With respect to the eligibility requirements in accordance with article 181(1)(f) of the CRR, we welcome the alignment proposed in the Guidelines between the AIRB and SA treatment with respect to the general principles on legal certainty and collateral valuation. We deem, in fact, that the legal certainty is independent of the approach used and the Guidelines in consultation warrant an overall coherence.
With respect to the eligibility of physical collateral which are movable and not in possession of the institution, we consider critical that the guidelines ask for a legal opinion confirming the legal effectiveness and enforceability of the loan according to the collateral agreement in the set of jurisdictions where the collateral could move during the lifetime. We are aware of the importance of enforceability of the collateral arrangement for movable properties, and we agree with EBA the need to discuss how to ensure that it is available for the institution. However, in our opinion, the requirement foreseen in paragraph 20 (d) is overly burdensome and its application would penalize some important sectors, such as shipping and aviation as well as the leasing business. It would be, in fact, operationally complex and costly to get a legal opinion for all the different jurisdictions in which the property could be moved, and as a final result, it would cause the exclusion of the eligibility for these type of collaterals.

In more details, we would like to raise some further concerns:

1) From our point of view, it is not clear why the effectiveness and enforceability of a collateral should also be assessed in the jurisdiction in which the debtor has his seat (page 30, paragraph 20, (a) and (c)). To our understanding, the legal effectiveness and enforceability of a collateral must be ensured only in the jurisdictions in which: (i) the (third party) guarantor is domiciled or the collateral is registered, as indicated in paragraph 20 (c); (ii) the bank has its seat, i.e. in its own legal area, as set out in paragraph 20 (a); and (iii) the collateral is located, as indicated in paragraph 20 (d) (see more comments on it below).

A review of the effectiveness and enforceability of a collateral security under the legal system of the debtor is not necessary and should not be required, with the exception of a possible accessory collateral right for which the existence of the secured main claim is a prerequisite.

In our understanding two situations can occur: i) the secured claim is governed by the same law of the state in which the institution is located or ii) it is exceptionally subject to third country law. In case the secured claim is governed by the law of the state in which the institution is located, the effectiveness and enforceability of the underlying contractual relationship must be ensured by a legally valid and enforceable bank agreement. By contrast, when the secured claim is exceptionally subject to third country law, e.g. British law, the bank will provide a legal opinion on the enforceability and the effectiveness of the secured contractual relationship.

Considering what stated above, we deem that the requirement of an assessment of the effectiveness and enforceability of the collateral also in the legal system of the country in which the debtor has his seat is redundant and would lead to a considerable and unnecessary financial expense.

2) With regard to movable assets to be collateralized we suggest that there should be no appraisal for the set of jurisdictions where the collateral could move during the lifetime of the loan as a consequence of the sale to third parties (para. 20 (d)).

We suggest to desist from trying to circumscribe the jurisdictions which should be taken into consideration and rather require the establishment of a risk based approach. Institutions would then have to assess in advance – on a risk based basis – which jurisdictions other than the place of registration/incorporation of the collateral could be relevant, and set the actions that can be taken in that case in order to mitigate noticeable/serious legal risks.

In the case of collateralized warehouses the claim of the bank always terminates with their ownership has been legally transferred to a third party acquirer. In this respect, a restriction could eventually be included in the respective collateral agreement, according to which the collateral may not be taken out of the country without the prior consent of the bank with a predetermined regular operation.

On the other hand, for pledges on single movable goods such as mobile cranes and construction machinery, (inland) ships, aircraft, or collective pledges on a number of movable goods such as motor vehicles, train components, etc., an appraisal of the legal validity and enforceability of the security according to the law of any conceivable state in which an airplane could land or a ship could dock or a moving machine could be used. We understand from the explanations in the background and rationale (items 3.3.1 paras. 16 et seq. and item 3.3.2 paras 18 to 20) as well as in the explanatory box to paragraph 20 and 21 that there is no intention to compel institutions to cover all jurisdictions where the collateral may potentially be moved throughout the term of the arrangement since this would be practically impossible and thus clearly disproportionate – and not commensurate to the risks involved.

