Banner
    Flying Or Driving: Which Is Safer?
    By Gerhard Adam | August 25th 2010 01:22 PM | 35 comments | Print | E-mail | Track Comments
    About Gerhard

    I'm not big on writing things about myself so a friend on this site (Brian Taylor) opted to put a few sentences together: Hopefully I'll be able...

    View Gerhard's Profile
    In reading one of the other posts a casual point was made regarding the relative safety of flying versus driving.  It is generally assumed that flying is, by far, the safest of the two modes of travel, but is this really true?  In looking at the data, it appears that the data is being skewed because of some strange assumptions that tend to favor flying.

    Let me be clear that I'm not suggesting that flying is unsafe.  In fact, I would argue that the two modes of travel can't legitimately be compared.  Why not compare flying with swimming?  or hiking?  However, since the safety argument is invariably made regarding flying and driving, it is this tacit assumption, that one is so fundamentally safer that it is beyond questioning, that I am challenging.

    One of the first assumptions in evaluating safety is based on the notion that passenger miles is a legitimate metric.
    "If we ignore property and bodily damage and focus on fatalities only, we should look at fatality rates per passenger mile traveled."
    http://www.newton.dep.anl.gov/askasci/gen99/gen99845.htm
    I would argue that this is a bogus metric. Consider that the point being made is that if 200 people travel 1000 miles together that it somehow represents 200,000 miles.  In truth, it doesn't.  It's only 1,000 miles. So using one set of statistics it is reported that the airlines experienced 572,885,732,000 passenger miles in 2005 (domestic U.S.).  The reality is a bit different.  With 10,090,274 flights averaging 867 miles, it turns out the actual distance flown is only 8,748,267,558 miles.  This is almost 66 times smaller than the original reported number.

    From a safety perspective it can hardly be argued that safety is increased based on the number of passengers (since the risk isn't carried individually).  The risk is intrinsic in the actual distance traveled, so it doesn't matter if the flight is full or empty when quantifying this value.

    If we consider the number of licensed drivers in the U.S. we have approximately 202.8 million (2006).  So once again, we have a factor of about 20 times difference in the number of drivers versus aircraft passengers.

    It is also important to consider what constitutes a fatality in these statistics.  Clearly when an aircraft crashes there's little doubt about the fate of the passengers, however in automobile statistics this isn't as clear-cut.  Once again, using the 2005 statistics, there were 43,510 people killed1, however nearly 6,000 of those were pedestrians or bicyclists.  While this is certainly an unfortunate statistic, it can hardly be argued that it is a result of the mode of travel (unless one wants to criticize walking).

    So to make a more proper comparison, we should consider the relative risk of something happening to a driver/passenger every time they enter a vehicle for transport.  After all, it doesn't really matter how long or short the trip is if we consider that it is just as likely for a plane to crash on take-off as it is on landing, just as a fatal automobile crash could occur just as readily in going to the corner store as in making a cross-country drive.

    If we take some arbitrary numbers, we can begin to approximate how these numbers relate.  With 10,090,274 domestic flights in the U.S., that becomes the number of possible events that could've resulted in accidents.  With cars, I'm going to make the following assumption that taking all the licensed drivers (202.8 million) and assuming that they only drive to and from work.  This would result in two trips per day for about 261 days or 522 trips per driver.  As a result, the total number of driving events would be a staggering 105,861,600,000 events.  On this basis, we can see that an automobile driver is 10,000 times more likely to be exposed to the circumstances leading to an accident.  This comparison indicates that every time we get into a vehicle we are incurring some risk, however, we engage in such a risk 10,000 times more frequently with cars than we do with aircraft.

    One of the obvious differences in assessing the two modes of travel is that automobile accidents tend to hold relatively steady from year to year, whereas aircraft fatalities can range from zero to several hundred, even with only one accident occurring.  As a result, it isn't really fair to compare the two values to each other, since any choice of aircraft fatalities is clearly arbitrary and can radically change based on the period being examined.

    It should be clear that when comparing these two modes of transportation an important element that is typically ignored is in comparing the relative amount of exposure to risk that each traveler incurs.  As a result, the assumption that air travel is orders of magnitude safer than an automobile is simply urban legend.  While one can certainly make the point that increased professionalism of pilots and the oversight in place with air traffic and ground controls also offer more protection, rather than the free-for-all we typically encounter on our highways, the simple reality is that we use our cars far more often than we fly and as a result, we incur a greater risk because of it.

    In the end, comparing automobiles and aircraft is an "apples and oranges" comparison, but that doesn't automatically translate into increased safety, nor a free pass in having the numbers examined. In fact, while driving has such a marketedly higher rate of participation, less qualified individuals, and approximately 30% of fatalities the result of alcohol, it's incredible that driving is actually as safe as it is.


