Note: I wrote this a few years ago for publication internally within Boeing. I have since decided to publish it more broadly, removing links to content that is only accessible in the Boeing intranet. I did not re-gather data about aviation accidents, so the data mentioned on this page are a few years old. Despite their age, I think they still make the point as clearly as ever. I also fixed broken links, and added some new links, for further reading. Enjoy! - BNM

Whenever there’s a commercial aviation accident, the topic comes up between me and my friends regarding the risks of flying. Aircraft systems, structures, operating procedures, engines, and other essential things, are apparently still imperfect. There is room for human error. Weather is still dangerous. Safety margins may be insufficient. And so on. As an “expert”, I am invited by my friends to remind them about the advances made in all of these areas, and about the astronomically small chances of being involved in a fatal aviation accident.

You probably already know that commercial flying is one of the safest things you can do. It’s vastly safer than driving, something that most of us do every day without a second thought. If you’re flying, the odds of an accident “totaling” the airliner you’re in (i.e. causing enough damage that the cost of repairing it would be higher than the value of the airplane) are literally one in a million... and even if that does happen, you still have something like a 95% chance of surviving such an accident with minor injuries at worst. Commercial flying is really astoundingly safe, when you think about it.

But what you may not realize, and what is really interesting to think about, is how commercial aviation is still getting safer and safer. Even though fatal accidents are rare, their frequency is still dropping. (Fatal accidents in small turboprops in the US, and in larger jetliners flown overseas, happen maybe once every year or two. Ground crew fatalities happen in the US a couple times a year. But if you look at flights in jet airliners in the US between November of 2001 and July of 2013 - that's twelve years - how many passengers were killed? Zero).

With each accident, the FAA and the manufacturers and the airlines all revise the way they do things. Airplanes are built a little differently, so that whatever failed last time will not fail again. Pilot training is improved, so as to make sure that pilots stay away from behaviors that have caused accidents. Mechanics may have to do additional work and check more things to make sure they have a thorough understanding of the health of the airplane. New FAA rules require new airplanes to have additional capabilities, and to demonstrate them via tests (often expensive and destructive) before the FAA will certify the airplane for commercial use. And so on.

Each of these things has an impact on air safety: When you know the cause of an accident, you can bet that it will NOT be the cause of any future accidents (although, occasionally, a problem will cause two similar accidents before preventive measures are put in place, and sometimes the preventive measures aren't super effective... but that's the exception rather than the rule).

Boeing (and each of the other airplane manufacturers) invests a lot of resources to understand what causes accidents, and then to share these findings with the rest of the world – via websites, conferences, magazine articles, awards, seminars, partnerships with the rest of the industry, blog posts, and frequent talks given to Boeing employees about different faces of commercial aviation safety: the development of structural inspection programs, the testing that proves that computer simulations of accidents accurately predict what actually happens, the required flight-testing on each new airplane, new regulations about crashworthiness and lightning-strike protection (including information such as: what caused them to be implemented, what they mean, what we're doing to comply with them), the information made available to the pilot in the cockpit and how it could be presented differently so that pilots can make better decisions more quickly and stay further from trouble... I really wish I could link to the huge lists of videos and transcripts and Powerpoints that are available internally at Boeing from all these insightful talks, given by brilliant people who have dedicated decades of their engineering careers to making aviation even safer. I watch them regularly (I even organize some of them) and it is truly amazing work.

Bottom line: Everyone at Boeing puts as much thought as they can into making airliners as safe as possible, and then do what they can to spread the word about their concerns and solutions. This is also true in other aircraft manufacturers, in the FAA, and in the airlines. These multiple parties are often in disagreement about how something should be designed, what kind of paperwork should be required to keep track of this or that, who should pay for what... but when it comes to safety, everyone knows that it's in their best interest to not cut any corners and to leave no concerns un-addressed.

From my own personal point of view: My group helps define how Boeing engineers should go about making sure that our airplanes are damage-tolerant. This means that if there is fatigue cracking, accidental damage, corrosion, etc (and there will be, because people and materials and manufacturing techniques are not perfect, and nothing lasts for ever), we determine how to make sure that the airplane will continue to fly safely despite this damage, and that the damage will be detected before it is fatal. (For example, a hole on the side or top of the fuselage will not bring an airplane down… but you already knew that). The rules that outline the specific requirements for damage tolerance are all “written in blood”; For each rule, I can tell you what fatal aviation accident led to its creation, and I can assure you that an accident like that will almost certainly not happen again. The rules that govern my job actually work, are actually effective in preventing future loss of life.

(Unlike, say, the rules about passenger screening at airports, which were supposedly put into effect to prevent terrorism, but - it is easy to show - are actually worse than useless . It’s probably the fact that I work with reasonable and effective rules every day that makes me especially frustrated when I encounter rules that are not so reasonable. But I digress).

Boeing's entire "structures design philosophy" basically looks at every conceivable way in which airplane structural mishaps could lead to an unsafe situation, every conceivable way to prevent that from happeneing, and every conceivable way to check whether it's working. Here is a very long presentation that covers all these topics, including all the kinds of testing we do, how we plan inspections and maintenance, what we have learned from the history of accidents, how we understand the airplane's environment (from loads to corrosion) and how we draw up requirements to make sure the airplane can survive whatever it encounters, how we categorize structure in terms of the consequences of it failing and the methods we use to ensure that any loss in structural capability will be detected before it can be fatal... People in my group insist that every new structures engineer at Boeing spend a week in a classroom so that we can go over these slides with them and lecture them about how this safety-mindedness is at the foundation of everything we do.

