Comparative Rail Safety

Using Wikipedia’s list of rail crashes and its UIC-sourced list of rail passenger-km by country, one can compare different countries’ mainline passenger rail accident fatality rates. The US turns out to be the least safe among the regions I’ve checked, even worse than India; much-maligned China comes out first.

I constructed the list below by averaging accident rates going back to 1991, to smooth out fluctuations coming from low-frequency, high-impact disasters. Crashes involving only freight trains are ignored, and pedestrians and car and bus passengers struck by passenger trains are included. Bombings are excluded, but sabotage incidents leading to accidents are included.

China: 876.22 billion passenger-km/year, 317 deaths over 20 years. This is one death per 55.3 billion passenger-km.

Japan: the UIC claims 253.55 billion passenger-km/year, which only includes JR companies. Figures including private railroads and excluding subways range from 360 to 395.9 billion passenger-km; I believe the higher number since it is slightly less dated. Over 20 years there have been 154 deaths, so this is one death per 51.4 billion passenger-km. Including subways would put Japan on a par with China.

EU-27: 386.24 billion passenger-km/year (presumably mainline only), 603 mainline deaths over 20 years. This does not include 155 deaths from a fire on a funicular. This is one death per 12.8 billion passenger-km, or 1 per 10.2 billion if the funicular fire is included. This varies a lot by country: the safest European countries, such as France and the Netherlands, are on a par with China and Japan, but the EU average is pulled down by Germany (due to Eschede) and the periphery.

South Korea: 31.3 billion passenger-km/year, 93 deaths over 20 years. This is one death per 6.7 billion passenger-km. Here the mainline-only rule is a problem because a) the Seoul subway is even more integrated with commuter rail than the Tokyo subway, and b) a subway fire in Daegu killed 198 people.

India: 838.03 billion passenger-km/year, 2,556 deaths over 20 years. This is one death per 6.6 billion passenger-km.

US: 27.26 billion passenger-km/year (both Amtrak and commuter rail), 159 deaths over 20 years. Note the rate is more than twice that of China per capita, let alone per rail passenger. This is one death per 3.4 billion passenger-km.

For comparison, the US road network has 33,000 accident deaths and 7.35 trillion passenger-km per year, which is one death per 220 million passenger-km.

On a closing note, China not only has the safest passenger trains, but also by far the busiest tracks. Freight density beats that of the US and Russia and passenger density beats that of any European country.


  1. Stephen Smith

    Frankly, I don’t know that I believe any statistics coming out of China. I know that this sort of fudging was endemic in the Soviet bloc, and while China has obviously liberalized economically more than, say, Romania, they don’t seem to have progressed politically very much.

    (I feel the same way about stats on prison populations – I realize the US jails a lot of people and that China probably doesn’t even have the resources to lock up as many people as we do, but it just seems lazy to take authoritarian governments at their word with things like this.)

    I’m also curious – do you know anything about British accident rates right after deregulation, and has it improved since? I’ve heard that a few high-profile crashes led to the adoption of some overly strict safety regulations, but I’m curiou as to whether these regs were really needed in the first place, and if removing them (which did happen, right?) had any effect, either.

    • Alon Levy

      I don’t know about British rates, but it’s probably been done before and could be Googled. If not, Wikipedia’s list plus a time series of passenger-km could help.

      I know what you’re saying about China. India might be the same – I’m not sure I believe it’s safer than the US. On the other hand, those numbers seem to be based on media reports; China is very public about flogging bureaucrats who fail or are corrupt, and I don’t know if they control the media there to the point of not talking about train crashes. So I’d trust the numbers more than I would numbers about executions or military spending. I think it’s likelier that there are many small accidents that are unreported or whose reports don’t make headlines that Wikipedia editors read than that China is lying here.

      • Stephen Smith

        I think the Chinese are a lot better at keeping information out of the press than you think. My dad is a demographer and he used to go to China and India to try to quantify the extent of female infanticide, and he said that the Chinese were much more obstructionist than the Indians. And outside of the provinces where it happened, I think that most Chinese – especially the rich, coastal citizens – are mostly in the dark about, for example, the schools that fell down during the last big earthquake.