    1 This also includes motorcycles, while the aircraft statistics do not include ballooning, gliders, etc.


    Comments

    You're misreading the chart you reference for passengers on airlines: that's 10M flights. Not passengers. Passengers * passenger-trip-length is where they're getting passenger miles.

    Gerhard Adam
    Actually that's a typo, so thanks for pointing it out.  It should have read with 10M flights...
    Mundus vult decipi
    kerrjac
    Good points all around.
    Another consideration is that not every flight trajectory can be driven by car. That is, if someone is wondering whether it's safer to drive or fly, they're probably not traveling overseas. In addition to taking out international flights, which have many passengers, you might also want to take out private flights. 

    Perhaps it might be useful as well to focus on individual travel routes, such as Chicago to New York, or Boston to DC.
    Gerhard Adam
    All true, but in the end, my primary point is that the two modes of travel are really not suited for comparison to each other regarding overall safety.

    Therefore the generally accepted notion that flying is safer than driving simply isn't true.

    Mundus vult decipi
    jlparkinson1
    Nice post. It's probably just another of these urban legends, I think...somebody says it and someone else repeats it and who knows where it goes from there.
    Gerhard Adam
    Thanks.  It is interesting how prevalent they become though.
    Mundus vult decipi
    Driving is far more dangerous. Driving displaces biking and walking, which are forms of exercise, and we do not get enough exercise. Diseases of the unfit are far more deadly than crashes. One estimate is that the exercise benefits of biking outweigh the crash risk by a factor of 10, maybe 20 (Mayer Hillman, haven't got an internet reference). Another study I have stumbled across says that non-cycling commuters have a 39% higher mortality rate.

    In contrast, almost nobody chooses between walking/biking and flying; the large distances take far too long by human power.

    Or are we pretending that gratuitous driving does not cause all those deaths, or that they don't count?

    Gerhard Adam
    I'm not sure what your point is.  We could argue that our diet is unhealthy, but that doesn't suddenly make it a legitimate comparison to airline safety.
    Or are we pretending that gratuitous driving does not cause all those deaths, or that they don't count?
    What does that even mean?  The simple reality is that we drive far more frequently than we fly and as a result the number of accidents (including fatalities increases accordingly).  If you want to argue about driving and the risks associated with it, then by all means you have a legitimate point.  If you want to use it as an argument to suggest that flying is somehow safer (because of behavior in cars), then you're wrong.
    Mundus vult decipi
    I think you have to be careful which driving you compare to airline travel for safety.

    Would you fly an airplane for a ten-mile trip? Most auto trips are shorter than that. However, this comparison apparently did not make that distinction, therefore, I think it is appropriate to point out that driving has other safety hazards, not widely recognized, that count against it, though these are only apparent in these shorter trips.

    Gerhard Adam
    Regardless of the intrinsic hazards, the simple fact is that there are 10,000 times more incidences of driving than flying.  As a result, the risks to which someone is exposed are correspondingly higher.

    The fact that cars lend themselves to shorter trips is one obvious reason why there are more incidences (and therefore potential for accidents).

    However, what is more important is that when air travel is considered the data is carefully selected to only consider commercial airlines.  The risk rises dramatically with general aviation.  In other words, it's still an "apples and oranges" comparison.

    Airlines fly at significantly faster speeds than cars.  Aircraft rarely encounter much other traffic with which they have to interact and when they do there are teams of professionals to ensure they stay out of each other's way (air traffic control and ground control).  This improves the safety of air travel but it isn't a proper comparison to automobiles.

    Similarly automobiles are subject to poorly trained drivers, alcohol related incidences, etc.  Therefore given the number of times people get into a car, it is surprising that it is as safe as it is.    They also have the advantage that at lower speeds the rate of accidents can still go up without there being a corresponding increase in fatalities.  In other words you're much more likely to walk away from a car crash than an airplane.
    I think you have to be careful which driving you compare to airline travel for safety.
    There's nothing special here.  When you get into a car or an airplane there is an intrinsic risk.  So the question is only ... what is the likelihood that I will survive this particular trip?  It doesn't matter whether it's a one mile car drive or a thousand mile car drive or airplane flight.  Risk does not increase with distance. 

    I'm not more likely to have a plane crash because it travels 1000 miles rather than 100.  Nor am I any more likely to have a car crash because of the distance.  Therefore distance doesn't count.  The only thing that counts is how frequently we engage in either activity.

    If people wanted to make an honest assessment, then it would be clear that the primary distinction between the two is the level of professionalism and experience exhibited by whoever is driving/piloting the craft.  Inexperienced pilots crash far more frequently than experienced pilots.  Similarly because the FAA regulates rest periods and the amount of flight time that can be accumulated, these are also factors that can work against the automobile driver if they get tired or are inattentive.

