We appreciate that the EBA thinks about the future of the EBA stress test and also asks the opinion of the banks.
In the new framework, there would be two results for the same adverse scenario, one from the bank leg (under bank responsibility) and one from the supervisory leg (sole responsibility of the supervisory authority)(see paragraph 54). The market participants would thus have two answers to the question of how resilient the Bank is in the given scenario. Even if the respective results are used separately (bank leg: transparency and market discipline by detailed publication for the market, supervisory leg: in the sense of P2G) and would differ in the level of detail of the publication, the question arises as to how much information the result of the labor-intensive bank leg would provide. To evaluate this result and differentiation from the supervisory result, market participants would have to view further information in addition to the granular publication of the banks. From the banks' point of view, it is questionable whether the transparency and relevance of the results would actually rise.
Furthermore, the proposed framework in general seems to increase the costs and decrease efficiency. Compared to the current methodology of the EBA stress test, not one methodology has to be calculated, but two methodologies: the supervisory leg and the bank leg. Consequently, for every risk type (e.g. credit risk, NII) two models need to be developed with different assumptions. These models have to be checked and kept up-to-date. For both methodologies, the results have to be generated and analyzed. Subsequently, two templates have to be filled in. Finally, the differences have to be analysed and explained which can be a difficult process. All these processes are difficult to automate as long as the EBA stress test keeps changing. Additionally, the ICAAP stress tests are performed around the same period increasing the workload even more. Hence, it is preferred to have a more constant EBA stress test such that processes can be automated and the efficiency increases and costs decrease.
Should the EBA adopt this new stress test framework, it would be important for the new stress test to be based on a clearly defined data set and requirements in advance. For this reason, we believe that early consultation, coordination and publication of the data and templates to be supplied by the banks in 2022 for the "supervisory leg" would be essential.
This element in the proposed framework aligns most with the current set-up of the EBA stress test. This supervisory leg is good for comparability between banks, but does not always reflect the business models of the separate banks. Some elements of the EBA stress test can be converted to a top-down model, since the information is already available to the supervisor, e.g. the information in the FINREP and COREP. When an entire top-down model is implemented, the understandability of the results by the bank may be a challenge. Banks do not have sufficient insight in the used models for full interpretation of the results. Consequently, banks cannot explain the results to the public (e.g. rating agencies) which might decrease trust in banks.
The bank leg seems to be a second supervisory leg, because it still makes use of the same methodology, scenario, and templates. If ICAAP is the basis for the bank leg, our preference would be to calculate the impact of the EBA stress test scenarios with the ICAAP models without constraints of the EBA stress test. We would like to calculate the result which would be visible in our balance sheet and ratios when the scenario really occurs. In case, our ICAAP models need to be changed to the EBA constraints and methodology, ICAAP is not suitable as a bank leg.
We think it is preferable to only publish one single CET1 capital depletion compared to multiple ones per different leg and scenario. Publishing only one single CET1 capital depletion reduces complexity to the public and makes communication of the results more clear and make the results easier to understand.
Since the EBA would like to determine P2G in the future based on the results of the supervisory leg, it is expected that this will be even more standardized and benchmark-based than the previous stress test. In this respect, special features of the institution, such as the application of national accounting rules (e.g. HGB), could suffer greatly from this and could lead to situations where the outcome does reflect even less the reality of the bank. Against this background, we believe that the "bank leg" is particularly important as a counterweight in the future, so that the P2G determined in the "supervisory leg" by the results of the "bank leg" should,in our view, also be part of the stress test and not be excluded from the outset.
When the bank leg is calculated based on the same methodology, templates, and scenarios as the supervisory leg, we think the results should be taken into account in the final supervisory outcome. It is still part of the supervisory stress test and not based on own internal bank specific scenarios and templates in which bank specific results can be taken into account. However, one supervisory outcome is preferred over two separate outcomes (one from the bank leg and one from the supervisory leg).
The proposed changes will cause significant additional work for banks. The banks would have to provide two sets of data (for their own calculation of the scenario in the bank leg and as input for the supervisory calculations in the supervisory leg). In addition, banks would have to quantify methodological deviations by using their own models on a stand-alone basis, and explain it to the market (e.g. generally at press conferences) (see also 75). Moreover, this explanation seems difficult, as the banks have no insight into the supervisory law and the supervisory procedure (challenger/benchmark models) would thus become less transparent for the banks (see paragraph 73).
The points mentioned above would provide banks with a better overview than the previous stress test- framework will result in significantly higher costs. From a banking perspective, it is questionable whether the cost savings possible in the new framework - mainly through the reduction of interaction between banks and supervisors - would be sufficient to drive this increase to compensate.
The advantage would be to draw the attention of the supervisory authority to certain institution- or market-specific features that are lost in the supervisory law. However, this would only be an advantage if it could also be discussed in the context of a dialogue and thus lead to a better common understanding.
The costs will significantly rise, because of:
- Implementing two separate models: one for the bank leg, one for the supervisory leg with different assumptions
- Calculating the results twice which takes time
- Analysing the results twice
- Filling in the EBA template twice
- Explaining the differences in the results can be time/cost consuming
- Relaxed constraints
- No sign off for supervisory leg
- Lighter QA interaction
Mitigate costs and increase benefits:
- It would help if the methodology and templates are stable. Then the process can be automated more.
see document attached
If the bank leg would have a common methodology, it would be closer to a supervisory stress test than to a bank stress test. Although the comparability will be higher than without a common methodology for the bank leg.
