We agree with the proposed definition, even though this definition would supposedly imply a not negligible increase of default rates and cure rates in jurisdictions where a more extensive definition of technical default was in place. The extent of such impact depends on the level of materiality thresholds that will be set along with the impact of probation period on multiple defaults.
Even though it is not object of this CP, we appreciated the variation introduced in the QIS provisions for exposures “other than retail”, switching from a relative “or” absolute threshold breached as preferred option outlined on EBA CP 2014/32 to a criteria of default as “both” triggers are breached. We believe this would better control the insurgence of “zero-losses defaults”. The system would be further enhanced whether similar provisions were extended to retail. A general adoption of a FIFO option, which is a standard payment allocation in Italy due to law prescriptions, where contracts do not provide different criteria of customers do not specify what payments are made explicitly for, would also provide a better balance in avoiding “zero-losses” defaults.
Insofar as the impact on models’ recalibration is concerned, different in extent under different options, we expect it to be larger on historical data than à regime. Thus some data cleasing on historical data might still be appropriate without assuming this to automatically require the introduction of specific margins of conservativism. For instance, as in some jurisdictions defaults were triggered at 180 days past due, than a backward simulation of the 90 days past due may itself lead to calibrate models on large amounts of default historically “cured” as the 180 days was approaching based on monitoring processes targeting this threshold. Some data cleasing and adjustments might be in such circumstances be required for calibrating unbiased risk parameters. This might be especially true for LGD where a longer time lag for calibration is expected.
As of Article 5 of RTS on the specification of the calculation of the specific and general credit risk adjustments (as adopted in REGULATION (EU) No 183/2014), Specific Credit Risk Adjustments relevant for default as specified in article 178(3) point (b) of CRR include those SCRA related to the credit risk of a single exposure or single obligor only.
We have some concern on the proposed article 25 of guidelines draft, for the part that assumes that every SCRA should be considered to be the result of a significant perceived decline in credit quality, thus innovating the CRR, not depending on the amount/materiality of the adjustment.
Explanatory text to this CP further extends this to situations where the amount of provisions is calculated collectively for a certain portfolio but all exposures in this portfolio are expected to lead to losses for the bank (note that this this rule is not explicitly included in the Guidelines). At this regard, the reference to expectations might be misleading since every exposure on the portfolio are expected to lead to losses with a certain probability.
Finally, insofar as SCRA can be individually or collectively assessed based on the decline in credit quality other than “significant” or on a portfolio bases based on forward looking expectations of specific geographic areas or industries, then assuming “significant decline” and extending the application from individually assessed SCRA to portfolio-assessed ones, would potentially imply triggering default according to CRR to large portions of portfolio.
As innovating article 5 of regulation 183/2014 with these guidelines may also raise some concern on the legal basis of such provisions, we would rather suggest to rely on concepts of “rebuttable presumption” in article 25 of the proposed regulation or rather to refer to such situations among other indicators of unlikely-to-pay as of article 47.
We agree that sales at significant discount are likely to occur under “unlikeliness to pay” conditions, and that ideally this should occur for positions already classified as defaulted.
We acknowledge CP provisions are mainly devoted to force the inclusion of such situations in risk parameters calibration, yet from a more general perspective we think a symmetric ruling for purchased would be most appropriate. Ideally, from the systemic point of view, default should never be discontinued by virtue of sales/purchases operations until a full restore of the credit worthiness is realized.
In this context, we think EBA might consider achieving asymmetry either relying to the general ruling for “unlikely-to-pay” potential triggering situations as of Article 47 for sales as well, or rather define symmetric objective ruling for credit origination/purchases, with the latter being more appropriate only if an effective and unbiased test and threshold can be set.
As we acknowledge objective ruling in this area is not straightforward, we think that the consulted test is subject to possible shortcomings.
We comment here on such shortcomings that might be mitigated by revising some provisions but that taken together might suggest excluding objective tests in the form proposed relying rather on more general Article 47 provisions.
