Response to discussion on the Significant Risk Transfer in Securitisation
Go back
The survey provides detailed information on number of trades, distributed tranches total size, pools total size, attachment and detachment points as well as geographic breakdown of total pool.
Similar to what is reported in the discussion paper, 2015 was the peak year in synthetic transactions but volume has remained steady in the subsequent years. The detailed survey information will be submitted in a separate response given the sensitivity of the information.
That said, the data generally appear to be accurate as far as they go but they are clearly incomplete as overall volumes are clearly much too low. In addition, members are aware of at least one synthetic securitisation in excess of EUR 8bn whereas the EBA data show the largest transaction at circa EUR 5bn.
We also observe that there are relevant ways in which the EBA did not break down the data. These include:
- The percentage of SRT traditional securitisations which were publicly offered vs. private placements.
- The percentage of SRT synthetic transactions which were public or privately issued CLNs, although it is acknowledged that this distinction is sometimes difficult to draw in the case of some synthetic transactions.
- The percentage of synthetic transactions which were structured as private/bilateral guarantees or CDSs without an associated issuance of notes.
Finally, we would note that the prevalence of calls as described in paragraph 50 of the Discussion Paper represents a considerable underestimate as compared to the experience of the members of the Joint Associations. In our experience, both cleanup and regulatory calls are included in the vast majority of transactions; we would have expected a prevalence at or near 100% for these types of call options.
This can be a very useful exercise because it acts as something of a checklist of the requirements for achieving SRT, but it is also obviously a very burdensome exercise and much more involved than the data submission requirements of other competent authorities. Members of the Joint Associations are of the view that the harmonisation of both the process and the form of notification (including the required level of detail) would be helpful in creating a level playing field. This, however, should take account of the type of transaction being notified and the institution's history. The EBA should consider a detailed notification template as standard but with the option to simplify the process where the transaction would not lead to a material RWEA reduction, or the transaction is of a type previously approved by the relevant competent authority in respect of that institution and there are no material structural differences. See our further answers below, especially our answer to Question 3.
Finally, the EBA has noted that very few deals have sought approval under Art. 243(4) or 244(4). In the experience of certain AFME members, this is likely to be because competent authorities are reluctant to commit themselves to analysing deals which do not meet the mechanistic, quantitative tests and providing timely approval. At least one member has waited an average of 11 months to receive a response on any SRT requests submitted pursuant to these provisions.
While we agree in general with the spirit of the proposals, there are several considerations members of the Joint Associations think should be taken into account in further refining the proposals, as follows:
- Timing of notification: The economics of some securitisation transactions will only be viable if SRT can be achieved. In respect of these transactions – which are predominantly but not exclusively synthetic securitisations – a framework that allows for ex ante notification, with feedback from the competent authority on SRT prior to closing, would be extremely helpful. We also view one month as a reasonable timeframe for this notification. It is, notably, significantly more workable than the current requirement for notification three months prior to closing imposed by the ECB. Obviously, information about the transaction may be incomplete at this stage as contractual terms and the structure continue to evolve and are negotiated right up to the closing date. See further discussion below under the heading Contents of ex ante notification".
That said, many transactions that might be capable of achieving SRT are nevertheless not done for that specific or exclusive purpose. There is no reason why a claim for SRT treatment should not be able to be made after closing of the transaction where the originator so chooses. This fact pattern is particularly likely to arise in respect of traditional securitisations, where SRT is most likely to be a secondary purpose of the transaction. On a traditional securitisation, the issue of SRT could also become relevant months or years after the closing of a traditional securitisation if part of the capital stack that was not sold to a third party investor at closing is later transferred.
If part of the purpose of this review of SRT is to increase the number of SRT traditional securitisations, it is essential to preserve the flexibility of originators to seek SRT treatment for an existing transaction after its closing.
In this respect, the focus should be on when the reduction in RWEA from an SRT transaction is recognised by the institution in its prudential capital calculations, which would normally follow the supervisor's final feedback. For this reason, a reasonable and predictable timeframe for receiving that feedback is an essential element of an efficient supervisory regime for securitisations.
- Contents of ex ante notification: While members of the Joint Associations generally support the idea of ex ante notification where appropriate, we feel it is important to clarify what documentation can realistically be made available as early as one month before expected closing/issuance. Members of the Joint Associations would be confident supplying competent authorities with a summary of the intention, key features and indicative size of the transaction, possibly in the form of a detailed term sheet. That said, there should be no expectation that the transaction size, pricing or other commercial terms are final since commercial terms may be subject to negotiation right up to the closing date. For the same reason, members of the Joint Associations are of the view that it is not constructive to provide draft deal documentation since this is highly likely to be negotiated right up until the closing date. Obviously, it will be impossible to provide final transaction documentation until after closing, as acknowledged in the Discussion Paper. Further comments on the proposed deadline of 15 days post-closing to provide final documentation appears below under the heading "Timing of final documentation and information".
Further, pre-notification should not be seen as committing the relevant institution to the transaction, as that would hamper its ability to carry out proper commercial negotiations with its prospective investor(s).
Finally, to the extent that transactions do not currently need to be notified (e.g. where the full deduction option is applied) a simplified notification regime should also be applied, with the institution explicitly able to assume SRT treatment and associated RWEA reduction in the absence of an objection by the competent authority within a prescribed timeframe (e.g. three months). The EBA might also consider a simplified/streamlined SRT process for small portfolios, where the time, cost and effort of completing the full SRT process would outweigh any benefits to the institution of doing the transaction. This might be appropriate where such transactions in aggregate do not exceed e.g. 0.1% of the total assets of the bank.
- Informal feedback from the competent authority: The Discussion Paper highlights at paragraph 70 that some market participants had specifically emphasised the advantages of receiving feedback ahead of the closing of the SRT transaction. The Joint Associations strongly agree that feedback prior to the closing of the transaction is extremely useful and helps to make the market much more efficient, especially if the transaction is novel or has features not previously reviewed by the competent authority in respect of the notifying institution. As it stands right now, however, a number of competent authorities do not provide this. Where transactions are being structured specifically to achieve SRT, significant transaction costs can be saved if feedback is obtained prior to closing. In addition, informal pre-closing feedback would help banks to manage reputational risk with investors by avoiding the need to call transactions prior to maturity because SRT is not achieved as expected.
Pre-closing informal feedback should highlight, at a minimum, any features of the transaction that the competent authority considers might hamper the achievement of SRT and the type of changes or analysis and information the competent authority will require in order to address the concerns they have highlighted. In the view of the Joint Associations, the one month ex ante notification should be the beginning of a conversation during which the originator and the competent authority exchange views about the transaction. The goal should be to arrive at a position of sufficient certainty in advance of closing that the originator can complete the transaction with a high degree of confidence that SRT will be achieved. Of course, to the extent that the originator has omitted information from the SRT notification, or the final transaction documentation varies materially from what has been described in the notification, it would be entirely appropriate for the competent authority to reassess SRT treatment.
- Nature and consequences of ongoing assessment: The Joint Associations are of the view that SRT should be assessed when applying for SRT only and should not require reassessment over the life of the transaction save in the case of implicit support, "step-in" or if there is a significant amendment or restructuring of the deal. To be clear, this should not prevent the bank from holding sufficient capital to reflect the risk of the positions it continues to hold, which may increase as the risk-weights of those positions increase. This approach is intended to reflect the fact that achieving "significant risk transfer" in fact merely means that the basis for calculating capital charges changes from calculating capital charges on the underlying portfolio to calculating capital charges using the securitisation framework. All relevant structural features of a transaction are known at the outset and should be captured in the documentation. It is entirely reasonable for the competent authority to wish to ensure that significant risk is transferred for a sufficient period of time, or that the first loss or mezzanine tranches are sufficiently thick to effectively transfer risk, but those concerns should be raised when the transaction is notified, rather than being subject to reassessment partway through the life of the deal.
Where SRT is initially achieved, the absorption of losses by the protection tranche(s) is expected and intended. The fact that losses materialise and are absorbed by the protected tranche(s) (thereby reducing its/their thickness) does not therefore mean that that risk was never initially transferred or that SRT treatment should be revisited. Rather, the absorption of actual losses by the protected tranche should result in a reassessment of the risk-weights of the retained tranches held by the originator. Provided that the originator adjusts its RWEA associated with the retained tranche(s) on an ongoing basis to reflect the then-current protection provided by the transaction, the basis of calculating its RWEAs (portfolio basis vs. securitisation framework) should not need to be reassessed.
One way in which this could be achieved, and (subject to our comments below on the specific proposals for commensurateness tests in the Discussion Paper) perhaps the most sensible application of a commensurateness test, would be to apply the commensurateness test essentially as a cap on the capital relief that can be achieved via a given SRT transaction. So, for example, if according to the commensurateness test eventually adopted 50% of the total expected losses and regulatory unexpected losses on the underlying portfolio had been transferred to investors, the effect of the commensurateness test would be to cap the capital relief resulting from that transaction at 50% of the capital that would otherwise have to be held. In this way, the cliff effects of the binary nature of SRT are alleviated while still achieving the regulatory objective of ensuring banks hold sufficient capital to reflect the risks of the positions to which they are exposed.
- Fixed timeframes for competent authority feedback: The Discussion Paper proposes specific timeframes and deadlines for originator submissions to competent authorities (i.e. notification one month before expected issuance, final documentation 15 days after closing). Subject to our comments elsewhere in this response, we consider that introducing fixed timeframes is helpful to promote predictability of the process. However, these timelines should also be realistic and linked to the final purpose, which is determination of when the RWEA effects of an SRT transaction should be recognised by the institution in its prudential capital calculations. In order to achieve the hoped-for predictability, it is necessary that all parts of the process should operate within predictable and consistent timeframes. To that end, the Joint Associations would urge the EBA to set out a defined timeframe for competent authority feedback on submissions, rather than simply saying it "should be provided within a reasonable timeframe after the submission of the final version of all information/documents". Where an ex ante notification process is followed, we have mentioned above that we believe the period between notification and closing should be used for the originator and competent authority to exchange views on the merits of the transaction with the aim of providing the originator with reasonable certainty as to the transaction's ability to achieve SRT. Obviously any views expressed by the competent authority would be subject to the final transaction documentation reflecting the information provided by the originator prior to closing. Of course, a final and binding determination would only be possible in the light of the final transaction documentation and this should be provided no later than three months following receipt by the competent authority of the final documentation needed for it to assess the claim of SRT.
- Timing of final documentation and information: Members of the Joint Associations are concerned that meeting the deadline of 15 days after closing to provide "a final version of all information/documentation" will be difficult to achieve. If all that is required at that stage is the final transaction documentation, this will of course not be problematic. Once the transaction has closed, 15 days is sufficient to provide the final executed documents. If, however, further extensive information and analysis is required by competent authorities (e.g. due to structural changes resulting from informal feedback between ex ante notification and closing or changes to the transaction not yet discussed with the relevant competent authority), this timeframe may be too ambitious for banks to meet. There would also be concerns around producing the SRT self-assessment proposed in the Discussion Paper within this timeframe. Based on current competent authority requirements for transaction notifications, some members of the Joint Associations have expressed the view that 30 days post-closing would be a more appropriate timeframe within which to provide final documentation and information.
The Joint Associations would be pleased to engage further with the EBA and national competent authorities on this subject once the form of SRT notification going forward (along with the details of any self-assessment or similar exercise) becomes clearer. We are confident that it will be possible to create a process that provides for timely delivery of relevant information in a timeframe that is manageable for market participants.
- Notification for time call exercises: In this respect, please see our answer to Question 9."
In addition, members of the Joint Associations support the development of separate templates for traditional and synthetic securitisation. These two types of transaction are materially different to one another and we believe that the development of separate templates for each type of transaction would help to reduce confusion and streamline the review process.
The contents of the ex ante notification should be as discussed in our response to Question 3. We would suggest that the existing ECB questionnaire for SRT notification is a good template to start from for the final submission to be made after closing of the transaction. Subject to our comments above in relation to simplified notification in some cases, this should be accompanied by a report demonstrating compliance with the relevant CRR articles and the SRT Guidelines issued by the EBA (as amended following the current re-examination). In addition, where any of the relevant structural features are present, a risk transfer self-assessment should also be included.
That said, to the extent ongoing monitoring is deemed necessary, members of the Joint Associations would strongly encourage the EBA to incorporate SRT monitoring into the COREP framework and cover it as a part of more general ongoing regulatory and economic capital monitoring. As much as possible this should be accomplished with standard investor reporting data. This would help to minimise the operational risk that would come with the need for ad hoc reporting. A new full assessment of SRT should only be necessary in the case of significant restructuring or amendment of the transaction as mentioned above in our answer to Question 3.
- General capital treatment: As a general matter, members of the Joint Associations believe that the SRT and capital rules actively discourage the use of securitisation – and in particular the use of securitisation to achieve capital relief. As a starting point, senior tranche risk weights are set (including the built-in non-neutrality of securitisation risk weights) to account for the fact that risk may not have been transferred as effectively as expected. This is despite the fact that, in order to achieve SRT, transactions have to be structured in such a way as to ensure that senior tranches will not suffer losses even under extreme stress scenarios. Then, where the portfolio does suffer a stress scenario, SRT is re-examined and sometimes lost, leading to sharp capital increases for the institution in question because capital must once again be held against the whole portfolio once SRT treatment is lost. We would encourage the EBA to look at these questions in a holistic way and consider whether existing rules unduly discourage the use of risk sharing transactions as a capital management tool for banks. Our comments above in response to Question 3 (and in particular under the heading nature and consequences of ongoing assessment) suggest an alternative approach that may reduce cliff effects and produce more sensible outcomes where protected tranches become thinner as a result of losses being realised during the life of the transaction.
Another way of looking at this would be to acknowledge that a principal effect of achieving SRT is to allow banks to calculate their capital requirement using the securitisation framework rather than calculating it on the basis that they continue to be exposed to the full risk of the underlying portfolio. Accordingly, banks could be permitted to continue calculating their RWEAs on the basis of the securitisation framework (adjusted for risk on an ongoing basis) throughout the life of the transaction, regardless of whether they continue to meet the initial SRT tests. SRT having been achieved at the outset of the transaction, one might for example incur progressively increased capital weights on the retained tranches as losses are incurred on the underlying portfolio and protection is eroded. This would avoid artificial cliff effects and more closely match capital requirements to actual risks held by banks. What is more, the non-neutrality of the securitisation framework would ensure that banks would never hold insufficient capital as a result of this approach. Indeed, this is already achieved by the new securitisation framework (see general introductory points above).
