Signing up could earn you gear and it helps to keep offensive content off of our site.
Visit our Flying shop
I have seen it claimed by a manufacturer of airliners, along with the media at large and many pilots, that flying is safer than driving.
Maybe they can say so because the two modes are not directly comparable, so it's hard to prove they're wrong. But I went and messed with some statistics, hoping to come to a conclusion myself, and I have to conclude that flying is not safer than driving.
I take no joy in this conclusion, because I love to fly. (I'm a brand new private pilot with about 70 hours. Just passed my flight test two days ago.)
But when I ran some numbers, I was forced to conclude that however you measure it,* flying is not safer than driving. This is especially true with general aviation. It important to remember this, not kid ourselves, and never get complacent about saftey.
The details for those who want more
The reason for the little * above is that there is one exception that the airlines/airliner manufacturers use that I consider to be statistical cheating: They measure fatalities/accidents/etc. per million passenger miles, a passenger mile being the multiplication of miles x passengers. Basically, they get 100-400 times the credit for each safe landing based on how many people are on board. The problem with this, from a statistical point of view, is that passenger miles are not accidents drivers. Hours, miles flown (i.e. airplane miles, not passenger miles), takeoffs, landing, etc. are all things that might drive accidents, but not how many people are on board. Anyway, all this is neither here nor there, because in general aviation, even using the misleading accidents per million passenger miles measurement would not make flying safer than driving. It only does so for the airlines.
Here are the conclusions I reached based on three ways of measuring it. I can provide the underlying data and references to anyone who may be interested:
When I considered fatalities per 100 thousand hours flown or driven (using all USA numbers), I got 2.1 for general aviation, 1.25 for air carriers, and around 0.06 for motor vehicles (i.e. driving). (I did have to make an assumption to try to convert driving miles to driving hours, since only miles are known. For all reasonable values, the order remains the same, i.e. with driving having the least fatalities.) This makes general aviation some 35 times more fatal than driving using this measure, and airlines 21 times more fatal.
When I considered fatalities per 100 million vehicle or aircraft miles, I got 23.32 for general aviation, 3.06 for air carriers, and 1.53 for motor vehicles. (Again, some assumptions were necessary to make things comparable, in this case an assumption about average velocity to convert general aviation hours to miles, since miles flown are not known in GA. However, as was the case in my assumption above, for all reasonable assumed values the order does not change: GA, then air carriers, then driving.) Using this measure general aviation is only 15 times more fatal than driving, and the airlines only twice as fatal as driving. But driving still comes out on top for safety (or at least survivability).
Only when you consider the funky fatalities per million passenger miles statistic do you get aviation suddenly looking better: GA was 0.05 over the period I examined, motor vehicles was 0.01, and air carriers was 0.0003. Here is where the airlines shine, when you add in a figure (number of people on board) that logically has no correlation with accidents.
So, as sorry as I am to disillusion anyone (okay, I'm not sorry; I like myth-busting), I think we should acknowledge that aviation, while not particularly hazardous, is still more dangerous than driving. When you think back on what it took for you to learn to fly, find the airport, land in a cross-wind, recover from a bounce, do a forced approach, etc., this should make intuitive sense anyway.
Let's continue to fly and love it, but let's not kid ourselves by minimizing the risks. They are real.
Chad Conrad Calgary, Alberta Click here to learn about the Calgary Flying Club
By the way, just to poke a little hole in my own argument, one could argue that if I don't want to given extra statistical "credit" for the number of people on board for a safe landing, I shouldn't do so for an accident either, meaning that I shouldn't consider the number of fatalities per thousand hours or million miles but rather the number of fatal incidents. I didn't do so, if I remember right, because those numbers were even harder to come by in a convenient and/or comparable form. If I were to do so it may make airline travel come out a bit better, but it wouldn't help general aviation much, and that's what most of us are interested in on this forum anyway.
The other potential analytical problem I can think of is that perhaps thousands of hours flown or millions of miles flown isn't even the right driver (denominator) to consider anyway. Perhaps the right one is the number of take-offs and landings, or even something else. Again, I went with the easiest to find numbers. For driving it is generally agreed that the right denominator is the number of miles driven, however. This highlights one of the biggest pitfalls in trying to compare different modes of transportation, and that is that they will kill you in entirely different ways that make it very hard to compare which is more dangerous. (You would almost have to run the numbers for each trip you were considering... "Okay, I need to get from Calgary to Seattle. Driving would be this many miles, which translates into X risk of dying. Flying would be this many hours/this many takeoffs and landings/whatever, which translates into Y risk of dying. So for this trip I should drive/fly." This of course, would be a bit ridiculous to do each time, and we may decide that it is worth some extra risk to fly anyway.)
One could spend a lot more time than I did analyzing the numbers, but I think it's generally fair to say that to automatically assume that air travel is safer than driving may be hanging on to a wishful, if comforting, myth. (In fact, I think this is almost incontestable if you exclude the airlines and consider only general aviation.)
Hummm. You present some interesting information here. I really did not feel GA was safer than driving but frankly never saw the math.
I've also wondered about the number of accidents in autos vs GA, not taking into account injuries.
Yup, the math is sobering. It won't stop me from flying, though!
Re. non-injury accidents (or even non-fatal injury accidents), I didn't investigate that too much, but I would guess that aviations accidents tend to be more fatal. So, it may well be the case that your odds of being involved in an accident are greater in a car, it's just that such incidents don't carry as much risk of killing you.
From my own experience, I've been in two car accidents in my life, one of them a high-speed highway incident (and both of them the other driver's fault) but obviously neither killed me...
Maybe aviation accidents will become less fatal with the adoption of such technologies as the ballistic parachute, though. (See this topic on this forum.)
Just a comment on a fact that has been brought up several times in this forum: One reason why the statistics can be so hard to come by is that the two modes of transportation don't always kill you in the same ways. The reason why I bring this up is that I have a friend who has been in three serious car accidents of which I am aware, having been the driver and sole occupant of the only involved vehicle in all three cases. In two of these cases, he statistically should've died, and in the other one, he statsitically should've been severely injured. Thankfully, he was uninjured in all three, except for one of them, in which he incurred injuries while climbing out of the roadside blackberry briars in order to get back to the roadway. It's difficult to say whether two of the accidents would've been comparable to respective kinds of aviation accidents, but the third, which is one of the two that he would've "normally" been killed in is the one I will comment on. On that morning, while he was driving to work with the rear windows rolled down, he heard some papers rustling in the back seat. Not wanting the wind to blow any of these papers out the window, he reached back to grab them. Momentarily distracted by this action, he did not notice an upcoming curve in the road. Later, when asked about the circumstances surrounding the accident, he replied, "I think it would be pretty safe to say that the road turned and I didn't." This highlights a primary danger of driving: when you do not stay on the road, there's a lot of stuff out there for you to hit.
