A Multi-view Model Framework for Stress Testing C&I Portfolios

Jimmy Yang and Kenneth Chen

This chapter will focus on a multi-view framework for stress testing commercial and industrial (C&I) portfolio. It will first discuss the challenges for stress testing wholesale portfolio, before introducing loss forecasting methods and a multi-view framework to properly assess and hedge model uncertainties. Besides probability of default (PD), the most important component of loss estimate in an expected loss (EL) framework, we will also explore several common methodologies to stress loss given default (LGD) and exposure at default (EAD). Finally, the chapter will conclude with a discussion of how to combine all these components to finalise recommended results from both a quantitative and qualitative perspective.

CHALLENGES FOR WHOLESALE STRESS TESTING

Estimating loss for the loan book under stress is problematic. On the one hand, the hypothetical scenario has not typically existed in history. Statistical models based on the historical relationship between loss and macroeconomic variables may not have fully reflected “accurate” sensitivity under the new regime. On the other hand, validation of models is also not straightforward. Unlike models used for baseline forecasting, where