- Credit events: The section on credit events and their definitions set out in the Discussion Paper is somewhat confusing. Members of the Joint Associations are of the view that the focus on Article 178 (which is based on likelihood of a loss being suffered) is incorrect, given that the effect of the SRT is that the originator is protected from suffering a loss should the underlying loan actually default, which is appropriately covered by Articles 215 and 216. Despite this, the recommendation focuses closely on the definitions of the various credit events set out in Article 178. The discussion of implications for SRT, on the other hand, mentions compliance with the requirements provided for in Part Three, Title II, Chapter 4 of the CRR. There is no mention of the possible conflicts between Article 178 and e.g. Article 215 and 216. To take the example of failure to pay, Article 178(1)(b) deems a default to have occurred when a payment is more than 90 days past due, but Article 216(1)(a)(i) regards there to have been a failure to pay "under the terms that are in effect at the time of such failure, with a grace period that is equal to or shorter than the grace period in the underlying obligation". A failure to pay within the meaning of Article 216(1)(a)(i) could therefore occur either before or after the obligation goes into default under Article 178(1)(b), depending on the terms of the obligation and any grace period it provides for. Similarly, the definition of restructuring in Article 216 is different from that set out in Article 178. Of course, if originators wish to define their credit events by reference to Article 178, that should also be permitted, but the effects of the transaction within the credit risk mitigation ("CRM") framework would continue to be governed by Part Three, Title II, Chapter 4 of the CRR.
Furthermore the proposal to require three credit events for synthetic securitisations to achieve SRT contradicts the credit risk mitigation rules under CRR Articles 213 and 215 CRR, as "restructuring" is not currently required for financial guarantees. As such, it is unclear how financial guarantees could qualify for SRT under this proposal as outlined in the EBA Discussion Paper.
We note, also, that the relevant provisions of the SRT requirements refer to the CRM framework in the CRR which are common to both synthetic securitisations and non-securitisation unfunded CRM. Any guidance on this point therefore has the potential to affect a large market beyond just securitisation SRT transactions and should be expressed not to affect non-securitisation unfunded CRM.
- Cost of credit protection: Members of the Joint Associations have no objection in principle to explaining their approach to price discovery and price appetite to competent authorities. However, we have the following comments about the factors to take into account when assessing the effect of this on SRT:
o As the market for SRT transactions operates largely on a private basis and because there is wide variation in asset type, obligor credit quality and originator, it is not straightforward to establish market comparables for the cost of first loss or mezzanine protection. Similarly, these features make it inappropriate to use credit indices or other market benchmarks as a basis for pricing of SRT transactions.
o Where premia are contingent, competent authorities should not disallow or haircut SRT simply because the lifetime premia paid are expected to exceed the lifetime EL. If anything, it would be surprising if the premia did not exceed the lifetime EL (assuming the protection completely covers tranches that map to the EL of the underlying portfolio) as the investor providing protection on the protected tranche expects to make a positive return on the transaction. Market pricing would therefore reflect the expectation that lifetime EL will be covered by the present value of the premia. Where the detachment point of the protected tranche is above the level of lifetime EL, the premia also need to account for the protection covering some or all of the UL against which the bank would otherwise set aside capital.
The proposed approach also ignores the time value of money. Any comparison of premium costs to portfolio characteristics should discount the future premium payments by a fair market rate. There is support in the high cost credit protection guidance issued by the Basel Committee in December 2011 for using the present value of premia as the appropriate comparator, rather than the absolute value. Ideally the EBA would issue clear guidance for how to determine a fair market rate and the appropriate curve to use, in order to take account of the impact of loss timing. This is particularly important for long-dated assets such as project finance or commercial real estate assets, where premia paid over the life of the deal will always exceed the amount of the protected tranche. In any case, if the originator demonstrates that they have conducted a fair auction with investors (or otherwise agreed a price on an arm's length basis) before agreeing a premium, it should be considered a fair market rate and it should not be a barrier to the achievement of (or a reason to haircut) SRT.
o Finally, comparing the cost of protection against income generated by the portfolio is misleading and should not be a barrier to achieving SRT and should not lead to haircuts from any SRT achieved. Looking at the income generated on particular loan products that may be in the portfolio is misleading because the interest rate on the loan is often only one aspect of the total return earned by the bank from the customer (especially true of corporate customers) on a relationship wide basis. In addition, any meaningful comparison would need to take account of other fees on services provided to such customers, because the all-in return may then become commercially reasonable. The reasonable cost of protection on the portfolio would then rise accordingly. There is no reason, though that this approach to corporate relationships should be a barrier to achieving SRT on the loans. In addition, the required rate on a loan is set at the time of origination of the loan. This could differ significantly from what the bank would require at the point in time when the risk sharing transaction is entered into. Market circumstances, cost of financial resources, and the creditworthiness of the borrower may all have changed in the interim. Once again, if the originator can demonstrate that they have conducted a fair auction with investors (or otherwise agreed a price on an arm's length basis) before agreeing a premium on the risk sharing transaction, it should not be a barrier to the achievement of (or a reason to haircut) SRT.
o Additionally, a pure comparison of premia and EL ignores the inherent differential in EL volatility and skewness of certain asset classes versus others. For example, although series 1 and series 2 below have the same EL (2.5%), their distributions are materially different, hence using a linear approach of premia vs. EL significantly understates the inherent protection value of guarantee 2 over guarantee 1. If premia are to be compared against EL (a proposal we think is unhelpful in principle, as described above) the EL should be adjusted for this purpose with an appropriate empirically derived scalar to reflect the volatility and skewness for different asset classes. Members of the Joint Associations would welcome the opportunity to work with the EBA to derive such appropriate scalars."
The members of the Joint Associations believe in general that requiring all four triggers in pro rata amortisation transactions seems excessive. There should be some triggers to convert to sequential amortisation but these should take account of the impact of the RWEA calculation on the senior tranche. If all four triggers are required as a general matter, market participants need flexibility to adapt or omit some triggers where they are not applicable to the particular asset type or transaction (e.g. granularity trigger would be inappropriate in a highly granular portfolio).
That said, members of the Joint Associations have a diversity of views about the appropriate approach to this. These include:
- Agreeing appropriate triggers on a transaction by transaction basis with their competent authority.
- Combining triggers I and II by including non-matured defaults as part of cumulative losses" for the purposes of I.
- Having a "menu" of triggers mapped to transaction features and asset classes to which they would be appropriate.
- Developing two relatively straightforward triggers, one of which is forward looking and the other of which is backward looking, to address the concerns around timing of losses and underestimating actual losses.
- Having a set of triggers to switch from pro rata to sequential amortisation, but also having triggers that allow switching back to pro rata amortisation (e.g. where the portfolio credit quality improves again over time or simply where the triggers for switching to sequential amortisation cease to be met).
- Eliminating the idea of binary triggers and instead allowing hybrid amortisation moving on a sliding basis from 100% pro rata to 100% sequential depending on the state of the figures used as triggers.
In any case, members of the Joint Associations consider it essential that market participants have flexibility to set the levels of the triggers in such a way as to take account of asset class constraints and market constraints, such as demands made by investors or by rating agencies in order to achieve appropriate ratings for the transaction where relevant."
Originators include time calls in securitisations for a wide variety of reasons and trigger them for a variety of reasons as well. Time calls permit banks to recycle securitised assets and use them for new transactions (especially relevant for banks with smaller volumes of eligible assets), allow banks to benefit from improved market conditions and reduce the cost of credit protection and provide them with a backstop in case the transaction becomes too expensive/uneconomic. This, in turn, may be because the transaction has ceased to provide an RWEA reduction or the cost of the RWEA reduction has become uneconomic (e.g. because the credit quality of the pool has improved or the coupon demanded by the market has decreased).
While members of the Joint Associations understand the regulatory concerns raised by the exercise of calls during the financial crisis to prevent investors from suffering losses, a general prohibition on the use of time calls in traditional securitisations and a lengthy non-call period in synthetic securitisations is not a proportionate response as it deprives banks of the legitimate benefits time calls can provide that are unconnected with supporting investors. The correct test is whether exercising the call is a reasonable decision for the bank to make purely in its own commercial interests (which may include putting the same assets into a new, lower cost – or longer-dated – transaction). If so, then the exercise of the call ought to be allowed, regardless of the effect on investors.
We believe that time calls should be no barrier to the achievement of SRT provided that (i) there is no structural incentive to exercise the call, such as a step up after the call date (as demonstrated in the originator's self-assessment of SRT); and (ii) the competent authority is pre-notified of the intention to exercise the call by the originator along with a confirmation that the call is being exercised in pursuit of its own commercial interests (including a short explanation of what those interests are). Provided the competent authority does not object within a given period of time (say, one month) then the originator should be permitted to proceed.
It may also be worth pointing out, in this context, that a time call on a synthetic securitisation is not analogous to repurchasing the whole portfolio at par. When protection is terminated pursuant to a time call on a synthetic securitisation, the assets in respect of which a credit event has already occurred continue to be worked out, with the protection provider continuing to absorb any losses pursuant to the terms of the protection. Similar measures are put in place for traditional securitisations, where only performing assets are repurchased at par.
As to non-call periods for synthetic securitisations, members of the Joint Associations agree that the EBA has a legitimate concern to prevent SRT transactions undertaken for year-end window dressing" concerns as described at the EBA's Public Hearing on 17 November 2017. However, inserting lengthy non-call periods has the effect of preventing banks from exiting transactions once they cease to be economically advantageous, which serves no legitimate purpose. Instead of the proposed non-call period in the Discussion Paper, we would recommend a minimum one year non-call period. A one year non-call period would exclude the possibility of engaging in transactions just designed to mislead the market by improving the apparent financial condition of the bank at year-end and in any case, investors would generally not accept a non-call period shorter than one year. The EBA proposal for a lengthier non-call period appears arbitrary and does not appear rooted in any Basel guidance of which we are aware. Going forward, the regulatory framework should not make reference to the WAL of the asset pool, which we view as irrelevant to the length of any non-call period.
Finally, members of the Joint Associations would be grateful if the EBA could issue guidance clarifying that time or regulatory calls can be used to repurchase only certain tranches without having to unwind the whole transaction. In some cases, this may be the most sensible thing for a bank to do. For example, where a traditional securitisation has both an SRT and a funding purpose, it is possible that the SRT purpose may cease to be economically viable leading to an incentive to redeem the expensive junior notes while leaving the relatively cheap senior notes outstanding for funding purposes."
We understand that the proposal to treat excess spread as a securitisation position subject to 1250% risk weight/capital deduction is restricted to any portion of synthetic excess spread committed on a yearly basis to loss absorption which is funded and will not apply to the committed but unfunded portion of synthetic excess spread. The same would apply to traditional securitisation excess spread. It would not be appropriate to treat excess spread on a traditional securitisation as a securitisation position subject to 1250% risk weight/capital deduction unless it is trapped" in a reserve account or other structured designed to provide credit enhancement. Indeed, on a traditional securitisation where excess spread is "actual" excess spread, it is fully dependent on the performance of the securitised portfolio rather than being contractually set. It should therefore not attract any capital charge.
We would also be grateful for additional clarity in respect of the way in which "excess spread should be taken into account within the risk transfer self-assessment analysis" in respect of a traditional securitisation. Barring the situation in which the originator guarantees a minimum level of excess spread on the transaction, or if it is trapped as described above, it seems to members of the Joint Associations that excess spread at whatever levels should not raise SRT concerns on a traditional securitisation. In order to achieve an efficient and predictable market, it is necessary that market participants should have greater clarity about the way in which excess spread would be taken into account."
We agree that synthetic excess spread should be capped at the level of 1-year regulatory expected losses, measured on the reference portfolio at inception – it being impractical to adjust the excess spread amount from year to year.
Furthermore, the requirement to use the trapping mechanism is potentially problematic when combined with the requirement to account for the trapped excess spread as a retained first loss tranche. In a benign credit environment where actual losses are below EL, trapped synthetic excess spread will build up and SRT may well be lost as a direct result of this mechanism as the value of the protection achieved by selling the first loss and/or mezzanine tranches would be reduced as the amount of trapped excess spread increased.
Also, the requirement in paragraph 123(a)(II) that excess spread not absorbed by losses in a given year should remain trapped in the transaction "in the form of a funded reserve account" is confusing. It suggests that there ought to be a separate cash funded reserve account established by the originator, which raises the question of who benefits from that fund after termination. We suspect this is just a terminological issue as the expectation in a synthetic would normally be that synthetic excess spread would be deducted from the periodic losses before those losses begin to be allocated to tranches. Accordingly, we would appreciate if the EBA could clarify that this would more appropriately be characterised as a "ledger" than an "account".
To the extent that the trapped excess spread on a synthetic is recognised as income/an asset by the bank, it may appropriate for it to be treated as a securitisation position with a 1250% risk weight and/or subject to capital deduction, at the bank's option. However, this then needs to be reflected in a reduction in the RWEA in respect of the more senior tranches (by factoring in the effect of the excess spread into the attachment point of each tranche for the calculation of the risk-weight when using SEC-IRBA or SEC-SA). The benefit to the senior tranches is realised in the context of the SEC-ERBA approach because the credit enhancement provided by the loss absorbing nature of the excess spread is reflected in a higher rating of the more senior tranches. On the other hand, in the context of the SEC-IRBA and the SEC-SA, treating trapped excess spread as a securitisation position subject to a 1250% risk-weight/capital deduction leads to unfair and overly conservative doubling up of capital requirements unless the attachment and detachment points of the first loss and mezzanine tranches are adjusted accordingly under SEC-IRBA and SEC-SA.
As with traditional securitisation, we would be grateful for clarification of the way in which "excess spread should be taken into account within the risk transfer self-assessment analysis" where this implies considerations beyond those articulated in paragraph 123(a) needing to be taken into account.
See also our response to Question 12 generally in respect of the treatment of excess spread.
11(a). In particular, do you agree with the proposal of only allowing a contractually fixed (pre-determined) excess spread commitment in synthetic transactions?
This is unproblematic in general, although we would clarify that it should be fixed as a percentage of the nominal amount of performing underlying assets each year rather than a fixed nominal amount of money. This is consistent with the idea that it should be limited to the 1-year expected losses on the portfolio.
11(b). In particular, do you agree with the proposal to only allow a ‘trap’ excess spread allocation mechanism in synthetic transactions?
No. See responses above in respect of Question 11 generally."
Members of the Joint Associations recognize the need to take account of the use of excess spread in a securitisation in the self-assessment an originator submits to its competent authority. However, rather than holding capital against untrapped excess spread or synthetic excess spread, whether by additional Pillar 1 funds or as a securitisation position, we believe excess spread should be recognised as a cost of protection as follows:
a. In a synthetic securitisation transaction with no excess spread feature, the bank would submit an analysis to the competent authority where the cost of protection would factor in – as a reduction in cost – the amount of losses that are offset by the benefit of the credit protection payments (as computed for the relevant loss scenarios).
b. In a traditional securitisation transaction where there is an excess spread feature, the bank would not take account of losses absorbed by the excess spread as a reduction in of the cost of protection (except for the portion of losses, if any, exceeding the excess spread amount).
c. In a synthetic securitisation transaction with a synthetic excess spread feature, similarly to (b), the bank would not take account of losses absorbed by the synthetic excess spread as a reduction in the cost of protection in its cost analysis (except for those exceeding the synthetic excess spread).
d. In all cases, only funded excess spread amounts would be subject to 1250% risk weighting/capital deduction.
