Making the Bank Securitisation Capital Rules Work for Europe

1 – Introduction

On 17th June 2025, the European Commission published the so-called “Securitisation Package”, a set of measures aimed at simplifying and improving the European Union (EU) securitisation framework. Part of the package[3] is a proposal to amend the Capital Requirements Regulation (CRR) which sets out how banks must calculate risk-weighted assets (RWA) for the securitisation positions (tranches) they hold.

This paper evaluates the above rule changes for representative securitisation transactions and considers how the changes might be made more consistent. The modifications we propose improve risk sensitivity, consistency in hierarchy of risks, and some technical flaws, resulting in a better level playing field across institutions and asset classes, supporting the European authorities’ objectives, as we understand them, to boost securitisation market activity through prudent, targeted rule changes.

Three important measures in the June 17th CRR proposal are (i) the introduction of a risk-sensitive risk weight floor (RWF) to replace the current fixed-value RWF, (ii) changes in the parameters of the p-factor which influences the degree of conservatism of capital charges when the floor does not bind, and (iii) the introduction of a new regulatory concept in securitisation, the notion of whether a senior tranche is ‘resilient’ or not.

On (i), in the current rules, capital calculated using the Securitisation-Internal Ratings-Based Approach (SEC-IRBA) or the Securitisation-Standardised Approach (SEC-SA)) is subject to an absolute RWF of 10% risk weight (RW) for Simple, Transparent and Standardised (STS) senior tranches and 15% RW for Non-STS or non-senior tranches. The RWF primarily affects low-risk senior tranches or upper mezzanine tranches with a sufficiently high attachment point. The use of an absolute or fixed-value RWF has significantly distorted the European market without an economic or regulatory justification.[4] The June 17th CRR proposal aims to rectify this by introducing a risksensitive RW floor equal to a percentage of pool RWs.

On (ii), under the SEC-IRBA and SEC-SA, the capital that a bank must hold against a securitisation exposure for which the RWF does not bind is computed using a formula known as the Simplified Supervisory Formula Approach (SSFA) which depends upon a parameter known as the p-factor. Under the SEC-IRBA, the p-factor is a weighted sum of variables representing deal characteristics, subject to a scaling, a floor and a cap depending on high-level deal features. Under SEC-SA, the pfactor is a fixed parameter which takes one of several values depending on the deal category. The June 17th CRR proposal would reduce the p-factor for specific sub-categories of securitisation positions, including those deemed to be ‘resilient’ (on which, see more below).

On (iii), some regulatory and prudential benefits are granted to senior tranches that satisfy additional criteria to be classified as ‘Resilient position’, a new notion designed by the European authorities to identify tranches that exhibit low levels of agency or model risk.

Under the June 17th proposal, the applicable RWF, and the p-factor parameters for SEC-IRBA and SEC-SA, depend on four binary categorisations:

(i) Whether the position is held by an Originator/Sponsor or not?

(ii) Whether the securitisation is Simple, Transparent and Standardised (STS) or not?

(iii) Whether the tranche is senior or not?

(iv) If the tranche is senior, whether it is a ‘resilient position’ or not?

Categorisation (iv) is an innovation. In this, the authorities explain that they wish to distinguish senior tranche exposures that exhibit low levels of agency or model risk. As explained in the CRR Explanatory Memorandum in the Securitisation Package, the criteria for a senior tranche position to be resilient are that:

1. The securitisation employs sequential amortisation (or pro-rata amortisation that switches to sequential if performance triggers are exceeded).

2. The granularity of the pool assets exceeds a minimum level.[5]

3. The senior tranche attaches above a minimum level.[6]

4. All credit enhancements are funded[7].

5. The bank holding the position is either an Originator or a Sponsor or, if the position is held by an Investor, the securitisation is STS. Note that three of the five characteristics of a resilient position (namely 1, 2 and 4) are already requirements for STS status under the existing rules.

If adopted, the proposed rules are likely, in our judgment, to stimulate the securitisation market and ameliorate some troublesome aspects of the existing rulebook. Nevertheless, to avoid distortion in capital charges vs. risk, to promote risk sensitivity, to develop a healthy market without arbitrage and unlevel playing field, there remain (i) aspects of the calibration and (ii) other features of the proposals that should be reconsidered if the proposed rules are to achieve the European authorities’ objective of prudently boosting securitisation activity.

For example, the distinction of whether the position is Originator/Sponsor-held strongly affects the p-factor parameters, and to a certain extent, the RWF to be applied. The authorities explain the emphasis on ownership by arguing that other holders, labelled ‘Investors’, are more subject to agency and model risk. We suggest that this justification is broad brush and is not rooted in evidence or an understanding of how agency and model risk operate in fact.

In our view, a simpler approach would have been preferable from the start, but we recognise that the legislative process is now underway. So, in this report, rather than a fundamental change to the riskweight formulas and functions[8], we focus on identifying targeted adjustments in the proposal that could serve the European authorities’ objective of implementing a prudent, risk-sensitive framework, allowing the development of securitisation across asset classes and Members States. Specifically, we suggest that the authorities:

(i) Replace “RWF = Scalar x KIRB or KA x 12.5” with the simpler, and more effective “RWF = Scalar x KSA x 12.5” to improve level playing fields between banks and between Member States. This has also the side benefit of reducing complexity.

(ii) Alter the definition of resilience by a. permitting Non-STS, Investor-held positions to be considered Resilient, provided they meet the attachment point, granularity and amortisation criteria. b. removing the “counterparty credit risk” criterion included in the Explanatory Memorandum of the Proposed CRR, in its discussion of “resilient positions”, which suggests an intention to prevent (re)insurers from participating in the Resilient market[9], or clarify that STS-eligible insurance can participate in such market,[10] c. for traditional securitisations, for SEC-IRBA and SEC-SA which do not recognise excess spread, to ensure consistency across banks and the Liquidity Coverage Ratio (LCR), setting the resilient attachment point for IRB banks[11] at the same level as for SA banks, i.e., AP = 1.5 KA for all banks, d. when SEC-ERBA is used, because excess spread is diverted to protect the senior tranche and such diversion are recognised by rating agency methodologies, there is no need for a resilient attachment point criterion for senior tranches. Indeed, the attachment point cannot be lowered arbitrarily as for a given rating, it is constrained by the rating methodology.[12]

(iii) Adjust the parameters of the Risk Weight Floor to be consistent with the low-risk indicators (Originator, STS and Resilient) so that STS implies generally lower RW than Non-STS, Resilient lower than Non-Resilient, and Originator/Sponsor lower than Investor, thereby enhancing banks’ incentives to structure deals satisfying STS and resilient position requirements, and introduce an absolute value cap in the RWF for consistency with nonsenior tranches,

(iv) Similarly, alter the p-factor parameters to make them consistent with the risk dimensions,

(v) Align the Securitisation-External Ratings Based Approach (SEC-ERBA) with (ii) & (iii).

The study is organised as follows. Section 2 sets out in a concise form the authorities’ June 17th CRR proposals. Section 3 shows the effect that the changes will have on SEC-IRBA and SEC-SA senior tranche Risk Weights (RWs). Sections 4, 5, 6 and 7 describe targeted technical modifications that would provide appropriate incentives for market participants, improve the level playing field across asset classes and Member States and reduce distortions. Section 8 explains how these would affect capital. Section 9 concludes the report.

2 – What do the authorities propose?

2.1 – SEC-IRBA and SEC-SA Proposals

This section summarises the June 17th securitisation capital rule proposals for European banks. The proposals include three significant innovations:

1. A risk-sensitive Risk Weight Floor (RWF) for all approaches,

2. Changes in the p-factor to reduce (but not eliminate[13]) capital non-neutrality in SEC-IRBA and SEC-SA,

3. A new distinction between ‘resilient’ and other senior securitisation positions.

Depending on whether a senior-tranche position is:

• held by an Originator/Sponsor or not,

• Simple, Transparent and Standardised (STS) or not, and

• resilient or not, the proposed rules imply different values for the capital charge parameters, specifically:

(a) The p-factor, employed in SEC-SA,

(b) The p-scaling factor that is applied to the p-formula()[14] used in SEC-IRBA to obtain the pfactor, and the minimum and maximum values for the p-factor, i.e., the p-floor and the p-cap.

(c) The scalar, i.e., the proportion of pool risk weight, and the minimum value (absolute floor) of the new risk-sensitive RWF.[15]

Here, a ‘resilient position’ is a new notion designed by the European authorities to identify tranches that exhibit low levels of agency or model risk. The criteria for a resilient position are (broadly):

1. It must be a senior tranche in a securitisation that either employs sequential amortisation or that includes specified performance-related triggers that determine a switch to sequential amortisation if conditions are appropriate. Requirements on the triggers are contained in the STS criteria in the Securitisation Regulation (SECR). In a synthetic securitisation, losses must also be allocated sequentially from the most junior to the most senior tranche, collateral must be returned to investors as tranches amortise, and the credit protection in respect of an exposure that is subject to a credit event must amortise only once the exposure has been worked out, or to the extent of any interim payment made.

2. On granularity, the maximum pool concentration (i.e., the aggregate exposure to any single obligor and its connected clients across the pool) must be no greater than 2%, the value currently applied in the STS criteria.[16]

3. The senior tranche must satisfy maximum thickness conditions (equivalent to the minimum thickness for its credit enhancement). Specifically, if the SEC-IRBA is employed, the senior tranche attachment point must exceed 1.1 times the sum of pool Lifetime Expected Losses[17] and pool Unexpected Losses. If the SEC-SA is used, the senior tranche attachment point must be no less than 1.5 times pool capital (with adjustment for delinquent assets), i.e., KA.

4. If the securitisation is synthetic, the tranche(s) providing credit enhancement to the senior tranche must comply with specified requirements of SECR relating to protection format. The relevant SECR requirements currently require (specified forms of) collateralisation except in relation to guarantees provided by (specified) sovereigns and supra-nationals. Under the SECR amendment proposals, eligibility under the relevant requirements will be extended to uncollateralised protection by insurers meeting specified safeguards.[18] The Proposed CRR Explanatory Memorandum, however, may prevent (re)insurers from participating in the Resilient market.

5. The senior tranche must either be held by an Originator or a Sponsor, or, if it is held by an Investor, the securitisation must be STS.

Condition 1 (sequential amortisation) is already included in the existing STS criteria (SECR Articles 21 and 26c). Condition 2 (granularity) is already included in the existing criteria for preferential STS capital treatment (‘STS+’) appearing in Article 243 of the CRR. Condition 3 (attachment point) attempts to prevent banks from holding such thick senior tranches that insufficient risk is transferred to credit protection providers, i.e., where the attachment point of the senior tranche is too low. Condition 4 (credit enhancement funding) is not relevant in this context since it applies to non-senior tranches, despite the purpose being to assign appropriate capital to a senior tranche.[19] It is unclear why the risk of a senior tranche should depend on the performance of another tranche.[20] The notion underlying Condition 5 (agency/model risk) is based on the assumption that, being held by an Originator or Sponsor reduces agency or model risk, and beyond that, the extremely prescriptive 100+ STS criteria also reduce agency and model risks for investors (but in the senior tranche only…).

For senior tranches, for SEC-IRBA, Table 2.1 summarises under the existing and proposed approaches the parameters to calculate the p-factor, i.e., the p-scaling factor to be applied to the p-formula[21], the minimum value (p-floor) and the maximum value (p-cap). The existing p-scaling factors are 0.5 and 1.0 for STS and Non-STS securitisation tranches. respectively.

The proposed p-scaling factors would be reduced to 0.3 for STS except for Non-resilient positions held by Investors (i.e., any bank holder other than an Originator or Sponsor) for which the p-scaling factor would remain at 0.5. The notion here is that for STS securitisations, agency and model risks are mitigated either by the holder being an Originator/Sponsor or by the position being ‘resilient’.

For Non-STS, being an Originator/Sponsor reduces the former p-scaling factor of 1.0 to 0.7 and being ‘resilient’ confers no advantage for senior tranches.

For STS securitisations, the p-factor is floored at 0.2, except when non-resilient and held by an Investor, in which case it remains at 0.3. Otherwise for Non-STS, irrespective of whether a tranche is resilient or non-resilient, held by an Originator/Sponsor or Investor, the p-floor is 0.3 as in the current rules.

The June 17th proposals introduce the notion of a p-cap, capping the p-factor to 0.5 for STS and 1.0 for Non-STS. The p-cap is a welcome introduction, as the Basel calibration of the ‘E’ parameter in the p-formula for the non-risk factor that is tranche maturity MT can lead to p-values abnormally high, disconnected from the risk of the underlying pool.

Table 2.1 summarises also, under the existing and proposed approaches, the Risk-Weight Floor (RWF) formula and its minimum value (absolute floor[22]) to be applied in computing SEC-IRBA capital for senior tranches. Note, that the proposed rules do not have a maximum value (absolute cap) for the RWF.

For SEC-IRBA capital treatment, the RWF in the current rules equals an absolute RW value of 10% for STS and 15% for Non-STS[23]. The proposed new rules constrain the RWF to be greater than both:

(i) A proportion of 10% and 15% of IRB pool capital, KIRB, multiplied by 12.5, for STS and Non-STS, respectively. The proportion coefficient is referred to in the proposed rules as the ‘scalar’.

(ii) Absolute floor, i.e., RW values of 5%, 7%, 10% and 12% for (a) Resilient STS, (b) Non- Resilient STS, (c) Resilient Non-STS held by an Originator or Sponsor and (d) other Non- STS.

Due to the combination of the risk dimensions for senior tranches (a) Originator / Sponsor vs. Investor, (b) STS vs. Non-STS and (c) Resilient vs. Non-resilient), there are theoretically 8 possible calibration cases. The one with the greatest numerical impact is the dimension of Originator / Sponsor vs. Investor, as the capital surcharge (via the p-factor) is reduced to 30% from 50% for STS, and to 70% from 100% for Non-STS if it is held by the Originator / Sponsor. There are no changes for Investor, except for the category STS and Resilient. The second most impactful risk dimension is STS vs. STS which affects primarily the RWF, via the scalar calibration. It also impacts the p-cap. The least important numerical impact is the resilient dimension, as it affects mainly the absolute floor of the RWF. Implicit in the order 5%, 7%, 10% and 12% of the absolute floor is the fact that STS Resilient is consider as ‘less risky’ than STS Non-resilient, itself less risky than Non-STS resilient, itself less risky than Non-STS Non-resilient.

For senior tranches, for SEC-SA, Table 2.2 summarises the existing and proposed p-factor and RWF parameter values to be used. The existing p-factor[24] is 0.5 and 1.0 for STS and Non-STS securitisation tranches. For senior tranches, the proposed p-factor is reduced to 0.3 for STS except for Non-resilient positions held by Investors for which the p-factor would remain at 0.5. For Non-STS, it is proposed that the current p-factor of 1.0 fall to 0.6 for Originator or Sponsor holdings and otherwise remain at 1.0; no advantage is conferred on Resilient positions.

