Hacker News Reader

Comment by harles | Hacker News Reader

There’s definitely bias in these numbers. I’m not sure what’s at play, but the lowest accidents per driver being from brands that no longer exist (Pontiac, Mercury, and Saturn) certainly looks suspicious. Maybe these are good proxies for age and filter out new drivers?

View on HN

The brands are proxies for the buyers, so it's biased only if you think the manufacturing of the car is the culprit.

Pontiac, Mercury & Saturn are more likely to be driven by older drivers or at least people who keep their cars in service for a long time (which doesn't work out so well if you are getting into accidents).

DUI is very much a white guy crime (check the stats), and very much tied to a sense of entitlement... sounds like a lot of BMW drivers, no?

> Pontiac, Mercury & Saturn are more likely to be driven by older drivers or at least people who keep their cars in service for a long time.

Heh. Nah. All those brands are being driven by my poor friends and relatives.

IDK I'd drive an M/T SC2 again if I could get one for the right price. It's a very specific level of scary-fun that is absent in modern vehicles.

That suggests to me you're disproportionately likely to avoid getting into accidents.

I'm going to go out on a limb here and say that your poor friends and relatives keep their cars in service for a long time.

More likely they're getting hand-me-downs from relatives or buying very old used cars. It's hard to keep an old car running if you don't either have capital or free time to replace aging parts.

In my experience, since getting a replacement vehicle is challenging, deductibles are crushing, and increased insurance costs are debilitating, the poorer you are, the more careful you are driving your car.

> being from brands that no longer exist

This may surprise you, but some people are poor and drive very old cars, some of whose brands no longer exist. They can only afford to incrementally fix them.

The article ranked the brands by incidents per 1,000 drivers.

Which is why they went with "rates". Right there in the title

We're waiting on the resurrection of Dodge Darts, Ford XL and Ford Sunliners so ours can get some new road relatives.

> This may surprise you …

That certain doesn’t seem in the HN spirit. Can you elaborate on what that has to do with brands being a proxy for hidden variables?

What he's saying is they may have calculated rates based on how many cars of that brand were sold rather than how many are still "in circulation".

Then the out of production brands would have infinity DUI, accidents, etc...

No, I think DP (1) didn't understand "rates" in the title then (2) critiqued methodology without opening the link

or, maybe, if you are the type of person who can only afford a car whose brand no longer exists, you are sufficiently risk averse that you are much less likely to get an accident. Makes sense to me. Also, it may be that a lot of these car brands only exists in a sort of suvivorship-biased place. For example, I drive a '94 Camry, in large part because I've never had a need to replace it, having never been in an accident.

There's something so interesting about post-hoc reasoning. Everything is always explainable. Once you have the facts, you can construct the rationale. Somehow, if the stats showed the inverse, I bet there'd be someone explaining how "people driving cheap cars don't care if they get a few bumps and scratches" and so on and so forth.

And all of those brands are GM or Ford platforms. You can basically find all of the parts, except for trim and body panels.

Saturn parts are notoriously difficult to source.

Yeah, that'd be my guess. People who have been driving the same vehicle for 20 years are probably not in it for the thrills. In contrast, Teslas are fashionable and probably disproportionately owned by younger drivers who like to "take it for a spin".

Another bias might be in product line-ups. Pickup trucks are often used in construction, hauling, snow plowing, and so on. All of these activities are accident-prone. So why is Ram #1 and Ford is way down the list? Probably because Ram only makes trucks, while Ford sells a lot of smaller vehicles too.

The _average_ age of a light vehicles in the US is nearly 13 years. That means half the cars on the road are pre-2010. These "dead brands" mostly died around 2009.

Not sure if it matters for your main point, but I think you're confusing "average" with "median".

Are you suggesting the median age, rather than the average age, for whatever a "light vehicle" is, is 13 years old?

Edit: nevermind, your comment makes more sense after re-reading the parent comment.

Survivorship bias; if the car brand hasn’t existed for almost 15 years, you’d _expect_ surviving examples to mostly be owned by responsible drivers.

Just because the cars are no longer made doesn't mean that the data is bad.

This could imply that people who drive old cars tend to drive them more carefully.

Right, but it does change how I’d expect the results to be attributed, with brand not being the relevant predictor.

It could also imply that they drive them less. People without money for new cars also generally might want to spend less on gas, or, might need to forgo driving until they're able to afford repairs. Several of the brands at the bottom of the list in particular are brands I see driven by older folks: Buick, Lincoln, Cadillac, etc. In rural areas at least, this often means retirees without a commute who may only take the car out for church on Sunday.

Other theories: Kids tend to inherit older cars from their parents, and may but fewer miles on them. Older cars might get used more as "the beater for around town" if the household has a newer primary vehicle. There could also be some geographical distribution that's a part of the puzzle.

The bias is they are measuring accidents per driver, not accidents per driven mile.

Classic case of Lying with Statistics.

Reply on news.ycombinator.com