We feel this approach ensures consistency with the treatment of future income on loan portfolios. The CRR explicitly confirms that capital need not be held against the possible reduction of future spread on a portfolio. We would submit that if an originator does not recognise any excess spread as income until the portfolio actually produces that cash, then there should be no need to recognise any potential losses that may be absorbed by that expected future excess spread stream.
No, for the reasons set out in our answers to the previous questions. Unrealised excess spread is not recognised as an asset and therefore there is nothing against which to hold capital. Rather, unrealised excess spread is simply future income that may or may not arise and it is not appropriate for such future income to become subject to Pillar I own funds requirements.
13(b). In relation to the further considerations for stakeholders’ consultation on the own funds treatment of excess spread, what would be the impact on SRT transactions if Pillar I own funds requirements were recognised as suggested in Section 3.2?
If Pillar I own funds had to be held against future excess spread (as opposed to only being held against trapped excess spread) this would lead to a significant reduction in the RWEA benefits it is possible to achieve from SRT deals, thereby making them less efficient and increasing costs to unacceptable levels. See an example at Annex I, which is based on a Joint Associations member's actual auto loan transaction and market pricing:
1. The transaction is a EUR 1bn rated traditional securitisation with seven classes of securities and a liquidity reserve account. It has a sequential pay structure with 10% clean up call and no time call.
2. In the example below, the risk weightings of the tranches are based on SEC-ERBA, assuming the transaction achieves STS status.
3. Mezzanine classes D, E and F are 95% sold to investors, meeting the mezzanine mechanical test for SRT. The coupons are based on market prices.
4. Retention requirements are met via retention of 5% of each tranche by originator.
5. We compare two scenarios:
a. Scenario assuming trapped deposit of excess spread equal to expected losses for the following year. Accordingly the funded excess spread account is totally consumed each year and no unused excess spread is carried over.
b. Scenario assuming no excess spread is trapped.
The conclusion of this analysis is that the trapped spread scenario results in a cost of protection 37% more expensive, with 37% less weighted average RWA reduction as compared to the non-trapped scenario.
We note that some ways of achieving protection of the investor from the counterparty risk of the originator would include depositing cash in a manner that is segregated from the bankruptcy estate of the originator or including downgrade triggers requiring the deposit of cash with another bank or investment of cash in eligible collateral held at an independent custodian. These methods are only viable if they do not impair the SRT analysis. We also note that depositing the cash with a third party bank even under these circumstances would fall foul of the eligibility conditions for STS treatment set out in Article 270 of the CRR, which we do not view as a sensible outcome.
In terms of continuity of servicing we would suggest that originator insolvency not be a termination event in itself. Rather, it should be a termination event only when combined with at least one servicing failure in the portfolio and the failure to appoint a back-up servicer.
- Clearly doing self-assessments should not be required in relatively simple cases or where it is clear that the structural features described in the Discussion Paper do not impair SRT, as acknowledged in the Discussion Paper. Further guidance on the situations in which self-assessments are required would be helpful.
- Additional guidance on appropriate back-loaded loss scenarios would be helpful in order to ensure consistent application of this rule across jurisdictions.
- It is unclear whether the reference to total losses" in the main body of paragraph 165 (as opposed to 165b. where it is clear that the reference is to expected losses) refers to expected losses, unexpected losses or both. It would be helpful to specify.
- The references to PD and LGD in paragraph 166a may not be applicable to all portfolios (e.g. SA or IRB-F portfolios) so this element of the analysis should only be required where they are applicable.
- Cash flow models are complex and time-consuming to produce and we would suggest that alternative methods of demonstrating achievement of SRT be permitted where appropriate. For example, for some originators the level of stress used in the commensurate risk transfer test can be tailored to reflect base and stress scenarios, making cash flow models unnecessary. Instead, the originator can demonstrate risk transfer under base and stress cases using a variety of appropriate internal models (which may include cash flow models or modified risk transfer simulation models). Actual model submission could also be arranged on request by the relevant prudential regulator.
- We are somewhat concerned that the articulation of the considerations as set out in the Discussion Paper leaves a large degree of discretion for national competent authorities in assessing the SRT implications of the various structural features and their interaction. While it is necessary to allow a significant degree of discretion in order to have appropriate case-by-case analysis, we think there would be value in periodic central reviews of these decisions by the EBA in order to identify and reduce differences between jurisdictions in the SRT decision outcomes for comparable transactions.
As to the appropriate standard for stress tests, members of the Joint Associations would emphasise that the tests need to be granular, loan-level stress tests appropriate to the particular portfolio rather than wide, institutional stress tests. Further, it is important that the tests actually applied should be designed by reference to the risk management principles adopted by the particular institution. Where relevant, it may be appropriate to use EU-wide or Basel stress tests as benchmarks against which to judge the individual institutions' stress tests, but careful thought should be given to the appropriateness of this in each case."
As a starting point, originators who do SRT transactions in order to obtain capital relief are automatically disincentivised from structuring transactions with thin protected tranches because that would result in very high risk-weights for the retained tranches especially in view of the built-in non-neutrality of the securitisation framework – thereby eliminating one of the main benefits of the transaction for the originator.
In addition, if there is to be a minimum first loss tranche thickness test, members of the Joint Associations would emphasise that this should apply for the purposes of initial assessment of SRT only. Even if the EBA chooses to assess SRT on an ongoing basis generally contrary to our suggestions above, it is important that any minimum first loss tranche thickness test not be tested in an ongoing manner. This is because – assuming a first loss tranche structured to meet the test precisely – SRT would be lost the moment there was a single loss on the portfolio, which may well happen soon after the transaction closes.
However, we have the following comments on the test proposed:
- Making the test one that is purely about minimum thickness of the first loss tranche would make SRT unnecessarily difficult to achieve. Instead, the EBA should consider a test based on minimum thickness of protection tranches in general (first loss and mezzanine) or a requirement to dispose of a higher threshold of the tranches representing expected losses.
- The test for the minimum thickness of the first loss or any other tranche(s) should be applied at the closing of the transaction only since the lifetime EL will change over the course of the transaction and it is generally not practicable to adjust tranche thickness after the transaction has closed. We also note that the definition of lifetime EL for the SEC-ERBA and SEC-SA will generally be zero on at closing unless the portfolio already contains impaired loans on day 1 (which will never be the case e.g. for STS securitisations).
- There is also an anomaly in the current mechanistic tests which is not corrected by the new definitions of first loss tranche and mezzanine tranche. This is illustrated below. In both examples, 80% of the EL and UL risk is sold to investors, which meets the substantive requirements of the EBA guidance. However, in the first example, the senior retained tranche has been divided into senior and mezzanine tranches (which is the most efficient structure under the SEC-ERBA and SEC-SA approach). Because the mechanistic test states that if mezzanine tranches are present, the originator must sell >50% of the risk weight of the mezzanine tranches, the mechanistic test fails, even though substantively the same amount of risk has been transferred to investors. We believe this is an unintended consequence of the mechanistic tests which should be corrected if the suggestions put forward in the Discussion Paper are ultimately agreed requiring that a significant portion of the EL and UL risk has been transferred.
- The minimum first loss tranche test requiring that the first loss tranche cover 100% of EL and 2/3 of UL should be modified to take into account a scenario such as the one illustrated below. In this graph, we see that the originator has transferred 100% of the EL and UL to investors. The first loss tranche does not meet the first loss test, but in substance, the combination of the mezzanine and first loss tranches sold is sufficient to transfer 100% of EL and 2/3 of UL. If not worded correctly, the first loss tranche test would inadvertently disallow the illustrated transaction.
Accordingly, the test should be restated or clarified to take into account transactions in which a sufficiently large portion of mezzanine risk is sold to cover UL, and where the attachment point of the sold mezzanine tranches encompasses the portion of EL not covered by a thin first loss tranche. Certain members of the Joint Associations find that the optimal cost of protection and risk transfer (and consequent RWEA reduction) can be achieved if they retain a small first loss tranche, while transferring a sufficient amount of mezzanine tranches to cover the rest of EL and most of UL. To prevent the abuse that this test is intended to cover, the EBA could specify that the mezzanine test will achieve SRT if the originator transfers any combination of first and mezzanine tranches with a thickness equal to EL + 2/3 UL. Another alternative would be to require banks using the mezzanine test to sell a greater percentage of mezzanine tranches than 50%, such that the percentage sold above 50% equals the amount of EL not covered by the first loss tranche.
Another way to approach this would be to articulate the test in terms of the detachment point of the transferred position combined with a maximum retained first loss size.
From an investor’s perspective, the focus is to take on an investment with an attractive risk/reward profile. Assuming the investor is not regulated under the CRR, the new regulation is not key in the assessment of risks and rewards. The focus is rather on achieving a target return on the investment considering the investor’s assessment of expected and unexpected losses of the underlying portfolio. If an originator is forced to sell thicker tranches/more junior risk to achieve SRT, such a sale will not be economic for the bank if investors do not agree to accept a commensurately lower return for their investment. However, those investors most interested in first and second loss tranches may have yield requirements higher than the maximum weighted average cost of protection required for a deal to make economic sense to the originator. In order to reduce the all-in cost of protection, it will become more necessary to structure transactions with multiple tranches in order to attract different investors with different yield requirements. If the above-mentioned anomaly in the first loss and mezzanine tranches is not corrected, it will be difficult, if not impossible, for originators to sell a thick enough portion of tranches at an economic level.
For example, under the new CRR and commensurate risk rules, a bank may determine it is necessary to sell a position with a zero attachment point and a 12% detachment point in order to demonstrate SRT and commensurate risk transfer, which is greater than the zero to 8% tranche thickness that bank would need to sell to meet the same tests under the current CRR. It is likely easier and less costly to divide such position into three tranches; a first loss and two mezzanine tranches, because the bank would then be in a position sell to different investors with different return requirements. The first loss investor may need, say, 12%, while the senior mezzanine investor may only require, say, 4%. On a weighted average basis, the total cost of selling these three tranches may be, say, 8%. However, if the originator cannot meet the mezzanine test in this example, they are forced to structure one single first loss tranche with a zero attachment and 12% detachment point. It is unlikely first loss investors will be willing to accept an 8% return for such a tranche, and the originator may be forced to pay, say, 9% or 10%.
Of course, an originator always has the option to seek specific approval for SRT under Article 244(4) or 254(4) if it cannot comply with one of the mechanistic tests. However, as mentioned above, this process is often lengthy and uncertain, with some members reporting average response times for obtaining approval from the competent authority of 11 months. Many originators will abandon the effort to do an SRT deal rather than to take the risk of closing a transaction, only to have to call it 11 months later (with all the attendant written off costs and reputation risk incurred with investors) if the regulator does not approve the transaction.
Members of the Joint Associations are not opposed to the commensurate risk transfer test in principle and believe it could play a valuable role as an alternative way of demonstrating SRT (and not a cumulative one required in addition to the existing mechanistic tests). Equally, as mentioned in our answer to Question 3, it could play a useful role in providing a ‘cap’ on the capital relief that can be achieved via SRT. That said, there are a number of issues that members have raised with the current proposal for the test.
Ratio 1 ≤Ratio 2
((UL+〖EL〗_1Y )-K_retained)/((UL+〖EL〗_1Y))≤((UL+(〖EL〗_1Y×WAL))-((T_j×〖Retained〗_j )+(T_m×〖Retained〗_m ))-(〖ES〗_1Y×WAL))/((UL+(〖EL〗_1Y×WAL)) )
Where Tj is the thickness of the junior tranche, Tm is the thickness of the mezzanine tranche (if present), Retainedj is the % retained of the junior tranche and Retainedm is the % retained of the mezzanine tranche (if present).
The main issues that have been highlighted with this proposed approach to the commensurate risk transfer test are:
1) For the case where a mezzanine tranche is 100% transferred to an investor and the excess spread and first loss tranche are both under CET1 deduction or have 1250% risk weights, increasing the thickness of the transferred mezzanine tranche has no impact on the Ratio 2 numerator. Economically, it would be expected that more risk has been transferred, but according to EBA proposed method, increasing the thickness of the transferred mezzanine tranche could make the test more difficult to pass. The reason for this is that the thickness of any retained senior tranche would be reduced and potentially also decreasing the risk weights of the retained senior tranche, thus reducing Kretained and increasing Ratio 1 (this is especially important to consider when SEC-IRB is introduced).
2) A further issue with the formula above, is that only the retained portion of the junior and mezzanine tranches are subtracted from the lifetime EL and UL (in addition to the lifetime excess spreads also subtracted), thus the formula does not account for any retained portion of the senior tranche that also covers lifetime EL and UL. For example in case study 2, the retained senior tranche covers 0.6% of the total 4.6% lifetime EL and UL; however, as the numerator of Ratio 2 only accounts for the mezzanine and junior tranches, this 0.6% of lifetime EL and UL retained is not accounted for. This causes an unexpected situation where although only 3.2% of the whole transaction has been transferred to investors (80% of the 4% thickness of the junior tranche), the calculation for Ratio 2 states that the ‘Lifetime EL + reg UL on transferred position’ is equal to 3.8%. The result is that Ratio 2 in the case study is equal to 82.6% instead of 69.6% if the retained senior portion had been accounted for, making the test easier to pass.
3) In respect specifically of transactions structured to pass the mezzanine test, it seems that the objectives of the commensurateness test are inconsistent with the minimum tranche thickness requirement as applicable to transactions with mezzanine tranches. On such transactions, the assumption is that the lifetime EL of a portfolio is retained by the originator whereas the majority of the lifetime UL is transferred. Accordingly, it seems illogical that the lifetime EL should be considered in Ratio 2 in respect of mezzanine transactions.
4) Ratio 1 consists of 1-year measures whereas Ratio 2 uses lifetime measures. The contrasting time horizons serve to make the comparison less meaningful and useful. We would suggest dropping the lifetime measures in Ratio 2 and replacing them with 1-year measures to align the time horizons of the two tests.
In respect specifically of traditional securitisations, it is unclear how lifetime excess spread is meant to be calculated given that actual excess spread must be used and not a fixed amount. We would suggest that under such circumstances, lifetime excess spread should be deemed to be zero for the purposes of this test.
- In principle, if this type of test is accepted, it shouldn’t matter whether a bank sells 20% of the risk or 100% of the risk, since the RWEA reduction achieved will be commensurate with the risk transferred.
