Data: Life Insurance
A Revolution in Underwriting
The race to develop accelerated products has driven life insurers to cautiously embrace the next generation of data.
- Hot Topic: Unconventional data driving accelerated underwriting is “the No. 1 topic” in the industry.
- Emerging Data: Everything from credit information, retail purchase history and even social media and facial analytics could soon impact underwriting.
- Accelerated Competition: In two years or less, every carrier in the industry will offer accelerated underwriting at a fully underwritten price, driven by competition.
The sweeping letter was a warning to the industry.
It spilled over six single-spaced pages and 2,000 words, putting life insurers on notice for the emerging use of unconventional data in their automated underwriting.
Data such as criminal and civil judgments. Credit information. Retail purchase history. Internet and mobile device usage. Geographic location tracking. Even social media and facial analytics, sources rarely used now but expected to be widely adopted in coming years.
After an 18-month investigation into insurers' underwriting practices, the New York State Department of Financial Services leveled a stern warning: Apply external data only if you can justify its actuarial validity and independently verify it does not discriminate or contain prohibited criteria.
But also tucked into that guidance was approval for using third-party data that has “the potential to result in more accurate underwriting,” the January letter read.
And with that, the influential regulator became the first to establish specific guidelines just as the exploration and application of nontraditional data in algorithms soars.
“The gist of that letter was insurance companies couldn't outsource whether [the data] was discriminatory to the vendor. It was on them, so they better know what they're doing,” said Tom Scales, head of life and health insurance at Celent.
The race to perfect fully underwritten, accelerated products using algorithms, predictive modeling and analytics as a substitute for paramedical exams and fluid tests has driven life insurers to increasingly embrace new forms of data.
Leveraging it enables carriers to provide a shorter, cheaper and more customer-friendly approval process amid rising consumer expectations in an Amazon world.
But that emerging data carries a host of privacy and regulatory concerns. It also presents accuracy and reliability issues that need to be addressed.
However, accelerated underwriting and external data remain “the No. 1 topic” in the industry, Scales said. “How can we change the way we underwrite? How can we do instant underwriting?”
Using alternative data from new sources such as social media and other digital footprints is “the next big thing” in life underwriting, said Mike Vogt, executive director of data, analytics and machine learning for technology consulting firm SPR.
“We are at the beginning of the curve with how insurers are applying unconventional data,” he said. “The biggest change and the biggest risk will be the information that we gather from social media and [artificial intelligence] will actually lead to more accurate risk predictions—at the expense of privacy.”
About 25 U.S. insurers offer accelerated underwriting using nontraditional data streams, and several more are testing platforms.
The objective is to skip the invasive medical tests whenever possible without losing precise risk assessment and fraud detection.
“It's a game-changer. Unless there's a regulatory challenge, we're 24 months from everybody doing it at a fully-underwritten price, at least up to a certain age, because your competitor is going to do it,” Scales said. “That's the heart of all this. It's not simplified issue.
“This is the same price as a regularly underwritten product. It's just underwritten differently. It's part of an ecosystem change.”
Insurers are using data analytics tools such as LexisNexis Risk Solutions, TransUnion TrueRisk Life Score and MassMutual's LifeScore360 to cull data and supply a mortality score from a wealth of sources.
Think of those scores as the mortality version of credit scores in the mortgage loan process. They have developed over the past five years, and in the case of LexisNexis, include information from more than 20,000 databases.
Meanwhile, a new frontier of alternative data is emerging from social media, facial analytics, retail purchases, public filings and epigenetics—the study of how environment and lifestyle choices such as diet, exercise and substance use influence mortality at the molecular level—to further understand and price risks. One day, genetics could join them.
The products people buy, the services they use and even the magazines they read can be highly predictive of policyholder longevity, analysts say. And so can the things they say and the photos they post on social media.
Only a “small handful of carriers” are using such information, said Samantha Chow, senior life insurance and annuity analyst at research and advisory firm Aite Group. But many insurers are exploring them.
“You're talking about everything from scoring data to social data to data from selfies and DNA,” she said. “Over the next couple of years, you'll see people utilizing more advanced scoring methodology using this type of data.