Sorry about the length af that last post, I usually try to keep them shorter.
If it's too long, which is usually less of a real problem than the opposite, then we don't have to finish reading it, do we?
Funny, this crazy website says I'm logged in as C. Craig Morris, who I am, but then lists me as some wimp "anonymous" in my posts.
As I've not seen it said yet, Congrats on the new ticket!
Next, I happen to believe you're thinking well, at least conceptually. I too believe study in risk is a vital role of a pilot. I might suggest you continue your thought process. Go decide what are the major things that get GA pilots killed and create a process to protect yourself from same. Nobody says you have to be in the same risk pool as everybody else.
As an example, I did some unusual attitute training and some advanced spin recovery training very early on. I'd suggest same. I also have my own fuel reserve requirements, and I obtained a lot more IFR training then the normal VFR intro before I started XC travel w/my family.
Just some thoughts, but again, I like thinking about risk, so thought to comment.
As an aside on your data as a whole, I disagree with an analysis based on hours in transit, and support analysis based on miles traveled. This as I never drive or fly for a fixed amount of time, but to a desired destination. That implies, to me, the comparative basis is not the medium of time, but distance. Hence risk study in miles...
FWIW, when I did my version of your study a few years ago, I concluded commercial flight was a solid 8x better then auto travel, but as I recall GA was at least as risky as auto, and could much more so depending the aircraft and travel methods used. As in a 182 has a lower incidence of fender bender then a 210.
This is Chad, not Craig, but I think you meant me. Thanks for your comments. Your thinking re. not being in the same risk pool as everyone else reminded me of a couple of good articles by Rod Machado on his web site. Basically he says that it isn't fair to compare the safety of automobiles and airplanes, because when flying airplanes we have a lot more control over our own fate than we do when driving a car. It's certainly another way to look at it anyway. He says:
"Comparing the statistical safety of airplanes to automobiles is more meaningful to those who cannot (or will not) attempt to influence the safety of their flight. Most pilots, thank goodness, donâ€™t fall into this category. While the airplane-automobile statistic might be meaningful for helpless passengers, a proactive pilot should find it of little value. If you'd like a statistical measure of your ability to fly safely, compare yourself to those who've flown safely for many years, not to automobile drivers."
On the other hand, I'm not sure if I totally buy into his line of thinking, because I think it is valuable to keep a good measure of humility and not assume that I am better than the average pilot. I will certainly keep training and try to be better than average, but it seems a bit arrogant to assume that I am.
Anyway, here are the articles:
Is Flying Dangerous?
Lies, Big Lies and Statistics
The second one has some practical saftey advice too.
What figures did you use and what are your sources. I am a business traveller and fly GA and commercial as well as driving. I have been trying to figure this out. Personally I am more interested in the fatalities per passenger mile. Can you give me your sources?
Date: 9/27/2004 6:47:29 PM
What figures did you use and what are your sources. I am a business traveller and fly GA and commercial as well as driving. I have been trying to figure this out. Personally I am more interested in the fatalities per passenger mile. Can you give me your sources?
I have included the data I used, with references to the source, along with my assumptions and analysis, in this spreadsheet:
Note that, for reasons I have explained above, I don't think fatalities per passenger mile is a meaningful number, but even if you use this statistic, GA comes out worse than driving.
I am not a statistician by trade, and I'm sure there are many ways to look at these numbers. What they tell me, though, is what should be instinctively evident to most people anyway: An activity where we propel ourselves forward much faster than when driving, and in which we leave the ground and then have to come back to the ground in a specific, controlled manner, is inherently more risky than driving. But flying is also immeasurably more rewarding than driving, so I'm going to keep flying and just try to make it as safe as possible. I think making it as safe as possible starts with acknowledging that it is hazardous, and that's part of why I did this exercise. Of course, the other reason I did it was that I was hoping it would come out the other way .
Anyway, this is just one man's analysis. Do read Rod Machado's articles that I have linked above for another, more optimistic, perspective on this.
I don't think that statistics can quantify your "chances of dying", especially for a GA pilot. Your chances of dying are primarily up to you. If you make a mistake, neglect an important detailj/procedure, lose concentration, or God forbid choose to fly into extremely adverse conditions, your chances of being killed are significantly better.
On the other hand, if you are dilligent about flying, safety, following the rules, etc, I'm sure your chances are much better.
When they post the "odds" regarding flying safety, they are simply giving you raw, random chance numbers. There is rarely anything random about it. If the mechanic made a mistake, your chances might not be good. If the airline captain had a bad day or makes the wrong decision, his passenger's chances are much worse.
As for weather, winter, thunderstorms, wind shear...those are not random occurances. If you fly in the winter, into a snowstorm, you have a good chance of icing up or hitting zero visibility on approach, and your chances are not good. In summer, a thunderstorm and/or wind shear can swat you to the ground. What are your chances if you avoid flying in/near a thunderstorm? It ain't a question of random.
As pilots, let's make sure we aren't making the numbers for ourselves and our passengers worse. The potentially most dangerous piece of machinery in flying is between our ears.
I beg to differ about using statistics to make decisions about risk of dying, but agree to the conceptual and logistical difficulties. The reason statistics work with human behaviors like flying and crashing is that human behavior is statistically extremely lawful. We all tell ourselves that WE won't do something wrong, and can point out certain things that we probably WON'T do wrong, but there are so many ways to screw up that we can be 100% sure that we will NOT eliminate them all.