For SEC-SA capital treatment, the RWF in the current rules equals an absolute RW value of 10% for STS and 15% for Non-STS (those values are thus both an absolute floor and an absolute cap). The proposed new rules constrain the RWF to be greater than both:

(i) A proportion (scalar) of 10% and 15% of SA pool capital (adjusted for delinquent assets), KA, multiplied by 12.5, for STS and Non-STS securitisations, respectively.

(ii) Absolute floor, i.e., RW values of 5%, 7%, 10% or 12% for a) Resilient STS, (b) Non- Resilient STS, (c) Resilient Non-STS held by an Originator or Sponsor and (d) other Non- STS.

SEC-IRBA and SEC-SA have many similarities, including the dependence on the three risk dimensions: 1) Originator / Sponsor vs. Investor, 2) STS vs. Non-STS and 3) Resilient vs. Nonresilient. Again, the combination of Investor and Non-STS and Non-resilient is not allowed due to the criteria on ‘resilience’, a notion reserved for ‘Originator / Sponsor’, except when it is combined with the STS label.

For non-senior tranches, the SEC-IRBA p-scaling factor and the SEC-SA p-factor remain at their current levels of 0.5 and 1.0 for STS and Non-STS, respectively. The SEC-IRBA p-factor benefits from the p-cap at 0.5 and 1.0 for STS and Non-STS, respectively. The p-floor is lowered to 0.2 from 0.3 for a non-senior tranche of an STS securitisation held by an Originator or Sponsor and otherwise remains at the current level of 0.3 (resilience not affecting the outcome).

For non-senior tranches, the RWF remains unchanged, i.e., takes a fixed value of 15% for both STS and Non-STS.

2.2 – SEC-ERBA Proposals

In SEC-ERBA, for tranche maturities[25] (MT), with a minimum value of 1 year and a maximum value of 5 years, the risk weight of a tranche is determined by a series of steps, the first one being a pro rata temporis interpolation of RW1Y, the one-year risk weight coefficient, and the RW5Y, the five-year risk weight coefficient.

5−MT RWTrancℎe = δ× RW1Y + (1 −δ) × RW5Y, where δ= min {max {( 5−1 ) , 0} , 1} (2.1)

For a given Credit Quality Step (CQS) of the tranche, the regulatory risk weight coefficients can be found in the CRR.

For non-senior tranches, an additional step is required whereby the value in equation (2.1) is adjusted for tranche thickness (i.e., multiplied by 1 minus thickness, with the senior tranche risk weight being a floor to this adjusted value.

We show in Table 2.3, the Proposed CRR “Table 4 (page 36)” calibration for SEC-ERBA for STS transactions, for the top 4 long term CQS. To simplify presentation in Table 2.3 Panel a), for STS transactions and for long term CQS, , we have grouped the originator positions (Resilient and Non- Resilient) with the Resilient investor position.

For Non-STS positions, the Proposed CRR has a “Table 2 (page 35)” with risk weight coefficients, that depend on the tranche rating (as expressed by its credit quality step (CQS)) and on the seniority of the tranche. It is to be noted that this latter table contains an obvious typographical error, in that the column headers ‘5-year’ and ‘1-year’ for ‘Non-senior (thin) tranches’ have been assigned wrongly and should be swapped (to match the lower 1-year risk weights compared to the higher 5-year risk weights).

In Table 2.3, to facilitate comparisons, we reproduce the top 4 CQS risk weight coefficients from “Table 2 (page 35)” of the Proposed CRR. These consist of fixed-value RWs for the lower CQS. However, for the top 2 CQS (CQS 1 (AAA) and CQS 2 (AA+)) formulae are employed like the risksensitive RWF with RWs equal to a scalar times SA pool RWs (adjusted for delinquent assets) with absolute-value floors. One would always expect that the 5-year RW would exceed the corresponding 1-year RW. Otherwise, an inconsistency arises.

Table 2.4 reproduces SEC-ERBA RWs from “Table 3 (page 35)” of the Proposed CRR for short-term CQSs. For the highest CQS a formula is employed like the risk-sensitive RWF.

2.3 – LCR Proposals

The June 17th Securitisation Package also proposes amendments to the Liquidity Coverage Ratio (LCR) Delegated Act to include the notion of resilience. Traditional STS securitisations, regardless of the underlying asset class,[26] are now eligible for inclusion in the liquidity buffers. ‘Resilient STS’ transactions are granted a lower haircut (15%), provided that their issue size meets a minimum amount of EUR 250 million and provided that their long-term credit rating is between CQS 1 and CQS 4. According to the proposal, “these criteria are key characteristics for the liquidity of resilient transactions to be exchanged against cash or sold in case of liquidity stress with no or little loss of value.”

Implicit in the ‘no or little loss of value’ is the view that a bank buyer will take on the purchased senior tranche. This assumes that the ‘resilient position’ sold by the bank seller remains a ‘resilient position’ by the bank buyer, so that there is no additional haircut to finance. However, if the bank seller is an IRB bank, the bank buyer will not be able to use the IRB parameters of the bank seller to assess resilience. Instead, resilience will be assessed using the SA parameters for SEC-SA (or the parameters relevant to SEC-ERBA).

In other words, if the attachment point criteria of resilience is based on IRB parameters (whether based on a ‘modified’ KIRB or KIRB itself) to enter the LCR liquidity buffer of an IRB bank, there is a high risk of rendering the liquid security into an instantly illiquid one as soon as resilience fails to pass the SA parameters for any bank buyer.

To avoid such a situation, for traditional securitisations, the ‘resilience criteria’ should be based exclusively on SA parameters, especially the attachment point criteria (see Section 5.1 for a discussion on the topic).

3 – What will the impact be?

This section analyses the impact of the proposed CRR for a set of representative securitisation transactions. Risk parameters are identified for four different asset types. Then these parameters are used to compute SEC-IRBA and SEC-SA RWs for senior tranches under current and proposed CRR rules.

Starting from the Pillar III disclosures of a large number of major European banks, Duponcheele and Perraudin (2022) provides estimates of Probability of Default (PD), Loss Given Default (LGD), Granularity for four different European asset types broken down by “all” and “high” credit qualities (see Table 3.1). The “all” credit quality asset figures are based on all loans for which the banks report estimates (excluding defaulted exposures). The “high” credit quality assets are based on loans for which the banks report estimates of PDs lower than 2.5%. Therefore, the “high” data can be considered as proxy for new securitisation issuances, whereas the “all” data can be considered as proxy for matured transactions. The “high” data is thus the most relevant for the securitisation market, whereas the “all” data provides a ‘what-if’ analysis.

The four different asset types considered are:

• Corporate

• Small and Medium Enterprises

• Residential mortgages

• Other retail

For the risk parameters estimated as just described, Internal Rating Based (IRB) capital is computed assuming the loan maturities M (in years) shown in column 6 of Table 3.1 (that are consistent with those in Duponcheele et al. (2014)). The senior tranche maturity (a non-risk factor needed for the SEC-IRBA and SEC-ERBA approach) is assumed to equal 5 years for all the asset types.

For the Standardised Approach, the risk weight of the pool (KSA ×12.5) is a reasonable estimate by the authors based on likely average pool values under the CRR 3. The delinquencies values in the column “w” for the “all” data are illustrative only to provide a ‘what if’ analysis, and to remain consistent with the increase in risk weights in the “all” vs. the “high” IRB data.

Also in Table 3.1 is the “Special case” of high-risk weight assets, such as pools of leveraged loans. The first row is relevant for ‘issuance’, whereas the second row is adjusted for long-term delinquencies (“w”), in a manner similar to the four asset classes mentioned above. The data for the ‘Special case’ is illustrative only.

Table 3.2 and Table 3.3 compare the RWs of securitisation senior tranches for the four representative pool asset types under the current and proposed regulatory capital rules and for SEC-IRBA rules (see Table 2.1) and for SEC-SA rules (see Table 2.2).

Table 3.2 shows the RW for a senior tranche under SEC-IRBA rules, with its attachment point set at the minimum under the rules, i.e., based on 1.1 times the ‘Modified KIRB’ (see Section 5.1), i.e., the sum of Life-Time Expected Losses and Unexpected Losses (UL).[27] Of the 8 logical possibilities of STS/Non-STS, resilient/non-resilient and Originator-Sponsor-owned/Investor-owned, one case, Non- STS-resilient-Investor-owned is ruled out by the proposed definition of Resilient position which requires either that a position be Originator-Sponsor-owned, or STS and Investor-owned.

In Table 3.2, for the four asset categories, under the current rules, the Resilient attachment point based on the ‘Modified KIRB’ is so low that the SSFA values, rather than the RWF, determines the RW of the senior tranche, for all 16 cases (“High”, “All”, “STS”, “Non-STS”). There are 3 exceptions regarding STS cases (“High SME”, “High Residential mortgage”, “All Residential mortgage”) where the RW is equal to the 10% fixed value RWF. However, the combined effects of lowering the p-scaling factor and the p-floor, as well as introducing a p-cap and a risk-sensitive RW Floor (including its minimum value) result in no senior tranche RW displayed being higher than the corresponding RW implied by the current rules.

The winners are the IRB retail framework asset classes (Residential mortgage and Retail – other). This is even more pronounced when held by the Originator / Sponsor. Investors only benefit materially if the position is both STS and resilient. There is little or no benefits for investors otherwise, especially for investors in Corporate and SME securitisations.

For the Special case of Leveraged loans, the Resilient attachment point is far too low for the RWF to override the SSFA, except with a high level of delinquencies. In that case the uncapped proportional RWF results in an increase of the RW of such senior tranche.

Table 3.3 shows the RW for a senior tranche under SEC-SA rules, with its attachment point set at the minimum under the rules, i.e., based on 1.5 times pool SA capital, adjusted for delinquent assets, KA. At this low attachment point, the SSFA produces a RW that is above the current fixed value RWF, except for residential mortgages, where the RWF is the constraint (hence the importance of introducing a risk-sensitive RWF for this category). The four asset categories and the Special case result in no senior tranche RW displayed being higher than the corresponding one implied by the current rules.

Under the proposed SEC-SA, Originator / Sponsors are the “winners”, across the board, whereas Investors do not benefit at all from the proposed securitisation reforms, except when a transaction is STS and Resilient. Benefits may occur only when the RWF is the constraint under the current rules, due to the introduction of the risk-sensitive RWF, and this will occur for a higher attachment point than the minimum set for the Resilient position criteria.

4 – What should be improved – use ksa in the rwf

The new proposals imply a framework that is much more complex than the existing one. This is justified by the greater risk sensitivity it permits which reduces distortions across asset classes and regulated entities. We believe that some targeted changes could be made which would simplify what is proposed while further reducing distortions.

A case in point is the use of different inputs to the risk-sensitive Risk-Weight Floor (RWF). A levelplaying field across Member States could be created by simplifying the RWF computation, by using KSA for both SEC-SA and SEC-IRBA approaches, rather than employing KIRB for SEC-IRBA as envisaged in the proposed rule changes.

If this suggestion were adopted, the risk-sensitive RWF for all banks in Europe would be given by:

(4.1) RWF= Scalar × KSA × 12.5

We do not believe that there is significant model risk in KIRB as calculated by European banks because of the rigorous regime of Margins of Conservatism (MoC) enforced by the Single Supervisory Mechanism (SSM) but it is nevertheless true that KSA as a wholly regulator-determined concept is more transparently free of model risk.

The use of KSA in the risk-sensitive RWF would also have the major advantage of creating a level playing field between SA and IRB banks. Because most Central and Eastern European banks, as well as smaller banks in Western countries, are almost all SA, this in effect would create a level playing field across banks business models and countries in different regions of Europe, a clearly desirable objective.

Note, also, that the use of KSA in the risk-sensitive RWF is preferrable to the use of KA. Indeed, KA is adjusted by the parameter w, the proportion of delinquent assets in the pool, which are implicitly risk weighted at 625%.[28] KSA is the standardised risk-weight of the non-delinquent assets (i.e., ‘good assets’). The value KSA is highly stable in a securitisation pool.[29] KA is an appropriate variable for the SSFA formula in SEC-SA, as delinquent assets will affect the riskier tranches, and we advocate the simplification of using KSA and not KA only for the RWF computation.

Indeed, situations when the RWF overrides the result of the SSFA formula means that the senior tranche is risk-remote. The tranches that do bear material risk are the ones where the RWF does not override the SSFA formula, and this is mainly for non-senior tranches. Therefore, by simplifying the computation of the RWF, there is no loss of risk-sensitivity. This pro-level-playing field RWF using KSA is still risk-sensitive to the risk weight of the underlying pool.

5 – What should be improved – define resilient better

5.1 – Change the Resilient Definition to Improve the EU Level Playing Field

5.1.1 – Attachment Point criterion

The definition of resilient position in the June 17th CRR proposal includes a minimum senior tranche attachment point. This is necessary because situations may arise with lower p-factor values in the SEC-SA and SEC-IRBA, in which banks employ very thick senior tranches and that inadequate capital is held against what would naturally be part of the senior mezzanine section of the seniority structure.[30]

To obviate this possibility, a minimum degree of credit enhancement for senior tranches may be enforced by specifying a minimum attachment point for senior tranches, above which a lower RWF may be safely applied. The 2022 Joint Committee (JC) report (see JC (2022)), proposed a complex approach to defining this minimum attachment point by estimating a threshold in the tail of the SSFA capital distribution. However, this method relied on elements of the p-factor and the SSFA formula itself (both of which introduce model risk).

The June 17th package adopts a simpler, model-independent approach to determine the minimum attachment point. This is preferable in that it relies less on models and is more transparent.

For SEC-SA, the minimum Attachment Point, APSA, is defined as follows:

(5.1) APSA = γSA × KA

Here, KA represents SA pool capital adjusted for delinquent assets.[31] The coefficient γSA is set to 1.5.

For SEC-IRBA, the minimum Attachment Point, APIRB, is defined as:

(5.2) APIRB = γIRB × Modified KIRB

Note, here, that the more usual notion of KIRB equals the sum of the 1-year Expected Loss (EL) plus the Unexpected Loss (UL), i.e., KIRB = EL+ UL. The quantity Modified KIRB alters the first term by multiplying it by the ‘initial’ Weighted Average Life (WAL)[32], a loose (one can say inaccurate) proxy for lifetime expected loss. The value of the coefficient γIRB, is then set at 1.1.

We predict that this approach will favour banks specialising in some asset classes, and handicap those specialising in others, without any obvious regulatory or economic justification. High-quality assets with high correlation risk (such as investment grade corporates or residential mortgages) typically have low KIRB. For these assets, the Minimum Attachment Point under SEC-IRBA will be materially lower relative to KIRB.

On the other hand, the securitisation of consumer loans or credit card receivables will become more challenging under this framework. For securitisations with pool assets of this type, the Expected Loss (EL) component tends to be high relative to the Unexpected Loss (UL). Indeed, when multiplied by the initial Weighted Average Life (WAL), the EL can even exceed the UL component.

In general, high spreads compensate lenders for the high Lifetime Expected Losses of such assets. Rating agency models recognise this and include structural features in their evaluation of traditional securitisations that allow the diversion of excess spread or interest proceeds to absorb losses. In synthetic securitisations, Synthetic Excess Spread (SES) serves the same risk-absorbing purpose.