        As for public flogging, again, I’m not sure this happens as much as they lead us to believe. There was one high profile case I remember where someone was actually put to death, and of course the head of the HSR head was sacked recently and it was a pretty big story, but I get the feeling that these were for public consumption, and that petty corruption is tolerated, and obfuscation of embarrassing facts is even encouraged if it’s done well. As for the HSR guy, if I’m right that they don’t actually punish officials that often, then their HSR lines are in a lot worse shape than a lot of people think, and he did something especially bad (what I’m hearing that the construction is shoddier than it was supposed to be and he and his cronies pocketed the difference).

        • Alon Levy

          Stephen, I think you misunderstood. I’m not saying China Railways keeps bad statistics. I’m saying that because my data source is narrative and sourced to media reports (which is why it misses most trespassers – though I believe the multi-hundred total per year has to include freight trains) it could miss crashes not known to Wiki editors.

          • Alon Levy

            Never mind, I misread your comment as “keep statistics,” instead of “keep statistics from the media.” Though I’m still unsure how effective it is at keeping individual major accidents under wraps. Unlike mulad, I’m fine with math at 3 in the morning; that’s when I prove theorems. It’s at 9 that I suck.

      • J B

        I doubt the Chinese gov’t is able to intentionally keeps rail accidents out of the media. Many Chinese newspapers report on industrial accidents and even corruption cases (though there may have been a clampdown on these recently), so I don’t see why not rail accidents, unless small accidents get ignored by everyone except local newspapers. If anything the trend is less to try to hide incidents, which is difficult and potentially embarrassing, and instead to present them in a way that doesn’t embarrass the government, so for example a lot of Chinese people believe starvation during the Great Leap Forward was due to bad weather rather than bad policies. China is open enough that news about major accidents- particularly rail accidents- would spread quickly even if the media wasn’t permitted to report on it. Accidents are much more difficult to hide than things like infanticide, since infanticide can be done in private, has no direct effect on those outside the family, and to compile statistics you need to record many separate incidences.

        • RR

          You need to realize 2 things. 1. China gov has control over all newspapers, they can fire editors at will, or even shut down the media for “reformation”. 2. When accident occure, they don’t usually release victims’ name, head counts are usually ridiculously low. Compensation package with NDA one time offer immediately on site to victims’ familiy, very often, it’s matter of choosing between carrot or stick.

          • Alon Levy

            For the record, the only service for which I’d trust the statistics of what’s reported on Wikipedia is China’s high-speed rail network, and that’s because HSR tends toward infrequent large accidents that are impossible to cover up, just like Wenzhou.

        • RR

          one more thing, rail transport is government owned/run in china. There is regulation if casualties above certain numbers, someone in the management camp will be fired. So you get the picture they have strong motive to cover up.

        • RR

          To Alon Levy:

          Many ppl believed Wenzhou high-speed rail accident was a cover up. 4 cars packed with passengers totally pancaked and only 3~40 deaths, rescued last for 8 hours, while at the same time damaged cars were crushed and buried on site. Search party found a 2 years old little girl alive right before they were ordered to stop the rescue mission.

          So you get the picture, when things get covered up, there won’t be secret file locked up somewhere.

          • Alon Levy

            Well, we don’t know the exact number of passengers on board those 4 cars, but they were the front 4 cars of a 16-car train that had 558 people on board, so that’s about 130 potential deaths. And we know there were survivors in those 4 cars, and also a large number of total injuries (more than 130, but there were 28 other cars involved in the collision).

