Insurers’ growing use of predictive data analytics is predicated on enhancing operational efficiency, improving decision-making and offering more detailed pricing of products. It’s that last practice that has provoked a firestorm of criticism.
Predictive data analytics leverages statistical techniques to analyze current and historical data to make more accurate predictions about the future. It is considered one of four breakthrough technology-related trends altering centuries-old ways of doing business, the others being social media, mobile and the cloud. Diverse industries are leveraging these tools to improve business forecasts, sharpen marketing methods, and tighten supply chains and compliance controls.
Data analytics also provides a way for industries to determine if consumers are browsing the Internet looking for better deals on a product they already have purchased from a company, such as car insurance. This knowledge arms the organization to contact the consumer and offer a lower price on the extant auto insurance policy, thereby minimizing the chance of losing the business to a competitor.
On the surface, this practice is called price optimization appears to be a “win-win” for the consumer and the insurer. The consumer is offered a reduced premium and the insurer maintains the business.
Not necessarily so, according to several state insurance regulators. Mandated by law to ensure nondiscriminatory auto insurance pricing practices (more precisely, practices that are not unfairly discriminatory), the regulators are concerned that price optimization is really unfair price discrimination, giving a bargain to one consumer (let’s call her Mary) and not to another (let’s call her Sue).
From a risk standpoint, Mary and Sue are exactly alike—women of the same age residing in the same state, with similar driving records, credit history, and so on. Both Mary and Sue have bought auto insurance from the same carrier and pay the same premium. This now changes: Mary is discovered by the carrier to be browsing multiple insurer websites looking for a better deal on her car insurance, prompting the insurer to gives her a 10 percent discount. Sue is not browsing for a better deal online and gets no such offer.
This difference in treatment is potentially discriminatory, regulators claim. “We’ve got no problem with an insurer modeling its expected costs related to a particular policyholder on traditional factors like driving history, and then pricing the risk accordingly,” said Lee Barclay, senior actuary at State of Washington’s Office of the Insurance Commissioner.
“But, this is modeling a consumer’s behavior insofar as their likelihood of renewing with an insurer,” Barclay added. “When you are modeling things other than the risk that will be transferred from the policyholder to the insurer, we have concerns.”
In Violation of State Law or Not?
Washington State is not the only U.S. state with this position. Many other states consider price optimization to be clearly discriminatory. Other states with slightly different rate filing regulations are not so sure. Consequently, the subject from a regulatory treatment standpoint has compelled the National Association of Insurance Commissioners (NAIC) to dig into the details.
In March, the NAIC released a preliminary draftof a white paper basically laying out the issues. A final draft is expected in late-August 2015. Barclay and the other regulators quoted in this article are members of the NAIC’s Casualty Actuarial and Statistical Task Force tasked with delivering the white paper.
The subject has heated up in recent months, with several insurers charging that the regulators are seeking to stifle their innovation. Other industries as diverse as retail, real estate, manufacturing and transportation are using data analytics to segment their customers for pricing purposes, why shouldn’t they? In an open, competitive marketplace, price discounting that is based on Mary’s online browsing habits is a way to retain her business by offering something she will want.
But, what about Sue, who now will be paying more premium than Mary is paying for the same insurance policy? Insurance consumer advocacy groups like theConsumer Federation of America argue that the practice discriminates low-income people, who tend to shop around less frequently than wealthier consumers, due to potentially fewer market options. Low-income people also tend not to switch insurers as frequently as those with fatter wallets.
In a letter to state insurance commissioners, the CFA blasted price optimization as having nothing to do with a driver’s risk, i.e., the traditionally accepted methods of calculating premiums based on projected costs, such as claims, overhead and profit. Rather, the practice bases insurer premiums on the maximum amount a consumer is willing to pay.
The NAIC would appear to agree with this position—to a point. “Price optimization is not necessarily risk-based pricing,” said Eric Nordman, the NAIC’s director of the regulatory services division and director of the center for insurance policy and research. “The law says that rates `shall not be excessive, inadequate or unfairly discriminatory.’”
He noted that two people “in the identical situation, who should be charged the exact same rates are now being charged two different rates, the only difference being one of them was shopping for a better price. This is not pricing for future losses based on historical information.”
By not offering the same price to comparative risks, the insurer would appear to be unfairly discriminatingagainst these policyholders. Since the company will generate more revenue by notproviding the cheaper-priced insurance across the board, the higher-priced rates may also be considered excessive under the law. And for the lucky consumers like Mary that got a better deal, the rate may also be consideredinadequate.
