Response to discussion Paper on the future of the IRB Approach
Go back
• Jurisdiction is currently labeled as the main and overarching driver for unintended differences in RWA. We are of the view that there are valid reasons why risk weights of, for example, the mortgage portfolios, are quite different per jurisdiction because of many underlying differences (e.g. social welfare, tax regimes, payment behavior, insolvency laws & bankruptcy laws, etc.). These differences are translated in differences in actual risks (and actual losses), which justifies these differences in risk weights. Alignment in back-testing methodologies and the evaluation process of back-testing results are possible answers to regain trust.
• At the moment the guidelines are in particular sparse on treatment of defaulted assets. It is recommended to focus on this topic as soon as the basic definitions have been established.
• In our view, the area that is one of the main driver of RWA variance between banks is the definition of default (DoD). The DoD also indirectly affects PD, LGD and EAD. Within the DoD default are numerous sub-areas such as, but not limited to: days past due; materiality threshold; and cure rates.
• In addition, you could consider to prioritise the RTS and GL on the economic downturn for the purpose of LGD modelling an area with a lot to gain in terms of strengthening the IRB framework and reduce the divergence in bank’s modelling practises.
• After the changes and additions to the modelling rules are final - we propose that each financial institution will draft its roll-out plan to be aligned with the regulator. It is recommended the largest models and the models representing portfolios in which the largest unintended variation is seen are redeveloped in a reasonable timeframe. In this way, we can show good progress towards the stakeholders and can re-gain the crucial trust in a timely fashion. On the other hand, for the models that cover smaller portfolios, a longer time path can be agreed upon.
• From the proposed timeframe (figure 2) it can be learned that the implementation of the revised IRB approach including supervisory approval is foreseen at the end of 2018.
We have concerns on the short timeline for several EBA deliverables, a substantial (if not all) current models have to be redeveloped, availability of historical data as well as capacity from a supervisory point of view. Therefore we propose roll-out plans (window 5 to 7 years) where the largest models are tackled first and that the industry starts with a exposure class (e.g. residential mortgages) and follow a pre-agreed sequencing.
• For banks that currently have no LGD in-default models, 2 years is too short. The development of these models require extensive modelling work. 2 years to develop such models from scratch based on a new concept, while also monitoring, updating and redeveloping existing models is quite ambitious.
• In any case, Europe should fiercely strive not to allow any changes on top of these EBA triggered changes, based on new BCBS guidelines.
• Defaults are considered at customer or facility level, and how to handle cross defaults (for Retail clients this can differ per jurisdiction)/ Multiple defaults/ Overdraft facilities.
• Further guidance on the changeover from non default to default status and (if applicable) a relation to forbearance (re)classification(s).
• Guidance on Unlikely to Pay criteria (e.g. fraud, other qualitative criteria).
• When calculating a default, a 3 months period would be a good operational translation of the 90 days criterion.
• How institutions should deal with the concepts of ‘interest holidays’ and ‘redemption pause’.
• There might be requirements to adjust historical data. How to handle situations in which some banks can make the required adjustment and some banks cannot. This could lead to an unlevel playing field. In such circumstances we propose to take into account a phase-out approach. The new methodologies might be based on historical (unadjusted) data, enriched with new data (every year new data is added), based on the new RTS and new GL requirements.
• In general the change will by definition be a material change according to regulatory articles. Additionally, the methodologies may change materially as well if the cure definition and multiple default interpretation is substantially different from the current interpretations.
• We need to ensure that these important and impactful changes will not end in vain due to the overshadowing BCBS proposals (revision of the SA framework for credit risk and more importantly the Capital Floors consultation).
• Finally we feel that there is a need for detailed guidance. As examples may serve Counting of days past/overdue, Downturn LGD and level of conservatism (versus data quantity, quality and modelling correctness), including how and where to apply levels of conservatism.
• (in combination with downturn), how to apply the levels of conservatism (aka MoC / margin of conservatism): for example per model step (multiple layers), or just one overall. When is the data (quality and quantity) of sufficient level and when is the modelling of sufficient level to have a minimum level of conservatism. How to address levels of conservatism between PD and LGD, etc.
