in terms of the scope of the publication, the question arises how significant the result of the labour-intensive bank leg would be. To assess this result and differentiate it from the supervisory result, market participants would have to view further information in addition to the banks’ granular publication in order to differentiate the bank's internal assessments from the constraints applying to the bank leg. From the banks’ perspective, it is questionable whether this would raise the level of transparency and relevance of the results.
One could argue that it is an advantage that the proposed framework provides banks with the opportunity to present their individual perspective in the interest of improved transparency for investors. However, the banks already have this opportunity: Nothing speaks against banks presenting an alternative result to the public that shows the bank’s own perspective of the effects of the given scenarios and the mandatory methodology that must be applied. In fact, market participants appear to be exclusively interested in the outcome that is used to set the P2G. Ultimately, the approach referred to as ‘bank leg’ is a bogus claim, as institutions by no means have the alleged ‘ownership of results’, must adhere to the methodology of the EBA in key aspects and may, at most, disregard some of the guidelines.
Moreover, the proposed approach lacks any transparent rules describing how the P2G setting should emerge as the outcome of the microprudential analysis. Aside from market transparency, this is the key function of the EU-wide stress test and should therefore be clearly worded.
Cost-efficiency: Via the reduced need for quality assurance and discussion with the banks, especially in the supervisory leg, the proposed framework primarily means a cost reduction for the supervisor. Even in the supervisory leg, there is no real benefit from reduced quality assurance, since the banks are in any case required to understand and reproduce, in detail, the outcomes of the supervisory leg. On the other hand, the bank leg imposes significant additional effort on the banks (preparation of the bank leg, reconciliation, potential additional downturn scenarios, potential additional multiple and exploratory scenarios, comparison with ICAAP stress test calculations). We perceive a reduction in effort solely on the supervisory side, not on the banking side: lower granularity of the publication of the outcomes for the individual banks and reduced dialogue with the banks in the context of the QA activities.
All in all, we expect high additional efforts and little benefit and, hence, reduced cost-efficiency of the exercise.
Box 2, text no. 44, also hints at the possibility of a full top-down approach after 2022 once the supervisor has collected more data. This paragraph gives cause for concern that the supervisor intends to substantially revise the data requirements and the calculation method that banks must use with every exercise, which would lead to high implementation costs for banks. In general, we are extremely critical of any expansion of the top-down approach and reject the idea of models based on the US approach (CCAR).
As set out in the Discussion Paper, we also regard the EU-wide stress test primarily as a microprudential exercise which aims to ensure sufficient individual capitalisation of the individual banks and provide a secure basis for the SREP decision (see also text no. 28 of the Discussion Paper). In our perception, the current ‘constrained bottom-up approach’ already fails to take sufficient account of the specificities of the banks’ business models. Up until now, this has always been justified with the additional conflicting macroprudential objectives. Hence, the current approach is something of a compromise that cannot do full justice to either of the two objectives. In this respect, we generally welcome the establishment of a clear hierarchy of objectives that clearly prioritises the microprudential aims.
We find it difficult to imagine that a supervisory leg that is largely based on top-down elements would offer sufficient detail for the main microprudential objective of the exercise. We are concerned that it would no longer be possible to determine the banks’ individual capitalisation with the due care, since the approach would lack sufficient detail and hence present an unreasonably distorted picture of the respective banks. For the same reasons we reject the use of such top-down elements in SREP decisions. The elements should be applied on a strictly limited basis and only where the respective institutions have no sufficiently robust internal models.
All in all, as regards the primary objective of the stress test as mentioned above, we do not consider the use of top-down elements to be an improvement of the status quo.
Ultimately, the approach referred to as ‘bank leg’ is a bogus claim since the institutions by no means have ‘ownership of results’, as claimed. On key points, they remain aligned with the EBA methodology and may at best disregard a few requirements.