    In the end, the point is that neither flying or driving is safer than the other.  It is an improper comparison.  However, it is equally clear that the biggest factor in safety is the human factor (however that has nothing to do with the mode of travel since both are equally like to have a crash if they behave badly).
    Mundus vult decipi
    Gerhard, I would like to respectfully disagree with at least a part of this assertion of yours:
    "I'm not more likely to have a plane crash because it travels 1000 miles rather than 100. Nor am I any more likely to have a car crash because of the distance. Therefore distance doesn't count. The only thing that counts is how frequently we engage in either activity."

    I will at least concede that the distance of a plane ride (commercial, not general aviation) doesn't increase the fatality risk as much as the number of takeoffs-landings ("trips"). This would be logical if we assume that most accidents occur at or near the origin or destination airports. However, I certainly do not agree that the distance of a car ride is not a major factor in the amount of risk assumed by the traveler. I would argue that, to a rough approximation, if you travel twice as far, you are twice as likely to be in a fatal accident. I mean, the more time you spend driving in the car, the more exposure you have to all the risks on the road which can cause you to be killed (tire blow, drunk/distracted driver, driving through intersections where someone else might run a red light, etc). Do you disagree?

    MikeCrow
    Jeff, I'd like to disagree with this:
    if you travel twice as far, you are twice as likely to be in a fatal accident
    According to the AAA 75% occur within 25 miles of home.
    Never is a long time.
    Perhaps, but what if that's because we spend 75% of our driving time within 25 miles of home? You can drive 100 miles in a day, and never get more than 25 miles from home.

    Gerhard Adam
    I would argue that, to a rough approximation, if you travel twice as far, you are twice as likely to be in a fatal accident. I mean, the more time you spend driving in the car, the more exposure you have to all the risks on the road which can cause you to be killed...
    I think you have to be careful here because you're actually switching arguments in several ways.  In the first case, you argue about traveling twice as far, but then you switch it to greater exposure to risks on the road.  These are not identical conditions.

    In other words, if we exclude other vehicles from being on the road, then I think one can conclude that the risks don't increase with distance in any appreciable way.  Therefore distance isn't actually a factor.  Instead, as you indicated in the second part of the argument, an increased exposure to OTHER drivers over an extended period of time, increases the risk of an accident.

    The point being that the risk isn't in the act of driving, but rather in the risk of other drivers being encountered.  This effectively increases the number of incidences per unit time depending on how heavy the number of interactions are. 

    Similarly we can make the same argument for flying, where the risk doesn't increase with distance, but rather the number of interactions we can expect (i.e. other aircraft, landings/takoffs, etc.).  Therefore, a comparable evaluation would result in examining the number of encounters/accident for automobiles versus those for aircraft.  This would ensure that flying doesn't arbitrarily gain "safety" credit for long-distances flown with no encounters. 

    If we don't adjust these values based on encounters and rely merely on distance we can reach the absurd conclusion that the Titanic logged 3,577,860 passenger miles safely before an "accident".

    [based on 2340 passengers/crew and 1529 miles traveled]
    http://www.webtitanic.net/framenumber.html
    http://www.encyclopedia-titanica.org/discus/messages/5921/90987.html?1223584080
    Mundus vult decipi
    Gerhard, this is a hard one to argue over the internet, but I'll give it a try :)

    I'm not switching arguments with regard to risk vs. distance. I am implying a correlation between the two--twice the distance means twice the "number of encounters", and therefore twice the chance of an accident, and therefore twice the chance of a fatal accident. And that's regardless of whether there are others on the road or not: even if I were driving the only car in the world, it's certainly possible I could get into a one-car fatal accident (although the risk is presumably much lower than it would be with other cars on the road with me). And the farther/longer I drive, my risk of a fatal accident goes up. Is it a linear increase? I don't know, but it certainly does go up.

    You are arguing that the risk of dying in a car accident is fixed for a given "trip", whether it's 2 blocks to the 7-11, or 3000 miles across the country. However, I'm saying the length (or time, which are roughly synonymous for the purposes of this discussion) of the trip is absolutely related to the risk.

    I'm not sure about airplanes though, so I was conceding that one. That is, I basically buy your argument that whether the trip is 100 miles or 3000 miles, the risk is more a function of the number of takeoffs-landings, not so much the distance of the trip (although there has to be at least a small component of the risk wrapped up in the distance since, if nothing else, that gives the airplane more opportunity to have a mechanical failure which could result in a fatal crash).

    Now what about you bringing in the AAA stat that 75% of accidents occur within 25 miles of home? That doesn't seem to have any bearing on our discussion. All that stat tells me is that we probably spend the lion's share of our driving time not too far from home. It certainly doesn't imply that if I'm far away from my house then I'm less likely to get in an accident, which seemed to be where you were headed with that.

