People with bad credit cost the insurance company more in claims. Not the same thing. And I would suspect that is because they don’t have the money to pay for repairs that someone with better credit would pay out of pocket rather than report it and have their rates go up.
Google for the aptly named “Hurt Report” (actually, I think it was done by a guy named Hurt), which breaks down motorcycle accident rates and the severity of injury suffered in motorcycle accidents by all kinds of factors.
One of the findings was that a rider’s level of education is inversely correlated with both the rider’s likelihood of getting into a motorcycle accident and the injuries the rider suffered if an accident occurred.
So, people with more education tend to be safer motorcycle riders than those with less education. It makes sense to me that this would carry over to driving a car, too, and many of the categories of jobs you listed as getting auto insurance discounts require more education than other categories of jobs.
Well, a counter argument is that people with poor credit can’t get into expensive cars and wreck 'em. They’re trashing the cheapos, but do you know how many 1995 Saturns you have to burn through in order to equal a 2004 Denali? I don’t buy it.
Insurers give discounts for good students. Not because they wanna encourage studious behavior, but because good students can generally be relied upon to PAY ATTENTION. In class, and on the road. Engineers are infamous for their attention to detail. People who get into credit trouble show, as a group, poor planning skills & lack of foresight. Possibly they are risk-takers. The Insurance agents & clergy that I mentioned above spend a lot of time with their heads on topics other than driving, while they are driving (this bit is anecdotal).
I’m not going to go digging for cites, but auto insurance rates are based on claims. If a specific group, such as the group of good students, is offered a discount, it is because the claims history shows that they can still make money that way. Insurance companies don’t offer discounts just because they think good students will have fewer claims, they know it. Ditto on the obverse for bad risks.
Interesting post, TaxGuy, I’ll go look for it.
This Year’s Model, I think we’re basically on the same page. That insurance companies base their decisions on data is not a point I’ve contested. I was just speculating on why that data might exist in the first place in an effort to explain why insurance companies don’t set rates using a Witch Board & Magic 8-ball as seemed to be the belief. 
That’s why I said “however you want to define race”. Whether “race” even exists as a genetic reality is debatable, but there are certainly objective definitions of race and/or ethnicity that could be used.
The point of identifying groups like “young drivers” or “women” or “engineers” for insurance isn’t to reflect any underlying scientific reason that people in those groups are safer drivers - the point is only to come up with a common factor that statistically makes drivers more of a risk. There’s no biological reason that someone should be more of a risk the week before his 25th birthday and less of a risk the week after, but insurers (presumably) have statistics to show that people over 25 are less of a risk, so they use age as a factor for premiums.
If they also have statistics to show that, say, people with at least one grandparent from China are better drivers on average, or people whose skin is lighter than shade X on some chart are worse drivers on average, that serves the same purpose. It doesn’t matter whether there’s a biological reason that Chinese or white people should be better or worse drivers. All that matters is that insurers can describe a group of people who present more or less of a risk.
The insurance company will, using whatever definition of “race” they used to identify people for their statistics. If their statistics group people by skin color, for example, then it won’t matter what you claim to be; your “race” will be determined by a color chart.
So plan that trip to Hawaii strategically. See the problem? Gonna buy in the summer or winter? Some criteria can be changed artificially. I’m sure there’s plenty of folks who manage to convince the DMV that they are much older than they really are, but that takes a bit more moxy than climbing into a tanning booth or pencil whipping your family tree.
Some of the people included in the original statistics surely would’ve gotten some sun recently, so that would already be accounted for (to some extent). Also, tanning can only change your skin color so much; the new Michael Jackson is not going to look like the old Michael Jackson again, no matter how much time he spends outside. Choosing the exact shades based on the limits of healthy tanning, and possibly correlating skin color with other physical traits, would let you account for tanning even further.
That said, I doubt insurers would actually use criteria that are as variable as skin color… but there are other objective criteria that aren’t so variable.
Again, the point isn’t to identify genetic “races”, but only to identify a group of people who are associated by physical traits and are more or less risky than average. If those groups don’t exist, insurers would have no reason to use those traits to decide premiums; if they do exist, then why should insurers be prevented from using them?
If you can figure out a way to change who your ancestors were, let me know - I’d like to be a wealthy prince. 
I have working in risk managment and insurance for the last, oh, 13 years or so but I am having real trouble imagining what possible objection you can have to rates being differentiated by all sorts of criteria - including age, occupation, sex, marital status, home city, income level, education level etc.
Unless it is transcribed by law - like race, and as will genetic testing information I predict - then surely it is completely fair to weight the premium of the assured to the risk he or she presents?
It would be unfair to do otherwise - but in the case of race and genetics I can see that the level or unfairness creating by legal preventing rating those factors is outweighed by the potential for abuse.
All these rating factors will be driven by statistical analysis which identify *statistical associations * **not ** cause and effect. Now it is the job of good product developers to try to get behind associations and discover the true risks factors. In a class like auto insurance - with a mass of statistics - this is possible, with other classes (say, satellite launch insurance) it is more difficult.
All rating engines for mass personal lines classes like auto and homeowners will be driven by backward looking analysis of actual claims incurred versus risk profile not theoretical forwarding looking guesswork as to likely risk.
Why should anyone get their knickers in a twist of this appoach?
I am guessing it is something to do with your weird insurance commissioner system, with political influence on rates permitted for personal lines insurance etc. proposition XXX giving you discounts etc etc. So much for a free market, eh?
We like to protect folks from capitalist predators in *some * areas. 
On better days I wouldn’t say this, but it seems like everyone feels they are entitled to be treated as an individual and not as part of a statistical group to which they belong. “Sure 16-17 year old drivers cause more and worse accidents than any other group, but* I’m * different.” Unless of course that group gets a perk. Then we all wanna be a number!