despite improvements over the
past two decades, extreme events
of all types frequently are missed
by the catastrophe models and by
insurance companies that rely heavily
on them to assess and manage risk.
latest model updates, along with recent events, clearly demonstrate the wide uncertainty surrounding these probability
estimates. It’s easy to be lulled into a false sense of security by
scientific research that sounds impressive, but what scientists
don’t know about hurricanes and earthquakes is much more
than what they do know.
Companies typically focus too much on the loss probabilities—particularly the 1-in-100- and 1-in-250-year return period
losses—and not enough on understanding the risk and the types
of scenarios that could cause large losses. Because the models
are based on historical data and scientific research, they will
continue to miss the Black Swans.
Insurance companies would benefit from additional tools to
help them assess possible losses from catastrophic events. To
be considered a Black Swan, an unexpected event must occur in a geographical area in which insurance companies have
exposure concentrations. A large magnitude event in an unpopulated area that doesn’t cause a lot of damage doesn’t produce
any negative consequences. Since exposure concentrations are
necessary for Black Swans, such concentrations should be managed to avoid them.
But how much is too much in terms of concentrations of exposures? Companies historically have developed simple rules of
thumb, such as limiting insured values or market shares by ZIP
code. Monitoring aggregations by ZIP code may be sufficient
for localized events such as hailstorms or small tornadoes, but
major tornado outbreaks, hurricanes, and earthquakes can affect much larger areas with high intensity.
A more scientific method for monitoring exposure concentrations is to identify potential extreme-event losses using
event “footprints.” For each region exposed to catastrophes, representative footprints can be created using the same scientific
information underlying the catastrophe models. These footprints then can be “floated” within the region (superimposed
on a company’s exposures) and the scenario losses calculated.
For terrorism loss estimation, using concentric circles as