Many variables are useful predictors of aviation risk. Here's one of my favorite (and more depressing) ones given by Richard Collins in the Dec Flying issue: every year, about 1 out of 500 registered GA aircraft is involved in a fatal accident, including crashes and ground accidents involving propellers (1 in 442 to be "exact"). Collins compares this statistic to the corresponding one for automobiles to show that it is many times larger for planes - which we spend much less time in than our cars. Another number Collins gives shows that almost 1 out of 1000 current instrument pilots dies each year in a botched instrument flight in IMC (1 in 1281 to be "exact"). These numbers speak for themselves. Collins finds the flying risk from 5 to 21 times greater than the driving risk, depending on how risk is defined and measured, overall.
GA planes will never be as safe as cars due to the much greater energy levels in flying as compared to driving, but that doesn't mean that GA has to remain as marginally safe as it really is now. The solution is for plane and equipment makers to design out some of the human error, like using sensors, avionics, and autopilots to prevent pilots from trying to take off loaded with ice (which SEEMS to have just happened in a high profile case), or controlled flight into terrain on an instrument approach (ditto), or stalling on short final on an instrument approach (which killed Sen. Wellstone), or flying into a thunderstorm, or flying head-on at 340 knots into each other while receiving radar services and talking to ATC (24 Jan 2003 midair over Denver, CO, for example; see NTSB report at http://www.ntsb.gov/ntsb/brief.asp?ev_id=20030203X00145&key=1). Technology is the key to further reduction of these risks, along with common sense and good judgment - of course.
Regarding relative safety, the Australian Transport Safety Board (I guess this is similar to the US NTSB) has published cross modal safety comparisons for different forms of transportation.
You can see them at:
- mark n.
John Szpara makes the same point that Rod Machado makes in his articles, that is that we determine our own fate, so the average stats are not relevant to us.
However, without meaning to cause any offense, this viewpoint seems a little arrogant to me. The statistics speak to how we pilots, on average, perform. As C. Craig Morris points out, we can tell ourselves that we will be better than average, that we will not commit the same silly mistakes that kill other pilots, but how can we be sure? I have occasionally let my speed get a little low on short final, even though I know that's a stupid thing, so I know that's something that can happen if you're not paying enough attention. I acknowledge that I might not remember to switch the tanks every 15 minutes when I'm flying a PA-28, especially since I fly C-172s most of the time, where such switching is not done. There are things I can do to lessen the risks, but to err is human, and I think the statistics are very valuable in pointing out how we mere humans tend to fail. I do hope that I will fly better than the average pilot, but I think I'd be living in a dream world if I were to assume that I will always do so.
I agree with Mr. Morris's assertion that that designing away more of the negative consequences of the human errors is the best way to make flying safer.
Mark, thank you for that link to the Australian cross-modal study! I was very interested by a number of things in the report:
(a) They agree that cross-modal comparisons are difficult because of the different data available for risk exposure (e.g. distance travelled vs. time travelled).
(b) Their relative safety rankings agreed with mine in that driving always came out as several times safer than GA flying no matter which measure you used.
(c) They divided non-GA flying into low capacity and high capacity transport, while my figures only considered "air carriers" as a whole. This probably explains in part why air carriers look a little more dangerous in my study. The rest of the difference appears to be attributable to the simple fact the Aussie and US figures are different: "No deaths have been recorded for High Capacity airline travel in Australia since the 1960s and none involved jet aircraft."
(d) The part of my analysis that I was the most worried about was my method for comparing GA (in which we only know hours flown) with driving and air carriers (in which we know distance flown). I was comforted that they did the same thing that I did: multiply by an average speed to convert GA hours to distance.
I still find it annoying that most of the comparisons use passengers x distance as the risk exposure (e.g. deaths per million passeger miles). If I'm trying to determine the risk that I will die for each mode, I fail to see how its relevant how many other people are in the vehicle with me. To me the relevant risk exposures are how far I'm travelling, how long I'm travelling, how many times I'm starting/stopping, etc., but not how many other people are making the trip with me.
But anyway, very interesting reading that largely confirms my original conclusions. Thanks again for sharing the link.
Passenger miles is a very useful and applicable base for these statistics. Let's say, for ease of comparison, each flight in the database had 100 passengers. Now, instead, let's say that each flight going forward will have only 1 passenger. The relative death rate remains the same: the passenger miles are divided by 100, but so are the deaths. Basically, the reason it's important how many people are on the plane with you, from a statistical perspective, is that if a plane transports 100 people 1000 miles safely, it's equivalent to 100 single passenger cars each completing a 1000 mile trip safely. Also, if that 100 passenger plane goes down, all 100 passengers make their way into the database as fatalaties.
It depends on how you're using the statistic, i.e. what question you're trying to answer.
From a public policy perspective, if you're a government, for example, trying to answer a question like "should we encourage our citizens to drive or fly", then maybe passenger miles is a useful measure, because it will give you the amount of the undesirable outcome (fatalities) compared with the amount of the desirable outcome (a person getting moved a mileâ€"a passenger mile). This seems to me to be a measure more of efficiency than of safety, though.
However, if your goal is to determine how safe it is to get onto that airplane, if, for example, you're trying to answer the question "should I, Chad Conrad, fly to Los Angeles today or drive, based on the probability that the trip will kill me?" then passenger miles are no longer a reasonable base to use. Whether there happen to be 20 other people or 100 other people on board the 737 that day with me has no effect on the probability that the plane will crash that day.
To measure the risk of a certain undesirable outcome, you need to compare the amount of the undesirable outcome (i.e. fatality) with the amount of something that leads to the undesirable outcome (the exposure). As a simple example, if we want to know the risk of dying from jumping from a fourth-floor window, we simply compare how many people died doing this jump to the number of people who did the jump. This ratio tells us the risk. It's important to get the denominator right because if we just compared the number of people who died jumping from a fourth-floor window with the number of people there are in the world, or the number of people who have looked out of a fourth floor window, then it would make the activity (jumping from fourth floor windows) appear to be much safer than it is. Anyway, the exposure measure in this case is number of jumps from fourth floor windows.
Relevent exposure measures for my example of trying to determine the risk of flying to LA might include hours flown, numbers of flights/take-offs/landings, miles flown, number of approaches in IMC, etc. All of these things can be seen to affect the probability that something bad will happen to the flight in that the more of them you have, the more the bad outcome happens, so they are measures of exposure. However, I still fail to see how the actual number of people moved is a measure of exposure. Vehicle miles I can see. Passenger miles I cannot.
