Anyone who has ever had a statistics course will hopefully remember the mantra of our professors regarding correlation between two or more factors — “correlation does not equal causation.”
For those of you fortunate enough to have avoided statistics in your studies, this simply means that even though two things occur in the same period of time or with the same frequency, we cannot necessarily say that one causes the other to occur. So if I slap on some coconut sunscreen at the beach and am stung by a bee (two events), I may conclude that bees are attracted, or even angered, by coconut sunscreen. As a result, I switch to the unscented sunscreen.
Yet I cannot say for certain that my choice of lotions was the causal factor for that particular bee on that particular day. Maybe I did something else to disturb said bee, like stepping on it. Indeed, I’m not even certain that coconut sunscreen does anything to affect a bee’s behavior. Should I find that bees still target me after the sunscreen switch, I am back at square one wondering what caused that bee to sting me.
In insurance, we often see underwriting decisions based on correlations. Apparently, there is a correlation between high credit scores and low incidence of auto accidents, thus many insurers will use credit scores to assess an individual’s risk potential. I doubt we will ever see a study proving that a high credit score invariably leads to safer driving (or vice-versa), but the correlation suggests that the two frequently co-occur. So even though common sense tells us that credit-worthiness should have nothing to do with driving behavior, we treat credit scores as a “causative” factor.
Now ISO’s Applied Informatix unit has announced a program to develop a Driving Behavior Database for Modelers (DBDM). According to Steven C. Craig, GM of Applied Informatix, this program “will define the relationship between losses, traditional policy information, and driving behavior. The risk information that will be uncovered will provide insurers with a platform for improving existing products and developing new ones.”
ISO says the database will be created by collecting risk-level telematics data transmitted from insurers’ covered vehicles and matching it with associated insurance data, including exposures, premiums, and losses. “The DBDM will offer analysts the greatest statistical value and flexibility when determining how driving behavior characteristics are predictive of loss,” the company adds.
I have no doubt that such a database will, indeed, provide more statistical evidence of correlation between certain behaviors and either safe or unsafe driving. The danger, however, is that finding correlations tempts us to jump to conclusions of causation. If, for example, statistics showed that in 75% of accidents, the vehicle’s car radio was on, we might conclude that having the radio playing is predictive of an accident — completely ignoring the many other factors that may or may not have influenced driver behavior.
Make no mistake, however. I believe it will be useful to uncover behaviors that occur with poor driving records. At the same time, though, someone must use common sense in analyzing the correlational data before creating rate structures that may be based on weak causational factors or even mere coincidence.
Drawing the wrong conclusions from such data could be dangerous. If I concluded from my bee sting episode that sunscreen itself (regardless of coconut content) was the causative factor, I would likely be left with second-degree sunburn and a very painful next few days.
This originally appeared on Insurance Networking News.
Register or login for access to this item and much more
All Information Management content is archived after seven days.
Community members receive:
- All recent and archived articles
- Conference offers and updates
- A full menu of enewsletter options
- Web seminars, white papers, ebooks
Already have an account? Log In
Don't have an account? Register for Free Unlimited Access