    Anyway, thanks for listening--I do enjoy a good discussion.

    Gerhard Adam
    Actually I don't disagree with you (although I'm not the one that introduced the AAA stats).  My ultimate point is that comparing air travel to automobile driving is an "apples and oranges" comparison and consequently doesn't represent a valid way to assess the relative safety of travel.
    And that's regardless of whether there are others on the road or not: even if I were driving the only car in the world, it's certainly possible I could get into a one-car fatal accident (although the risk is presumably much lower than it would be with other cars on the road with me). And the farther/longer I drive, my risk of a fatal accident goes up. Is it a linear increase? I don't know, but it certainly does go up.
    Agreed.  However, this is equally true with air travel so it doesn't seem to change anything if we consider private pilots rather than simply comparing professional pilots and commercial air travel.  After all, if we wanted to do a more rigorous comparison, then we should select the same standards and criteria for both modes, such as comparing (for example) Fed Ex air travel versus Fed Ex ground.  This would represent the same basic kinds of activity with a corresponding level of professionalism to render a comparison that would match the two modes of travel better.

    There is certainly no question that ground transportation is subject to far more random encounters and unpredictability than air transportation, which is precisely why I indicated that given that there was a much higher risk, then we should be experiencing far more accidents and fatalities than we already do.
    You are arguing that the risk of dying in a car accident is fixed for a given "trip", whether it's 2 blocks to the 7-11, or 3000 miles across the country. However, I'm saying the length (or time, which are roughly synonymous for the purposes of this discussion) of the trip is absolutely related to the risk.
    You need to be careful in assigning a probability as you have, since there is nothing to suggest that risk increases with distance.  In other words, I'm at no greater risk of having an accident one mile from my home than I am at 3000 miles from home.  Whatever risk is involved has nothing to do with distance.  The probabilities are not dependent on distance and consequently have no "memory" of time or distance.  Your argument as you have presented it is often a belief that gamblers have in assuming that future outcomes are determined by past behaviors (i.e. increasing the probabilities with increased events).

    What you are suggesting is that the longer I travel, the more encounters I have with others, and therefore I increase the likelihood that I may have an encounter with someone that proves fatal.  However, for that to be true, we would have to have something that indicates that we regularly experience accidents at some rate of encounters (i.e. 1 accident every 20,000 encounters).  If that were true, then we could also assess the risk of accident simply by the amount of time spent on the road.  In general, these probabilities of risk don't hold because they aren't actually random and consequently do not have an equal probability of occurring simply because of time spent on the road.

    As a simple example, let's consider driving a series of 10 blocks and consider what the probabilities of an accident are.  During the first block I would have 'x' probability of some encounter with another vehicle that could result in an accident.  However, in traveling the second block, there is no increase in the probability of an accident, because there is no "historical" connection to the probability of the first block.  Neither will there be for any of the other blocks in which I travel, therefore the probability of an accident is no greater traveling the 10th block than it was in traveling the 1st block.  To suggest otherwise is like arguing that a series of 10 heads must increasing the probability that a toss of tails is going to follow.

    You can certainly suggest that the probabilities are higher with more cars, or at intersections, or using whatever criteria you like, but then you'd also have the responsibility of establishing whether such risks have measurable probabilities associated with them, or whether they are simply a "gut feel". 
    Mundus vult decipi
    Gerhard - Maybe a better comparison is driving somewhere alone vs. flying solo in a small aircraft, because in both cases the purpose is the same, to get from point A to point B and back. I have heard that the risk per passenger mile for small aircraft is about the same as riding a motorcycle, in other words, riskier the driving. I'll see if I can find some statistics.

    Gerhard Adam
    You're absolutely right.  As someone that drives and flies, I can tell you that flying is dramatically more dangerous.  It's precisely why we don't have (and never will have) flying cars.  Operating in three dimensions is obviously more risky. 

    It's pretty obvious that you never hear of someone dying because they ran out of gas in a car, and yet it's surprising how frequently it happens in aircraft.

    Mundus vult decipi
    SynapticNulship
    It's precisely why we don't have (and never will have) flying cars.  Operating in three dimensions is obviously more risky.

    Ahh, but didn't you just say in that you can't compare the safety of planes and cars?  I think operating in 3D is risky only in some ways.  For example, for machines, flying is _easier_ than driving.  Autonomous flying vehicles are many decades ahead of autonomous ground vehicles.
    Gerhard Adam
    Ahh, but didn't you just say in that you can't compare the safety of planes and cars?
    Well, actually that's being kind of fast and loose with my comments.  The point was that comparing commercial airlines statistics with driving wasn't a valid comparison.  If I take a person in a car and then compare that individual in a plane (assuming the comparable experience), then we could certainly judge the relative safety of that individual in the two modes of transportation.

