Response to discussion Paper on the future of the IRB Approach
Go back
Yet, the sequential implementation seems difficult to achieve, since each of the different elements that have to be analysed are interrelated. For instance, a change in the definition of default would have immediate consequences on LGD.
The implementation planning seems ambitious and the impacts on current models difficult to evaluate since the most important and significant planned modifications are not defined precisely enough (ex. Default materiality thresholds, ex. period of time and conditions to be taken into account for multiple defaults to be considered as a unique default). The significance and the complexity of the changes in the definition of default will determine the level of difficulty to implement the revised models in the information systems, and to reconstitute historical data on the basis of which revised risk parameters will have to be calibrated.
It is difficult to take position on the proposed timing without having precise enough elements necessary for impact studies on the existing models. Most definitions given at that stage are too wide.
Yet, we consider that the planned achievement of the works in 2018 seems ambitious, especially for those portfolios for which the reconstitution of historical data on the basis of new definitions could be long, even before considering the implementation of the revision of the models themselves, and not counting the necessary adjustments to make in the operational processes. We think that implementation timings and deadlines should rather be adapted, following specific discussions with NCAs.
More generally, the ability of NCAs to certify the revised models, simultaneously on a short period of time, raises interrogations.
At last, it does not seem coherent to review the principles of the assessment methodology of IRB models to be used by NCAs, with a planned application by 2015, when the planned modifications for the rebuilding of the models are not stabilized, and require for most of them complementary details and specifications. The result would be that a revised assessment methodology of IRB models RTS would be in force, while consultations on some issues with strong impacts would still be on course.
The methodology of calculation for repayment allocations (FIFO or LIFO)
The definition of “technical defaults”
The final definition of materiality threshold to use to determine a situation of default
The definition of the return to non-defaulted status on expert bases
The contagion mode to entity exposures
The ability to extend to 180 days the past due criteria for real estate exposures (strong impact on historical data)
The new definition of multiple defaults requires precisions on the definition of the “short term” during which the exposure leaves and comes back to the defaulted status.
ASF agrees that the proposed modifications will reduce the divergences across current models, and also reduce the margins of interpretation.
Yet we also consider it might be dangerous to treat in a too homogeneous manner very different types of credit activities (banking loans, leasing, factoring, consumer credit…). For instance there is little common point between a small amount 3 years lease to an SME, in which the credit institution owns the asset financed, and a significant classic banking loan to a large corporate.
We want to attract the attention of the regulator on the fact that a too strong homogenization would in fine reduce the IRB models risk sensibility.
Other aspects could be clarified in the field of the IRB approach revision, so that to clarify definitions and perimeters:
Actualization rates to use for the calculations?
The starting point of the loss amount calculation (should events previous to the default be taken into account, and which events?).
For PD: the definition of low default portfolios should be detailed.
For LGD: the definition of “indirect costs” should be detailed.
How should the refunding costs be taken into account in addition to the actualization costs?
The definition of LGD downturn should be detailed.
Yet, the proposed revised definitions would deserve to be more detailed to allow effective impact studies on the current models.
The achievement of the improvement of the models comparability will also depend on the Guidelines and recommended methodologies that will come along with the new regulatory products. For instance, some credit institutions currently use LGD parameters with estimations based on the average weighted by the amount of defaults (cf. CRR). A new methodology implying estimations based on the average weighted by the number of defaults would require close examination to ensure that it does not reduce the risk sensibility.
It also raises the issue of the ability for NCAs to absorb in that period of time the certification demands.
Take the stress tests into account will have a strong impact on the capital requirements. It is essential to precise if the stress test taking into account is to be treated in pillar one or in pillar two.
Concerning benchmarking requirements, we also consider that precisions should be given on the benchmarking processes and modalities before examining the disclosure requirements.
We would like to underline the difficulty of the benchmarking process concerning some portfolios for which information is hardly available, such as low defaults portfolios. New benchmarking requirements still raise a series of other questions, for instance: to which entities would the benchmarking reporting be dedicated? Would the benchmarking require acquiring external data? How could their cost be evaluated?
A too strict convergence objective could lead to a less risk sensitive approach, especially if expert based analyses were no longer authorized. It would result in a “one size fits all” model, with reduced risk sensitiveness to the real risks of the activity. A systemic risk could even be taken if all credit institutions of one same activity line were to use the same type of risk analysis model to evaluate the risk on LDP.
The requirement to use TTC approach would have operational impacts since the PIT approach remains in use in other fields, such as provisions calculation.