That’s all great. But what if you want to get this across to someone and you don’t have the time to dive into all these details? Can this be communicated in a couple of easy graphs and statistics?

Why, yes, it can!

Below are my two favorite graphs about this.

(I made both images; the first with data from Jet Information Services, Inc, the second with data from the New York Times. But don't give me too much credit; others have displayed this same data in pretty much the same way. I just created my own images, rather than displaying theirs, for the sake of respecting their intellectual property rights. Luckily for me, you can't copyright data).

The first graph shows three things:
- The number of hull-loss accidents per year; Those are accidents when the airplane is “totaled”, i.e. damaged so badly that repairing it would exceed the cost of the airplane…
- The number of commercial flights per year, and…
- The number of hull-loss accidents DIVIDED by the number of flights; This gives you a statistical accident rate, i.e. the probability that any individual flight will end in a hull-loss accident.
(This only covers commercial flights in the US, in jets weighing 30 tons or more… basically anything that’s no older than a 707 and no smaller than an Embraer 170).

As you can see, the number of hull-loss accidents in the US has been hovering roughly between 15 and 25 per year, since the dawn of jet aviation. The number of flights in the US, however, has risen from less than one million per year to about twenty million per year. This means that the chances of any individual flight encountering a hull-loss accident is a small fraction of what it was back in the 707 days, literally about one in a million.

And as several recent hull-loss accidents have demonstrated, they can be perfectly survivable when they happen to modern airplanes operated according to the most up-to-date procedures. When everyone on board survives these nasty-looking accidents, most newspapers call it a 'miracle', and most engineers disagree.

Now, if you look at that accident-rate curve, you see that it starts out relatively high, drops pretty steeply over the 60s, still decreases a bit in the 70s, and by the 90s it looks very low and very flat. This might imply that there have not been major improvements in air transport safety in the past 20 years, because things are so safe, they can hardly get safer.

Right?

Wrong.

If you zoom in to that “low, flat” period of the curve, you’ll see the same trend in falling accident rates. It’s like a “half life” kind of exponential decay; Even when there is practically nothing left, rates still fall to half of their previous values every certain period of time. Of course, by that later period, there is so little there that when you zoom into it, you can see each individual accident.

This graph speaks for itself. The data are of a slightly different kind: It shows fatal accidents (not hull-loss accidents) in any size commercial airplane (including small turboprops, anything with paying passengers) operated by US airlines (i.e. including accidents that take place outside the US). But the numbers are similar, and the trend is exactly the same "half-life decay":

Each block represents one accident, and each pile of blocks is for one year. Note that the blocks are scaled so that the top of each pile represents the number of fatal accidents per million flights that year; if there were 15 fatal accidents in a year with 10 million flights, then the blocks are scaled so that the top of the pile is at "1.5". In other words, the blocks get slightly smaller as there are typically more flights per year and each block becomes a smaller fraction of the overall total flights, statistically. But even if that were not done, you can see how the number of blocks decreases pretty dramatically.

Also worth noting is that many of the accidents that caused 1-2 fatalities involved only ground crew who were refueling the airplane, handling luggage, etc, and not passengers. The flight line can be a dangerous place.

As the two graphs above show, air travel is not only ridiculously safe, it’s still getting safer! All this hard work by regulatory agencies (coming up with new rules that prevent accidents from happening again, even when they increase the cost of flying), Boeing (building our airplanes better, even when this requires mountains of paperwork that literally take years to complete), and the airlines (more closely monitoring more things about the health of each airplane and about the pilot’s operating procedures), is all paying off. The numbers don’t lie. Flying just gets safer and safer and safer.

To learn more, I would highly recommend you check Boeing’s aviation safety website, the FAA's safety webpages and lessons-learned and statistics databases, all of which are very informative, written in accessible language, and easy to navigate. Do check them out.

If you dig around, there’s even more great data out there to be found. For example, you can see here and here how each new generation of airliners is progressively safer than the previous one (other than for an initial spike of accidents right after each new airplane type is first introduced… with the remarkable exception of the 777, which has only caused one single fatality in its entire operational history. We must be doing something right. I expect that we’re still doing it on the 787).

Next time you ask yourself whether flying is safe, whether the work done by manufacturers and regulators and airlines and controllers and maintenance shops is paying off, whether anyone is cutting any corners when it comes to safety... Have no doubt: Everyone in the commercial aviation industry is working as hard as they can to keep you as safe as they can, literally eliminating every foreseeable disaster scenario, and right now they are more successful than ever. YEARS go by between fatal jetliner accidents in the US, and this number of years will just keep increasing, even though the number of flights that take off and land every day keeps going up and up. Deaths due to cows happen more often than deaths in commercial airplanes.

So whenever someone says "Have a safe flight!" when you go to the airport, do what I always do. Reply: "They're all safe" :]

Blue skies and tailwinds!

- Bernardo