I don't think that's right. We shouldn't think of death in terms of "deaths per mile" because human costs aren't amortized per mile. Every death is a death. Killing someone then driving in circles shouldn't make you a less dangerous driver.
I don't think it should be thought of as "deaths per mile" but rather "deaths per car passing by" which allows one to compare the relative danger of driving in different locations.
This implies that the only road users are cars and that vehicle volume is fixed. Factors like vehicle speed, modal mix, lane widths, and parking geometry all affect the safety profile of a street. "Cars passing by" is too lossy a metric to be relevant. Decreasing vehicle volume is always an option.
Imagine a town that has 100 cars in it and 10 traffic deaths per year vs a town that has 10,000 cars in it and also 10 traffic deaths per year. Both towns have the same population. Without any sort of normalization the towns look the same, but clearly very different things are going on in those scenarios and a per mile or per car or whatnot will help you identify that. Obviously it doesn't give you any sort of root cause but it is an indicator that a deeper investigation is needed.
Then just note the vehicle volume and don't divide your deaths by volume? When you divide your deaths by volume you're implying that there's some value in a metric that amortizes deaths by vehicle volumes. Automobile volume is just one of several variables and simplifying this equation is why the US has a much higher traffic crash and fatality rate than any other developed country in the world.
What is the point of "noting" and not dividing? Sure, you can normalize by all sorts of different metrics. That doesn't mean you shouldn't normalize at all.
Nonsense. Mostly because nobody actually "kills someone and then drives in circles".
Human life being valuable does not change that certain activities carry a certain amount of risk, and so of course the more you do that activity the more likelihood the risk manifests. It is valuable information to know if one area has a hugely disproportionate number of traffic deaths compared to overall traffic metrics.
e.g. let's say I'm picking between two cities for my next job. Every day I will drive 25 miles to work, regardless of which city I choose. I look at the stats and see that in X city, there is 1 traffic death per 10,000,000 miles driven. In Y city, there is 1 traffic death per 100,000 miles. Therefore if I live in Y city I am 100x more likely to be involved in a fatal traffic accident. That is very relevant info to me and doesn't cheapen the value of human life at all.
Of course "driving in circles" is an exaggeration, but I think they're trying to say that if a place is structured to encourage lots of mileage (maybe there are things with relatively little utility, like surface parking, causing destinations to be very spread out; maybe there's no public transit option so more people are driving), then there will be more fatalities at a given per-mile rate. If you optimize _purely_ for the per-mile rate you might be missing out on ways to reduce fatalities by reducing mileage.
By not designing everything around car drivers, it's possible to have such things as "15-minute cities" where the aim is to have amenities withing a 15 minute walk or cycle. That reduces traffic delays as there'll be more people choosing to walk/cycle, plus the car journeys will be much shorter. They also reduce traffic collisions ("accidents" is a loaded term and should not be used for RTCs as it implies no fault), so it's really a win-win.
