|
Extreme Value Models Task Force
Bob Guth, Leader Goals
In today's business environment, extreme events are
commanding more attention. These events have very low frequencies
(e.g., once-a-century) but extraordinarily high costs. Research has
shown that the Normal Distribution, which is pervasively used in
actuarial work, underestimates the frequency of these events,
sometimes significantly. Consequently, actuaries need to have more
expertise in this aspect of risk management.
The Extreme Value Models subgroup of the Risk
Management Task Force has two goals:
- Increase the actuarial profession's awareness of these
extreme risks and of the pitfalls of using simplistic
methods to assess these risks.
- Provide education and tools needed to quantify, manage, and
price the risks associated with extreme-valued outcomes.
In order to accomplish
these goals, the subgroup will:
- Identify, develop, and publicize resources for actuaries
concerning the quantification, management, and pricing of
relevant risks.
- Investigate the statistical distributions describing the
frequency of extreme-valued outcomes, including
non-parametric and non-symmetric distributions.
- Monitor and/or sponsor research in this area.
- Provide continuing education opportunities about
applications of extreme-valued risk measurement and
management.
Recent Activities
The Extreme Value Models subgroup helped organize a session on Extreme
Value Theory at the 2004 Enterprise Risk Management Symposium.
In 2003 we sponsored a contest encouraging
actuaries to consider possible, but perceived-to-be unlikely,
events. Read the  winning entry,
a thought provoking essay on negative interest rates by Ken Faig!
Sample Calculations
Steve Craighead has developed an Excel workbook that can be used to
become familiar with calculations stemming from Extreme Value
Theory. To begin applying these techniques to your applications
follow this  link.
For instructions on use of the workbook follow this
link.
Jonathan T Wang has prepared a discussion on using
extreme value theory to model operational risk.
 Click here for details.
|
|
|