Response to consultation on Guidelines PD estimation, LGD estimation and treatment of defaulted assets
Go back
The significant evolution of the institutions models would require numerous validation processes by the regulator.
The most impacting part of the proposal would probably be the LGD estimation (in default and non -default), since it implies the collection of new statistical series of data and the building of new models. Above the main points of operational cost on IT systems, it raises the issue of planning and complying with the deadlines, considering implementation and validation delays.
Homogeneity is already required for pools of exposures in CRR (Art 170 (3)c)), then why would a MoC be required? What would be the criteria for homogeneity?
The requirement to calculate the one-year default rates at least quarterly will be difficult to implement and will request as a consequence IT development.
Due to the fact that the parameters are backtested and updated on an annual basis, it seems more approriate to have an annual review, but based on quarterly data.
The proposed formula for the calculation of Economic Loss seems not relevant for operations quitting the status of default.
Ex: in the case of 3 unpaid month instalments, that are repaid in the following month.
Observed recuperation would equal 0
Economic loss would equal EAD at the date of entry in default (if no fees)
LGD would equal 100% whereas there is eventually no loss.
To reduce heterogeneity, benchmarks would have to be realized by type of exposure, business models, localization… which seems too complex.
Specific business models such as consumer credit, leasing, guarantees…could be strongly impacted by the introduction of benchmarks of pools and grades and maximum PD levels: if same grades are applied for all business line, those specialized credit activities will be concerned by very few grades. It would be difficult to duly monitor the large scale of institutions business models / activities with a too low number of grades.
As a general comment, it has to be underlined that there is no definition of homogeneity in the EBA paper. It is important to have a clear definition in order to achieve homogeneity (the same issue is underlined further concerning LGD).
The “rating philosophy” is mostly the same for all the exposures.
The definition of Category C (General estimation errors) is less detailed than the three others, and not precise enough. No concrete example is given (not even in the annex I p. 23).
The proposed methodology for the estimation of the MoC does raise operational concern since we understand that two calculations will be required: the estimation with the available data, plus the estimation with the corrected data. This would be burdensome, time consuming and complicated to implement.
The superposition of two levels of MoC (by type and global), in case there are several MoC required, seems significantly complex, and could reintroduce variability in the different models.
The proposed approach with a very granular and analytic vision on margin of conservatism may lead to the aggregation of numerous margin of conservatism that would have a significant impact in capital requirement, and most of all will put in question the value of operationality of risk parameter. Therefore, and in order to avoid duplicates, entities should have to possibility to some extend to assess and apply margin of conservatism in a global way, or even not to apply any margin if it can be demonstrated that the deficiency itself lead to a conservative outcome.
Question 1: Do you agree with the proposed requirement with regard to the application of appropriate adjustments and margin of conservatism? Do you have any operational concern with respect to the proposed categorization?
For most specialized credit institutions, the impacts on current models would be very material. Due to the proposed Guidelines, they would have to modify consistently their methodology and so, to change all their models (For example currently for one consumer credit institution: 7 models of PD and 7 models of LGD). The costs of complying with these Guidelines will be very heavy, especially on IT systems.The significant evolution of the institutions models would require numerous validation processes by the regulator.
The most impacting part of the proposal would probably be the LGD estimation (in default and non -default), since it implies the collection of new statistical series of data and the building of new models. Above the main points of operational cost on IT systems, it raises the issue of planning and complying with the deadlines, considering implementation and validation delays.
Question 2: Do you see any operational limitations with respect to the monitoring requirement proposed in paragraph 53?
General comment on paragraph 44:Homogeneity is already required for pools of exposures in CRR (Art 170 (3)c)), then why would a MoC be required? What would be the criteria for homogeneity?
The requirement to calculate the one-year default rates at least quarterly will be difficult to implement and will request as a consequence IT development.
Due to the fact that the parameters are backtested and updated on an annual basis, it seems more approriate to have an annual review, but based on quarterly data.
Question 3: Do you agree with the proposed policy for calculating observed average default rates? How do you treat short term contracts in this regard?
General comment on paragraph 113:The proposed formula for the calculation of Economic Loss seems not relevant for operations quitting the status of default.
Ex: in the case of 3 unpaid month instalments, that are repaid in the following month.
Observed recuperation would equal 0
Economic loss would equal EAD at the date of entry in default (if no fees)
LGD would equal 100% whereas there is eventually no loss.
Question 4: Are the requirements on determining the relevant historical observation periods sufficiently clear? Which adjustments (downward or upward), and due to which reasons, are currently applied to the average of observed default rates in order to estimate the long-run average default rate? If possible, please order those adjustments by materiality in terms of RWA.
Benchmarks for number of pools and grades and maximum PD levels doesn’t seem relevant as regard to the heterogeneity of risk profiles and business models across the EU. It could be punishing for institutions which have a low risk profile.To reduce heterogeneity, benchmarks would have to be realized by type of exposure, business models, localization… which seems too complex.
Specific business models such as consumer credit, leasing, guarantees…could be strongly impacted by the introduction of benchmarks of pools and grades and maximum PD levels: if same grades are applied for all business line, those specialized credit activities will be concerned by very few grades. It would be difficult to duly monitor the large scale of institutions business models / activities with a too low number of grades.
As a general comment, it has to be underlined that there is no definition of homogeneity in the EBA paper. It is important to have a clear definition in order to achieve homogeneity (the same issue is underlined further concerning LGD).
Question 5: How do you take economic conditions into account in the design of your rating systems, in particular in terms of: d. definition of risk drivers, e. definition of the number of grades f. definition of the long-run average of default rates?
Currently, for some consumer credit institutions for example, economic conditions are only taken into account in the calculation of the long-run average of default rates.Question 6: Do you have processes in place to monitor the rating philosophy over time? If yes, please describe them.
The specialised activity of consumer credit mostly includes short term contracts. The requirement of an additional MoC would introduce undue divergence and variability in the models for this industry compared to other credit activity.Question 7: Do you have different rating philosophy approaches to different types of exposures? If yes, please describe them.
Currently, some consumer credit institutions for example make studies to observe migration between the different risk pools.Question 8: Would you expect that benchmarks for number of pools and grades and maximum PD levels (e.g. for exposures that are not sensitive to the economic cycle) could reduce unjustified variability?
For most specialized credit activity, the types of credit granted is usually “monoline”, meaning there is only one type of portfolio / type of exposure: ex. retail portfolio for consumer credit.The “rating philosophy” is mostly the same for all the exposures.
Question 9: Do you agree with the proposed principles for the assessment of the representativeness of data?
The EBA proposal doesn’t give any element on the calculation methodology of the Margins of Conservatism (MoC), which could eventually introduce variability among institutions, opposite the EBA objective.The definition of Category C (General estimation errors) is less detailed than the three others, and not precise enough. No concrete example is given (not even in the annex I p. 23).
The proposed methodology for the estimation of the MoC does raise operational concern since we understand that two calculations will be required: the estimation with the available data, plus the estimation with the corrected data. This would be burdensome, time consuming and complicated to implement.
The superposition of two levels of MoC (by type and global), in case there are several MoC required, seems significantly complex, and could reintroduce variability in the different models.
The proposed approach with a very granular and analytic vision on margin of conservatism may lead to the aggregation of numerous margin of conservatism that would have a significant impact in capital requirement, and most of all will put in question the value of operationality of risk parameter. Therefore, and in order to avoid duplicates, entities should have to possibility to some extend to assess and apply margin of conservatism in a global way, or even not to apply any margin if it can be demonstrated that the deficiency itself lead to a conservative outcome.