Chad's got it exactly right. There's no one "correct" measure and nothing based on a single exposure measure comes close to the truth due to the multiple exposure factors that must be taken into account. In GA, the number of real (as opposed to instructional) instrument approaches is more predictive of risk than distance or time alone by far. Yet the crude measures of "risk" that policy makers rely on invariably give only one - very misleading - measure such as miles, passenger miles, hours, takeoffs, etc. A simple thought experiment reveals the problem of using only distance or time alone. Given the astronomical miles traveled by the space shuttle, and the few miles travelled by skateboarders, one might compute the usual kind of policy making risk measure of accidents or fatalities per unit miles traveled and conclude that skate boarding is more risky than space shuttle flight - an absurd conclusion. A more meaningful comparison would be, if I take a skate board ride or a space flight, which is more likely to kill me and how much more likely. Alternatively, one might ask, how is my life expectancy likely to be affected if I choose the hobby of space flight as compared to skateboarding? The latter seems to me to be the only way to compare and even combine the multiple risk factors involved in even our most mundane choices like what to have for breakfast - or what not to have. If I choose to drive fast and aggressively, or fly with little regard for regulations and conventions, or to eat and drink with reckless abandon, or to work in a hazardous occupation, or to live in a dangerously polluted environment (like mine here in Washington, DC), or to carry a handgun (as opposed to keeping one where neither I nor anyone else can access it in the wrong frame of mind), then how much life expectancy will I have to pay the piper? And how much will those losses add up to across all involved parties for the policy makers to consider? Difficult but not impossible questions - and closely related to personal freedom.
Cconrad, I think you're missing the point. If the question you're trying to answer is whether you should drive or fly to your destination, then passenger miles is a perfectly acceptable base on which to compare the two statistics. If you change the airline statistic to be number of crashes per flight (to remove what seems to be confusing you, i.e. the number of passengers on YOUR flight), you'd get an approximately identical answer for your probability that you will die on your flight. As a simple example, let's assume there were 1000 flights in a year, each going 1000 miles, with 100 passengers each, and 10 crashed with every passenger dying on each crash (I know this magnitude isn't realistic, but it makes the numbers easy to work with). Then let's say you were trying to answer the question "what's my probability of dying if I take this 1000 mile flight?". With a passenger-mile base, you get 100*1000*1000=100,000,000 passenger miles, and 100*10=1000 deaths, for a probability of 1000/100,000,000=1/100,000 deaths per passenger mile. Since you'll be flying 1000 miles, you get (1/100,000)*1000=1/100 or 1% chance of dying. Now, instead of this base, let's use your number-of-flights base (with disregard to the number of passengers on each flight), it's simply 10 crashes out of 1,000 flights, or a 10/1000=1% chance of dying. As you can see, you get the same result.
I understand perfectly the importance of selecting a good base (i.e. denominator), but using passenger-miles for this purpose works quite well.
By the way C. Craig, space shuttle flight vs. skateboarding is a non-germane example. They don't both serve the same purpose, i.e. to transport you to the same destination. If you're going to be doing both of those for the same length of time, then deaths per hour or something of that nature is probably more applicable than deaths per mile. However, with the example of taking an airplane or a car to Los Angeles, both serve the same purpose, and I'll be traveling the same number of miles for each. My time, however, will be much different for the two trips, which is exactly why deaths per hour is a very poor measure to use for comparing these two choices.
There's plenty of confusion among experts in this area.
Start with passenger miles traveled (pmt) as supposedly better than vehicle miles traveled (vmt). That depends on the numerator, among other things. If one persists in defining risk in terms of a mere ratio, and the numerator is fatalities, then pmt is better than vmt, because in a given vehicle mile traveled, there is proportionally more opportunity for a fatality to occur the more passengers aboard. So, if airline A carries twice as many passengers as airline B, but crashes equally often in terms of vmt (say once every million vmt), then A will appear twice as risky as B in terms of vmt (with twice as many fatalities, on average) but equally risky in terms of pmt. That point was made above. But, if Chad only wants to know what's the chance that this flight will crash, and he must choose between airlines A and B, then he wants to know how many fatal crashes occur per million vmt, or per million takeoffs, or per million flight hour, etc. Any one ratio alone is misleading out of context of the others.
As an extreme example of how misleading an irrelevant risk ratio may be, suppose an airline racks up 99% of its vmt or pmt on coast to coast high altitude flights, thus touting a wonderful safety record given the preponderance of such innocuous flights, but spends the remaining 1% of flights at low altitudes over mountainous terrain crossing cold fronts with frequent instrument approaches to minimums. While the airline no doubt compares favorably to other airlines routinely flying more in the latter conditions, which airline will really be the safer choice? There is no way to use a simple ratio to answer such a question, because it requires that the two compared trips be identical in all respects except the denominator of that simple ratio. That never happens, yet such misleading ratios persist because they do.
Regression models of the form g(Y) = f(X0, X1, X2, ..., Xk) are a more sophisticated way to express and predict the "numerator" (like fatalities or fatal accidents) as a function of various exposure measures taken together simultaneously, but require a lot of information (ie, the exposure measures) not readily available for various reasons. Another interesting concept of risk is the cumulative risk of a lifetime of some portfolio of trips by various modes, or of all known risk factors considered together simultaneously. Confusing, but challenging, too. This is the kind of thinking behind people's lifetime decisions to fly GA planes or not, ride the airlines or drive, drink red wine or not, and so on, but the experts continue hawking misleading "risk ratios" that ignore complex interactions of multiple exposure factors, some beneficial and many not - whether involving transportation safety or health in general.
Anonymous: It's arbitrary to declare that the purpose of a trip is to get from point A to point B when in fact, the purpose of the trip may be to experience recreation in one form or another - as with space travel (coming to a spaceport near you soon) versus skate boarding. But regardless of the purpose, no single ratio (and thus "denominator") suffices at all except for rhetorical (eg, advertising, marketing, self-promoting) purposes.
