Response to consultation on draft RTS on performance-related triggers in STS on-balance -sheet securitisations
Go back
Rather, the focus of this trigger should be on observed divergences of defaults or losses (as applicable) from the defaults/losses indicated by the modelling, and should not be affected by scheduled amortisation. Accordingly, AFME members agree that specifying that this trigger is to be measured by reference to the original portfolio size is appropriate. One important point of clarification should be made, however. The reference to "the time of origination" should be replaced with a reference to the closing of the transaction (ie, the time the securities are issued or, for a securitisation which does not involve the issuing of securities, the time the initial securitisation positions are created). This is to avoid any confusion that "origination" refers to the time of origination of the underlying exposures themselves.
The rules should also contemplate the possibility of transactions where the underlying portfolio is rampedup over a period of time after the initial closing date of the transaction. In such a case, the trigger should be measured by reference to the lesser of the maximum securitised portfolio amount on the closing date and the gross amount of the securitised exposures included in the portfolio (ie, prior to taking any repayments or disposals into account) as at the end of the ramp-up period.
commonly used in synthetic securitisations for several years. In contrast, and as noted above, the proposed additional backward-looking trigger set out in Article 3 of the Draft RTS is not included in the SRT Report, and it has rarely been used previously in synthetic securitisations in the market. Accordingly, including this trigger would be a change from the experience with current practice in the SRT market.
That said, if this trigger were to apply instead of the requirement to include one of the additional backwardlooking triggers set out in Recommendation 2 to the SRT Report (ie, that SRT can be achieved without one of those additional triggers), AFME members broadly agree with the principle underpinning this trigger.
As a point of clarification, it is not clear how the reference to "the outstanding balance of the pool of underlying exposures" in Article 256(1)–(2) of the CRR should be interpreted in the context of a synthetic securitisation. This is because in a synthetic securitisation, the tranches would usually be written down by initial loss amounts (as, indeed, is impliedly required by Article 26e(2) of the EUSR) shortly after the occurrence of a credit event, without there necessarily being a corresponding reduction in the outstanding balance of the defaulted exposure. This therefore creates a mismatch where the sum of the outstanding balance of the tranches may be less than the actual outstanding balance of the securitised portfolio. This means that (i) the attachment point of the first loss tranche would become greater than zero, and (ii) the detachment point of the protected tranche would be artificially higher.5 While we acknowledge that the EBA's mandate under Article 26e(5) of the EUSR does not extend to clarifying the interpretation of Article 256(1)–(2) of the CRR, to avoid this anomaly causing issues for the application of the proposed trigger, it may be helpful to clarify that for the purpose of the calculation in Article 256(2) of the CRR and calculating the outstanding balance of the pool of underlying exposures, the outstanding balance of a defaulted exposure should be deemed to be equal to the protected amount minus the initial credit protection amount.
formulation of that trigger as set out in Article 3 of the Draft RTS, and not the inconsistent description set out in Paragraph 13 of the Consultation Paper.
AFME members have a number of concerns with the alternatives option for the additional backwards-looking trigger. First, we have the same concerns outlined in the Introductory Comments and the response to Q2 arising from the fact that this trigger is not one of the ones already identified in the SRT Report. Secondly, the way this trigger is formulated, once some losses have been incurred, the likelihood of the trigger level being reached increases on an exponential basis purely as a result of scheduled amortisation of the portfolio. Unlike the option discussed in Q2, where the detachment point only decreases as a result of losses, because the alternative trigger calculates the losses as a percentage of the amortised protected tranche balance, that percentage will increase as amortisation increases. This is an undesirable outcome, as if the rate of losses is not increasing, there is no reason why expected amortisation should move the transaction closer to a switch to sequential amortisation. As such, the proposed trigger is not fit for purpose and does not meet the regulatory requirement of protecting the retained senior tranche in the event of rising defaults and losses.
Thirdly, because the trigger does not distinguish between EL and UL, or indeed between those potential losses which were taken into account when structuring the transaction and those which were not, the occurrence of any losses allocated to the protected tranche will eventually lead to the trigger being hit as the portfolio amortises, even if those losses are entirely consistent with base case modelling. Given that that modelling is an important part of the SRT assessment process, it is not appropriate for the triggers to be structured in a way which does not take that into account. Fourthly, AFME members agree with the EBA's observation in Paragraph 13 of the Consultation Paper that the effect of the alternative option will vary depending on the transaction structure. In particular, it will be much more restrictive for a first loss transaction than for a mezzanine transaction having the same detachment point, despite the fact that in such a scenario a first loss transaction provides greater risk transfer from the originator's perspective. That is, for a first loss transaction, all losses will contribute to the erosion of the protected tranches (and thus eventually lead to the trigger being hit), whereas for a mezzanine transaction, there would be no erosion of the protected tranche until the first loss tranche has been completely exhausted.
In the case of a mezzanine transaction, this also means there is a step-change in the impact of losses. Those which fall into the first loss tranche have no impact on the trigger, whereas losses which are allocated to the mezzanine tranche have a leveraged impact on the performance of the trigger. Given the efficiency benefits provided by pro-rata amortisation to the transaction overall, this may result in originators pursuing mezzanine transactions in place of first loss transactions where otherwise that may not have been the case. It would be an unfortunate outcome of the STS framework if it incentivised originators to pursue transactions involving less risk transfer for the purpose of being permitted to apply a lower risk-weight to the retained senior tranche.
