FIGURE 2: Distribution of Losses for 100-Year
Characteristic Event (Hypothetical Company)
estimated loss in millions of dollars
SOuRCE: Au THOR
| New York | Rhode Island | Massachusetts |
insurance companies. While the catastrophe models may simulate a $4 billion
loss for this hypothetical company, it’s
probably in the extreme tail of the ex-ceedence-probability curve and not used
for decision making.
The probability of the $4 billion loss
indeed may be low, but its occurrence
seems much less remote when caused
by a Category 3 storm with a certain
track. Given this perspective, the effects
of such storms seem more plausible
than their low probabilities originally
make them appear.
footprints has been employed to estimate the aggregation of insured values
exposed to a terrorist attack. A ground
motion footprint for earthquakes and a
wind footprint for hurricanes would be
more appropriate than a circle. This can
be illustrated using a hypothetical Northeastern company.
The Northeast of the United States is
vulnerable to hurricanes. But at less than
10 percent, the annual frequency of landfall events is quite low—there have been
only nine hurricanes that made landfall
in the region since 1900. While the data
on these storms are very sparse, scientists believe the most intense event was
the Great New England Hurricane of
1938. It’s a matter of scientific disagreement whether this storm was a strong
Category 2 or a Category 3 hurricane at
landfall, but it’s known that the storm
was fast moving and quite large in terms
of the diameter of its eye (which is typical for Northeast hurricanes).
Using available scientific informa-
tion, it’s possible to create a windfield
footprint for this type of event. While sci-
entists don’t know the exact probability
of the 1938 storm occurring again, it’s
reasonable to assume it’s around 1 per-
cent and that this event is characteristic
of a 100-year event in the Northeast. The
graphic in Figure 1 shows this type of
event with a track typical for the region.
Statistics can Fool you
Without a doubt, trying to estimate loss
probabilities is a valuable exercise, but
probable maximum losses don’t illuminate
the full profile of risk in a given area. An
overemphasis on the probabilities obscures
specific scenario losses and the thoughtful
mitigation of risk from Black Swans.
In an article that appeared in Edge
Magazine in 2008, Taleb cautioned, “
[Sta-tistics] is the ‘logic of science’; it is the
instrument of risk-taking... you can’t be
a modern intellectual and not think prob-abilistically—but... let’s not be suckers.”
When the data and scientific knowledge are limited, statistics can fool you.
Shifting some resources from catastrophe
models to other methods for assessing the
risk will lead to enhanced understanding
of the loss potential and more robust risk
management strategies over time.
KAREN CLARK is chief executive
officer of the risk management
consulting company karen Clark & Co.
She is the founder of AiR, a Boston-based catastrophe modeling firm.
AIRAC (All-Industry Research Advisory
Council), “Catastrophe Losses: How the
Insurance System Would Handle Two $7
Billion Hurricanes,” Oak Brook, Ill., 1986.
Taleb, Nassim, “The Fourth Quadrant:
A Map of the Limits of Statistics,”
Edge Magazine, Sept. 15, 2008.