Chad, here's one more eplanatory example for you that I think I didn't explain in enough detail earlier. Let's say we start with x and y, the average deaths per passenger-mile for airline travel and vehicle travel respectively. Let's assume each airline trip had an average of 100 passengers and each vehicle trip had only 1 passenger. YOU are claiming that the ratio of x/y is not a valid comparison. I am trying to explain that it is, and here's why. Let's say I wanted to now modify my data as though each airline carried only one passenger per trip. My original x deaths per passenger-mile when each airline carried an average of 100 passengers per trip now becomes x deaths per (passenger/100)miles, i.e. (x deaths)/((passengers/100)*miles)=100*x deaths per passenger-mile. In YOUR analysis, you stop there, but you shouldn't. Because also now, if I want to change my data as though each airline trip had only one passenger, then each time one crashed, it killed only one person, not 100 as before. So I need to divide my deaths by 100, and I get 100*x (deaths/100) per passenger-mile. As you can see, the 100's cancel and you're left with again x deaths per passenger-mile. So whether each airline carries 100 passengers or 1 passenger per trip, the number of deaths per passenger-mile is unchanged, and in either case directly comparable to the number of deaths per passenger-mile by vehicle. Think about it and run through a few numerical examples.
One way to begin is to say there are only 4 flights in your database, with one of them crashing. Now, let's ask, "what's my probability (just based on this limited data) of dying if I take a flight?". First, if I analyze it by chance of crashing per take-off, then it's clearly 1/4, or a 25% chance of dying. If I instead analyze it by the number of passengers, then we need some information about how many passengers were on each flight. Let's again assume each had 100 passengers. 100 died out of 400 passengers, or again a 25% chance of dying. If each plane had 1 passenger, then it's 1 death out of 4, or a 25% chance of dying. The only time you get something much different is if the plane or planes that crash are carrying a far different number of passengers than an average flight. With the number of years of experience you're looking at in your database, I think it's safe to assume that the number of passengers per crashed flight is roughly equal to the number of passengers per flight overall, so that should not introduce a bias.
C. Craig, I think you're a little confused too. Your example of airline A and airline B is a great one, with airline A carrying twice as many passengers on average as airline B. If they both crash once every million vmt, then they will both show the same safety rating as each other, whether looking at vmt or pmt. It's true, as you said, that airline A will have twice as many pmt, but it will also have twice as many deaths. If airline B has x deaths and y pmt (i.e. x/y deaths per pmt), then airline A has 2*x deaths and 2*y pmt (i.e. 2x/2y deaths per pmt, or again x/y deaths per pmt).
I was not saying that the only question worth asking is which mode of transportation is safer for a given trip, but that was the question that Chad threw out there saying that deaths per pmt is a poor base for comparison.
Slow down, anonymous, you're getting a little reckless. No need to rush. We've all spent a lot of time thinking about these issues and made many valid points. My goal is to achieve breakthroughs by brain storming with others. PMT is no panacea. Get over it.
C. Craig, nothing is going to be perfect, but pmt has been made out to be a much more misleading base than it is, which in itself, is quite misleading (although I believe it's been an honest mistake). No comment on your Airline A and Airline B example, huh?
Thanks for your concern, but I have nothing to "get over" as you so eloquently put it. I have no vested interest here, other than trying to correct some incorrect statistical usage, which ironically is being used to attempt to prove incorrect statistical usage. You may have put a lot of time into thinking about these issues, but without outside influence, you may be wasting that time thinking about them incorrectly in some cases.
OK, here's my final answer. As Albert Einstein once said, a theory should be as simple as possible, but no simpler. All risk ratio concepts like fatalities/pmt and fatal accidents/vmt oversimplify, because they omit many nontrivial factors known to affect risk thereby reducing the comparison to a pseudo-scientific rhetorical exercise. With auto versus airline trips, for example, relevant factors for the auto trip include the mass and CG of the auto, whether it has airbags (front, side, curtain), whether occupants are belted, their age and gender (these are big factors), their fatigue level, their propensity towards speeding and aggressive driving, road conditions, weather, etc. With the airline trip, relevant factors include the airline, the aircraft, number of takeoffs and landings, pilots'skill and physiological state (eg, fatigued or ill), enroute and destination weather, and ground travel to/from origin and destination airports. All this is obvious, but ignored anyway in simplistic ratios used to supposedly compare risk across or even within travel modes.
You are right when you say all of those are obviously factors. However, equally as obvious is the fact that many of those pieces of data are not readily available. So what would you have us do? Perhaps we should just ignore the statistics altogether because they aren't perfect? NO, they are calculated based on averages of very large numbers. Of course they will not tell you the exact percentage chance of dying on a specific flight or in a specific vehicle with various characteristics. They do, however, give a pretty darned good picture of relative risk by travel mode in general.
I thought the point of this thread was to discuss the general relative risk of flying and driving, not the relative risk of flying on a cloudy day with a particular aircraft and cargo and a pilot who hasn't had enough sleep and had a drink two hours beforehand versus a driver who is in a bad mood and is driving an SUV with the radio blaring that has side-curtain airbags.
By the way, I like how you side-stepped addressing your example of Airline A and Airline B and simply added the "here's my final answer" clause to your last post.
If I were in your shoes, I'd simply thank the guy for showing me that I was misapplying a concept. You don't seem like an obtuse fellow, so I'm pretty sure that you now understand the error, but decline to address it.
By the way, my name is Sean Bailey. I didn't mean to hide behind a veil of anonymity. I'm just not a member.
OK, Sean, maybe I still don't quite understand what the problem is with the A,B example, which I thought was merely a restatement of your own example showing what pmt adds to vmt. And thanks for lifting the unintended "cloaking," which I've also struggled with in some past posts. It's an interesting thread thanks to contributors like you.
Date: 2/16/2005 6:34:41 PM
Cconrad, I think you're missing the point. ...
Wow. I was still trying to figure out how to respond to Seanâ€™s post of 16 Feb 05 18:34 and then this explosion of posts happened .
Firstly, Sean, thanks for identifying yourself. I know itâ€™s a little thing, but I think people should always take responsibility for their opinions. (Iâ€™ll defend to the death your right to say anything, provided you identify yourself.) If you didnâ€™t want to register on the forum (a very quick and easy task), why not at least type your name at the bottom of the posts? It can become confusing trying to figure out which Anonymous is which!
Second, I think I may have to concede that passenger miles are not as misleading as I was making them out to be. I didnâ€™t actually miss your original point (which you restated essentially the same way very well several times). Itâ€™s just that it seemed so wrong to me that the number of people on board should be involved in the calculations that I purposefully ignored the argument.