Finally, because Article 26e(2) requires that interim credit protection payments are based on the higher of the accounting impairments and the loss given default under Chapter 3 of Title II of Part Three of the CRR, the general expectation is that these losses will be conservative (ie, higher than the realised credit protection amount at the end of the workout period). One effect of this is that the trigger will be activated earlier than it would be if based on realised losses alone, and indeed in circumstances where there is a reasonable expectation that once those interim losses are converted into realised losses, the trigger level would not have been reached. This later point supports the proposition that it is appropriate for it to be possible to revert back to non-sequential amortisation if a backward-looking trigger level (particularly this one) is reached. In this regard please see also our response on Q6, below.
In light of the above concerns, of the two backward-looking triggers proposed in the Draft RTS, AFME
members consider that the "alternative option" is not fit for purpose and will lead to undesirable outcomes by incentivising banks to pursue mezzanine transactions involving less risk transfer. In contrast, the EBA's "preferred" option does not suffer from these deficiencies and is workable in practice, notwithstanding the general concerns expressed in relation to the entire approach set out in the Introductory Comments, above.
To return to the purpose of the performance-related triggers in the first place, the concern with pro-rata amortisation is that it will result in the protected tranches amortising away before losses crystalise, and thus before they can be borne by the protected tranches.6 The triggers to switch to sequential amortisation should, therefore, be calibrated in such a way as to avoid this outcome. By definition, however, that therefore means that the appropriate level for the trigger depends on the portfolio and transaction structure, and how it is expected to perform in the various scenarios which the originator is required to model as part of the SRT assessment process (see, eg., Section 3.3 of the SRT Report). The triggers should be designed such that they don’t result in a switch to sequential amortisation unnecessarily when the securitised portfolio performs within its expected range, but only when it starts to diverge negatively and materially from the expected case.
In light of this, AFME members strongly urge the EBA not to set the trigger on a "one size fits all" basis for all synthetic STS securitisations. While such an approach may be superficially attractive on the basis that it appears to be simple, it would undermine the very purpose of the trigger, and create significant inefficiencies in the SRT framework. It would also increase the risk of transactions being structured to fit the mandatory trigger level rather than providing a framework to encourage prudent transactions.
Further, while AFME members recognise that, as set out in Paragraph 104(b) of the EBA Report on the STS Framework for Synthetic Securitisation (EBA/Op/2020/07) (the "STS Report"), and unlike in the case of traditional STS securitisation, the criteria for synthetic STS securitisation are intended to provide adequate protection for the originator as well as investors, it is also important to ensure that nothing in the STS framework cuts across the SRT assessment process. That is, while having STS criteria which are consistent with the requirements for SRT does no harm, the STS framework is not the place to "gold plate" the SRT rulebook, or establish additional requirements that are not driven by the need to provide appropriate protection for investors, as is the case here given that the "protection" afforded by the performance-related triggers is primarily for the senior tranche retained by the originator.
(i) Where the originator applies the IRB Approach and estimates a regulatory PD for all exposures in the portfolio the trigger would occur if the exposure-weighted average PD of the portfolio compared to the corresponding value at the time of origination is greater than a given percentage (the "PD Approach"). What that percentage should be is not stated in the rule, and so is presumably for the originator to determine.
(ii) Where the originator does not estimate a regulatory PD for all exposures in the portfolio, then the trigger would occur if the ratio of the outstanding amount of underlying exposures assigned to higher credit risk buckets to the outstanding amount of the underlying portfolio is greater than a given percentage (the "Risk Bucket Approach"). Again, what that percentage should be is not specified, and so is presumably for the originator to determine.
For the purpose of applying the Risk Bucket Approach, it is therefore necessary to determine which exposures should be allocated to the higher risk buckets. Article 4(4) of the Draft RTS attempts to do this, and again draws a distinction between where the originator applies the IRB Approach and where it applies the Standardised Approach as follows:
(A) Where the originator applies the IRB Approach to the exposures, the originator should use the rating grades for the relevant type of asset under Articles 153(5), 170(1) and 170(3), as applicable, and assign each exposure to the appropriate rating grade. However, the rule does not specify which rating grades are to constitute the "higher credit risk buckets" for the purpose of the trigger.
(B) Where the originator does not apply the IRB Approach to the exposures, it should use the differentiation of credit risk as recorded by it in its financial statements in accordance with the applicable accounting framework. No further guidance is provided on how this is to occur.
(C) Regardless of whether (A) or (B) applies, all exposures classified as in default within the meaning of Article 178(1) of the CRR, all exposures to a credit-impaired debtor and all other exposures entailing higher credit risk shall be assigned to the higher credit risk buckets.
AFME members have many concerns with this proposed trigger. First, as with the proposed additional backwards-looking triggers, it does not sit neatly with the forward-looking triggers set out in the SRT Report, although, at least where the PD Approach applies, it is similar to forward-looking trigger (i) from Recommendation 2 of the SRT Report. However, whereas the SRT Report expressly refers to the "weighted average 1-year PD of the underlying portfolio", the Draft RTS refers simply to the "exposure weighted average PD of the underlying portfolio". It is not clear if this is intended to be different.7
Secondly, it is not actually clear whether the PD Approach or the Risk Bucket Approach is intended to apply where the originator does estimate a PD for all exposures in the portfolio, but they fall into different asset types (eg., retail and non-retail). Given that the trigger works very differently depending on which approach is followed, this lack of clarity should be rectified.
We also note that the two forms of the trigger work quite differently in practice. In the case of the PD Approach, by being based on the weighted average PD of all securitised exposures, whether or not the trigger is hit depends on the performance of the portfolio as a whole, and a deterioration in the PD of some exposures may be offset in this regard by an improvement in the PD of other exposures. In contrast, the Risk Based Approach looks only to the proportion of the portfolio which is allocated to the higher credit risk buckets, and thus applies a binary test which ignores the relative performance of individual exposures allocated to the lower credit risk buckets.