In principle, it is conceivable to raise the compatibility of the tools and processes for internal stress tests and scenario analysis in the context of the normative internal ICAAP perspective and the EU-wide stress test in the long run. Under certain conditions, for instance the stability of the EU-wide stress test methodology and data requirements as well as sufficient, individual flexibility for banks, harmonisation could result in time and cost savings. Individual modules (e.g. for the simulation of impairments) are often employed for both purposes. The use of ICAAP for the bank leg in the proposed form would entail the disincentive that the ICAAP and hence the internal processes would have to be fully aligned with the EU-wide stress test methodology and an additional option to selectively switch a large number of constraints on and off would have to be created in order to allow further internal calculations. Ultimately, there is a danger that the further development of the stress test infrastructure is too closely aligned with the needs of EU-wide stress testing, and that internal requirements and flexibility receive insufficient consideration. Furthermore, an overly narrow alignment of the ICAAP with the EU-wide stress test would create further dependencies. Regular adjustments of the ICAAP to new methodological requirements under EU-wide stress testing cannot be the aim. Experience with EU-wide stress testing in recent years has shown that new issues always arise in the given methodology which are difficult to integrate into the ICAAP process in the short term.
We certainly appreciate the idea of a joint outcome by bank and supervisor - as is the case in the status quo. In the course of the dialogue, a proper appraisal of the bank specificities could take place and the supervisory P2G setting in the context of the SREP could have a demonstrable connection with public communication. However, this would require the above-mentioned dialogue to take place on equal footing. In the past, this has been a shortcoming in the case of banks that are subject to direct ECB/SSM supervision.
We are convinced that the specific conditions of the individual institution, including the application of national accounting standards, should play a key role in any supervisory assessment, and that this is the only way the P2G can be adequately set in the SREP. These specific conditions can only be adequately included if sweeping conservative constraints are dropped in general, or at least in the context of the P2G setting. Otherwise, certain institutions will be generally placed in a worse situation.
In the proposed bank and supervisory leg framework that we do not consider appropriate, as described in our answer to question 1, it would therefore be of importance that the results of an appropriately designed bank leg are adequately assessed by the supervisor and are included in the supervisory result. Hence, a recalibration of the P2G set in the supervisory leg should be investigated on the basis of the results from the bank leg and should not be not be ruled out from the outset.
In our opinion, expenses on the banking side in the current procedural model are primarily driven by the pre-exercise review of the methodology in regard to amendments and their implementation, data provision and quality assurance of the mapping onto the template structure, the implementation of the calculations and analysis of specific effects. In this context, the dialogue with the supervisor plays a minor role. However, if the supervisor questions the outcomes or allows them to be overwritten, this dialogue is necessary to ensure proper comprehension.
Under the proposed framework, banks would have to calculate the stress test multiple times for the bank leg in order to set up a reconciliation along the eliminated constraints (i.e. at least one stress test calculation per constraint). In addition, banks would have to process the bank leg results and the reconciliation effects for publication as required. Compared to the current approach, both of these factors are associated with significant additional effort and hence substantially higher costs. The reconciliation between the two legs thus quickly becomes more time-consuming than the calculation of the legs itself. The elimination of the dialogue with the supervisor in the QA context would not lead to any simplification for the banks, since the latter would still have to fully understand the results of the supervisory leg while being further removed from supervisory explanations than before.
Banks would have to provide several sets of data (for their own scenario calculations in the bank leg and as input for the supervisor’s calculations in the supervisory leg). In addition, banks would have to quantify, publish and explain to the market the methodology deviations caused by the use of their own standalone models (e.g. potentially in the context of press conferences, cf. text no. 75). Moreover, such explanation appears difficult since the banks would have no insight into the supervisory leg, and the supervisor’s approach (challenger/benchmark model) would become less transparent for the bank (cf. text no. 73). It is unlikely that the cost savings assumed in the new framework, which are essentially to be achieved through reduced interaction between banks and supervisor, would be sufficient to compensate this increase.
Hence, it seems that only the supervisor’s time and effort will be reduced: lower granularity of the publication of the outcomes for the individual banks and reduced dialogue with the banks in the context of the QA activities.
However, in our opinion, precisely this dialogue with the supervisor in the QA context is a crucial element in order to arrive at an appropriate outcome within the framework of the subsequent microprudential assessment of the capital adequacy and resilience of the institutions.