Furthermore, a limitation to a number of jurisdictions specified in the arrangement would not necessarily help in this context: in fact, in case of a number of moveable collaterals such as ships, rolling stock or airplanes, contractual restrictions on the jurisdictions (expressly or implied) would severely diminish the value of the object for the collateral provider. Contractual restrictions would only be acceptable if limited to few and specific jurisdictions.

In this context, it should also be taken into consideration that the legal risk of the invalidity and unenforceability of a collateral arrangement is not a particularly high one in comparison to other risks. Rigid requirements regarding the scope of possible jurisdictions to be covered apart from that of the applicable law and the place of incorporation of the movable good could thus effectively impose very burdensome if not impossible obligations with very little effective value.

3) Shipping finance exception (similar considerations can be provided for the aircraft financing):

In shipping finance transactions the owner of the vessel is in most cases also the borrower (obligor) under the loan agreement and the mortgagor of the ship mortgage. The jurisdiction of the relevant ship register can be, but not necessarily is, the same place of the owner's/borrower's registered office. In this context, with respect to a ship mortgage, the legal opinion in the jurisdiction in which the bank (institution) is incorporated (20 (a)) is irrelevant as it is almost excluded that the ship mortgage will be enforced in this jurisdiction.
Thus, only legal opinions under the jurisdictions of the ship/mortgage register (enforcement) and the borrower's registered office (capacity) are required in most ship finance transactions.

Furthermore, as stated above for all movable collaterals, legal opinions stating the effectivity and enforceability against the obligor in the set of jurisdictions where the collateral could move during the lifetime of the loan, such as those foreseen in paragraph 20 (d) of the consultation paper, are not viable. Neither the loan agreement nor the mortgage/deed of covenants under a ship finance transaction can in any way (other than by sanction/general compliance with law provisions) limit the mobility of the financed/mortgaged vessel. In fact banks generally finance merchant vessels that do not just go from point A to point B, but rather sail the world's oceans, as any specific route is exclusively in the decision of its owner/manager. As a consequence, to comply with the provision of paragraph 20 (d) approximately 120 legal opinions would be required for each ship finance transaction, that is to say for each country with a sea port.

Eventually, it is worth to note that at least during the last 20 years, and in a recent extremely difficult environment for the shipping finance sector with numerous enforcements, there is not even a single case among at least the German ship finance banks where a ship mortgage was not recognized in an enforcement proceeding by any jurisdiction whatsoever.

In conclusion, we deem that for the shipping finance sector it is absolutely essential that the requirement as set out in para. 20 (d) of the consultation paper would not be implemented.
Question 3: Do you agree with the proposed clarification regarding the calculation of realised LGD on exposures covered by eligible on-balance sheet netting or master netting agreements?

We agree with the proposed clarification regarding the calculation of realised LGD on exposure covered by eligible on-balance sheet netting or master netting agreements. In our opinion the calculation is consistent with the usage of the adjusted exposure value by the netting effects for the computation of both the numerator and denominator of the realised LGD, without including into the economic loss calculation any recoveries after default.
Our main concern stems principally from the interaction of such recognition with the finalised Basel III framework. Indeed, the application of the input floors to the LGD and PD and the haircut on the AIRB models proposed within the finalised Basel III revision framework are, in our opinion, overly punitive, as the formula used to determine such floors considers the haircut according to the comprehensive approach (under the foundation approach), hence leading to an increased capital absorption in the presence of better quality collateral. The increased capital absorption would also impact the pricing and could result in a reduction of lending.

Further to this, the new framework introduces floors on internal haircuts equal to the ones to be used under the standardized and FIRB approaches. These floors, in addition to the LGD’s ones, would create a strong disincentive for banks to model internal haircuts. Thus the collaterals subject to the largest devaluation would be penalized while the better quality collaterals would not receive any benefit (given that the input floors are applied at single transaction level). Ultimately, this encourages banks to apply the standardized and FIRB parameter for the collateral evaluation even under the AIRB approach.

