Stochastic modelling of the loss given default (LGD) for non-defaulted assets

In the Basel framework of credit risk estimation, banks seek to develop precise and stable internal models to limit their capital charge. Following the recent changes in terms of regulatory requirements (Basel regulation, definition of the Downturn…), it is prudent to think about innovative methods to estimate the credit risk parameters with the constrains of models’ stability, robustness, and economic cycles sensitivity.

This paper introduces a different recovery forecasting methodology for LGD (loss given default) parameter and explores stochastic forecasting with details of how to calibrate the model. Three classical methods of recovery forecasting based on Chain-Ladder are presented for comparison and to contest and the stochastic methodology. Finally, a regulatory calibration of the LGD for non-defaulted assets is proposed to include Downturn effects and margins of prudence.