ASF share the view of the EBA that modelling downturn LGD and CF is a difficult task. Against this background, ASF recognises that the EBA proposal is constructive, especially in its analysis of the payment recovery process in case of default. We agree with the limit that is mentioned in the credit activity. We agree with the limit that is mentioned in the credit activity.
Yet, we estimate that the proposed method suffers from its complexity. We agree with the limit that is mentioned in the credit activity. ty and cost. It carries the risk that the efforts to implement the approach in the modelling process are burdensome for uncertain results, since the correlation between the proposed factors and the real causes of the losses is not established.
The model component approach requires eventually a heavy workload, and at the end, seems not to achieve more interpretable results compared to the alternative approaches.
The objective of modeling of the two parameters LGD and CCF is different, and based on different starting point of observation. For instance, for residential loans, the Credit conversion factors is to be determined at the date of default, whereas the reality of the loss may be known only after several months, or even after years in case of legal procedures.
Another limit is that some specialised credit activity – for instance Consumer credit activity, and especially in revolving credit - the downturn conditions are not specifically linked to external macroeconomic parameters.
We consider that the presentation of the approach is not immediately understandable, but that the proposed example is relevant and brings clarity, especially for residential loans activity. It would be better to have other explicit examples for credit activity.
We notice that it is specified that the model components is based on historical defaults which are not risk factors (input factors for non-defaulted LGD estimation).
We consider that some elements of the proposed list of economic factors to be considered deserve to be more precisely defined. For instance, “default rates and credit losses from external data” seems difficult to collect, and the “tax benefits” impact seems difficult to model and turn into usable statistical data. The collection of external data concerning other institutions (for example default rates) is especially difficult in the case of specialised activities, that are sometimes “niche” activities.
Furthermore, institutions need to find out additional economic factors, besides the only factor: ex. ”consumer price index” for other retail.
Another example of operational concern is the collection of “regional specific indexes” if it is supposed to be infra-national data.
These elements should be more precisely determined in the RTS.
A solution would be that the NCAs are responsible of the collection and delivery of these data. Otherwise it would be difficult for the institutions to justify to the supervisor that they cannot respect the RTS requirements.
We globally agree with the proposed approach, but we underline that it would be quite difficult to implement due to the required length of historic.
We consider that one year duration would be suitable.
ASF consider that the proposed methodology is clear, but wish to raise the issue of the availability of some statistical data, especially on a 20 years historical period. For instance, on the residential loans market, infra-national data may not be available on such a long period.
ASF understands the approach that consists in considering the worst level in a period of 20 years for each economic factor. Yet, we think that a methodological issue is raised by the fact that some economic factors may not be correlated one to each other: they do not react to the same causes and may reach their worst level at different time of the period. This could lead to difficulties in determining a coherent final downturn scenario based on economic factors.
The Guidelines should be more precise on this issue, and would be more efficient if some examples were given.
We consider that the articulation between the RTS and the Guidelines is not efficient. The definition of the elements to consider that are listed in the RTS, and the methodology to estimate LGD and CF given in the Guidelines, are directly linked. We would then recommend that the elements of the Guidelines are included in the RTS. According to us, this would serve the objective of reducing the variability of the models results.
We consider that the Model Component approach, whereas relevant on the paper, could lead for some specialised activities to non-coherent results, or at least raise implementation issues in terms of complexity (access to data) and cost.
Considering this risk, and according to the principle of proportionality, we would advocate for the possibility given to specific business models and/or risk profiles to use a simpler pertinent alternative approach in case it seems not possible / justified to use the Model component approach.
For some activities, the economic downturn is not always the key parameter of a downturn LGD. For instance, considering credit activities, the evolution of the legal background and external elements other than strictly economic may have a major impact on LGD. It would not be efficient to impose the complex and costly use of the Model component approach to these institutions as it is predictable that it will not lead to results relevant for credit risk analysis.
We do not consider that the “Reference value” approach would be more efficient, as it is based on the same methodology.
Therefore, we would support the ability to use the “Supervisory add-on” approach, but we consider that the proposed actualisation rates are not reasonable. They correspond more to severe stress tests than to a regular credit risk analysis modelling. The application of such a level of actualisation rates would reduce excessively the risk sensitivity of the models, denying significantly their advantages compared to the IRB – F or even the standardised approach.
We would support the ability to use the “Supervisory add-on” approach, in “the spirit of proportionality”.
But we underline the fact that this approach relies on a balanced calibration of the add-on by the supervisors in order to guarantee that it is determined considering the specificities of the business models. Otherwise it would annihilate the risk sensitivity of the internal models.
Some Consumer finance institutions currently perform an approach that is close to the supervisory add-on approach.