“How soon depends. How scary is it? It's not about changing how they underwrite. It's about being more accurate in their underwriting, pricing and improving the overall experience.”
Over the next couple of years, you’ll see people utilizing more advanced scoring methodology using this type of data. How soon depends. How scary is it?
The embrace of alternative data is growing rapidly.
Five years ago, nearly all carriers told Scales they were either not ready for accelerated underwriting, “or they called me an idiot” just for asking about it, he said.
Just a year ago, many insurers were still debating whether they should offer a fully underwritten product using external data.
But the “life underwriting revolution,” as Scales called it, has arrived in earnest.
“They're tightening up the experience to the point where they expect their take-in rate is going to go up because it's going to be easy,” he said.
The old way is not. Forcing consumers to fill out applications that exceed 25 pages on average and undergo a medical exam has contributed to long-term declining sales trends.
“While these tests help ensure efficient underwriting, they add considerable time to the application process, and are undesirable and uncomfortable for insurance applicants,” said Jeff Heaton, vice president and insurtech data scientist at Reinsurance Group of America, in a written response to questions. “The desire is to use new data sources that can be obtained simply by obtaining the permission of the applicant.”
Such data streams can become effective underwriting tools, especially as a supplement to the four traditional sources: applications, Medical Information Bureau reports, prescription drug histories and motor vehicle records.
Unconventional data sources range from census information, public filings (homeownership records, bankruptcies, property deeds, tax filings and licensures), credit information and geographical data (community-level mortality, addiction and smoking data).
More alternative data such as digital footprints, social media and purchase histories—encompassing everything from consumers' grocery store receipts to the publications they subscribe to and services they use—may soon augment current rating factors.
Much of it is supplied by data brokers and insurtechs that collect and analyze information, then package it and sell it.
Just one out of 160 insurers operating in New York told the state's regulators it was using social media, retail purchases or internet activity in underwriting, The Wall Street Journal reported in January.
Then-New York DFS Superintendent Maria T. Vullo told the newspaper that the objective of the letter was to establish ground rules before the use of such information became mainstream. Many expect that to happen soon.
But algorithms and models are only as accurate and unbiased as the information that feeds them, as the letter from the regulator warned.
“The relationship between the vendors and the carriers is going to have to change, and there's going to have to be more transparency,” Celent's Scales said. “But that was inevitable.”
Privacy and regulatory concerns have given many insurers and vendors pause.
Insurtech data analytics firm Carpe Data “definitely explored” solutions for life underwriting, said CEO Max Drucker. But it was dissuaded by the contentious privacy debate.
“There is a lot of controversy around social data to make life insurance decisions,” Drucker said. “We focus on small commercial because no one is arguing about privacy or whether it's right or wrong.”
The digital lifestyles of the 21st century have given birth to new ways to evaluate risk.
Did a consumer post photos on Facebook showing him smoking or skydiving? Did an applicant post a video of herself on Instagram while at the gym or tweet about her love of kale?
“Posting a bunch of pics of drag racing, ice climbing, partying a lot … that indicates a pattern of risky behavior,” Vogt said. “But posts of running half-marathons and healthy eating paints a very different picture.”
But using social media in life underwriting has proven challenging.
The cost, time and complexity of monitoring such information at scale and converting it into a metric or score remain obstacles to wide utilization.
“They're not to the point where they can look at your social media posts and other kind of data that isn't scorable for underwriting,” Scales said. “Some of this comes back to regulation. You've got to be able to actuarially prove the decision you're making is not only accurate, but is not discriminatory.”
But they are testing it, according to John Lucker, a principal with Deloitte Risk &Financial Advisory and global advanced analytics market leader.
“I've talked to numerous life insurers and they are experimenting with it,” he said.
The capability already exists to apply other alternative data in rating applicants.
In fact, some vendors and insurtechs have expressed frustration with insurers that have explored emerging opportunities, but have not put them into practice.
“These insurtechs have great solutions, and the carriers come out and invest in them,” Aite Group's Chow said. “Then they put them in their basement so no one else can benefit from their tools and give them only a part of the data they need to do their proof of concept.
