Reduce Your Odds Of Dying In A Plane Crash

Wednesday, July 1st, 2009

A round-up and mash-up of data on fatal plane accidents. Cross-referencing the data (unscientifically) reveals the statistically most dangerous flights to take.

UPDATE 15th August: Visitors pointed out a serious error in this diagram relating to the number of fatal accidents by airline. I’ve fixed this now. Thanks to all who pointed it out.


Books and Store

Our Beautiful Books - Information is Beautiful Information is Beautiful Store

Show Comments ( )

  • Peter Douglas

    Yeah thanks mate. Now I panic if I can’t get a seat in the rear third of the plane.

    • david


  • Charlie Emberson

    Good job.

  • Jons

    Kind of baffled by the fatalities as a ratio to the number of planes in service. For example the DC-9 is 19:1, surely this can’t mean that one DC-9 out of every twenty will crash…

  • Garrett

    I didn’t see much meaningful information in these charts. Is the Yugo safer than a Volvo because they are in fewer accidents? It’s safer to fly in Uganda than the US? United is more dangerous than Joe Schmoe’s Airline? I have to question the assumptions behind every graph. How about something like deaths per 100,000 miles flown by country/airline/model?

    • david

      Fair comments. This was meant as an ironic extrapolation of the data, rather than a guide. I like your suggestion for deaths per 100,000 per model or airline. Any tips for data source?

  • Andy

    …how can someone die from blogging?

  • http://PlaneCrashwrong Steve B

    Your graph lists Qantas as having 11 fatal air accidents, but you’re WAY wrong there. It’s had 0.

    Your Wikpedia “source” was “List of accidents”, not “List of fatal accidents”. If you’d cross referenced it with the more detailed Wikipedia article: you’d see that the “fatal accidents” all have descriptions like so:

    “September 23 – Qantas Flight 1, a Boeing 747, overshoots the runway upon landing in Bangkok, Thailand; none of the 410 people on board is seriously injured.”

    • david

      Thanks. I’ll check this.

  • Steve

    You say that the higher ratio implies a safer plane but the 777 with zero fatal accidents has the lowest possible ratio of zero. According to your heading, this is the most fatal plane.

    It also seems there should be some accounting for flight miles and/or number of flights. My preference would be for flights (take-off/landing) instead of miles since it correlates with the risk of a flight and does not make larger planes flown mostly on long flights appear safer by virtue of flying more miles but fewer trips.

  • Peter Manson

    Your “reduce-your-odds-of-dying-in-a-plane-crash” graphic contains an error.

    Under the heading “Safety Record – Fatal accidents by airline (5 or more)” you have summarized information from Wiki’s “List of accidents and incidents involving airliners by airline” page – which includes all incidents, not just those involving fatalities.

    Factually correct information is beautiful, sloppiness isn’t so appealing.

    Peter Manson

  • Kent

    Just echoing what Garrett said. Much of this data needs to be normalized by number of miles flown or number of flights made. For example, are there more crashes in the summer because there is a lot of bad weather or simply because more flights are made?

  • Blackmore

    As the above poster said, Qantas has never had a fatal accident involving a jet aircraft. They did have a number of fatal crashes involving propeller craft, though the last of those happened in 1951. I only even noticed this error because I was surprised how large the Qantas logo was in the fatal accidents section, and due to being Australian happened to know what an excellent safety record Qantas actually has.

  • david

    @blackmore, @saddened, @petermanson

    Yes, this is a serious error. Thank you for pointing it out. Quantas does indeed have a spotless safety record. I have corrected and redrawn the image. Please accept my apologies.

  • Nik Sargent

    I love your visualisations – they really appeal to me – it’s a shame Microsoft ever launched the curse that is Excel on the world; seems that’s how all data comes these days.

    However, I don’t understand your representation of number of fatalities per plane in service, which you say is better if higher… My instinct tells me the other way round – i.e. if i have 1 fatality and 100 planes in service (1/100), that is not as good as 1 fatality with 200 planes in service (1/200) as that suggests each latter plane is half as likely to have a fatality. So, isn’t lower better?

    Maybe you just quoted it as “number of planes in service per fatality” without changing the order of the words? Then higher would be better… :)

  • peter

    What’s going on with the data in the first slide?

    None of the data in your google docs spreadsheet agrees with your cited source. Your numbers are closer to the number of accidents, not fatal accidents. The title of the slide is “Density of Fatal Accidents 1942-2009″ so I assume you are stating that, for example, the US had 2613 fatal accidents but your source only lists 1079.

    So did you pick the wrong data or title your slide incorrectly?

  • peter

    Also, the sixth slide, “Bad Flights: statistically speaking,” falls victim to the fallacy of composition.

