Predictive modeling describes how insurance companies develop rates that match individual risk traits to expected losses. Companies utilizing good predictive models have a distinct advantage over those who don't. For consumers, expect to be asked more questions about homes, cars, and other aspects of your lives; businesses have unique rating characteristics as well. We're headed to a world where rates are unique to every consumer.
Predictive modeling is the insurance industry's take-away from the data mining work that is occurring throughout a broad cross-section of businesses today. Many industries benefit from data mining: bar coded discount tags from the grocery store, pharmacy, sporting goods store, and other places we shop are providing companies we use more information about our buying habits. Data mining yields benefits to many industries.
In insurance, the large volume of data today allows actuaries (insurance statisticians) to seek and test new factors for correlation to insurance losses, then weigh these factors in rating algorithms. Insurance companies have always shared some industry data on losses, though larger companies collect and analyze their own data, and now in many new ways. Mining data for factors that contribute to predictability makes sense.
As an example, "account credits" are a new rating factor here in Massachusetts home and auto insurance, popular since the fix-and-establish system with rates by the Insurance Commissioner changed to a market based system. These account credits provide significant savings off both auto insurance and home insurance. Why? Is it because consumers who buy their home and auto insurance with the same carrier tend to remain customers longer? Or because there are cost savings with billing or other account maintenance items? Or is one indicative of a decent credit score? "Yes" to all of the above. Because it means "yes" to all of these factors, it becomes even more powerful. Mathematically, it's akin to the magic of compound interest. Credits on credits on credits make for deep discounting.
There are other credits (and charges) that are less obvious but contribute to a final rate. One predictive factor that is NOT permitted for rating Massachusetts auto insurance is credit history. But the owners of homes generally have good credit. The bank already knew that when they lent the money. Companies and their actuaries have discovered other indicators of good credit: including buying higher liability limits, shopping and buying ahead of an existing policy's expiration, and paying in full. These proxies for credit are the industry's way of developing a leaner rate while still following the law. Other rating factors contribute too: if you're a member of a motor club, if you've graduated from college, how long you've lived at your current address, if you've had (even) not-at-fault accidents, and other questions that you'd think have nothing to do with auto insurance rates. Combine these with traditional metrics such as annual mileage, years of driving experience, moving violations and at-fault accidents, and auto insurance rates become unique to every driver.
For homeowners, some of the unexpected rating factors include gun ownership, dog ownership (broken down further to breed), existence of home generators, and credit history (which is permitted in home insurance). Combine those with expected rating factors such as age of home, existence of alarms, and proximity to the ocean, and models become more predictive with every piece of relevant information. Businesses are increasingly subject to additional questions for rate development: what insurance professionals know as 'supplemental applications'. These collect more details to find the most competitive and accurate rate. Companies using these have a distinct pricing and product advantage those who don't, and they know it.
The old methods of calculating an insurance rate with a pencil and a calculator are out; multi-variant rating is in.
At Gordon Insurance, we use a library of checklists to aggregate the rating factors from our many companies to ensure you're getting the best rate possible. Many questions we ask will result in: "why do they need to know that?" We don't always know. But some actuary somewhere found a correlation to expected claims. Privacy laws and our national tradition of privacy will push back against this trend. But the trend toward pricing based on multi-variant models is accelerating, as a combination of proprietary calculations and quickly secured public information makes more data available. So to the question, "does it affect me?": Absolutely. By working with Gordon Insurance, you'll find the best fit with the company that likes your profile better than anyone else.
If you have any questions about how insurance works and how it is determined, feel free to contact us. We love answering insurance questions and helping customers find the best possible rates to meet their insurance needs.