Following from this, there does not appear to be any good reason for preventing the originator from using the Risk Bucket Approach, even if it does estimate the PD for all exposures in the underlying portfolio. Such a trigger would be in line with market practice and is an approach which aligns with how many investors approach their analysis of the originator's risk framework. Similarly, where the originator does not calculate a regulatory PD for all exposures, there does not appear to be any good reason for not allowing the originator to base the trigger on the weighted average accounting probability of default. As discussed in the Introductory Comments, there is nothing in Article 26c(5) of the EUSR which specifies that only one additional backward-looking trigger and one forward-looking trigger can be specified in the RTS. In light of that, AFME members remain of the view that providing a menu of triggers from which the originator can select the most appropriate triggers for a given transaction would be a much better approach.
Thirdly, as also noted above, it is not at all clear how the credit risk grades are to be assigned where the originator is required to use those recorded in its financial statements. Presumably it is not sufficiently granular simply to refer to IFRS Stage 1, 2 and 3, and it should be clear that the originator can make such allocation based on other factors such as IFRS 9 PDs or rating buckets. It would also be appropriate to permit the originator to use other metrics, such as internal ratings, to determine the credit risk grades, as is indeed common in many existing transactions in the market, and this should be stated explicitly in the RTS. The current reference to the applicable accounting framework in the Draft RTS is simply too vague to be workable in practice.
Fourthly, a change in PD (or allocation of exposures to credit risk buckets) may be driven by factors other than an actual change in the performance of the securitised portfolio. That is, a change in the rating model or various add-ons introduced or removed by the regulator following a model review may lead to a change in the exposure-weighted PD of the portfolio, even though there is no observed increase in the rate of defaults. Given that such changes would generally not be caught by a regulatory call,8 it would be disadvantageous for the originator if one outcome of such a review was to trigger a switch to sequential amortisation for a portfolio that is actually performing as expected. This also means that the output of the trigger may be unstable, as it would depend on what add-ons may or may not be applied at different times over the life of the transaction.
In light of the above, AFME members feel that this trigger is somewhat over-engineered, and runs the risk that it could have unintended consequence in certain cases. We invite the EBA to consider whether the trigger could be simplified and made more flexible so as to require a forward-looking trigger based on either an increase in an exposure-weighted credit risk metric or an increase in proportion of exposures categorised in a higher credit risk bucket above a predefined threshold, where such metric or buckets are based on the relevant regulatory or accounting framework, or internal credit risk metrics used by the originator.
AFME Members do welcome the fact that the actual trigger levels are not specified for either the PD Approach or the Risk Bucket Approach, which ensures that the originator is able to calibrate the triggers at a level which is appropriate for the transaction, and also avoid creating tension with the SRT assessment process.
AFME members also welcome the proposal in Article 4(6) of the Draft RTS that if, following the activation of this trigger, the exposure-weighted PD or percentage of exposures allocated to higher credit risk buckets falls back below the threshold, the transaction should be permitted to switch back to non-sequential amortisation. However, it should be clarified that this only requires the level to fall below the threshold, and not to return to the level at the time of closing of the transaction. Further, AFME members consider that the requirement that the trigger be cured for four consecutive quarters is overly conservative, and that a shorter period should be sufficient.
In addition to the concerns outlined above, AFME members have a number of additional technical comments in relation to the proposed trigger:
(i) First, defaulted exposures should be excluded for the purpose of the forward-looking test on basis that
the default has already occurred, and the effect of that default should be captured in the backward-looking triggers. A forward-looking trigger should only take into account factors where there is a degree of uncertainty as to the outcome. Including the 100% PD for defaulted exposures in this calculation would also significantly skew the overall calculation and would require the trigger to be set at a much higher level to avoid it being hit solely as a result of an expected level of defaults, thereby reducing the sensitivity of the trigger to unexpected deterioration in the overall portfolio.
(ii) Secondly, it should be clarified that the reference to "exposure-weighted" should be interpreted as requiring weighting by the protected amount in the securitisation. It is quite common for synthetic securitisations to reference only a part of a larger exposure, and it would therefore not be appropriate to use the overall exposure amount for this calculation.
(iii) Thirdly, it should be clarified what is to happen if the PD Approach applies initially, but the originator subsequently ceases to estimate a regulatory PD for all exposures in the portfolio, as has been the case in some previous transactions in the market.
(iv) Fourthly, the requirement in Article 4(5) of the Draft RTS that all exposures "entailing higher credit risk" shall be assigned to the higher risk buckets appears to ride roughshod over the method for allocating exposures to buckets. What exactly is meant by exposures "entailing higher credit risk" that is not already covered by the methodology in Article 4(4)? Indeed it is not clear that Article 4(5) as a whole serves any purpose that is not already covered by Article 4(4).
(v) Fifthly, for the same reasons set out in our response to Q2, above, the reference to "time of origination" in Articles 4(1) and 4(2) of the Draft RTS should refer to the closing of the securitisation.
At the same time, there are some types of triggers which cannot be cured. Examples include triggers based on the cumulative final losses or cumulative gross defaults measured as a percentage of the initial portfolio amount.
