NATIXIS // 2021 Universal Registration Document

3 RISK FACTORS, RISK MANAGEMENT AND PILLAR III Risk management

In terms of exposure, the backtest applies historical Mark-to-Market (MtM) values of static backtesting portfolios to test the predicted futureMtM values. In terms of risk factors, exposuresare backtested ex post by gathering the historical prices achieved, even though retro-backtestingcapacities have been introduced for a sub-groupof products. The ex-post approach involves duplicatingthe transactions in the dedicated backtesting portfolios directly in the front-office systems. The ageing process is thus similar to that of real portfolios. Retro-backtesting uses a specific tool to retro-backtest MtM on the main product categories. Main internal models used by Natixis 3.2.4.8 Main internal models: PD, LGD, CCF and volatility discounts (EUCRE)

The results are factored in at transaction level for historical and retroactive approaches and at several aggregated levels, including the product type and the counterparty for the historical approach only. The transactionsconsideredfor static backtestingduplicated in the front-office portfolios were chosen in such a way as to ensure the representativenessof portfoliosand are discountedon an annual basis. Present and future prices are taken from the CCR (Counterparty Credit Risk) model engine, current prices being reconciled every quarter with those of the front-office systems, acting as reference.

Modeled input

Portfolio

Number of models Description/Methodology

PD

Sovereigns

1

Expert analysis-based rating models using macroeconomic criteria and the assessment of legal and political risks Expert analysis-based rating models based on quantitative criteria (accounting balance sheet) and qualitative criteria (questionnaire). Model by type of counterparty and by geographic area Expert analysis-based rating models by business sector for corporates and statistical models for SMEs (scores) Qualitative model based on internal and external defaults. The assessment of LGD during periods of decline is included insofar as all defaults are included for the LGD model Qualitative models based on internal and external defaults by type of counterparty. LGD assessed in this model include defaults occurring in periods of decline Statistical models (decision trees or assessment of recoverable assets) by type of financed asset. The safety buffers included in the LGD models serve to cover periods of decline (primarily via bootstrap techniques) Models used to assess assets on resale. Asset disposal assumptions are based on adverse scenarios Model calibrated on internal defaults and segmentation by type of product and type of counterparty Expert analysis-based rating models by type of financed asset Statistical models by business sector

Banks

3

Corporates (incl. SMEs)

12

Specialized Financing 6

10

Retail SMEs

LGD

Sovereigns

1

Banks

1

Corporates (incl. SMEs)

2

Specialized Financing 4

1

CCF

Corporate Financing (incl. SMEs), financial institutions and sovereigns Financial and other collateral

Volatility correction

4

Stochastic models based on historical market prices with assumptions based on internal data and expertise

The model mainly applies the following elements: pricing models for reassessing products on various trajectories; V modeling of the close-out netting framework; V distribution models of the factors distributed and called during V the valuation stage.

Main models used for counterparty risk Calculating the EEPE (Effective Expected Positive Exposure) requires simulating the potential changes in risk factors at future dates over a large number of scenarios using a Monte Carlo approach. The Mark-to-Market value of each transaction is then remeasuredat each simulated time horizon and under each scenario. For offsetting, exposure is then calculated by price aggregation taking into account the legal framework (master agreements and credit support annex) allowing price offsetting and offsetting for collateral exposures. The exposure value for a given counterparty is equal to the sum of the exposure values calculated for all offsetting sets.

128

NATIXIS UNIVERSAL REGISTRATION DOCUMENT 2021