Regulatory requirements on credit risk models have increased substantially. Banks need credit models for computing regulatory minimum capital under the Basel II / III accords. A reform of loan loss provisioning (IFRS 9) mandates the use of credit models for impairment models. Finally, stress testing has become an integral part of risk management, both internally and externally. Banks are required to run internal stress tests regularly and, in addition, have to evaluate the impact of scenarios provided by regulatory bodies like the Fed and EBA.
Fulfilling all these requirements poses a challenge to banks, especially to small and medium-sized banks that cannot afford hiring large modeling teams. It is intensified by the fact that Basel, IFRS and stress tests require credit risk estimates of different nature which often leads to confusion on how risk estimates generated for one application can be re-used for other purposes, e.g. whether risk parameters estimated by Basel can be applied for IFRS 9 and vice versa.
This workshop starts with a short overview of the current regulatory environment. The focus in this part is on characterizing risk parameters estimates that are required for each risk application (Basel, IFRS 9, stress tests), what they have in common and in what respect they differ. The heart of the workshop is providing a framework that allows a bank to compute PD, LGD and EAD consistently in an approach based on combining loan-level with macroeconomic modeling. This framework is generic and applicable also in situation where loan-level data histories are short.
After its formal introduction, the framework will be illustrated with numerous examples. Some stylized but realistic loan models are provided as starting
point together with a time horizon that was used for estimating these models. This could be the models a bank has estimated on its available historic data. To compute credit risk parameters for Basel, IFRS and stress tests, publicly available data from the IMF, OECD, and the US Fed (data for the CCAR 2020 stress test) is used. It is shown in detail how to fulfill the requirement of Basel and IFRS 9 and how to compute the impact of the Fed’s CCAR 2020 scenarios on bank capital and loan loss provisions.
A huge challenge in the current environment is building stress test models for inflation and energy price shocks where in the past 10-20 years, it is very difficult to observe any sensitivity of default rates with respect to inflation for many portfolios due the benign inflationary environment. Ideas to tackle this problem by proxy models are discussed and illustrated with real-world data.