• LGD in-default: aspects such as over which period or confidence interval this needs to be calculated; should it be the best guess (making this a point-in-time estimate) or an average of different default scenarios?
• Also execution of the roll-out plans and the “end-state” percentage of the portfolio reported on IRB, should be aligned and national discretions be removed. Hence, the supervisor should ensure that the roll-out plan is actually adhered to.
• (copy of answer to Q2:) We think that there are valid reasons why risk weights, of for example the mortgage portfolios, are quite different per jurisdiction because of many underlying differences (e.g. social welfare, tax regimes, payment behavior, insolvency laws & bankruptcy laws). These differences are translated in differences in actual risks, which justifies these differences in risk weights. Alignment in back-testing methodologies and back-testing results are possible answers to regain trust.
• We would favour clear rules for when the data quality and (more important in this respect) the data quantity is sufficient. In order words, when is the data too limited to build an internal model.
• How should institutions deal with expert models. Guidelines with respect to levels of conservatism should be clear as well.
• Rules on how to pool data are welcome too.
• Finally, it might make sense to come up with more risk sensitive FIRB LGD models, looking into the potential differences between markets (jurisdiction) and into the differences between exposure classes.
• We appreciate the desire for a best practise assessment in which regulators take stock and learn from lessoned learned in other jurisdictions.
• However, given the current SA revision proposals it should be noted that it might be extremely challenging to align the SA exposure classes with the IRB exposure classes. We see extra challenges for non-banks financial institutions.
• However, it could be valuable to provide guidance on PIT methodologies in the light of IFRS 9 and expected loss calculations.
• We understand that reducing the unintended variances will require a substantial effort for all IRB banks. In the Netherlands the IRB portion is very high, therefore the efforts for Dutch banks is very substantial. As we strongly believe that IRB is the superior method we wish to put this effort into this process. We need to ensure that the BCBS proposals do not make our efforts in vain (by flooring the IRB risk weights to the SA risk weights via the Capital Floor).
• We would welcome a strong link between EBA and BCBS, so European institutions are not required to process impactful changes twice.
• We favor a decent roll-out plan, with a pre-determined sequencing (largest models first). This way, during the process, EBA can demonstrate that indeed the EBA initiative reduced variance (for the first exposure class that was tackled, which will give a positive signal to the market). All stakeholders would know and understand which models will be updated based on the new disclosed methodology, this makes the explanation easier. On the one hand, the first success can be celebrated even earlier, while on the other hand we can take more time for those exposure classes that are scheduled in the end (beyond 2019).
• Please take notice of the IIF RWA Task Force (IRTF).
• Requirements that should be adhered to in order to report portfolios on IRB, such as the Use Test, should be supervised in the same fashion across Europe without national discretions, creating a level playing field.
2. What would you consider the areas of priorities?
• In general the proposed priority setting seems fine. The fundamental components, such as the definition of default, PD, LGD and conversion factors should established first, before subsequently the other aspects can be worked out in detail.• Jurisdiction is currently labeled as the main and overarching driver for unintended differences in RWA. We are of the view that there are valid reasons why risk weights of, for example, the mortgage portfolios, are quite different per jurisdiction because of many underlying differences (e.g. social welfare, tax regimes, payment behavior, insolvency laws & bankruptcy laws, etc.). These differences are translated in differences in actual risks (and actual losses), which justifies these differences in risk weights. Alignment in back-testing methodologies and the evaluation process of back-testing results are possible answers to regain trust.
• At the moment the guidelines are in particular sparse on treatment of defaulted assets. It is recommended to focus on this topic as soon as the basic definitions have been established.
• In our view, the area that is one of the main driver of RWA variance between banks is the definition of default (DoD). The DoD also indirectly affects PD, LGD and EAD. Within the DoD default are numerous sub-areas such as, but not limited to: days past due; materiality threshold; and cure rates.
• In addition, you could consider to prioritise the RTS and GL on the economic downturn for the purpose of LGD modelling an area with a lot to gain in terms of strengthening the IRB framework and reduce the divergence in bank’s modelling practises.