Furthermore, we recommend to align the standardised to the IRB approach with respect to the collateral that can be considered as eligible for credit mitigation purposes. This is the more important as the calculation of the output floor is based on the standardised approach (the difference between risk weighted assets under the IRB and the standardised approach does not reflect only discrepancies due to modelling practices but also inconsistencies between the two frameworks in terms of credit protection recognition).

As to how UniCredit currently recognise collateral in your LGD estimates, UniCredit does so in line with the EBA Guidelines on PD and LGD estimation. Specifically, in case of eligible FCP, and other FCP, for the Secured – unsecured approach: (i) the collateral is allocated to each transaction; (ii) the haircut volatility values are calculated; (iii) LGD like loss on secured components are calculated. The unsecured component is calculated and estimated in the same way as in the overall approach.

Alternatively, the overall approach can be used where: (i) the collateral is allocated to each transaction; (ii) the financial collateral presence is flagged and then treated as secured; (iii) the recoveries are used for LGD calculation as secured; (iv) collateral is included in variables (in particular flags and VTL).
We welcome a clear definition of handling collaterals to be eligible also for an assignment to multiple partners/obligors, as this it is a standard procedure in many countries, e.g. in Germany . It represents an important clarification with respect to the current TRIM guideline, where the collateral should be only assigned to a facility or at facility level.

With reference to the Substitution Approach, we understand the rationale of not allowing the migration of the guaranteed part of the exposure class towards the guarantor asset class for modelling purpose (e.g. large corporate with banks guarantee, the covered part would not migrate on banks but it remains as large corporate although with banks PD and LGD). In order to avoid potential misunderstanding, we would suggest to clarify, as done in the context of the Public Hearing held in Paris on April 15th 2019, that for regulatory reporting purposes the relevant migration is admitted consistently with the principle behind the following extract of paragraph 29 (a) (ii):

“when direct exposures to the guarantor are, or would be, treated under the IRB approach, substitution of both the PD and LGD risk parameters of the underlying exposure with the corresponding PD and LGD of a comparable direct exposure to the guarantor in accordance with paragraph 36.b of these Guidelines (i.e. ‘substitution approach’); […]”

Furthermore we would suggest to clarify that, even in the presence of the migration of the exposure class for regulatory reporting, in case of mismatch of the default classification between the guaranteed obligor and the guarantor (the former in default and the latter in performing status) the default classification should remain the guaranteed obligor’s one. This clarification would better explain the treatment required in the following extract of paragraph (29) (a) (ii):

“[…] in particular, in case the institutions have not received the permission of the competent authority to use own LGD estimates in accordance with Article 143(2) of Regulation (EU) No 575/2013 for direct exposures to the guarantor, institutions should substitute the LGD of the underlying exposure with the LGD value specified according to Article 161(1) of that Regulation; under this approach the following applies to defaulted exposures:
- the ELBE should be the expected loss which would have been assigned to the guaranteed part of the exposure after the substitution of the PD and LGD parameters in case the obligor was in a non-defaulted status;
- the LGD in-default should be such that the risk weight assigned to the guaranteed part of the exposure is the same as the risk weight which would have been assigned to the guaranteed part of the exposure after the substitution of the PD and LGD parameters in case the obligor was in a non-defaulted status;”.
Finally, to make comparable ELBE and EL for the guaranteed obligor in default and the guarantor in performing status, the removal of the Margin of Conservatism (as foreseen by EBA GL) in adopting the EL of the guarantor as ELBE should be considered. For the same reason the EL of performing should be adjusted in order to reflect the current economic condition as required for ELBE (and also to ensure consistency with IFRS9). This would represent the main challenge in applying the approach as it would entail an extension of the approaches currently required for ELBE also to EL for performing assets.