That’s three strikes. Or is it?
The NAIC’s preliminary draft points out that “expected retention” is just one of many rating factors that companies customarily use when modeling a driver’s risk. Others include expected profitability, premium volume and expenses. The difference is that expected retention in past was a subjective determination, along the lines of what expected market demand might be. Data analytics, on the other hand, presents the means to actually quantify expected retention.
Is this any different than merely mulling the potential for a customer to switch insurers?
The NAIC draft contemplates the distinction. It even goes so far as to note the benefits of price optimization, which may improve rate stability, limit policyholder disruption, and lower an insurer’s long-term cost for providing coverage.
But, the draft elevates the more dire threat of potential discrimination against lower income consumers, finding some credence in the argument by consumer groups that Sue may not be shopping for insurance simply because she earns less money than Mary.
“People who are less likely to switch insurers will wind up being overcharged in relation to the risk—that’s the key issue,” Barclay explained. “One group [of consumers] ends up subsidizing another group, paying more than their fair share.”
(Editor’s Note: In the draft white paper, two key areas are yet to be filled in: a section on best practices and principles to be identified by the task force and the concluded section, “Recommendations and Next Steps.”)
Hard Lines and Dotted Lines
Several states like New York and California are playing hardball, sending letters to insurers in their domiciles strictly prohibiting the use of price optimization. Many other states remain on the fence, waiting for the full release of the NAIC’s white paper. The preliminary draft provided no guidance to regulators on the subject, noting that best practices and other recommendations were in the process of development.
Without clear directions, states are carefully weighing their options, based on their interpretation of their respective laws. The State of Oregon, for instance, permits prospective rate filings to include factors indicating that the current rate, if continued, would result in a disruption to its book of business.
“We’ve looked at some recent [rate] filings that use proprietary models to assess rate shock, and they appeared to pass muster,” said David Dahl, casualty actuary at the State of Oregon Insurance Division. “In one case, the filing specifically noted `policyholder retention.’ We requested additional information to be sure that this was, in fact, an accurate predictor.”
The insurer provided the information and the department found its use to not be unfairly discriminatory. “Our statutory authority is to ask for what we need in order to do our own [rate] analyses,” Dahl explained. “Our review is pretty extensive. We currently feel we are getting what we need and haven’t decided if we need any further guidance. But, we are still studying the issue. The jury is still out.”
As Dahl’s comments indicate, each state has its own perspective on the subject. The NAIC preliminary draft recognizes these varying views, noting that some state responses may be to disallow the use of price optimization, whereas other states may want to request more specific disclosures in the rate filings to ascertain possible discriminatory practices, as Oregon did. The association also acknowledged that some states might determine to pursue an altogether different direction—to allow price optimization as a rating factor.
A Delicate Balance
As the draft seems to concede, the subject remains open to each state’s interpretation of its laws as they relate to pricing discrimination. Not that clarifying the distinctions will be easy.
“We have a dual role—to protect consumers from unfair pricing practices first and foremost, but also to provide an insurance market of available insurance to the public,” said Richard Piazza, chief actuary, at the Louisiana Department of Insurance. “Sometimes these obligations clash. That’s possibly the case right now with price optimization.”
Piazza commented that the state’s laws lean to the benefit of consumers in this Solomon-like decision. History proves his point. “In Louisiana, if you’re over the age of 65, there is justification actuarially to charge these consumers more for auto insurance—they tend to get into more accidents,” he said. “But, at the same time, these people also are on fixed incomes in many cases. If you raise the rates, they may not be able to afford insurance. So our legislature, in its wisdom, decided to flatten the rates for people that age and older.”
Is the same result in store for carriers using data analytics to price auto insurance, since the practice may have a similar effect on another low-income class of drivers?
“It’s a good question,” Piazza said. “The Consumer Federation of America says these individuals tend to not shop their business as much as people who earn more. If that’s the case, and they are going to be end up paying $20 more per month than someone who actually is shopping online for car insurance, then that is a concern.”
But, he pointed out that the subject is not as black-and-white as it appears. “In our state, the law says you can’t discriminate against someone based on their race, creed or national origin—that’s clear and obvious,” Piazza explained. “But, what if people who are shopping online for a better rate are actually doing it because they’re making less money and are deeply concerned about the cost. Would pricing the business at a more affordable rate be discriminatory or in the individual’s best interest?”
It’s a difficult question, one that Piazza and his fellow actuaries continue to debate as they prepare to deliver the final version of the white paper in a few months. Stay tuned.
This article was originally published at Carrier Management.