Flying and Driving after the September 11 Attacks
Safety in Numbers
The risks of flying and driving are influenced by different parameters. Whereas the risk of driving depends most strongly on the distance traveled, the risk of flying is primarily affected by the number of takeoffs and landings. A study carried out by Boeing indicates that out of 7,071 worldwide airline fatalities during the interval between 1991 and 2000, 95 percent happened either during takeoff and climb after takeoff, or during descent and landing. Conversely, only 5 percent of the fatalities resulted from accidents that occurred at cruising altitudes. Consequently, as we and others have pointed out before, the risk of flying depends mostly on the number of flight segments involved in the trip, not on the distance traveled.
In gathering the statistics for flying, we considered the scheduled domestic passenger operations of 10 major U.S. airlines: Alaska, America West, American, Continental, Delta, Northwest, Southwest, TWA, United and USAirways. (The commuter affiliates of these airlines were not included.) Because the number of airline fatalities varies greatly from year to year, we used the data compiled by the National Transportation Safety Board for a 10-year period from 1992 through 2001. To calculate the probability that a particular passenger would be killed on a nonstop (one-segment) flight, we divided the number of passengers killed during 1992–2001 (433, including the 232 aboard the four hijacked flights) by the product of the total number of nonstop segments (54,061,237) and the average number of passengers per nonstop segment (101.9). The resulting value is 78.6 X 10–9, or roughly eight in a hundred million.
The probability of a fatality on a one-stop (two-segment) flight can be calculated by combining the probabilities of a fatality on either segment. Roughly speaking, the probability of becoming a fatality on a two-segment flight is just two times the probability of becoming a fatality on a one-segment flight. (In actuality, because one must survive the first segment to become a fatality on the second, the full probability calculation is more complicated. But given the very low probabilities involved here, the simple approximation is quite accurate.) Similarly, the probability of a fatality on a three-segment flight is approximately equal to three times the probability for a single-segment flight, and so on.
When one decides between flying and driving, the latter option usually involves being the driver (as opposed to a passenger). Because the susceptibility to injury varies with the position of the occupant in the vehicle, we included only drivers in this analysis. Also, we tallied just cars, light trucks, vans and sport utility vehicles, ignoring heavy trucks, buses and motorcycles. Furthermore, we considered travel just on rural
interstate highways—the safest driving environment—because those constitute the most likely setting when one chooses to drive as an alternative to flying. To gauge the risks of such motoring, we used statistics from the year 2000, the most recent data available in detail.
To calculate the probability of fatality per kilometer of driving, we divided the number of driver fatalities on rural interstate highways in 2000 (1,511) by the estimated distance traveled on those roads by cars, light trucks, vans and SUVs (345 X 109 kilometers). The resulting value is
4.4 X 10–9, or about 4 in a billion per kilometer.
Armed with these two risk estimates, one for driving and the other for flying, we can specify something we call the indifference distance—the distance at which the two modes of travel are equally risky. For distances shorter than the
indifference distance, driving is safer; for distances longer than the indifference distance, flying is safer. The indifference distance for driving versus a nonstop flight can be calculated by dividing the risk of flying a nonstop segment
(78.6 X 10–9) by the risk of driving a kilometer (4.4 X 10–9). The result is 18 kilometers. For one-stop and two-stop flights, the indifference distances are 36 kilometers and 54 kilometers, respectively. Thus for any distance that is long enough for flying to be an option, driving even on the safest roads is more risky than flying with the major airlines.
Astute readers will note that our calculations do not include the trip to an airport (for flying) or the travel on local roads on the way to a rural interstate (for driving). True, we've overlooked this complication. But in many circumstances, the risks for these portions of the journey for the two modes of long-distance travel may be about the same. So we don't believe that our estimates of indifference distance would change all that much, even if such factors were fully accounted for.
Just how much safer is flying than driving? For an average-length nonstop flight (which works out to 1,157 kilometers), the risk of flying is just the 78.6 X 10–9 value derived above. The risk of driving those same 1,157 kilometers is 1,157 X 4.4 X 10–9, or 5,091 X 10–9. Dividing 5,091 by 78.6, we estimate that driving the length of a typical nonstop segment is approximately 65 times as risky as flying. Driving farther than 1,157 kilometers would be more than 65 times as risky; driving shorter than 1,157 kilometers, but longer than the 18-kilometer indifference distance, would be between 1 and 65 times as risky as nonstop flying (neglecting the drive to the airport and the travel on local roads on the way to the interstate).