However, the proposed CRR does not recognise excess spread (a key feature for asset classes such as auto loans, consumer loans, and credit cards) in the calculation of the Modified KIRB. This omission would hamper or even prevent the securitisation of such portfolios. Thus, the Proposed CRR for APIRB will distort markets, without any regulatory or economic rationale in our understanding.[33]

The fact that ‘initial’ WAL is such a poor proxy for lifetime expected loss means that the proposals as they relate to APIRB introduce noise/model risk into the securitisation rules. The notion of resilience should not depend on elements affected by model risk. Initial WAL is subject to model risk.

A design compatible with equation (5.1) for the minimum Attachment Point would be to keep KIRB as the base for the calculation as follows:

(5.3) APIRB = γIRB × KIRB

Appendix 1 goes into more details, should this option be retained for synthetic transactions.

This said, it is not free from model risk. Depending on one’s confidence in Margins of Conservatism, one may be concerned that model risk may also affect estimates of EL and UL, the two components of KIRB affected by model risk. There will also be a mismatch under the calculation of the Output Floor, i.e., a senior tranche that is resilient under IRB may lose its ‘resilient’ status once SA parameters are employed in assessing ‘resilience’.[34]

For traditional securitisations, the distortions introduced by APIRB provide strong arguments against the currently proposed approach. More importantly, the use of APIRB in the LCR (see Section 2.3) has the potential to create ‘fake liquidity’. To avoid such a situation, for traditional securitisations, the ‘resilience criteria’ should be based exclusively on SA parameters, especially the attachment point criteria.

A better design for the minimum Attachment Point would be to align IRB and SA banks by anchoring the minimum attachment point to the standardised capital metric, i.e., to use a formula such as: (5.4) AP= γ× KA

We suggest allowing γ to vary for different asset classes. Specifically, γ= 1.5 (the value already proposed for the SA) would apply to all asset classes except:

• Auto loans and leases,

• Very short-maturity assets (e.g., trade receivables with maturity < 1 year). For those two categories, we would set γ = 1.1 (the γ value in the Proposed CRR for the IRB approach), although more quantitative regulatory research may be warranted to confirm that this estimate is appropriate.[35]

This single, transparent rule would:

1. Eliminate the effect of winners and losers between IRB and SA banks and IRB banks specialised in different asset classes,

2. Remove unnecessary model complexity, and

3. Preserve risk sensitivity by using KA while maintaining a level playing field across Member States,

4. Make ‘resilience’ the same across approaches and a static characteristic of senior tranches rather than a potential source of capital volatility.

As far as SEC-ERBA is concerned, the criteria on the attachment point could be discarded altogether. Indeed, most traditional securitisations use diversion mechanisms of the interest proceeds (containing excess spread) to protect the senior tranche, and such mechanisms are the source of lowering the attachment point of the senior tranche. As only CQS 1 and CQS 2 rating will be allowed to use a risksensitive RWF, there is no need for an attachment point safeguard. Removing the attachment point criterion in SEC-ERBA (due to the presence of another mitigant, i.e., the external rating that limits the artificial lowering of the attachment point) would simplify the framework, without a loss of risksensitivity in the framework.

Finally, the focus on the attachment point by detractors of the reforms is misplaced. It will impact only a few securitisations, as can be seen in Table 3.1 and 3.2. Indeed, there is a mathematical relationship between the scalar of the RWF, the p-factor and the γ of the Attachment Point of the senior tranche. This is presented in Appendix 2. It can be shown that for most tranches attaching at the limit γ× KA or γ× KIRB, as proposed in the regulation, the RWF will be driven by the SSFA and not by the scalar, and thus the main constraint will be the p-factor. In other words, having an Attachment Point criterion is required only against aggressively low values of p-factor, i.e., p values that are mainly below 0.3.

5.1.2 – Other relevant criteria

The other two relevant criteria for resilience are minimum granularity (to reduce concentration risk in the pool) and a securitisation structure that amortises. The calibration of lower risk weights (or, in a risk-sensitive approach, the factor of proportionality (scalar)) relies on analytical tools that assess risks in the underlying pool. These tools are based on assumptions regarding pool granularity and the allocation of cash flows and/or losses. Therefore, if a lower RWF is to apply, the definition of a resilient position should reflect these assumptions.[36] That is the purpose of those two criteria.

In summary, the definition of resilience should be modified so that it is rooted in robust financial and mathematical principles, and, to the extent possible, the same criteria should apply to all.

Last but not least, we have assumed here that the term “senior” that is embedded in the concept of “Resilient position” is what is commonly accepted in the capital markets and in the regulation up to now, i.e., a tranche that has a detachment point of 100%. However, a surprising new definition is included in the Proposed CRR, which states in the Explanatory Memorandum: “Moreover, an amendment should also be introduced in the definition of the senior position, where an additional condition/clarification should be introduced that the senior tranche needs to attach above KIRB/KA.”

The reason behind this new definition is linked to a new SRT assessment, not in the scope of this paper. However, we expect this counterfactual definition to be highly problematic for the market as it will generate transactions that will have no ‘senior’ tranche, contrary to economic reality.[37] The riskweight of such economically-senior tranches will have to be based on non-senior calibration, creating major swings and pro-cyclical effects during an economic crisis. This is highly undesirable.

5.2 – Allow Banks as Investors to Invest in “Resilient Non-STS”

Of the five ‘resilience’ criteria, the distinction between Originator/Sponsors and Investors should be eliminated. The authorities assert this distinction is justified by the lower agency and model risks that Originators/Sponsors face compared to Investors and link resilience to whether positions are held by Originators/Sponsors or not.

The June 17th CRR proposal (see European Commission (2025)) states: “Only securitisation positions which feature reduced agency and model risks are eligible. This includes (i) positions by originators, both in STS and non-STS securitisation (as originators have more detailed knowledge of and control over the underlying exposures and the securitisation origination process than investors); (ii) positions by sponsors, both in STS and non-STS securitisations (as sponsors have access to more information than investors and agency risks are smaller than risks associated with investors’ positions).”

One should reflect on what ‘agency risk’ means? Agency costs arise when an originator either

(i) chooses assets to securitise that are of lower credit quality than securitisation tranche holders may guess by looking at the evidence presented at origination or (ii) manages the pool loans ineffectively so that the credit quality of the pool is worsened. These phenomena generate higher average pool PDs and LGDs than one might have predicted based on ex-ante evidence and this constitutes ‘agency risk’.

Well-established and regulated European banks are unlikely, we would argue, to engage in such practices. Indeed, this is why the delinquency rates in European securitisation loan pools in past downturns have not been enough to threaten the credit performance of highly rated tranches. The US loan market, particularly, in the period of the 2003-2007 mortgage price bubble, has been much more subject to such agency-risk abuses of the type documented in, for example, Ashcraft and Schuermann

(2008a) and (2008b).

To quantify the magnitude of agency effects for securitisation capital, Perraudin (2013) computes the increase in pool regulatory capital, as measured by KIRB, that is implied by the higher, postsecuritisation PDs, using the results of Elul (2011) who estimates US mortgage portfolio PDs for loans either (i) held on balance sheet or (ii) securitised. Perraudin (2013) finds KIRB for securitised loans to be around 10-20% greater than that for loans retained on balance sheet. Such estimates of agency risk are conservative as far as the European loan market is concerned since they are based on the loosely regulated, pre-GFC US subprime market. But they do shed light on prudent degrees of non-neutrality that may be built into a securitisation framework (through choices of SEC-IRBA and SEC-SA p-factor parameters (which, in turn, are used in calibrating the SEC-ERBA).

What do agency risks imply for whether regulation should favour Originator/Sponsor holders versus Investors? We would argue that a bank’s actions that generate agency risks will be little affected by the fact that it holds the senior tranche. The senior tranche will only bear losses in extreme states of the world. A bank engaged in a securitisation is much more focussed on shedding mezzanine and, possibly, junior tranche risk exposure. In so doing, it will still have an incentive to cherry-pick loans and cut corners in monitoring, if it is subject to these temptations in the first place.

In any case, past regulatory changes have been made to control agency risk. The minimum 5% retention implemented after the crisis was supposed to mitigate just this effect. And the p-factor is supposed to cater for model risk.[38] The STS label was introduced to cater for complexity risk, transparency risk, operational risk and credit portfolio risk quality in the underlying pool (and some model risk with the homogeneity requirements).

It is, thus, surprising that the proposed CRR adds to the previous text: “This includes […] (iii) investor positions - in STS securitisations only (because STS criteria largely mitigate the agency and model risks). Investor positions in non-STS securitisations are excluded as the agency and model risks are not reduced.” This paragraph reveals a clear issue in how policymakers envision the future of the European securitisation market. It appears that the “Non-STS” segment will be categorised exclusively as “Non-resilient” for Investors.

This represents a policy misstep that could significantly shape the market for years to come. Without a mechanism to promote “resilience,” the market will gradually and inevitably drift toward features that are undesirable in the long term — namely, non-granular asset pools, senior tranches with insufficiently high attachment points, and pro-rata structures lacking performance triggers.

We see such outcomes as undesirable. If “resilience” is to be introduced as a new regulatory concept, then banks as investors should be allowed to invest in Resilient Non-STS structures. A resilient structure is, by definition, more conducive to financial stability than a non-resilient one. Better Non- STS structures (i.e., Resilient Non-STS) could be encouraged by defining the granularity threshold for resilience at 3% concentration risk for Non-STS (compared to 2% for STS) (See Duponcheele et al.

(2024a) for a discussion on granularity). This approach would reflect that STS transactions, by definition, already meet this criterion.

It is also worth considering whether allowing bank investors access to Resilient Non-STS transactions (even with the same 2% granularity threshold as STS) was in fact the original intent of policymakers. Notably, “Summary Table 2” on page 13 of the Proposed CRR provides a calibration for Resilient Non-STS. If the intention were to prohibit bank investments in Resilient Non-STS, that section of the table would logically be left blank, but it is not. This suggests that at least some policymakers favour differentiating between Resilient and Non-Resilient within the Non-STS category. It is clearly in Europe’s interest to foster a securitisation market with resilient senior tranches in both STS and Non- STS transactions.

There is also a sound economic rationale. Smaller banks may find it difficult or even impossible to comply with the STS homogeneity criteria, simply because their portfolios are too small. Yet this should not prevent them from accessing the benefits of traditional securitisation. Supporting such participation would promote cross-border capital flows within the EU, especially in Member States where the banking landscape remains fragmented and reliant on smaller institutions.

5.3 – Allow (Re)insurers to Participate in the “Resilient STS” SRT Market

The current resilience criteria in the Proposed CRR present a significant issue for (re)insurers. As of 17th June 2025, the European Commission has finally taken steps to address the exclusion of (re)insurers from participating in the STS SRT market by introducing clause (aa) in the Proposed SECR Article 26e(8). However, the Proposed CRR includes an Explanatory Memorandum mentioning a criterion for “counterparty credit risk (only relevant for synthetic transactions)” within the definition of resilient positions, which effectively excludes (re)insurers from participating in Resilient STS SRT transactions.

This exclusion would confine (re)insurers to operating only in the existing Non-STS market or in Non-Resilient STS transactions. Worse, they could potentially be excluded from their existing market, the Non-STS market as many Originator/Sponsors held senior tranches would meet the Resilient position characteristics. The reforms are aimed at relaunching the securitisation market with more investors, including (re)insurers; they are not aimed at blocking mainly EU (re)insurers from intervening in Europe.

We see no reasonable economic justification for the explanation in the Explanatory Memorandum. It may simply reflect a drafting oversight.[39] In case it is not, let us address the point.

The Proposed CRR (See European Commission (2025a)) stipulates: “Counterparty credit risk (only relevant for synthetic transactions). Only credit protection supported by high quality collateral or in the form of guarantees provided by sovereigns or supra-nationals is allowed. This reduces the counterparty credit risk associated with the credit protection to which the originator is exposed, enables the originator to quickly compensate the losses incurred in SRT structures and contributes to the effectiveness of the risk transfer. The focus of this requirement is to protect the originator (and the originator’s exposure to the senior position), since in synthetic securitisations the senior position is usually retained by the originator.”

It is worth noting that the legal text refers to SECR Article 26e(8) (amended to include the new point (aa)) which allows insurers to participate in the STS market on an uncollateralised basis). This amendment does not reference the counterparty credit risk criterion. Nevertheless, we believe that both the spirit (Explanatory Memorandum) and the letter (Articles) of the law should be aligned.

By way of background, the Proposed CRR defines five criteria for identifying a ‘Resilient position’, four of which were inherited from the Joint Committee of the ESAs' report published in December

2022. The JC report (see JC (2022), aiming to align the CRR regulatory capital requirement for senior tranches with the de facto exclusion of insurers under SECR Article 26e(8), proposed introducing a “Counterparty Credit Risk (Synthetic only)” criterion, like the one mentioned above.

On this, Duponcheele et al. (2024a) argues: “We consider that this criterion should be removed altogether as it contributes potentially to financial instability. Indeed, the regulatory focus has been the protection of an originator against counterparty credit risk of the protection seller. While the requirement for collateral for STS securitisations appear well intentioned, little attention has been paid to how this collateral is financed by protection sellers, and who benefits. […] Within the context of the CMU, it cannot possibly have been the intention of the policymakers to exclude highly regulated, well capitalised and well-diversified (re)insurers from providing credit protection to banks when executing an STS transaction, while supporting such risk transfer when executing a Non-STS transaction.”

The issue is further examined in Bell et al. (2025a). The exclusion of insurers from participating in the STS SRT market on an uncollateralised basis was not debated when the synthetic STS framework was hastily introduced as part of the Capital Markets Recovery Package (CMRP) during the COVID-19 crisis. It is now clear that the European Commission intends to address this oversight in the Proposed SECR by inserting a new clause (aa) in Article 26e(8), allowing insurers to participate in the STS market on an uncollateralised basis, subject to four key risk mitigants.[40]

However, it is likely that the section on resilience criteria in the Proposed CRR Explanatory Memorandum was drafted before the modifications to the SECR Article 26e(8). As a result, the combination of both texts would imply that insurers may only participate in the non-resilient STS market, since the resilient category remains restricted by the “Counterparty Credit Risk (Synthetic only)” criterion, originally formulated in December 2022. To enable insurers to access the resilient STS market, a simple cross-reference to Article 26e(8) in the Proposed CRR Explanatory Memorandum would be sufficient to align the two texts.

Nevertheless, we believe it would be preferable to remove this criterion entirely. Doing so would not only simplify the definition of what constitutes a resilient position but would also eliminate a criterion that lacks any sound justification, as explained in detail in Bell et al. (2025a). Furthermore, beyond the specific issue of insurer participation, the credit quality of a senior position is not at all dependent on the presence or absence of collateral or the credit quality of that collateral. It is determined solely by the quality of the underlying assets.

From this perspective, we believe the “Counterparty Credit Risk (Synthetic only)” criterion is not relevant to assessing the resilience of a senior tranche.[41] Therefore, combined with the above discussion in Section 5.2, being an Originator/Sponsor, or an Investor, is a policy-based risk dimension. It is not a criterion for Resilience either. It is a risk dimension relevant to calibration. This means that the notion of a ‘Resilient position’ can be simplified to three vital criteria: 1) sequential amortisation of tranches, 2) sufficient pool granularity, and 3) sufficiently high attachment point for the senior tranche.