Why have banks so far failed to avail themselves of this opportunity to additionally publish their own risk assessments?
This is certainly due to the major challenge associated with explaining the different results arising from their own and the supervisor’s approaches and to present the underlying, often complex facts to the different recipients in an intelligible form. There is a risk that the recipients may be uncertain which of the results is “more correct”. However, there is virtually no satisfactory answer to this obvious question posed by external third parties, even if a future concept contains parallel supervisory and bank legs. Ultimately, this would further reduce the relevance of the exercise to both market participants and investors.
The EBA proposal also entails further costs consisting of the massive disincentive posed by an alignment of the internal models with the supervisory stress models, which the EBA intentionally puts up with in view of the prevailing cost pressure. The resilience of the European banks is likely to suffer as a consequence.
In principle, it should be possible to drop all constraints.
In our view, the following constraints are currently the main contributors to an unwarranted distortion of stress test results:
- The static balance sheet approach (even in the envisaged revised form)
- Pass-through constraints applying to the passing on of interest rate changes (NII calculation, text nos. 381, 384, 355, 359 of the methodological note)
- Caps on the recognisable interest income (NII calculation, text no. 362 of the methodological note)
- No accounting for possible recovery from defaulted engagements (credit risk, text no. 34, text no. 57, text no. 143 of the methodological note)
- Application of the haircut approach for items held with a trading intent and their related economic hedges (market risk, text no. 259 of the methodological note)
We are in favour of the first alternative (“banks have a discretion to relax certain methodological constraints”), which nevertheless permits a ‘slight’ individualisation of the procedure. At least, this allows for some inclusion of bank specificities.
In this context, comparison can only be made with the ICAAP stress test under the normative perspective, since this perspective includes the regulatory risks and relevant parameters. There would be no synchronisation even if all constraints were dropped. Aside from the issue of different models, there are major differences in the data landscape and the application of accounting principles, especially where P&L items are concerned.
Nevertheless, the most likely common ground could probably be found for the simulation of losses in market value. However, in many cases, losses in market value are not recognised as a shock in the first year of the scenario case. Instead, they are also simulated for the following years in accordance with the assumed market movements on the basis of the position at the starting point and are also recognised there. A plethora of individual internal bank methods is available for all other risks (esp. net commission income, net interest income, non-interest income P&L) that differ fundamentally from the methodology of the EU-wide stress test.
We believe that the availability of supervisory expectations and benchmark parameters (e.g. for PDs) is useful for the implementation of stress tests, even if we can resort to our own models. If these were available at a sufficiently early stage of the process (ideally directly in conjunction with the publication of the scenario definition), we would be in a better position to question our own models and prepare for a substantive discussion with the supervisor at equal footing. However, such ‘assumptions’ must not be binding requirements.
The main benefit of the bank leg (compared to the supervisory leg), if properly structured, would consist of the institutions’ ability to reflect developments in the scenarios from their own perspective and take account of their respective individual characteristics, e.g. with regard to markets, products and customers. The relevance and comparability criteria (at a purely economic level, not at the granular field level) would benefit immensely. However, the benefits associated with the bank leg turn into drawbacks if ownership of outcomes is merely a claim, while the results actual fail to represent the perspective of the banks.
In our opinion, the specific risk profiles of the individual banks have so far been addressed exclusively in the context of quality assurance. It is not in the interest of the participating banks if, due to reduced QA activities, the supervisor is even less concerned with the institutions' views than before and hence fails to ‘accept’ these views in any way.
Although we are certainly prepared to take responsibility for the outcomes we produce, this should under no circumstances reduce the dialogue between supervisor and bank. Nevertheless, we reiterate that the results of the bank leg do not generally reflect the bank's own view.
On the contrary: Aside from a more technical level (STAR portal, RAG flags, troubleshooting in templates, elimination of comprehension issues), interaction between bank and supervisor should place greater focus on the substantive-technical dialogue between the supervisor and the institute.
In order to arrive at a result that is more realistic than the outcome achieved in the current EU-wide stress test, the static balance sheet assumption would have to be dropped. As a minimum, envisaged portfolio expansion or expiry should be taken into account.