Despite the fact that many performance-related triggers are capable of being cured, it is relatively rare for synthetic securitisations in the market to permit a switch back to non-sequential amortisation where that occurs. In the majority of cases, once a trigger is activated, the switch to sequential amortisation is permanent. Of course this makes the calibration of the trigger levels all the more important, and also means that the originator is wary of setting the level for certain types of trigger too low. For example, in a portfolio which may be expected to experience many defaults but very low losses, a trigger which is based on the cumulative unmatured defaults would need to be set quite high to avoid the risk that sequential amortisation would be triggered even though actual losses are quite low. In contrast, if it were possible to switch back to non-sequential amortisation where the trigger falls below the relevant threshold, the implications of activating the trigger would be less, meaning that it may be possible to set the trigger at a lower level in the first place.
As noted in our response to Q5, above, AFME members are in favour of the proposal to allow a switch back to non-sequential amortisation if the forward-looking trigger is cured, albeit that we consider that the requirement for the trigger to be cured for four consecutive periods is overly conservative. This is particularly the case given that the performance of the forward-looking trigger may be affected by short term market conditions which resolve themselves without actually leading to a significant increase in defaults. A good example of this is the experience during the Covid-19 pandemic, when many portfolios experienced a significant increase in weighted average PDs or significant downward migration of exposures into higher credit risk grades, which has subsequently been largely reversed without having experienced significantly increased defaults.
AFME members also consider that it would be appropriate to permit a switch back to non-sequential amortisation if the backward-looking triggers are cured (where that is possible given the nature of the trigger). Indeed, this is perhaps even more appropriate for the backward-looking triggers than the forward-looking trigger as a backward-looking trigger will only cure once the circumstances giving rise to the trigger have been fully taken into account by the tranches, such that a "reset" of the amortisation mechanics going forward can proceed with certainty. In contrast, the forward-looking triggers always involve a degree of uncertainty and a reduction in the trigger metrics below the relevant threshold is obviously not a guarantee that the future performance of the remaining exposures could not cause the trigger level to be exceeded again in the future.
In terms of timing, AFME members consider that it should be sufficient for the trigger to be cured for two consecutive quarters.
As noted in our response to Q5, above, we also consider that it should not be necessary for the originator to describe in detail the process by which exposures are allocated to credit risk buckets for the purposes of the Risk Bucket Approach.
When the SRT Report was published in November 2020, the menu of backward and forward-looking triggers set out in Recommendation 2 largely reflected the types of triggers which had been common for many years in synthetic securitisations across the EU. In contrast, the proposals in the Draft RTS, particularly in relation to the additional backwards-looking trigger, appear to reflect a change of thinking from the EBA as to what are the appropriate triggers for this purpose. If that is the case, then it would be very helpful if the EBA could provide further clarity on its expectations in this regard as far as the SRT assessment process is concerned. It is clearly not appropriate to have two sets of overlapping but different requirements that apply for synthetic STS securitisations as apply for synthetic SRT securitisations, particularly given that virtually all synthetic STS securitisations are also SRT securitisations.
As to the substance, and as set out above, AFME members consider that it would be more appropriate for the RTS to provide a menu of backward and forward-looking triggers from which the originator can select the most appropriate triggers for a given transaction, a selection which will form part of the SRT assessment process.
Further, there should also be an explicit acknowledgement in the RTS that the originator is not prevented from including additional performance-related triggers beyond whatever is prescribed in the RTS. While this may be impliedly the case given the reference to "at a minimum" in Article 26c(5), given the importance of being able to include such triggers to satisfy the SRT assessment process, it is appropriate to clarify that doing so does not fail to meet the "standardisation" requirements of the STS framework.
Q1. Do you agree with the specification made in Article 2?
Yes. AFME members are of the view that this is the only sensible way in which this trigger can be applied.4 If, instead, the trigger referred to the outstanding amount of the underlying portfolio from time to time, then, at least in theory, scheduled amortisation of the portfolio would of itself increase the likelihood of the trigger applying, even if no further defaults or losses occur. To avoid such an impact, the trigger would need to be set at a higher percentage than would otherwise be the case, which would undermine the purpose of the trigger.Rather, the focus of this trigger should be on observed divergences of defaults or losses (as applicable) from the defaults/losses indicated by the modelling, and should not be affected by scheduled amortisation. Accordingly, AFME members agree that specifying that this trigger is to be measured by reference to the original portfolio size is appropriate. One important point of clarification should be made, however. The reference to "the time of origination" should be replaced with a reference to the closing of the transaction (ie, the time the securities are issued or, for a securitisation which does not involve the issuing of securities, the time the initial securitisation positions are created). This is to avoid any confusion that "origination" refers to the time of origination of the underlying exposures themselves.
The rules should also contemplate the possibility of transactions where the underlying portfolio is rampedup over a period of time after the initial closing date of the transaction. In such a case, the trigger should be measured by reference to the lesser of the maximum securitised portfolio amount on the closing date and the gross amount of the securitised exposures included in the portfolio (ie, prior to taking any repayments or disposals into account) as at the end of the ramp-up period.
Q2. Do you agree with the aim of Article 3 with regard to ensuring that the credit enhancement of the senior tranche does not fall below a certain threshold because of the non-sequential amortisation?
The triggers set out in Recommendation 2 to the SRT Report reflect the types of triggers which have beencommonly used in synthetic securitisations for several years. In contrast, and as noted above, the proposed additional backward-looking trigger set out in Article 3 of the Draft RTS is not included in the SRT Report, and it has rarely been used previously in synthetic securitisations in the market. Accordingly, including this trigger would be a change from the experience with current practice in the SRT market.
That said, if this trigger were to apply instead of the requirement to include one of the additional backwardlooking triggers set out in Recommendation 2 to the SRT Report (ie, that SRT can be achieved without one of those additional triggers), AFME members broadly agree with the principle underpinning this trigger.