- In analysing the impact of this test on existing SRT positions, members of the Joint Associations have noted that, in many cases, the increased capital on senior retained tranches required by the new capital calibrations under the revised CRR causes their SRT transactions to fail, even though the positions sold to investors exceed EL + 2/3UL, which is the minimum suggested in the Discussion Paper. The effect of the new calibrations is to require an originator to sell a much larger portion of first loss and/or mezzanine tranches in order to reduce the risk weighting of the senior retained tranches below the threshold required by the test. For any particular transaction, some or all of the following negative impacts then may apply:
o It increases the cost of protection above the bank’s cost of capital, making the deal uneconomic to execute.
o It causes the total cost of protection to exceed the EBA guidelines on the cost of credit protection.
o It requires the originator to sell total tranche thicknesses far in excess of EL + 2/3UL, which makes such test redundant.
- The test in option 2 is even more problematic for rated and/or standardised portfolios. This results from the fact that the non-neutrality factor in the capital calibrations is more acute with securitisations involving the SEC-SA or SEC-ERBA approaches because of the need to divide the senior retained portion of an SRT securitisation into a series of senior and subordinated tranches, rather than one single unrated senior tranche.
Condition 1 should be amended to omit the 1-year excess spread element. This is because the amount of the first loss piece generally equals about the EL for the transaction's initial year, with excess spread paid thereafter at a rate that roughly equals the EL for the upcoming year. Including the first loss tranche and 1-year excess spread therefore amounts to double-counting. Calculation of the provisions on the underlying portfolio generally follows the same logic.
Condition 1 should also be amended to clarify that 50% of UL refers to 50% of regulatory UL (as opposed to lifetime UL).
- The double capital penalty inherent in the SEC-IRBA and SEC-SA approaches for non-auto transactions with high excess spread portfolios, which distort the capital weighting of the tranches, which results in a need to sell a greater number or portion of tranches than is necessary to cover EL and UL.
- To the extent excess spread is required to be recognised as a securitisation position (trapped or otherwise), the credit enhancement feature of this trapping should be reflected in the attachment point calculations for the approaches, similar to the EBA’s proposal for taking into consideration the purchase price discount credit enhancement effect.
- The EBA’s proposal to require a 50% LGD assumption for standardised portfolios when calculating EL is a blunt instrument that will cause a lot of collateral damage with many asset classes, such as commercial real estate, project finance or other portfolios secured with collateral that effectively reduces the LGD of such portfolios. The 50% requirement will heavily penalise these portfolios unnecessarily and unfairly, especially in the light of the double capital penalty mentioned above, if ultimately the 50% LGD requirement is adopted. There may also be further problems resulting from the fact that some portfolios may be mixed, with both standardised and non-standardised assets. See also our answer to Question 30 in respect of the very conservative results of the SEC-SA approach hampering the use of synthetic securitisation as a technique.
It should be noted, in the context of the NPL market, that:
- The key outcome from the originating bank's point of view is that the NPLs are effectively transferred off of its balance sheet. The ability for the originating bank to achieve SRT is academic, because typically these transactions are not direct securitisations, but straight disposals.
- The key issue for banks, then is their ability to derecognise the assets, which is often complicated by that originating bank keeping some form of control over the assets (e.g. as servicer) even after they have been transferred to the acquirer.
Other key aspects of the NPL market to bear in mind include:
- The size of portfolios is typically smaller for NPL portfolios as compared to other traditional securitisations.
- The NPL market typically is characterised by more specialised market participants (borrowers, investors and special servicers) since deep expertise in servicing the particular type assets involved in essential to be able to optimise the recovery process on the underlying portfolio.
- The representations and warranties from sellers will typically need to be much more detailed and bespoke than for a performing portfolio.
The secondary market in NPLs could be further developed if certain obstacles are removed, such as:
- Barriers to entry for NPL investors: For example, some EU jurisdictions require loan servicers to hold a licence. This is a barrier to entry for entities that would otherwise be interested in becoming investors in NPLs.
- Limitations in the transferability of loans: Banks are often unable to sell loans containing restrictions on transfer/assignability of the loan. Policymakers could (at the national level) make such clauses unenforceable by operation of law once the loan ceases to perform.
- Disclosure limitations: in some jurisdictions, banks are not permitted to make disclosures to potential acquirers of loans that a mandate or conciliation procedure is underway, which makes it difficult or impossible for banks to make the necessary disclosures for acquirers to perform appropriate due diligence on the portfolio and means the bank can't give the relevant representations and warranties as to disclosure on the portfolio a reasonable acquirer would require of them.
- Requirement for banking licenses: Certain jurisdictions require a banking license to extend new credit facilities, which restricts the ability of the SPV to extend further credit as part of the management of the debt portfolio.
- ECB eligibility: permitting banks to use the senior tranche of NPL-backed ABS as eligible collateral for ECB repo operations would encourage investment.
It should be stressed that it is important to ensure the “effective transfer of risks”, otherwise IFRS 9 (when it becomes applicable) could impose additional loan loss provisioning related to the expected losses of the transferred exposure.
We note as well that there is a wide discrepancy in this case between the results of the SEC-IRBA and the SEC-SA. De facto for NPL pools, the SEC-SA results are conservative; consequently, the SEC-SA cannot be used by market participants as it would make transactions unprofitable. This is quite detrimental for the development of the NPL securitisation market or the securitisation market of pools with material delinquency buckets. In our view this unfairly discriminates against standard banks.
The over conservatism of the SEC-SA for NPL pools stems in our view from the way the W factor is taken into account in the SEC-SA. This issue is not specific to NPLs but it is amplified in the case of NPL pools. This has already been described (with a simple correction proposed) in a paper called: “How to revive the European Securitisation Market: a Proposal for a European SSFA” by Georges Duponcheele et al. published in November 2014.
According to the paper: “It is to be noted that this official version of the SA in the BCBS 269 proposals contains a conceptual error in that it includes the delinquent assets (via KA) in KP. This in effect means that the capital associated with provisions for the delinquent assets is allocated to the senior tranches via the exponential function and the parameter α. To correct this conceptual error, KP should simply equal pool capital for performing loans and should not be a function of W. The delinquent assets should only affect the 100% capital charge threshold KT. It is important to distinguish between KT and KP to avoid making errors.” We therefore reiterate the need to correct the way the W factor is taken into account in the SEC-SA as detailed in the extract above.
In addition, first loss risk retention being done on a nominal / notional basis (i.e. option (d) of Art. 405(1)) for NPL securitisations may not allow for any downside risk exposure to the retention party where the purchase price discount is significant.
Application of the maximum risk weight for senior securitisation should take into consideration the purchase price discount applied. A possible method of doing this (assuming that the entire securitised portfolio is non-performing) could be:
SEC-IRBA:
Adjusted KIRBR Formula = [Max.(0, EL-PPD) + RWA (based on NPL Formula) x 8%)] / [EAD (based on GBV) – PPD]
Max Senior Tranche Risk Weight = Max. [(Adjusted KIRBR x 12.5), 15%]
SEC-SA:
Adjusted KSA Formula = [RW x Max. (0, (EAD (based on GBV) – PPD) x 8%)] / [EAD (based on GBV) - PPD]
Max Senior Tranche Risk Weight = Max. [(Adjusted KSA x 12.5), 15%]
Question 1: Does the data on synthetic and traditional SRT securitisation transactions correspond with your assessment of SRT market activity in the EU? Do you have any observations on these data?
As part of the response to this question, IACPM has asked member firms to update the 2015 and 2016 volume survey information that was provided to the EBA on synthetic securitisations. This survey was completed by 17 member firms that engage in synthetic securitisations. These firms provided data from 2008 to the present for large corporate, SME, trade finance and other transactions for this response letter. We acknowledge there may be additional data forthcoming from active market participants that we will certainly share with the EBA.The survey provides detailed information on number of trades, distributed tranches total size, pools total size, attachment and detachment points as well as geographic breakdown of total pool.
Similar to what is reported in the discussion paper, 2015 was the peak year in synthetic transactions but volume has remained steady in the subsequent years. The detailed survey information will be submitted in a separate response given the sensitivity of the information.
That said, the data generally appear to be accurate as far as they go but they are clearly incomplete as overall volumes are clearly much too low. In addition, members are aware of at least one synthetic securitisation in excess of EUR 8bn whereas the EBA data show the largest transaction at circa EUR 5bn.
We also observe that there are relevant ways in which the EBA did not break down the data. These include:
- The percentage of SRT traditional securitisations which were publicly offered vs. private placements.
- The percentage of SRT synthetic transactions which were public or privately issued CLNs, although it is acknowledged that this distinction is sometimes difficult to draw in the case of some synthetic transactions.
- The percentage of synthetic transactions which were structured as private/bilateral guarantees or CDSs without an associated issuance of notes.
Finally, we would note that the prevalence of calls as described in paragraph 50 of the Discussion Paper represents a considerable underestimate as compared to the experience of the members of the Joint Associations. In our experience, both cleanup and regulatory calls are included in the vast majority of transactions; we would have expected a prevalence at or near 100% for these types of call options.
Question 2: Are you aware of any material supervisory practices that have not been covered in the EBA analysis?
In addition to the areas identified (heterogeneity in procedures and heterogeneity in treatment of structural features), there is also significant heterogeneity in the content of the information competent authorities (including different joint supervision teams acting as part of the SSM) require from originators in their SRT notifications. At least one competent authority requires a report substantiating compliance with virtually every article in the relevant chapters of the CRR as well as the SRT Guidelines issued by the EBA, including cross-references to relevant provisions of the transaction documentation.This can be a very useful exercise because it acts as something of a checklist of the requirements for achieving SRT, but it is also obviously a very burdensome exercise and much more involved than the data submission requirements of other competent authorities. Members of the Joint Associations are of the view that the harmonisation of both the process and the form of notification (including the required level of detail) would be helpful in creating a level playing field. This, however, should take account of the type of transaction being notified and the institution's history. The EBA should consider a detailed notification template as standard but with the option to simplify the process where the transaction would not lead to a material RWEA reduction, or the transaction is of a type previously approved by the relevant competent authority in respect of that institution and there are no material structural differences. See our further answers below, especially our answer to Question 3.
Finally, the EBA has noted that very few deals have sought approval under Art. 243(4) or 244(4). In the experience of certain AFME members, this is likely to be because competent authorities are reluctant to commit themselves to analysing deals which do not meet the mechanistic, quantitative tests and providing timely approval. At least one member has waited an average of 11 months to receive a response on any SRT requests submitted pursuant to these provisions.
Question 3: What are your views on the proposals on the assessment process set out above? Are any other changes necessary to further improve the process?
In general, the Joint Associations are supportive of the recommendations. We consider in general that having a harmonised procedural framework for the assessment of SRT would be of significant assistance in establishing a level playing field across Europe in respect of SRT transactions. In particular, we consider that the requirement for explicit feedback within a reasonable and predictable timeframe by competent authorities to originators in respect of SRT is highly desirable and likely to promote greater certainty and predictability in respect of SRT transactions where there is a novel element to the transaction. This feedback should be both informal (before closing where ex ante notification is given) and formal (after the transaction is closed). In both cases, timeliness of the feedback is an essential element in order for it to be useful. We would caution, however, that this may become unnecessarily cumbersome where originators are undertaking a number of very similar transactions. Therefore, the EBA should consider an exception to the general rule of requiring explicit feedback where the transaction is of a type previously approved by the relevant competent authority in respect of that institution and there are no material structural differences.While we agree in general with the spirit of the proposals, there are several considerations members of the Joint Associations think should be taken into account in further refining the proposals, as follows:
- Timing of notification: The economics of some securitisation transactions will only be viable if SRT can be achieved. In respect of these transactions – which are predominantly but not exclusively synthetic securitisations – a framework that allows for ex ante notification, with feedback from the competent authority on SRT prior to closing, would be extremely helpful. We also view one month as a reasonable timeframe for this notification. It is, notably, significantly more workable than the current requirement for notification three months prior to closing imposed by the ECB. Obviously, information about the transaction may be incomplete at this stage as contractual terms and the structure continue to evolve and are negotiated right up to the closing date. See further discussion below under the heading Contents of ex ante notification".
That said, many transactions that might be capable of achieving SRT are nevertheless not done for that specific or exclusive purpose. There is no reason why a claim for SRT treatment should not be able to be made after closing of the transaction where the originator so chooses. This fact pattern is particularly likely to arise in respect of traditional securitisations, where SRT is most likely to be a secondary purpose of the transaction. On a traditional securitisation, the issue of SRT could also become relevant months or years after the closing of a traditional securitisation if part of the capital stack that was not sold to a third party investor at closing is later transferred.
If part of the purpose of this review of SRT is to increase the number of SRT traditional securitisations, it is essential to preserve the flexibility of originators to seek SRT treatment for an existing transaction after its closing.
In this respect, the focus should be on when the reduction in RWEA from an SRT transaction is recognised by the institution in its prudential capital calculations, which would normally follow the supervisor's final feedback. For this reason, a reasonable and predictable timeframe for receiving that feedback is an essential element of an efficient supervisory regime for securitisations.
- Contents of ex ante notification: While members of the Joint Associations generally support the idea of ex ante notification where appropriate, we feel it is important to clarify what documentation can realistically be made available as early as one month before expected closing/issuance. Members of the Joint Associations would be confident supplying competent authorities with a summary of the intention, key features and indicative size of the transaction, possibly in the form of a detailed term sheet. That said, there should be no expectation that the transaction size, pricing or other commercial terms are final since commercial terms may be subject to negotiation right up to the closing date. For the same reason, members of the Joint Associations are of the view that it is not constructive to provide draft deal documentation since this is highly likely to be negotiated right up until the closing date. Obviously, it will be impossible to provide final transaction documentation until after closing, as acknowledged in the Discussion Paper. Further comments on the proposed deadline of 15 days post-closing to provide final documentation appears below under the heading "Timing of final documentation and information".
Further, pre-notification should not be seen as committing the relevant institution to the transaction, as that would hamper its ability to carry out proper commercial negotiations with its prospective investor(s).
Finally, to the extent that transactions do not currently need to be notified (e.g. where the full deduction option is applied) a simplified notification regime should also be applied, with the institution explicitly able to assume SRT treatment and associated RWEA reduction in the absence of an objection by the competent authority within a prescribed timeframe (e.g. three months). The EBA might also consider a simplified/streamlined SRT process for small portfolios, where the time, cost and effort of completing the full SRT process would outweigh any benefits to the institution of doing the transaction. This might be appropriate where such transactions in aggregate do not exceed e.g. 0.1% of the total assets of the bank.
