Stress testing is an ongoing concern for many banks across the planet and some risk teams are even struggling to find a starting point for entering into a comprehensive stress test program.
This short article is the outcome of a presentation that was recently delivered as a speech on stress testing in Singapore.
This short article is the outcome of a presentation that was recently delivered as a speech on stress testing in Singapore.
Stress Testing
The presentation below focuses on the types of models which may be employed as part of a comprehensive stress test for the trading book of a bank. Various reports that risk teams may find useful were also exemplified within the framework and the presentation swings in several articles or blogs that have already been published on this site.
You can view this "Prezi Stress Testing Roadmap" by clicking on the image below and for slow internet connections you may need to be patient.
You can view this "Prezi Stress Testing Roadmap" by clicking on the image below and for slow internet connections you may need to be patient.
Stress Testing Presentation | Martin Davies [click image to view]
Papers that feature in this work include:
[] Nested Archimedean Copulas Meet R [ LINK ]
[] A Thesis on Garch Modelling [ LINK ]
[] Designs for Lasso Latin Hyper Cube [ LINK ]
[] Nested Archimedean Copulas Meet R [ LINK ]
[] A Thesis on Garch Modelling [ LINK ]
[] Designs for Lasso Latin Hyper Cube [ LINK ]
The hurdles of stress testing
We have to be real about stress testing in banks today and what I am seeing at present is pretty disappointing across the board. In some cases the activity is a bit futile and in others it is causing more harm than good for the bank overall.
There are three roadblocks that need to be navigated for a 'comprehensive' stress test and they can loosely be stated as:
[1] Stress Testing is focused deeply on the Value at Risk effort and if the statistical model is flawed in a bank, risk teams might be better off correcting this problem rather than trying to stress test it.
[2] Stress testing trades, portfolios and books separately without aggregating the results will not consider the cross correlation effects that exist between these positions. A comprehensive stress test such as the those that are recommended in Basel III, must take risk teams beyond simply adding up exposure values.
[3] The modelling tools in some banks are more often than not unsatisfactory, and we have to accept stress testing really is beyond the capabilities of what Microsoft Excel can deliver.
We can summarize these three hurdles as the VaR engine, Aggregation / Correlation, Modelling and the presentation above aims to address these roadblocks.
Additionally, we are finding that many banks aren't supporting their stakeholders with appropriate stress test reports. Reporting fictitiously large but improbable outcomes from a singly traded position doesn't help managers "risk on" or "risk off" in the bank.
There are 10 major hurdles which need attention in a comprehensive stress test, these were discussed deeply during the speech and have been listed below.
The main presentation above attempts to address these hurdles and roadblocks.
There are three roadblocks that need to be navigated for a 'comprehensive' stress test and they can loosely be stated as:
[1] Stress Testing is focused deeply on the Value at Risk effort and if the statistical model is flawed in a bank, risk teams might be better off correcting this problem rather than trying to stress test it.
[2] Stress testing trades, portfolios and books separately without aggregating the results will not consider the cross correlation effects that exist between these positions. A comprehensive stress test such as the those that are recommended in Basel III, must take risk teams beyond simply adding up exposure values.
[3] The modelling tools in some banks are more often than not unsatisfactory, and we have to accept stress testing really is beyond the capabilities of what Microsoft Excel can deliver.
We can summarize these three hurdles as the VaR engine, Aggregation / Correlation, Modelling and the presentation above aims to address these roadblocks.
Additionally, we are finding that many banks aren't supporting their stakeholders with appropriate stress test reports. Reporting fictitiously large but improbable outcomes from a singly traded position doesn't help managers "risk on" or "risk off" in the bank.
There are 10 major hurdles which need attention in a comprehensive stress test, these were discussed deeply during the speech and have been listed below.
10 Stress Testing Hurdles | Martin Davies [click image to view]
The main presentation above attempts to address these hurdles and roadblocks.
No comments:
Post a Comment