As a point of clarification, it is not clear how the reference to "the outstanding balance of the pool of underlying exposures" in Article 256(1)–(2) of the CRR should be interpreted in the context of a synthetic securitisation. This is because in a synthetic securitisation, the tranches would usually be written down by initial loss amounts (as, indeed, is impliedly required by Article 26e(2) of the EUSR) shortly after the occurrence of a credit event, without there necessarily being a corresponding reduction in the outstanding balance of the defaulted exposure. This therefore creates a mismatch where the sum of the outstanding balance of the tranches may be less than the actual outstanding balance of the securitised portfolio. This means that (i) the attachment point of the first loss tranche would become greater than zero, and (ii) the detachment point of the protected tranche would be artificially higher.5 While we acknowledge that the EBA's mandate under Article 26e(5) of the EUSR does not extend to clarifying the interpretation of Article 256(1)–(2) of the CRR, to avoid this anomaly causing issues for the application of the proposed trigger, it may be helpful to clarify that for the purpose of the calculation in Article 256(2) of the CRR and calculating the outstanding balance of the pool of underlying exposures, the outstanding balance of a defaulted exposure should be deemed to be equal to the protected amount minus the initial credit protection amount.
Q3. Do you agree with the trigger set out in the Article or would you prefer the alternative option in order to meet the aim of this additional backward-looking trigger? Please justify your answer, providing, if possible, evidence of the outcome of both triggers based upon your past experience.
Our comments on the alternative option for the additional backward-looking trigger are based on theformulation of that trigger as set out in Article 3 of the Draft RTS, and not the inconsistent description set out in Paragraph 13 of the Consultation Paper.
AFME members have a number of concerns with the alternatives option for the additional backwards-looking trigger. First, we have the same concerns outlined in the Introductory Comments and the response to Q2 arising from the fact that this trigger is not one of the ones already identified in the SRT Report. Secondly, the way this trigger is formulated, once some losses have been incurred, the likelihood of the trigger level being reached increases on an exponential basis purely as a result of scheduled amortisation of the portfolio. Unlike the option discussed in Q2, where the detachment point only decreases as a result of losses, because the alternative trigger calculates the losses as a percentage of the amortised protected tranche balance, that percentage will increase as amortisation increases. This is an undesirable outcome, as if the rate of losses is not increasing, there is no reason why expected amortisation should move the transaction closer to a switch to sequential amortisation. As such, the proposed trigger is not fit for purpose and does not meet the regulatory requirement of protecting the retained senior tranche in the event of rising defaults and losses.
Thirdly, because the trigger does not distinguish between EL and UL, or indeed between those potential losses which were taken into account when structuring the transaction and those which were not, the occurrence of any losses allocated to the protected tranche will eventually lead to the trigger being hit as the portfolio amortises, even if those losses are entirely consistent with base case modelling. Given that that modelling is an important part of the SRT assessment process, it is not appropriate for the triggers to be structured in a way which does not take that into account. Fourthly, AFME members agree with the EBA's observation in Paragraph 13 of the Consultation Paper that the effect of the alternative option will vary depending on the transaction structure. In particular, it will be much more restrictive for a first loss transaction than for a mezzanine transaction having the same detachment point, despite the fact that in such a scenario a first loss transaction provides greater risk transfer from the originator's perspective. That is, for a first loss transaction, all losses will contribute to the erosion of the protected tranches (and thus eventually lead to the trigger being hit), whereas for a mezzanine transaction, there would be no erosion of the protected tranche until the first loss tranche has been completely exhausted.
In the case of a mezzanine transaction, this also means there is a step-change in the impact of losses. Those which fall into the first loss tranche have no impact on the trigger, whereas losses which are allocated to the mezzanine tranche have a leveraged impact on the performance of the trigger. Given the efficiency benefits provided by pro-rata amortisation to the transaction overall, this may result in originators pursuing mezzanine transactions in place of first loss transactions where otherwise that may not have been the case. It would be an unfortunate outcome of the STS framework if it incentivised originators to pursue transactions involving less risk transfer for the purpose of being permitted to apply a lower risk-weight to the retained senior tranche.
Finally, because Article 26e(2) requires that interim credit protection payments are based on the higher of the accounting impairments and the loss given default under Chapter 3 of Title II of Part Three of the CRR, the general expectation is that these losses will be conservative (ie, higher than the realised credit protection amount at the end of the workout period). One effect of this is that the trigger will be activated earlier than it would be if based on realised losses alone, and indeed in circumstances where there is a reasonable expectation that once those interim losses are converted into realised losses, the trigger level would not have been reached. This later point supports the proposition that it is appropriate for it to be possible to revert back to non-sequential amortisation if a backward-looking trigger level (particularly this one) is reached. In this regard please see also our response on Q6, below.
In light of the above concerns, of the two backward-looking triggers proposed in the Draft RTS, AFME
members consider that the "alternative option" is not fit for purpose and will lead to undesirable outcomes by incentivising banks to pursue mezzanine transactions involving less risk transfer. In contrast, the EBA's "preferred" option does not suffer from these deficiencies and is workable in practice, notwithstanding the general concerns expressed in relation to the entire approach set out in the Introductory Comments, above.
Q4. Which level of the trigger would you consider more appropriate and why?