For the recognition of unfunded credit protection, UniCredit is currently adopting both the Substitution Approach and the LGD modelling approach, according to the type of UFCP.
We already highlighted in Question 4 the interconnection envisaged between the Guidelines and the implementation of the finalised Basel III standards regarding the input and output floors application, and the related need to foresee an alignment of the eligibility criteria between the Standardized and IRB Approaches on the AIRB perimeters where these floors are adopted. In our opinion both of them are extremely relevant elements that will need to be addressed in addition to the four listed issues (see box page 36) related to the eligibility and treatment of unfunded credit protection once the EU will implement the finalised Basel III framework .

As to the four listed issues (see box page 36) related to the eligibility and treatment of unfunded credit protection we deem the ones which would warrant an improvement are the following two:
i) treatment of unfunded credit protection provided by guarantor, when the direct exposures to the guarantor are treated under the foundation approach (FIRB). Indeed, the implementation of the new Basel framework foresees the dismissal of the advanced IRB approach for specific asset classes: as a result, banks with supervisory approval will use the foundation IRB (F-IRB) approach for banks, other financial institutions and large and mid-sized corporates (consolidated revenues > €500m) and the same treatment will be foreseen for unfunded credit protection provided by guarantor, when the guarantor falls within the abovementioned asset classes.

This treatment will entail the application of a fixed LGD value and predefined haircut penalizing above all unfunded credit protections provided by guarantors with better quality. Therefore, as already stated in question 4, we reiterate the need to align (in addition to Standardized Approach versus IRB in light of the Output Floor adoption) Advanced and Foundation IRB framework with respect to the collateral eligibility in order to mitigate the impact of the adoption of the FIRB haircuts on the input floor. Additionally, in case of substitution with guarantors under FIRB approach, for the covered part, unless an overlapping coverage is provided also with a funded credit protection, the LGD Unsecured to be applied would be equal to 40% (in the presence of Large Corporate guarantors) or 45% (in the presence of banks or other guarantors different from large corporate under the FIRB approach), whereas under the AIRB an LGD equal to 25% (as per para. 85 of the Finalised Basel III standards) input floor for fully unsecured part would be applied. Therefore, in the presence of FIRB guarantors under the substitution approach, the LGD for the covered part of the exposure would lead to a LGD sensibly higher than the input floor applied for fully unsecured exposure thus potentially hindering the mitigation effect both for the substitution and modelling approaches. More specifically:
- for the substitution, even in the presence of a guarantor with a lower PD compared to the guaranteed obligor (that is something to be expected ,), the LGD might potentially be higher than the LGD estimated by the model for unsecured exposures (that is another expected result considering the lower input floor for AIRB set at 25%) thus potentially offsetting the lower PD of the guarantor and removing all the mitigation effect;
- for the modelling approach, the same effect on substitution will apply also on the RW floor, thus removing the mitigation effect determined by the modelling approach on internal LGD estimates.
As a consequence, in order to avoid the unwarranted effect of not incentivizing the collection of high quality guarantees and protections thus ultimately worsening the asset quality of the banks portfolio, and ensuring the adoption of FIRB estimates as required by Finalised Basel III standards an alternative approach should be thought.

In order to recognize a mitigation effect specifically for large corporate and banks guarantors in case the underlying guaranteed exposures are treated at AIRB, a possible solution might be the adoption of an approach “secured-unsecured-like” (thus within the Modelling approach) envisaging the computation of a “LGD Secured” on the tranche of the AIRB exposure covered by large corporate or banks guarantors equal to the Expected Loss resulting from the adoption of internal PD and FIRB LGD (40%-45%). Indeed if the guarantor will pay the guarantees, the resulting LGD on the covered tranche of the exposure would be 0 otherwise, in case it fails in meeting its obligation thus going to default, the LGD would be equal to the LGD of the guarantor itself. This LGD secured value would contribute together with the guaranteed obligor’s PD at customer level and the internal LGD estimates on the other tranches of the AIRB exposures not covered by that large corporate / banks guarantor, to the computation of the RWA. In this way the mitigation effect would be at least recognized in proportion to the quality of the guarantor creditworthiness (the lower its PD the lower the EL; realizing therefore what we referred to as the “LGD Secured” for the part of the AIRB exposure covered by large corporate or banks guarantor). Clearly the LGD attributed to the LGD part not covered by the guarantees should be estimated without the inclusion of the cash flows stemming from the guarantors (or at least not recognizing the mitigation effects of the Unfunded Credit Protection estimated by the LGD model).