Furthermore, in the event of an adverse scenario, banks will generally not continue operations without modifications and will specifically adjust their refinancing mix, reduce their new business to some extent, inter alia by imposing conditions appropriate to the situation (e.g. re-margining of counterparties) and also adjust their market risk position.
The envisaged market discipline will depend to a lesser extent on detailed templates, but rather on the relevance and comparability of the results. We therefore suggest a significant reduction of the publication granularity. In particular, we request a streamlining of the TRA_CR_IRB and TRA_CR_STA templates, see also question 17.
In continuation of text no. 51, the banks would ideally be in a position to decide themselves whether and to which extent they wish to publish the results of the bank leg.
The publication of two different outcomes from the bank leg and the supervisory leg may cause confusion both for the markets and for the broader public. These stakeholders will not have the resources necessary to critically evaluate any reconciliation between the two legs. Furthermore, the question arises how the bank should prepare a reconciliation of the result from the bank leg to the supervisory leg without having recourse to the detailed outcomes from the supervisory leg. A mere switching on and off of methodological relaxations, as suggested in text no. 59, will lead to large numbers of superfluous parallel calculations. In text no. 77, the EBA itself cites numerous reasons that speak against the parallel publication of the results produced in the two legs.
In our opinion, past experience with EBA stress testing has shown that the granular details are not of relevant interest to the public and do not produce any added value. Irrespective of our fundamental criticism of the approach publicating of two different results, we would therefore only disclose a few key figures to the public and reject granular disclosure. On top of this, granular disclosure in the baseline scenario would presumably be very close to the planning.
We refer to our response to questions 15 und 17. In our opinion, past experience with EBA stress testing has shown that the granular details are not of relevant interest to the public and do not produce any added value. We therefore believe that limitation of the public disclosure to key figures is appropriate.
However, the disclosure of supervisory calculation results and calculation methods to the respective banks, especially the challenger models employed by the supervisor, must be viewed as a separate issue. If this information is not disclosed to the public, it must at least be available to banks in a quality-assured form to facilitate its use in the presentation of reconciliation effects.
In addition, any top-down adjustments compared to the published methodology (and any challenger models used for this purpose) would also have to be disclosed for the supervisory leg in a transparent manner. Otherwise, this would create a supervisory black box which does not allow for a critical examination of the projections. The supervisors would elude any assessment of their model methodology.
From our point of view, it is doubtful that high data granularity is required in the transparency templates in order to gain the required insights. In our opinion, the detailed statement of the impairment effects (stages 1, 2, 3) and the breakdown into FIRB, AIRB and STA represents an overload of information and should no longer be prepared in the future.
Very few banks manage operations according to COREP asset classes. A natural discrepancy therefore arises between the information published in annual reports and the current disclosure in the updated credit risk transparency templates. The elimination of this outline according to COREP asset classes in the transparency templates could therefore result in clearer messages and greater relevance to markets and investors.
We believe that, although market participants appreciate transparency, they also appreciate clear messages. In principle, a clear result is therefore preferable, i.e. only one final stress test result. However, this would only make sense if the single result actually reflects the current earnings and risk profile of the respective bank and represents the outcome of a close dialogue between the bank and the supervisor (see question 4).
If two results were published, a granular explanation of the cause of the discrepancy would be necessary. The only way to ensure that no misinterpretations arise is to provide additional qualitative and quantitative information. However, this would lead to considerable additional effort on the part of the banks, which does not appear justified from a current perspective.
We would be prepared to take responsibility for the results of a bank leg of which we have methodological ownership and which actually reflects our internal perspective. Moreover, the bank leg should also be highly relevant to the official result of the stress test and affect the P2G setting by the supervisor. Under these framework conditions, we principally consider a control mechanism that ensures a common starting point to be sufficient.
Such an explanation would be highly misleading for market participants. It would be presented as a reconciliation between supervisory and bank results when, in fact, it would be a reconciliation with little added value between more or less supervisory-driven results which are imposed on the banks solely for cost reasons.
