expensive classification, such as blue-collar work, with a less
expensive one, such as clerical work. Roofers could be classified
incorrectly as less expensive carpenters.
Many factors considered in pricing and underwriting, such
as class codes and locations, are being used not only in underwriting and pricing but also at policy inception, Johnson said.
It is logical to assume that insurers are benefiting from premium auditing by applying the results to underwriting and
pricing, but this isn’t easy for all carriers. “It might have to do
with the way they are organized,” she said, because knowledge
sharing is less likely to happen when auditing and underwriting
are kept separate. Underwriters might take it under advisement
as they have to balance many business priorities, including
agents who also get involved in pricing, she explained.
The Price Is Right
Determining the right price for each risk in a workers’ comp
portfolio has always been challenging. Amid fluctuating
markets and underwriting cycles, the long-tail line is sensitive
to ever rising costs and changing legislation and regulation on
a state-by-state basis.
Through the lens of predictive modeling, insurers can view
traditional factors in new ways, add in non-insurance factors,
and determine the degree to which factors should be considered to develop detailed market segmentation opportunities to
underwrite more profitable policyholders. Predictive modeling identifies more marketing and cross-selling opportunities
as well, said Dean.
The process makes use of often buried internal database
information to reveal new factors for pricing. Most carriers (
historically almost all carriers) fail to capture much of the non-rating
information collected for underwriting, Stoll said. “Anything not
needed to rate the policy was only in paper files, and the underwriter knew it and the actuaries didn’t,” he explained.
Predictive modeling is using data based on experienced
underwriters’ conventional wisdom and finding a lot of it
to be true, Wu said. “We are validating the traditional business wisdom about what kind of information can be used for
segmenting.” For example, just as credit scoring for private-passenger auto is indicative of lower risk, workers’ comp
underwriters rightly assume that more efficiently operating
companies should have better workers’ comp programs.
The Evolution of Predictive Modeling
Workers’ comp predictive modeling started after the successes in personal auto,
personal home, and small commercial
insurance product, said Peter Wu, a director
at deloitte Consulting. “there is a natural
evolution of predictive modeling.”
Insurance predictive modeling began
about 20 years ago. before predictive model-
ing, businesses had databases of information,
but it was impossible for a human being to
dig out the facts and connect the relation-
ships, said Curtis Gary dean, a professor of
actuarial science at ball State university.
Progressive was the first u.S. insurer to
embrace predictive modeling, initially in the
form of credit scoring in the late 1980s and
early 1990s, said brian Stoll, a director at
towers Watson. the other large personal
automobile insurers that followed suit im-
mediately began gaining market share and