    It looks like you just took the most dangerous locations, the most dangerous airline, and the most dangerous plane and mixed them all up to assume that they would make for the most dangerous combination. It’s misleading, if not irresponsible, to interpret and present data this way.

  • Andy Crossett

    Interesting graphs and statistics.
    I echo the suggestions of correlating to the number of takeoffs/landings.
    Also, the ratio of accidents per number of aircraft in service could be misleading.

    For example, the DC9 series may have had more aircraft in service in the past than it does now. So the number of aircraft in service should somehow be normalized to take into account how many were in service at the time.

    To clarify, and end of life aircraft may have few still in service currently, but still has all the accidents over its product lifecycle.

    Great work, though!
    Best regards!

  • SamM

    The possibility that you will potentially die in an airplane increases by 100% once you set foot in one. The possibility that you will potentially die in an airplane crash increases by 100% once that airplane takes off. Just sayin’…

  • David TallEnglishman

    Hi there

    Great set of images. Couple of comments.

    Your Final Destination Density Chart is pretty meaningless without knowing what percentage of flights went to that country. What if 0.1% of flights to the US crashed whereas 80% of flights to Papua New Guinea crashed? That cannot be deduced from your chart.

    Same with the plane type. Your chart suggests that the 737 is the least safe type, whereas it may actually be the safest (e.g., say 3 million flights with 60 crashes is much safer than 10 flights with 7 crashes).

    Think about what you want to convey!

    Just one minor additional comment: I don’t think it’s supposed to be “Pre-fight check” at the top!

  • Tom in Raleigh

    When normalizing your data by passenger mile or seat-mile, you will likely find that the accident density of the United States resolves to a fairly low accident rate. If these were represented, then, as circles sized relative to each other, the circle for Russia/former USSR would be much larger, and two countries–Nigeria and Indonesia–would grab the viewer by the throat and say “for the love of God, don’t fly in these countries unless you want to die very soon.”

    The best I could do for data was at, for a quick and dirty search. You might find something useful at

    I think your data sources are quite good, but they just don’t go far enough to normalize the data. Good luck with tweaking with this!

  • Manu

    “Ratio of fatal accidents to number of planes in service (the higher the better)”
    Am I crazy or is this the other way around?

    If you meant, “the higher the better [chance of being in an accident],” perhaps this needs to be restated. As it stands, it resembles the sort of thing Tufte rants about in his criticism of the NASA slide using the word “significant” in contradictory ways.

  • Alex Zuzin

    Some of the data here doesn’t make any sense to present in 2009, e.g. the odds of dying in a DC-9 for an average Internet user are equal to their odds of actually seeing one. ;)

  • Bob Chapman

    I don’t see any comparison based upon type of plane versus number of air miles flown.

    A plane model with fewer air miles for the entire fleet of that make should have a lower rate than a plane model with more air miles for the fleet. Planes on the ground don’t usually have accidents.

    There should also be some comparison with number in fleet versus air miles of fleet, to get an idea of a model is being worked more than others.

  • Garrett

    This is a good place to start for basic airline safety statistics:

    For non-airline, look at the Nall report from the Aircraft Owners and Pilots Association (

  • Thomas R

    This link seems to argue against the Embraer stat. of 0 crashes (or did I misread your graph?)

  • Abbie Mitchell

    At the top of the page, where it’s supposed to say “Pre-flight check,” it says “pre-fight” check.
    Just thought I’d mention it…

  • Michael Ross

    “Ratio of Fatal Accidents to number of planes in service”

    In service NOW? Or in service EVER? It looks like you have used the former but it should be the latter, if anything.

    Either way, since the number of planes in service changes in time, the statistic is pretty meaningless. Accidents are distributed over time. The risk of EACH single accident on a particular date with X planes of that type in service on that day is meaningful, but you cannot combine statistics for ALL accidents occurring with that plane model.

    You would have to “sum”, for each and every day, the ratio of accidents that day based on the number of planes in service that day. Again, meaningless.

  • Edwin Hermann

    You forgot Air New Zealand’s Mt Erebus crash in 1979. It killed 257 people yet Air NZ is not mentioned in the list of airlines with 50 or more deaths. Not sure how accurate your data is.

  • Todd Curtis

    I like the visualization of the data. Airline safety statistics can make the eyes glaze over if you’re not careful, so anything that summarizes data in an easy to understand fashion is a good thing. I’m a bit partial to the way I display data at, but there is always room online for different ways to deal with describing airline risks.

  • Todd Curtis

    A recent video that uses the method at lists that top five airliner models with the best accident history:

  • Salo

    Thanks muchly for your work on the aviation deaths, but wouldn’t the country of origin be more relevant to the likelihood of a crash than the destination country?