- Informal feedback from the competent authority: The Discussion Paper highlights at paragraph 70 that some market participants had specifically emphasised the advantages of receiving feedback ahead of the closing of the SRT transaction. The Joint Associations strongly agree that feedback prior to the closing of the transaction is extremely useful and helps to make the market much more efficient, especially if the transaction is novel or has features not previously reviewed by the competent authority in respect of the notifying institution. As it stands right now, however, a number of competent authorities do not provide this. Where transactions are being structured specifically to achieve SRT, significant transaction costs can be saved if feedback is obtained prior to closing. In addition, informal pre-closing feedback would help banks to manage reputational risk with investors by avoiding the need to call transactions prior to maturity because SRT is not achieved as expected.
Pre-closing informal feedback should highlight, at a minimum, any features of the transaction that the competent authority considers might hamper the achievement of SRT and the type of changes or analysis and information the competent authority will require in order to address the concerns they have highlighted. In the view of the Joint Associations, the one month ex ante notification should be the beginning of a conversation during which the originator and the competent authority exchange views about the transaction. The goal should be to arrive at a position of sufficient certainty in advance of closing that the originator can complete the transaction with a high degree of confidence that SRT will be achieved. Of course, to the extent that the originator has omitted information from the SRT notification, or the final transaction documentation varies materially from what has been described in the notification, it would be entirely appropriate for the competent authority to reassess SRT treatment.
- Nature and consequences of ongoing assessment: The Joint Associations are of the view that SRT should be assessed when applying for SRT only and should not require reassessment over the life of the transaction save in the case of implicit support, "step-in" or if there is a significant amendment or restructuring of the deal. To be clear, this should not prevent the bank from holding sufficient capital to reflect the risk of the positions it continues to hold, which may increase as the risk-weights of those positions increase. This approach is intended to reflect the fact that achieving "significant risk transfer" in fact merely means that the basis for calculating capital charges changes from calculating capital charges on the underlying portfolio to calculating capital charges using the securitisation framework. All relevant structural features of a transaction are known at the outset and should be captured in the documentation. It is entirely reasonable for the competent authority to wish to ensure that significant risk is transferred for a sufficient period of time, or that the first loss or mezzanine tranches are sufficiently thick to effectively transfer risk, but those concerns should be raised when the transaction is notified, rather than being subject to reassessment partway through the life of the deal.
Where SRT is initially achieved, the absorption of losses by the protection tranche(s) is expected and intended. The fact that losses materialise and are absorbed by the protected tranche(s) (thereby reducing its/their thickness) does not therefore mean that that risk was never initially transferred or that SRT treatment should be revisited. Rather, the absorption of actual losses by the protected tranche should result in a reassessment of the risk-weights of the retained tranches held by the originator. Provided that the originator adjusts its RWEA associated with the retained tranche(s) on an ongoing basis to reflect the then-current protection provided by the transaction, the basis of calculating its RWEAs (portfolio basis vs. securitisation framework) should not need to be reassessed.
One way in which this could be achieved, and (subject to our comments below on the specific proposals for commensurateness tests in the Discussion Paper) perhaps the most sensible application of a commensurateness test, would be to apply the commensurateness test essentially as a cap on the capital relief that can be achieved via a given SRT transaction. So, for example, if according to the commensurateness test eventually adopted 50% of the total expected losses and regulatory unexpected losses on the underlying portfolio had been transferred to investors, the effect of the commensurateness test would be to cap the capital relief resulting from that transaction at 50% of the capital that would otherwise have to be held. In this way, the cliff effects of the binary nature of SRT are alleviated while still achieving the regulatory objective of ensuring banks hold sufficient capital to reflect the risks of the positions to which they are exposed.
- Fixed timeframes for competent authority feedback: The Discussion Paper proposes specific timeframes and deadlines for originator submissions to competent authorities (i.e. notification one month before expected issuance, final documentation 15 days after closing). Subject to our comments elsewhere in this response, we consider that introducing fixed timeframes is helpful to promote predictability of the process. However, these timelines should also be realistic and linked to the final purpose, which is determination of when the RWEA effects of an SRT transaction should be recognised by the institution in its prudential capital calculations. In order to achieve the hoped-for predictability, it is necessary that all parts of the process should operate within predictable and consistent timeframes. To that end, the Joint Associations would urge the EBA to set out a defined timeframe for competent authority feedback on submissions, rather than simply saying it "should be provided within a reasonable timeframe after the submission of the final version of all information/documents". Where an ex ante notification process is followed, we have mentioned above that we believe the period between notification and closing should be used for the originator and competent authority to exchange views on the merits of the transaction with the aim of providing the originator with reasonable certainty as to the transaction's ability to achieve SRT. Obviously any views expressed by the competent authority would be subject to the final transaction documentation reflecting the information provided by the originator prior to closing. Of course, a final and binding determination would only be possible in the light of the final transaction documentation and this should be provided no later than three months following receipt by the competent authority of the final documentation needed for it to assess the claim of SRT.
- Timing of final documentation and information: Members of the Joint Associations are concerned that meeting the deadline of 15 days after closing to provide "a final version of all information/documentation" will be difficult to achieve. If all that is required at that stage is the final transaction documentation, this will of course not be problematic. Once the transaction has closed, 15 days is sufficient to provide the final executed documents. If, however, further extensive information and analysis is required by competent authorities (e.g. due to structural changes resulting from informal feedback between ex ante notification and closing or changes to the transaction not yet discussed with the relevant competent authority), this timeframe may be too ambitious for banks to meet. There would also be concerns around producing the SRT self-assessment proposed in the Discussion Paper within this timeframe. Based on current competent authority requirements for transaction notifications, some members of the Joint Associations have expressed the view that 30 days post-closing would be a more appropriate timeframe within which to provide final documentation and information.
The Joint Associations would be pleased to engage further with the EBA and national competent authorities on this subject once the form of SRT notification going forward (along with the details of any self-assessment or similar exercise) becomes clearer. We are confident that it will be possible to create a process that provides for timely delivery of relevant information in a timeframe that is manageable for market participants.
- Notification for time call exercises: In this respect, please see our answer to Question 9."
Question 4: Could you provide suggestions as to whether and how the template for SRT notification by the competent authority to EBA provided in Annex I of the EBA Guidelines should be amended to reflect the new EU securitisation framework and the STS securitisation product?
Members of the Joint Associations are of the view that it would be most appropriate for national competent authorities and the EBA to formulate a proposed template for SRT notification in the first instance. Once a proposal has been formulated, the Joint Associations would be pleased to engage with the EBA and provide their views on the proposed template.Question 5: Should a standardised SRT notification template be developed, for submission by originators to competent authorities, in order to facilitate the SRT assessment process? If yes, should this template be different for traditional and synthetic securitisation? (Please provide examples of templates, as appropriate).
In general, members of the Joint Associations support the development of standardised templates. This would help to level the playing field for institutions in terms of the work necessary to notify their competent authorities of an SRT transaction and address the issues described in our answer to Question 2. That said, there is a large diversity of types of securitisation transaction that may be used to achieve SRT. It is therefore important that any standardised template(s) developed should be sufficiently flexible to accommodate the full range of transactions (with different asset classes, features, geographies, local laws, etc.) in respect of which it might need to be used.In addition, members of the Joint Associations support the development of separate templates for traditional and synthetic securitisation. These two types of transaction are materially different to one another and we believe that the development of separate templates for each type of transaction would help to reduce confusion and streamline the review process.
The contents of the ex ante notification should be as discussed in our response to Question 3. We would suggest that the existing ECB questionnaire for SRT notification is a good template to start from for the final submission to be made after closing of the transaction. Subject to our comments above in relation to simplified notification in some cases, this should be accompanied by a report demonstrating compliance with the relevant CRR articles and the SRT Guidelines issued by the EBA (as amended following the current re-examination). In addition, where any of the relevant structural features are present, a risk transfer self-assessment should also be included.
Question 6: Could you provide suggestions as to how a template for monitoring SRT compliance should look like (e.g. by potential amendments of the current COREP framework)?
Please see answer to Question 3 above in respect of the need for ongoing monitoring of SRT.That said, to the extent ongoing monitoring is deemed necessary, members of the Joint Associations would strongly encourage the EBA to incorporate SRT monitoring into the COREP framework and cover it as a part of more general ongoing regulatory and economic capital monitoring. As much as possible this should be accomplished with standard investor reporting data. This would help to minimise the operational risk that would come with the need for ad hoc reporting. A new full assessment of SRT should only be necessary in the case of significant restructuring or amendment of the transaction as mentioned above in our answer to Question 3.
Question 7: Do you agree with the assessment of the SRT implications of all the identified structural features? Are any material aspects missing from this representation?
Assessment of SRT implications from identified structural features:- General capital treatment: As a general matter, members of the Joint Associations believe that the SRT and capital rules actively discourage the use of securitisation – and in particular the use of securitisation to achieve capital relief. As a starting point, senior tranche risk weights are set (including the built-in non-neutrality of securitisation risk weights) to account for the fact that risk may not have been transferred as effectively as expected. This is despite the fact that, in order to achieve SRT, transactions have to be structured in such a way as to ensure that senior tranches will not suffer losses even under extreme stress scenarios. Then, where the portfolio does suffer a stress scenario, SRT is re-examined and sometimes lost, leading to sharp capital increases for the institution in question because capital must once again be held against the whole portfolio once SRT treatment is lost. We would encourage the EBA to look at these questions in a holistic way and consider whether existing rules unduly discourage the use of risk sharing transactions as a capital management tool for banks. Our comments above in response to Question 3 (and in particular under the heading nature and consequences of ongoing assessment) suggest an alternative approach that may reduce cliff effects and produce more sensible outcomes where protected tranches become thinner as a result of losses being realised during the life of the transaction.
Another way of looking at this would be to acknowledge that a principal effect of achieving SRT is to allow banks to calculate their capital requirement using the securitisation framework rather than calculating it on the basis that they continue to be exposed to the full risk of the underlying portfolio. Accordingly, banks could be permitted to continue calculating their RWEAs on the basis of the securitisation framework (adjusted for risk on an ongoing basis) throughout the life of the transaction, regardless of whether they continue to meet the initial SRT tests. SRT having been achieved at the outset of the transaction, one might for example incur progressively increased capital weights on the retained tranches as losses are incurred on the underlying portfolio and protection is eroded. This would avoid artificial cliff effects and more closely match capital requirements to actual risks held by banks. What is more, the non-neutrality of the securitisation framework would ensure that banks would never hold insufficient capital as a result of this approach. Indeed, this is already achieved by the new securitisation framework (see general introductory points above).
- Credit events: The section on credit events and their definitions set out in the Discussion Paper is somewhat confusing. Members of the Joint Associations are of the view that the focus on Article 178 (which is based on likelihood of a loss being suffered) is incorrect, given that the effect of the SRT is that the originator is protected from suffering a loss should the underlying loan actually default, which is appropriately covered by Articles 215 and 216. Despite this, the recommendation focuses closely on the definitions of the various credit events set out in Article 178. The discussion of implications for SRT, on the other hand, mentions compliance with the requirements provided for in Part Three, Title II, Chapter 4 of the CRR. There is no mention of the possible conflicts between Article 178 and e.g. Article 215 and 216. To take the example of failure to pay, Article 178(1)(b) deems a default to have occurred when a payment is more than 90 days past due, but Article 216(1)(a)(i) regards there to have been a failure to pay "under the terms that are in effect at the time of such failure, with a grace period that is equal to or shorter than the grace period in the underlying obligation". A failure to pay within the meaning of Article 216(1)(a)(i) could therefore occur either before or after the obligation goes into default under Article 178(1)(b), depending on the terms of the obligation and any grace period it provides for. Similarly, the definition of restructuring in Article 216 is different from that set out in Article 178. Of course, if originators wish to define their credit events by reference to Article 178, that should also be permitted, but the effects of the transaction within the credit risk mitigation ("CRM") framework would continue to be governed by Part Three, Title II, Chapter 4 of the CRR.
Furthermore the proposal to require three credit events for synthetic securitisations to achieve SRT contradicts the credit risk mitigation rules under CRR Articles 213 and 215 CRR, as "restructuring" is not currently required for financial guarantees. As such, it is unclear how financial guarantees could qualify for SRT under this proposal as outlined in the EBA Discussion Paper.
We note, also, that the relevant provisions of the SRT requirements refer to the CRM framework in the CRR which are common to both synthetic securitisations and non-securitisation unfunded CRM. Any guidance on this point therefore has the potential to affect a large market beyond just securitisation SRT transactions and should be expressed not to affect non-securitisation unfunded CRM.
- Cost of credit protection: Members of the Joint Associations have no objection in principle to explaining their approach to price discovery and price appetite to competent authorities. However, we have the following comments about the factors to take into account when assessing the effect of this on SRT:
o As the market for SRT transactions operates largely on a private basis and because there is wide variation in asset type, obligor credit quality and originator, it is not straightforward to establish market comparables for the cost of first loss or mezzanine protection. Similarly, these features make it inappropriate to use credit indices or other market benchmarks as a basis for pricing of SRT transactions.
o Where premia are contingent, competent authorities should not disallow or haircut SRT simply because the lifetime premia paid are expected to exceed the lifetime EL. If anything, it would be surprising if the premia did not exceed the lifetime EL (assuming the protection completely covers tranches that map to the EL of the underlying portfolio) as the investor providing protection on the protected tranche expects to make a positive return on the transaction. Market pricing would therefore reflect the expectation that lifetime EL will be covered by the present value of the premia. Where the detachment point of the protected tranche is above the level of lifetime EL, the premia also need to account for the protection covering some or all of the UL against which the bank would otherwise set aside capital.
The proposed approach also ignores the time value of money. Any comparison of premium costs to portfolio characteristics should discount the future premium payments by a fair market rate. There is support in the high cost credit protection guidance issued by the Basel Committee in December 2011 for using the present value of premia as the appropriate comparator, rather than the absolute value. Ideally the EBA would issue clear guidance for how to determine a fair market rate and the appropriate curve to use, in order to take account of the impact of loss timing. This is particularly important for long-dated assets such as project finance or commercial real estate assets, where premia paid over the life of the deal will always exceed the amount of the protected tranche. In any case, if the originator demonstrates that they have conducted a fair auction with investors (or otherwise agreed a price on an arm's length basis) before agreeing a premium, it should be considered a fair market rate and it should not be a barrier to the achievement of (or a reason to haircut) SRT.
o Finally, comparing the cost of protection against income generated by the portfolio is misleading and should not be a barrier to achieving SRT and should not lead to haircuts from any SRT achieved. Looking at the income generated on particular loan products that may be in the portfolio is misleading because the interest rate on the loan is often only one aspect of the total return earned by the bank from the customer (especially true of corporate customers) on a relationship wide basis. In addition, any meaningful comparison would need to take account of other fees on services provided to such customers, because the all-in return may then become commercially reasonable. The reasonable cost of protection on the portfolio would then rise accordingly. There is no reason, though that this approach to corporate relationships should be a barrier to achieving SRT on the loans. In addition, the required rate on a loan is set at the time of origination of the loan. This could differ significantly from what the bank would require at the point in time when the risk sharing transaction is entered into. Market circumstances, cost of financial resources, and the creditworthiness of the borrower may all have changed in the interim. Once again, if the originator can demonstrate that they have conducted a fair auction with investors (or otherwise agreed a price on an arm's length basis) before agreeing a premium on the risk sharing transaction, it should not be a barrier to the achievement of (or a reason to haircut) SRT.
o Additionally, a pure comparison of premia and EL ignores the inherent differential in EL volatility and skewness of certain asset classes versus others. For example, although series 1 and series 2 below have the same EL (2.5%), their distributions are materially different, hence using a linear approach of premia vs. EL significantly understates the inherent protection value of guarantee 2 over guarantee 1. If premia are to be compared against EL (a proposal we think is unhelpful in principle, as described above) the EL should be adjusted for this purpose with an appropriate empirically derived scalar to reflect the volatility and skewness for different asset classes. Members of the Joint Associations would welcome the opportunity to work with the EBA to derive such appropriate scalars."