Firstly, a higher threshold (10%) would better prevent unintended results since discount on defaults, if not subject to SCRAs already triggering default, would normally be much higher than such a threshold.
Secondly, under proposed guidance the “loss test” should be performed, and thus default would be triggered, either at the level of the overall transferred credits or at the level of individual credits depending on the price setting scheme of the transaction.
At this regard we think that seldom a real “facility-based” price setting can be envisaged when the operation involves a large number of obligations. Risk Parameters figures might be biased and price setting schemes be strategically oriented so that losses are fictitiously allocated to a smaller number of credits. This is particularly true if both “already default” and performing credits are sold within the same transaction, with the latter not triggering new defaults. Banks under Foundation IRB will have an incentive to concentrate losses on a small number of transactions or on “already default” ones. Banks under retail IRB/Advanced IRB will also have an impact on forward-looking LGDs estimates, and thus incentives are less clear. Banks under Standardised-approach are most likely to be not biased in the definition of the price-setting schemes. We believe that homogeneity of application would be enhanced by introducing some counterbalancing of incentives extending rules to purchased credits and possibly by identifying a standard “price splitting” rule independently on the price-setting scheme assuming proportionality of the price (and thus of the discount) to the net balance sheet value of each individual credit included in the transaction.
Furthermore, default classification implies default propagation to unsold credit obligations to the same counterparty. This can be particularly tricky when default is triggered at overall sold portfolio level where not necessarily every transferred credit would otherwise have been assessed as “unlikely to pay”. Propagation of default can be undue even in some circumstances where material loss is tested at the level of the individual credit: 1) where a “transaction-based” definition of default is applied for retail facilities, such automatic propagation to other sold or unsold obligations should be excluded; 2) where a “counterparty-based” definition of default applies, then propagation should take into consideration the materiality of sold credit experiencing loss to the overall counterparty position including other obligation with no to little discount and unsold obligations.
Since proposed objective rules would trigger the “unlikely-to-pay” classification, we suggest that guidance as of article 37 should be revised in order to allow that for banks applying a “transaction-based” definition the loss test performed at specific-credit level triggers default on the specific credit, unless general pulling triggers operationalized as of article 47 (h) apply.
Similarly, when the test is performed globally on all sold credits, under a “transaction-based” definition defaults shall be triggered on sold credit facilities only; extensions to unsold credits should be regulated by potential “pulling effect” triggers operationalized as of article 47 (h).
Finally, when a “counterparty-wide” definition of default applies, as we earlier suggested, the test shall be performed rather considering jointly all sold obligations and adjusted for materiality. For instance assuming that materiality is triggered when either obligations sold at material discount are over a certain percentage (20%) of the overall credit obligation or when the sum of the discounts exceeding the threshold of the loss test plus any past dues amounts on other sold and unsold credits exceed standard materiality thresholds for past dues.
We would rather consider effective interest rate applicable at the moment before signing the arrangement as economically more relevant expecially as the level of threshold set is rather low (1%). Were the customer’s original effective rate we would suggest a material revision of the threshold set (for instance, 5%).
As we commented on Question 4, we support a symmetric treatment of sold and purchased credits. A default shouldn’t be discountinued with credit transfer operations unless the credit worthiness of the counterparty is restored.
As we acknowledge and commented earlier from the sell side perspective, objective ruling in this area is not straightforward. We thus support to restore symmetry by rather requiring internal operationalization as of article 47 for sold credit as well. Were EBA confirming the adoption of objective triggers for sold credit , possibly modified in order to limit shortcomings, similar ruling might be adopted for purchased credits as well.
We agree that a probation period shall be set in order to avoid multiple defaults. We think that proposed thresholds for distressed restructing and for other defaults are appropriate in most but not every case. Among such cases, we think that since materiality thresholds as of EBA CP 2014/32 may still intercept situations of zero-loss defaults, then were default triggered by past due only a lower probation period, if any, should be allowed were past dues amounts fully repaid.