• After the changes and additions to the modelling rules are final - we propose that each financial institution will draft its roll-out plan to be aligned with the regulator. It is recommended the largest models and the models representing portfolios in which the largest unintended variation is seen are redeveloped in a reasonable timeframe. In this way, we can show good progress towards the stakeholders and can re-gain the crucial trust in a timely fashion. On the other hand, for the models that cover smaller portfolios, a longer time path can be agreed upon.
3. Do you consider the proposed timeframe reasonable? In particular do you consider reasonable the proposed timeline for the implementation of the changes in the area of: a. definition of default; b. LGD and conversion factor estimation; c. PD estimation; d. treatment of defaulted assets; e. CRM?
• First of all, it is difficult to provide detailed feedback without studying the RTS and GL (EBA indicates that on several topics work has not started yet). Nevertheless, we wish to make the following points:• From the proposed timeframe (figure 2) it can be learned that the implementation of the revised IRB approach including supervisory approval is foreseen at the end of 2018.
We have concerns on the short timeline for several EBA deliverables, a substantial (if not all) current models have to be redeveloped, availability of historical data as well as capacity from a supervisory point of view. Therefore we propose roll-out plans (window 5 to 7 years) where the largest models are tackled first and that the industry starts with a exposure class (e.g. residential mortgages) and follow a pre-agreed sequencing.
• For banks that currently have no LGD in-default models, 2 years is too short. The development of these models require extensive modelling work. 2 years to develop such models from scratch based on a new concept, while also monitoring, updating and redeveloping existing models is quite ambitious.
• In any case, Europe should fiercely strive not to allow any changes on top of these EBA triggered changes, based on new BCBS guidelines.
4. Are there any other aspects related with the application of the definition of default that should be clarified in the GL?
Some clear examples on the application of new default definition will be helpful. Especially, whether:• Defaults are considered at customer or facility level, and how to handle cross defaults (for Retail clients this can differ per jurisdiction)/ Multiple defaults/ Overdraft facilities.
• Further guidance on the changeover from non default to default status and (if applicable) a relation to forbearance (re)classification(s).
• Guidance on Unlikely to Pay criteria (e.g. fraud, other qualitative criteria).
• When calculating a default, a 3 months period would be a good operational translation of the 90 days criterion.
• How institutions should deal with the concepts of ‘interest holidays’ and ‘redemption pause’.
5. Do you have experience with adjustments of historical data? What are the methods that you used to adjust historical data, including both internal and external data?
• Although many institutions have experience with adjustments of historical data to some degree, the proposed strengthening of IRB will require some fundamentally differences to data collection and data processing, which will have a major impact. Therefore we propose to take into account the importance of the changes (estimated impact on RW or harmonization) versus the effort it will require for the institutions to process this change.• There might be requirements to adjust historical data. How to handle situations in which some banks can make the required adjustment and some banks cannot. This could lead to an unlevel playing field. In such circumstances we propose to take into account a phase-out approach. The new methodologies might be based on historical (unadjusted) data, enriched with new data (every year new data is added), based on the new RTS and new GL requirements.
6. To what extent is it possible to adjust your historical data to the proposed concept of materiality threshold for the purpose of calibration of risk estimates?
• Please see answers to question 57. What is the expected materiality of the changes in your IRB models that will result from the proposed clarifications as described in section 4.3.2?
• Obviously, when the objective is reducing the variances in practises between banks, it is expected that banks will be required to adjust their practises substantially to accommodate some changes while less effort is needed to accommodate other changes. We feel it is important that there is a good collaboration between the banking industry, regulator and EBA in order to arrive at changes that reduces the unintended variances and will strengthen the IRB framework, without creating too much changes that lead to unneeded RW fluctuations.• In general the change will by definition be a material change according to regulatory articles. Additionally, the methodologies may change materially as well if the cure definition and multiple default interpretation is substantially different from the current interpretations.
8. Do you consider the direction of the proposed changes adequate to address the weaknesses and divergences in the models across institutions?
