turchin comments on Open thread, Jul. 18 - Jul. 24, 2016 - Less Wrong
You are viewing a comment permalink. View the original post to see all comments and the full post content.
You are viewing a comment permalink. View the original post to see all comments and the full post content.
Comments (123)
Why car safety is not advertise as its main quality?
Tesla suffered its first fatal accident in self-driving mode after driving 130 million ml, while the average mileage between fatal accidents in the US is 90 million ml. This is presented as evidence of the safety of Tesla.
However, the safety of cars of different classes of security has 1000 times difference.
Kia Rio has one fatal accident at about 10 million ml, and Subaru Legacy has less than one per billion km (in fact zero).
The latest data on the risks of different car models is here: http://www.iihs.org/iihs/topics/driver-death-rates
I did some calculations based on presented data and typical car driving 20 000 ml a year assumption.
Dodge Caravan has the risk of a fatal accident on a 1 to 10 billion miles. (I saw it in similar sheet before.)
These cars are 3-5 times more expensive than the Kia, and due to the greater mass, strength and quality provides great security.
I think part of the promise of a BMW is that it's a safe car. However rightly or wrongly 5-star safety ratings is the common measurement of safety and not fatal accidents per mile.
But isn't the fatal accident per mile is better predictive measure?
It might be or it might not be because observational studies don't tend to be good at analysing causation.
In general marketing executives don't focus on using metrics that are good predictive measures. They also don't want to advertise less fatalities but more safety.
I don't think so; different types of car are bought by different people and driven differently. E.g. a person who buys a Lamborghini or Ferrari probably likes to drive fast and show off; a person who buys a Volvo probably drives a lot more carefully.