Primary tabs

Arch Mortgage Insurance dac

We note a discrepancy with other market reports such as EBA / IACPM (2015) estimate of €40bn of synthetic tranches issued in 2014 and €60bn in 2015, compared with €45bn listed in this Discussion Paper (“DP”) for 2014-15 and €19bn for 2016. It would seem that the DP methodology fails to capture a significant share of deals, or that many have not been granted SRT, or that the industry has an exaggerated view of the use of synthetic deals for capital relief.
Also, it appears from data in this DP that with €19bn originated in 2016 SRT deals have failed to make significant inroads, compared to past volumes and more importantly compared to the €13 trillion of Euro loans outstanding.
As there is no shortage in the demand for capital from banks, the expectation is that either there is a large backlog of transactions waiting to be analysed by supervisors, or that the cost and the uncertainty of the SRT approval mechanism restrain banks from even applying.
It would have been useful to know how many institutions have completed SRT deals in the observation period. Our expectation, backed by Pillar III disclosures, is that such number is very small, questioning the “level playing field” auspicated by the CMU.
We also note that 85% of securitisations that achieve SRT are synthetics, therefore the proposed EC rule to exclude only traditional securitisations that achieve SRT from the leverage ratio calculation (art 429a point m) goes against market practice and is likely to remain underutilised if not amended.
Mortgage insurance is a loan-level tranched cover where a guarantor typically protects the first loss of a mortgage loan; such first loss size being the portion of the loan in excess of a threshold LTV, normally 80%. Article 234 CRR requires the use of the securitisation approach even for loan-level tranched covers, although the securitisation framework was developed and is mostly utilised for loan portfolios.
The new Basel d424 paper contains specific provisions for capital relief through “mortgage insurance” which would eliminate the need to use the burdensome securitisation approach to calculate the capital benefit for such tranched cover. We would welcome a clarification so that said Basel provisions can be applied at the soonest.
Art 234 CRR provides for a single loan which is subject to a tranched cover to be treated as a (synthetic) securitisation. An ongoing protection programme such as mortgage insurance (or such as the UK Help to Buy guarantee scheme) could thus generate thousands of synthetically securitised exposures every year, each one possibly representing an individual securitisation to be reported upon. A reporting standard should be devised so that tranched exposures of a similar nature (at least by type of protection, asset class and year of origination) can be grouped together for reporting purposes.
We note the concern on the use of non-contingent premium payments (138 a), however we think the analysis should be more refined.
The DP implies that upfront premium may be used as a way to defuse losses and inappropriately circumvent solvency rules. This concern may be founded when particularly risky or defaulted assets are being protected, and protection might be bought as a way to spread large, expected losses over time.
The most common case of protection purchase is however for performing loans. For performing loans, cumulative (upfront) premium is usually larger than expected losses because it has to cater for a significant part of unexpected losses. Spreading losses through non-contingent premium is therefore not a reasonable arbitrage strategy for performing assets.

There are sometimes practical implications in the use of payment structures different from contingent premium that should not be hindered by regulation excessively focused on preventing misbehaviour.
For example, a programme of mortgage insurance which protects the first loss on a mortgage (such as the Help to Buy guarantee scheme in the UK, or several other similar, privately funded schemes) would be treated as loan-level synthetic securitisation. Such a programme could produce thousands of synthetically securitised mortgages a year with a single lender.
In those circumstances, it is administratively much easier to collect a single upfront premium on each loan rather than monthly monitoring and calculating each loan contingent premium, and such single upfront premium will be agreed by the protection buyer and seller based on ex-ante expectations of losses and prepayments as much as it is done with contingent payments. Therefore an outright exclusion of non-contingent payments is in our opinion inappropriate.

In particular, if premium is paid upfront and it is a) fully paid by a third party such as the borrower or b) fully expended by the protection buyer on day one or c) fully deducted from capital for the portion that has been accrued and not yet expended, we see no reason to exclude these transactions from SRT recognition. This approach would be consistent with the provision to deduct from capital the unfunded portion of committed excess spread (section 3.2). Failing to recognize this similarity could lead to arbitrage opportunities between the two mechanisms.