• The proposed changes will fundamentally strengthen the IRB framework, and therefore we support this EBA process.• We need to ensure that these important and impactful changes will not end in vain due to the overshadowing BCBS proposals (revision of the SA framework for credit risk and more importantly the Capital Floors consultation).
• Finally we feel that there is a need for detailed guidance. As examples may serve Counting of days past/overdue, Downturn LGD and level of conservatism (versus data quantity, quality and modelling correctness), including how and where to apply levels of conservatism.
9. Are there any other aspects related with the estimation of risk parameters that should be clarified in the EBA guidelines?
• Downturn (LGD), including the determination of a full cycle, how to enrich data (downturn factor) if the data does not represent a full cycle, etc.• (in combination with downturn), how to apply the levels of conservatism (aka MoC / margin of conservatism): for example per model step (multiple layers), or just one overall. When is the data (quality and quantity) of sufficient level and when is the modelling of sufficient level to have a minimum level of conservatism. How to address levels of conservatism between PD and LGD, etc.
• LGD in-default: aspects such as over which period or confidence interval this needs to be calculated; should it be the best guess (making this a point-in-time estimate) or an average of different default scenarios?
10. Do you have dedicated LGD models for exposures in default that fulfil the requirements specified in section 4.3.4.(ii)?
• Some of the Dutch banks have and some do not have dedicated LGD models for exposures in default.11. Do you consider the direction of the proposed changes adequate to address the weaknesses and divergences in the treatment of defaulted assets across institutions?
• Overall, the proposed changes would quite adequately address the weaknesses and divergences in the treatment of defaulted assets across institutions. It, however, remains to be seen how the actual GL would pan out and how this can be implemented in bank’s practises.12. What else should be covered by the GL on the treatment of defaulted assets?
• The notion of “uncertainty” should be elaborated.13. What are the impacts for the institutions that should be considered when specifying the conditions for PPU and roll-out?
• Changes to the roll-out plan will probably not impact most Dutch banks substantially. The PPU is being used by most larger European banks to report sovereign exposure to EU member states on low risk weights. We think it is important that – regarding PPU – an level playing field is ensured, while the capital allocated to any risk type should reflect the actual risk.• Also execution of the roll-out plans and the “end-state” percentage of the portfolio reported on IRB, should be aligned and national discretions be removed. Hence, the supervisor should ensure that the roll-out plan is actually adhered to.
14. Do you expect that your organisational structure and/or allocation of responsibilities will have to be changed as a result of the rules described in section 4.3.5?
• No, compliance to the regulations as proposed in the draft RTS would not require significant changes in the organisational structure.15. Do you agree that CRM is a low priority area as regards the regulatory developments?
• It is of lower priority in comparison with the other aspects.16. Are there any other significant intra-EU or global discrepancies?
• Jurisdiction is seen as unintended driver for RWA-differences. We think that there can be valid reasons for differences across jurisdictions.• (copy of answer to Q2:) We think that there are valid reasons why risk weights, of for example the mortgage portfolios, are quite different per jurisdiction because of many underlying differences (e.g. social welfare, tax regimes, payment behavior, insolvency laws & bankruptcy laws). These differences are translated in differences in actual risks, which justifies these differences in risk weights. Alignment in back-testing methodologies and back-testing results are possible answers to regain trust.
17. Do you agree that the area of disclosures needs to be strengthened, in particular with regard to disclosures related with the benchmarking exercise, for instance by publishing them on the EBA website?
• Although we agree that information should be widely and publicly available, we do not agree with the publication of benchmarking exercises on the EBA website. Quantitative information publicly published by banks should always be accompanied by the qualitative explanation of this information. The solvability of banks cannot not simply be compared across banks based on quantitative information alone. The solvability of the bank is also determined by more qualitative factors such as the process of restructuring defaulted clients. In general, EBA’s benchmark exercise itself concluded that banks apply different definitions and assumptions. The results can therefore be misinterpreted very easily.18. Would you support EBA Guidelines targeted at disclosure requirements related with the IRB Approach and taking into consideration the proposals of the Basel Committee on those requirements? Which current disclosure requirements should be given the priority? What should be the timetable for such Guidelines?