AFME Members do not consider that it is appropriate to specify in the RTS a single percentage as the level at which the trigger would apply for all transactions. There is great variety in the types of portfolios securitised using synthetic securitisation, and determining the correct level for the trigger requires careful analysis of the portfolio in question. For example, in a very high quality but relatively non-granular portfolio, it may be appropriate for the trigger to be set at a lower level (on the basis that very few losses are expected, and even a small increase above the expected amount of losses may be a cause for concern), while in a highly granular portfolio of lower quality exposures, a higher level of erosion of the protected tranche may be entirely consistent with expectations at closing, such that continuing with non-sequential amortisation is appropriate. This principle applies to both the preferred and the alternative additional backward-looking triggers.To return to the purpose of the performance-related triggers in the first place, the concern with pro-rata amortisation is that it will result in the protected tranches amortising away before losses crystalise, and thus before they can be borne by the protected tranches.6 The triggers to switch to sequential amortisation should, therefore, be calibrated in such a way as to avoid this outcome. By definition, however, that therefore means that the appropriate level for the trigger depends on the portfolio and transaction structure, and how it is expected to perform in the various scenarios which the originator is required to model as part of the SRT assessment process (see, eg., Section 3.3 of the SRT Report). The triggers should be designed such that they don’t result in a switch to sequential amortisation unnecessarily when the securitised portfolio performs within its expected range, but only when it starts to diverge negatively and materially from the expected case.
In light of this, AFME members strongly urge the EBA not to set the trigger on a "one size fits all" basis for all synthetic STS securitisations. While such an approach may be superficially attractive on the basis that it appears to be simple, it would undermine the very purpose of the trigger, and create significant inefficiencies in the SRT framework. It would also increase the risk of transactions being structured to fit the mandatory trigger level rather than providing a framework to encourage prudent transactions.
Further, while AFME members recognise that, as set out in Paragraph 104(b) of the EBA Report on the STS Framework for Synthetic Securitisation (EBA/Op/2020/07) (the "STS Report"), and unlike in the case of traditional STS securitisation, the criteria for synthetic STS securitisation are intended to provide adequate protection for the originator as well as investors, it is also important to ensure that nothing in the STS framework cuts across the SRT assessment process. That is, while having STS criteria which are consistent with the requirements for SRT does no harm, the STS framework is not the place to "gold plate" the SRT rulebook, or establish additional requirements that are not driven by the need to provide appropriate protection for investors, as is the case here given that the "protection" afforded by the performance-related triggers is primarily for the senior tranche retained by the originator.
Q5. Do you agree with the specification of the forward-looking trigger in Article 4? In your view, will the possibility of switching back to non-sequential, as set out in paragraph 6, be detrimental for the simplicity of the specific transaction and the objective of standardisation of STS on-balance-sheet securitisations?
The forward-looking trigger set out in the Draft RTS is somewhat difficult to follow, due to some imperfections in the drafting. However, we understand that the trigger is intended to work as follows:(i) Where the originator applies the IRB Approach and estimates a regulatory PD for all exposures in the portfolio the trigger would occur if the exposure-weighted average PD of the portfolio compared to the corresponding value at the time of origination is greater than a given percentage (the "PD Approach"). What that percentage should be is not stated in the rule, and so is presumably for the originator to determine.
(ii) Where the originator does not estimate a regulatory PD for all exposures in the portfolio, then the trigger would occur if the ratio of the outstanding amount of underlying exposures assigned to higher credit risk buckets to the outstanding amount of the underlying portfolio is greater than a given percentage (the "Risk Bucket Approach"). Again, what that percentage should be is not specified, and so is presumably for the originator to determine.
For the purpose of applying the Risk Bucket Approach, it is therefore necessary to determine which exposures should be allocated to the higher risk buckets. Article 4(4) of the Draft RTS attempts to do this, and again draws a distinction between where the originator applies the IRB Approach and where it applies the Standardised Approach as follows:
(A) Where the originator applies the IRB Approach to the exposures, the originator should use the rating grades for the relevant type of asset under Articles 153(5), 170(1) and 170(3), as applicable, and assign each exposure to the appropriate rating grade. However, the rule does not specify which rating grades are to constitute the "higher credit risk buckets" for the purpose of the trigger.
(B) Where the originator does not apply the IRB Approach to the exposures, it should use the differentiation of credit risk as recorded by it in its financial statements in accordance with the applicable accounting framework. No further guidance is provided on how this is to occur.
(C) Regardless of whether (A) or (B) applies, all exposures classified as in default within the meaning of Article 178(1) of the CRR, all exposures to a credit-impaired debtor and all other exposures entailing higher credit risk shall be assigned to the higher credit risk buckets.
AFME members have many concerns with this proposed trigger. First, as with the proposed additional backwards-looking triggers, it does not sit neatly with the forward-looking triggers set out in the SRT Report, although, at least where the PD Approach applies, it is similar to forward-looking trigger (i) from Recommendation 2 of the SRT Report. However, whereas the SRT Report expressly refers to the "weighted average 1-year PD of the underlying portfolio", the Draft RTS refers simply to the "exposure weighted average PD of the underlying portfolio". It is not clear if this is intended to be different.7
Secondly, it is not actually clear whether the PD Approach or the Risk Bucket Approach is intended to apply where the originator does estimate a PD for all exposures in the portfolio, but they fall into different asset types (eg., retail and non-retail). Given that the trigger works very differently depending on which approach is followed, this lack of clarity should be rectified.
We also note that the two forms of the trigger work quite differently in practice. In the case of the PD Approach, by being based on the weighted average PD of all securitised exposures, whether or not the trigger is hit depends on the performance of the portfolio as a whole, and a deterioration in the PD of some exposures may be offset in this regard by an improvement in the PD of other exposures. In contrast, the Risk Based Approach looks only to the proportion of the portfolio which is allocated to the higher credit risk buckets, and thus applies a binary test which ignores the relative performance of individual exposures allocated to the lower credit risk buckets.