We also note the DP further requirement (138 c) that the lender illustrates the process used to evaluate and purchase protection. That should provide sufficient confidence for the supervisor that (upfront) pricing was appropriate.
It is not clear from the text in (92 a) if all four conditions must be included as triggers for the switch to sequential, and the EBA should clarify if this is the case.

We also consider the requirement in (92 a iii) as possibly impractical in a number of cases. To the extent the trigger is determined by a worsening of an externally assessed portfolio rating (eg ECAI model), the provision might be acceptable. It is questionable whether intervening model changes should be allowed. This requirement is much less acceptable if the protection seller should rely on an internal assessment of the portfolio conducted by the protection buyer, which is hard to verify for the protection seller. Similarly to the model changes in the external ratings approach, it is questionable whether changes in regulation or reviews of PD, LGD such as TRIM with no connection to actual portfolio performance should affect an in-force transaction. It could be argued that the conservativism of the non-neutral securitisation approach is exactly designed to accommodate for future model changes.

In the case of retail loans, banks will usually lack information to determine rating migrations unless the loan is already in arrears or in default. For these assets, trigger 92 a iii should not be required.
A connection between the portfolio WAL and the minimum period before exercising a time call seems reasonable. Time calls are used to improve the efficiency of the transaction and this element should also be kept in mind, otherwise deals may not be justified. In a balanced approach, WAL should be based not on final legal maturity alone but also considering statistical features such as prepayments. Also, we suggest that a time call should always be exercisable after a sufficient period of time, say 5 years, has elapsed from inception regardless of WAL. This would reduce penalisation of long term loans such as mortgages and would be consistent with CRR Art 250 on maturity mismatch.
Part a): This proposal looks reasonable, fixed excess spread typically being expressed as a percentage of outstanding balance of performing loans

Part b): We note that the discussion in para 121 about the pros and cons of trapping vs “use it or lose it” is essential but unfortunately too short to allow a proper analysis of EBA assumptions and motivations.
Excess spread in a use-it-or-lose-it format is beneficial to the bank (protection buyer) over the trapping approach as the former allows the bank to book profits without a future risk of them being claimed back to cover losses in the future. KfW deals included synthetic excess spread in this format. The DP alludes to this scenario in para 125 but the wording is unclear and it does not elaborate much.
Looking at the proposed text in para 123, the prerequisite is excess spread providing credit enhancement. The definition of Credit Enhancement (CE) in CRR Art 3 (65) is very wide [1] and revolves around the improvement of the “credit quality” of a securitisation position. “Credit quality” is however not defined in the CRR. If improvement in the “credit quality” is to be intended as a lowering of the risk weight applied to the securitisation position, this should be explicitly stated in para 123. It would follow that deals based on SEC-IRBA and SEC-SA are not affected by this rule (and therefore can adopt “use it or lose it” excess spread) since excess spread is not an input in the formula that determines the capital absorbed by a tranche, as alluded in para 125. In those types of deal, excess spread can benefit the bank by reducing the cost of protection but cannot change its solvency position.
Also, note that the benefit of excess spread to the originator is only limited to the impact it has on the retained tranches. Any tranche which is sold will be risk weighted 0 and most of the first loss tranches have to be sold in order to achieve SRT. Para 121 argues that trapping provides better CE but to the extent the tranche is sold that benefit goes to the protection seller, not the bank, and so it can reduce the cost of protection but should not affect the originator capital position post transaction. So it is not entirely clear what EBA aims to achieve by favouring the supposedly “stricter” trapping – better protection of first loss investor? But these investors are usually sophisticated enough to make their own risk assessment and are hardly ever banks.
Together with the requirement that excess spread be trapped, EBA is advocating a capital build-up for trapped excess spread. Clearly, if use-it-or-lose-it was allowed, there would be no build up and therefore no capital charge. It is difficult to ascertain what EBA wants to achieve from excess spread. It would seem that trapping is mandatory because it is the only way to justify a capital add-on, but the whole rationale has not been adequately explained and it is therefore extremely difficult to comment upon.

