←back to thread

307 points MBCook | 1 comments | | HN request time: 0s | source
Show context
jiveturkey ◴[] No.42151256[source]
Seems flawed. Tesla Model Y was the best selling model worldwide for 2023. (I think #3 if limited to US.) The study only covers 2017-2022, but we can infer that for the entire Brand, Teslas sold quite well over at least that latter part of that period.

Now if there are more Teslas on the road vs other vehicles (note they excluded car model years earlier than 2017, another fatal (heh) flaw in the study), it makes sense they would have more fatalities.

So this should be normalized "per capita" to vehicle counts if we want to extract any brand-related causality, in the same way as the data is already normalized to miles driven.

I enjoy hating on Tesla as much as the next person, but come on.

replies(2): >>42151484 #>>42151584 #
FactKnower69 ◴[] No.42151584[source]
If you had clicked through to the article before writing your comment, you would know that the stat being compared is "fatal crash rate per billion miles driven", and that the fatal crash rate for Teslas is 2.0x the national average
replies(1): >>42179092 #
1. jiveturkey ◴[] No.42179092[source]
If you had read my comment before replying, you would know that I recognized and even stipulated that the report was already normalized to miles driven.

I am suggesting that per mile driven and per vehicle brand on the road are not the same thing. Now that I think about it, miles driven per brand would be an interesting data point.

Just as an example of why per-brand might be useful, let's imagine that teslas are driven half the miles of the average for all other brands. so if that were the case and if the number of fatalities were equal, then that would result in tesla having 2x the fatality rate per mile. but at the same time, this doesn't let us appreciate the amount of time tesla spends on the road vs other brands. so miles might be half but time spent might be closer to 1 (equal time). and maybe that is because teslas are concentrated in urban-suburban interface environments in high congestion areas, whereas other brands are spread out across the country where for the same amount of time spent, twice as many miles are covered and congestion is less.

Data like that is needed to properly evaluate this result. And while I asked for per-vehicle-on-the-road, and my example didn't match that, you can still see how the simple per-mile-driven is insufficient.

They also could have noted the demographics of the drivers -- male/female and years of driving experience, things like that.

So again, I find the implications of this article to be very unfair.