Is flying statistically safer than driving?

(I am assuming we are comparing USA flights vs driving in USA)

I was reminded of this question today by this report by Forbes which rates professional driver as a less dangerous occupation (measured by deaths) than the occupation of pilot (both are in the top ten)

We have all heard the statement that flying is statistically safer. What evidence shows this?

If I were to do this statistical comparison, the way I would do it is by comparing the deaths per time spent traveling. That is, find estimates of the total amount of time spent flying or driving. For flying this would be the total number of passengers of all flights in a year multiplied by the total time of those flights. similarly I would get an estimate of the total amount time all Americans spent driving in that year. By comparing the number of deaths we would get an estimate of the probability of dying per minute spent driving vs flying. The time is important. Certainly, all other things being equal, one hour spent driving is more dangerous than five minutes spent driving. But is one hour spent driving definitely more dangerous than one hour spent flying?

Anyone know if this type of comparison or anything similar has been done?

Based on 2004 data analyzed here.

Per million hours:
General Aviation:


Commercial Aviation:

Per 100 million miles:
General Aviation:


Commercial Aviation:

Looks like the answer from a fatality standpoint is no. Flying (especially general aviation) is not safer than driving. At best, commercial aviation on a per mile traveled basis it is similar to driving. However your chances of being involved in a non-fatal accident are much greater while driving.

It looks that way if you accept this analysis, and the information being compared as a valid study.

The math and premise of this study (PDF) from the January 2003 issue of American Scientist make more sense to me. It accounts for 10 years of data instead of just one since aviation fatalities vary by a large amount from year to year.

The study also asserts that since 95% of plane crashes occur during takeoff or landing, the risk of flying depends almost entirely on the number of flights involved in a trip. The length of the trip isn’t significant; a long flight has about the same risk as a short flight. But with a car the risk of fatality depends upon how many miles are driven.

Bear in mind that driving involves a lot more non-fatal acidents than flying does.

Airline & former USAF safety guy …

Several thoughts …

When dealing with anything involving aviation, you need to be very careful to separate out general aviation (or military aviation) from commercial air carrier aviation. The difference in any safety fact you care to name is between 10 and 40 to one. And unless you’re in the habit of riding in Cessna’s, the general aviation statistics aren’t relevant to you. (GenAv encompasses a lot more than just light-plane pilots flitting about on weekends. But all the other categories of GenAv are even less accessible to the typical Doper and hence can be ignored for the OP’s purposes.)

Taking Hbns’s data at face value, he says

That’s 6.5 fatalities per million *aircraft *flight hours. If we assume 100 passengers per aircraft (which is low but the math is easy) we get .065 fatalities per million passenger flight hours.

Which changes commercial aviation from being about 10x as dangerous as driving per hour to 1/10th as dangerous per hour.

Naturally, cars aren’t always driven solo, so we’d need to apply the same correction to them to be fair & accurate. But the average occupancy of cars is (IMO) somewhere less than 2, and the actual average passenger load is somewhere North of 100, so the end result of more precise calcs will still be about as I outlined above.

As Crazyhorse’s post & cite say, distance is probably a better metric for driving. If you drive to work on a Monday in stop & go traffic & later make the same trip on Saturday afternoon, are you really exposed 2x or 3x the risk of death in heavy traffic at 20mph as you are on Saturday at highway speeds? Almost certainly not.

Boeing’s data is a bit muddled in that most of climb & descent are all-but as safe as cruise. But yes, the main risk the first & last few minutes of flight. A joking aphorism in the aviation biz has it that “You should avoid the edges of the air.” Stay in the middle and you’ll be fine. But bad things happen when you get near the edges.

The recently much-discussed Air France 447 accident was an example of getting too close to the top edge of the air (plus bunch of other problems), which resulted in them getting to the bottom edge of the air in a bad state. And as soon as they got to the bottom edge, everybody was killed. Stay in the middle, People!!!

Commercial piloting in air carrier aviation is not really much more dangerous than any other white-shirt-and-tie office job. It’s just that so few office workers are killed on the job that the few pilots who do get killed stand out statistically. There’s certainly more risk exposure, but not much actual difference in bad outcomes.

Once you add in the bush pilots and crop dusters & emergency medical helicopter guys and heli-loggers and such, now you’re talking about a job with real risk. Again the numbers are 20 to1 or worse than commercial air carrier statistics.

