While predictive modeling serves primarily as a way to contain medical costs on the claims side of the ledger and improve overall the quality of care, predictive modeling also serves an important function on the underwriting side of the equation.
Robert Conger, who co-authored with Russell Greig a Towers Perrin report on predictive modeling in the workers' compensation area, says the ultimate test for the underwriting value of predictive modeling is in establishing so-called "right" pricing to smooth out the peaks and valleys of the underwriting cycle.
"Some may ask whether workers' compensation pricing can be improved much, given that it has long been thought to have the most accurate pricing structure of any line of business and has benefited from a heavy reliance on insurance data," Conger says.
But such data and classifications have their limitations.
For example, a coding such as Auto Service or Repair Centers and Drivers includes major repair shops as well as car detail shops. Such a classification could produce an accurate price for any employer with the right mix of all these elements, but individual employers will most likely vary from the average.
Predictive modeling, according to Conger, can verify or refute the kind of schedule rating plan adjustment needed to account for such variations. "It does not eliminate the need for underwriter judgment, but creates a more analytically based framework in which judgments can be applied and evaluated," he says.
Conger warns that predictive modeling in the underwriting sphere requires significant upfront investment. "The whole process is very data-intensive and requires linking policy, claims and external data at the individual risk level," he says.
So-called data-scrubbing, or verifying the completeness and accuracy of all the databases involved, also remains a challenge. "It also takes time to properly analyze the results, depending on the number of factors being reviewed," Conger says.
Most carriers remain reluctant to disclose the extent of their use of predictive modeling for fear of losing competitive advantage, but it appears to be a practice gaining acceptance.
Conger says that if enough competitors started using predictive modeling in a line of business, the effects on the marketplace could be profound. "As companies identify better risks that can be written at lower rates, the best risks will migrate to those insurers recognizing them," Conger says.
And on the flip side, worse-than-average risks uprated by sophisticated insurers may flee to carriers slower to innovate, he added.
August 1, 2008
Copyright 2008© LRP Publications