Thomas Lord [12.28.07 12:43 PM]

Geeze, what's to discuss? But, seeing no comments I wanted to make clear at least one feedback point:

Nice piece! Very densely packed -- almost a trot to a much more extensive literature. I find that kind of thing very useful especially when (plausibly) packed into a year-in-review.

(I like the paper, too, though I wish it had a proper masthead somewhere. Did I just miss it?)

-t

  Ross Stapleton-Gray [12.28.07 06:34 PM]

I think this is a weak point in the essay:
More recently in Massachusetts, consumer advocates protested when automobile insurance companies decided to assign rates based not only on driving records, but on actuarial data that had nothing to do with driving: debt, credit rating, and so forth. The companies’ number-crunchers had found a correlation between these behaviors and safety in driving. But who wants to be treated like a number?
If, in fact, there's a (strong?) correlation, then "nothing to do with driving" isn't at all the case, or at least it's "predictive of driving behavior," which means that the insurance companies are doing a better job of shifting the onus of paying for road mishaps to those most likely to incur it. As far as being treated like a number, that's kind of the business that actuaries are in... I don't need my insurance company to be nice, just efficient.

  Thomas Lord [12.29.07 11:00 AM]

Ross,

"shifting the onus of paying [...] to those most likely to incur [the expense]" is exactly not the business of insurers. They are there to pool risk, not to presumptively collect hypothetical liabilities.

When someone gets a lot of speeding tickets you can say "Ok, that person is deliberately devaluaing the insurance pool so its fair to charge them more."

How do people who have a certain kind of credit rating put the pool at greater risk? What do they themselves do, that you are measuring, that justifies price discrimination in insurance?

-t

  Ross Stapleton-Gray [12.30.07 08:25 PM]

Well, Andy's telling me there's an algorithm that seems to accurately predict their effective cost. If that's the case, AND the insurers have the latitude to discriminate in pricing, they'd be foolish not to.

But that suggests one of the major ways to "defuse" some of the privacy problems: remove the ability to discriminate on pricing/acceptance. Enact guaranteed universal health care, and the incentive to dig into someone's medical history/predispositions (the latter being accessible as an extrapolation from relatives' DNA, so it's even more insidious) plummets.