 |
|
|
About: News Archive
"Assessing the Risk in Risk Assessments" |
| 16-May-2003 |
| |
Steve Figlewski of NYU Stern School of Business - who is well-known to readers as an IAFE Board Member and the Editor of the Journal of Derivatives - gave a terrific and thought-provoking talk for the IAFE’s April Monthly Meeting. Approximately 85 people attended his presentation titled “Assessing the Risk in Risk Assessments”, which was held on April 23 at Goldman Sachs in New York City and was generously sponsored by PricewaterhouseCoopers. Common procedures for assessing financial risk exposure, such as Value at Risk, involve predicting the lower tail of the probability distribution of asset value or return. Steve pointed out that the statistical estimation error in the calculation is typically ignored. In practice, a "tail event" may represent a truly rare occurrence, or it may simply be a not-so-rare occurrence at a time when the predicted volatility is an underestimate of the true volatility. The problem grows worse the further in the tail one is trying to predict, especially when volatility varies over time.
Steve’s presentation explored model risk through a review of an extensive simulation study. Even in the "plain vanilla" case, with normal distributions and no time variation in parameters, the risk of a large loss can be significantly underestimated due to ordinary sampling error in estimating the volatility. Volatility that varies stochastically over time makes the problem distinctly worse, and fat-tailed return shocks are even more serious. Pros and cons of alternative approaches to tail estimation, such as historical simulation, were also discussed. Open discussion with the audience following the formal presentation raised a variety of interesting issues regarding estimation risk in real-world risk management.
To view Steve's PowerPoint presentation from this event, please click here. |
|
|
|
|
|