As I recall the fatality statistics for office & retail workers, they exclude deaths due to crime, i.e. retail store hold-ups and disgruntled office workers shooting up the place. I wonder how the numbers would work out if we added those in? Is USPS desk clerk in fact a highly dangerous occupation since they’re exposed to both those crimes?

Late add: I followed the link in Hbns’s post. It has a good discussion of the various statistics and which numbers are really meaningful & which aren’t. I don’t necessarily endorse all the author’s thinking, but it’s a good intro for the OP into the topic. Note the main thrust of the article is determining risk to GenAv pilots & riders, not commercial passengers.

You have to be careful about comparing two situations and assuming both can be considered random. Flying on a commercial airline is a pretty random event, although some airports are more dangerous than others. But driving a car involves different risk levels, depending on the skills and driving habits of the driver.

I read somewhere [paraphrased] ‘You are more likely to crash in a car than in an airplane; but in an airplane crash you are more likely to die than in a car crash.’

The OP did say they wanted to compare by the amount of time spent traveling. Whether that is the best measure of airline versus automobile safety I would thing is more a topic for great debates. A per minute fatality rate is what the OP mentioned. I thought the statistics I linked were pretty well presented for the OP to generate such a result.
I have no reason to doubt the analysis, the author does a fair job of questioning the results and pointing out the flaws.

Airline data was averaged from a ten year sample similar to the Sivak study.


It is completely sensible to me that takeoff and landing are the riskiest times during an airline flight, but several things about that segment of the paper are unclear to me. I will have to dig deeper into it.

I am not certain if Sivak is comparing passenger trips to passenger trips for autos or simply flights to automobile trips. One plane can account for a couple hundred passenger trips, while one auto typically accounts for around 2.

Secondly, I am not certain why Sivak has bounded the bottom end?

Certainly there are few if any domestic commercial flights of anywhere near 18km. What purpose does this serve? Does it rule out a lot of auto accidents? It is commonly said that you are more likely to die in an auto accident close to home (whether that is true or not is a whole other topic so lets just assume it doesn’t come into play one way or the other).

I’ll read the rest of the report and see if it answers my questions further on, or if any of the citations shed light on them.

Found the original article, Flying and Driving after the September 11 Attacks

Not quite understanding how they arrived at the 433 passengers killed. When I look here I get a total of 1149 fatalities.

I guess I am just too easily confused by the methodology. I’m not expecting commercial aviation to be more dangerous than driving. Far from it, but something about the Sivak and Flannagan calculations just seems off. Can’t put my finger on it, maybe someone else with a better mind for statistical analysis can tell me I am just being paranoid.

Missed the edit window.

I just noticed they sampled just 10 airlines for their data set. So throw out 252 fatalities from my total, leaving 897.
(Alaska Airlines 83, Valujet 105, and American Eagle 64)

Anyone know where they might have gotten the fight segment totals from?

Given Sivak’s POV that per-mile is the correct way to score auto accident risk & per-flight is the correct way to score airplane accident risk I interpret him to mean this:

Your exposure to a fatal accident on any single airline flight is the same as your exposure from driving 18km. If you drive farther (all at once or spread out over several days) you’ve been exposed to more risk than one airline flight. Said another way, for every 18km you drive, you take the risk equivalent to one airline flight.

For a typical American who drives or rides ~12K miles or ~18K Km per year, they’d need to fly 1,000 airline flights per year before their risk of dying on the plane gets as big as, or bigger than, their risk from their ordinary driving.

Quoth LSLGuy:

Like, they got too close to the Sun, so the wax holding their wings together melted?

Very metaphorically, yes.

But poetically, no. Icarus was doomed for his motivation: to fly up to the realm of the gods. The sun & the wax was just a plot device to deliver his comeuppance.

The Air France guys had no such motivation. Conversely, there was an accident involving a Bombardier CRJ a few years ago which some lay folks over-interpreted as re-enacting the parable of Icarus.

They flew as high as the jet was certified to fly (in other words, a completely normal & legit maneuver) just because they had the rare opportunity to do so. And something failed, they handled it poorly, and ended up being killed for their errors. Which would never have happened but for the (arguable) hubris of climbing so high.