In addition to the comments below, we would like to know how the interaction between economic and regulatory risk transfer is envisaged. When regulatory risk transfer is evidenced by the quantitative test (Article 244 (2) (a) CRR), regulators expect demonstration not only of regulatory risk transfer, but also of economic risk transfer to the same extent. This actually results in mandatory and parallel SRT evidence of economic and regulatory risk transfer. When considering the EBA’s proposal, we notice that the focus is primarily on regulatory risk transfer enriched with some economic analysis, but not on an equal economic and regulatory risk transfer test. We would like to know whether our understanding is consistent with the EBA’s intention.
Moreover, we have the impression that the EBA intends to define certain transaction features as standard features (e.g. amortisation, credit events). However, we wonder whether the standard will be mandatory for all future transactions and to what extent any divergent features are possible. In this context, we would appreciate a clear distinction between transaction features that are regarded as mandatory and features that banks might adapt to their needs and discuss with the regulator.
We understand that this discussion paper is supposed to be one of the sources from which the EBA draws the content of the report on SRT it has to deliver to the European Commission within two years of the date of application of the amended CRR. This is why the proposals made in this paper will become effective only after the Commission has incorporated them into a delegated act. For this reason, the deadline for application is largely uncertain. We, nevertheless, would like to suggest already at this early stage that, in order to protect the legitimate expectations of banks as originators, the modified STS requirements be applied only to transactions closed after the date of application of the aforementioned delegated act.
Comments to question 1:
We appreciate and agree with the market overview of synthetic and traditional securitisation transactions. Nevertheless, it is surprising, from our point of view, that Spain and Germany have no transactions that would need to be reported according to the EBA Guidelines. Furthermore, the number of non-SRT securitisation transactions would be a useful piece of information for the EU market.
Supervisory practices differ across Europe with respect to the use of economic and supervisory parameters (i.e. LGD) in the quantitative test (Art. 244 (2) CRR). We would like to know whether there is a common understanding that F-IRB institutions have to use supervisory parameters and A-IRB institutions are allowed to use economic ones. Heterogeneity has also been observed in the stress scenarios.
Further specification of the parameters and examples of the stress scenarios to be used would be helpful for originators and will facilitate the SRT assessment process.
We appreciate the proposals on the standardisation of the SRT assessment process. In this context, we would also like to emphasize that, once the proposals are agreed, they are binding on the relevant regulators and do not give them any further flexibility, e.g. ex ante notification period should not be extended by the responsible regulator: “ex ante notification by the originator of the SRT transaction at the latest 1 month before the expected issuance”. This applies particularly when considering that the ECB has currently set a period of 3 months.
Also, we welcome the EBA’s proposal that regulators should provide explicit point in time feedback to the originator even where no permission is required (e.g. quantitative test).
We hold the view that a standardised SRT notification template will help to facilitate the SRT process. Such a template should be different for traditional and synthetic securitisation in order to avoid further explanations of common features of synthetic but not of traditional securitisation, or vice versa.
Irrespective of the pure reporting of the SRT monitoring results, we would like to know whether the ongoing monitoring should be performed to the same extent as the initial assessment at the beginning of the transaction (or following any significant structural changes or SRT breach). The initial assessment is a lifetime assessment and includes highly comprehensive analysis covering difference scenarios and stress situations, whereas the regular monitoring (e.g. monthly) is embedded in the ongoing regulatory and economic capital calculation. This two-stage approach seems reasonable to us, considering that transactions are embedded in the bank’s capital monitoring processes. Without doubt, a comprehensive ad hoc assessment is required in the event of any significant structural changes or the imminent endangerment of SRT.
In our view, there is no reason why many of the forms of non-contingent premiums mentioned in paragraph 138 are to be banned in the synthetic securitisation market. In the true sale market, on the other hand, the way premiums are structured appears to be viewed as unproblematic, even if they are non-contingent (premiums for the equity or junior tranche are, for example, usually structured as non-contingent premiums). We therefore suggest basing the assessment less on the “contingent” or “non-contingent” criteria and more on whether the portfolio covers the transaction premiums and costs. The definition of “contingent premiums” should in any case be geared only to the current size of the tranche and not to its current risk as well.
With regard to paragraph 151, it should be made clear that there can be cases where there are only definitions “a” + “b” as credit events. Since “c” is often connected with derivatives and thus makes representation in the balance sheet as a financial guarantee more difficult, this definition should be allowed, but not be mandatory.
The securitisation market has a large number of safeguards for more senior tranches. These are different for externally rated securitisations of purchased receivables than for externally rated term loan ABSs or unrated synthetic securitisations.A rigid requirement for mandatory use of triggers would in this case not do justice to this heterogeneity, particularly as the types of triggers are defined in concrete terms, but not their level. We therefore suggest that the amortisation mechanisms outlined be considered as a non-exhaustive list of examples, which do not have to be applied mandatorily.
