Title: From ratings to credit losses, in good and bad times: Correlation modeling of credit risk and macroeconomic variables
Authors: Libor Pospisil - Moody's Analytics (United States) [presenting]
Abstract: In order to perform calculations set out in the new international accounting standards, IFRS9 and CECL, financial institutions need models that would project losses on credit portfolios under hypothetical macroeconomic scenarios. While we can relate this kind of calculation to the well-known topic of stress testing, the IFRS9 and CECL applications often have a limitation: financial institutions may only have agency ratings as the credit risk measures of their portfolios. Agency ratings are typically considered Through-the-Cycle (TTC) risk measures, not reflecting the current economic environment. Projecting losses if only rating is known therefore must include two components: accounting for how the current economic environment affects credit risk as of now, leading to Point-in-Time (PIT) view of credit, and quantifying effects of the hypothetical scenarios that describe a possible future path of the economy. We present several time series and correlation models, which allow us to project credit losses when only rating of a portfolio is known. We then relate these models to the theory of TTC and PIT credit measures, as well as to the standard stress testing methods for credit portfolios.