Insurer use of predictive models growing: Towers Watson

Property and casualty insurers have increased their use of predictive modeling across nearly every line of business over the past year, suggests a new survey from Towers Watson.

Despite that, most of those businesses still lack a comprehensive, company-wide approach for using predictive modeling for all “core functions,” the global professional services firm said.

Overall, insurers are often not applying their data-driven analytics uniformly throughout the business, while overall usage fluctuates significantly by line of business and company size, according to the firm’s fifth annual survey, which was based on a survey of 59 U.S. P&C insurance executives.

“The survey demonstrates a real willingness by carriers to embrace predictive modeling programs, but in many instances, the actual investment in, or execution to establish these frameworks, has been incomplete or targeted to specific business lines or operational needs,” Brian Stoll, director of the P&C practice at Towers Watson commented in a press release.

“It could be due to the financial crisis that insurers put many investments on hold and renewed focus on the expense side of the balance sheet, or maybe a narrow vision of predictive modeling’s applications and potential,” he noted.

“Perhaps data, people or cultural challenges are a factor, or some are only applying data-driven analytics when an area is underperforming. Whatever the reasons, a compelling case can be made that well-executed predictive modeling provides better pricing guidance to underwriters,” he added.

More personal lines carriers – 80% of personal auto and 62% homeowners – are using modeling techniques than commercial lines carriers – 33% general liability and 32% commercial property, according to Towers Watson’s research.

Forty-five percent of respondents that write personal lines auto now have formal usage-based insurance plans (a 10-percentage-point increase from last year), the survey also noted.

In addition, 12% of commercial auto carriers have either launched UBI products or plan to do so in the next year, an increase over last year, when none of the respondents had launched commercial auto UBI programs and only 4% of commercial auto carriers had plans to start one, Towers Watson said.

Specialty lines carriers expressed an interest in applying predictive modeling to their business with 45% saying they plan to do so in the future, according to the study.

Among those using predictive modeling, nearly all said that it has led to favourable bottom line results with positive impacts on rate accuracy (85% personal, 96% commercial), profitability (80% personal, 78% commercial) and loss ratio improvement (80% personal, 74% commercial).

Large carriers were more likely to use models over small carriers, which noted that they tend to differentiate themselves through service and claims matters rather than by using models.

Companies did report seeing less significant benefits to their top-line results, such as modeling’s positive impact on the expansion of their underwriting appetite (45% personal, 48% commercial) and on market share (35% personal, 39% commercial).

“Respondents are really tailoring their programs to focus on specific market realities,” Klayton Southwood, senior consultant with Towers Watson noted in its release.

“Personal lines carriers operate in a highly competitive, mature market, so it’s not surprising a high percentage have adopted many aspects of modeling. On the other hand, commercial lines carriers face less intense pricing pressure in some segments, in part due to heterogeneous risks and the heightened reliance on individual risk underwriting expertise, particularly in large risk/specialty lines.”

In terms of communication, carriers could improve how they communicate with agents about modeling, Towers Watson said. Less than half provide their agents with relevant insights concerning modeling, and companies often don’t involve agents in the model-building process (7%), explain pricing models (17%) and communicate model changes in advance (17%), according to the firm.

“Companies using captive agents (50%) believe their agents have more understanding of modeling’s impact on risk appetite, compared to participants using independent agents (30%); respondents using captive agents (50%) also say their agents approve of data-driven analytics more than participants using independent ones (33%),” it noted.