We agree in general with the EBA's proposed approach for defining the scope of application of the securitisation purchased receivables approach.
However we deem that, every time the Bank enters into a securitization deal as Investor (i.e. non-Originator), irrespectively it be Servicer or not Servicer, the “top down” approach defined in this RTS should be applicable. Indeed in the case of an Investor operating as a servicer, the securities exposures are not originated by the Investor Bank and therefore the IRB internal estimates would suffer from a limited representativeness between the data adopted for the estimation and the perimeter of application. Therefore we would deem more reasonable to foreseen the following combination:
1) Originator and Servicer: IRB Approach;
2) Originator and not Servicer: RTS “top down approach”
3) Investor and not Servicer: RTS “top down approach”
4) Investor and Servicer: RTS “top down” approach (since the securitized exposure are not originated by the Investor Bank and IRB estimates might be not fully representative, especially for Mid-Corporate, Retail portfolios)
n general we do not see the necessity to foresee exceptions to the approach defined in the RTS for specific type of obligor, e.g. large corporate. Indeed Large Corporate (that, following the definition adopted within the Basel 4 paper - i.e. “Basel III: Finalising post-crisis reforms” December 2017 - are related to Corporate with consolidated turnover above 500 mio EUR), although covered by PD models developed leveraging on publicly available information like financial statement, in order to ensure the information completeness required by EBA GL on modelling (i.e. EBA/RTS/2016/03 and EBA/GL/2017/16), typically includes also qualitative/soft information that plays an important role in the rating assignment. In the context of adopting a “top down” approach, asking for diversified and large in number pool of securitized exposures (e.g. revolving exposures, trade receivables, etc) the adoption of internal IRB models for PD estimation might become unduly burdensome in terms of costs for gathering information in case the Bank is not servicer and complex rating assignment process for the PD attribution, thus falling under the criteria reported in Article 6, par. 3.
Thus we deem that no exceptions should be foreseen for large corporate given the potential complexity on PD estimation side, but rather to allow also for pool of large corporate securitized exposures fulfilling the requirement of the RTS to benefit from the higher simplicity and standardization of the “top down approach”. Moreover, with regard to LGD, irrespectively from the adoption of an IRB PD estimates (assuming it were not unduly burdensome), due to Basel 4 reform, in the future the LGD will be purely based on Foundation values. Consequently, it is deem advisable not to create specific exceptions to the top down approach, with the aim of estimating Expected Loss values for homogenous clusters of the securitized portfolio, also considering that the adoption of the Foundation LGD values, that will be mandatory for Regulatory Purposes once Basel 4 paper will enter into force, will allow to easily retrieve the PD value.
We basically agree that in a context of investor (i.e. credit protection provider) positions in Synthetic securitizations, Article 4 principles would result hardly applicable. Thus in order to allow Banks to exploit the possibility to invest into synthetic securitization we deem that EBA shall define proper rules considering the specificities of the synthetic securitization, in particular by creating the proper regulatory background in order to make available the information set for the adoption of a “top down” approach.
We deem that a more detailed definition of proxy data is definitively necessary. In addition we deem fundamental that the EBA explains its views on different types of proxy data and their relationship to margins of conservatism. In particular, from an pure modelling standpoint, we deem that data obtained from specialized data warehouses (such as the European Data Warehouse) will necessarily be a primary type of data for institutions to be used in developing top down models. Indeed, these data should be considered as broadly representative and relevant for the purpose of developing inputs to be applied to receivables originated and serviced by third-parties compared to the Bank that should calculated the K-IRB. On the other hand, data directly related to the portfolio of securitized exposures in which the bank is investing may be much more limited either in terms of number of exposures or number of historical data points. Consequently, the availability of data different from proxy ones may be too limited for a sound estimation process, making the recourse to proxy data like the ones related to specialized data warehouses potentially unavoidable. Therefore data sourced from specialized ware houses, by being treated as proxy" data and subject to margins of conservatism, may result in unduly high risk estimations, which would hinder the adoption of the top down approach and the expected prominent role in the context of the SEC-IRBA approach adoption. Therefore it should be evaluated a more granular taxonomy that might be: primary date (i.e. directly related to the securitized exposures), Level 1 proxy data (i.e. related to specialized data warehouses, on which EBA might provide a specific list of recognized Data Provider), Level 2 proxy data (other), where for Level 1 the inclusion of specific margin of conservatism should be basically limited (and specific Guidelines from EBA on how to quantify this MoC for Level 1 and Level 2, would be very useful in order to foster modelling practices harmonization across the banking system)."