As already mentioned in our General Comments, we, however, welcome the description of standard features. These should be defined clearly to obtain a high degree of legal certainty when used. But it should naturally also be possible to use other features that would then have to be checked in the course of a risk transfer self-assessment.
Paragraph 102 (c) merely refers to the weighted average life (WAL) for determining the first call date. What is standard practice, however, is calculating the Macaulay duration to determine the call date or using the WAL for simplicity’s sake. We therefore suggest referring in the wording to both the Macaulay duration and the WAL.
We see no reason why paragraph 123 (a.I) should require the use in every case of an amount fixed for each year at the start of the transaction. This is at odds with the current understanding of excess spread (a firmly agreed excess spread amount would thus have to be paid even where the entire portfolio has defaulted). We would not therefore support this proposal.
We would like to make clear that in our view Excess Spread is to be calculated on the basis of the performing portfolio as of a fixed annual date.
Paragraph 123 (a.II) is, in our view, at odds with paragraph 121, which says that the trapped allocation mechanism tends to counteract successful SRT.
Paragraph 123 (a.III) should be deleted. How to apply the excess spread, when performing the quantative SRT test, is already explained and commented somewhere else. The same is true for the commensurateness test.
The requirement in paragraph 123 (a.IV) is, in our view, at odds, with (a.i), as the latter calls for specification of the excess spread amount at the beginning of the tenor. How the regulatory EL will evolve in the future is not yet clear at that point, however. We fail to understand the reason given for limiting excess spread. If correct computation of excess spread is questioned, the consequence could, alternatively, be a check by supervisors on the relevant calculations, and not dropping/limiting correctly computed excess spread amounts. Synthetic securitisation of portfolios with high excess spread would otherwise be virtually impossible.
Originators using the full deduction option are required to report the fact that they opted for full deduction. We believe that the information that has to be reported should be proportionate. It should be clarified that in the case of full deduction the originator is not obliged to report the information needed to analyse whether there is an effective risk transfer. This information is not necessary, as the originator fully deducts the retained positions from its own funds. Thus, it would not make economic sense to analyse the effectiveness of a significant risk transfer.
The representation of prepayments called for in paragraph 166 (c) requires an extremely high level of modelling detail. Such detailed modelling is neither necessary nor desirable (a model should be geared to abstracting essential features). A general requirement for this appears unreasonable.
With regard to paragraph 168, we should like to point out that the EU stress test usually defines requirements for the level of losses and the timing of their realisation. The freedom to back-load losses called for in paragraph 166 (b) then no longer exists. These provisions contradict each other.
In addition, the EU stress tests have specific focuses and do not stress all portfolios evenly. It should be made clear that these EU scenarios are not the benchmark when it comes to assessing a transaction internally.
The requirement in paragraph 169 does not appear practicable where, for example, the cash-flow model is integrated into an institution’s IT architecture (e.g. database connection to user interface).
As we understand it, standard transactions are subject only to the quantitative tests (Figure 17), whereas non-standard transactions are subject to the originator’s quantitative self-assessment (chapter 3.2.10). In this context, is the originator’s quantitative self-assessment the same as the additional risk transfer self-assessment (paragraph 82)?
No, a specification of the minimum thickness of mezzanine tranches is of no relevance. If a transaction wants to claim SRT, the presented tests qualify as safeguards to ensure appropriateness of mezzanine tranche thickness(es).
With respect to the new CRR securitisation framework, senior risk weights tend to increase dramatically, whereas mezzanine risk weights tend to be rather stable and first loss tranches remain to be deducted/with a 1250% risk weight. EL and provisioning of the underlying portfolio remain unchanged.
Originator’s perspective: In essence, the originator’s focus is on achieving economically efficient SRT. With senior risk weights being impacted most by the new regulation, originators might tend to structure senior tranches so as to keep risk weights on retained senior positions as low as possible. This will likely be achieved by higher amounts of subordinated tranche notional (including excess spread). In this respect, we rather expect senior tranche size to shrink and mezzanine as well as first loss tranche size (including excess spread) to increase.
Investor’s perspective: The investor’s focus is on taking 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 (economic) expected and unexpected losses of the underlying portfolio in order to assess the riskiness of the investment. We expect the impact of the new CRR securitisation framework on the investor’s minimum tranche thickness requirements to be marginal.
The following answer relates solely to transactions that are structured to pass the mezzanine test.