6 – What should be improved? revise the calibration

6.1 – The Three Dimensions and Calibration Principles

In the Proposed CRR, the calibration of the risk-sensitive RWF is a function of three dichotomies, namely:

1. whether or not the bank holder is an Originator/Sponsor,

2. whether or not the securitisation is STS,

3. whether or not the position is resilient.

On the first dichotomy, the European Commission’s impact assessment of the proposed CRR and proposed SECR (see European Commission (2025c)) states: “increasing the risk sensitivity of the existing capital framework through targeted reductions to the risk weight floors has been a central recommendation by the EBA. In their analysis, the EBA fully acknowledged the reduced model and agency risk associated with some types of transactions (i.e. tranches retained by originators) and proposed that the reduced risks should be recognised through lower risk weigh floors for these types of transactions, when accompanied by a set of safeguards ensuring their resilience.”

In this framework, an entity other than an Originator is effectively classified as an ‘Investor’. When a bank acts as a Sponsor, the question arises as to whether it should be treated as an Investor or as an Originator within this dimension. The concept of Sponsor is particularly relevant for Asset-Backed Commercial Paper (ABCP) conduits. Given that Sponsors typically possess deeper knowledge of the underlying assets, it might appear reasonable to classify them alongside Originators.

Nevertheless, the Originator/Sponsor dimension is ultimately a proxy for the underlying factor of interest: the quality of origination. In the EU, regulated banks must comply with the EBA loan origination guidelines. Therefore, EU-based Originators/Sponsors have access to the necessary information to ensure sound origination practices.

Building on this, the Investor category could be further refined by distinguishing between securitisations backed by assets originated by:

• EU banks (subject to EBA origination guidelines), and

• Other institutions, such as non-bank financial institutions (NBFIs) or non-EU banks (not subject to the EBA guidelines).

Such a refinement would offer a more logical approach. Securitisation is a technique that transforms the risk of underlying assets. It does not itself create or mitigate risk. The quality of a securitisation market is determined by the quality of the securitised assets. Therefore, differentiating between regulated and non-regulated origination would be a welcome enhancement which regulators may reconsider in future.

The second dichotomy is whether a securitisation carries the STS label. Duponcheele et al. (2024b) discuss the rationale behind the EU's introduction of this distinction.

The third dichotomy is whether a position qualifies as Resilient. This concept was discussed in Section 5. We assume here that ‘resilience’ is an objective status, determined by three key criteria:

1. A sufficiently high attachment point, defined as AP= γ × KA,[42]

2. A minimum level of granularity in the underlying asset pool, and

3. A securitisation structure that amortises sequentially (or pro-rata with performance triggers that switch to sequential amortisation).

Based on these three dimensions, the calibration of the Risk Weight Floor (RWF) should follow these principles, ceteris paribus:

• The RWF should be lower for an Originator-held tranche than for a Non-Originator tranche.[43]

• The RWF should be lower for an STS than for a Non-STS securitisation.

• The RWF should be lower for a Resilient position than for a Non-Resilient position.

The risk-sensitive Risk Weight Floor (RWF) is itself subject to a fixed minimum value, the “floor of the floor”. To avoid confusion in terminology, we refer to this fixed level as the “Minimum Value RW”. Based on this minimum, we can derive the Implied Sensitivity Threshold. It represents the risk weight of the underlying asset pool below which the Minimum Value RW overrides the risk-sensitive RWF. Below this threshold, the RWF ceases to be sensitive to the credit quality of the pool and becomes dependent only on the three structural dimensions: Originator, STS status, and Resilient position.

Since the rationale for introducing a risk-sensitive RWF is to facilitate the securitisation of highquality assets (as reflected in the low-risk weight of the underlying asset pool), it is important that the Implied Sensitivity Threshold be explicitly calculated and aligned with the calibration principles outlined above.

In the Proposed CRR, the calibration of the p-factor is also a function of the same three dichotomies as for RWF Floor. Although the p-factor has three distinct roles (capital non-neutrality, asset correlation, smoothing of a cliff-effect – see Duponcheele and Perraudin (2025)), the focus of the Proposed CRR is on “capital non-neutrality”. It states: “A (p) factor of 1 should be interpreted as a 100% higher capital requirement or a doubling of the capital requirement for all securitisation positions, compared to the capital requirement of the underlying non-securitised assets, while a (p) factor of 0.3 results in a 30% higher capital requirement. [,,,] The targeted amendments therefore differentiate between positions in STS and non-STS securitisations, originators/sponsors and investors positions, and senior and non-senior positions.”

The resilient dichotomy is mostly absent in the Proposed CRR p-factor calibration, intervening only for the combination of the dimensions STS and Investor.

6.2 – The Calibration in the Proposed CRR: Strengths and Inconsistencies

6.2.1 – Strengths and Inconsistencies of the Proposed Risk-Sensitive RWF

Table 6.1 shows the eight possible combinations of the three dichotomies. The table displays Implied Sensitivity Threshold (IST), the level of RWs below which the risk-sensitive RWF is overridden by the Minimum Value RW in the proposals. The IST RW is easily calculated as the ratio of the Absolute Floor expressed as RW to the scalar expressed as percentage.

The calibration in Table 6.1 is in line with some of the calibration principles mentioned earlier, but not all.

The STS scalar (10%) is materially lower than the Non-STS scalar (15%), a difference of 5% based on the STS label, a ratio of Non-STS to STS of 1.5, the same ratio under the current rules for Non- STS RWF to STS RWF.

For Originators/Sponsors, the Minimum Value RW increases from 5% RW to 7% RW to 10% RW to 12% RW from Case 1 to Case 4. This means that the IST is 50% RW, 70% RW, 67% RW and 80% RW respectively. The Case 3 IST is lower than the Case 2 IST, this is not consistent with the calibration principles.

Some inconsistencies remain which could adversely affect the policy objective of shaping the European securitisation market.

• Case 1 (Originator – STS – Resilient) has the same scalar as Case 2 (Originator – STS – Non-Resilient). This means that for any portfolio above the IST of 70% RW, there is no incentive to create a Resilient position. SME loans typically have 75% RWs under the Standardised Approach. As the p-factor is undifferentiated between Resilient and Nonresilient positions between Case 1 and Case 2, and as the attachment point of a Resilient position will be above that of a Non-Resilient position, there is a clear incentive to create Non-Resilient positions in SME securitisations. The solution to remove this wrong incentive is to follow the calibration principles by ensuring that the scalar is lower for Resilient than for Non-Resilient positions.

• Case 1 (Originator – STS – Resilient) has the same scalar as Case 5 (Non-Originator – STS – Resilient). Case 1 (Originator – STS – Resilient) has the same scalar as Case 6 (Non-Originator – STS – Non-Resilient). Case 1 (Originator – STS – Resilient) has the same Minimum Value as Case 5 (Non-Originator – STS – Resilient). In each case, the former should be lower.

• Similarly, Case 2 (Originator – STS – Non-Resilient) has the same scalar as Case 5 (Non- Originator – STS – Resilient) and Case 6 (Non-Originator – STS – Non-Resilient). Case 2 (Originator – STS – Non-Resilient) has the same IST as Case 3 (Originator – Non-STS – Resilient). Case 4 (Originator – Non-STS – Non-Resilient) has the same IST as Case 8 (Non-Originator – Non-STS – Non-Resilient). Case 5 (Non-Originator – Non-STS – Resilient) has the same scalar as Case 6 (Non-Originator – Non-STS – Non-Resilient). In each case, the first of the two compared should be lower.

• Case 7 (Non-Originator – Non-STS – Resilient) exists as a dimension, but it is not allowed in the Proposed CRR. The text describing the agency risk criteria for resilience stipulates: “Investor positions in non-STS securitisations are excluded as the agency and model risks are not reduced.” It is clear to all market investors in securitisations that a senior tranche of a CLO of leveraged loans to European corporates would qualify as being in ‘Case 7’ and is far safer than a senior tranche of a low granularity CMBS with refinancing risk with the same underlying credit quality, that would qualify under ‘Case 8’. To amalgamate transactions that would qualify in Case 7 as being the same as Case 8 is not desirable. It would be much better to authorise this classification and to make the table consistent with the calibration principles, both for the scalar and the IST.

6.2.2 – Strengths and Inconsistencies of the Proposed p-Factors

Table 6.2 shows the eight possible combinations of the three dichotomies and display the SEC-IRBA p-factor parameters and the SEC-SA p-factors.

In SEC-IRBA the Proposed CRR does not differentiate the p-factor parameters (p-scaling factor, p-floor and p-cap) based on the Resilient position status of the senior tranche (Case 1 and Case 2, Case 3 and Case 4, Case 7 treated the same as Case 8), except for STS positions held by Investors (Case 5 vs. Case 6). The same applies to the p-factor in SEC-SA.

Another inconsistency is the fact that the SEC-IRBA p-scaling factor for Originators/Sponsors holding Non-STS positions (Case 3 and Case 4) is numerically different (0.7) compared to SEC-SA p-factor (0.6) for the same case.

Finally, Case 7 being not allowed, senior tranche exposures with those characteristics will apply the p-factor parameters of Case 8, which remain unchanged in the Securitisation compared to the current rules. This penal treatment that has affected only the European market is expected to continue for another five years. We think that this is a policy mistake to push for a de-facto Originator/Sponsor friendly market and an Investor unfriendly market; a market needs both sides.

6.3 – Targeted Amendments to Make the Proposed Architecture Consistent

6.3.1 – Targeted Amendments to the Proposed RWF

This section 6.3 considers a recalibration of the risk weight floor and p-factor that would be consistent across all risk dimensions.

Table 6.3 presents a modified calibration for the risk-sensitive RWF which respects the architecture of the Proposed CRR while following the calibration principles set out above. Bold text are numerical amendments to the proposed rules that would improve incentives in the European market.

The objective here is to shape the market in accordance with policy objectives: Originator being better than Non-Originator, STS being better than Non-STS, Resilient being better than Non-Resilient. This applies to both the RWF, the Minimum Value, and the Implied Sensitivity Threshold (IST).

The process employed in deriving the numerical values presented in Table 6.3 is as follows.

• We began with the benchmark established by the Proposed CRR for the riskiest combination: Non-Originator, Non-STS, and Non-Resilient (Case 8). This combination remains unchanged. The second benchmark is the combination Non-Originator, STS, and Non-Resilient (Case 6). This case also remains unchanged, as it reflects a material distinction between STS and Non- STS transactions.

• To align Case 5 with the calibration principle relative to Case 6, we reduce the scalar from 10% to 8%, while keeping the Minimum Value unchanged from the Proposed CRR. The missing calibration for Case 7 is interpolated between Cases 6 and 8. We assign it a scalar of 12% and a Minimum Value of 9% RW. These adjustments ensure internal consistency across the Non-Originator dimension (Cases 5–8), both in terms of scalar, Minimum Value, and IST.

• Next, we align Case 1 with Case 5 by setting a scalar of 6% (i.e., below the 8% of Case 5) and a Minimum Value of 3% RW (i.e., below the 5% RW in Case 5). These values preserve the IST as defined in the Proposed CRR, at 50% RW. We then align Case 2 with Case 6 by setting a scalar of 8% (i.e., below the 10% in Case 6) and a Minimum Value of 5% RW (i.e., below the 7% RW in Case 6).

• Case 3 is calibrated in relation to Case 2, with a scalar of 10% (i.e., above the 8% in Case 2) and a Minimum Value of 7% RW (i.e., above the 5% RW in Case 2). These values remain below those in Case 7, ensuring consistency. Finally, Case 4 is aligned with Case 3, with a scalar of 12% (i.e., above the 10% in Case 3) and a Minimum Value of 9% RW (i.e., above the 7% RW in Case 3). These remain below the corresponding values in Case 8.

Once the calibration in Table 6.3 is made consistent, one may ask whether these values are reasonable. This can be assessed by referring to Table 3.1 of Agarwal and Perraudin (2024). That study presents scalar ranges per asset class. According to their findings, a suitable floor generally lies around 10% of pool RWs.

Duponcheele et al (2024a) discusses the treatment of STS and Non-STS transactions and supports a central scalar value of 10% with a ±5% variation around this value, suggesting a 7% scalar for STS transactions and 12% for Non-STS. They conclude that by setting the calculation base at KSA “for all investing banks, the level playing field that this simple proposal creates would strengthen the European Single Market. If they choose, regulators could introduce greater differentiation by replacing the 10% factor of proportionality we propose for the risk-sensitive RW Floor described above with 7% and 12%, respectively, for STS and Non-STS securitisations.”

The 6% scalar in Case 1 (Originator – STS – Resilient) and the 8% scalar in Case 2 (Originator – STS

– Non-Resilient) yield an average scalar of 7% for Originator – STS combinations. Likewise, the 8% scalar in Case 5 (Non-Originator – STS – Resilient) and the 10% scalar in Case 6 (Non-Originator – STS – Non-Resilient) result in an average scalar of 9% for Non-Originator – STS combinations. Based on these figures, we conclude that the scalar values for STS combinations, as presented in Table 6.3, are reasonable.

Turning to Non-STS transactions, the 10% scalar in Case 3 (Originator – Non-STS – Resilient) and the 12% scalar in Case 4 (Originator – Non-STS – Non-Resilient) yield an average scalar of 11% for Originator – Non-STS. Similarly, the 12% scalar in Case 7 (Non-Originator – Non-STS – Resilient) and the 15% scalar in Case 8 (Non-Originator – Non-STS – Non-Resilient) result in an average scalar of 13.5% for Non-Originator – Non-STS. We conclude that the scalar values for Non-STS combinations in Table 6.3 are also reasonable.

The difference between the 6% scalar in Case 1 (Originator – STS – Resilient) and the 10% scalar in Case 3 (Originator – Non-STS – Resilient) is 4%, which is close to the 5% difference proposed in the June 17th CRR proposals. Similarly, the difference between the 10% scalar in Case 6 (Non-Originator

– STS – Non-Resilient) and the 15% scalar in Case 8 (Non-Originator – Non-STS – Non-Resilient) is 5%, as in the Proposed CRR. These observations support the reasonableness of the scalar differentiation between STS and Non-STS combinations.

The other two risk dimensions —Originator / Non-Originator and Resilient / Non-Resilient— show a sensitivity about half that of the STS / Non-STS risk dimension. In Table 6.3, the scalar difference between Originator and Non-Originator is 2% (rising to 3% for Non-STS – Non-Resilient combinations). Similarly, the scalar difference between Resilient and Non-Resilient positions is also 2% (and 3% for Non-STS – Non-Resilient).

Therefore, the scalar sensitivities in Table 6.3 appear well-calibrated. The STS / Non-STS distinction remains the most significant driver of scalar differentiation, while the Originator / Non-Originator and Resilient / Non-Resilient dimensions (though meaningful) appropriately exert a smaller influence. This structure ensures a proportionate and conceptually sound calibration across risk dimensions.

SEC-IRBA and SEC-SA risk-sensitive RWFs should also have a cap, calibrated to be consistent with the caps in SEC-ERBA. We propose a value of 20% RW (see Section 7 for this).

