Changes announced by the FCA last week are going to force insurers to pause for thought in the way they use big data in insurance pricing. The changes affect insurance at every level but it is the lens of the customer layer that will be the only one that counts to the FCA.
The FCA talks about why this is important to it because it needs “to make the markets we regulate work for consumers, cutting across our consumer protection, competition, and market integrity operational objectives.” This is important now because “issues of fairness in pricing are likely to become increasingly prevalent and complex in the future, particularly as firms’ use of new technologies and data becomes more sophisticated”.
So, what does all of this mean to insurers and the senior managers who work there?
In simple terms, the insurance industry has lagged other financial services markets when it comes to thinking about price and vulnerable customers. Essentially, when the Insurance industry thinks about price it, thinks about the people it wants to attract.
This new set of regulation will force it to think about the people that it might be adversely affecting. The industry will be forced to answer the question “can you explain why you gave that customer a worse price than the other customer?”.
In a world of actuarial tables, this should have been relatively simple. You take the base price for insurance and then you adjust according to certain numbers on the tables.
But pricing science has moved on at a pace. The recent big data arms race coupled with the use of sophisticated pricing algorithms has brought us to a point where no one truly understands why a certain price has been calculated. The computer has made the decision.
But if one person gets a better price from the computer then it must, by definition, mean that someone gets a worse price. And that is the person the FCA and senior managers should be worried about:
These are the six tests the FCA will apply when they start looking at pricing in general insurance. In the case of car insurance, it is essential, so that leaves only five tests. With the computer having made pricing decisions, will companies be able to know who got worse prices? Can you produce a good reason (based on actuarial science) as to why they have been disadvantaged?
At issue here is the ability to be able to explain. In a complex world with complex algorithms stitched together by machine learning it is very, very difficult to understand why one group gets a lower price and another gets a higher price.
And even if you could “explain”, would society consider this egregious? Let’s just say that the statistical model shows that left-handed people (I am one) are worse risks than right-handed people. You can get some data which approximates which hand is dominate and you can then use this in the rating tables. Does this pass the fairness test? You haven’t been able to explain what the difference is between a left-handed driver and a right-handed driver, it’s just a computer saying something. As a left-handed member of society, I’m going to say that isn’t fair until you can produce some evidence that proves why I am less able than my right-handed friends. Small point if I am less able then am I more vulnerable?
Bottom line: is it worth it for the few percentage points you might gain? Let’s look at this another way. Is all this big data and advanced analytics worth it? Is the fractional, unexplainable gain giving you the level of competitive gain that a simpler, more explainable model might give you? It is certainly something to add to the risk register!
Insurance pricing is rapidly becoming a science, over the last decade Consumer Intelligence has observed a radical change in the way that insurance is priced by insurers and valued by customers. In this paper, we explore the current best practice in insurance pricing and give you some guidance about where the industry is about to go next.
Comment . . .