Following from this, there does not appear to be any good reason for preventing the originator from using the Risk Bucket Approach, even if it does estimate the PD for all exposures in the underlying portfolio. Such a trigger would be in line with market practice and is an approach which aligns with how many investors approach their analysis of the originator's risk framework. Similarly, where the originator does not calculate a regulatory PD for all exposures, there does not appear to be any good reason for not allowing the originator to base the trigger on the weighted average accounting probability of default. As discussed in the Introductory Comments, there is nothing in Article 26c(5) of the EUSR which specifies that only one additional backward-looking trigger and one forward-looking trigger can be specified in the RTS. In light of that, AFME members remain of the view that providing a menu of triggers from which the originator can select the most appropriate triggers for a given transaction would be a much better approach.
Thirdly, as also noted above, it is not at all clear how the credit risk grades are to be assigned where the originator is required to use those recorded in its financial statements. Presumably it is not sufficiently granular simply to refer to IFRS Stage 1, 2 and 3, and it should be clear that the originator can make such allocation based on other factors such as IFRS 9 PDs or rating buckets. It would also be appropriate to permit the originator to use other metrics, such as internal ratings, to determine the credit risk grades, as is indeed common in many existing transactions in the market, and this should be stated explicitly in the RTS. The current reference to the applicable accounting framework in the Draft RTS is simply too vague to be workable in practice.
Fourthly, a change in PD (or allocation of exposures to credit risk buckets) may be driven by factors other than an actual change in the performance of the securitised portfolio. That is, a change in the rating model or various add-ons introduced or removed by the regulator following a model review may lead to a change in the exposure-weighted PD of the portfolio, even though there is no observed increase in the rate of defaults. Given that such changes would generally not be caught by a regulatory call,8 it would be disadvantageous for the originator if one outcome of such a review was to trigger a switch to sequential amortisation for a portfolio that is actually performing as expected. This also means that the output of the trigger may be unstable, as it would depend on what add-ons may or may not be applied at different times over the life of the transaction.
In light of the above, AFME members feel that this trigger is somewhat over-engineered, and runs the risk that it could have unintended consequence in certain cases. We invite the EBA to consider whether the trigger could be simplified and made more flexible so as to require a forward-looking trigger based on either an increase in an exposure-weighted credit risk metric or an increase in proportion of exposures categorised in a higher credit risk bucket above a predefined threshold, where such metric or buckets are based on the relevant regulatory or accounting framework, or internal credit risk metrics used by the originator.
AFME Members do welcome the fact that the actual trigger levels are not specified for either the PD Approach or the Risk Bucket Approach, which ensures that the originator is able to calibrate the triggers at a level which is appropriate for the transaction, and also avoid creating tension with the SRT assessment process.
AFME members also welcome the proposal in Article 4(6) of the Draft RTS that if, following the activation of this trigger, the exposure-weighted PD or percentage of exposures allocated to higher credit risk buckets falls back below the threshold, the transaction should be permitted to switch back to non-sequential amortisation. However, it should be clarified that this only requires the level to fall below the threshold, and not to return to the level at the time of closing of the transaction. Further, AFME members consider that the requirement that the trigger be cured for four consecutive quarters is overly conservative, and that a shorter period should be sufficient.
In addition to the concerns outlined above, AFME members have a number of additional technical comments in relation to the proposed trigger:
(i) First, defaulted exposures should be excluded for the purpose of the forward-looking test on basis that
the default has already occurred, and the effect of that default should be captured in the backward-looking triggers. A forward-looking trigger should only take into account factors where there is a degree of uncertainty as to the outcome. Including the 100% PD for defaulted exposures in this calculation would also significantly skew the overall calculation and would require the trigger to be set at a much higher level to avoid it being hit solely as a result of an expected level of defaults, thereby reducing the sensitivity of the trigger to unexpected deterioration in the overall portfolio.
(ii) Secondly, it should be clarified that the reference to "exposure-weighted" should be interpreted as requiring weighting by the protected amount in the securitisation. It is quite common for synthetic securitisations to reference only a part of a larger exposure, and it would therefore not be appropriate to use the overall exposure amount for this calculation.
(iii) Thirdly, it should be clarified what is to happen if the PD Approach applies initially, but the originator subsequently ceases to estimate a regulatory PD for all exposures in the portfolio, as has been the case in some previous transactions in the market.
(iv) Fourthly, the requirement in Article 4(5) of the Draft RTS that all exposures "entailing higher credit risk" shall be assigned to the higher risk buckets appears to ride roughshod over the method for allocating exposures to buckets. What exactly is meant by exposures "entailing higher credit risk" that is not already covered by the methodology in Article 4(4)? Indeed it is not clear that Article 4(5) as a whole serves any purpose that is not already covered by Article 4(4).
(v) Fifthly, for the same reasons set out in our response to Q2, above, the reference to "time of origination" in Articles 4(1) and 4(2) of the Draft RTS should refer to the closing of the securitisation.
Q6. According to market practice, is it common that performance-related triggers can change several times the amortisation system of the tranches throughout the life of a synthetic securitisation? If so in your view, please provide concrete examples of triggers, distinguishing between backward-looking and forward-looking triggers.
