NATIXIS_PILLAR_III_2017_EN

CREDIT RISK Credit risk: internal ratings-based approach

The change in the portfolio’s credit quality over one year is also analyzed by looking at internal rating migrations. Additional indicators are also calculated to verify the internal risk ranking (Gini Index, average rating, previous year’s ratings, ratings of counterparties that have defaulted) and provide statistics as a supplement to the qualitative analyses. CANO Meetings are chaired by the heads of the Individual Risk and Consolidated Credit Risk departments within the Risk division, or by their representatives. The follow-up on the decisions made during Committee Meetings are presented at subsequent meetings, particularly if thresholds have been breached and this situation has not been rectified. All of these analyses are also presented each quarter to the Chief Risk Officer and sent to the regulator. Monitoring and backtesting of internal LGD, CCF and ELBE under the advanced method The LGD, ELBE (see glossary) and CCF (see glossary) levels for the different lending scopes are backtested at least once a year (based on internal data), as are the rating models and the associated PD, to verify the reliability of the estimates over time. LGD, CCF and ELBE backtesting is carried out by the Risk divisions teams to: The parameters of the models for the specialized financing and collateral (financial or other) scope are regularly updated, so that they reflect actual conditions as accurately as possible. Both the market parameters and the recovery parameters are updated. The losses and estimates produced by the models are compared based on historical data covering as long a period as possible. The indicators defined for backtesting are used both to validate the model and measure its performance. Two types of indicators are used: verify that the model is correctly calibrated; a assess the model’s discriminating power; a assess the model’s stability over time. a

population stability indicators: these analyses are used to verify a that the population observed is still similar to the population that was used to build the model. The model may be called into question if the segmentation variables or the LGDs result in excessively large distribution differences. All of these indicators are compared against the benchmark indicators (usually those calculated when the model was built or those issued by external data or agencies). These analyses are applicable to both expert appraisal-based models and statistical models; model performance indicators: the model’s performance is a measured to validate the segmentation and also to quantify, overall, the differences between the forecast and actual figures. This is achieved by using statistical indicators, which are compared against those calculated during modeling. Losses given default models (internal LGD) are calculated: on a statistical basis for the corporate asset class; a based on internal and external histories and an external a benchmark for banks and sovereigns; using stochastic models if there is a claim against a financial a asset. The results of the backtesting may result in the risk parameter’s recalibration, where appropriate. A backtesting report is produced once backtesting is complete. This report includes: all the results for the backtesting indicators used; a any additional analyses; a an overall opinion of the results in accordance with the Group’s a standards. The report is then submitted to the internal validation teams (Model Risk Management) for an opinion, then presented to the various committees to inform the bank’s management.

5

TABLE 28: BACKTESTING OF LGDS AND PDS BY EXPOSURE CLASS R

Figures resulting from backtesting

Observed default rate

Observed LGDs

Model LGDs

Estimated PD

Sovereigns

31.30% 37.91% 29.35%

48.20% 50.44% 40.45%

0.23% 0.26% 0.42%

6.52% 1.15% 0.88%

Financial institutions

Corporates

This table provides a general summary of the system’s over an extended period and for a significant, representative performance but differs from the annual backtests carried out percentage of each exposure class. The results come from data within the Group, which are conducted on a model-by-model warehouses used for modeling. This is based on all performing basis and not overall by portfolio. However, this table allows a customers for default rates and PD, and on all customers in comparison of estimates and actual results for each internal input default for LGD.

75

NATIXIS Risk report Pillar III 2017

Made with FlippingBook flipbook maker