Is GA Flying Safer Than Driving?

The airline industry likes to point out that, despite some passengers' apprehension, flying in an airliner is actually safer than driving a car. Some pilots of small planes 'recall' this argument, and mistakenly apply it to General Aviation (i.e. flying that isn't the scheduled airlines or military, typically "little planes" and "biz jets").

Is it true? Is our "little plane" flying safer than driving a car?

Raw stats

For GA accident stats, we'll use the 2005 Nall Report (available at http://www.aopa.org/asf/publications/nall.html), which reports on accidents from calendar year 2004.

For driving, we'll use 2004 statistics from the National Highway Traffic Safety Administration (NHTSA), available here.

According to the Nall report, in 2004:

  • there were 1413 GA fixed-wing accidents
  • of these accidents, 290 involved fatalities
  • these fatal accidents accounted for 510 fatalities
  • the FAA estimates 25.9M GA flight hours
According to the NHTSA, in 2004:
  • there were 6.1M accidents
  • 1.9M involved injuries
  • 38253 involved fatalities
  • there were 42636 fatalities in motor vehicle accidents, breaking down to:
    • 33134 "occupants" (car/truck drivers and passengers)
    • 4008 "motorcycle riders"
    • 5494 "non-occupants" (e.g. pedestrians)
  • an estimated 2.9T vehicle miles traveled
  • an estimated 10.0B motorcycle vehicle miles traveled.
So in 2004, 510 people died in 290 GA accidents, while 42636 died in 6.1M motor vehicle accidents. Of course, to determine risk level, we have to convert this into a rate, as there was a whole lot more driving done in 2004 than there was GA plane flying.

Comparison

Let's compare the rate of fatal accidents to the amount of flying/driving done.

For GA, there were 11.2 fatal accidents and 19.7 fatalities per million hours of flying.

For motor vehicles, there were 1.32 fatal accidents and 1.47 fatalities per 100M miles.

Note, though, that about 13% of these fatalities were pedestrians, and so weren't actually engaged in driving. Also note that motorcycles have a much higher fatality rate than cars/trucks, namely 38.8 fatalities per 100M miles, over 25 times higher than the average, so these numbers are a bit inflated when it comes to determining the risk of an occupant of a car or truck.

Of course, the next hurdle we face is that our flying accident rate is based on the number of hours, while the driving accident rate is based on miles driven. To directly compare them, we have to either pick an average speed for GA aircraft and compare by hours, or else pick an average speed for cars, and compare by miles. We'll try both of these approaches.

GA flying covers small training aircraft capable of cruising at 100mph, and business jets capable of cruising at several hundred miles per hour, so choosing an average cruise speed is difficult, but for the sake of argument, we'll choose 150mph. This gives us a comparison of:

  • GA: 7.46 fatal accidents and 13.1 fatalities per 100M miles
  • driving: 1.32 fatal accidents and 1.47 fatalities per 100M miles
So when compared on a mile to mile basis, flying has 5.6 times as many fatal accidents, and 8.9 times as many fatalities (these number would be even worse for flying if we took out motorcyle and pedestrian fatalities).

How about if we compare on an "hour to hour" basis? That requires an assumption of an average speed for autos. We'll choose 40mph. This leads to the following numbers:

  • GA: 11.2 fatal accidents and 19.7 fatalities per million hours
  • driving: .528 fatal accidents and .588 fatalities per million hours
On this basis, flying has 21 times the number of fatal accidents and 33.5 times the number of fatalities per hour of operation.

How do the GA numbers compare to commercial airline travel?

The NTSB provides accident statistics for "part 121" air carriers (scheduled airline travel) at http://www.ntsb.gov/Publictn/2001/SR0101.pdf Airline fatalities are concentrated into a small number of accidents. It's not uncommon for an entire year to pass with 0 airline fatalities, so it's not really possible to generate meaningful statistics by looking at just one year. Instead, we'll sum up all the fatalities and estimated flight hours and miles flown for US carriers over the ten years from 1991-2000. Since we're evaluating the risk to passengers, we'll only consider the fatalities of those aboard the plane (and not victims on the ground, the airliner equivalent to "pedestrians").

The results are:

  • 938 fatalities in 31 fatal accidents
  • 145 million flight hours
  • 59.7 billion miles flown
If we convert these into rates similar to the ones provided above for GA and cars, we get:
  • .2 fatal accidents and 6.5 fatalities per million flight hours
  • .05 fatal accidents and 1.57 fatalities per 100 million miles flown
Note that given that these fatalities result from a very small number of incidents, these statistics vary tremendously from year to year. If we summed the accident data for just the four years 1997-2000, the fatality rate would be 1/4 of the 10-year average shown above.

What about commercial vehicle travel?

The GA flying data above excludes results from airline travel. The safety record for airline travel is so far superior to GA, that to combine the stats together would skew the results for GA travel, boosting the appearance of GA's safety record.

The motor vehicle stats above, however, do include commercial motor vehicle travel. Is it possible that commercial motor vehicle travel is similarly skewing the results for motor vehicles, making cars seem safer than they really are?