6.3.2 – Targeted Amendments to the Proposed p-Factors

Based on these three dimensions, the calibration of the p-factor parameters (p-scaling factor, p-floor and p-cap) should follow these principles, ceteris paribus:

• They should be lower for an Originator-held tranche than for a Non-Originator tranche.[44]

• They should be lower for an STS than for a Non-STS securitisation.

• In the logic of Table 6.2, we apply no differentiation between Resilient position and Non- Resilient position, without any exception (unlike Table 6.2, Case 5). The notion of Resilient position affects only senior positions, whereas the p-factor (as a capital surcharge and factor covering asset correlation risk) is mainly relevant for non-senior tranches. The Proposed CRR is thus logical on making no differentiation based on the Resilient position status (Cases 1 and 2, Cases 3 and 4, Case 7 treated as Case 8). There is no particular logic to differentiate Case 5 and Case 6.

Table 6.4 presents a modified calibration for the SEC-IRBA p-factor parameters and SEC-SA pfactors which respects the architecture of the Proposed CRR while following the calibration principles set out above. Bold text are numerical amendments to the proposed rules that would improve incentives in the European market.

The objective here is to shape the market in accordance with policy objectives: Originator being better than Non-Originator, STS being better than Non-STS, Resilient and Non-Resilient being undifferentiated. This applies to both the SEC-IRBA p-factor scaling, the p-floor, the p-cap and the SEC-SA p-factor.

The process employed in deriving the numerical values presented in Table 6.4 is as follows:

• We begin by aligning the SEC-IRBA p-scaling factors of Cases 3 and 4 (former value of 0.7) with the SEC-SA p-factor values (0.6). The difference between the p-scaling factor of Non- STS vs. STS is thus 0.3 (= 0.6 – 0.3).

• We merge the SEC-IRBA p-scaling factor of Case 5 and Case 6, by taking an average value of 0.4. The difference between Originator / STS and Investor STS is thus 0.1 (=0.4 – 0.3), in line with the risk dimension Originator vs. Non-Originator. The same process is applied to the SEC-SA p-factor of Case 5 and Case 6, resulting in a p-factor of 0.4.

• We apply the Originator Non-STS vs STS difference of 0.3, to the Investor Non-STS vs STS, resulting in a p-scaling factor of 0.7 (instead of the former 1.0) for Cases 7 and 8. The same process is applied to the SEC-SA p-factor of Case 5 and Case 6, resulting in a p-factor of 0.7 (instead of the former 0.7).

• Finally, the p-floor of Case 5 and 6 is aligned based on the STS status, to the Cases 1 and 2.

The most notable effect of removing the inconsistencies in the Proposed rules regarding the p-factor is the reduction for Cases 7 and 8. Duponcheele and Perraudin (2025) indicate a scaling factor 0.65 for Non-STS. Thus, both the 0.6 value for Originator Non-STS and the 0.7 value for Investor Non-STS are deemed reasonable calibration levels. Table 6.4 offers a proportionate and sound calibration across the two risk-dimensions relevant to the p-factor.

7 – What should be improved? refine the SEC-ERBA

7.1 – SEC-ERBA (and IAA) for STS Securitisations

One may adjust the SEC-ERBA RWs to be consistent with those proposed in Section 6 for SEC- IRBA and SEC-SA. Specifically, this involves modifying the proposed CRR SEC-ERBA calibration shown in Table 2.3 to be consistent with Table 6.3.

A complication is that the SEC-ERBA RWs have a maturity dimension lacking in SEC-SA and differing from that in the SEC-IRBA. We start in Panel a) of Table 7.1 by devising a calibration constructed with specific assumptions, that there should be a 5-year tranche maturity coefficient, RW5Y, and that other fixed values in the tables are the main drivers of calibration.

However, this maturity adjustment does not exist in the proposed SEC-IRBA RWF, nor in the SEC- SA RWF nor in the Proposed CRR Table for SEC-ERBA for CQS 1 for Originator/Sponsor. There is thus a better, more consistent Panel b) in Table 7.1.

The drawback of Panel b) Table 7.1 is that it is not compatible with the structure of “Table 4 (page 36)” in the Proposed CRR. This can be fixed by making a conservative choice in the parameters of the risk-sensitive RWF. This is presented in Panel c) of Table 7.1. (Note that the “Originator” position Panel b) denotes Originator or Sponsor, whereas in Panel c) of Table 7.1 the “Originator” position also includes Investor in a securitisation compliant with Article 243(3) (i.e., ‘Resilient’)).

Panel c) of Table 7.1 is thus the ‘best trade-off’ between the various consistency constraints, and is, therefore, the ‘preferred’ solution. Moreover, please note that, to ensure overall consistency, the 1year risk-weight coefficient, RW1Y, for non-senior tranches in the Proposed CRR (not reproduced here) should still be increased for CQS 2 to 18% RW (from 15% RW), and for CQS 3 to 20% (from 15% RW).

Turning to short-term ratings, for short-term CQS, the consistency of risk weights under SEC-ERBA for long-term ratings in Table 7.1 requires an increase in the value used for non-senior tranches to 20% RW (instead of 15% RW as in the current CRR and in the Proposed CRR). Also, there is no distinction between Originator(/Sponsor) and Investor in the Proposed CRR. We take the view that there should be one, and that the minimum value, scalar and cap should be made compatible with RW1Y in the long-term ratings table, in Table 7.1.

Should regulators decide to have no distinction between Originator(/Sponsor) and Investor in Table 7.2, the calibration for CQS 1 should be based on the calibration for Originator(/Sponsor) as the main users, as this would principally apply to ABCP conduits. Should policymakers decide to have only one value for the short term CQS 1, one can chose the most conservative one.

7.2 – SEC-ERBA (and IAA) for Non-STS Securitisations

To maintain a level playing field between regulatory methods, we propose to adjust the Securitisation External Ratings-Based Approach (SEC-ERBA) and the associated Internal Assessment Approach (IAA) grids proposed by the European Commission to the calibration of the Risk-Weight Floor for SEC-IRBA and SEC-SA.

The comments made previously in Section 2 about replacing KA by KSA is particularly relevant for SEC-ERBA. We assume that this important and basic adjustment is implemented. The result of the adjustment, based on the Table 6.3 of this paper, is shown in Table 7.3.

The highest risk weight in KSA for non-delinquent credit assets is 150%, which corresponds to an uncapped CSQ 1 and CQS 2 RW1Y of 18% (for Originator/Sponsor). This would be higher than the CQS 3 RW1Y of 17%. Thus, for consistency purposes, CQS 1 and CQS 2 RW1Y should be lower than CQS 3 RW1Y, which in Table 2 of the Proposed CRR (page 35) is set to 17% for “senior tranche, originator and sponsor”, and to 25% for “senior tranche, investor position.” There is, thus, an implicit cap to the 1-year risk weight coefficient, and this should be made explicit in CQS 1 and CQS 2.

If one turns to the 5-year CQS 2 RW5Y, it takes a fixed value of 18% RW. With a KSA of 150% RW, and the calibration of Table 6.3 using a 12% scalar, the RW5Y is 18% RW, as per the Proposed CRR. However, if one uses the Proposed CRR calibration with a scalar of 15% for originator, RW1Y at 22.5% RW would be above the RW5Y of 18%. Moreover, one can see that for investors, the implicit CQS 1 cap at 25% RW, based on the CQS 3 RW1Y is above the 20% RW for CQS 1 RW5Y. And even if this is solved, then further inconsistencies would appear with the last column, the Non-Senior (thin) tranches RW1Y that are set at 15% RW. They should be set at the cap.

To ensure consistency, the RW1Y for originators and investors should itself be proportional to the risk weight of the pool, and subject to a cap. The thin tranche RW1Y should be increased above the cap for senior tranches. The same would apply for RW5Y. Finally, to ensure that RW5Y is greater or equal to RW1Y, a time coefficient should be introduced.

This would give Table 7.3 Panel a). Please note that this was constructed with specific assumptions, that there should be a tranche maturity component for RW5Y, and that other fixed values in the tables are the main drivers of calibration.

However, this maturity adjustment does not exist in (i) the proposed SEC-IRBA RWF, (ii) the SEC- SA RWF, or (iii) the Proposed CRR Table for SEC-ERBA for CQS 1 for Originator/Sponsor. There is, therefore, a better, more consistent Panel b) in Table 7.3.

Nevertheless, the issue with Panel b) of Table 7.3 is that the notion of Resilience would be only for CQS 1 and CQS 2 categories. Thus, a conservative simplification of Panel b) can be found in Panel c), where the notion of Resilience is removed for Non-STS transactions. Please note that, to ensure overall consistency, the 1-year risk-weight coefficient, RW1Y, for non-senior tranches in the Proposed CRR (not reproduced here) still requires an increase for CQS 1 to 20% RW (from 15% RW), and for CQS 2 to 25% (from 15% RW).

As regards short-term ratings, there is no distinction between Originator(/Sponsor) and Investor in the Proposed CRR for short term ratings. We take the view that there should be one, and that the minimum value, scalar and cap should be made compatible with RW1Y in the long-term ratings table. Note that Panel a) of Table 7.4 is fully compatible with Panel b), and Table 6.3. Should policymakers decide to have only one value for the short term CQS 1, one can chose the most conservative one.

The consistency of risk weights under SEC-ERBA for long-term ratings leads to a requirement to increase the value used to non-senior tranches for CQS 1 to 20% RW (instead of 15% RW as in the current CRR and in the Proposed CRR).

8 – How these improvements would affect capital

The tables in this section show the effects that our suggested changes would have on the RWs for the representative transactions we consider in Section 3. Specifically, Table 8.1 shows how the suggestions of Sections 4 to 7 would alter senior tranche RWs for different deal characteristics under the SEC-IRBA framework. While Table 8.2 shows equivalent results under the SEC-SA framework.

The results in these tables should be compared with those shown in Table 3.2 (for SEC-IRBA) and Table 3.3 (for SEC-SA) which correspond to the effects of adopting the June 17th package proposals.

What are the key differences between what we propose and what the European authorities presented in the June 17th package?

• In Table 8.1, the five panels show results under different assumptions about (i) the minimum attachment point for a senior tranche to constitute a resilient position (adopting as scale variable either the complex concept of Modified KIRB, the standard KIRB concept or KA and trying different scalar multipliers in this context), and (ii) the scale variable in the risk sensitive Risk Weight Floor (RWF) formula. A broad conclusion here is that the choice of scale variable in the RWF minimum and the attachment point expressions does substantially change the pattern of senior tranche RWs for the different categories (STS or not, resilient position or not, Originator/Sponsor or not). It would, therefore, be sensible for the authorities to tie the RWF to simpler SA quantities (ideally of a static nature such as KSA) and the same for the minimum attachment points (in this case KA). Even if level playing field considerations are deemed irrelevant, linking minimum attachment points to KIRB would appear more sensible than the more complex modified KIRB.

• Unlike the results based on then June 17th proposals shown in Table 3.3, the senior tranche results based on own modified proposals shown in Table 8.1 reflect the three low-risk factors (STS, Resilient position and Originator/Sponsor held) in a smoother fashion in that adding any one of the three low-risk characteristics leads to a reduced RW. The notion that investorheld positions are only riskier if they are non-STS seems to us contrived. High- and low-risk factors work cumulatively not as a form of ordering based on some notion that information is discretely available or not.

• A third key difference between the results shown in Table 8.1 and those in Table 3.3 is that RWs are reduced for Non-STS, non-resilient senior tranches relative to current RWs. The notion that the current capital charges remain appropriate for non-STS exposures is hard to square with the fact that STS did not exist in the European market at the time of the Great Financial Crisis (GFC). However, no highly rated European securitisation tranches exhibited delinquencies except some exposed to US subprime tranches and some Commercial Mortgage Backed Securitisations (CMBS) which are subject to refinancing risk.

• A fourth difference is that the June 17th proposals do not permit the category of Investorowned, non-STS resilient, since the proposed definition of resilient requires that positions either be Originator/Sponsor owned or that they be STS. Again, a risk-based view of regulation will tend to link RWs to different risk drivers in a cumulative way (more low-risk factors implies lower RWs). Reintroducing the possibility of Investor owned, Non-STS resilient positions, leads to more plausible RWs that are smoother in the underlying risk drivers.

9 – Conclusion

The June 17th package is a significant attempt by the European authorities to reboot securitisation capital regulation. The proposed changes have the potential to boost activity significantly and, thereby, to free up lending headroom to the region’s banks at a time when Europe has pressing energy and digital transitions, and defence financing needs.

We seek in this paper to identify small, targeted changes in the proposals that would improve outcomes further. Where possible, we aim to simplify the proposals and to reduce distortions across institutions and asset classes. Since the quantitative proposals and criteria have come without a prior public consultation, it is important, we believe, that regulators be open to considering improvements.

We suggest that the authorities:

(i) Replace “RWF = Scalar x KIRB or KA x 12.5” with the simpler, and more effective “RWF = Scalar x KSA x 12.5” to improve level playing fields between banks and between Member States. This has also the side benefit of reducing complexity.

(ii) Alter the definition of resilience: a. By permitting Non-STS, Investor-held positions to be considered Resilient, b. By removing the “counterparty credit risk” criterion included in the Explanatory Memorandum of the Proposed CRR, in its discussion of “resilient positions”, which suggests an intention to prevent (re)insurers from participating in the Resilient STS market[45], or update the discussion to contemplate STS-eligible insurance,[46] c. For traditional securitisations, for SEC-IRBA and SEC-SA which do not recognise excess spread, to ensure consistency across banks and the Liquidity Coverage Ratio (LCR), set the resilient attachment point for IRB banks at the same level as for SA banks, i.e., AP = 1.5 KA for all banks, d. When SEC-ERBA is used, because excess spread is diverted to protect the senior tranche and such diversion are recognised by rating agency methodologies, there is no need for a resilient attachment point criterion for senior tranches. Indeed, the attachment point cannot be lowered arbitrarily as for a given rating, it is constrained by the rating methodology.

(iii) Modify the parameters of the Risk Weight Floor to be consistent with the low risk indicators (Originator vs Investor, STS and resilient vs non-resilient) so that STS implies generally lower RW than Non-STS, Resilient lower than Non-Resilient, and Originator/Sponsor lower than Non-Originator/Sponsor, thereby enhancing banks’ incentives to structure deals satisfying STS and resilient position requirements, and introduce an absolute value cap in the RWF for consistency with non-senior tranches,

(iv) Similarly, alter the p-factor parameters to make them consistent with the risk dimensions,

(v) Align the Securitisation-External Ratings Based Approach (SEC-ERBA) with (ii) & (iii).

Finally, we note that 506d in the Proposed CRR states the Commission will assess the securitisation market by 4 years after the proposals come into force and “report on the appropriateness and effectiveness of the Union prudential securitisation framework.” In addition, it states “the Commission shall consider whether a more fundamental change to the risk-weight formulas and functions would make it possible to achieve more risk sensitivity, achieve more proportionate levels of capital nonneutrality, mitigate cliff effects and address structural limitations of the current framework, taking into account the historic credit performance of securitisation transactions in the Union and the reduced model and agency risks of the securitisation framework.”

