Statistical thinking

The earlier post about car insurance touched a nerve.

That prompted this post.

It is often possible to find a proxy or a relationship between behaviors.

For example, insurance companies found a relationship between people whose address-of-record had a Detroit Zip Code and an increased risk of having the car stolen. The difference was large enough that it justified segregating that population from the general population.

The Governor demands that "non-driving" data cannot be used in setting rates.

I am just spit-balling here, but what if insurance companies discovered a relationship between drivers who run red-lights and an increased rate of vehicle theft?

A side issue is that many police are loath to write traffic tickets for things like running red lights. I have seen light changes where they would have to write eight tickets for every red light. But many intersections have cameras for that kind of thing.

There might be pushback regarding the fact that the only place you find many intersections with lights is metropolitan areas and only the large cities have the automated cameras. That does not negate the fact that it is "fair game" for insurance companies to apportion the cost of covering theft risk based on documentation of any driver running red lights.

It also creates an incentive for insurance companies (or their customers) to file law suits against cities if the automatic cameras are disabled. Currently, I don't give a rip whether they work or not.

But I will care if automatic cameras are a statistical link used by the insurance companies to more accurately apportion costs. If the cameras are inactivated either by politics or vandalism, then my rates will rise because those costs cannot be rationally apportioned.

It is what the Governor demanded: That the price of vehicle insurance be based only by "driving" data.