Moreover, the removal of the possibility to grant permission for the data waiver from CRR would introduce more rigidity in the models implementation by imposing an intangible period of five years to implement IRB models.
2. What would you consider the areas of priorities?
See Q33. 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?
The proposed prioritization of the different regulatory products (definition of default, LGD, conversion factors estimation, treatments of defaulted assets and credit risk mitigation) seems a logical prioritization order.Yet, the sequential implementation seems difficult to achieve, since each of the different elements that have to be analysed are interrelated. For instance, a change in the definition of default would have immediate consequences on LGD.
The implementation planning seems ambitious and the impacts on current models difficult to evaluate since the most important and significant planned modifications are not defined precisely enough (ex. Default materiality thresholds, ex. period of time and conditions to be taken into account for multiple defaults to be considered as a unique default). The significance and the complexity of the changes in the definition of default will determine the level of difficulty to implement the revised models in the information systems, and to reconstitute historical data on the basis of which revised risk parameters will have to be calibrated.
It is difficult to take position on the proposed timing without having precise enough elements necessary for impact studies on the existing models. Most definitions given at that stage are too wide.
Yet, we consider that the planned achievement of the works in 2018 seems ambitious, especially for those portfolios for which the reconstitution of historical data on the basis of new definitions could be long, even before considering the implementation of the revision of the models themselves, and not counting the necessary adjustments to make in the operational processes. We think that implementation timings and deadlines should rather be adapted, following specific discussions with NCAs.
More generally, the ability of NCAs to certify the revised models, simultaneously on a short period of time, raises interrogations.
At last, it does not seem coherent to review the principles of the assessment methodology of IRB models to be used by NCAs, with a planned application by 2015, when the planned modifications for the rebuilding of the models are not stabilized, and require for most of them complementary details and specifications. The result would be that a revised assessment methodology of IRB models RTS would be in force, while consultations on some issues with strong impacts would still be on course.
4. Are there any other aspects related with the application of the definition of default that should be clarified in the GL?
See Q65. 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?
See Q66. 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?
We consider that the following points are not defined clearly enough and would require additional precisions in order to measure the impact of the planed modifications: The methodology of calculation for repayment allocations (FIFO or LIFO)
The definition of “technical defaults”
The final definition of materiality threshold to use to determine a situation of default
The definition of the return to non-defaulted status on expert bases
The contagion mode to entity exposures
The ability to extend to 180 days the past due criteria for real estate exposures (strong impact on historical data)
The new definition of multiple defaults requires precisions on the definition of the “short term” during which the exposure leaves and comes back to the defaulted status.
7. 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?
See Q98. Do you consider the direction of the proposed changes adequate to address the weaknesses and divergences in the models across institutions?
See Q99. Are there any other aspects related with the estimation of risk parameters that should be clarified in the EBA guidelines?
Some of the proposed modifications have already been subject to consultation, but as their calibration is not detailed yet, they do not allow an effective impact study on existing models as many questions remain open (ex. would the new definitions be compatible with the LGD and PD classes currently used in the existing models? Will historical data have to be reconstituted?).ASF agrees that the proposed modifications will reduce the divergences across current models, and also reduce the margins of interpretation.
Yet we also consider it might be dangerous to treat in a too homogeneous manner very different types of credit activities (banking loans, leasing, factoring, consumer credit…). For instance there is little common point between a small amount 3 years lease to an SME, in which the credit institution owns the asset financed, and a significant classic banking loan to a large corporate.
We want to attract the attention of the regulator on the fact that a too strong homogenization would in fine reduce the IRB models risk sensibility.
Other aspects could be clarified in the field of the IRB approach revision, so that to clarify definitions and perimeters:
Actualization rates to use for the calculations?
The starting point of the loss amount calculation (should events previous to the default be taken into account, and which events?).
For PD: the definition of low default portfolios should be detailed.
For LGD: the definition of “indirect costs” should be detailed.
How should the refunding costs be taken into account in addition to the actualization costs?
The definition of LGD downturn should be detailed.
10. Do you have dedicated LGD models for exposures in default that fulfil the requirements specified in section 4.3.4.(ii)?
See Q1211. Do you consider the direction of the proposed changes adequate to address the weaknesses and divergences in the treatment of defaulted assets across institutions?
See Q1212. What else should be covered by the GL on the treatment of defaulted assets?