Overall AIRB Exposure = 100
Unsecured part = 50 (LGD applicable 35%)
Part covered by FIRB banks guarantor = 50
PD of the customer = 1,25%
PD bank guarantor = 0,8%
LGD bank guarantor = FIRB Value = 45%

LGD will be calculated as the combination of Secured / Unsecured (weighted by 50:50 unsecured: covered exposures) with an LGD Secured equal to 0,8%*45%=0,36%

LGD* = (LGDU * 50 + LGDS * 50) / 100 = 35% * 0,5 + 0,36% * 0,5 = 17,68%

The LGD* equal to 17,68% would be used together with the guaranteed customer PD equal to 1,25% for RWA computation. Clearly the application of the Basel input floor would be relevant for all the tranches not covered by the large corporate or banks FIRB guarantor, since for the latter the LGD Secured will be computed using an internal IRB PD that in turn will be subject to the Basel PD input floor - equal to 0,05% - and a Regulatory set LGD for Foundation. This would be equivalent to say that the LGD secured for large corporate / banks guarantors in the context of an IRB approach is subject to an input floor equal to 0,05% * 40% or 45% for large corporate and banks guarantor.

ii)On the use of an appropriate risk weight function for the purposes of computing the risk weight under the substitution approach, in our opinion it is important to ensure, for Regulatory Reporting purposes, the adoption of the risk weight function related to the guarantor.
As explained above, we consider inappropriate to use the cash flows derived from ineligible UFCP for credit risk mitigation purposes. Therefore we agree with the parallel treatment of ineligible UFCP and FCP. Nevertheless, banks should define specific cash flows tracking rules aimed to recognize them within the modelling activity.
Furthermore, in case of not eligible collaterals, the LGD estimation can be computed:
• separating secured cash flows related to not eligible collaterals, and considering them in the LGD calculation as unsecured cash flows;
• otherwise, if the separation between eligible and not eligible cash flows is not possible, all the recoveries are to be treated as unsecured. When the ineligible collaterals exclusion would drive to a biased LGD estimation, a dummy variable (presence of such collateral) and/or coverage variable (e.g. VtL) can be added to sterilize the effect. This would equally ensure that an LGD estimate not inclusive of the UCP mitigation would be recognized on the part of the exposures not supported by the UCP.
In general, we deem relevant to assess if the presence of both ineligible UFCP and FCP in the historical data drives to statistically significant mitigation effects. This would indeed be an objective criteria in order to understand if the presence of cash flows from ineligible credit protections might introduce a bias in the estimates and deserve appropriate treatment.
We understand the rationale of the proposed rules for the application of the substitution approach; however, in case the separation of cash flows deriving from UFCP was not possible, the substitution approach application could be implemented. The guarantor presence/absence can, in fact, be managed in the estimation phase by means of a risk driver, eliminating the mitigation effect in the application phase through a dummy variable equal to zero.