    The primary risk of operating in 3D is the height component since that tends to be the least forgiving.
    Autonomous flying vehicles are many decades ahead of autonomous ground vehicles.
    Whether they are or not isn't particularly relevant, because we'll never see traffic in the air that is comparable to that on the ground.

    Mundus vult decipi
    I found one set of risk calculations at http://www.meretrix.com/~harry/flying/notes/safetyvsdriving.html. They are informal calculations using data from "official" sources, so I think they give a reasonable estimate of relative risk.

    General aviation (small planes): 7.46 fatal accidents per 100M miles flown

    Driving: 1.32 fatal accidents per 100M miles

    Commercial aviation (ten year average):
    .05 fatal accidents per 100M miles flown

    Gerhard Adam
    That sounds about right.  I would expect commercial aviation to have a much safer track record, simply because of the professionalism and experience of most commercial pilots.  In addition, with the strict controls around airports, any encounter with other planes is also more tightly controlled to avoid a mishap.

    The other thing about general aviation faring worse than driving is that there is less margin for error, so we would expect to see more fatalities from flying (higher speeds and fewer opportunities for survivable crashes).

    Thanks for the link
    Mundus vult decipi
    barryleiba
    I have quibbles with some of your details (for example, one can't just count the number of instances of driving, because it's clear that driving 5 miles doesn't have the same risk as driving 500, even if you only enter the car once), but I agree with the overall point that making any broad comparison between the risk of driving/riding and the risk of flying is silly.

    What does make sense is trying to answer a more specific question, such as this: "I'm in New York, and I have to get to Chicago.  Statistically, am I safer if I drive, or if I fly?"  You can even add "if I take the train" to that, as well.  That's a sensible question, and, had we the right data, we might be able to come up with a valid, not-silly answer.

    Of course, I, at least, don't have the data.
    Gerhard Adam
    ... driving 5 miles doesn't have the same risk as driving 500 ...
    I have to disagree.  What's the basis for correlating risk with distance?  Certainly you might argue driver fatigue or the duration of the exposure to other vehicles, but essentially there is no reason to believe that one's risk is any less entering a freeway one block from your home versus entering it after a 500 mile cross country drive.

    Since the majority of accidents occur within a few miles of home, I would argue that it is precisely the short trips that represent the greatest risk of accidents.  Clearly we could argue about the speeds involved and whether they are lower risk, etc.  However, the essential point is that the oft-quoted airline statistic is basically nonsense.

    While I agree that the framing of the question is often silly, part of my point was that the airlines awards themselves "safety points" because they transport more people at one time and over longer distances.  Since there is zero correlation between distance flown and the likelihood of crashing, that skews the statistics to favor flying over driving.
    Statistically, am I safer if I drive, or if I fly?"
    Actually the only proper comparison would be to say "if I drive myself or if I fly myself" versus "if I'm driven or if I'm flown".  The former case would making driving (probably) a safer bet because of vastly more experience driving than flying whereas, the latter would likely end up being comparable (especially if we match similar modes of transportation - bus&airplane).



    Mundus vult decipi
    barryleiba
    What's the basis for correlating risk with distance?  Certainly you might argue driver fatigue or the duration of the exposure to other vehicles, but essentially there is no reason to believe that one's risk is any less entering a freeway one block from your home versus entering it after a 500 mile cross country drive.
    Oh, heavens; the mind boggles.  I'm not sure how to answer this, because I can't fathom how you can deny it.  Let's just simplify to "entering a freeway" as an example.  If entering a freeway is a point of risk, and you enter freeways 20 times on your 500-mile drive, you have 20 points of risk, compared with one or two on your local errand.

    You can call it correlation with distance, with time, or frame it some other way, but I can't see how you think that, say, driving around the block once, ten times... and driving around the block ten times at once... have any significantly different risk characteristics, just because in the first case you went in the house, came back out, and got into the car anew each time.

    Every minute you're on the road, every intersection you pass through, every merge you make, every time a truck passes you, every time someone changes lanes in front of you, and so on... you are at risk of a collision.  How can you possibly claim that the risk is (essentially) constant per trip, regardless of the length of the trip or where you go?  That makes no sense at all.

    Keep in mind that I agree with your basic premise that we're not comparing flying and driving reasonably, so there's no point in reminding me of that.
    Actually the only proper comparison would be to say "if I drive myself or if I fly myself" versus "if I'm driven or if I'm flown".
    What?  This is also silly.  I'm going to Chicago.  Either I'm going to drive, or I'm going to catch a ride on a commercial airline.  Each of those carries some risk of death that's statistically quantifiable, given access to the data.  Those are what make sense to compare, in deciding the death-risk of my Chicago trip.  Why on Earth does my flying myself have any bearing on it?
    Gerhard Adam
    Either I'm going to drive, or I'm going to catch a ride on a commercial airline.  Each of those carries some risk of death that's statistically quantifiable, given access to the data.  Those are what make sense to compare, in deciding the death-risk of my Chicago trip.  Why on Earth does my flying myself have any bearing on it?
    Because the reality is that flying is safe primarily because professionals engage in it.  If the same degree of professionalism were required of vehicle drivers, then corresponding levels of safety would also be attained.  You can clearly see the safety statistics begin to get worse when general aviation is included.