Question 8: Do you agree with the proposed safeguards related to the use of pro-rata amortisation?
Members of the Joint Associations agree in principle that there should be triggers to require sequential amortisation of any tranches providing protection. We welcome the articulation of these proposed triggers in general, but would suggest that the EBA publish high level principles and parameters in respect of the levels at which these triggers should be set in order to ensure consistency in practices across the Union. The Joint Associations would welcome the opportunity to engage with the EBA in developing such principles and parameters.The members of the Joint Associations believe in general that requiring all four triggers in pro rata amortisation transactions seems excessive. There should be some triggers to convert to sequential amortisation but these should take account of the impact of the RWEA calculation on the senior tranche. If all four triggers are required as a general matter, market participants need flexibility to adapt or omit some triggers where they are not applicable to the particular asset type or transaction (e.g. granularity trigger would be inappropriate in a highly granular portfolio).
That said, members of the Joint Associations have a diversity of views about the appropriate approach to this. These include:
- Agreeing appropriate triggers on a transaction by transaction basis with their competent authority.
- Combining triggers I and II by including non-matured defaults as part of cumulative losses" for the purposes of I.
- Having a "menu" of triggers mapped to transaction features and asset classes to which they would be appropriate.
- Developing two relatively straightforward triggers, one of which is forward looking and the other of which is backward looking, to address the concerns around timing of losses and underestimating actual losses.
- Having a set of triggers to switch from pro rata to sequential amortisation, but also having triggers that allow switching back to pro rata amortisation (e.g. where the portfolio credit quality improves again over time or simply where the triggers for switching to sequential amortisation cease to be met).
- Eliminating the idea of binary triggers and instead allowing hybrid amortisation moving on a sliding basis from 100% pro rata to 100% sequential depending on the state of the figures used as triggers.
In any case, members of the Joint Associations consider it essential that market participants have flexibility to set the levels of the triggers in such a way as to take account of asset class constraints and market constraints, such as demands made by investors or by rating agencies in order to achieve appropriate ratings for the transaction where relevant."
Question 9: Do you agree with the proposed safeguards related to the use of time calls? Do you agree with the different approach to time calls in traditional vs. synthetic transactions?
The Joint Associations disagree in general with the approach in respect of time calls.Originators include time calls in securitisations for a wide variety of reasons and trigger them for a variety of reasons as well. Time calls permit banks to recycle securitised assets and use them for new transactions (especially relevant for banks with smaller volumes of eligible assets), allow banks to benefit from improved market conditions and reduce the cost of credit protection and provide them with a backstop in case the transaction becomes too expensive/uneconomic. This, in turn, may be because the transaction has ceased to provide an RWEA reduction or the cost of the RWEA reduction has become uneconomic (e.g. because the credit quality of the pool has improved or the coupon demanded by the market has decreased).
While members of the Joint Associations understand the regulatory concerns raised by the exercise of calls during the financial crisis to prevent investors from suffering losses, a general prohibition on the use of time calls in traditional securitisations and a lengthy non-call period in synthetic securitisations is not a proportionate response as it deprives banks of the legitimate benefits time calls can provide that are unconnected with supporting investors. The correct test is whether exercising the call is a reasonable decision for the bank to make purely in its own commercial interests (which may include putting the same assets into a new, lower cost – or longer-dated – transaction). If so, then the exercise of the call ought to be allowed, regardless of the effect on investors.
We believe that time calls should be no barrier to the achievement of SRT provided that (i) there is no structural incentive to exercise the call, such as a step up after the call date (as demonstrated in the originator's self-assessment of SRT); and (ii) the competent authority is pre-notified of the intention to exercise the call by the originator along with a confirmation that the call is being exercised in pursuit of its own commercial interests (including a short explanation of what those interests are). Provided the competent authority does not object within a given period of time (say, one month) then the originator should be permitted to proceed.
It may also be worth pointing out, in this context, that a time call on a synthetic securitisation is not analogous to repurchasing the whole portfolio at par. When protection is terminated pursuant to a time call on a synthetic securitisation, the assets in respect of which a credit event has already occurred continue to be worked out, with the protection provider continuing to absorb any losses pursuant to the terms of the protection. Similar measures are put in place for traditional securitisations, where only performing assets are repurchased at par.
As to non-call periods for synthetic securitisations, members of the Joint Associations agree that the EBA has a legitimate concern to prevent SRT transactions undertaken for year-end window dressing" concerns as described at the EBA's Public Hearing on 17 November 2017. However, inserting lengthy non-call periods has the effect of preventing banks from exiting transactions once they cease to be economically advantageous, which serves no legitimate purpose. Instead of the proposed non-call period in the Discussion Paper, we would recommend a minimum one year non-call period. A one year non-call period would exclude the possibility of engaging in transactions just designed to mislead the market by improving the apparent financial condition of the bank at year-end and in any case, investors would generally not accept a non-call period shorter than one year. The EBA proposal for a lengthier non-call period appears arbitrary and does not appear rooted in any Basel guidance of which we are aware. Going forward, the regulatory framework should not make reference to the WAL of the asset pool, which we view as irrelevant to the length of any non-call period.
Finally, members of the Joint Associations would be grateful if the EBA could issue guidance clarifying that time or regulatory calls can be used to repurchase only certain tranches without having to unwind the whole transaction. In some cases, this may be the most sensible thing for a bank to do. For example, where a traditional securitisation has both an SRT and a funding purpose, it is possible that the SRT purpose may cease to be economically viable leading to an incentive to redeem the expensive junior notes while leaving the relatively cheap senior notes outstanding for funding purposes."
Question 10: Do you agree with the proposed safeguards on the use of excess spread in traditional securitisation?
In general, members of the Joint Associations agree with the proposed safeguards on the use of excess spread in traditional securitisation. Allowing its use in SRT transactions is essential, as without an excess spread mechanism it would be difficult or impossible to construct an economically viable SRT securitisation of assets with high expected losses (and correspondingly high interest income) such as unsecured consumer loans.We understand that the proposal to treat excess spread as a securitisation position subject to 1250% risk weight/capital deduction is restricted to any portion of synthetic excess spread committed on a yearly basis to loss absorption which is funded and will not apply to the committed but unfunded portion of synthetic excess spread. The same would apply to traditional securitisation excess spread. It would not be appropriate to treat excess spread on a traditional securitisation as a securitisation position subject to 1250% risk weight/capital deduction unless it is trapped" in a reserve account or other structured designed to provide credit enhancement. Indeed, on a traditional securitisation where excess spread is "actual" excess spread, it is fully dependent on the performance of the securitised portfolio rather than being contractually set. It should therefore not attract any capital charge.
We would also be grateful for additional clarity in respect of the way in which "excess spread should be taken into account within the risk transfer self-assessment analysis" in respect of a traditional securitisation. Barring the situation in which the originator guarantees a minimum level of excess spread on the transaction, or if it is trapped as described above, it seems to members of the Joint Associations that excess spread at whatever levels should not raise SRT concerns on a traditional securitisation. In order to achieve an efficient and predictable market, it is necessary that market participants should have greater clarity about the way in which excess spread would be taken into account."
Question 11: Do you agree with the proposed safeguards constraining the use of excess spread in synthetic securitisation? In particular, do you agree with: a. The proposal of only allowing a contractually fixed (pre-determined) excess spread commitment in synthetic transactions? b. The proposal to only allow a ‘trap’ excess spread allocation mechanism in synthetic transactions?
It is unclear why the trapping mechanism is required when use-it-or-lose-it would be more effective at achieving SRT since it transfers more of the 'timing-of-losses' risk to the investor. For this reason, members of the Joint Associations would strongly urge the EBA to permit (and indeed encourage) the use-it-or-lose-it approach in SRT transactions. As the Discussion Paper says at paragraph 121 …when loss volatility is factored-in, the trapped allocation mechanism effectively counteracts SRT to a larger extent than the use-it-or-lose-it allocation mechanism."We agree that synthetic excess spread should be capped at the level of 1-year regulatory expected losses, measured on the reference portfolio at inception – it being impractical to adjust the excess spread amount from year to year.
Furthermore, the requirement to use the trapping mechanism is potentially problematic when combined with the requirement to account for the trapped excess spread as a retained first loss tranche. In a benign credit environment where actual losses are below EL, trapped synthetic excess spread will build up and SRT may well be lost as a direct result of this mechanism as the value of the protection achieved by selling the first loss and/or mezzanine tranches would be reduced as the amount of trapped excess spread increased.
Also, the requirement in paragraph 123(a)(II) that excess spread not absorbed by losses in a given year should remain trapped in the transaction "in the form of a funded reserve account" is confusing. It suggests that there ought to be a separate cash funded reserve account established by the originator, which raises the question of who benefits from that fund after termination. We suspect this is just a terminological issue as the expectation in a synthetic would normally be that synthetic excess spread would be deducted from the periodic losses before those losses begin to be allocated to tranches. Accordingly, we would appreciate if the EBA could clarify that this would more appropriately be characterised as a "ledger" than an "account".
To the extent that the trapped excess spread on a synthetic is recognised as income/an asset by the bank, it may appropriate for it to be treated as a securitisation position with a 1250% risk weight and/or subject to capital deduction, at the bank's option. However, this then needs to be reflected in a reduction in the RWEA in respect of the more senior tranches (by factoring in the effect of the excess spread into the attachment point of each tranche for the calculation of the risk-weight when using SEC-IRBA or SEC-SA). The benefit to the senior tranches is realised in the context of the SEC-ERBA approach because the credit enhancement provided by the loss absorbing nature of the excess spread is reflected in a higher rating of the more senior tranches. On the other hand, in the context of the SEC-IRBA and the SEC-SA, treating trapped excess spread as a securitisation position subject to a 1250% risk-weight/capital deduction leads to unfair and overly conservative doubling up of capital requirements unless the attachment and detachment points of the first loss and mezzanine tranches are adjusted accordingly under SEC-IRBA and SEC-SA.
As with traditional securitisation, we would be grateful for clarification of the way in which "excess spread should be taken into account within the risk transfer self-assessment analysis" where this implies considerations beyond those articulated in paragraph 123(a) needing to be taken into account.
See also our response to Question 12 generally in respect of the treatment of excess spread.
11(a). In particular, do you agree with the proposal of only allowing a contractually fixed (pre-determined) excess spread commitment in synthetic transactions?
This is unproblematic in general, although we would clarify that it should be fixed as a percentage of the nominal amount of performing underlying assets each year rather than a fixed nominal amount of money. This is consistent with the idea that it should be limited to the 1-year expected losses on the portfolio.
11(b). In particular, do you agree with the proposal to only allow a ‘trap’ excess spread allocation mechanism in synthetic transactions?
No. See responses above in respect of Question 11 generally."
Question 12: Do you agree with the proposed way to treat the excess spread commitment in synthetic securitisation transactions for the purposes of the quantitative assessment of SRT and commensurate risk transfer?
The Discussion Paper says at paragraph 123. a. III. that “the total (unfunded plus funded) excess spread amount committed on a yearly basis should be considered within the SRT tests as well as the test of commensurate risk transfer, as a securitisation position subject to 1250% risk weight / capital deduction”. As stated above, members of the Joint Associations agree in principle that trapped excess spread on a synthetic securitisation must be treated as a securitisation tranche subject to 1250% risk weighting/capital deduction, at the bank's option. They wish to confirm that this would only apply to funded excess spread amounts and not to committed but unfunded synthetic excess spread amounts. It is worth mentioning that the amount of future, unfunded excess spread cannot be known in advance because the transaction could be terminated early and also because the asset amortisation schedule is inevitably subject to uncertainty.Members of the Joint Associations recognize the need to take account of the use of excess spread in a securitisation in the self-assessment an originator submits to its competent authority. However, rather than holding capital against untrapped excess spread or synthetic excess spread, whether by additional Pillar 1 funds or as a securitisation position, we believe excess spread should be recognised as a cost of protection as follows:
a. In a synthetic securitisation transaction with no excess spread feature, the bank would submit an analysis to the competent authority where the cost of protection would factor in – as a reduction in cost – the amount of losses that are offset by the benefit of the credit protection payments (as computed for the relevant loss scenarios).
b. In a traditional securitisation transaction where there is an excess spread feature, the bank would not take account of losses absorbed by the excess spread as a reduction in of the cost of protection (except for the portion of losses, if any, exceeding the excess spread amount).
c. In a synthetic securitisation transaction with a synthetic excess spread feature, similarly to (b), the bank would not take account of losses absorbed by the synthetic excess spread as a reduction in the cost of protection in its cost analysis (except for those exceeding the synthetic excess spread).
d. In all cases, only funded excess spread amounts would be subject to 1250% risk weighting/capital deduction.
We feel this approach ensures consistency with the treatment of future income on loan portfolios. The CRR explicitly confirms that capital need not be held against the possible reduction of future spread on a portfolio. We would submit that if an originator does not recognise any excess spread as income until the portfolio actually produces that cash, then there should be no need to recognise any potential losses that may be absorbed by that expected future excess spread stream.
Question 13: In relation to the further considerations for stakeholders’ consultation on the own funds treatment of excess spread: a. Do you agree that the unrealised/unfunded component of the excess spread commitment should become subject to Pillar I own funds requirements? b. What would be the impact on SRT transactions if Pillar I own funds requirements were recognised as suggested in Section 3.2?
13(a). In relation to the further considerations for stakeholders’ consultation on the own funds treatment of excess spread, do you agree that the unrealised/unfunded component of the excess spread commitment should become subject to Pillar I own funds requirements?No, for the reasons set out in our answers to the previous questions. Unrealised excess spread is not recognised as an asset and therefore there is nothing against which to hold capital. Rather, unrealised excess spread is simply future income that may or may not arise and it is not appropriate for such future income to become subject to Pillar I own funds requirements.