As an additional comment, we would suggest to clarify whether the substitution approach can be applied only if the execution costs are expected to be negligible, or if it is necessary to shift to the modelling approach should the costs be significant. With respect to this, we suggest in particular to clarify the following sentence of the Guidelines: “the institution may reasonably expect that the direct costs of exercising the unfunded credit protection are negligible with respect to the amount covered by the unfunded credit protection.”.
We agree with the proposed rules for the application of the modelling approach, where the unfunded credit protection is considered as a risk driver in the model development, in terms of dummy variable (presence of this collateral) and/or coverage variable (e.g. VtL).
A back-testing might be done by treating the Expected Loss of the guarantor as a sort of “LGD Secured” for the tranche of the exposure covered by the UFCP, which contributes to the overall estimated LGD (combining Secured and Unsecured) in order to compare this with the Realized LGD observed on the overall facility supported by personal guarantees. This kind of back-testing leads to consider the substitution approach similarly to the modelling approach (since it would consider the mitigation effect in the LGD modelling).
Overall, we agree with the proposed guidance for the estimation of the LGD of comparable direct exposure towards the guarantor. Indeed, we evaluate positively the requirement for consistency in the application of the substitution approach: “splitting the part of the exposure covered by a given unfunded credit protection in two parts and applying to one part the ‘substitution approach’ and to the other part the ‘modelling approach’ should not be allowed“.

We deem important to establish which is the appropriate criteria to choose which unfunded credit protection to use for the purposes of substituting the risk parameter in case of multiple unfunded credit protections that cover the same part of the original exposure. For the same situation, we would equally have some concerns on the possibility to satisfy the requirement foreseen by paragraph 37 (c), related to the calculation of the risk weight floor as the criteria to be used for the calculation of each risk weight direct exposures to the guarantor in case of multiple unfunded credit protection on the same part of the exposure is not clear. In particular, we wonder whether the effect of the other existing unfunded credit protections on the same part of the exposure (in accordance with paragraph 40) should even be considered or not.

Finally, we would recommend to add more details in the final guideline on how to consider the effect of credit mitigation in case of an exposure that includes an unsecured portion, although more collaterals (funded and unfunded) cover the whole exposure (see Figure 4).
Unfunded Credit Protection (UFCP) provided to a portfolio of loans rather than to an individual exposure is subject to a different mechanism, even in the event the credit worthiness of the guarantor remains unchanged. In particular, once the loss is realized and the guarantee activated two cases are possible:
a) in case the guarantee is provided to a single loan, the guarantee
isk mitigation is no longer active and it terminates its coverage effect;
b) whereas, when the guarantee is provided to a portfolio of loans, the occurred loss of a single loan does not lead to the termination of the risk mitigation effects because the contractual guarantee continues to be effective until its legal maturity (that shall not be shorter than the covered portfolio Weighted Average Life). This is a specific characteristic of all securitization transactions.

As a matter of fact, article 4(61) of the CRR states that a securitization is “a transaction or scheme, whereby the credit risk associated with an exposure or pool of exposures is tranched, having both of the following characteristics: (a) payments in the transaction or scheme are dependent upon the performance of the exposure or pool of exposures; (b) the subordination of tranches determines the distribution of losses during the ongoing life of the transaction or scheme”.

UniCredit considers the portfolio guarantees eligible as Unfunded Credit Protection based on the rules defined under Chapter 4 of the CRR for own synthetic transactions portfolio.
In particular, the applicable guarantee scheme for synthetic securitizations is the third case described in the explanatory box, i.e.: guarantee provided both in a form of “guarantee rate” covering only a part of credit losses on each loan, and “guarantee cap rate” up to a certain percentage of the portfolio”. In these transactions the risk of loans is divided in one or more tranches, the risk is transferred to third party; therefore the credit risk parameter applied on the part retained by a bank fully reflects the CRM effects of the UFCP. In order to recognize the CRM effects of the guarantee in these transactions, UniCredit adjusts the risk parameters leveraging on the “Substitution Approach” or applying the SA risk weight of the guarantor that UniCredit would assign to direct comparable exposure.

With regard to the treatment of portfolio guarantees, we consider that the effect of the UFCP should be independent of the eligibility rules applicable on the guarantor, and we recommend to allow the adjustment of the relevant parameters in case of both SA and IRB underlying portfolios. As a consequence, the UFCP schemes described in the draft guidelines shall operate and be used on both SA and A-IRB portfolios.
Costanza Bufalini