For many performance-related triggers, it is possible that after the trigger has been activated, the relevant metrics will fall back below the relevant threshold. This can be for a variety of reasons. For example, in the case of triggers based on cumulative losses, this could be because the trigger level was initially reached as a result of initial losses which end up being higher than the realised final losses at the end of the workout process. In the case of triggers based on cumulative unmatured defaults, this would be because once those defaults are worked-out, they cease to be unmatured. In the case of triggers based on probability of default or expected losses, a trigger could be cured by an improvement in the quality of the securitised portfolio. In other cases, such as where the trigger is based on the detachment point of the protected tranche, the trigger could self-cure after a period of sequential amortisation, which has the effect of increasing the detachment point back above the relevant level.At the same time, there are some types of triggers which cannot be cured. Examples include triggers based on the cumulative final losses or cumulative gross defaults measured as a percentage of the initial portfolio amount.
Despite the fact that many performance-related triggers are capable of being cured, it is relatively rare for synthetic securitisations in the market to permit a switch back to non-sequential amortisation where that occurs. In the majority of cases, once a trigger is activated, the switch to sequential amortisation is permanent. Of course this makes the calibration of the trigger levels all the more important, and also means that the originator is wary of setting the level for certain types of trigger too low. For example, in a portfolio which may be expected to experience many defaults but very low losses, a trigger which is based on the cumulative unmatured defaults would need to be set quite high to avoid the risk that sequential amortisation would be triggered even though actual losses are quite low. In contrast, if it were possible to switch back to non-sequential amortisation where the trigger falls below the relevant threshold, the implications of activating the trigger would be less, meaning that it may be possible to set the trigger at a lower level in the first place.
As noted in our response to Q5, above, AFME members are in favour of the proposal to allow a switch back to non-sequential amortisation if the forward-looking trigger is cured, albeit that we consider that the requirement for the trigger to be cured for four consecutive periods is overly conservative. This is particularly the case given that the performance of the forward-looking trigger may be affected by short term market conditions which resolve themselves without actually leading to a significant increase in defaults. A good example of this is the experience during the Covid-19 pandemic, when many portfolios experienced a significant increase in weighted average PDs or significant downward migration of exposures into higher credit risk grades, which has subsequently been largely reversed without having experienced significantly increased defaults.
AFME members also consider that it would be appropriate to permit a switch back to non-sequential amortisation if the backward-looking triggers are cured (where that is possible given the nature of the trigger). Indeed, this is perhaps even more appropriate for the backward-looking triggers than the forward-looking trigger as a backward-looking trigger will only cure once the circumstances giving rise to the trigger have been fully taken into account by the tranches, such that a "reset" of the amortisation mechanics going forward can proceed with certainty. In contrast, the forward-looking triggers always involve a degree of uncertainty and a reduction in the trigger metrics below the relevant threshold is obviously not a guarantee that the future performance of the remaining exposures could not cause the trigger level to be exceeded again in the future.
In terms of timing, AFME members consider that it should be sufficient for the trigger to be cured for two consecutive quarters.
Q7. Do you agree that the information that the originator shall provide under Articles 7 and 26d of the Securitisation Regulation includes the information needed by the investor providing protection to understand and verify the functioning of the performance-related triggers in an STS on-balance-sheet securitisation?
Yes. AFME members are of the view that all the information which investors require to monitor and verify the performance-related triggers is provided by the existing Article 7 and 26d reporting frameworks in the EUSR. No additional reporting should be required.As noted in our response to Q5, above, we also consider that it should not be necessary for the originator to describe in detail the process by which exposures are allocated to credit risk buckets for the purposes of the Risk Bucket Approach.
Q8. Since as a first step before specifying the triggers above, the EBA reassessed the triggers included in recommendation 2 on Amortization Structure of the EBA 2020 Report on significant risk transfer in securitisation (see Section 5.2), and some elements from them were taken on board in the draft RTS, stakeholders are also invited to comment on the suitability of other triggers included in that recommendation for the purpose of these draft RTS.
Please see Introductory Comments and responses to the other questions above.When the SRT Report was published in November 2020, the menu of backward and forward-looking triggers set out in Recommendation 2 largely reflected the types of triggers which had been common for many years in synthetic securitisations across the EU. In contrast, the proposals in the Draft RTS, particularly in relation to the additional backwards-looking trigger, appear to reflect a change of thinking from the EBA as to what are the appropriate triggers for this purpose. If that is the case, then it would be very helpful if the EBA could provide further clarity on its expectations in this regard as far as the SRT assessment process is concerned. It is clearly not appropriate to have two sets of overlapping but different requirements that apply for synthetic STS securitisations as apply for synthetic SRT securitisations, particularly given that virtually all synthetic STS securitisations are also SRT securitisations.
As to the substance, and as set out above, AFME members consider that it would be more appropriate for the RTS to provide a menu of backward and forward-looking triggers from which the originator can select the most appropriate triggers for a given transaction, a selection which will form part of the SRT assessment process.
Q9. Do you have any other comments on these draft RTS?
The RTS should include an explicit grandfathering provision which provides that they only apply to transactions executed after the RTS enter into force, and that synthetic securitisations which closed prior to that date can still achieve STS status if they meet the requirements of Article 26c(5) on the basis of a plain reading of those requirements in the absence of the RTS. Such grandfathering should apply for the remaining life of those existing transactions.Further, there should also be an explicit acknowledgement in the RTS that the originator is not prevented from including additional performance-related triggers beyond whatever is prescribed in the RTS. While this may be impliedly the case given the reference to "at a minimum" in Article 26c(5), given the importance of being able to include such triggers to satisfy the SRT assessment process, it is appropriate to clarify that doing so does not fail to meet the "standardisation" requirements of the STS framework.