    The point here isn't to assess one individual's level of risk, but to address the larger question of whether flying is intrinsically safer than driving.  My point is that it is not. 

    Flying involves all modes of being airborne, just as traffic accidents include motorcycles, pedestrians, and bicyclists.  So to have the air safety statistics cherry-picked to only include commercial flights flown by professionals (and then multiplying the safety factor by the number of passengers) creates a false illusion of safety.  It certainly doesn't suggest that air travel is necessarily more risky than other modes of travel, but it certainly doesn't warrant the flippant attitude that air travel is so obviously safer than driving as to brook virtually no questioning.




    Mundus vult decipi
    barryleiba
    The point here isn't to assess one individual's level of risk, but to address the larger question of whether flying is intrinsically safer than driving.  My point is that it is not.
    I agree with you.

    You further say that the very question, in the general case, makes no sense, and is not answerable because there's no reasonable way to compare the general cases.  I agree fully with that.

    I was posing an alternative, more specific question, distinct from your thesis.  I maintain that my specific question makes sense, can be answered, and is a reasonable comparison: there are two ways I'm likely to choose to get to Chicago; three, if you count trains.  Compare my likelihood of death (or, perhaps, serious injury) from those three.  You might also have a fourth alternative, flying yourself.  That's fine, but it doesn't apply to me.

    I think this argument has about played out, at this point....
    "Why not compare flying with swimming? or hiking? However"

    Because swimming and hiking do no involve the use of internal combustion /jet engine's or tons of metal, plastic and rubber traveling at speeds humans are incapable of producing on their own?

    By the way i do no mean to be a jerk and take shots at a well thought out and written article, but that analogy is just wrong.

    I liked the article, and agreed with most of it, though I agree more with Barry Leiba's points. But thanks for writing the article.

    Anyway, I think it would help to reframe the question one more time. I don't believe people want the answer to "I'm going from New York to Chicago, should I drive, ride an airline, or get flying lessons and fly myself there?". Usually there are other constraints and the choice of transport is non-optional.

    Personally I say, "my lifestyle is that I fly about 3-10 times per year and I drive just about every day with a mix of short trips, longer commutes, and weekend jaunts for about 15,000mi per year. I'd like to die of natural causes, but I know not everyone is so lucky, so am I more likely to die in the air or on the road? Should I pay attention to the airlines and planes with the best safety record, or should I pay more attention to my driving habits?" and "am I better off in a terrestrial vehicle driven by an amateur (me) and surrounded by a bunch of wackos, or am I better off in an insane gravity-defying flying machine contraption under the control of a team of experts?"

    I do agree that passenger-miles are a dubious way to answer either my questions or the original ones in the article, so I really appreciate the discussion. I would also say that people really should look at swimming and hiking too. It's most useful to say: "Here's all the things I do regularly; for which ones should I spend the most money and effort to mitigate risk?"

    For me, the only metric I can use is the total number of accidents, and this graph http://www.anesi.com/accdeath.htm, tells me most of what I need to know. I get the biggest payback from getting enough sleep and avoiding stress before driving, not driving while intoxicated or texting, and getting someone to hold the ladder while I am painting. The rest of it I am just not going to sweat.

    Some people may skew their own risks by flying several times a week or not owning a car, and they can make their own judgments about how nervous to be while flying (or swimming). But I think if you live an "average" American lifestyle, then that graph tells you unambiguously what you should worry about.

    Have a safe New Year!

    Gerhard Adam
    Despite the direction that some of the comments have taken, let's note that the only real purpose I had in posting this was to dispel the presumptive position that "flying is safer". 

    This view was invariably stated as if it were a "truism" and beyond reproach.  So, hopefully, this discussion has done its part in quelling that position and raised the recognition that such simplistic assessments should not go unchallenged.
    Mundus vult decipi
    Wow... they will put ANYTHING on the internet, it seems. This article barely classes as psuedo-science.

    If 500 people want to travel from Point A to Point B (say, 1,000miles), they have a choice of all flying in one airplane, or driving in 250 cars (assume 2 people/car for this argument - the exact number isn't important). There is a certain risk of death associated with each option... a certain "expected number of deaths" - a very solid statistical concept. Vastly simplifying, but without loss of the core principal, the risk for the plane ("X") is the chances of it crashing * 1 flight * 1,000 people dying in that flight. For the cars ("Y") it is the chance of any one car having a fatal crash on that trip * 250 trips * 2 people/car.