We agree with the underlying assumptions of the mezzanine test presented in paragraphs 173 and 174, i.e. that the originator intends to cover the EL over the lifetime of the transaction, while placing the majority of UL associated with the mezzanine tranche with the investor. Since first loss tranche plus excess spread are traditionally retained by the originator, we acknowledge the introduction of the minimum thickness test as a helpful measure to ensure that the lifetime EL of a portfolio is covered by the originator, whereas its UL is transferred. Given this intention, we miss transparency on why the lifetime EL is considered in Ratio 2 of the commensurate risk transfer test. In fact, paragraphs 216 and 218 reveal that Ratio 2 aims at measuring “the risk transferred to the third parties” obtained “as the share of losses transferred to third parties for the life of the transaction”. We understand that the commensurateness test is not directed solely at transactions that intend to pass the mezzanine SRT test. However, it is our understanding that the test of commensurate risk transfer contradicts the minimum tranche thickness test, particularly in the event that the mezzanine SRT test applies. This is due to the fact that the latter requires the lifetime EL not to be part of the transferred risk, whereas the commensurateness test considers it to be transferred.
Furthermore, we have difficulty understanding the reason why the commensurateness risk transfer test compares a 1y measure, Ratio 1, with a lifetime measure, Ratio 2. Even though it is clear why comparing capital savings to transferred risk takes place, it remains questionable to actually base the compared figures on different time horizons.
Synthetic excess spread has the high-level aim of increasing the high initial efficiency of a transaction and thus enhancing the transaction’s economics. Please note that the focus is on keeping the structure efficient for the originator, while keeping the risk profile of the placed portfolio, represented as the risk level of the mezzanine tranche, constant. The latter is achieved by the originator when posting excess spread in the amount of the expected loss on an annual basis. This procedure is comparable to the way an originator would annually provision the portfolio in the event that no securitisation takes place. Under current as well as future IFRS 9 accounting rules, booking the lifetime provisions of a performing portfolio at day 1 is inappropriate in terms of risk assessment. Given these conditions, we believe it is inappropriate to count the lifetime EL in one ratio and not to account for it in the other.
A transaction’s efficiency typically decreases when the portfolio nominal is reduced by redemptions and prepayments after the replenishment period. As a consequence, the average lifetime capital relief tends to be lower compared to its day 1 (and 1y horizon) equivalent. The EBA should thoroughly assess which horizon is best suited to meeting the purpose of the commensurateness test. The result will have significant implications for the likelihood of passing it.
In order to tackle the presented issues best, we believe that the lifetime component needs to be dropped in Ratio 2. If so, cutting out the lifetime component in Ratio 2 leaves the adjusted commensurateness test with ratios that are based solely on a 1y horizon. This would lead, in addition to the currently existing SRT test and the presented new test in condition 1 of Option 2, to a third test based on a 1y horizon. Taking into consideration the standard mezzanine test and condition 1 of Option 2’s new comprehensive test, the adjusted commensurateness test appears to be the least appropriate test. We therefore propose skipping the commensurateness test.
Furthermore, consideration of the lifetime excess spread as a retained position is an inappropriate assumption if, for example, in the case of a true sale there are no retained positions, i.e. the originator also has no claim to the excess spread. Ratio 1 would then be 100%, while Ratio 2 would be less than 100%, although the originator has outplaced all risks.
Yet even if the originator were to keep the claim to the excess spread, it would merely have retained chances and no risks, but would not pass the test. Consideration of the lifetime excess spread in Ratio 2 should thus be rejected (for true sales and synthetic securitisation).
In addition, including EL in risk assessment is generally problematic. The test can usually only be passed if not only UL, i.e. risk, but also EL is outplaced with investors to a considerable extent. Since, because of information asymmetry, investors always take a conservative approach to EL and EL always has to be paid for via premium payments, this requirement leads to a substantial increase in costs and thus endangers SRT elsewhere.
1y excess spread payments should not be considered in the calculation of condition 1. Generally, the amount of the first loss piece equals approximately the EL for the transaction’s initial year, the first excess spread (paid about one year after transaction start) equals EL for the second year, etc. Following this approach, EL is double-counted in this formula. Also, please consider that the calculation of provisions on the underlying portfolio follows the same logic.
In order to avoid any unnecessary complexity of the framework, no differing SRT tests should be used.
In NPL securitisation transactions, originators will often provide senior funding. So that this can also be treated as a senior tranche under the new CRR, it must obtain a risk weight below 25%. This is usually unobtainable for NPL securitisation transactions, however, as KIRBR is very high.
An NPL portfolio has defaulted in full (PD = 1) and has an average LGD of 42%. For this portfolio, KIRBR would be at least 42%, irrespective of whether the institution uses the IRB Foundation Approach or Advanced Approach.
Where KIRBR is 42%, the attachment point of the senior tranche must be around 85% to obtain a risk weight below 25%.
Senior tranches in NPL securitisation transactions are often considered for regulatory purposes as mezzanine tranches. This prevents application of the mezzanine test, for example.