    Eg. I live in Australia, and all the aircraft used and owned by our national carrier Quantas and local carriers Jetstar and Virgin etc are all serviced and maintained locally in Australia.

    I feel (without justification to offer!) that local service and ownership would be a significant contributor to better safety. Local Australian ownership means the age of the aircraft is lower and local repairs mean a better likelihood of quality spare parts and quality work, as opposed a greater risk of questionable second hand parts from (say) Russia.

    If I’m right, then to rate Australia as a destination as you have done, you would need to work out where the aircraft that crashed here came from and who serviced them…



  • jemauvais


    “Thanks muchly for your work on the aviation deaths, but wouldn’t the country of origin be more relevant to the likelihood of a crash than the destination country?”

    Actually I think the country of destination is equally relevant. Crashes due to mechanical issues should be an issue with the originating country but crashes due to pilot error or poor instrument approach design would be more relevantly linked to the destination rather than the origin of the flight.

    American Airlines Flight 965, Korean Air Flight 801, and USAF IFO-21 are examples of what I mean.

  • Capt Tom Bunn LCSW

    Many things have changed in the last fifty years of aviation. We transitioned from props to jets. We have ILS to provide precision guidance to the runway. We have automatic landing which takes pilot error out of landing in bad weather. We have the Ground Proximity Warning System to warn when headed toward a crash. We have TCAS to prevent mid-air collisions.

    When you go back too many years, you mush together aviation as it is today with how it was years ago. Few, if any, who see this site will be interested in history. Rather, they want to know about CURRENT safety. When the focus of these statistics goes back before 1990, that is far too removed from the level of safety we have today. So your presentations fail to reflect what people want to know that can help them make a decision when flying.

  • Richard Whitehouse

    Yeah, this is just pure statistical crap. There is no referencing – you imply that the USA has the highest risk, where as it is actually the case that it just has more flights per year. Without any reference to per flight or per mile, this diagram is useless, and represents a complete and utter lack of understanding in the source matter.

  • Karl Wohlgemuth

    Are you aware that driving a car for more than 20 km is already more dangerous than a direct flight? If you have arrived at the airport by car you have already survived the most dangerous part of the journey. So why even think about plane crashes? ;-)

  • gifts for her

    The number of accidents per destination apparently was not set in relation to the number of flights and the number of accidents per model was set in relation to the number of planes in service (I hope not currently but overall) but not to the time of service of each of these planes. The latter especially means that newer planes seem safer just because they didn’t have so much time to crash.

  • Collin Webster

    Aeroflot never used DC-9. Perhaps you meant DC-10 (Aeroflot Cargo).
    The Clipart for DC-9 is a Fokker F-70.
    See Wikipedia for further information.

  • cuvy

    So, what are the stats on people dying in car crashes while driving from/to an airport to catch a flight?

  • Brian

    You need to include relative proportions rather than absolute proportions in your plane crash data (i.e. plane crashes per # of flights). Or better yet, include both.

  • Tanya

    I liked the visualization a lot, but if I start to think carefully and analyze the info…
    It is like biased statistical review..
    You cannot rank mortality for different planes without taking into account how many passenger they carry…of course one crash of a huge airplane will have more deaths then several crashes of small planes… and in this case pictures are working to enforce the wrong perception,… as I see it.

    If you wanna make a person a fool – give him/her just a piece of all available information…

  • Chris

    Since the point of the graphics is to discuss fatalities, shouldn’t the focus be on the number of fatalities rather than the number of fatal accidents?

  • Randall Wisby

    I have a hard time believing that a person is more likely to die in an airplane crash than to be stung by a bee. Most people I know have been stung by a bee. Over my lifetime I have only known a small handful of people who have died in plane crashes. I highly doubt that figure.

    However, otherwise this is fascinating stuff!

  • Ashley

    I’ve actually flown from the US to Russia through Aeroflot in August (not sure the type of plane, though). Won’t be showing this chart to my grandmother any time soon.

  • Ira Higgelbottom

    Great info i will take this into consideration

  • Johnson

    I don’t really think this chart is doing the public a service. They are likely to take it seriously. The ratio of fear to actual danger is probably higher for air-flight than anything else.

    I really love the work you do but this is both useless and paranoia-inducing.

  • Harold

    How do I know that this information is actually valid? For all some viewers know, this could all be made up to make people fear/relax about these topics. Also, I’d like to know how these figures that calculate the chance of people dieing in different ways are made up. I mean, how the heck are you supposed to find out the chances of being attacked by a mountain lion? 0_o

  • Chris

    Bad data is bad data.

    What is the usefulness of a total plane crash number by country/plane? 0. You need the per flight number. . .i.e. is the 737 a deadly plane or the most popular plane ever?