We believe that this future review, for once flagged up well in advance, will be an important opportunity for the industry and regulators to develop a convincing evidence base, technical tools and capital formula to permit a simpler and more scientific approach to securitisation capital than has so far been achieved. Ideally, this would be adopted jointly at EU and BCBS levels.[47]

References

Antoniades, A. and N. Tarashev (2014) “Securitisations: tranching concentrates uncertainty,” BIS Quarterly Review, December, (Here)

Ashcraft, Adam B. and Til Schuermann (2008a) “Understanding the Securitization of Subprime Mortgage Credit,” Federal Reserve Bank of New York Staff Reports, March, No. 318, (Here)

Ashcraft, Adam and Til Schuermann, (2008b) “The Seven Deadly Frictions of Subprime Mortgage Credit Securitization,” The Investment Professional, Fall, 2-11.

Bell, Ian, Michael Bennett, Georges Duponcheele, Tamar Joulia-Paris and Véronique Ormezzano, (2025a). “Strengthening Financial Stability through Insurance-based Credit Risk Transfer”, Report, May. (Available on PCS and IACPM websites).

Bell, Ian, Michael Bennett, Georges Duponcheele, Tamar Joulia-Paris, and Véronique Ormezzano

(2025b) “Insurance-based Credit Risk Transfer: a resilient risk-bearing capacity untapped in the EU”, SUERF Policy Note, no 375, SUERF, September.

Agarwal, Swapnil and William Perraudin (2024) “Calibration of a Securitisation Capital Floor,” July. (Here)

Duponcheele, Georges, William Perraudin, Alexandre Linden and Daniel Totouom-Tangho (2014) “Calibration of the CMA and Regulatory Capital for Securitisations,” Risk Control Research Note, April. (Here)

Duponcheele, Georges and William Perraudin (2022) “Impact of the SA Output Floor on the European Securitisation Market,” 22-65a. (Here)

Duponcheele, Georges, Marc Fayemi, Jérémy Hermant, William Perraudin and Frédéric Zana (2024a) “Rethinking the Securitisation Risk Weight Floor,” Report, May. (Here)

Duponcheele, Georges, Marc Fayémi, Fernando González Miranda, William Perraudin and Alessandro Tappi (2024b) “European Competitiveness and Securitisation Regulations,” Report, August. (Here)

Duponcheele, Georges, Marc Fayémi, Fernando González Miranda, William Perraudin and Alessandro Tappi (2024c) “Securitisation Reform to Boost European Competitiveness”, Policy Brief, no 976, SUERF, September. (Here)

Duponcheele, Georges and William Perraudin (2025) “How to Calibrate Securitisation Capital Rules”, Report, April. (Here)

Elul, Ronel, (2011) “Securitization and Mortgage Default,” Working Paper 09-21R, Federal Reserve Bank of Philadelphia, May.

European Systemic Risk Board SRB (2025). “Unveiling the impact of STS on-balance-sheet securitisation on EU financial stability,” May. (Here)

Joint Committee (2022) “Joint Committee advice on the review of the securitisation prudential framework (Banking),” Joint Committee of the European Supervisory Authorities, December. (Here: JC 2022 66)

Joint Committee (2025) “Joint Committee Report on the implementation and functioning of the Securitisation Regulation (Article 44)”, Final Report, EBA, EIOPA, ESMA, March. (Here: JC 2025 14)

European Commission (2025a) “Proposal for a Regulation of the European Parliament and of the Council amending Regulation (EU) No 575/2013 on prudential requirements for credit institutions as regards requirements for securitisation exposures”, June. (Here)

European Commission (2025b) “Proposal for a Regulation of the European Parliament and of the Council amending Regulation (EU) 2017/2402 of the European Parliament and of the Council of 12 December 2017 laying down a general framework for securitisation and creating a specific framework for simple, transparent and standardised securitisation”, June. (Here)

European Commission (2025c) “Impact Assessment Report, Review of the Securitisation Framework”, Commission Staff Working Document, June. (Here)

Keys, Benjamin, T. Mukherjee, A. Seru and V. Vig, (2010) “Did Securitization Lead to Lax Screening? Evidence from Subprime Loans,” The Quarterly Journal of Economics, February.

Perraudin, William (2013) “Agency Cost Arguments for Non-Neutrality in Securitization Capital,” Risk Control Research Note, No. 13-81a, July.

Standard & Poor’s (2024) “Default, Transition, and Recovery: 2023 Annual Global Financial Services Default And Rating Transition Study,” S&P Global, June. (Here)

Zana, Frederic and Eric Rossignol (2025), “Capital-neutral securitisation risk weights”, Capital Management, Cutting edge, Risk.net, September.

Appendix 1: IRB attachment point calibration

Section 5.1.1 mentions that the proposed IRB Attachment Point using Modified KIRB (see equation (5.2)) could be simplified by use of KIRB (as in equation (5.3)). For synthetic securitisation, in which the Originator almost always retains the senior tranche, equation (5.3) has the merit of simplicity relative to (5.2) and is compatible with the variables of the existing SEC-IRBA formula.

For synthetic transactions only, where the senior tranche is almost always retained by the Originator / Sponsor, we suggest in this Appendix 1 that for the minimum Attachment Point for equation (5.3) a possible value for γIRB be set at 1.3. (5.3) APIRB = γIRB × KIRB

Two approaches that can be employed to assess the reasonableness of this value. The first is explained in Agarwal and Perraudin (2024). That paper computes the capital requirement of thin tranches using the parameters for three different asset types, namely (i) Small and Medium Enterprises (SME), (ii) Low Risk Weight Residential Mortgages, and (iii) Other Retail[48] using the parameters on European loan backed securitisation presented in Duponcheele et al. (2014).[49]

Based on these calculations, the authors estimate the capital floor, and for different γ values between 1.0 to 2.0, in which they show that that the capital floor is correlated with KIRB in that, for lower KIRB, they observe a lower capital floor. Moreover, they observe that the average capital floor as a percentage of the KIRB for the three asset classes is below 10% for γ of 1.5 and 2.0 using either the Risk Control Internal Research or the Calibration Data parameter values (see Table 4.3 of their paper reproduced below, with Panel (a) and Panel (b), respectively).

For SME loans with attachment point of 1.5 × KIRB, the capital floor is marginally above 10% but is reduced by half when γ is 2.0. Their findings suggest that for the asset classes we consider, a suitable floor is in the region of 10% of pool RWs.

Table 4.3 (Agarwal and Perraudin): Floor as Percentage of KIRB Panel (a) Risk Control Internal Research Panel (b) Calibration Data Gamma Gamma Asset Asset x1.0 x1.5 x2.0 x1.0 x1.5 x2.0 KIRB KIRB Type Type SME 28.0 13.3 5.9 7.06 SME 21.8 10.6 4.9 5.70 RMBS 15.9 4.5 1.1 3.65 RMBS 13.9 3.9 0.9 3.37 Auto 1.3 0.3 0.1 7.82 Auto 1.6 0.4 0.1 9.08 Average 15.1 6.1 2.4 Average 12.4 5.0 2.0 Note: See the note to Table 4.1.

Even for an intermediate γ value of 1.3, the interpolated floor would still be in the region of 10% of pool RWs for mortgages, well below 10% for Auto. Nevertheless, for SMEs, the γ value of 1.3 would be too low to be compatible with a proportional floor in the 10% area.

The second approach is derived from Zana and Rossignol (2025). Table B of their paper provides curves of tranche PD to assess the proper level of credit enhancement ratio. In this respect, they use the level of CE corresponding to a 0.30% PD to 0.05% PD on the pool (table B), which corresponds to rating BBB to A, and then divide it by the KIRB of the pool for the given PD (we assume no recovery). The result is a set of multiples, that depend on the quality of the pool:

Table A1.2 uses a result from the article that links the tranche PD with the level of CE with the exponential formula PDtrancℎe = p[1] ∙e−p[2]CE.

Table A1.2: Ratio CE/KIRB based on Pool PD

Pool PD 0.20% 0.50% 1.00% 2.00% 5.00% Retail 1.41 1.16 1.06 1.03 1.06 Tranche PD 0.30% (BBB) 1.62 1.35 1.21 1.15 1.16 Tranche PD 0.15% (BBB+) 1.96 1.65 1.45 1.35 1.31 Tranche PD 0.05% (A) Corporate 1.05 0.99 0.91 0.93 0.77 Tranche PD 0.30% (BBB) 1.21 1.16 1.04 1.06 0.85 Tranche PD 0.15% (BBB+) 1.46 1.43 1.25 1.26 0.96 Tranche PD 0.05% (A)

The couples (p[1], -p[2]) have been calibrated for different pools, so that they can easily be used to estimate the ratio. They are provided in Table A1.3.

Table A1.3: couples (p[1], -p[2]) based on Pool PD Pool PD 0.2% 0.5% 1.0% 2.0% 5.0% (0.3,-94.83) (0.2259,-59.4) (0.3735,-49.94) (0.9295,-45.35) (8.188,-44.24) Retail pool (p[1], -p[2]) Corporate pool (p[1], -p[2]) (0.3,-79.08) (0.1702,-41.72) (0.3738,-37.77) (0.469,-28.61) (4.3986,-33.17)

The average ratio of the minimum CE / KIRB corresponding to a tranche rated BBB or BBB+ (PD of the tranche between 0.15% and 0.3%) is 1.1, and it is 1.2 if we extend it to tranches rated A. Hence, based on this data a multiple of 1.15 x KIRB is an appropriate level for the minimum CE for the resilience criteria.

Taken together, those two different approaches would indicate that a choice of γIRB at 1.3 would be quite conservative, excluding for SME, for equation (5.3). Table A1.4 show the effect of such change and can be compared to Table 8.1 Panel b).

Table A1.4: 1.3 × KIRB attachment point and RWF Formula = Scalar × KSA ×12.5 KIRB-based Current Proposed Attach- Originator / Sponsor Investor Credit KIRB ment STS Non-STS STS Non-STS STS Non-STS quality Asset Class (RW) Point Resilient Non-resi. Resilient Non-resi. Resilient Non-resi. Resilient Non-resi. Corporate 52.59 5.47 10.00 16.10 4.20 5.60 7.00 9.00 5.60 7.00 9.00 12.00 SME 58.22 6.05 10.00 15.87 5.10 6.80 8.50 10.20 6.80 8.50 10.20 12.75 Residential mortgage 19.40 2.02 10.00 18.06 3.00 5.00 9.14 9.14 5.00 7.00 11.33 12.00 Retail - other 45.27 4.71 16.61 42.94 7.08 7.08 21.71 21.71 11.68 11.68 26.92 26.92 High Corporate -58.00 -44.00 -56.53 -44.11 -44.00 -30.00 -44.11 -25.48 SME Percentage change -49.00 -32.00 -46.43 -35.71 -32.00 -15.00 -35.71 -19.64 Residential mortgage versus current RWs -70.00 -50.00 -49.38 -49.38 -50.00 -30.00 -37.24 -33.55 Retail - other -57.39 -57.39 -49.44 -49.44 -29.64 -29.64 -37.29 -37.29 Corporate 80.50 8.37 10.00 22.08 4.20 5.60 9.70 9.70 5.60 7.00 11.78 12.00 SME 89.29 9.29 10.86 21.04 5.10 6.80 10.86 10.86 6.80 8.50 10.92 12.75 Residential mortgage 28.93 3.01 10.00 25.55 4.10 5.00 12.82 12.82 6.84 7.00 15.95 15.95 Retail - other 69.07 7.18 21.57 57.65 8.80 8.80 28.51 28.51 14.93 14.93 35.64 35.64 All Corporate -58.00 -44.00 -56.09 -56.09 -44.00 -30.00 -46.64 -45.66 SME Percentage change -53.05 -37.40 -48.37 -48.37 -37.40 -21.75 -48.10 -39.40 Residential mortgage versus current RWs -58.97 -50.00 -49.81 -49.81 -31.61 -30.00 -37.59 -37.59 Retail - other -59.23 -59.23 -50.54 -50.54 -30.79 -30.79 -38.18 -38.18 Leveraged Loans 223.68 23.26 32.17 36.04 13.01 13.01 32.17 32.17 13.01 15.00 32.17 32.17 Special Lev. Loans with w 253.75 26.39 38.05 38.05 15.38 15.38 38.05 38.05 15.38 15.38 38.05 38.05 case Leveraged Loans Percentage change -59.56 -59.56 -10.73 -10.73 -59.56 -53.37 -10.73 -10.73 Lev. Loans with w versus current RWs -59.56 -59.56 0.00 0.00 -59.56 -59.56 0.00 0.00

Appendix 2: resilient positions relations

The European Commission Securitisation Package, published on 17th June 2025 contains various calibration. We will use the following terminology to describe such calibration.

Scalar: this is the percentage of the risk weight of the pool that is applied when calculating a risk-sensitive risk weight floor tranche, before the application of minimum value. For example, if the risk weight floor is set to be equal to 15% times the pool capital times 12.5, then the coefficient of proportionality of 15% is the “scalar”.

p-factor: this is the smoothing coefficient that is used in the decreasing exponential function (‘KSSFA()’) of the capital formula. In its current design the exponential function starts at an attachment point equal to one times capital and decreases continuously based on the p-factor, resulting in an area under the curve equal to ‘p-factor’ times the pool capital for low risk-weight pools and low p-factors. For higher risk-weight pools and higher pfactor, the exponential curve does not converge to zero at the time of reaching the senior tranche detachment point of 100%, resulting in an area under the curve that is marginally lower than ‘p-factor’ times pool capital.

Gamma (γ): this is the coefficient that is applied to pool capital, when describing a sufficiently high attachment point of a senior tranche to define such tranche as a possible ‘resilient position’. In the Securitisation Package proposal, for the standardised approach, when the pool capital is defined as KA, this value is equal to 1.5.

Under the current design, there is a mathematical relationship between a) the area of the exponential curve for a senior tranche attaching at ‘Gamma times pool capital’ and detaching at 100%, and b) the pool capital. The ratio of a) over b) defines an minimum value for the scalar. In other words, in the scalar is lower than this theoretical value, it will never have an effect for a tranche attaching at the point of ‘resilience’; instead the output of the exponential function KSSFA() will always be higher.

For a theoretical negligible pool risk weight, Table A2.1 provides the theoretical minimum Scalar for various pfactor (columns) and Gamma values (rows). For a given p-factor and a given Gamma, if the Scalar in the regulation is set below this theoretical minimum Scalar in Table A2.1, the RW Floor will not override the KSSFA function at the point of resilience.

