> This is a strange metric to use though, because you're unlikely to walk (say) 20 miles to work, while you might drive that distance. Unilaterally choosing to walk or bike to work rather than drive in such a situation also typically means moving closer to work.
If the decision you're trying to make right now is how to get to work, then holding the mileage constant is the right way to look at it. If you're trying to decide where to live then, yes, the analysis gets a lot more complex.
Well of course where you live plays into what mode of transportation you'll choose, and the mode of transportation you'd prefer will often drive where you live. For example, I live a long walk from work (or a short bike trip, which is what I do 99% of the time; walking is mostly reserved for nice lazy evenings or blizzards). Yes, the rent is more expensive, but I'm also saving a lot of money on not having to buy/maintain/insure/park a car. And my daily commute is only around 4 miles round trip each day.
So it does make sense to evaluate commuting mortality on a per trip basis, e.g. "what is the mortality associated with day of going to the office". The point of a trip isn't to go X miles, it's to go accomplish Y thing, such as going to work, running an errand, etc. Sparse suburbs shouldn't get "credit" and have their mortality normalized away on distance for everything being spread out. The sparseness itself is a problem in that it's requiring such long mileage for such trivial trips as "going to work" or "getting groceries".
Perhaps, but let me throw in an additional consideration. Per-passenger-mile statistics treat a mile driven within a city (which involves complex interactions with pedestrians/cyclists/other drivers) the same as a mile driven on a freeway (relatively uncomplicated).
If you happen to be in a position right now to be able to walk to work at all, most of your driving is probably intra-city, because you probably both live and work in the same city. However, I suspect per-passenger-mile stats weigh freeway driving more heavily than city driving: on a typical car commute into a city from the suburbs, most of your mileage is not within the city proper, but on the freeways surrounding it.
If the decision you're trying to make right now is how to get to work, then holding the mileage constant is the right way to look at it. If you're trying to decide where to live then, yes, the analysis gets a lot more complex.