13(b). In relation to the further considerations for stakeholders’ consultation on the own funds treatment of excess spread, what would be the impact on SRT transactions if Pillar I own funds requirements were recognised as suggested in Section 3.2?
If Pillar I own funds had to be held against future excess spread (as opposed to only being held against trapped excess spread) this would lead to a significant reduction in the RWEA benefits it is possible to achieve from SRT deals, thereby making them less efficient and increasing costs to unacceptable levels. See an example at Annex I, which is based on a Joint Associations member's actual auto loan transaction and market pricing:
1. The transaction is a EUR 1bn rated traditional securitisation with seven classes of securities and a liquidity reserve account. It has a sequential pay structure with 10% clean up call and no time call.
2. In the example below, the risk weightings of the tranches are based on SEC-ERBA, assuming the transaction achieves STS status.
3. Mezzanine classes D, E and F are 95% sold to investors, meeting the mezzanine mechanical test for SRT. The coupons are based on market prices.
4. Retention requirements are met via retention of 5% of each tranche by originator.
5. We compare two scenarios:
a. Scenario assuming trapped deposit of excess spread equal to expected losses for the following year. Accordingly the funded excess spread account is totally consumed each year and no unused excess spread is carried over.
b. Scenario assuming no excess spread is trapped.
The conclusion of this analysis is that the trapped spread scenario results in a cost of protection 37% more expensive, with 37% less weighted average RWA reduction as compared to the non-trapped scenario.
Question 14: Are there any other safeguards or alternative regulatory treatments to address risks retained through excess spread in traditional and synthetic securitisation?
As described in our response to Question 12, excess spread used to absorb losses before it becomes cash flow to be distributed according to the waterfall should be recognised as a cost of the transaction and this cost should be justified by the originator in the context of its self-assessment of the cost of credit protection. The analysis should make transparent that the expected spread income minus costs on the underlying portfolio is sufficient to cover any excess spread. This is already an important safeguard to make sure that excess spread is not providing credit enhancement beyond what is covered by the spread income on the loans.Question 15: Should there be a specific treatment in those transactions featuring excess spread in which the originator, instead of achieving SRT in accordance with one of the SRT tests specified in the CRR, chooses to deduct all retained securitisation positions from CET 1 or apply a risk weight of 1250% to all of such securitisation positions (‘full deduction option’), in order to be allowed to exclude the securitised exposures from the calculation of risk-weighted exposure amounts?
The full deduction option should be retained as an additional straightforward alternative to meeting one of the SRT tests set out in the CRR. In this context, there is an even stronger argument for not treating excess spread as a securitisation tranche since the originator would already be holding sufficient capital to absorb losses of 100% on the portions of the underlying portfolio that it retains.Question 16: What are your views on the use of originator’s bankruptcy as an early termination clause? How does this clause interact with the resolution regime (i.e. the BBRD framework)? Should this clause be banned?
Early termination upon originator insolvency is, as the Discussion Paper points out, relatively market standard. It is included in the ISDA Master Agreement used to document most credit derivative CRM deals and in any case is often a requirement of investors. That said, members of the Joint Associations believe that there is nothing necessary about including originator insolvency as an early termination event, per se, and would not support it being required by the regulatory framework. The risk of originator insolvency for the protection provider is typically that insolvency will lead to the originator ceasing to pay the credit protection premium or ceasing to service the portfolio to an appropriate standard. Provided that the protection premium continues to be paid and the investors can remain comfortable that (a) the counterparty risk they assume on any cash prepaid to the originator to fund the protection is appropriately mitigated and (b) servicing will continue to an appropriate standard, originator insolvency need not be an early termination event.We note that some ways of achieving protection of the investor from the counterparty risk of the originator would include depositing cash in a manner that is segregated from the bankruptcy estate of the originator or including downgrade triggers requiring the deposit of cash with another bank or investment of cash in eligible collateral held at an independent custodian. These methods are only viable if they do not impair the SRT analysis. We also note that depositing the cash with a third party bank even under these circumstances would fall foul of the eligibility conditions for STS treatment set out in Article 270 of the CRR, which we do not view as a sensible outcome.
In terms of continuity of servicing we would suggest that originator insolvency not be a termination event in itself. Rather, it should be a termination event only when combined with at least one servicing failure in the portfolio and the failure to appoint a back-up servicer.
Question 17: Do you agree with the proposed originator’s self-assessment of risk transfer? Should such assessment be formulated differently?
The Joint Associations agree in general with this proposal, although it would be helpful to provide additional guidance and specification as follows:- Clearly doing self-assessments should not be required in relatively simple cases or where it is clear that the structural features described in the Discussion Paper do not impair SRT, as acknowledged in the Discussion Paper. Further guidance on the situations in which self-assessments are required would be helpful.
- Additional guidance on appropriate back-loaded loss scenarios would be helpful in order to ensure consistent application of this rule across jurisdictions.
- It is unclear whether the reference to total losses" in the main body of paragraph 165 (as opposed to 165b. where it is clear that the reference is to expected losses) refers to expected losses, unexpected losses or both. It would be helpful to specify.
- The references to PD and LGD in paragraph 166a may not be applicable to all portfolios (e.g. SA or IRB-F portfolios) so this element of the analysis should only be required where they are applicable.
- Cash flow models are complex and time-consuming to produce and we would suggest that alternative methods of demonstrating achievement of SRT be permitted where appropriate. For example, for some originators the level of stress used in the commensurate risk transfer test can be tailored to reflect base and stress scenarios, making cash flow models unnecessary. Instead, the originator can demonstrate risk transfer under base and stress cases using a variety of appropriate internal models (which may include cash flow models or modified risk transfer simulation models). Actual model submission could also be arranged on request by the relevant prudential regulator.
- We are somewhat concerned that the articulation of the considerations as set out in the Discussion Paper leaves a large degree of discretion for national competent authorities in assessing the SRT implications of the various structural features and their interaction. While it is necessary to allow a significant degree of discretion in order to have appropriate case-by-case analysis, we think there would be value in periodic central reviews of these decisions by the EBA in order to identify and reduce differences between jurisdictions in the SRT decision outcomes for comparable transactions.
As to the appropriate standard for stress tests, members of the Joint Associations would emphasise that the tests need to be granular, loan-level stress tests appropriate to the particular portfolio rather than wide, institutional stress tests. Further, it is important that the tests actually applied should be designed by reference to the risk management principles adopted by the particular institution. Where relevant, it may be appropriate to use EU-wide or Basel stress tests as benchmarks against which to judge the individual institutions' stress tests, but careful thought should be given to the appropriateness of this in each case."
Question 18: Are you aware of circumstances where institutions have entered into a structured risk transfer transaction which is not captured by Articles 243 or 244 CRR? For example, where the accounting treatment has meant a transaction is not considered for SRT assessment, or where transactions economically similar to SRT transactions do not fall into the definition of a ‘traditional securitisation’ or ‘synthetic securitisation’.
Members have not made us aware of any such transactions.Question 19: Do you agree with the proposed specification of the minimum first loss tranche thickness for the purpose of the first loss test?
Members of the Joint Associations agree that a test for minimum thickness of protection tranches may be workable, but are not convinced that it would be a helpful addition to the framework and we would favour omitting any such test from the regulatory framework for SRT.As a starting point, originators who do SRT transactions in order to obtain capital relief are automatically disincentivised from structuring transactions with thin protected tranches because that would result in very high risk-weights for the retained tranches especially in view of the built-in non-neutrality of the securitisation framework – thereby eliminating one of the main benefits of the transaction for the originator.
In addition, if there is to be a minimum first loss tranche thickness test, members of the Joint Associations would emphasise that this should apply for the purposes of initial assessment of SRT only. Even if the EBA chooses to assess SRT on an ongoing basis generally contrary to our suggestions above, it is important that any minimum first loss tranche thickness test not be tested in an ongoing manner. This is because – assuming a first loss tranche structured to meet the test precisely – SRT would be lost the moment there was a single loss on the portfolio, which may well happen soon after the transaction closes.
However, we have the following comments on the test proposed:
- Making the test one that is purely about minimum thickness of the first loss tranche would make SRT unnecessarily difficult to achieve. Instead, the EBA should consider a test based on minimum thickness of protection tranches in general (first loss and mezzanine) or a requirement to dispose of a higher threshold of the tranches representing expected losses.
- The test for the minimum thickness of the first loss or any other tranche(s) should be applied at the closing of the transaction only since the lifetime EL will change over the course of the transaction and it is generally not practicable to adjust tranche thickness after the transaction has closed. We also note that the definition of lifetime EL for the SEC-ERBA and SEC-SA will generally be zero on at closing unless the portfolio already contains impaired loans on day 1 (which will never be the case e.g. for STS securitisations).
- There is also an anomaly in the current mechanistic tests which is not corrected by the new definitions of first loss tranche and mezzanine tranche. This is illustrated below. In both examples, 80% of the EL and UL risk is sold to investors, which meets the substantive requirements of the EBA guidance. However, in the first example, the senior retained tranche has been divided into senior and mezzanine tranches (which is the most efficient structure under the SEC-ERBA and SEC-SA approach). Because the mechanistic test states that if mezzanine tranches are present, the originator must sell >50% of the risk weight of the mezzanine tranches, the mechanistic test fails, even though substantively the same amount of risk has been transferred to investors. We believe this is an unintended consequence of the mechanistic tests which should be corrected if the suggestions put forward in the Discussion Paper are ultimately agreed requiring that a significant portion of the EL and UL risk has been transferred.
- The minimum first loss tranche test requiring that the first loss tranche cover 100% of EL and 2/3 of UL should be modified to take into account a scenario such as the one illustrated below. In this graph, we see that the originator has transferred 100% of the EL and UL to investors. The first loss tranche does not meet the first loss test, but in substance, the combination of the mezzanine and first loss tranches sold is sufficient to transfer 100% of EL and 2/3 of UL. If not worded correctly, the first loss tranche test would inadvertently disallow the illustrated transaction.
Question 20: Do you agree with the proposed requirement of the minimum first loss thickness for the transactions assessed under the mezzanine test (i.e. transactions including mezzanine securitisation positions)? Do you consider this requirement relevant for all the approaches for calculation of securitisation own funds requirements (including e.g. SEC-ERBA)?
See discussion and examples provided in response to Question 19. While we agree with the EBA's suggestion that the first loss tranche should generally map to lifetime EL and the mezzanine tranche should map to lifetime regulatory UL, there are a number of transaction structures that do not fit neatly into this arrangement (e.g. where the first loss tranche is calibrated on the assumption that lifetime EL on the portfolio will be covered by actual excess spread). Many of these structures clearly and effectively transfer significant risk of EL and UL to third parties but will not necessarily meet the particular tests set out in the discussion paper.Accordingly, the test should be restated or clarified to take into account transactions in which a sufficiently large portion of mezzanine risk is sold to cover UL, and where the attachment point of the sold mezzanine tranches encompasses the portion of EL not covered by a thin first loss tranche. Certain members of the Joint Associations find that the optimal cost of protection and risk transfer (and consequent RWEA reduction) can be achieved if they retain a small first loss tranche, while transferring a sufficient amount of mezzanine tranches to cover the rest of EL and most of UL. To prevent the abuse that this test is intended to cover, the EBA could specify that the mezzanine test will achieve SRT if the originator transfers any combination of first and mezzanine tranches with a thickness equal to EL + 2/3 UL. Another alternative would be to require banks using the mezzanine test to sell a greater percentage of mezzanine tranches than 50%, such that the percentage sold above 50% equals the amount of EL not covered by the first loss tranche.
Question 21: Is a specification needed of the minimum thickness of tranches constituting mezzanine securitisation positions for the purpose of the mezzanine test?
Not per se, no. If there is to be a test, it should ideally be articulated to more explicitly achieve the underlying assumptions of the legislation. That is to say, selling any combination of tranches should be acceptable so long as the total risk sold is subordinated to the retained tranche(s) (if any) covers a sufficient proportion of losses to be sensible from a prudential perspective. Taking this approach would deal with the concerns raised in our response to Question 19 above.Another way to approach this would be to articulate the test in terms of the detachment point of the transferred position combined with a maximum retained first loss size.
Question 22: What impact do you expect the new CRR securitisation framework to have on tranches’ minimum thickness?
From the originator’s perspective, the new capital calibrations increase the capital consumption of retained tranches in an SRT transaction. For a given thickness of tranches sold to investors, all else being equal, an originator will achieve less capital relief under the new rules than the current rules. In order to achieve a similar level of capital relief, originators aiming to meet the first loss test will be forced to sell thicker junior tranches. In addition, if the new commensurate risk transfer tests are ultimately adopted, these will further serve to reduce the efficiency of capital relief achieved for most SRT deals, as even more junior risk may need to be sold to achieve SRT. Originators aiming to meet the mezzanine test will be forced to sell mezzanine tranches with higher attachment and detachment points in order to keep the mezzanine risk weight at a constant level and limit the increase in senior risk weights. In this case it is likely that more of the portfolio’s UL would be allocated to retained first loss positions leaving less of it allocated to the mezzanine piece. This seems contrary to the intention of the mezzanine SRT test. In addition, with more risk retained it is harder to pass the commensurateness tests (whether Option 1 or Option 2). In the end, a more likely effect in most cases is that banks will cease to do SRT deals because it is likely the incremental cost of selling additional tranches or mezzanine tranches with higher attachment and detachment points makes the transaction uneconomic compared to the bank’s cost of capital, and the sale of such tranches might cause the bank to fail the high cost of credit protection rules as proposed by the EBA. See discussion below on the effect of the Option 1 and 2 tests.From an investor’s perspective, the focus is to take on an investment with an attractive risk/reward profile. Assuming the investor is not regulated under the CRR, the new regulation is not key in the assessment of risks and rewards. The focus is rather on achieving a target return on the investment considering the investor’s assessment of expected and unexpected losses of the underlying portfolio. If an originator is forced to sell thicker tranches/more junior risk to achieve SRT, such a sale will not be economic for the bank if investors do not agree to accept a commensurately lower return for their investment. However, those investors most interested in first and second loss tranches may have yield requirements higher than the maximum weighted average cost of protection required for a deal to make economic sense to the originator. In order to reduce the all-in cost of protection, it will become more necessary to structure transactions with multiple tranches in order to attract different investors with different yield requirements. If the above-mentioned anomaly in the first loss and mezzanine tranches is not corrected, it will be difficult, if not impossible, for originators to sell a thick enough portion of tranches at an economic level.