We furthermore consider the respective reconciliation calculation as extremely time-consuming, cost-intensive and, otherwise, as an approximation at best. A further challenge consists of the conveyance of the increasingly complex message to market participants, who may not wish to concern themselves with numbers of reconciliation effects in the mid double digits and the appropriateness of such effects. As explained above, confusion of the recipients and misinterpretation of the results cannot be ruled out if identical scenarios and identical starting points lead to significantly different results, depending on the respective risk assessment.
However, the scope of the respective effects also depends on the order of elimination of the constraints. This may have to be taken into account in the comparison of effects between the individual banks.
The current EBA proposal is a reconciliation with little added value between more or less supervisory-driven results which are imposed on the banks solely for cost reasons. Apart from this, however, we believe that a top-level reconciliation of the effects of the individual constraints on the capital ratio (e.g. CET1) would make sense. Based on the bank leg results, the effect on the capital ratios would be shown and explained in technical terms for each applied constraint.
Among the three alternatives presented, we consider the bucket approach under alternative (ii) and alternative (iii) to be the least suitable, since bucket margins always lead to a distorted perception in alternative (ii) and alternative (iii) does not offer any transparency regarding the actual stress effects.
In our view, it is important that the supervisor's derivation of the P2G is fully transparent to the institutions in order to allow for better integration of this capital requirement into our bank management.
We reject publication of the P2G since the latter exclusively reflects supervisory expectations. In the context of the banking package review, a deliberate decision was taken not to impose any requirements in this regard. See recital no. 16, CRD V. Publication of the stress test results should suffice.
Publication of the P2G makes any underrun of the respective target capitalisation fully transparent and present for the market. In the public perception, the recommendation character would be replaced with a de facto obligation. This could boost procyclical effects in crisis situations, as we are seeing in the context of the COVID-19 pandemic. If the P2G is not reached, the public would already know before the breach of the MDA threshold that there could be a problem, which would make it more difficult for the bank to respond.
An overly specific disclosure of possible measures in the event of an adverse scenario would limit the institution's scope for action and adversely affect the costs of certain measures, or even rule them out. The same reasons apply to the non-disclosure of recovery & resolution plans and banks’ internal liquidity and capital emergency plans.
No. The bank and supervisory leg framework in the proposed basic alternative leads to a diluted message (two different results) which reduces the relevance of the exercise. Transparency and comparability are still not given, or are further complicated by the introduction of a bogus bank leg. Moreover, with the bank not involved in the determination of the supervisory leg, the quality of appropriate P2G setting in the SREP suffers. Overall, this will lead to a significant extra effort on the part of the banks, without any immediate discernible added value. We are therefore extremely sceptical about the proposed framework, not only as a whole, but also from a cost-efficiency perspective.
All in all, we believe that the alternative option of having a single outcome as described in box 1 (dialogue/synthesis/convergence) – similar to the status quo – is preferable, depending on the specific model design. A uniform result sends a clear message to the market and increases relevance. There is room for improvement where the transparency of the P2G setting is concerned.
Bank leg Supervisory leg Overall exercise
Relevance Better Worse Worse
Comparability Worse Worse Worse
Transparency Neutral Worse Worse
Cost-efficiency Much worse Neutral Much worse
Relevance: Although, in the proposed basic model, the bank leg (involving common methodology or a certain number of dropped constraints) is marginally more relevant to the banks’ internal control, the supervisory leg result is, however, much less transparent to the bank, resulting in an overall decrease in relevance. On top of this, investors and the public are left confused by the different results. All in all, relevance declines significantly.
Comparability: The supervisory leg still fails to ensure that the specificities of the business model are included in the result figures, leading to systematic discrimination against certain business models. By comparison, the proposed bank leg achieves only marginal improvements. Although the results are standardised, they are not comparable in the sense of a (fair) level playing field. All in all, the exercise delivers less comparable information.
Transparency: There is merely an apparent increase in transparency for the public when it comes to the banks' view of the respective scenario and the results derived from it, while transparency of the supervisory leg results, which is the actual relevant result for the SREP, actually declines. Altogether, the generated transparency will be lower than the current level. Real transparency of relevant results is still not achieved.
