BPCE - 2019 RISK REPORT Pillar III

5

CREDIT RISKS

RISK MEASUREMENT AND INTERNAL RATINGS

Review of internal ratings-based models

The Groupe BPCE Risk division is responsible for reviewing the Group’s internal models whenever a new model is being developed or an existing model changed. It also performs the annual review of backtests on credit, market and ALM risk models. The validation team conducts independent analyses in compliance with a charter and procedures that describe interactions with the modeling entities and the steps of the review. This review is based on a set of qualitative and quantitative criteria, and mainly addresses the following points: documentation; • methodology, including the validity of assumptions; • performance; • robustness; • compliance with regulations. • The Risk division maps out all Group internal rating models, clearly indicating their scope in terms of Group segments and entities, as well as their main features, including a general score derived from the annual model review characterizing the performance and freshness of each model (age/year of development). New models were recently added and are in the process of being approved by the ECB. The models in question are PD rating models for “individual retail” customers and LGD estimation models for “individual retail” and “professional retail” customers. The new methodology for PD rating models aims to improve predictive power over customers without payment incidents. The new LGD calculation methodology aims to distinguish losses in the event a customer is downgraded to “disputed” (material loss) from losses in the event a customer is quickly restored to “performing” status (non-material loss Model mapping

The level of detail in the review is adjusted for the type of work examined. In any event, it must at least include a document review focusing on the quantitative aspects of rating systems. For a new model or a major change to an existing model, in addition to this review, the computer codes are checked and additional tests are run (comparative calculations). The scope of the Validation division’s involvement may be expanded before and after an investigation of data quality, system implementation and operational integration. In conclusion, the review generates an opinion on the validity of the models and the associated inputs for credit and counterparty risks, and for models authorized for use in determining capital requirements. It also generates an opinion on compliance with prudential regulations. Where necessary, the review is accompanied by recommendations.

stemming primarily from admin costs). These models will be added to the inventory once they have been approved by the ECB. Efforts were also focused on overhauling the models used to rate “professional retail” customers and to estimate exposure at default (EAD) for the both the “ individual and professional retail” customer segments. These new models, developed in 2018, were approved by the ECB in 2019. The following table lists the internal credit models used by the Group for risk management purposes and, where authorized by the supervisor, to calculate capital requirements for the Banque Populaire and Caisse d’Epargne networks, Natixis and its subsidiaries, Crédit Foncier and Banque Palatine.

Number of CCF/EAD (Exposure At

Number of LGD (Loss Given Default) models

Number of PD (Probability of Default) models

Default) models

Exposure class

Description/ Methodology

Description/ Methodology

Description/ Methodology

Portfolio

Portfolio

Expert criteria including quantitative and qualitative variables

Expert criteria including quantitative and qualitative variables/economic and descriptive variables Portfolio with low default risk

Application of regulatory inputs

Sovereigns and affiliates

Sovereigns and affiliates Multilateral development banks Municipalities (communes), departments, regions, social housing agencies, hospitals, etc. OECD or non-OECD banks, brokers/dealers

Sovereigns, central governments and central banks

1

1

1

Expert criteria t Portfolio with low default risk

1

Expert criteria/statistical modeling (logistic regression) Portfolio with low default risk

Public sector

10(NA*)

Expert criteria including quantitative and qualitative variables

Application of regulatory inputs

1

3

Expert criteria

Banks

1

Institutions

84

RISK REPORT PILLAR III 2019 | GROUPE BPCE

www.groupebpce.com