Response to consultation on draft RTS on the determination by originator institutions of the exposure value of SES in securitisations
Go back
Since currently the simplified model approach seems to be more conservative than the full model approach, Banks should be given the option to utilize the full model approach when they have the operational capacity to do so.
Typically though, the average time that takes for a revolving loan to be repaid is materially shorter both than the proposed “longest permitted maturity” and indeed the five year threshold proposed. In our view the proposed treatment in the asset model would overestimate the exposure amount of SES over the lifetime of the transaction.
Institutions should have the option of incorporating revolving exposures in their asset model during a replenishment period with assumption for their remaining maturity based on historical evidence for similar asset types.
It is worth contemplating extending the option of utilizing through the cycle probabilities of default to standardized institutions based on the merit of suitability rather than that of prudency as described in point b) ii).
In contrast with the trapped SES mechanism, heavily skewed loss distributions tend to reduce the available SES for back loaded scenarios. In the example provided in the draft RTS, the credit losses per annum over the WAL of the transaction that account for ~95% of the lifetime losses, imply a very low level of fluctuation around the average ratio of credit losses over outstanding balance, which according to historical evidence is materially higher.
For UIOLI mechanism transactions, it can be demonstrated that the front loaded scenarios are not sensitive to the increase of the variation of the credit loss ratio, but the back loaded scenarios are sensitive to it, leading to an amount of available to absorb losses SES that is significantly lower than the one in the example. Please note that heavily skewed loss distributions tend historically to be the norm rather than the exception.
For trapped SES mechanisms, the calculation of exposure value as proposed is expected to be much more consistent with observed examples.
As an alternative to the proposed UIOLI mechanism, an approach can be envisioned that would follow these principles:
• Create credit loss distributions scenarios, based on different credit loss rate variation parameters (2-3 multiples of the standard deviation of historical loss rates for the underlying asset type could be utilized)
• Calculate the arithmetic average for front loaded and back loaded scenarios for each level of variation.
• Assign weights to the averages of the previous bullet point, or use again a simple arithmetic average, for a single figure to be derived.
The methodology is simple enough for standardized banks to implement, although we would favor a value of the scalar factor that is less than the base 0.8 value proposed.
As evident from Table 7 of the draft RTS, the exposure value of SES –as in currently stands- is relatively comparable under the simplified method and under the full model method. As noted in Q6 & Q7 though, we believe that the scenario assumptions used to derive the full model figure lead to an overly conservative estimation of SES. Thus, we believe that the 0.8 factor used in the simplified model should be recalibrated alongside the full model in order to keep on producing similar results, to provide a level playing filed amongst institutions.
Since we already implement a similar methodology for existing transactions after receiving relevant supervisory guidance, we are able to verify that under the rolling window approach outstanding transactions remain economically viable.
a) With our role as originators we already have been affected by the introduction of capital requirements for available SES through the current supervisory approach. Although no grandfathering was provided for existing transactions, the Bank has tolerated the increased capital cost. Introducing a penalizing treatment for future SES would further adversely impact the economics of outstanding transactions and most probably result in an economic decision to exercise regulatory calls.
b) Synthetic securitisations have proved to be a useful tool for releasing capital and redirecting it to facilitate the execution of the Bank’s business plan, while at the same time financing the real economy. We further note that this would have a disproportional effect in SME synthetic securitisations where the use of SES is more prevalent, when compared to transactions including other asset classes (e.g. large corporate exposures).
Imposing additional capital costs would compromise the growth of the European securitisation market which is a stated goal amongst market participants.
c) Finally, returning to a technical issue, the different complexity of the two methodologies proposed should also mean a different level of oversight over their application. We believe that, for the simplified method, the proposed annual review would pose an unwarranted burden on the validation units of the Institution. An initial approval of the asset model and losses model utilized, accompanied with a mandate for revision where adequate reason is provided, would in our view suffice.
Q1. Do respondents find the provisions clear enough or would any additional clarification be needed on any aspect?
The definitions are clear enough.Q2. Do you agree with the possibility of choosing between the full and the simplified model approaches in a consistent manner?
We believe that an institution should have the flexibility to make that decision on a transaction level. Not all transactions have the same level of complexity and different asset types might be easier to model than others.Since currently the simplified model approach seems to be more conservative than the full model approach, Banks should be given the option to utilize the full model approach when they have the operational capacity to do so.
Q3. Instead, would you favour that the RTS consider only one method (i.e. the full model approach or the simplified model approach) for the calculation of the exposure value of the synthetic excess spread of the future periods?
No, because for more complex transactions the application of the Full Model Approach would be challenging for some banks but the Simplified Model Approach is very conservative for uniform use.Q4. Do you agree with the specifications of the asset model made in Article 3?
Revolving exposures for specific asset types are many times included in the selection pool by utilizing assumptions based on behavioral models for early repayments and rollovers. This is mainly due to the fact that, in some cases, revolving facilities are open loans with no stated maturity.Typically though, the average time that takes for a revolving loan to be repaid is materially shorter both than the proposed “longest permitted maturity” and indeed the five year threshold proposed. In our view the proposed treatment in the asset model would overestimate the exposure amount of SES over the lifetime of the transaction.
Institutions should have the option of incorporating revolving exposures in their asset model during a replenishment period with assumption for their remaining maturity based on historical evidence for similar asset types.
Q5. Do you agree with the specifications for the determination of the relevant losses made in Article 5?