The FHWA results for 2004 can be found summarized here: http://www.fhwa.dot.gov/policy/ohim/hs04/htm/vm1.htm

Out of 2.96 trillion vehicle miles driven, .226 trillion of these miles were by "SINGLE-UNIT 2-AXLE 6-TIRE OR MORE TRUCKS" and "COMBINATION TRUCKS". This would include all tractor trailers, buses, etc.

First, note that this represents 7% of all vehicle miles traveled. So if commercial truck and bus travel was unfairly improving the motor vehicle safety record, the maximal possible impact would be if these vehicles had 0 fatal accidents in 2004, in which case the impact couldn't exceed 7% of the overall fatal accident rate.

So in the case of motor vehicles, the percentage of travel that could be described as "commercial" is such a small percentage of the whole, that it's really not possible for it to significantly boost the motor vehicle stats, assuming commercial motor vehicle travel was even shown to have a better safety record than cars. On the other hand, if the truck accident rate was much higher than the overall average, it could be unfairly raising the motor vehicle accident stats, making cars appear to be more dangerous.

The NHTSA data lists 761 occupants killed in "Large Trucks" and 556 in "Other Vehicles" (which includes buses and farm and construction equipment). Together, these account for about 4% of overall motor vehicle fatalities.

But while the NHTSA report lists 761 occupants killed in Large Trucks, in accidents involving these vehicles, 4006 "other vehicle occupants" were killed, as well as 423 "non-occupants" (pedestrians, presumably).

So accidents involving large trucks account for 12.6% of fatalities, while only accounting for 7% of vehicle miles traveled. This suggests that commercial motor vehicle travel is actually worsening the stats for cars slightly. However, in accidents involving trucks, the casualties were mostly the occupants of other vehicles and pedestrians, so commercial motor vehicles are, in themselves, a significant risk to car travelers, but the numbers just aren't large enough to make a significant difference.

Conclusion

Choosing "mile to mile" as the more appropriate comparison for differing modes of transportation (and overlooking that small planes often takeoff and land at the same airport, without ever really "going anywhere"), let's review the fatality rates:
  • driving: 1.32 fatal accidents and 1.47 fatalities per 100 million miles
  • airlines: .05 fatal accidents and 1.57 fatalities per 100 million miles
  • GA: 7.46 fatal accidents and 13.1 fatalities per 100 million miles
So mile per mile, GA flying has about 5 times as many fatal accidents, and 9 times as many fatalities, as compared to travel by motor vehicle. The airlines have about the same fatality rate as driving, but a much lower fatal accident rate (by virtue of a large number of fatalities per accident).

But what's the important stat?

The numbers above provide both the rate of fatal accidents, i.e. accidents in which there are 1 or more fatalities, and also the rate of occurence of fatalities. When we consider these numbers, what we're typically hoping to come to terms with is something like "what's the relative risk of me, a single passenger, taking a trip via each of these modes of transportation?"

Neither of these stats perfectly encapsulates what we're looking for. The best approximation to this would be the product of:

  1. the rate of fatal accidents (i.e. how likely is it that somebody will die on this trip)
  2. in fatal accidents, what percentage of passengers are killed (i.e. what are the odds that I'll be one of those fatalities?)
For 1) above, we do have the statistics, but how do we gauge 2)? We know how many accidents there are, and how many fatalities, so we can calculate the average number of fatalites per fatal accident. This yields:
  • driving: 1.1 fatalities per fatal accident
  • GA: 1.7 fatalities per fatal accident
  • airlines: 31 fatalities per fatal accident
The above statistic depends on 2 factors:
  • average number of passengers on-board
  • percentage of passengers killed in fatal accident
Airline accidents have a far greater number of fatalities per accident for an obvious reason, namely because they carry a larger number of passengers per trip.

It may be possible to get a better idea of risk to a single passenger by comparing fatalities per fatal accident to average number of passengers on-board. This assumes that there's no correlation between fatal accident rate and passenger counts, which may not be true (e.g. is the probability of fatal accident equal for a large jet as for the smallest planes still used for scheduled airline service?)

Caveats

What are the assumptions that went into this?
  • government estimates of GA flying hours and miles driven
  • a "guessed at" average GA airspeed of 150mph
  • a "guessed at" average car speed of 40mph
These rate-based numbers depend on estimates of miles traveled and hours flown. The hours flown data come from the FAA's "GA Survey", and are derived from responses to a voluntary survey sent to a sample of pilots. All the usual caveats of survey and sampled data apply.

The miles driven data are compiled by the FHWA from data provided from individual states. I'm not sure what methodology is used to generate these numbers.

What else?

  • we're including motorcycle and pedestrian fatalities in the motor vehicle risk, which inflates the apparent risk of driving in a car
  • we're also including "pedestrian" fatalities in the GA stats. We don't know how big these are, but suspect that they're insignificant.
Harry Mantakos / harry@meretrix.com