“I've seen this happen a couple of different times.”
So startups have begun changing their business model seeking partnerships with carriers rather than complete funding.
Some of those insurtechs offer facial analytics, in which selfies are examined to glean health and lifestyle information by studying the lines and contours on a person's face.
Color variations in the whites of someone's eyes can show the presence of certain conditions. Lines around the mouth can indicate a smoker. Dark spots can offer information on how someone is aging and even their life expectancy.
About 20 insurers reportedly are testing Lapetus Solutions facial analytics technology. One of them, Gen Re, launched a prototype app in February that allows consumers to purchase insurance by uploading a selfie to calculate their estimated age, gender and body mass index.
Insurtechs have also embraced epigenetics and how environmental and lifestyle factors potentially can “turn on or off” positive or harmful genes.
“Epigenetics can tell you about basically everything,” Chow said. “It's about as close as you can get to DNA testing. And it's not intrusive, especially for people afraid of needles” because tests use a saliva sample instead of drawing blood.
Wearables—which John Hancock applies with its Vitality program, now part of all of its life policies—are also “something that everybody's thinking about,” she added.
They track heart rate, sleep patterns, exercise and other factors that anyone who owns an Apple Watch or Fitbit already records.
Insurers can build customer loyalty and engagement while gleaning information to improve mortality experience.
Eventually, life insurers will have enough data at their disposal to even preapprove some consumers, much like a credit card company does, Scales said.
At What Cost?
But will consumers trade their privacy for convenience?
The more capabilities presented by emerging data, the larger the privacy and regulatory issues loom for insurers. The Cambridge Analytica/Facebook scandal only exacerbated fears that Big Brother—and your insurance carrier—are watching.
Many consumers view the thought of their insurer scrolling through their Instagram posts and shopping histories as a violation of privacy. Others balk at reporting biometric data from their Apple Watch, even for a discount.
Lucker wonders if the value of social media will eventually be negligible, with a growing segment of privacy-conscious consumers locking their accounts or scrubbing them of certain behaviors.
But the industry and analysts say customers have long had to voluntarily surrender personal information—such as medical records—in order to gain coverage. Some ask if granting access to that data is not a preferable trade if it means skipping an invasive medical exam and fluid screens.
And in the meantime, younger generations grow ever more comfortable posting their entire lives on social media. They might not be so bothered by insurers, or anyone else, probing their accounts.
“Life insurance doesn't work without data,” Scales said. “They're so fundamentally built on the fact that if you have triggers that are going to cause you to die early, they can't give you a cheap rate.”
The industry needs data, but it also needs to know from where that data originates. And as the New York regulator warned, insurers need to be transparent in disclosing the reason for a rejection or other adverse action in accelerated underwriting.
That might not be easy considering the large volume of external data that goes into carriers' complex algorithms. It is one of the many sensitive issues involving next-generation information.
Those uncomfortable topics may be why some insurers have chosen not to address emerging data publicly. AIG declined an interview request through an outside public relations firm. Scor did not respond to an email. Swiss Re said no one was available for comment.
MassMutual released a statement.
“MassMutual's ongoing investments in data science—along with advancements in technology and data aggregation—have enabled us to improve on all of the traditional ways we've utilized data, while also providing consumers with quicker and easier access to the peace-of-mind and financial protection that life insurance provides,” the company said. “Benefits include the opportunity for fewer and less invasive tests for the underwriting process and an enhanced ability to accurately assess risk. Additionally, and importantly, MassMutual is committed to the fair and responsible use of any external data in life insurance underwriting.”
The New York DFS declined an interview request, citing a lack of available staff. It then declined to answer written questions about the guidance letter and the investigation that preceded it, saying it does not comment on specific details of an investigation.
While insurers explore more alternative data, they continue to venture further into information sources that can be easily and quickly obtained that won't pose regulatory issues. Electronic health records, including individual clinician records, and financial information purchased from third parties are increasingly a common source.
One day soon, consumers' Instagram accounts and retail purchases may join them.
“This is just applying technology and applying data to be the logical, next evolutionary step,” Scales said.