Predictive Modeling and Your Insurance Premiums
Predictive modeling is a analytic process that uses data mined from a variety of sources to create a statistical model of future behavior. Increasingly, it’s being used in the insurance industry to decide what to charge you for almost every kind of coverage from car insurance to life insurance, homeowners insurance and workers compensation. Its use is reported to be up by 10% across all lines of insurance except commercial property/BOP where it remains relatively flat. Basing insurance premiums on predictive modeling is about as far as you can get from the old days when companies relied on the collective judgment of underwriters and actuaries supported by a smattering of data to decide if you were a good a risk and what to charge to you. The transition from human judgment to algorithms has a number of obvious advantages for insurance companies, but what are the implications for you, the consumer?
According to a 2012 survey by Towers Watson, of the 69 US and Canadian property and casualty insurers queried, 85% said they are using or plan to use predictive modeling. Most cited improved top- and bottom-line results as the reason. The process improved rate accuracy, had a positive impact on loss ratio and helped retain policyholders at renewal time. All good things for the insurance industry.
Consumers can hope to derive some benefits from this reliance of technology, too. By streamlining the underwriting and claims processes, consumers gain more responsive service. Predictive modeling could also determine what types of claims drive up costs and what types of claims should be investigated for fraud. Anything that reduces insurance fraud, which costs all of us billions of dollars each year, is a good thing.
On the downside, experts caution that an overreliance on predictive modeling without a solid application of human judgment and business experience could result in too many decisions being made based on “data noise.” As one critic pointed out, predictive models aren’t reality. In her article of the same name, senior director of knowledge resources and ethics counsel at American Institute for CPCU Donna Popow writes that “some aspects of predictive modeling can be concerning.” She points to the widespread use of credit scores to determine auto insurance premiums. Popow rightly points out that a person’s poor credit rating could be the result of the poor economy and have nothing to do with his driving record, yet predictive models would place that person in the pool of people most likely to file a claim, resulting in higher premiums.
The validity of a predict model will always depend on the quality and quantity of the data parsed. As the use of predictive modeling to set insurance rates becomes more prevalent, watch for consumer advocacy groups to demand more transparency about the factors used to justify pricing.