Table A2.1: Theoretical minimum Scalar, as a function of the p-factor and Gamma combination. Theoretical p-factor minimum Scalar (%) 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 1.0 20.0% 30.0% 40.0% 50.0% 60.0% 70.0% 80.0% 90.0% 100.0% 1.1 12.1% 21.5% 31.2% 40.9% 50.8% 60.7% 70.6% 80.5% 90.5% 1.2 7.4% 15.4% 24.3% 33.5% 43.0% 52.6% 62.3% 72.1% 81.9% 1.3 4.5% 11.0% 18.9% 27.4% 36.4% 45.6% 55.0% 64.5% 74.1% 1.4 2.7% 7.9% 14.7% 22.5% 30.8% 39.5% 48.5% 57.7% 67.0% 1.5 1.6% 5.7% 11.5% 18.4% 26.1% 34.3% 42.8% 51.6% 60.7% 1.6 1.0% 4.1% 8.9% 15.1% 22.1% 29.7% 37.8% 46.2% 54.9% 1.7 0.6% 2.9% 7.0% 12.3% 18.7% 25.8% 33.3% 41.3% 49.7% 1.8 0.4% 2.1% 5.4% 10.1% 15.8% 22.3% 29.4% 37.0% 44.9% 1.9 0.2% 1.5% 4.2% 8.3% 13.4% 19.4% 26.0% 33.1% 40.7% 2.0 0.1% 1.1% 3.3% 6.8% 11.3% 16.8% 22.9% 29.6% 36.8% 2.1 0.1% 0.8% 2.6% 5.5% 9.6% 14.5% 20.2% 26.5% 33.3% 2.2 0.0% 0.5% 2.0% 4.5% 8.1% 12.6% 17.9% 23.7% 30.1% 2.3 0.0% 0.4% 1.6% 3.7% 6.9% 10.9% 15.8% 21.2% 27.3% 2.4 0.0% 0.3% 1.2% 3.0% 5.8% 9.5% 13.9% 19.0% 24.7% 2.5 0.0% 0.2% 0.9% 2.5% 4.9% 8.2% 12.3% 17.0% 22.3% 2.6 0.0% 0.1% 0.7% 2.0% 4.2% 7.1% 10.8% 15.2% 20.2% 2.7 0.0% 0.1% 0.6% 1.7% 3.5% 6.2% 9.6% 13.6% 18.3% 2.8 0.0% 0.1% 0.4% 1.4% 3.0% 5.3% 8.4% 12.2% 16.5% 2.9 0.0% 0.1% 0.3% 1.1% 2.5% 4.6% 7.4% 10.9% 15.0% Gamma 3.0 0.0% 0.0% 0.3% 0.9% 2.1% 4.0% 6.6% 9.8% 13.5% Note: A theoretical negligible risk weight of 0.001% was used for the construction of Table A2.1

How can Table A2.1 be interpreted? In the Securitisation Package, for the Standardised Approach, the Gamma is set at 1.5 for all transactions. Thus, for issuers of STS Resilient positions, the p-factor is set at 0.3. Table A2.1 provides for this combination of Gamma and p-factor a theoretical minimum Scalar of 5.7%, which is below the Scalar of 10% proposed in the package. This means that, theoretically, conditional on a proportional risk weight floor being above the minimum floor value, this value will apply.

For issuers of Non-STS Resilient positions, the p-factor is set at 0.6. Table A2.1 provides for a Gamma of 1.5 and this p-factor combination a theoretical minimum Scalar of 26.1%, which is well above the Scalar of 15% proposed in the package. This means that, theoretically, conditional on a proportional risk weight floor being above the minimum floor value, this value will not apply, and instead the function KSSFA will drive the risk weight of the resilient position. That function will drive it until the senior tranche attachment point is about 1.9 times pool capital or above.

This effect is accentuated as the p-factor increases. For investors of Non-STS Resilient positions, the p-factor is set at 1. Table A2.1 provides for a Gamma of 1.5 and this p-factor combination a theoretical minimum Scalar of 60.7%, which is well above the Scalar of 15% proposed in the package, i.e. more than four times what would be expected. This effect shows that the p-factor of 1.0 combined in the SSFA formula that is added to a risk-weight of 1250% up to one-times pool capital is clearly problematic. This has been discussed in the paper “How to Calibrate Securitisation Capital Rules” by Duponcheele and Perraudin (2025). For an investor, the senior tranche would need an attachment point of at least 3.0 times pool capital to see the benefit of a proportional risk weight floor with a scalar of 15%.

Appendix 3: resilient and STS collateralisation criteria

The JC report (December 2022)

For Synthetic STS framework, the (misconceived) idea that an unfunded senior position is resilient if the nonsenior positions of the same transaction are fully collateralised was first suggested in a report titled “Joint Committee advice on the review of the securitisation prudential framework (Banking)” published by the Joint Committee (JC) of the European Supervisory Agencies (ESAs) in December 2022 (See JC (2022)).

This report included a “[…] substantial, but still targeted, recommendation aimed at improving the risk sensitiveness of the framework by recognising the reduced model and agency risk associated to originators […].” The JC advice elaborated “on why a reduction of the risk weight floor for senior tranches retained by originators may support further growth in the SRT market in a prudent manner, if accompanied by a set of appropriate safeguards. This would in particular promote the issuance of resilient securitisations, which can qualify for a more beneficial capital treatment, without jeopardising financial stability.”

But in exchange for a reduction in the RW Floor, the JC wanted ‘thicker’ mezzanine tranches sold to the nonbank investor market. The JC argued that “[…] a reduction of the risk weight floor if accompanied by robust eligibility criteria may contribute to a prudentially sound structuring of transactions by giving the originators an incentive to engage in more resilient transactions (e.g. based on granular pools) and increasing the risk sharing through a transfer of thicker junior and/or mezzanine tranches.”

The JC proposed four eligibility criteria (or safeguards), the first one being the securitisation tranches should amortise sequentially (or pro rata with appropriate performance-related triggers switching to sequential), the second one on counterparty risk (being the collateralisation requirement discussed here), the third one to ensure that the credit enhancement of the senior tranche is sufficiently thick, and the fourth one ensuring that the pool is sufficiently granular.

While we agree on the principles of the first, third and fourth eligibility criteria in that report, the second one is problematic. Those criteria – especially the second one (under discussion here) – were considered in the paper “Rethinking the Securitisation Risk Weight Floor” (see Duponcheele et al. (2024a)), published in May 2024, and briefly mentioned in the paper “European Competitiveness and Securitisation Regulation”, published in August 2024 (see Duponcheele et al. (2024c)).

The second eligibility criteria in the JC report stated: “The credit protection agreement by which the transfer of risk is achieved shall comply with the criteria specified in paragraphs 8 to 10 of Article 26e SECR for STS synthetic securitisation. […][50]” The Securitisation Regulation (SECR) Article 26e(8) refers to eligible guarantors for STS transactions without collateral (i.e., ‘unfunded’) and contains a point (a) that lists only public sector entities (such as sovereigns or supranationals), meeting certain conditions, as eligible guarantors, thus defacto excluding all private sector entities, included regulated ones, such as (re)insurers.

The JC justified its restriction by stating: “This safeguard is key as it permits the originators to compensate quickly the losses incurred in SRT structures. In the absence of such a safeguard the protection buyer (i.e. the originator which retains the senior note) risks to bear losses on the subordinated protected tranche.” This reflects a concern that is sometimes mentioned by regulators on the risk of late payment or non-payment of the credit protection amount when a borrower or counterparty in the securitised pool defaults.[51]

The JC added: “Though the losses on the senior tranche are not affected by this criterion if the guarantor defaults[52], this feature is nevertheless deemed beneficial for lowering the risk weight floor associated to the senior tranche as it contributes to the effectiveness of the risk transfer.” This is an odd way of thinking about risk. Indeed, although the JC recognises here that the collateralisation of the non-senior tranches has no bearing whatsoever on the losses of the senior tranche, it stills deems it ‘beneficial’ to justify a reduction in the riskweight of the senior position and disregards any other risk mitigant such as counterparty risk mitigation (CRM), itself a core component of the Capital Requirements Regulation (CRR), the mechanics for which under the standardised approach to credit risk (unless the protection seller is preferentially treated public body) give rise to a residual risk weight, reducing the capital relief associated with the protection for the protection buyer. Instead, it links the issue directly to SECR: “A few remarks are worth noting: a) the criterion is focused on protecting the originator from counterparty credit risk of the protection seller. Therefore, only the criteria specified in Article 26e (8) to (10) of the SECR for STS synthetic securitisation which are relevant for this purpose should apply.[…]”.

The JC report (March 2025) and ESRB report (May 2025)

The topic was revisited (see Joint Committee (2025)) by the Joint Committee on 31st March 2025 in the report “Joint Committee Report on the implementation and functioning of the Securitisation Regulation (Article 44)”, not in the context of a ‘resilient position’, but in the context of ‘STS’ securitisation. The report lists three ‘pros’ and three ‘cons’ of the participation by (re)insurers in the STS market.[53] The report mentions among the ‘pros’ supporting their participation, the existence of the credit risk mitigation rules in the CRR: “According to the CRR in case of unfunded credit protection the banks have to capitalise against the insurers and reinsurers counterparty risk. In addition, as per Article 249 of the CRR, these entities are subject to minimum rating requirements when providing credit risk mitigation to securitisation positions. In summary, the investor side of OBS[54] securitisations will be broadened, thereby deepening, and enlarging the market.”

Nevertheless, the same argument was used about one month later to argue in the opposite direction, in a report published on 5th May 2025 by the European Systemic Risk Board (ESRB) (see ESRB (2025)), entitled “Unveiling the impact of STS on-balance-sheet securitisation on EU financial stability”. The reasoning was developed in the concluding paragraph of the report: “To mitigate counterparty credit risk, Article 249 of the CRR requires providers of unfunded credit protection to have a minimum credit quality step both at the time the credit protection is initially recognised and throughout the duration of the securitisation. It also imposes a capital charge on the protected tranche, thereby reducing the capital relief the originator would receive through STS securitisation. However, this minimum credit quality step requirement under Article 249 of the CRR introduces a cliff-edge effect, as a downgrade below the eligibility threshold would render the synthetic securitisation ineligible for both the STS label and SRT. As a result, banks would lose the associated capital relief. This effect would be amplified if the change in regulation led to (re)insurers becoming major providers of credit protection for synthetic STS securitisations. Adverse developments specific to the (re)insurance sector that would lead to rating downgrades of (re)insurers might thus create large spillover effects to the banking sector.”

It is clear from the context that both the JC and the ESRB reports refer to a previous version of the CRR (CRR2), which has been substantially changed since 1st January 2025 when a new version (CRR3) came into force, and the issue raised has lost its relevance, as it has been addressed by the European Commission. Indeed, while Article 243(9) has not changed between both versions, Article 201(1)(g) to which it refers has, and since the 1st January 2025, the ‘cliff-edge effect’ is no longer relevant for the European SRT market.[55]

CRR Article 249(3) states: “[…] the eligible providers of unfunded credit protection listed in point (g) of Article 201(1), shall have been assigned a credit assessment by a recognised ECAI which was credit quality step 2 or above at the time the credit protection was first recognised and is currently credit quality step 3 or above.”[56]

Prior to 1st January 2025, all (re)insurers were included in point (g) of Article 201(1). Until that date, the unfunded SRT market, which already exists for non-STS transactions, had thus an embedded downgrade risk.[57]

To address the rating downgrade-related ‘cliff-edge effect’[58], the EC and the co-legislators updated CRR2 Article 201(1),[59] with the effect that point (g) no longer applies to “regulated financial sector entities” within newly added point (fa) in the list of “eligible providers of unfunded credit protection”. The new CRR3 Article 201(1), in force since 1st January 2025 adds: « For the purposes of the first subparagraph, point (fa), of this Article, ‘regulated financial sector entity’ means a financial sector entity meeting the condition set out in Article 142(1), point (4)(b).” And this point (4)(b) includes entities within the scope of prudential consolidation under the EU Solvency II Directive (2009/138/EC) or equivalent, as it states: “the entity is subject to prudential requirements, directly on an individual or consolidated basis, or indirectly from the prudential consolidation of its parent undertaking, in accordance with this Regulation, Regulation (EU) 2019/2033, Directive 2009/138/EC, or legal prudential requirements of a third country at least equivalent to those Union acts;”

And since no minimum rating requirement applies to “regulated financial sector entities” within CRR3 Article 201(1) point (fa), this also applies to (re)insurers providing SRT guarantees.

These changes flow through from the general credit risk mitigation framework to the securitisation specific credit risk mitigation requirements, under CRR Article 249(3). Since the relevant have been in force since 1st January 2025, the ‘cliff-edge effect’, by which banks would lose the benefit of the SRT recognition, is no longer there for EU (re)insurers subject to Solvency II.

Conclusion

To conclude, in the June 17th CRR proposal from the European Commission, the requirement of collateralisation of non-senior positions to define whether a senior position is ‘resilient’ or not is not logical per se. It goes back to a December 2022 report by the JC, prior to the elimination of the rating cliff-edge effect by the European Commission in the CRR3, implemented in law since the 1st January 2025. The collateralisation requirement for non-senior positions should not be part of the criteria of a senior ‘resilient position’; it has no theoretical basis nor practical basis.

Moreover, the collateralisation requirement for Resilient positions is only mentioned in the ‘Explanatory Memorandum’ of the June 2025 17th CRR proposal from the European Commission; it is not mentioned in the proposed CRR recitals, nor in the proposed CRR articles. Neither is it mentioned in the June 17th SECR proposal. Maybe it could be one of those cases, where it was never the final intention to include it, and the relevant content in the ‘Explanatory Memorandum’ is a remnant of a previous draft and might be deleted or amended in a final version.

Last, but not least, looking forward, the current Proposed SECR may inadvertently reintroduce the ‘rating cliff’ that existed in the CRR 2, and which the ESRB has warned about. There is a real risk that it might be so. Indeed, the second safeguard of the new proposed Article 26e(8)(aa) is about robustness and states that a (re)insurer “[…] has been assigned to credit step 3 or better;”. The wording implies an ‘on-going test’, and there is a risk that the STS capital relief of a bank might be disallowed upon a downgrade to CQS 4 or below. This is potentially problematic, and it reintroduces a risk that the ESRB warned about (although the warning was based on CRR 2, which itself has been superseded by CRR 3). As policymakers want to build a robust market, they should not change the CRR 3 risk mitigations without a rating cliff, but might add a barrier to ensure that only the robust (re)insurers can underwrite those transactions. This would require a small adjustment in the language so that the words “has been assigned to credit step 3 or better;” are replaced by “[…] has been assigned a credit assessment by a recognised ECAI which was credit quality step 2 or above at the time the credit protection was first recognised;”. In essence, this change takes the ‘best’ part of Article 249, without the negative effects of the on-going requirement causing the rating cliff. This would lead to both Robustness and Financial Stability.

Notes

[1] Michael Bennett is Chief Underwriting Officer at the Mortgage Division of Arch Insurance (EU) dac; Dr. Georges Duponcheele is a Senior Credit Portfolio Manager at Great Lakes Insurance SE (part of Munich RE); Dr. Fernando González Miranda is a Senior Advisor at Qatar Central Bank and until recently was a Senior Advisor at the European Central Bank (ECB); Jérémy Hermant is Senior Advisor to Alantra; Dr. William Perraudin is the Managing Director of Risk Control Limited; Véronique Ormezzano is Director of VYGE Consulting; Dr. Frédéric Zana is Head of Quantitative Structuring at CACIB. All authors can be contacted via LinkedIn. The views expressed are the authors’ own and not necessarily those of their firms.