Therefore, we favour the complete removal of bans on “use it or lose it” mechanism.
At the very least in para 123 the text “providing first loss credit enhancement” should be explicitly limited to SEC-ERBA transactions where excess spread improves the rating of some of the tranches and consequently the rating/risk weight of the retained tranches. Even in this case, it seems that EBA is questioning the way rating agencies give credit to excess spread, without providing much background information on EBA concerns and the impact on the bank solvency.
EBA could also consider to make trapping as the only acceptable solution (together with the build-up of a capital position) where the originator retains more than [20%] of the junior tranche which benefits from excess spread.
Note that the imposition of a trapping mechanism would create an arbitrary differentiation between synthetic and cash transactions which is likely to produce arbitrage or a skew in the adoption of one solution over the other. In facts, there is no explicit excess spread position in a cash transaction while losses are absorbed by excess spread on the equivalent of the “use it or lose it” basis, and normally not of “trapping”.
Also, in 123 a III) it is not clear if “total excess spread” refers to the amount of a single year or the cumulative amount for the expected life of the transaction or else. To avoid future discussions between banks and supervisors, clarification is needed. This accumulation approach could prevent the bank from earning excess spread until the end of the transaction, which is unlikely to be acceptable.
Finally, in the example in (106 a) we note fixed excess spread should be calculated on the balance of performing loans only.

[1] CRR Art 3 (65): “credit enhancement” means a contractual arrangement whereby the credit quality of a position in a securitisation is improved in relation to what it would have been if the enhancement had not been provided, including the enhancement provided by more junior tranches in the securitisation and other types of credit protection
Part a): We believe this text requires further expansion by EBA as there are too many possible interpretations about “commitment”, “unrealised” and “unfunded” and the period over which these statements are to be interpreted (e.g. one year or transaction life).

Part b): Section 3.2 is quite large and presumably reference is to para 124-127. Specifically, 126 suggests that an own fund requirement may be needed to account for the future, variable excess spread that may be pledged to the transaction. The assumption is that such future excess spread, though not specifically accounted for, would be available to the bank to be used against adverse scenarios had it not been pledged. We believe that for the time being, where the amount of pledged excess spread in existing and expected future transactions is negligible, the inclusion in Pillar I would lead to endless conversations with no material benefit and the clear disadvantage of maintaining another open area of uncertainty to the detriment of the CMU. Considerations on the amount of pledged excess spread would be well suited for the ICAAP analysis in Pillar II and, if the practice became material and better evidenced, it might at a later stage be moved to Pillar I.
From a credit protection provider, the key concern should be with underlying asset performance, independent of the originator’s fortunes. However, the originator is typically also the servicer of the assets, and its bankruptcy may lead to a deterioration of servicing which could adversely impact the credit protection provider.
Servicing risks following bankruptcy include: Key servicing staff moving to other players; cost cutting and therefore increase in case workload; lower staff motivation in pursuing cases, which may lead to more leniency in dealing with debtors, lower recoveries and also lead to more borrower defaulting as they fear less the consequences; restructuring/reorganizations leading to disruption in processes and missed legal recovery deadlines; possible loss of key documents, especially if premises are vacated or moved; higher complexity in legal recovery cases due to the bankruptcy status of the originator.
Originators could propose hot or cold back-up servicers in order to allay the concerns of protection sellers, but this would introduce a certain cost against a typically unlikely bankruptcy scenario.
Therefore early termination clauses based on originator’s bankruptcy should be maintained as an option.
The proposed text in para 165-166 should reflect that self-assessment complexity should be proportionate to the transaction complexity (para 164).
In para 166 the requirement based on loan PD and LGD cannot be applied by standardised lenders. Unless EBA intends to exclude those lenders from the potential benefits of SRT transactions, such requirement should be amended.
Giuliano Giovannetti
+447717716543