A couple of the blogs I read mentioned this LA Times article today - Insurers learn to pinpoint risks -- and avoid them. In particular RiskProf and Workers Comp Insider covered the article. RiskProf felt that the article missed the point and that better risk assessment means better precision in pooling risks and less subsidies for those who chose riskier behavior. RiskProf also pointed out that new, and hard to estimate, risks tend to put off insurers but this is temporary. As a friend of mine put it "there's no such thing as a bad risk, only a bad price" so once the risk is known there will be a price. Finally RiskProf wanred against over-regulation as a solution, preferring competition, arguing that capping rates causes insurers to stop writing policies when the rate no longer covers the risk. Meanwhile Jon over at "Workers Comp Insider" feels that the article makes some good points - that micro-segmentation defeats the pooling that makes insurance work and that losers (people identified as bad risks and made to pay a higher premium) will outnumber winners (good risks who get discounts) and that this will cause potentially severe repercussions for Insurers. Interestingly the students at UC Berkeley's Services Science Management and Engineering class asked exactly this question when I talked about micro-segmentation in insurance.
I'm summarizing both - read the posts for details - but I found both the article and the responses interesting so I thought I would add my points.
Why should losers outnumber winners?
There seems to be no particular reason why those whose rates rise because they are bad risks should outnumber those whose rates fall because they are better ones.
Why should those who take good decisions subsidize those who don't?
Should non-smokers pay more for life insurance so that smokers do not? I think most people would say no. So why should people who buy safer houses, live in safer places, drive safer cars pay more for insurance so that those who take more risks do not?
I think regulators should focus on those things someone cannot change like genetics and on ensuring that anything used to assess price can be justified and supported mathematically.
Clearly not everything that impacts risks can be controlled by an individual and I think regulators could make a case for preventing the use of aspects outside your control from impacting your risk, forcing insurers to manage that kind of risk in a pooled way. Regulators should also push for causal relationships (smoking causes lung cancer therefore smoking increases risks and can be used to raise rates) while recognizing that very strong correlations that some logical "root cause" should also be allowed (the level of responsibility people take over meeting their credit obligations, for instance, can be inferred from their credit data and a lack of responsibility might tend to cause more claims on auto insurance - poor credit behavior does not cause accidents but the correlation is very strong). Regulators should clamp down hard on anything that cannot be shown to have a real mathematical relationship - they should aim to replace bad judgment with good math.
I think the industry, and regulators, can encourage price transparency so customers can understand what they get
Hiding behind fine print is not OK - insurers need to explain what the pricing options are, how they are calculated and what customers can do about it. This need to build defensible and explicable models (see this discussion of legal issues with analytics in the predictive analytics FAQ) is well established in credit and in some areas of insurance and allows people to find out why they are not getting the best rate (some states require this explanation and I think that's a good thing).
Some insurers will rush to low risks only, others will specialize in high risk.
Someone will figure out how to tell the difference in risk between drivers in a particular category and price accordingly. This will cause others to be adversely-selected against and competition will cause a change in behavior. New products, like Pay as you drive insurance, will target segments for whom the standard risk models cause high prices.
It seems to me that insurers are pooling risks, ideally "like risks", but also spreading risk evenly over time.
Even if I have a high risk of loss (say I own a sea-level property on the coast) I am not going to get a loss every year. The insurance company is spreading my losses evenly across time as well as across similar policy holders. This element of pooling is not going away.
- No matter how fine grained insurance companies try and get they still need a statistically significant base in each segment so there is a natural limit to their size.
The insurance industry is busily adopting Enterprise Decision Management or EDM to improve the precision, consistency and agility of their underwriting decisions. Nothing in this article, or the responses to it, makes me think this will (or should) change - especially the growing use of predictive analytics. There's lots more in the Insurance section of this blog.