Société Générale / Risk report 2021 - Pillar 3 2020

6 CREDIT RISK RISK MEASUREMENT AND INTERNAL RATINGS | SOCIETE GENERALE GROUP | PILLAR 3 - 2021 85 TABLE 30: MAIN CHARACTERISTICS OF MODELS ANDMETHODS – WHOLESALE CLIENTS Parameter modeled Portfolio/ Category of Basel assets Number of methods, models Methodology Number of years default/loss WHOLESALE CLIENTS Sovereigns 1 method. Econometric method, low default portfolio. Public sector entities 4 models according to geographic region. Statistical (regression)/expert methods for the rating process, based on the combination of financial ratios and a qualitative questionnaire. Low default portfolio. Probability of Default (PD) Financial institutions 11 models according to type of counterparty: banks, insurance, funds, financial intermediaries, funds of funds. Expert models based on a qualitative questionnaire. Low default portfolio. Specialised financing 3 models according to type of transaction. Expert models based on a qualitative questionnaire. Low default portfolio. Large corporates 9 models according to geographic region. Mainly statistical models (regression) for the rating process, based on the combination of financial ratios and a qualitative questionnaire. Defaults observed over a period of 8 to 10 years. Small- and medium-sized companies 19 models according to the size of the Company and the geographic region. Mainly statistical models (regression) for the rating process, based on the combination of financial ratios and a qualitative questionnaire, behavioural score. Defaults observed over a period of 8 to 10 years. Public sector entities – Sovereigns 6 models according to type of counterparty. Calibration based on historical data and expert judgments. Losses observed over a period of more than 10 years. Large corporates – Flat-rate Approach 24 models Flat-rate approach according to type of collateral. Calibration based on historical data adjusted by expert judgments. Losses observed over a period of more than 10 years. Large corporates – Discount Approach 16 models Discount approach according to type of recoverable collateral. Statistical calibration based on historical market data adjusted by expert judgments. Losses observed over a period of more than 10 years. Loss Given Default (LGD) Small- and medium-sized companies 15 models Flat-rate approach according to type of collateral or unsecured. Statistical calibration based on historical data adjusted by expert judgments. Losses observed over a period of more than 10 years. Project financing 9 models Flat-rate approach according to project type. Statistical calibration based on historical data adjusted by expert judgments. Losses observed over a period of more than 10 years. Financial institutions 5 models Flat-rate approach according to type of counterparty: banks, insurance, funds, etc. and the nature of the collateral. Statistical calibration based on historical data adjusted by expert judgments. Losses observed over a period of more than 10 years. Other specific portfolios 6 models: factoring, leasing with option to purchase and other specific cases. Statistical calibration based on historical data adjusted by expert judgments. Losses observed over a period of more than 10 years. Credit Conversion Factor (CCF) Large corporates 5 models: term loans with drawing period, revolving credits, Czech Corporates. Models calibrated by segment. Defaults observed over a period of more than 10 years. Expected Loss (EL) Real estate transactions 2 models by slotting. Statistical model based on expert judgments and a qualitative questionnaire. Low default portfolio. MONITORING THE PERFORMANCE OF INTERNAL MODELS The performance level of the entire wholesale client credit system is measured by regular backtests that compare PD, LGD and CCF estimates with actual results by portfolio, thus making it possible to measure the prudence of the risk parameters used by the IRB approach. “Governance of the modelling of credit risk”). These results justify the implementation of remedial plans if the system is deemed to be insufficiently prudent. The results of backtests and remedial plans are presented to the Expert Committee for discussion and approval (see section The results presented above cover the entire Group portfolios. Backtests compare the estimated probability of default (arithmetic mean weighted by debtors) with the observed results (the historical annual default rate). The historical default rate was calculated on the basis of performing exposure over the period from 2007 to 2019.