Factors of individual accounts can include payment history,
loss control efforts, and how many and which insurance lines
are being purchased—such as grouping workers’ compensation
with commercial auto. Just as the age of drivers is a factor in
personal-auto pricing, the age of an employee is a factor in pricing workers’ comp coverage.
Predictive modeling can handle more data, past and present, and can enhance or go beyond the experience modifier by
differentiating what previously was “bucketed together” in the
experience mod, Stoll said.
Predictive modeling can be used also to enhance segmentation in the NCCI’s classification plan. The NCCI classification
codes represent risk potential by occupational type. Despite having about 580 classifications that are constantly updated, the
classification plan remains limited.
“Historically, pricing was done on a univariate class basis,
and then adjusted for individual risks,” said Stoll. “Predictive
modeling … produces much more accurate pricing at the indi-
vidual risk level.”
Predictive modeling is also more nimble to new informa-
tion like policyholder changes to improve price risks, unlike the
experience mod, which is based on loss data from the past two
to three years, he said. “Adding new employees is more predic-
tive than the last two years with fewer employees because more
employees mean more risk exposure.”
Workers’ comp carriers are split about their confidence re-
garding the appropriateness of payroll as the standard industry
exposure base, according to the Towers Watson survey, with
58 percent of respondents saying they are not or are somewhat
confident and the remaining 42 percent saying they are highly
Changes in operations and other individual risk factors can
differentiate individual risks within a class, said Stoll, and are
more responsive than the experience modification factor. He
offered that income levels not only are predictive of the amount
of wage replacement benefits, but also the more generously
employees are compensated, the more likely is high employee
morale, which can positively affect claim outcomes and speed
of return to work. Whether the employer provides health insurance also can affect cost shifting.
One way to apply the predictive modeling results is to
build the difference between the indicated and standard premi-
um into a schedule rating plan credit/debit, making adjustments
to specific schedule factors to arrive at the target price, Stoll
said. The more pragmatic approach frequently used by carri-
ers, he said, is starting with what the predictive model says the
price should be (or the price agreed on with the agent) and then
comparing it with the rating plan premium. “The ratio of the
two tells you how much debit or credit you want to give the risk
and then charge back into the targeted premium.”
Larger insurance companies, however, were initially skepti-
cal about the ability of predictive modeling to go beyond the
NCCI classification plan, and worked on other commercial lines
first, Stoll said. Now the national carriers have recognized the
benefit of predictive modeling in enhancing price accuracy.
For niche carriers, for example, accounting for the nuances
in similar businesses can make a big bottom-line difference.
“They know those niches better than competitors,” he said,
“and if they capture their information electronically, [they]
can build their industry expertise directly into their pricing
through predictive modeling.”
Reducing Losses from the Beginning
Nothing makes more sense than preventing losses before they
occur. Even with the best safety and risk management efforts,
a work-related incident occurs and the claims process begins.
Apply predictive modeling in this area and cost savings eventually can cause a domino effect all the way back to underwriting
and premium auditing.
The recent uptick in frequency of lost-time workers’ compensation claims might be explained partially by medical-only
claims unnecessarily becoming lost-time claims. In the past two
decades, despite countless system improvements, medical costs
have increased cumulatively from the 1991 to 2010 accident
years by an estimated 238 percent, according to NCCI. Workers’