But I suppose youâ€™re right, thereâ€™s nothing wrong with multiplying the numerator and the denominator by the same factor (in this case passengers). In fact, I think I was erroneously comparing fatalities per passenger mile with fatalities per vehicle mile, which definitely would not be directly comparable, whereas as you pointed out, it is probably okay to compare fatalities per passenger mile with fatal events per vehicle mile.
However, I have to say that I still donâ€™t know what is gained by adding the passengers into the calculations. For my sample problem from my recent post (should I fly or drive to LA), I think what I really want to know is something like fatal events per take-off and landing. I really donâ€™t want all those passengers muddying the water, because, despite the fact that youâ€™ve demonstrated that their inclusion is innocuous, they are still not a risk exposure either.
P.S. I suppose I should redo my spreadsheet (linked up near the top of this thread in my message that was posted 27 Sep 04 20:22) to include some fatal crashes per flight type of information. Or may I couldnâ€™t find that informationâ€¦
P.P.S. I almost forgot the other point I wanted to make: My original, original point was to try to prove or disprove that general aviation is safer than driving, as some pilots have tried to tell their spouses and friends. All of the numbers I found (again, refer to the spreadsheet linked above) disprove this by a very large margin, no matter which base/exposure measures you choose. I think weâ€™ll all agree on that. But, as Iâ€™ve said before, it wonâ€™t stop me from flying. I just hope that recognizing it as hazardous will help me stay alive.
By the way, to my credit, way back last year I did admit that I may be making just the statistical mistake Sean has charged me with:
Date: 8/3/2004 12:42:54 AM
By the way, just to poke a little hole in my own argument, one could argue that if I don't want to given extra statistical ''credit'for the number of people on board for a safe landing, I shouldn't do so for an accident either, meaning that I shouldn't consider the number of fatalities per thousand hours or million miles but rather the number of fatal incidents. I didn't do so, if I remember right, because those numbers were even harder to come by in a convenient and/or comparable form. If I were to do so it may make airline travel come out a bit better, but it wouldn't help general aviation much, and that's what most of us are interested in on this forum anyway. ...
Chad, based on all of the data I've seen, it certainly does look like GA is more dangerous than driving. Also, it looks like commercial airlines, as a means of getting from point A to point B, is considerably safer (as the airlines have stated) than driving from point A to point B.
To your credit, you were absolutely on the right track when you pointed out that statistical mistake. But then the thread seemed to veer from that point and I didn't think the myth-busting (concerning commercial flying) had been busted in most readers'and posters'minds. I just wanted to get the math right for everyone, which is what compelled me to post.
More interesting to me would be looking at the data broken into length-of-trip subsets. I would guess that deaths per vehicle or passenger mile in cars is largely unchanged based on the length of trip (i.e. I'd guess that a 5 mile trip across town poses close to the same risk as 5 miles of my cross-country trip). However, I would also guess that deaths per passenger mile is considerably higher for shorter plane rides than for longer ones. This is because a short trip still involves one take-off and one landing (probably a large part of the total risk), but tallies only a fraction of the miles (i.e. making the denominator in deaths per passenger mile smaller, and in turn making the overall fraction larger). I don't think, however, that it would show commercial flights to be riskier than driving even for relatively short flights. I haven't done this analysis, so I am only speculating, but it seems like a more interesting distinction to me.
As for this thread's purpose in encouraging attention to safety in GA, and everywhere for that matter, I hope it is 100% effective.
When I first posted, I only intended to add a quick 2 cents, and did not see the need for adding any identification, but as I proceeded to add many more dollars worth, I felt a little accountability might be in order. I appreciate the give and take.
Date: 2/17/2005 10:38:36 PM
... More interesting to me would be looking at the data broken into length-of-trip subsets. I would guess that deaths per vehicle or passenger mile in cars is largely unchanged based on the length of trip (i.e. I'd guess that a 5 mile trip across town poses close to the same risk as 5 miles of my cross-country trip). ...
Actually, when reading a discussion on this point somewhere (sorry, I can't recall where, but I think it was while I was first gathering the data for this last year), I believe I read the opposite. That is, the risks from driving increase with each hour and mile. This makes intuitive sense to me too, since with each passing mile more of the same hazards (traffic, dangerous curves in the road, etc.) get thrown at you. But since I can't recall the source, this may just be my own opinion.
Date: 2/17/2005 10:38:36 PM
... However, I would also guess that deaths per passenger mile is considerably higher for shorter plane rides than for longer ones. This is because a short trip still involves one take-off and one landing (probably a large part of the total risk), but tallies only a fraction of the miles (i.e. making the denominator in deaths per passenger mile smaller, and in turn making the overall fraction larger). ...
This I would agree with. There is probably almost no risk in level cruise in an airliner compared with the risks at take off and landing.
So, if you accept both of these premises (that driving gets more dangerous with each mile, but flying not), this would make flying safer the longer the trip.
On an interesting side note, the Australian study that someone provided a link to earlier in this thread noted a remarkably high fatality rate associated with walking, but when you read the notes it turns out that a large number of people killed while they were pedestrians were also intoxicated. It shows how statistics can mislead. I think walking is quite safe, but the study sure didn't indicate that until you got to the footnote about walking drunk.
I think you are in fact agreeing with me about the cars, without realizing it. I stated that I'd guess the deaths PER MILE are fairly constant in road vehicles regardless of trip length. In other terms, I'd guess you would encounter roughly the same total hazard by driving 1000 miles to another state versus driving 200 5-mile trips.
I agree that airliners are probably relatively safer for long trips versus short trips, however, I'd still guess that they're still safer for short trips than road vehicles.
Date: 2/18/2005 12:07:36 PM
I think you are in fact agreeing with me about the cars, without realizing it. ...
Yup, I was just reading too fast. My eyes skipped over it quickly and read "I would guess that deaths in cars are largely unchanged based on the length of trip," so I thought you were saying that the risk did not go up with each mile driven, which didn't sound right.