The proposed changes seem to be adequate to address the weaknesses and divergences in the treatment of defaulted assets across institutions, with the reserves expressed in questions 7, 8 and 9.Yet, the proposed revised definitions would deserve to be more detailed to allow effective impact studies on the current models.
The achievement of the improvement of the models comparability will also depend on the Guidelines and recommended methodologies that will come along with the new regulatory products. For instance, some credit institutions currently use LGD parameters with estimations based on the average weighted by the amount of defaults (cf. CRR). A new methodology implying estimations based on the average weighted by the number of defaults would require close examination to ensure that it does not reduce the risk sensibility.
13. What are the impacts for the institutions that should be considered when specifying the conditions for PPU and roll-out?
The five years period for roll out could appear to be too short in some cases. For instance when the model is to be implemented in several entities or jurisdictions.It also raises the issue of the ability for NCAs to absorb in that period of time the certification demands.
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?
The main evolution is the required independency between the credit risk control units and the units in charge of the validation of the models. Specialized credit institutions that are subsidiaries of larger groups may satisfy the request in relying on the group level validation units. In smaller credit institutions, the impact on the organization may be stronger. It is essential that the principle of proportionality mentioned by the EBA should be respected.Take the stress tests into account will have a strong impact on the capital requirements. It is essential to precise if the stress test taking into account is to be treated in pillar one or in pillar two.
15. Do you agree that CRM is a low priority area as regards the regulatory developments?
CRM is a lower priority area as regards the regulatory developments. However, it would be useful to give attention to the CRM issue while considering the GL on LGD. Yet, its impact on capital requirement is potentially strong. CRM will then also have to be subject to close examination.16. Are there any other significant intra-EU or global discrepancies?
No comment.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?
See Q1918. 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?
See Q1919. Would you like to see any modification of the reporting framework implemented in terms of IRB exposures?
We think that impact studies concerning templates of reporting should also take into account the technical aspect of the ability of software editors to integrate the changes.Concerning benchmarking requirements, we also consider that precisions should be given on the benchmarking processes and modalities before examining the disclosure requirements.
We would like to underline the difficulty of the benchmarking process concerning some portfolios for which information is hardly available, such as low defaults portfolios. New benchmarking requirements still raise a series of other questions, for instance: to which entities would the benchmarking reporting be dedicated? Would the benchmarking require acquiring external data? How could their cost be evaluated?
20. What would you consider an appropriate solution with regard to the definition and treatment (modelling restrictions) of the low default portfolios?
We consider that it could be difficult to achieve the revision of IRB approach objective of homogenization as regards to Low Default Portfolios. Statistical approaches are not always relevant and the risk evaluation concerning LDP portfolios are often the result of expert based approaches (ex. Factoring activities).A too strict convergence objective could lead to a less risk sensitive approach, especially if expert based analyses were no longer authorized. It would result in a “one size fits all” model, with reduced risk sensitiveness to the real risks of the activity. A systemic risk could even be taken if all credit institutions of one same activity line were to use the same type of risk analysis model to evaluate the risk on LDP.
21. How would you ensure appropriate use of the IRB Approach in a harmonised manner without excessive concerns of the so called ‘cherry picking’?
PPU authorization should also be given regarding the aspects of feasibility and IT cost proportion of the integration of some exposures classes in the IRB models.22. Do you see merit in moving towards the harmonisation of the exposure classes for the purpose of the IRB and the Standardised Approach?
If the perspective of convergence between IRB and standardized approach is maintained, with requirements to benchmark IRB model results with standardized approach results, then the exposures classes’ convergence would be essential to ensure the comparability of the different results.23. Would the requirement to use TTC approach in the rating systems lead to significant divergences with the internal risk management practices?
Even if considered as less reactive and fine than PIT approach, the TTC approach is already mainly in use in current IRB models, according to supervisor’s recommendations. Yet, ASF would welcome the reopening of discussions on the use of TTC or PIT approaches.The requirement to use TTC approach would have operational impacts since the PIT approach remains in use in other fields, such as provisions calculation.
24. Do you agree that the possibility to grant permission for the data waiver should be removed from the CRR?
We consider that the removal of the possibility to grant permission for the data waiver from CRR would have a negative impact. For some activity lines and some portfolios, the constitution of a five years historical data series is complicated to achieve. The possibility to grant permission for the data waiver is important as it allows adjustments in the treatment of certain portfolios for which historical data are not especially relevant in the risk analysis.Moreover, the removal of the possibility to grant permission for the data waiver from CRR would introduce more rigidity in the models implementation by imposing an intangible period of five years to implement IRB models.