(1) What assumptions are made using the traditional "excess wind procedure"=
=20
of property ratemaking? How do these assumptions measure up to current=20
conditions?
(2) When simulation models were introduced, what challenges were introduced=
?
(3) Most member of the financial community accepted modeling technology wit=
h=20
relatively little controversy as it evolved over the years. What factors=20
explain this phenomenon?
(4) What are benefits to consumers of achieving better estimates of loss=20
exposure unique to the individual property owner?
(5) What could be involved in the regulatory process of model validation?
(6) Many have argued that models can be externally validated without the=20
need for full disclosure of their inner workings. What tests might be=20
involved in this process?
(7) What safeguards have been suggested to reduce the risk of manipulation=20
of models?
Answers:
(1) Several assumptions are made regarding the 20-30 year period used in th=
e=20
excess calculation:
a. Catastrophic activity was normal - frequency, location, and intensity of=
=20
hurricanes vary over time. Thus, a short experience period would be suspect.
b. Population demographics were stable - coastal areas, particularly in=20
Florida, have been growing at a relatively faster pace than inland.
c. Insured losses by peril were stable - level of fire protection has=20
increased since 1970.
d. Changes in coverage or construction practices did not affect the ratio o=
f=20
wind to non-wind losses - property insurance coverages and building=20
practices have changed significantly. Changes in construction techniques=20
have been dramatic, but the degree of wind resistance is debatable.
(2)
a. Raw insurance data for modeling was often exposure (amount of coverage)=20
which, unlike premiums or losses, was not reported in financial statements=20
or other externally verifiable sources.
b. Raw data was often sent to the modeling company in policy level detail=20
and processed in the modeler's proprietary computer program, making it=20
difficult to follow calculations in the manner traditional methods had=20
allowed.
c. Seismic or meteorological simulations and damage functions at the core o=
f=20
models are extremely complex and difficult to follow without extensive=20
technical expertise.
(3)
a. The modeling process represented a clear technological improvement over=20
available alternatives.
b. Modeling was not wholly foreign to insurers (e.g. similar techniques had=
=20
been used for years to perform economic forecasts used in investment=20
decisions).
c. Insurers had the benefit of a competitive environment for modeling=20
services.
d. Insurers lacked the type of safety net available to individual consumers=
=20
in the form of government-sponsored programs such as "Windpools," leaving=20
them little choice but to accept the use of techniques favored by their=20
reinsurers and investors or face a loss of reinsurance or equity capital.
(4)
a. Comprehensibility of Prices - this will help consumers better understand=
=20
and control insurance losses.
b. Rational Behavior - when cost reflects the economically correct long-ter=
m=20
price, consumers will take that cost into account and act accordingly.
c. Fair Pricing - subsidies will be reduced, and consumers will be rewarded=
=20
for loss mitigation.
d. Reduced Information Risk - risk of uncertainty would be reduced, which=20
will lead to lower prices and/or greater availability.
e. Stable Pricing - models are less susceptible to variation since they use=
=20
(i) long-term seismic or weather data and (ii) all available information on=
=20
the risk to develop loss estimates.
(5)
a. Developing familiarity with the technical background of models.
b. Defining a clear process and specifying policies related to the use of=20
models in ratemaking.
c. Making reasonable efforts, within the constraints of state open records=20
or "sunshine" laws, to safeguard legitimate trade secret information.
(6)
a. Comparison of predicted wind fields in a series of simulated storms=20
versus those actually observed in recent events.
b. Comparison of statistics such as (i) mean minimum barometric pressure=20
(ii) mean wind speed, or (iii) the number of storms of certain categories=20
for a large number of simulated events versus actual historical averages.
c. Comparison of relative damage estimates by type of structure to actual=20
observed damage in recent storms.
d. Comparison of predicted losses for individual events to actual insurer=20
losses.
(7)
a. Require a legal affidavit, attesting that the user has not manipulated=20
assumptions.
b. Require a formal opinion from the modeler on the proper execution of the=
=20
model when run by the insurer.
c. Modelers could provide regulators with rate ranges that reflect (i)=20
geographic (ii) building structure, and (iii) deductible options.