    Which option do we expect to result in more dead people - which is larger - X or Y? A very sensible comparison. And, usually, the answer is that you expect fewer dead people on average if they fly... flying is safer.

    Further, passenger-miles is the most sensible way of comparing different modes of transport, without getting into the utterly unmanageable complexity of comparing different components of the travel (getting going, the bit in the middle, arriving, etc). If someone drives 10miles, he is at a certain risk; if he drives 100miles, he is at a higher risk, for a variety of factors - fatigue, more points of risk, more time on the road; probably not 10 times higher, but higher. Similarly, the most risky parts of a plane ride are the take-off and landing, but there is risk to the bit in the middle, so distance is also a factor. You might have been able to find some stats to show that take-offs and landings are by far the riskiest, so for flying in isolation passenger-miles may not be a great metric (in comparing large planes and small planes, you might want to normalize by take-offs, for example)... but, when comparing flying with driving you need an apples-for-apples comparison, and passenger-miles is the easiest way to answer our original question - is it safer to fly or drive from A to B?

    If you wanted to attack the comparison intelligently, a much more sensible/interesting approach would be to argue that getting to the airport needed to be considered, and that there are spin-off health consequences, etc. While I agree that "simplistic assumptions" should be challenged, it should happen with FACTS and SCIENTIFIC METHOD, not the Voodoo that You Do.

    Please take a BASIC (high-school level) statistics course before writing anything else.

    Gerhard Adam
    You simply don't know what you're talking about.  By your argument, all things being equal, two planes capable of transporting 100 people and 200 people respectively would render the second plane twice as safe (in passenger miles).
    ...the risk for the plane ("X") is the chances of it crashing * 1 flight * 1,000 people dying in that flight. For the cars ("Y") it is the chance of any one car having a fatal crash on that trip * 250 trips * 2 people/car.
    Again ... false comparison because you're comparing one event against 250 events and extrapolating them to the whole.  Rephrase the question and consider what the consequences of any ONE accident occurring and you suddenly find that your expected loss of life is 2 people for any incident, but 1000 people for the airplane crash.

    In the end, none of those numbers tells you anything about the actual level of safety.  It's simply unfairly comparing the number of events and then presuming that the effect on one person is the same as 1000 people.

    A more appropriate way to look at your example is to say ... what are the odds of 1000 people dying in a plane crash versus the odds of 1000 people dying in 250 car trips.
    Mundus vult decipi
    Gerhard Adam
    If someone drives 10miles, he is at a certain risk; if he drives 100miles, he is at a higher risk, for a variety of factors - fatigue, more points of risk, more time on the road; probably not 10 times higher, but higher.
    It's always interesting how this assumption is so readily made, despite having no basis for making it.  Where does this higher risk originate from?  If I travel 9 miles, does it occur during the 10th mile?  How about during the 8th mile?  Or pick any mile?  What differentiates the risk at any particular mile that would justify your adding them together and claiming that they are higher.

    You may think that I'm simply being silly in the mathematical comparison, but you stated that at 10x the distance, the risk was higher for travel.  Therefore, you explicitly linked it to distance and my question is to ask ... at what distance do we see this increased risk manifest if we can't actually identify any point along the way that would be responsible.

    By claiming that the risks increase, you're claiming that the risks are cumulative over distance.  This would require that we assume a fixed probability for any given mile, then the total distance would be the sum of all the probabilities, you might consider why mile 8 should have a "memory" of the probability for mile 2. If I'm on a 100 mile trip, and I've already traveled 99 miles, why should my risk be any higher during the last mile as it was for the first [based on distance alone].

    Of course, since you bring up fatigue, then your point is that the probability of an accident has nothing to do with distance at all, but rather the attentiveness of the driver.  Fair point, but you're simply changing the metrics now.  After all, is a drowsy driver less accident prone if he begins that way, or only after he becomes tired?  Is it the monotony of the drive or the lack of sleep?  Certainly we wouldn't consider an airplane piloted by a drunk pilot safer simply because its an airplane.  We recognize the responsibility of the individual "driving" the vehicle.

    All the points regarding the qualifications of the "driver", the density of the traffic, the variabilities in individual behaviors, etc. etc. etc were all brought up in the article, but again, have nothing to do with distance traveled.

    So let's see these high school statistics of which you speak so fondly.
    Mundus vult decipi
    I'm not sure why I'm breaking my policy of not arguing with those who don't know, but here I am...