In general, a framework for KIRB calculation is necessary also for securitization of NPL exposures considering that the securitization can be adopted as one of the suitable instruments for NPL reduction, that in the current historical phase is also a political priority at EU level. In the context of NPL portfolio securitization the three following possible cases might occur:
1) the Bank is Originator and Servicer
2) The Bank is Originator but not Servicer
3) The Bank is Investor (Servicer or not Servicer).
The case 1) would fall under the adoption of internally developed IRB LGD in-default estimates. Indeed, even if the NPL portfolio is de-recognized with a true sale securitization, the Bank continues to operate as a servicer (i.e. it continues to manage the workouts according to its recovery process) thus the adoption of LGD estimates is possible.
The case 3) is expected to be a very limited situation except for Banks specialized (or having specialized business divisions) in investing in troubled credit assets originated by other banks. In this case, irrespectively if the Bank is servicer or not servicer for the same reasons reported in Q1, a “top-down” - like approach is required, foreseeing the estimation of LGD values for pool of homogenous cluster of non-defaulted exposures. In this regard the adoption of the draft RTS might have limitations if it is required to consider time-in-default and trend of recoveries realized so far as risk drivers in line with what required by EBA GL on PD estimation, LGD estimation and treatment of defaulted assets. Indeed, in addition to the availability of the information for each NPL exposure in the securitized portfolio, it is necessary to have these two information regularly updated in order to capture the behaviour of the recovery in the ongoing review of estimation. Considering the potentially limited availability of this information in case of Investor positions, it would be advisable to extend the facilitations foreseen within the Article 5 on risk differentiation and to envisage simple solutions for the calculation of the LGD in-default for these cases. Possible approaches might be:
- leveraging on LGD retrievable from “top down” approach estimation for performing assets with the inclusion of standardized conservative adjustment in order to account for the defaulted nature of the securitized exposures;
- leveraging on average values of internal IRB LGD estimates for the same pools of homogenous exposures of the securitized portfolio but calculated on the internal portfolio of the Bank, with the inclusion of Margin of Conservatism being this approach an approximation;
- adoption, if available and not unduly burdensome in terms of costs and data collection, of data on recoveries provided by the servicer.
The case 2) is basically similar to the case 3). Indeed the recovery process is carried out by a third servicer making the adoption of the internal IRB LGD estimates on line-by-line basis not feasible both for potential informative limitations and lack of representativeness with respect to the servicer’s workout processes. Thus the same considerations expressed for case 3) are valid. The only one possible exception might be the Bank, although not being servicer, also in its internal workout process outsources the recovery process to a third party and the latter one is also the recovery servicer of the NPL securitization. In such a case the adoption of the internal IRB LGD in-default estimates might be valid subject to an information flow from servicer to Originator Bank allowing the application of the internal LGD in-default estimates on a line-by-line basis. In this case, and unless the adoption of this approach might result unduly burdensome (particularly in terms of costs) as for the criteria reported in Article 6, par. 3 (in which case the adoption of a “top-down” –like approach would be needed), the internal LGD in-default estimates can be adopted and we ultimately fall under the same situation of case 1).
We report below the additional further comments:
- with regard to Article 5 - General Conditions for risk differentiation, we deem relevant to provide more details regarding the information falling under “originator’s underwriting standards”, “servicer’s recovery practices” and “servicer’s standards”. Indeed, since one of the main purposes of the “top down” approach for securitization is to apply easier methods given the lack of information, risk drivers falling under this categories should be better specified, envisaging a minimum list, in order to avoid potential inconsistent application of the regulation as well as potential inconsistent assessment by the different National Competent Authorities. In this regard a possible approach might be to leverage on the already issued EBA RTS on the homogeneity of the underlying exposures in securitisation under Art. 20(14) and 24(21) of [Regulation (EU) XXX/201X … laying down a general framework for securitisation and creating a specific framework for simple, transparent and standardised securitisation]. In that context the mapping risk factor and asset category envisages the creation of homogeneous clusters, allowing for the adoption of simple top down approach (thus “fitting for purposes”) and ensuring a proper consistency and synergies with other Regulatory products related to the new securitization framework
- with regard to articles 6 and 7 it would be useful to further explain the reason why the condition in Article 6, par. 2, point a. is relevant only for Retail.
- more in general, it would be extremely relevant to report a more clear mapping of the sections related to EBA GL on PD estimation, LGD estimation and defaulted assets treatment tailored for the adoption of the “top down approach” in the context of securitisation. Indeed, the current draft, although providing extremely useful indications, might be enhanced in this sense in order to avoid mis-specifications and mis-interpration by different National Competent Authorities within the approval process. Indeed EBA GL on PD and LGD modelling results particularly strict in providing clarification of CRR Articles. Therefore, within the correct approach of providing specifications tailored for top-down approach versus CRR, it is deemed relevant to make the similar exercise also with respect to the detailed requirements of EBA GL on PD and LGD modelling.