• The benchmarking exercises itself might have a very positive effect on the confidence of stakeholders in the IRB approach. Therefore, EBA could consider this one of the top priorities. However, it should be very careful with the actions to take based on the data of the benchmarking exercises. EBA should consider that banks are very complex institutions that cannot be compared based on benchmarking exercises alone.19. Would you like to see any modification of the reporting framework implemented in terms of IRB exposures?
• The focus should be on reducing the unwanted variation in risk weights.20. What would you consider an appropriate solution with regard to the definition and treatment (modelling restrictions) of the low default portfolios?
• First of all there should be no labelling like “low default portfolio” versus “high default portfolios”. The proper terminology should be low data versus sufficient data portfolios, which can be different per institution.• We would favour clear rules for when the data quality and (more important in this respect) the data quantity is sufficient. In order words, when is the data too limited to build an internal model.
• How should institutions deal with expert models. Guidelines with respect to levels of conservatism should be clear as well.
• Rules on how to pool data are welcome too.
• Finally, it might make sense to come up with more risk sensitive FIRB LGD models, looking into the potential differences between markets (jurisdiction) and into the differences between exposure classes.
• We appreciate the desire for a best practise assessment in which regulators take stock and learn from lessoned learned in other jurisdictions.
21. How would you ensure appropriate use of the IRB Approach in a harmonised manner without excessive concerns of the so called ‘cherry picking’?
• In the current practice an institution needs to disclose a roll-out plan, in which the institution needs to evidence why certain portfolios would stay on SA (too small, to burdensome, etc). The new European rules on PPU are clear, in this respect there is no opportunity of cherry picking.22. Do you see merit in moving towards the harmonisation of the exposure classes for the purpose of the IRB and the Standardised Approach?
• Yes. If SA and IRB portfolios are going to be compared, then it would be best to apply consistency in this aspect.• However, given the current SA revision proposals it should be noted that it might be extremely challenging to align the SA exposure classes with the IRB exposure classes. We see extra challenges for non-banks financial institutions.
23. Would the requirement to use TTC approach in the rating systems lead to significant divergences with the internal risk management practices?
• To the extent that we can oversee it, we tend to answer no.• However, it could be valuable to provide guidance on PIT methodologies in the light of IFRS 9 and expected loss calculations.
24. Do you agree that the possibility to grant permission for the data waiver should be removed from the CRR?
• We take a neutral stand regarding the removal of the possibility to grant permission for the data waiver from the CRR although from a transparency point of view it is good to know what minimal data requirements are.25. Are there any other aspects of the IRB Approach not discussed in this document that should be reviewed in order to enhance comparability of the risk estimates and capital requirements?
• We would welcome more guidance on how to handle provisioning according to accounting standards (IFRS / IFRS9) and regulatory capital reporting.• We understand that reducing the unintended variances will require a substantial effort for all IRB banks. In the Netherlands the IRB portion is very high, therefore the efforts for Dutch banks is very substantial. As we strongly believe that IRB is the superior method we wish to put this effort into this process. We need to ensure that the BCBS proposals do not make our efforts in vain (by flooring the IRB risk weights to the SA risk weights via the Capital Floor).
• We would welcome a strong link between EBA and BCBS, so European institutions are not required to process impactful changes twice.
• We favor a decent roll-out plan, with a pre-determined sequencing (largest models first). This way, during the process, EBA can demonstrate that indeed the EBA initiative reduced variance (for the first exposure class that was tackled, which will give a positive signal to the market). All stakeholders would know and understand which models will be updated based on the new disclosed methodology, this makes the explanation easier. On the one hand, the first success can be celebrated even earlier, while on the other hand we can take more time for those exposure classes that are scheduled in the end (beyond 2019).
• Please take notice of the IIF RWA Task Force (IRTF).
• Requirements that should be adhered to in order to report portfolios on IRB, such as the Use Test, should be supervised in the same fashion across Europe without national discretions, creating a level playing field.