Traffic safety wonks often congratulate themselves and Detroit (not always in that order) on the splendid safety record of American highways. They almost always cite fatalities per mile driven as the definitive measure of safety. But is "fatality risk per mile" the only, or the most meaningful, measure of public safety? It has been accepted for over 50 years as the definitive metric, but imho the metric itself is carcentric and oversimplified.
Risk per mile is a useful figure for the spindocs :-) Americans now drive (collectively) 2.5 trillion miles per year -- this figure has been climbing alarmingly for the last 50 years. Obviously there are more people in America every year; but the increase in miles driven is not tied to population growth alone. Each "average person" person is driving more miles per year.
Risk per mile for automobile passengers is impressively low, thanks to endless technological innovations making it "safe" for unskilled and inattentive people to operate high-speed heavy machinery (cars). But "unnecessary miles driven" (a figure not recognized by most safety experts) exerts a countervailing force, by constantly increasing exposure.
The "risk per mile" figure will always look better, the faster the mode of transport and therefore the more miles any human being can experience in a lifetime. We can construct a reductio ad absurdum which illustrates this problem: Big Dog Space Cruises, Inc., offers you a ticket on a round trip to Alpha Centauri in their new FTL spaceship; you ask about their safety record. Say they have made 2 round trips so far (about 16 light years) and on the first trip there were no problems; but on the second trip, the astronaut died. They can tell you with honest pride that they have had only one fatality in 16 light years (almost 100 trillion miles) of travel -- their risk per mile is astonishingly low, far safer than cars, planes, biking, horseback, walking, etc.
This way of measuring risk always makes cars look better than biking or walking (and planes look better than cars). But in the real world, it is "risk per trip", or "risk per hour" or "risk per year of this lifestyle" which mean more to people. How willing would you or I be, for example, to board that hypothetical space craft?
To put it yet another way, people who prefer to walk or bike obviously change their travel habits and make trips within a smaller radius from their home. For example, they do not choose to live 40 miles from their workplace. Ownership of a car encourages people to take more and longer trips. If travelling in a car is 100 times safer per mile (I just made up this figure, btw, it's not real), but the pedestrian travels 100 times fewer miles per year than her neighbour who drives a car, then neither is safer than the other.
Some studies have suggested that people tend to seek a certain average duration of trip (i.e. an acceptable commute to work hovers around 1 hour) regardless of their travel mode. This seems to indicate that people think and plan in hours, not miles: the cyclist, the ped, and the driver will, all other things being equal, make trips of about the same duration in their daily business. There are always, of course, people on the far edges of the curve (such as a fellow I met in TX who commutes 100 miles each way to work every day, or a person who lives right above the shop where they work).
Risk per mile metrics also do not take into account the public health impact of a sedentary vs an active means of transport. Various medical commentators have noted that cycling, for example (but regular walking also) is beneficial to general health. Some large-sample studies suggest that cyclists live at least 2 years, and possibly as much as 10 years, longer than non-cyclists. Anyway, "risk/mile" stats do not reflect the health risk of lack of exercise (for a car centred life) or the benefit of regular mild exercise (for a bike or ped). Nor does this metric reflect the "fallout" effect, i.e. not only do drivers incur a general health risk from lack of regular exertion, but carcentric planning & an excess of cars on the road spoils both urban and rural amenity and discourages people from walking and biking, thus "making them into drivers".
Hillman has pointed out that risk/mile statistics make no account of whether the risk is to self or to others, an important ethical and social distinction. For a truck driver, the chance of surviving an impact in which the truck kills another person is over 90 percent. For a car driver it's about 50 pct (i.e. you are as likely to kill the other person as to kill yourself). For peds and bikes the likelihood of killing another person by your means of transport, regardless of your own degree of recklessness, is very low, even lower for peds than for bikes. So in terms of who's "safest" for those around them, the pedestrian wins hands-down.
Risk per mile stats make no measure of how much risk is imposed on others (which should carry a heavier ethical weight) and how much is imposed only on the traveller. In other public health issues we make a much stronger distinction (as in the case of public smoking and the risks of 2nd hand smoke) between self- imposed risk and inflicted risk.
Hillman also points out that risk/mile metrics lead to an obssession with reduction of the collision count per mile driven (this is exactly the metric used by the SCPD OTS grant) -- but this is not necessarily associated with improved safety. If peds and bikes avoid roads that have been made dangerous for them, or cease to bike and walk altogether, then this avoidance of danger is not the same thing as increased safety -- even though it may result in lower injury/fatality stats and make risk/mile look "better".
Lifetime risk and hourly risk seem to me essential metrics which should at least augment, or preferably replace risk/mile in public discourse. And "risk to others" should be weighted more heavily than "risk to self" when we discuss safety issues. IMHO.