The phone rings. “Hello, John Smith here.”
“Hello, Mr. Smith, this is the insurance company. We regret to inform you that your application for life insurance has been declined.”
“But why? My physician says I’m in perfect health, my body fat percentage is 18 percent, I work out every day, both of my parents lived to be 108, I work as a librarian, and my hobby is knitting. How can you possibly decline me?”
“Well, Mr. Smith, our data mining discovered you keep photos of orchids on your smartphone and our algorithm says orchid viewers have a higher mortality rate. Sorry. Have a nice day.”
Lest anyone think the previous exchange is fanciful or seems a bit paranoid, it is already happening. It is known as using accelerated underwriting using external data and over two dozen U.S. insurers are currently using it. What data is being used to decide your insurance fate? The underwriting decision is not only based on your health, but on the products you buy, which selfies appear on your smartphone, the people you have lunch with and even the magazines you read. A recent article in Best’s Review predicts every insurer will be doing this type of accelerated underwriting within a couple of years.1
The credit industry has been using these external data algorithms for quite some time. It’s the reason you receive fraud alert messages. However, it doesn’t always work properly. After having a purchase blocked by one of my credit card issuers due to possible fraudulent activity, I had to spend time on the phone explaining that just because a person from St. Louis was now buying gas in Des Moines didn’t mean the card was stolen.
Another problem with an over-reliance on algorithms is it can override the human factor. In recent years I have had to jump through numerous hoops to open bank accounts because the credit bureaus all have my post office box sometimes listed as my residential address, and this flags my account as suspicious and an automatic turndown from their computer model. I can usually get this corrected by getting in touch with a real person that eventually passes me along to their supervisor’s supervisor, whom apparently is the only one with the power to override the software.
One side effect of this machine data gathering for insurers will be less of a need for underwriters. After all, if your computers can do all of the underwriting by mining data from a smartphone, you don’t need a human underwriter to make the call.
Both regulators and academics say that safeguards are necessary. First and foremost is the insurer must still be responsible for the underwriting decision and they must designate actual humans that can override the algorithm’s decision. A consumer must be able to say they don’t want to share certain personal data and the insurer cannot decline them solely for this. In addition, if a consumer opts out of participating in this type of underwriting the insurer’s pricing must not be punitive or extreme. And, of course, the data must be protected from theft.2
Should this type of underwriting be allowed? The regulators could stop it, but even the New York insurance department approved the use of accelerated underwriting using external data. Consumers could stop it, but many consumers already give complete access of their personal data to almost anyone that asks for it. A consumer could opt out, but then they would likely pay a far higher premium.
The industry sees this as a win because fraud will be reduced as well as claims in general, insurer payrolls can be cut, and many consumers will pay less for insurance. Whether it is a win for consumers is for you to determine.
- J. Roberts. A revolution in underwriting. Best’s Review. May 2019. pages 44-48.
- H. Albrecher et al. 2018. Insurance: Models, Digitalization, and Data Science. SFI Research Paper 19-26.