- Business cycle fluctuations and macro economic information have been given second order importance
- Future house prices should be a key driver in a PD model
- Falling prices and unemployment have dramatic impact
- Traditional credit risk models are good at comparing clients, but not so good at determining the likelihood that either or both will defaults
- Including aggregate data in individual regression leafs to biased and inconsistent estimates of the impact of the macro factors
- Unemployment will not touch all individuals and you need a way to spread the pain
- Just sticking macro factors in PD models is the wrong approach to modelling and forecasting their likely future impact
- Best way to account for external macro factors is to model the future path of aggregate default statistics by tying them to macroeconomic and portfolio level information of a matching level off aggregation. Then benchmark to forecast aggregates.
- A model that assumes constant elasticities of slopes across the business cycle is likely to incorrectly determine the PD for certain groups under changing economic conditions
- Assuming participants manage their operations competently, management action should be incorporated into the models used to conduct stress testing
- Strong incentive to outsource stress testing activities of a lender and that incentive is negatively associated with the size of the institution
- Source of errors:
- Naïve purely statistical or historical view of what constitutes stress
- Non economist view with taking into account the interrelationship between variables
- Simply replicate past events. Past superlatives like the oil crash of 1973 are not realistic anymore. he external shock needed to generate theses types of outcomes today will have to be of comet strike severity
- Greatest current threat (2007) for mortgages is the outbreak of inflation.
