Predictive analytics: Changing the way insurers think about their homeowners risks

By Robert Curry September 27, 2012

There was a time when homeowners underwriting and rating basically followed the old adage "a house is a house is a house."

But over the past ten years or so, the homeowners market has grown increasingly concentrated, leading insurers to fight for market share to create growth. In today’s market, thinking a house is a house is a house would be disastrous to an insurer’s bottom line. To compete, insurers have been trying to get a better understanding of how closely their premiums match their risks. And they’re looking for better ways to predict future losses so they can identify their best risks and price all the risks in their book of business more accurately.

The companies lagging behind may become vulnerable to adverse selection — good risks lured away to other insurers offering lower premiums and bad risks not paying the proper premiums to cover losses.

As the pressure to remain competitive mounts, companies continue to look for different variables that might factor into the potential for loss. Changes in homeowners underwriting and rating continue to accelerate.

The use of advanced analytics is leading the way in increasing predictability and developing new underwriting and rating approaches for homeowners. In the ongoing search for better ways to develop equitable rates that accurately reflect the experience of a group of insureds, more and more companies are looking to create their own predictive models, work with consultants to do so, or purchase models from vendors.

One way companies are using advanced analytics in homeowners is to develop rating plans that rate separately for the individual perils covered by the homeowners policy, as opposed to the traditional approach of rating all perils as a package.

Why are by-peril rating factors more accurate? The relative importance of perils can vary for a number of reasons. For example, hail can be big in Nebraska or hail alley but not so much in other places. Fire can be more important in areas without adequate public protection. By-peril rating factors allow for a more explicit recognition of the effects of perils varying by location, and they react more dynamically to changing peril contributions over time.

By looking at advanced data sets from multiple sources, carriers can develop models that consider different types of perils as they relate to homeowners coverage. For example, our environmental module in ISO Risk Analyzer Homeowners provides loss costs for nine separate perils at the level of census block group by Public Protection Classification for a base policy. The by-peril rating factors module looks at three key homeowners rating factors — amount of insurance, deductible, and age of construction — for the same nine perils.

The combination of the two modules offers increased flexibility for insurers that want to examine the effects of perils individually and perform their own by-peril analysis. Insurers can use the output of ISO Risk Analyzer Homeowners as stand-alone analytics to help classify, segment, and price their homeowners risks. Or they can use the model's detailed output for each peril and major component of risk, such as weather, to enhance their current analytic systems and better meet their objectives.

ISO is now developing two additional modules for ISO Risk Analyzer Homeowners — a building characteristics module and an occupant module. Together, the four modules will offer insurers a complete set of policy tools for homeowners. Carriers can use the modules to jump start their initiatives for improving pricing accuracy. The information will let an insurer go from “a house is a house is a house” to “this house.”

* * *

In addition to predictive modeling tools for homeowners, the ISO Risk Analyzer suite also offers tools for personal auto and commercial auto. For an in-depth look at the entire suite of predictive models, register for our in-person workshop taking place October 10, 2012, at ISO’s headquarters in Jersey City, N.J.