BACKOFFICE
Strategies and technologies to boost
behind-the scenes efficiencies
Too Much Information?
In theIr quest to accumulate data to feed Into theIr
predictive models, underwriters are bumping up against regulatory strictures and privacy concerns. Indeed, one of the most
useful predictive tools for underwriting—credit scores—is the
subject of a great deal of legislative attention. While insurers say
credit-related variables such as payment history are highly predictive indicators of potential future risk, states such as california, massachusetts, hawaii and maryland have either fully or
partially banned the use of credit scores. other states, including
michigan, have legislation pending that would restrict its use.
at hearings held in may by the house subcommittee on financial Institutions and consumer credit, representatives of the
insurance industry defended the use of credit scoring in underwriting. dave snyder, VP and associate general counsel for the
american Insurance association (aIa), said the use of credit-based insurance scores by property/casualty insurers was necessary to accurately assess and price risk. “In the midst of the
financial turmoil and its related chaos, the u.s. property/casualty insurance sector is stable, secure and strong,” snyder says.
“there are good reasons for this. Insurance scoring has played a
major role in creating this positive market for all concerned.”
Proponents of credit scoring note the majority of states
have taken some legislative or regulatory action to protect consumers and govern how credit score information is used. ann
Weber, VP, state government relations for the Property casualty Insurers association of america, notes that insurers are legally prohibited from considering a person’s income, race, age,
address, marital status or nationality when calculating an insurance score.
“While it may be a surprise to some to hear that insurers
use credit information to help set insurance rates, it’s a long-established practice that insurers have used this for the past
two decades to offer lower premiums to most consumers,”
Weber says. “every serious and reputable study on the issue,
including a 2007 study released by the federal trade commission, has borne out the correlation between insurance
scores and risk.”
nonetheless, with the legality of credit scoring under attack,
insurers may want to consider adding additional variables that
demonstrate the characteristics of personal responsibility,
which may have been deemed insignificant in light of the use
of credit scores. —Bill Kenealy
After the model becomes available to
underwriters, management should
gauge acceptance periodically because
underwriters may simply stop using a
model they find difficult or ineffective.
“If it’s contrary to what the underwriter
understands, then they’re not going to
trust what’s coming out of the model,”
Diers says, adding that often just a few
tweaks can make it more acceptable.
Once underwriters join in the predictive analytics process, they become valuable in shaping future releases of the tool,
says Oltman. “It allows us to go back and
forth with underwriting to get something
that’s statistically relevant, but at the same
time keep it manageable at the underwriting desk level,” he says.
enough companies are “asking the right
questions and doing the work that I would
call it reasonably well established,” he says.
Applying predictive models to other
lines such as personal umbrella liability
strikes Trasancos as a natural extension
of the work in auto and homeowner’s.
“You’re dealing with fundamentally the
same risk because it’s still the same
household,” he says. “If the stability, be-
havioral and structural characteristics of
the household are X for auto, they
should be X for home and they should
be X for umbrella. They should be the
same.”
Though these are the early days for
predictive analytics in life and commer-
cial lines, carriers are accepting of the
eventual need for the technology, lest
they lag behind competitors, says Clark
Troy, senior analyst for the Boston-based
Aite Group LLC.
“There’s a lot of theoretical accep-
tance of predictive modeling,” agrees
Diers, “but there’s a hesitancy to be
first.” That makes it important to figure
out what groups within the company
are ready to use predictive modeling,
she says.
DATA BARRIER
Besides acceptance, having enough us-
able data can present the biggest barrier
to predictive analytics, observers agree.
“What’s holding it back is the lack of
data and accessibility to that data, says
Guidewire’s McCully. “Without that,
you don’t have much to go on.”
Many carriers still rely on paper ap-
plications that yield little in the way of
structured data for models, says Aite’s
Troy. Using predictive analytics is not
impossible without electronic applica-
tions but nearly so, observers agree.
Even with automated systems gathering information, departments sometimes overestimate the quality of their
data, notes Diers. “Do we have the data
we think we have?” she cautions.
“There are carriers out there that I
would call data rich, but information
poor,” says Toth. “They’ve got a large
amount of data that’s available, but
they just haven’t figured out how to
harness it in a usable and serviceable
way.”
Besides internal data, carriers can
draw upon a wealth of information
from sources outside the company, in-
surance executives note. That third-par-
ty data includes details of consumers’
past claims from underwriting ex-