As we fully agree that some unlikely to pay indicators are inherently obligor-wide and that cases of propagation of default to the obligor shall be ruled as of article 47 (h), we partially agree with the proposed level of application of the definition of default since we would have rather preferred a greater flexibility (ref. option c. of paragraph 5.1.D.a of CP).
We acknowledge CRR provisions might have admitted a much stricter interpretation of “obligor vs facility” option, assuming it was consistently spread across the institution without taking into consideration any specificity. We appreciate this guidance will allow a greater flexibility allowing mixed-schemes particularly valuable for more complex institutions.
We do not fully agree that this flexibility finds a limit in the last proposition of article 74 where it refers to the fact that “there is evidence that the number of situations where the same clients are subject to different definitions of default at different levels of application is kept to a strict minimum”. The reason for this is that we would consider valuable, in a context where the definition is facility-specific, that some interrelated facilities could be considered on a comprehensive basis.
Retail businesses in some jurisdictions are often financed through credit lines that may be used as overdrafts supported by advances on notes. These different facilities belong to the same transaction and past dues are normally reported on the riskier usage line (the current account) but jointly generated, making ideally advances never default. Such transaction may involve different usage lines/current accounts and default may occur on one line, the other or both depending on the past due registered by virtue of automatic detection.
Consumer credit, depending on the business model of an institution or subsidiary for more complex institutions, may face similar tricky situations. For instance, a credit card would eventually be charged on the current account without prior check of available funds, in so doing making the latter to default while past due was triggered by credit card usage.
Installment lines are normally more segregated, but unsecured lines may follow the same scheme where expired installments might be charged on the account without prior check of available funds, as credit cards normally are. The same would seldom happen for mortgage or other secured lines installments. Such situation are obviously particularly cumbersome to deal with for EAD/CCF modelling purposes as well, sometimes leading to an aggregate definition of CCFs.
As it is allowed that a “facility-based” definition of default is applicable for retail, and this can be appropriate for risk reasoning and not only for reasons related to the complexity of an institution or to its business model, and taking aside inherently obligor-wide unlikely to pay triggers, it would be sound allowing institutions to transparently avoid such situations and adopted a definition of default that is the most appropriate. For instance, establishing a layer of facilities where the definition shall be transaction based and a more undifferentiated layer of facilities where the definition should better be considered on a comprehensive basis, would be valuable in our view, as this choice will be well supported, documented and applied consistently over time.
This objective, where considered valuable by EBA as well, might be achieved in four ways:
i) allowing the co-existence of different level of definition of default for the same counterparty as long as the involved facilities belong to different IRB portfolios (such clause would allow segregating mortgages, for instance);
ii) allowing an interpretation of “facility” that may encompass more than one credit/usage line and exposure where they are inherently interrelated;
iii) allowing explicitly both;
iv) exclude explicit limitation to flexibility from guidance, leaving CAs to assess the appropriateness of the application depending on the specificities of jurisdictions, institutions, business model, etc.
The principle introduced in article 47 (h) and 80 is appropriate. As the guidance include objective thresholds in many areas for sake of homogeneity in application, we would have considered a stricter rule appropriate in this area as well. In substance, we would have considered appropriate option d. at paragraph 5.1.D.b of the CP with the form of a general reference to “supervisory reporting framework” regulation similarly to the reference to IFRS stage 3 as default trigger.
Since the accompanying document refers to “possible future changes in the supervisory reporting framework” as one reason for not introducing here such prescription, a general reference would make regulation resilient to such changes.
Of course, a stronger/fixed threshold rule do not exclude institutions will address more complex situations under article 47(h) related internal procedures. This not meaning that we would have endorsed option c. at paragraph 5.1.D.b since such procedures may rather involve more complex propagation-assessment schemes that a simple lower threshold wouldn’t be able to capture.
We fully agree with the approach proposed and that this will enhance significantly homogeneity of practices.