Please note that, for a standardized bank that would use SCRAs as a proxy for future loses, IFRS9 accounting rules dictate that for stage 1 exposures the calculation would be based on point in time parameters. That would mean that during adverse periods of the economic cycle, these loss estimates could be significantly higher than those derived from internal model parameters, thus creating a unleveled playing field between standardized and IRB institutions.It is worth contemplating extending the option of utilizing through the cycle probabilities of default to standardized institutions based on the merit of suitability rather than that of prudency as described in point b) ii).
Q6. Do you agree with the calculation of the exposure value of synthetic excess spread for future periods made in Article 6?
No, we believe that, especially for UIOLI transactions, averaging over the front-loaded, even-loaded and back-loaded scenarios, would lead to an overly conservative SES exposure value.In contrast with the trapped SES mechanism, heavily skewed loss distributions tend to reduce the available SES for back loaded scenarios. In the example provided in the draft RTS, the credit losses per annum over the WAL of the transaction that account for ~95% of the lifetime losses, imply a very low level of fluctuation around the average ratio of credit losses over outstanding balance, which according to historical evidence is materially higher.
For UIOLI mechanism transactions, it can be demonstrated that the front loaded scenarios are not sensitive to the increase of the variation of the credit loss ratio, but the back loaded scenarios are sensitive to it, leading to an amount of available to absorb losses SES that is significantly lower than the one in the example. Please note that heavily skewed loss distributions tend historically to be the norm rather than the exception.
For trapped SES mechanisms, the calculation of exposure value as proposed is expected to be much more consistent with observed examples.
Q7. Shall the average of the scenarios be made in a different way for UIOLI and trapped mechanisms (e.g. back-loaded and evenly-loaded only for UIOLI mechanisms, and front-loaded and evenly-loaded for trapped mechanisms)?
As a general principle though and in line with our answer in Q6, evenly distributed loss scenarios are in contrast with historically observed loss distributions and, in that sense, they tend to distort the measurement. As a result, omitting the even loaded scenarios from the proposed methodology would by itself be a step in the right direction.As an alternative to the proposed UIOLI mechanism, an approach can be envisioned that would follow these principles:
• Create credit loss distributions scenarios, based on different credit loss rate variation parameters (2-3 multiples of the standard deviation of historical loss rates for the underlying asset type could be utilized)
• Calculate the arithmetic average for front loaded and back loaded scenarios for each level of variation.
• Assign weights to the averages of the previous bullet point, or use again a simple arithmetic average, for a single figure to be derived.
Q8. Do you agree with the specification of the simplified model approach made in Article 7?
Yes.The methodology is simple enough for standardized banks to implement, although we would favor a value of the scalar factor that is less than the base 0.8 value proposed.
Q9. Do you consider that the formula can be further simplified (e.g. by using the maturity of the credit protection multiplied by a conservative scalar instead of WAL)?
We disagree with any further simplification of the formula.Q10. Do you agree with the scalar assigned for UIOLI mechanisms? If not, please provide empirical evidence that justifies a different scalar based on the different loss absorbing capacity of UIOLI vs trapped mechanisms.
No, we believe that it is overly conservative.As evident from Table 7 of the draft RTS, the exposure value of SES –as in currently stands- is relatively comparable under the simplified method and under the full model method. As noted in Q6 & Q7 though, we believe that the scenario assumptions used to derive the full model figure lead to an overly conservative estimation of SES. Thus, we believe that the 0.8 factor used in the simplified model should be recalibrated alongside the full model in order to keep on producing similar results, to provide a level playing filed amongst institutions.
Would you favour that approach? If so, how do you think that this rolling-window approach for calculating UIOLI SES will affect the efficiency and viability of synthetic transactions in comparison with the current supervisory practices? Please justify your response with specific illustrative examples or data.
We believe that including the current supervisory practice of the rolling window approach in the regulation as a means to cover for SES, is the least disruptive course of action.Since we already implement a similar methodology for existing transactions after receiving relevant supervisory guidance, we are able to verify that under the rolling window approach outstanding transactions remain economically viable.
Q12. Do you agree with the treatment of the ex-post SES of future periods in the RTS? If not, please provide rationale and data supporting your views
We do not agree with it because that would deviate in essence for the current regulatory treatment. CRR does not impose capital requirements for expected losses in any other circumstances, as the prevailing view is that expected losses are covered by the excess of interest income over costs and unexpected losses are covered by capital.Q13.Do you have any other comments on these draft RTS?
Please note the following as general remarks to the draft RTS:a) With our role as originators we already have been affected by the introduction of capital requirements for available SES through the current supervisory approach. Although no grandfathering was provided for existing transactions, the Bank has tolerated the increased capital cost. Introducing a penalizing treatment for future SES would further adversely impact the economics of outstanding transactions and most probably result in an economic decision to exercise regulatory calls.
b) Synthetic securitisations have proved to be a useful tool for releasing capital and redirecting it to facilitate the execution of the Bank’s business plan, while at the same time financing the real economy. We further note that this would have a disproportional effect in SME synthetic securitisations where the use of SES is more prevalent, when compared to transactions including other asset classes (e.g. large corporate exposures).
Imposing additional capital costs would compromise the growth of the European securitisation market which is a stated goal amongst market participants.
c) Finally, returning to a technical issue, the different complexity of the two methodologies proposed should also mean a different level of oversight over their application. We believe that, for the simplified method, the proposed annual review would pose an unwarranted burden on the validation units of the Institution. An initial approval of the asset model and losses model utilized, accompanied with a mandate for revision where adequate reason is provided, would in our view suffice.