> Punitive insurance rates based on unreasonable inferences, especially for mandatory insurances. Like say doubling your auto insurance rate for buying more than a few beers per week. Or your health insurance going up from buying a power tools.
Is it just supposed to be self-evident that those inferences are unreasonable?
I've always thought that this was an interesting argument. If there is some form of correlation with beer consumption and and car accidents, wouldn't it make sense to adjust your estimated risk based on that information?
I do find it self-evident that that would be a bad thing, but I also have a hard time putting my finger on why.
The problem is that "correlation" seems objective and mechanical, but the model itself carries the bias by choosing which overly simplistic factors are relevant.
Directing focus at "people who drink a lot of beer" means considering people who who drink a lot of beer at home as guilty by association, ultimately due to the subjective priorities of whomever pushed for that model.
Obviously in the expected value sense, charging on correlations is lucrative for the company (as is any justification for raising prices on a set of customers if your competitors do it too). But in the exact same way as saying certain zip codes are more likely to default on a loan, which we rightfully reject.
Your model sucks if you are arbitrarily choosing factors. You choose the factors with the most significant correlations, because those correlations are least likely to be "overly simplistic".
Why would we reject that zip codes are more likely to default on a loan? Seems like information I would like to be aware of if I was a home lender.
I certainly look at crime rates of a community before I live there. While a bad crime rate certainly doesn't make potential neighbors "guilty by association", it certainly increases the likelihood that one of my neighbors might be actually guilty.
> If there is some form of correlation with beer consumption and and car accidents, wouldn't it make sense to adjust your estimated risk based on that information?
Nope. Because it's flawed reasoning. If many people who get into accidents were driving drunk and everyone who drives drunk buys beer it might seem logical to increase rates for everyone who buys beer, but people who drive drunk are only a small percentage of the people who are beer buyers. That kind of reasoning seems more likely to be a weak justification to raise rates for a large number of people than a reasonable response to a trend.
This is assuming that auto insurance isn't a competiitve industry. If a company attempts to raise rates because of a trend that doesn't actually exist, they will inevitably not be competitive with companies that recognize that the trend doesn't exist, and thus it won't change prices for the consumer.
If the insurance companies could arbitrarily raise rates due to a trend that doesn't exist, than they would have already done so. These companies know their margin and they don't bid above that if they want to be competitive.
> If a company attempts to raise rates because of a trend that doesn't actually exist, they will inevitably not be competitive with companies that recognize that the trend doesn't exist,
that assumes that all companies involved aren't doing the same thing. Corporations figured out a long time ago that when one of their competitors does something that makes them more money at the expense of their customers they could start doing the same thing to their own customers and profits increase for everyone without risking prices being driven down by a truly competitive market. The insurance industry in particular is has a long history of shady practices from good old fashioned collusion and price fixing to new techniques like data mining to charge customers different rates depending on where they live, what jobs they have, or how often they're willing to change insurance companies.
If you have reason to believe that the insurance industry is colluding to artificially inflate prices, that's a criminal accusation. If you want to make criminal accusations you start with evidence.
In my view, the insurance industry looks competitive, which means that even though these shady practices happen they can't effectively dictate the entire market.
Buying beer is legal. Driving while sober is legal. Assuming you have no record of driving drunk, any such alleged correlation should not be used to inflate insurance rates, unless you're also willing to say that other correlations of increased risk are also fair game, even if they're based on race or sexual orientation or income or education level or politics or any other characteristic that can be measured and grouped into risk categories.
One of the tendencies of the neo-puritanism that has become prominent in the last decade is a real willingness to abandon any boundaries that have kept corporations from using certain kinds of information against individuals -- boundaries that were in large part legislated during the civil rights era.
It's not illegal to be a male, and my insurance premium still raises because of it.
It's the job of the insurance company to accurately assess risk and charge me that plus their margin. If the companies can more accurately assess risk, than that makes insurance a less volatile and therefore cheaper market.
Insurance companies don't have access to the actual root causes of accidents. They have no measure of my driving skill or risk tolerance or attention span. They just estimate based on some really primitive data they have about me. What's the harm in including more data?
Is it just supposed to be self-evident that those inferences are unreasonable?
I've always thought that this was an interesting argument. If there is some form of correlation with beer consumption and and car accidents, wouldn't it make sense to adjust your estimated risk based on that information?
I do find it self-evident that that would be a bad thing, but I also have a hard time putting my finger on why.