This news clip on a new study is provocative and illustrative of the right way to approach risk. Finds among other things that the youngest drivers are 4 times more likely to be in auto accidents, and 3 times more likely to die in them, than adults. Taken out of context, the latter ratios are somewhat misleading, because the fatality risk is higher for boys than girls - the study may discuss this. Also, accident risk goes up with number of occupants for these young drivers - which, like the age and gender effects, has interesting ramifications for use of pmt as a measure of risk exposure. I'll try to get the study and see if they accounted for the dramatic improvements in crashworthiness and crash avoidance that have been designed into autos over this period when, concurrently, the regulations were changed to more graduated entry into driving. Most likely, both contributed to the improvements. I wonder if aviation might benefit from similar effort on both fronts, ie, more attention to design enhancements that help avoid or survive crashes and more graduated entry into the real world of flying.
Youngest Drivers' Death Rate Decreases
By Elizabeth Williamson
Washington Post Staff Writer
Thursday, February 24, 2005; Page B01
Fatal car crashes among the nation's youngest, most accident-prone drivers decreased sharply in the decade after most states enacted laws limiting their access to a driver's license, a new study shows.
Auto deaths involving 16-year-old drivers fell 26 percent between 1993 and 2003, a period when 46 states and the District enacted graduated licensing laws that allow fewer 16-year-olds to drive, according to the study released today by the Arlington-based Insurance Institute for Highway Safety.
Among 16-year-olds who have full driving privileges, the rate of fatal crashes has not fallen, and it remains higher than that of any age group. Researchers said the difference between the two groups points to the effect of the new laws, which keep most 16-year-olds from receiving unrestricted licenses and which are intended to curb risky practices, such as carrying teenage passengers and driving at night.
The study shows that "withholding a license is going to improve the crash picture among 16-year-olds," said Susan A. Ferguson, senior vice president for research at the institute and an author of the study. "That's a success story, because the fewer of them who drive, the fewer of them who die."
The dramatically higher crash and fatality rates for teenagers in their first year behind the wheel have prompted highway safety advocates to fight for laws that prevent them from driving unsupervised, at least until they turn 17. Sixteen-year-olds are four times as likely than adult drivers to become involved in a crash and three times as likely to die in one, national statistics show.
Motorists are at the greatest risk during their first year of driving -- with high speed, driver error and multiple passengers contributing to crashes, previous research cited by the Insurance Institute shows .
Although the rules vary by state, graduated licensing programs extend the learner's period, the time that teenagers must drive with an adult in the car. Further, they limit driving under conditions linked to crashes. The first graduated licensing program was introduced in Florida in 1996. By 2003, 47 jurisdictions had the programs, including the District, Maryland and Virginia.
Lawmakers are considering legislation to further restrict teenage drivers. In Maryland, the state Senate gave tentative approval yesterday to a bill that would prohibit new drivers from carrying passengers younger than 18 during the first six months of an 18-month provisional license. The measure, which is expected to win final approval today, has not cleared the House of Delegates.
In Virginia, the Senate and House have approved measures that would restrict cell phone use by teenage drivers, and are working to resolve differences between two versions of the bill.
According to the study, 938 16-year-old motorists nationwide were involved in fatal wrecks in 2003 -- more than one-fourth fewer than in 1993, when 1,084 died in such crashes. That is despite an 18 percent increase in the number of 16-year-olds in the nation, the study found. At the same time, the percentage of fully licensed 16-year-olds declined -- from 42 percent in 1993 to 31 percent in a decade.
Also striking, the institute's Ferguson said, was a 39 percent drop in fatal crashes involving a 16-year-old drivers carrying other teenage passengers.
"The probability of being in a crash increases with each additional in the car," she said. The figures show that "there's a fundamental change in the way young passengers are being transported."
In the region, crash rates among teenagers have been more erratic, pointing to differences in the graduated licensing programs, as well as the difficulty in drawing conclusions from relatively few accidents.
In 1993, there were 27 fatal crashes in Virginia that involved a 16-year-old driver, eight such crashes in Maryland and none in the District. In 2003, there were 12 fatal accidents involving 16-year-old drivers in Maryland, 16 in Virginia and one in the District.
This year, a string of deadly accidents involving new drivers has moved legislators to try to toughen the region's licensing laws, and to place new restrictions on the youngest drivers -- an effort that Ferguson applauds.
"I think there's clearly more we need to do," she said.
I, like you, am drawn to the technological solutions as a way of carrying on where training leaves off.
If mass-produced cars can have front and side airbags, curtain airbags, sensors that prevent oversteer, understeer and brake lock-up, radar-driven cruise control, etc., then why can't my small GA plane have airbags, a parachute, a radio altimeter, weather and traffic radar, TCAS, TAWS/GPWS/whatever it's called, etc.?
Seriously, when I am fortunate enough to stop renting 1970s airplanes some day, I hope I can own something with at least airbags and a parachute.
It's interestng to compare the safety-relevant features of cars made in the 1960's to those made today. In spite of the bright engineering making mass-produced autos available to nearly everyone, they didn't even put in seat belts. Many minor crashes by today's standards killed people back then, eg, hitting a pole at 30 mph. But have aircraft designs changed accordingly over the same period? Hardly. Then they did have lap belts, and now they have shoulder belts - which make a big difference in survivability - but that's about it. My guess is that arrogant American engineers (or CEOs) will again wait for a Japanese company to solve the obvious problems, eg, integrating autopilots, gps, airspeed gauges, fuel burn, etc, to "discourage" and even prevent recurrent causes of crashes like flying too low and/or too slow on instrument approaches, VFR into IMC loss of control or CFIT, etc.
Why should we "good" pilots care? Two reasons: (1) we're not always as good as we may have been led to believe and should value our own lives and those of our family and friends, and (2) the cost of aviation will drop drop substantially with a growing GA market as more people fly who won't now because they've made a legitimate life decision to exclude that unacceptable risk.
Fly or drive, an interesting question, which I have thought a lot about through the years, as I have circled the globe by air many times.
You have made some good points, but even comparing risk by hours has some drawbacks, because if you fly coast-to-coast by air, you subject yourself to crash risk for only 5 hours or so, whereas if your drive, you subject yourself to crash risk for what?, 50 or 60 hours?
My favorite statistic to reflect on as I fly is ''flights per fatal hull-loss'', rather than ''passenger miles per fatality''. It varies each year of course, but for air carriers, it is often 2 to 3 million flights per fatal hull-loss. In a year when it is 2,000,000 for example, that means that if I took one flight a day, I could expect to fly 5,500 years between crashes with fatalities (if I stick to air carriers). That makes me feel fairly safe, although I still get nervous every single time the plane starts it takeoff roll. I am not in control! (Also, statistically, the crash flight might just as well be at the beginning of the 5,500 years than at the middle or end.)