For example, under the new CRR and commensurate risk rules, a bank may determine it is necessary to sell a position with a zero attachment point and a 12% detachment point in order to demonstrate SRT and commensurate risk transfer, which is greater than the zero to 8% tranche thickness that bank would need to sell to meet the same tests under the current CRR. It is likely easier and less costly to divide such position into three tranches; a first loss and two mezzanine tranches, because the bank would then be in a position sell to different investors with different return requirements. The first loss investor may need, say, 12%, while the senior mezzanine investor may only require, say, 4%. On a weighted average basis, the total cost of selling these three tranches may be, say, 8%. However, if the originator cannot meet the mezzanine test in this example, they are forced to structure one single first loss tranche with a zero attachment and 12% detachment point. It is unlikely first loss investors will be willing to accept an 8% return for such a tranche, and the originator may be forced to pay, say, 9% or 10%.
Of course, an originator always has the option to seek specific approval for SRT under Article 244(4) or 254(4) if it cannot comply with one of the mechanistic tests. However, as mentioned above, this process is often lengthy and uncertain, with some members reporting average response times for obtaining approval from the competent authority of 11 months. Many originators will abandon the effort to do an SRT deal rather than to take the risk of closing a transaction, only to have to call it 11 months later (with all the attendant written off costs and reputation risk incurred with investors) if the regulator does not approve the transaction.
Question 23: Do you have any comments on the test of commensurate risk transfer proposed under Option 1?
As explained in our responses to Questions 19, 20 and 21, the proposed minimum first loss tranche thickness test inadvertently causes several AFME members' SRT transactions to fail under the Option 1 approach before the commensurate risk transfer test is considered.Members of the Joint Associations are not opposed to the commensurate risk transfer test in principle and believe it could play a valuable role as an alternative way of demonstrating SRT (and not a cumulative one required in addition to the existing mechanistic tests). Equally, as mentioned in our answer to Question 3, it could play a useful role in providing a ‘cap’ on the capital relief that can be achieved via SRT. That said, there are a number of issues that members have raised with the current proposal for the test.
Ratio 1 ≤Ratio 2
((UL+〖EL〗_1Y )-K_retained)/((UL+〖EL〗_1Y))≤((UL+(〖EL〗_1Y×WAL))-((T_j×〖Retained〗_j )+(T_m×〖Retained〗_m ))-(〖ES〗_1Y×WAL))/((UL+(〖EL〗_1Y×WAL)) )
Where Tj is the thickness of the junior tranche, Tm is the thickness of the mezzanine tranche (if present), Retainedj is the % retained of the junior tranche and Retainedm is the % retained of the mezzanine tranche (if present).
The main issues that have been highlighted with this proposed approach to the commensurate risk transfer test are:
1) For the case where a mezzanine tranche is 100% transferred to an investor and the excess spread and first loss tranche are both under CET1 deduction or have 1250% risk weights, increasing the thickness of the transferred mezzanine tranche has no impact on the Ratio 2 numerator. Economically, it would be expected that more risk has been transferred, but according to EBA proposed method, increasing the thickness of the transferred mezzanine tranche could make the test more difficult to pass. The reason for this is that the thickness of any retained senior tranche would be reduced and potentially also decreasing the risk weights of the retained senior tranche, thus reducing Kretained and increasing Ratio 1 (this is especially important to consider when SEC-IRB is introduced).
2) A further issue with the formula above, is that only the retained portion of the junior and mezzanine tranches are subtracted from the lifetime EL and UL (in addition to the lifetime excess spreads also subtracted), thus the formula does not account for any retained portion of the senior tranche that also covers lifetime EL and UL. For example in case study 2, the retained senior tranche covers 0.6% of the total 4.6% lifetime EL and UL; however, as the numerator of Ratio 2 only accounts for the mezzanine and junior tranches, this 0.6% of lifetime EL and UL retained is not accounted for. This causes an unexpected situation where although only 3.2% of the whole transaction has been transferred to investors (80% of the 4% thickness of the junior tranche), the calculation for Ratio 2 states that the ‘Lifetime EL + reg UL on transferred position’ is equal to 3.8%. The result is that Ratio 2 in the case study is equal to 82.6% instead of 69.6% if the retained senior portion had been accounted for, making the test easier to pass.
3) In respect specifically of transactions structured to pass the mezzanine test, it seems that the objectives of the commensurateness test are inconsistent with the minimum tranche thickness requirement as applicable to transactions with mezzanine tranches. On such transactions, the assumption is that the lifetime EL of a portfolio is retained by the originator whereas the majority of the lifetime UL is transferred. Accordingly, it seems illogical that the lifetime EL should be considered in Ratio 2 in respect of mezzanine transactions.
4) Ratio 1 consists of 1-year measures whereas Ratio 2 uses lifetime measures. The contrasting time horizons serve to make the comparison less meaningful and useful. We would suggest dropping the lifetime measures in Ratio 2 and replacing them with 1-year measures to align the time horizons of the two tests.
In respect specifically of traditional securitisations, it is unclear how lifetime excess spread is meant to be calculated given that actual excess spread must be used and not a fixed amount. We would suggest that under such circumstances, lifetime excess spread should be deemed to be zero for the purposes of this test.
Question 24: Do you have any comments on the test of SRT and commensurate risk transfer proposed under Option 2? In particular, is the 50% threshold for SRT therein needed and appropriate?
Please see comments on Question 23 above. In addition, we have the following comments on this test:- In principle, if this type of test is accepted, it shouldn’t matter whether a bank sells 20% of the risk or 100% of the risk, since the RWEA reduction achieved will be commensurate with the risk transferred.
- In analysing the impact of this test on existing SRT positions, members of the Joint Associations have noted that, in many cases, the increased capital on senior retained tranches required by the new capital calibrations under the revised CRR causes their SRT transactions to fail, even though the positions sold to investors exceed EL + 2/3UL, which is the minimum suggested in the Discussion Paper. The effect of the new calibrations is to require an originator to sell a much larger portion of first loss and/or mezzanine tranches in order to reduce the risk weighting of the senior retained tranches below the threshold required by the test. For any particular transaction, some or all of the following negative impacts then may apply:
o It increases the cost of protection above the bank’s cost of capital, making the deal uneconomic to execute.
o It causes the total cost of protection to exceed the EBA guidelines on the cost of credit protection.
o It requires the originator to sell total tranche thicknesses far in excess of EL + 2/3UL, which makes such test redundant.
- The test in option 2 is even more problematic for rated and/or standardised portfolios. This results from the fact that the non-neutrality factor in the capital calibrations is more acute with securitisations involving the SEC-SA or SEC-ERBA approaches because of the need to divide the senior retained portion of an SRT securitisation into a series of senior and subordinated tranches, rather than one single unrated senior tranche.
Condition 1 should be amended to omit the 1-year excess spread element. This is because the amount of the first loss piece generally equals about the EL for the transaction's initial year, with excess spread paid thereafter at a rate that roughly equals the EL for the upcoming year. Including the first loss tranche and 1-year excess spread therefore amounts to double-counting. Calculation of the provisions on the underlying portfolio generally follows the same logic.
Condition 1 should also be amended to clarify that 50% of UL refers to 50% of regulatory UL (as opposed to lifetime UL).
Question 25: Should the SRT test be different depending on asset classes? Should it differ across STS and non-STS transactions?
No differentiation based on asset class is necessary per se (or indeed desirable, as the test should be standardised as much as practical), however there should be a differentiation in the test according to the capital approach an originator is required/chooses to use, in order to take into account:- The double capital penalty inherent in the SEC-IRBA and SEC-SA approaches for non-auto transactions with high excess spread portfolios, which distort the capital weighting of the tranches, which results in a need to sell a greater number or portion of tranches than is necessary to cover EL and UL.
- To the extent excess spread is required to be recognised as a securitisation position (trapped or otherwise), the credit enhancement feature of this trapping should be reflected in the attachment point calculations for the approaches, similar to the EBA’s proposal for taking into consideration the purchase price discount credit enhancement effect.
- The EBA’s proposal to require a 50% LGD assumption for standardised portfolios when calculating EL is a blunt instrument that will cause a lot of collateral damage with many asset classes, such as commercial real estate, project finance or other portfolios secured with collateral that effectively reduces the LGD of such portfolios. The 50% requirement will heavily penalise these portfolios unnecessarily and unfairly, especially in the light of the double capital penalty mentioned above, if ultimately the 50% LGD requirement is adopted. There may also be further problems resulting from the fact that some portfolios may be mixed, with both standardised and non-standardised assets. See also our answer to Question 30 in respect of the very conservative results of the SEC-SA approach hampering the use of synthetic securitisation as a technique.
Question 26: Could you provide, on the basis of SRT transactions that are part of your securitisation business, an assessment of the impact in terms of SRT achievement of the proposed requirements under both Option 1 and Option 2, taking into account the new CRR securitisation framework (Securitisation Regulation package)?
Certain individual members of the Joint Associations have elected to respond separately to this question rather than submitting a joint response in this letter. Please refer to those individual responses.Question 27: Do you agree with the assessment of the market practice of NPL transfer? Are there material aspects that are not covered in this representation?
In general terms, we do. The main reason behind the scarcity of NPL securitisations is the wide bid-ask price, which is in part caused by the costs of NPL resolution from the investor perspective. Factors such as the collateral costs, NPL management costs and judicial costs are influenced by divergent national regulation, and should be harmonised in the EU (via a Regulation rather than a Directive).It should be noted, in the context of the NPL market, that:
- The key outcome from the originating bank's point of view is that the NPLs are effectively transferred off of its balance sheet. The ability for the originating bank to achieve SRT is academic, because typically these transactions are not direct securitisations, but straight disposals.
- The key issue for banks, then is their ability to derecognise the assets, which is often complicated by that originating bank keeping some form of control over the assets (e.g. as servicer) even after they have been transferred to the acquirer.
Other key aspects of the NPL market to bear in mind include:
- The size of portfolios is typically smaller for NPL portfolios as compared to other traditional securitisations.
- The NPL market typically is characterised by more specialised market participants (borrowers, investors and special servicers) since deep expertise in servicing the particular type assets involved in essential to be able to optimise the recovery process on the underlying portfolio.
- The representations and warranties from sellers will typically need to be much more detailed and bespoke than for a performing portfolio.
Question 28: What conditions/initiatives would, in your view, facilitate the well-functioning of the NPL securitisation market?
In order to boost the securitisation of NPLs, the creation or extension of public guarantees targeting more junior risk (even 50% of the junior tranche as the ECB has proposed) would boost demand.The secondary market in NPLs could be further developed if certain obstacles are removed, such as:
- Barriers to entry for NPL investors: For example, some EU jurisdictions require loan servicers to hold a licence. This is a barrier to entry for entities that would otherwise be interested in becoming investors in NPLs.
- Limitations in the transferability of loans: Banks are often unable to sell loans containing restrictions on transfer/assignability of the loan. Policymakers could (at the national level) make such clauses unenforceable by operation of law once the loan ceases to perform.
- Disclosure limitations: in some jurisdictions, banks are not permitted to make disclosures to potential acquirers of loans that a mandate or conciliation procedure is underway, which makes it difficult or impossible for banks to make the necessary disclosures for acquirers to perform appropriate due diligence on the portfolio and means the bank can't give the relevant representations and warranties as to disclosure on the portfolio a reasonable acquirer would require of them.
- Requirement for banking licenses: Certain jurisdictions require a banking license to extend new credit facilities, which restricts the ability of the SPV to extend further credit as part of the management of the debt portfolio.
- ECB eligibility: permitting banks to use the senior tranche of NPL-backed ABS as eligible collateral for ECB repo operations would encourage investment.
Question 29: Which, in your view, are the core structural features that should be assessed within the SRT assessment of NPL securitisation transactions? Are the proposals on selected structural features of securitisation transactions proposed in this document (see Section 3.2.2) equally valid for NPL securitisation transactions?
Regarding the securitisation of NPLs, we understand the proposed framework is also appropriate. The only additional feature that may be needed in an NPL context is a mechanism to ensure alignment of interests between the servicer and investors.It should be stressed that it is important to ensure the “effective transfer of risks”, otherwise IFRS 9 (when it becomes applicable) could impose additional loan loss provisioning related to the expected losses of the transferred exposure.
Question 30: Do you agree with the proposed way of implementing the SEC-IRBA and SEC-SA approaches for the calculation of securitisation tranche capital in the presence of a non-refundable purchase price discount? Do you envisage other ways to implement the mentioned approaches in the presence of a non-refundable purchase price discount?
We agree with the way of implementing the SEC-IRBA. We note that this method should be generalised to cases where there is a non-refundable purchase discount irrespective of the type of underlying assets, i.e. not just used for NPL securitisation.We note as well that there is a wide discrepancy in this case between the results of the SEC-IRBA and the SEC-SA. De facto for NPL pools, the SEC-SA results are conservative; consequently, the SEC-SA cannot be used by market participants as it would make transactions unprofitable. This is quite detrimental for the development of the NPL securitisation market or the securitisation market of pools with material delinquency buckets. In our view this unfairly discriminates against standard banks.
The over conservatism of the SEC-SA for NPL pools stems in our view from the way the W factor is taken into account in the SEC-SA. This issue is not specific to NPLs but it is amplified in the case of NPL pools. This has already been described (with a simple correction proposed) in a paper called: “How to revive the European Securitisation Market: a Proposal for a European SSFA” by Georges Duponcheele et al. published in November 2014.
According to the paper: “It is to be noted that this official version of the SA in the BCBS 269 proposals contains a conceptual error in that it includes the delinquent assets (via KA) in KP. This in effect means that the capital associated with provisions for the delinquent assets is allocated to the senior tranches via the exponential function and the parameter α. To correct this conceptual error, KP should simply equal pool capital for performing loans and should not be a function of W. The delinquent assets should only affect the 100% capital charge threshold KT. It is important to distinguish between KT and KP to avoid making errors.” We therefore reiterate the need to correct the way the W factor is taken into account in the SEC-SA as detailed in the extract above.
In addition, first loss risk retention being done on a nominal / notional basis (i.e. option (d) of Art. 405(1)) for NPL securitisations may not allow for any downside risk exposure to the retention party where the purchase price discount is significant.
Application of the maximum risk weight for senior securitisation should take into consideration the purchase price discount applied. A possible method of doing this (assuming that the entire securitised portfolio is non-performing) could be:
SEC-IRBA:
Adjusted KIRBR Formula = [Max.(0, EL-PPD) + RWA (based on NPL Formula) x 8%)] / [EAD (based on GBV) – PPD]
Max Senior Tranche Risk Weight = Max. [(Adjusted KIRBR x 12.5), 15%]
SEC-SA:
Adjusted KSA Formula = [RW x Max. (0, (EAD (based on GBV) – PPD) x 8%)] / [EAD (based on GBV) - PPD]
Max Senior Tranche Risk Weight = Max. [(Adjusted KSA x 12.5), 15%]