Will Self-Driving Cars Have Less Than 1,000 Accidents Per Year?

This week, the self-driving car contingent has rallied, this time behind the argument that driverless cars will be safer than what we currently have. The event that kicked this off was a self-driving car that killed a pedestrian in Arizona. There really isn’t any evidence that self-driving cars will be dramatically safer than cars with drivers, other than repeating the claim over and over. But the real question is will driverless cars be safer than mass transit?

There’s a 2013 paper look at the fatality risks of various modes of transit (including cars with drivers; pdf). Here’s the key figure:


Bus transit is about 66 times safer, while urban rail is about 30 times safer (and that safety estimate includes things like physical assaults and ‘suicide by train’). Given around 30,000 automobile deaths per year, if everyone driving took mass transit instead, that would be around 1,000 deaths per year (probably a little less).

I’m skeptical that Silicon Valley, whose products require constant updating and patching, would be able to design a system–or more likely, multiple interacting systems–that could do as well. What one might call an edge case or a bug is also called a traffic fatality. If driveless cars can’t do significantly better than this*, maybe we should be spending money on making communities more mass transit friendly–aka ‘suburban renewal’–and relying on proven technology, rather than building and maintaining infrastructure to preserve a mass transit unfriendly building pattern.

Suburban sprawl is a powerful drug.

*In a trivial sense, they could: every car moves at five miles per hour. Not what I think most people have in mind though.

This entry was posted in Automobiles, Transportation. Bookmark the permalink.

1 Response to Will Self-Driving Cars Have Less Than 1,000 Accidents Per Year?

  1. AndrewD says:

    You might find this discussion interesting albeit with a UK flavour:- http://www.antipope.org/charlie/blog-static/2018/03/test-case.html

Comments are closed.