If you have to pick an air carrier plane to fly, fly the Boeing 777 or Airbus A340. They are the only airliners that I know of that have yet to have their first hull-loss crash with fatalities, an amazing statistic for each plane, considering the billions of passenger miles they have racked up since their introductions.
One last thought, more or less agreeing with you that driving is safer is this: If you DO crash, at least in a car crash, there is a better than average chance that you will survive. In a plane crash, although occasionally there are survivors, more often than not, it's 100% of the passengers that bite the dust.
Thanks for starting the interesting discussion
(retired oil-field trash)
Factor out pattern accidents and flying should be safer!!!!!
Mmm. You would have to compare distances. The assumption is that you are driving the same distance you are flying; so per passenger mile *is* correct. Or, worst case, divide the per-passenger mile by the avg number of passengers.
I do recall however that more then 75% of all passengers invovled in commercial ''crashes'survive. Generally because planes don't crash because the wings tend to fall off, they crash because something goes horribly wrong on take-off or landing... when you are near an airport with a whole *bunch* of nifty stuff for helping people.
Hey folks - read the Nall Report and do the math. It makes things pretty clear. GA flying is many times more dangerous than driving. Also consider that the fatality rate is even higher than the Nall numbers report since the total GA flight hours come from what pilots report and most exagerate their flight time. Richard Collins talked about this years ago and, to his credit, was brutally honest about aviation safety (unlike some in av media).
A lot of the dead pilots in these accident stats probablythought of themselves the same way most us do - that they are conservative, proficient and well trained aviators.
I love general aviation more than most things but I never BS people and preach about how safe it is.
A great old discussion!
I can't cite references but my understanding is that GA overall is about as safe a motorcycling. That's how it was described to me at uni in 2007. Take out the usual higher-risk groups though and the numbers look a lot better. Part of that is that the per centage of accidents that are human error apparently sit consistently at around 70%.
As for GA vs RPT - I think its a no-brainer and I like the measure mentioned above - safe flights per fatal hull-loss. Despite that, I am impressed with the OPs willingness to shake down the numbers and exercise initiative.
Richard Collins is the Guru! A term I'm sure he would hate but it's true.
Thanks, John in Brisbane
Also just to add ... with the exception of twin engine training, I think the training period is very safe and that it's the initial few hundred hours that are the most dangerous, particularly for people who can't fly enough hours to retain proficiency.
thanks, John in Brisbane
although i am not a pilot myself and was in the air cadets as a young chap and having only flown a beech baron about 30 minutes in the early to late 1970 s i have flown in a wide range of passenger aircraft from 747 to 767 a330 l1011 and dc10 in my personal opinion i belive that flying is safer than driving a car.i went up in gliders two times although i see crashes on tv by watching the national geographic channel about air disasters i am often asked by people aint you affraid to fly my answer is no i am not affraid,to say the truth i would rather fly than drive a car(remember the nice thing about flying a big passenger airliner or heavy metal jet you cruise at a nice speed ) also flight has interested me i have been fascinated by airplanes since i took my first airplane ride when i was a small child.
I'm a pilot with 3500 hours.
Of course flying isn't safer than driving... at 3 times the speed, there is 9 times the kinetic energy to deal with in a crash. Aircraft by design are lightweight, and therefore much less crashworthy than an automobile even if the energy were the same.
The failure modes in an automobile are much more benign also. Unless you happen to be driving through a very bad neighborhood at just the wrong moment, the consequences of an engine failure, even well managed are much riskier in an aircraft than in a car. An engine failure 30 seconds after liftoff in an airplane is exponentially more dangerous than the same event in a car.
Factor in weather and the relative complexity of flying as compared to driving and it's a wonder that the stastics aren't much worse than they are.
This should not be surprising to anyone, airplanes leave the ground and cars do not.
Most of the risk in driving is uncontrollable by the driver, such as the risk of colliding with another driver who is intoxicated, or distracted. In flying we get to choose our level of risk for the most part -- given a good attitude, good training, and a little experience, and much (but not all) of the risk can be mitigated -- this is not possible to teh same extent in an automobile.
Finally, let me comment on the passenger miles vs hours flown vs miles flown discussion. For the individual on the aircraft, it seems to me that the sensible denominator is "per departure". In other words, if I, today, buckle myself into this vehicle, what are the odds that I will DIE before I get out of it. This is the most personal and immutable of the stastics, and really it is the only one that an individual cares about. GA doesn't fare particularly well in this category. If 400 people get on an airplane and it flys 3000 miles in one hop, you can see that your actual risk of NOT walking out of the airplane you just walked into could be 800,000 (400*3000) times as much as this "passenger miles" value.
Fly because it's convenient, flexible, and fun, but don't kid yourself into thinking you're safer than driving.
I'm a pilot with 2500 hours and one accident to my name, a ground loop in a Stearman.
The thought that invariably passes through my mind each time I get into my Citabria or C-182, a commercial airliner, or an automobile is, “will this be my last ride?”.
Disregarding miles or hours traveled but only considering experiencing which mode of travel is most survivable for each trip taken.
I don’t know how to figure this out statistically, but hopefully someone may have some idea how to approach this that might reveal some meaningful information.
I know this is like comparing apple to oranges to potatoes, but psychologically the thought experience each time I embark on a mode of travel is very real.
Statistics ?in a day maybe a few thousands flight happening but in a day how many cars going to roads?Maybe billions tomes.It is something like every 1000 flights 1 crash and 1 million drive 1 crash and less fatal.Plus and more importantly car crushes happens mostly because of driver mistake while airplanes due to technical problem.
It depends but the Ratio of the automobile and aircraft is huge. But plane accidents especially jetliner crashes is more like mass transit disaster but the survivability ain't that good. Having a technical problem on the air is way different to the ground and thats why having an air crash is mostly unsurvivable. By the way did you know that airsafe.com released a data for 2011 regarding air crashes.
Check the link provided.
Take extra care to avoid activities that might detract from flying.
Copyright © 2010 FLYING. All rights reserved. Reproduction in whole or in part without permission is prohibited.