In general, we think that the addition of the bank leg is still not reflecting the bank stress, because the same templates and scenarios are used. For this reason, we think it is not useful to add a bank leg.
ICAAP is very different from the supervisory leg, e.g. static balance sheet assumption (e.g. anticipation on capital impact of future regulations and new business), assumptions on margins and reference rate, and ICAAP includes a more economic approach. Furthermore, the ICAAP contains very bank-specific risks and events. In some cases, the models may be more realistic In some cases, the models may be more realistic, but they may also produce completely different results (e.g. FIRB institutions with separate supervisory and economic parameters). In the case of promotional banks these risks and events are even more specific and thus the gap between the two approaches bigger. Therefore, ICAAP can only be a supplement.
If there would be general guidance and use of own models without constraints, the comparability between banks would be low.
The constraints which do not represent the reality for a bank can be relaxed. The purpose of the bank leg can be increased by making use of the internal stress models with internal assumptions which are then used to calculate the EBA scenario. The bank should have the possibility to represents the results in a way which represents reality (e.g. including Cost of Hedging and Own Credit Adjustment).
That is to be welcomed. However, the process has also been intensive in the past because there were sometimes errors in the templates or difficulties in understanding them. The unnecessarily complex presentation of partial results in the templates leads to more QA.
The supervisory leg adds value to the comparability of banks in times of stress. However, the bank leg does not add much value to the supervisory leg, since it is still a constrained approach which does not necessarily reflect the real banks situation in times of stress. The internal models of the banks reflect this reality best, but make comparison difficult.
We that the disclosure of information on both the supervisory leg and bank leg will lead to confusion of the market. The differences between the two need to be explained and banks can damage their reputation when the bank leg results in better results for the banks than the supervisory leg. This may very well happen, since less constraints are used in the bank leg compared to the supervisory leg. Due to the very specific nature of promotional banks, the outcomes might be even further apart than between 'regular' banks.
If only the results of bank leg are going to be disclosed, then the market might still try to compare the results of the banks to each other. They would like to know how the differences arise and hence banks have to provide a lot of details on the methodology. This might not be understood by the market because of complexity. This might result in an stress test which is not easy to understand for the market.
If the question relates to disclosing only the supervisory leg, then this is a good option if comparability is important. Only publishing the supervisory leg does not lead to confusion between different methodologies.
If the question is related to disclosing less information on the supervisory leg, then we are indifferent.
Make more clear where impact comes from in capital ratio and leverage ratio: credit risk, market risk, NII, etc. and how it has impact: OCI, P&L. Some information can be summarized in a graph to see the development over the years better. Maybe work with conditional formatting with cells which have a high impact.
We thinks the transparency on exposures and credit risk impairment is very detailed. A potential drawback is that the information is not easy to read and understand.
We think that the disclosure of information on both the supervisory leg and bank leg will lead to confusion of the market. The differences between the two need to be explained and banks can get reputational damage when the bank leg results in better results for the banks than the supervisory leg. Additionally, the differences between the two legs could be larger for promotional banks due to their highly specific business models.
Disclosure of the bank leg is not preferred in order to avoid confusion caused by differences between the bank leg and the supervisory leg.
The challenges of explaining the differences between the bank leg and the supervisory leg would be to make the differences understandable to the market which may not be familiar with the details of the EBA stress test. The detailed assumptions are different for the supervisory leg and the bank leg. These assumptions might not be understandable to the market. Explaining why these assumptions are changed by the bank is a challenge.
If the standalone impact needs to be calculated for all constraints which are changed, the cost-efficiency of the stress test would decrease significantly. The models have to be amended step by step and the results have to be generated constraint by constraint.
We think that the disclosure of information on both the supervisory leg and bank leg will lead to confusion of the market. The differences between the two need to be explained and banks can suffer reputational damage when the bank leg results in better results for the banks than the supervisory leg. This may very well happen, since less constraints are used in the bank leg compared to the supervisory leg.
If the final P2G would be disclosed, the market will be looking for an explanation of different P2G levels across banks. Therefore, the way how the P2G is calculated should be disclosed as well. Furthermore, the impact of the stress test for many promotional banks is higher in terms of capital ratio decrease in %-points compared to commercial banks because of the higher starting capital ratios. This results in a relative high P2G which could be wrongly interpreted by market participants if a comprehensive explanation of the methodology is missing.
See question 23.
If the management actions are disclosed, the market might know when an institution is in trouble by checking when specific management actions are undertaken. This may damage the reputation of the instition and lead to a reaction in the market. Furthermore, communication of these actions might undermine their effectiveness.
See table of question 27.
see attached document
Relevance will increase, because a broader sensitivy of the bank is calculated. However, the cost-efficiency and transparancy will decrease. Considering the proposed framework, all scenarios have to be calculated for both the bank leg and the supervisory leg. This results in more calculation and analysis time which decreases cost-efficiency. Moreover, transparancy will decrease, because it is difficult to explain what the results and the differences between these results mean to the public.
We think this is a valuable addition. First of all, the impact of the shock can be explained and second the impact of other scenarios can be approximated, EBA has to keep in mind that the template should not become too detailed (e.g. also asking more granular data for other greeks), because then the template becomes difficult to read and understand and cost-efficiency decreases.
We think this is a valuable addition for seperate stress tests, but not as an addition to the EBA stress test unless the scenario provided in the EBA stress test reflects for example an environmental scenario.