Cost-efficiency: For the banks, the bank leg involves a significant additional effort. However, the bank leg is not included in the derivation of the P2G. The benefit associated with the exercise declines from the banks’ perspective. All in all, this signifies a decline in cost-efficiency. The supervisor benefits from savings arising from the reduction of its quality assurance activities, while the information needed to establish a P2G will still be obtained. Consequently, improved cost-efficiency only arises from the supervisory perspective. The banks will also have to bear the main calculation expense in the supervisory leg. Since the supervisor intends to reduce its QA activities, it will generally become more difficult for the banks to understand the supervisory mark-ups and the P2G that is set on this basis. Hence, the banks will also face higher expenses and lower cost-efficiency in the supervisory leg.
From the banks' point of view, further scenarios would not be feasible, even in the current setup. The outlined new setup would further aggravate this situation.
In our view, the two scenarios of different severity would significantly increase the efforts required to calculate and check the plausibility of the results. Yet the gain in knowledge would appear rather small. Altogether, cost-efficiency would suffer substantially.
One might assume that two scenarios are generally better suited to identify the different vulnerabilities of the business models of the respective banks. However, two scenarios are unlikely to be enough to map all of the relevant individual vulnerabilities of various business models. In this context, additional scenarios translate directly into additional effort – both for the banks and the supervisor. However, additional insights would only be gained by banks whose vulnerabilities differ from the majority of institutions.
This also raises the question whether there should be a supervisory leg and a bank leg for both adverse scenarios. This would be consistent with the logic of the proposal, but would once again lead to substantially higher expenses for the banks.
We believe that it would be more efficient to choose a universal scenario in the context of a standardised exercise, such as the EU-wide stress test, which addresses different vulnerabilities (perhaps over time). In its definition of the P2G for the remaining ‘outliers’, the supervisor should also include the effects from the banks’ internal ICAAP scenarios which reflect the specificities of the respective business models.
As regards market price risks, the CSV_MR_FULL_REVAL template already records simple sensitivities for affected banks in the context of the current exercise. In this respect there is a broad understanding of the meaning of these sensitivities.
However, in regard of the rest of the exercise, it is not so easy to define exactly what the sensitivities signify. On top of this, the sensitivity results would be difficult to interpret: What does GDP sensitivity mean? Does it include repercussions on the market risk factors? Should further second-round effects be considered? How does a change in GDP translate into a change in new business and margins beyond the static balance sheet assumption?
In addition, too many non-linearities reduce the significance of the sensitivities. Investors basing their assessments on this information could produce misleading results and cause confusion rather than additional transparency. To cite two examples: The effects connected with the transition of impairments from stage 1 to stage 2 cannot be adequately reflected via sensitivities, and even a product as simple as a floating rate note causes unexpected non-linearities when interest rates are low.
In any event, the associated methodological and operational costs would be very high, while the gain in knowledge and the contribution to transparency and relevance would be minor due to limited interpretability. We would therefore categorically reject any supplementing of the stress test with additional sensitivity analyses.
The implementation of scenarios which can take effect on time scales as disparate as days or more than 3 years places very high demands on the methodology and flexibility of the infrastructure.
The specificities of the individual banks become more significant over long time horizons. In our opinion, it is mostly impossible to quantify an institution’s ability to adapt its business model and the corporate organisation, for instance to changing markets, new technologies, etc. In this case, general guidelines by the supervisor would at best lead to systemic statements, but not to valid statements for the individual banks. Against this background, we believe that this would rule out the disclosure of results for individual institutions. In addition, to the best of our knowledge, the previous scenarios were generally derived from business cycle theory, which generally assumes cyclical trends. Seen from this perspective, after three years for instance, it is not only possible that the problems have subsided, but that a positive countermovement is already following on from the economic stress. The ‘stress’ would thus revert into a positive scenario.
Since the statements on ‘climate risks’, for instance, are still rudimentary and are obviously in the early stages of consideration, a conclusive assessment of the approach appears close to impossible.