In EBA/GL/2017/16 paragraph 73 it is explained that in the calculation of the one-year default rates the number of obligors is used. Is it always mandatory to count one legal entity as one obligor, or can a group of legal entities for which there is no difference in risk due to an appropriate guarantee (i.e. those will default together) be counted as one obligor?
In corporate portfolios there can be customers that consist of multiple (and even many) legal entities. In some situations, these entities will always default together, e.g. because of the existence of appropriate guarantees. The PD calibration will depend on whether the group of legal entities that default together is counted as one obligor or each individual legal entity is counted as one obligor for the calculation of the one-year default rate. Counting each legal entity separately, would bias the PD towards companies with large group structures and it would increase correlation between defaults.
Without more guidance on this topic there is a risk that each credit institution will have its own definition and interpretation of what an obligor is and correspondingly would have different PD estimates for an exactly equal portfolio. This would contribute to an increased variability in RWA across institutions.
According to Article 172(1)(d) of Regulation (EU) No. 575/2013 as amended by Regulation (EU) 2019/876 (CRR2), each separate legal entity to which the institution is exposed shall be separately rated. As such, the existence of a group of connected clients defined in Article 4(1)(39) CRR or the existence of third-party rating support as described in paragraph 62 of the Guidelines on PD estimation, LGD estimation and the treatment of defaulted exposures (EBA/GL/2017/16) is instead to be considered in the assignment of these separate obligors to rating grades or pools. As a matter of fact, the mere existence of an ‘appropriate guarantee’ or the existence of a rating transfer does not necessarily imply that the obligor and the third party represent a single risk with only joint defaults possible. In particular, the default status should be assessed according to the internal policies of the institution, as specified in paragraph 61 of said Guidelines, and is not necessarily automatic.
According to Article 180(1)(a) and Article 180(2)(a) CRR, institutions shall estimate PDs by obligor grade from long-run averages of one-year default rates. On this regard, Chapter 5.3 of the above mentioned Guidelines clarify the CRR requirements for the purpose of calibrating PD estimates. More in detail, for the purpose of calculating the one-year default rate referred to in Article 4(1) (78) of the CRR, paragraph 73 of the GL clarifies that the denominator “consists of the number of non-defaulted obligors with any credit obligation observed at the beginning of the one-year observation period”.
It follows that, irrespective of “the treatment of individual obligor clients and groups of connected clients” for the purpose of assigning an exposure to obligor grades or pools in accordance with Article 172(1) (d) of the CRR, each separate legal entity to which the institution is exposed constitutes an obligor for the purpose of calibrating PD estimates to the “long-run averages of one-year default rates” as defined in Article 4(1)(78) CRR.
However, where there is a significant proportion of connected obligors with a high default correlation within a considered PD model, institutions should analyse the level of risk of such obligors compared with other obligors in that model and, where necessary, reflect the difference in the level of risk in the model through appropriate risk drivers and rating criteria, for instance through different calibration segments.
In that case of high default correlation of connected obligors institutions should ensure that the lack of independence of defaults does not lead to biased results or wrong conclusions during the development or the validation of the PD model where one year default rates are computed, in particular if this high default correlation of connected obligors in the model implies a violation of the assumptions required by the underlying statistical methodologies (e.g. in the validation of model performance) in accordance with article 180(1)(d).
Please note that the differences in the treatment of unfunded credit protection and rating transfer are further clarified in Q&A 4598.