[2] The authors thank those who provided comments including Sophie Asselot, Raffaela Baechler, Alexander Batchvarov, Ian Bell, Ralf Fischer zu Cramburg, Jo Goulbourne Ranero, Tamar Joulia-Paris, Michael Heckl, Sara Millard, Harry Noutsos, Eric Rossignol, Stephan Tilke, Laurent Vanoverberghe, Ansgar West, Kirk Willison, Yves Zerkovitz. Any errors remain our own.

[3] The Securitisation Package is made up of proposed changes to the Capital Requirements Regulation (CRR), to the Securitisation Regulation (SECR), to the Liquidity Coverage Ratio LCR) Delegated Regulation, and to the Solvency II Delegated Regulation. It is a holistic review aimed at addressing both offer and demand issues. We address in this paper issues pertaining to CRR, SECR and LCR DR, as they related to banks’ prudential regime.

[4] As explained in Duponcheele et al. (2024a) and (2024b), these distortions arise because the ratio between the RWF applied to a very low-risk senior tranche and the risk weight of the underlying pool of securitised assets can be very high. This renders entire asset classes (low-risk residential mortgages being an obvious example) uneconomical to securitise.

[5] This requirement may alternatively be expressed as “maximum concentration”.

[6] This requirement is equivalent to “minimum credit enhancement for senior tranches,”, “sufficiently thick non- senior portion of the capital structure” or “sufficiently high attachment point of the senior tranche”.

[7] Except for guarantees provided by sovereigns or supra-nationals, as per 2021 Capital Markets and Recovery Package.

[8] In the June 17th CRR Proposal, a new Article 506d states that the framework will be reviewed within 4 years of the law being implemented: “In particular, the Commission shall consider whether a more fundamental change to the risk-weight formulas and functions would make it possible to achieve more risk sensitivity, achieve more proportionate levels of capital non-neutrality, mitigate cliff effects and address structural limitations of the current framework, taking into account the historic credit performance of securitisation transactions in the Union and the reduced model and agency risks of the securitisation framework.”

[9] The concern derives from a reference, in the Explanatory Memorandum of the Proposed CRR, to a requirement for collateralisation, other than where guarantees are provided by sovereigns or supra-nationals. This reference should be deleted or updated to accommodate STS-eligible insurance under the Proposed Securitisation Regulation (SECR), as it is not compatible with the mechanics of the proposed articles in the CRR that, as a technical matter, allow (re)insurers to access the ‘resilient STS’ market. As it stands, the Explanatory Memorandum introduces a technical inconsistency with the law itself.

[10] Under the Securitisation Package, in the Proposed Securitisation Regulation (SECR), (re)insurers will be allowed to participate in the STS market on an uncollateralised basis, subject to 4 safeguards (Article 26e(8), point (aa)).

[11] For SEC-IRBA, if the administrative constraints of the CRR Article need to refer to KIRB, we propose an alternative in Appendix 1 of this paper.

[12] The main beneficiaries of such a change would be rated transactions backed by consumer loans and auto loans and leases.

[13] The reduction of p-factor depending on the type of exposure aims at reducing excessive capital non- neutrality, while avoiding cliff effects, notably on mezzanine tranches.

[14] We call the p-function in SEC-IRBA, the variable component of ‘p’, as set in the CRR regulation, i.e., p-function() = (A + B / N + C x KIRB + D x LGD + E x MT). The p-function is multiplied by the p-scaling factor before p-floor and p-cap are applied. 15The proposals contain no maximum value (absolute cap). This should be introduced and is discussed below.

[16] Thus, the minimum granularity, N, equals 50.

[17] Use of the product of EL and the WAL for the initial reference portfolio is ill-advised if the test is dynamic since EL is likely to decline while the WAL remains the ‘day 1’ WAL

[18] This is described in the Proposed Securitisation Regulation (SECR) Article 26e(8), with the new point (aa).

[19] In the overall prudential framework, immediate enforcement of guarantees is a feature, and pre-funding in SRT has been a way to ensure immediate enforcement of losses on non-senior tranches. However, the type of guarantees or insurance contracts used by regulated insurance companies for SRT transactions do not leave discretion as to the timely payment. It is a peculiar situation of credit insurance which works differently from other insurance contracts and needs to be recognised. Bell et al. (2025a), Appendix A4 (The Fear of the Insurance sector risk vs. the Quantification of the risk) discuss this issue.

[20] The Proposed SECR in the Securitisation Package will include a new point (aa) in Article 26e(8) to add (re)insurers, subject to four specified safeguards, as eligible uncollateralised protection providers. The reference to a requirement for collateralisation in the Explanatory Memorandum of the Proposed CRR, thus, creates an inconsistency with the Proposed SECR.

[21] In 2013, the Basel Committee on Banking Supervision (BCBS) introduced the p-factor for SEC-IRBA capital calculations employing the quantity pIRBA, that was then subject to a floor, the p-floor: pIRBA = A+ (B/N) + (C× KIRB) + (D× LGD) + (E× MT) Here, N is the effective number of loans, KIRB is the pool capital charge under the IRB Approach, LGD is the Loss Given Default rate of the underlying pool, MT is the maturity of the tranche, and parameters A, B, C, D, and E are provided differently based on three categories (i) exposure type (retail or wholesale), (ii) seniority and (iii) granularity (only applicable to wholesale exposure type). The BCBS never published information about how it had calibrated pIRBA and the calibration was adopted without any modification or challenge in CRR 2. The look-up table is provided in CRR Article 259. The calibration of the parameter E is particularly problematic for retail assets, as it is applied to a factor, the tranche maturity MT, which has very little to do with risk, rather than depending on pool maturity which is a genuine risk factor. The BCBS Simple, Transparent and Comparable (STC) framework, implemented as the EU STS framework, introduced in 2016 the notion of a p- scaling factor, as shown in Table 2.1.

[22] Also called informally as the “floor of the floor”.

[23] A fixed value is both an absolute floor and an absolute cap; this is how the current fixed value RWF is represented in Tables 2.1 and 2.2.

[24] Because the p-formula in SA is an implied fixed value of 1, the SA p-scaling factor and p-factor are reduced to the same values.

[25] As noted in an earlier footnote, unlike pool maturity, tranche maturity (MT) is unconnected to risk. Under the Proposed CRR, we note that the European Commission has capped the p-factor in SEC-IRBA at 1.0 for Non- STS, to reduce the impact of the extremely high values that are generated by the Basel “calibration” of this non- risk factor when applied to retail asset classes. How and why Basel calibrated an effect that is disconnected from risk remains unclear and why it has remained in regulations more than a decade later is a mystery.

[26] Currently, only senior STS tranches backed by high quality residential mortgages, auto loans and leases, consumer loans, portfolios made mainly of SME loans are eligible.

[27] KIRB = Pool UL + 1-year EL, thus the Attachment point is 1.1 × ((KIRB – EL) + M × EL), with the pool maturity M value floored at 1 year and capped at 5 years.

[28] The formula in the legislative text is: KA = w × 0.5 + (1 −w) × KSA, and since 0.5 x 12.5 = 625% RW, this value is the implied risk weight for delinquent assets.

[29] Another argument in favour of using KSA instead of KA, is that the RWF is only relevant when the SSFA formula, which employs KA, generates a level so low that the RWF becomes relevant. In other words, even with delinquent assets and the p-factor as a surcharge, the delinquent component is not sufficient to generate a value greater than the RWF. It would be strange then to evaluate capital for the senior tranche with a scale variable that includes delinquent assets, when the delinquent assets will only affect the non-senior liabilities of the SPV.

[30] This can occur for example with very low granularity pools, or low values of the p-factor, or both.

[31] CRR Article 261 ought to be reviewed, considering CRR Article 256 on the Attachment Point, to ensure consistency with respect to defaulted assets between the treatment of tranches already written down and tranches that have not yet been written down. Economically the impact of an asset default before a write-off should not be higher than a pure write off (100%, i.e., the maximum capital loss).

[32] The text refers to the “Initial” WAL, which is static. It does not make sense to apply this value on an on-going basis. The Modified KIRB is a mixture of dynamic and initial values, which makes no financial sense at all.

[33] Moreover, APIRB will introduce ‘fake’ liquidity into the LCR for traditional securitisation positions. This issue may be resolved easily by switching to APSA for the definition of ‘Resilient position’ and using a lower haircut.

[34] There is a technical solution for this, i.e., ‘resilience’ is determined under the IRB approach, but the status is not reassessed when calculating the Output floor. The attachment point criteria must thus be drafted as a standalone criterion to be disapplied for the purpose of the Output floor, rather than embedded in various conditions.

[35] Aggarwal and Perraudin (2024) show that, in calibrating a risk-sensitive Risk Weight Floor, for some short- tenor, high-granularity asset classes such as auto loans, RWF scalars well below 10% of pool capital are achieved for senior tranches attaching distinctly beneath 1.5 times pool capital. This may justify an asset-class distinction in the definition of resilient position.

[36] If they were not part of the criteria, there would be an obvious regulatory arbitrage, applying low risk weight floors to pools backed with a very low granularity level (below 10), where the senior tranche would be undercapitalised (see Duponcheele et al. (2024a)). Of course, this is a theoretical construct, as there would be no market in real life for such structures. And supervisors should question the use of the securitisation framework for such structures and/or request supervisory Pillar II adjustments.

[37] On the new “senior position” definition, since KSA is based on a prudential approach calibrated on the average performance of the asset class, the proposal for SEC-SA results in the paradoxical outcome that the securitisations of the safest assets (like prime autos), AAA-rated and senior, typically attaching below KA, are treated as “non-senior”, with commensurate junior capital requirements.

[38] In fact, the p-factor has three roles, the most important of which is to cater for asset correlation risk in a pool.

[39] It appears that the CRR criteria for Resilient positions were defined before the European Commission decided to amend STS criteria to allow insurers to participate in the STS SRT market. Indeed, the proposed CRR specifically indicates that “only two out of five criteria are new, as the criteria on amortisation mechanism, concentration/granularity and counterparty credit risk are already STS criteria (as set out in the Securitisation Regulation)”. That sentence is not compatible with the Proposed SECR which modifies Article 26e(8) by introducing a point (aa), in which the European Commission has introduced safeguards specifically to mitigate counterparty risk, and also to ensure the solidity of the protection to the non-senior tranches.

[40] The proposed text in the new SECR 26e(8)(aa) needs to be amended to address some operational issues and market scope, to ensure that it works as intended. For example, the fourth risk mitigant states “iv. assets under management by the insurance or reinsurance undertaking exceed 20 billion euro”; but ‘assets under management’ is not a term used by non-life undertakings. ‘Total assets’ is a metric available from an undertaking’s audited balance sheet. Also, undertakings often book transactions in their subsidiaries. For this criterion to work, it should be reworded to expand the eligibility to undertakings that either themselves have total assets more than EUR 20 billion or are part of a group with consolidated assets of that amount where the group parent undertaking is subject to Solvency II or Solvency II-equivalent supervision. Currently, not a single EU booking entity active in the market could satisfy the letter of the proposed law, while quite a few would be eligible in the spirit of the law if the above changes were adopted. There is another operational issue with the first risk mitigants (i.) in SECR 26e(8)(aa) regarding the use of internal models which will require addressing but that lies outside the scope of this note.

[41] For a more in-depth discussion, see Appendix 3 on the Resilient and STS Collateralisation Criteria.

[42] With one potential exception, as described in equation (5.3) for IRB Originators/Sponsors of synthetic transactions.

[43] To simplify the dimension, “Originator” means ‘Originator/Sponsor’. And “Non-Originator” means ‘Investor’.

[44] To simplify the dimension, “Originator” means ‘Originator/Sponsor’. And “Non-Originator” means ‘Investor’.

[45] The concern derives from a reference, in the Explanatory Memorandum of the Proposed CRR, to a requirement for collateralisation, other than where guarantees are provided by sovereigns or supra-nationals. This reference should be deleted or updated to contemplate STS-eligible insurance under the Proposed Securitisation Regulation (SECR), as it is not compatible with the mechanics of the proposed articles in the CRR that, as a technical matter, allow (re)insurers to access the ‘resilient STS’ market. The Explanatory Memorandum introduces a technical inconsistency with the law itself.

[46] Under the Securitisation Package, in the Proposed Securitisation Regulation (SECR), (re)insurers will be allowed to participate in the STS market on an uncollateralised basis, subject to 4 safeguards (Article 26e(8), point (aa)).

[47] The existing proposals contain new concepts such as resilient positions (flagged up as a possibility in the past but not widely discussed) and a heavy reliance on distinctions (Originator/Sponsor versus Investor) which are not evidence-based as far as one can tell. We would hope that a well-prepared adjustment to the current rules four years from now will be able to base simplified distinctions on genuine risk factors in an evidence-based manner.

[48] The other retail in this study is considered as auto loans.

[49] They also perform calculations using data collected and analysed as part of Risk Control’s internal research for (i) SME, (ii) Residential Mortgage-Backed Securities, and (iii) Auto.

[50] This eligibility criterion #2 added a sentence related to rating constraints on originators.

[51] As stated in Bell et al. (2025a), “(re)insurers have been present in the EU SRT market on an unfunded basis since 2018, and claims posted by banks have been paid by insurers in a timely manner. This is consistent with the experience of the broader loan-by-loan credit insurance market. Moreover, the terms of the unfunded credit protection contracts used in SRT securitisations have broadly originated from funded credit protection contracts. They are very different from those used in credit insurance more generally and provide (re)insurers, as protection providers, with far fewer opportunities to challenge their liability for losses incurred on the relevant tranche. Furthermore, unlike traditional insurance, in an unfunded credit protection the (re)insurer pays ‘first’ meaning the (re)insurer pays out on initial/interim losses on protected tranches before the bank/originator recoups recoveries (e.g. through the sale of assets and the work-out process).”

[52] This addresses a different risk, counterparty default risk, which also be quantified: “The average default rate for insurance companies between 1981 and 2023 remains the lowest of any financial or nonfinancial sector.” It is 0.06% for insurance companies, 0.42% for banks, and 1.88% for nonfinancial entities. In the EU, there has been no insurer default of credit (re)insurers since 1992 (see Standard & Poor’s (2024)).

[53] For an extensive discussion on the ‘pros’ and ‘cons’ mentioned in the JC (2025) report, see Bell et. al. (2025a).

[54] OBS: on-balance sheet, and alternative term for Synthetic SRT.

[55] This point was highlighted in Bell et. al. (2025b).

[56] ‘CQS 2 or better’ means a rating ‘at least A-’, whereas CQS 3 or better means a rating ‘at least BBB-‘.

[57] In terms of quantification of that risk, for the European SRT market, that risk was not material. See Appendix A4 of Bell et al. (2025) for a quantification of the downgrade risk.

[58] Only securitisation CRM by non-preferential providers was associated with a specific credit rating requirement (CQS 2 at inception, 3 ongoing). Non-securitisation CRM by non-preferential providers required only the existence of a rating of any kind (ECAI or if the institution recognising the CRM was IRB, internal), though the protection would only be useable if it generated a lower capital requirement than the underlying exposure.

[59] Source: “Regulation (EU) 2024/1623 of the European Parliament and of the Council of 31 May 2024 amending Regulation (EU) No 575/2013”, Official Journal of the European Union, June 2024. (Known as CRR3).