I think a lot of these don't suffer direct comparisons well. eg, the greatest risk of accidents to a bus, comes from other passenger vehicles where the bus is

I think a lot of these don't suffer direct comparisons well. eg, the greatest risk of accidents to a bus, comes from other passenger vehicles where the bus is more likely to win that particular dispute. The greatest risk of accidents to a train, comes from the trains own inertia, where the train is much more likely to lose. The greatest risk of accidents to an aircraft is the ground, in which the ground has an undisputed record. So I think it's safe to assume busses would be more likely to have accidents, but less likely to have fatalities. I think it actually gets worse if you try to normalise the results. There is greater risk in a car journey from LAX to JKF, than there is for an air journey along the same route. However, if I tried to commute by air to work (a 2km commute), the risk per km would rocket, as take-offs and landings are the most dangerous part of each journey, so with an atypical number of take-offs per km, I'd get an atypical risk per km. The wiki article does go on to point out the difference in usage of these statistics. per-journey does make more sense for air insurance, as the difference in risk between long-haul and short-haul isn't really significant - the take-off and landing are still the riskiest phases, there's just a larger period of boredom between them for long-haul. So 10 short flights vs 1 10x longer flight might normalise to the same hours & kms, but 10 short flights means 10 landings etc (10x the chances to miscalculate fuel, 10x the entries into & out of congested airspace, etc) But comparing per-journey between a car and a plane is rather meaningless, as they are on average very different journeys. commuting in a car isn't two phases of high risk separated by one phase of near-immunity.