An important but usually neglected statistic in evaluating the relative danger of different modes of transportation is the risk of death or injury per
trip, rather than per mile or per year. On a per mile basis cars are as much as 8 times safer than bicycles (
http://cycleseven.org/dangerous-cycling-statistics), but as some of you have pointed out, it's foolish to compare cars and bicycles directly. Similarly, the per-mile safety of air travel is a little bit misleading because almost all airline accidents occur at takeoff or landing -- so the risk of death is actually determined by the number of flights, not by how far they go. This leads to interesting number games where one can theoretically calculate a distance that is safer to drive than to fly. Some people have done this and come up with numbers as high as 600 miles, depending on what roads are used (
http://ipmall.info/risk/vol4/winter/halperin.htm). That source does a very good job of teasing apart the different ways to measure transportation risk.
Ultimately, the point of travel is not to stack up as many thousands or millions of miles as possible. It's to get to places. So even if biking really is eight times more dangerous than driving on a per mile basis, it's safer to ride to a store 3 miles away than drive to one 30 miles away. For a person deciding whether to cycle or drive to work, it's likely that an identical distance ridden is more dangerous than the same distance driven. But it probably still isn't
very dangerous, and moreover, if you make a point of living at a distance where bicycle commuting is possible, your commute mileage will likely be far less and so the relative risk
per trip is comparable or even lower when bicycling. I'm sure that on a per-mile basis walking is much more dangerous than driving -- and maybe even more dangerous than biking -- but no one avoids it for safety reasons. These number of aggregate traffic fatalities is high because we drive 3 trillion miles per year which makes the risk look very large but the risk to individual drivers, bikers, walkers, etc. is very low.
Furthermore, there is good reason to believe that the risk of cycling decreases as the population of cyclists increases (
http://www.sirbikesalot.com/uploads_images/2010/2010-04-30_bicycle_chart_01.jpg). Presumably a number of factors are at play, but one of them is likely just that the normalization of bicycling makes drivers more apt to notice cyclists, pass safely, and not get that insane psycho driver road rage fury that seems to be the norm when people notice one GOD DAMN BICYCLIST SLOWING UP TRAFFIC ARGH IM PISSED. It's a lot harder to feel that way when there are cyclists everywhere, little kids riding to school, an old guy in a reclinocycle etc.
For the record, the safest per-trip way to travel by a very wide margin is public buses (by an entire order of magnitude iirc). But mysteriously, you don't see many folks turning in their safe-n-secure Lexus gx460s for a bus pass. The reason is of course that transportation safety, from the perspective of an individual, is very high in all modes. So the idea of choosing one transportation mode over another for reasons of safety is almost laughable.