Price Optimization and Insurance

I spend a fair amount of time on LinkedIn posting and reading about insurance. Not long ago, there was a discussion about Instacart’s announcement that it would discontinue the use of an AI-powered tool that enabled retailers to charge customers different prices for the same products, the apparent goal being to charge as much as possible to those willing to pay for it. In other words, maximize profits.

This was allegedly in response to investigations by at least two consumer organizations that claimed to have evidence that a number of well-known retailers were using, or at least testing, AI pricing systems. According to a CBS News report, one of the investigations found that prices varied by as much as 23%–identical product, different customers.

The issue in this LinkedIn discussion revolved around the question of whether this practice is, or should or shouldn’t be, legal in the property/casualty insurance industry. Insurance pricing, beyond general anti-discriminatory practices that apply to virtually all industries (e.g., race, color, creed, national origin, religion, etc.), is more highly regulated than most other industries.

“Price optimization” is the term most often used to describe pricing variations of identical products among different individual customers. In a general sense, the term can be applied to pricing variations based on individual customer demand, competitive reasons, and varying or fluctuating costs with the goal being to maximize revenue and/or profits.

Most likely, everyone reading this article has benefited or been a “victim” of price optimization. No one is happy if they believe the latter has occurred, but everyone loves a bargain if the former happens. Price optimization in one form or another has existed for decades in the travel industry by airlines, hotels, cruise ships, etc.

More recently it has become increasingly common in industries that provide cable TV, internet, and cell services. For example, every year without fail, our cable TV service increases our monthly fee. My wife then calls, complains, and the provider backs off the increase to prevent us from cancelling service, sometimes offering a “new customer” program with broader channels or even a reduced price. Many people probably never call and just accept the higher price. That’s price optimization.

Given the rapid development of data and prescriptive analytics and artificial intelligence, absent any legislative or regulatory restrictions, we can likely expect the use of price optimization algorithms to increase in most industries.

This brings us specifically to the property/casualty insurance industry.

Insurance Journal began to cover this development as early as November 2014, reporting on Maryland becoming the first state to declare price optimization to be illegal in that state. On October 31, 2014, the Maryland Insurance Administration (MIA) issued Bulletin 14-23 “Unfair Discrimination in Rating: Price Optimization” advising insurers that the use of “price optimization” in Maryland was a violation of §27-212(e)(1) of the state insurance code.

The Maryland bulletin broadly defined price optimization to be essentially any practice of varying rates or premiums based on factors other than the risk of loss. The MIA cited a 1997 Maryland Court of Appeals case (Insurance Commissioner v. Engelman, 345 Md. 402, 413) to support its decision.

Perhaps the most distasteful form of price optimization arises from the application of a common economic principle called “price elasticity,” where a provider of a product or service charges the highest price the market will bear without losing business. When this practice is applied at the granular level to individual consumers or groups, it potentially triggers long-standing state insurance laws requiring that insurance rates not be unfairly discriminatory. What is “unfair” is determined by regulatory agencies and the courts.

The Maryland bulletin stated that, “If an insurer’s analysis indicates that a policyholder is likely to switch to another insurer, that policyholder will be charged a lower premium than a policyholder who is considered unlikely to switch to another insurer.”

The bulletin went on to describe one price optimization model that considered whether a policyholder had complained to the insurer. If so, this indicated that the policyholder was not likely to accept a premium increase. All other factors being equal, this could mean that this policyholder could be charged a lower premium than a policyholder with identical risk of loss characteristics, a violation of unfair discrimination statutes.

Less than three months later, the state of Ohio issued a similar bulletin that warned insurers against the use of price optimization that constitutes unfair discrimination in insurance pricing as it violates state law. Less than a month after that, California followed suit, as did New York a month later. In May of 2015, Florida became the fifth state to ban discriminatory price optimization. By my count, at least 18 states plus the District of Columbia have expressly banned price optimization as a rating or premium development tool.

Are all P/C product premiums, at the policyholder level, based exclusively on the risk of loss? Probably not.

Premiums cover insurer costs with some margin for profit for most insurers. Those costs include both risk-based loss costs and operational costs. Two insureds with identical risk exposures may have very different operational costs for the insurer, and accounting for such differences would pretty much inarguably be proper and not unfair.

In addition, many rating plans allow for judgmental credits or debits within certain filed ranges. Is it possible that a credit may be applied to the premium of a prospective customer based more on the competitive or retention pleading of an agent or broker than on the real perception that they will be a profitable policyholder? Probably. Risk is not always quantifiable, but hopefully insurers will focus on pricing equity that complies with the spirit of anti-discriminatory pricing laws.

What do you think? If you’re reading this article online, hopefully you are aware that there is a Comments section where you can express your opinions and share tales from the trenches, even debate issues like this. If you’re reading this article in print, head to the online version and join the conversation. You may be surprised at how much you can learn from other Insurance Journal subscribers.

P.S. — If you are active on LinkedIn, feel free to connect with me. I will be posting and commenting considerably more this year on the LinkedIn platform, including a discussion on this specific topic with links to additional resources.