    Your second post...
    Let us make the assumption that the risk of every mile is the same... no "memory". This isn't a good assumption if we know something about the specific route (e.g. is it more dangerous to be merging onto a busy highway at mile 3 than driving a quiet, open country road at mile 98), but it is a good assumption for the generalized case, where we have no reason to believe any given mile is more or less dangerous than any other, and we just want to get the general principal). Let us say that there is a 1:1,000,000 chance of being killed on each mile (fictious number... no idea what it is). Then, if we drive the two miles, we have two opportunities to be killed, so the risk is 2:1,000,000... the risk is higher. Those of us familiar with statistics make assertions like mine without back-up because they are self-evident, trivially obvious, and almost axiomatic. Those who try and determine which mile the risk is in don't understand statistics (if you could pin-point the risky mile, DUH!, you just wouldn't drive that mile, and there would be no risk) It doesn't matter if you drive a given 1 mile route twice, two 1 mile routes once each, one 2 mile route, or we are considering a sub-set of a 10,000mile around-the-country odyssey.

    Your first post....
    1. I guess I'd better give back all the money I've been paid for risk modelling and managing remote-area aviation deployments over the past decade. (Sarcasm). Yes, I do know what I'm talking about. And, I seem to remember these concepts being covered in my 6th grade extra lessons (seriously... although it may have been 5th grade), so maybe I was wrong about high-school stats.
    2. The risk (in this context) of a form of transport is the risk to an individual who chooses to travel that mode of being killed on a particular trip. Alternatively (and completely equivalently) you could postulate that, if equal numbers of people were to choose to use each of two or more forms of transport, how many people would you expect to die using each form (if 1,000,000 people each drove, flew, and walked from New York to LA, how many would die on the trip in each group... if 500 people died driving, the risk of driving is 500:1,000,000; if 100 died flying, 100:1,000,000; if 2,000 died walking, 2,000:1,000,000, etc). Risk can either be observed as 1,000 individuals making 1 trip all together, or 1 individual making 1,000 trips, or 2 individuals making the trip together 500 times... statistically these are the same (this makes an unstated but obvious assumption that the number of people vs the number of trips choice does not change the dynamics or characteristics of the mode of transport... it is simply a mode of observation - this is a good assumption in this setting, where many, many people fly, drive, etc anyway). So, if a car carrying 2 people has a 1:1,000 chance of crashing, the risk of dying while driving that route is 2:1,000, or 1:500 (the risk to each individual of dying... cars carrying 1, 4 or 8 people would have different risks to the individual, assuming the risk of each car crashing remained the same... but we are getting into "advanced" statistics here). If the chance of a plane carrying 500 people crashing is 1:100,000, the risk to each individual is 500:100,000, or 1:200 - you have a 1:100,000 chance of 500 people dying, or a 1:200 hundred chance of one person dying (yes, you would never get just one person dying, but that misses the whole point of statistics). This would make planes much more dangerous, as, if equal numbers of people flew and drove, you would expect 2.5 time more people to die in plane crashes if equal number of people drove and flew. However, we find the risk of dying in a commercial plane crash is between 20 and 60 times lower the in driving the same route (the range is because different assumptions and datasets are used by different people, but they all show you have a much lower risk of dying in planes.)

    The mode of crash is irrelevant - all we are interested in is the chance of a crash. The discussion of amateur driver vs. professional pilot, or fatigued driver vs. drunk pilot, etc is of not use to anyone... well, no-one who wants to do anything useful. True, the next phase in the discussion is to look at the modes of crash for each form of transport, but within a particular form of transport, so as to see which risks can be reduced, and then (if you are feeling very ambtious) to try determine which form of transport can be MADE INTO the safe. (I suspect that, even with full automation of the private road fleet, flying as it is today will still be safer than road can ever realistically become... but that is gut feel, and something which I would love to have a discussion on with someone who understands risk.)

    Well, I have spent enough of my time on outreach and education. Hopefully I've helped one or two readers avoid the misunderstandings of the article. I'm outa here, not to return.

    Gerhard Adam
    Then, if we drive the two miles, we have two opportunities to be killed, so the risk is 2:1,000,000... the risk is higher. Those of us familiar with statistics make assertions like mine without back-up because they are self-evident, trivially obvious, and almost axiomatic.
    Well, it's good to know that you don't feel any need to establish correlation and causation.  You can simply assume it as "axiomatic".  You make statements like the "risk is higher" without establishing the slightest reason to accept that statement.

    Whenever someone tells me something is "trivially obvious", that's usually a good sign that they're totally wrong and haven't got a shred of evidence to support their position.
    Those who try and determine which mile the risk is in don't understand statistics...
    No, it generally means they don't understand driving.
    ...if 500 people died driving, the risk of driving is 500:1,000,000; if 100 died flying, 100:1,000,000; if 2,000 died walking...
    This was definitely an interesting assertion.  So, in your "axiomatic" way you simply declare that death is a function of distance, regardless of the mode of travel. 

    Yeah ... I think you should give the money back.
    Mundus vult decipi