    • Tom West

      re Great Britain… there was a recent report on this:
      Fatal trains accidents on Britain’s railways. It doesn’t include pass-km, but you can find these here (table TSGB0699).
      A few points:
      1) For the period 1991 to 2009, there were 67 fatalites and 734.3bn pass-km, making one fatality for every 10.96 bn pass-km.
      2) for 2000-2009, the figure is one fatality for every 36.66 bn pass-km.
      3) The Ladbrooke Grove crash(1999) prompted the roll-out of TPWS acorss the network, which brings trains to a hault if they try and pass a signal at danger (or pass certain poinst to fast). TPWS doesn’t gaurentee the train will stop before the signal – it may stop in the area beyond the signal (UK signals are always some distance in front of the area they pr0tect, so that small overuns pose no safety risk). Every fatal train accident in the UK since 1999 has been caused by track problems rather than signalling issues.
      3) Ever since the Clapham Junction crash (1989), there has been a push to phase out the older rail vehicles. The majority of rolling stock now dates form the ’90s or later, and there vrutually done dating from before the ’80s. Modern rolling stock is a *lot* more durable in a crash. Case in point: Grayrigg, 2007. Trains derails ona faulty set of points (switches) at 96mph / 155km/hr,smashes through some caternary masts, and rolls down an embankment. However, the train dated from 2001, and all the carriages remained structurally intact, with the result that only one person died.
      4) With regard to accident rates before and after privitisation… if you plot the rolling ten-year average of fatalties/bn-pass-km, there’s no post-privitisation drop – it just has a steady rise (improvement).

  2. mulad

    According to the FRA’s Office of Safety Analysis, the fatality rate is far, far higher than that in the U.S. In 2011, Amtrak trains and equipment were involved in 132 fatalities, while the rail system as a whole had 739. The biggest chunk appears to come from trespassers, though grade crossings are also a big problem.

    The number of onboard passengers being killed is very low in comparison to those, however — Amtrak had two passenger fatalities last year, and 96 since 1991. Adding in commuter rail nearly doubles that to 178.

  3. Miles Bader

    What’s the point of distinguishing between “mainline rail” and “subways” anyway?

    In Tokyo at least, they’re pretty much the same in most respects (equipment, infrastructure, operating practices, scheduling) — and indeed are heavily interlined, using equipment from bo th — so there seems little justification in including one but not the other.

    • Alon Levy

      The point is that the UIC statistics I’m using are mainline-only. For Japan I could also find statistics for private railways, but I couldn’t find any passenger-km numbers for subways anywhere except the US and a few European cities. It’s too bad, because in Korea as I said there’s even less of a distinction between subways and mainline commuter trains than in Japan. Not including the Daegu subway fire, subways seem to have a better record overall than mainline trains.

  4. Cap'n Transit

    How about safety per passenger-year? That’s what we really care about, because the built environment forces people to travel more in one country/region than another.

    • Alon Levy

      I thought about that, but since the traffic in question in this post is predominantly intercity (except in Japan), passenger-km is the more accurate metric. Building fast transportation can induce people to make longer commute trips, but it’s not going to induce them to travel from New York to Washington rather than to Philadelphia, or from Paris to Marseille rather than Lyon. This is especially true for rail, because for both driving and air, the most dangerous segments are the beginning and the end (takeoff and landing, or driving on local roads), whereas on a train station approaches seems no more dangerous than travel between stations.

  5. Matt Sauer

    See also David Cay Johnston’s Free Lunch for a discussion of the lethality (and limited liability) of the US rail system.

  6. Pingback: TheCityFix Picks, June 3: Brazillian BRT, Mobile Wallet Technologies, C40 Cities Mayors Summit | TheCityFix
  7. Joseph E

    Does this include all commuter rail systems, or only a select few?

    Metrolink (Greater Los Angeles) commuter train have been involved in 244 fatalities over 15 years from the mid-1990’s, according to this LA Times article from 2008:,0,2843534.story

    Most of those deaths were pedestrians crossing the tracks, or vehicles running the light, though there were 17 train passengers killed in the infamous Chatsworth crash. Metrolink has about 500 million passenger-km per year now, and less in the 1990s, so the fatality rate is nearly 1 death per 30 million passenger-km, actually worse than the road network. Amtrak looks great by comparison.

    I wonder if China and India include all the pedestrians and cyclists killed at grade crossings. But certainly the European and Japanese data should be excellent. Does Japan keep its mainlines grade-separated, to prevent collisions at crossings? That seems to be the main source of deaths (along with suicides) here in LA.

    • Alon Levy

      It should include everything classified as commuter rail in the US, including Amtrak. As I said, grade crossing accidents below a certain level of severity are out by the nature of the data set I’m using. So are suicides (which dominate the same in Japan, where some busy commuter lines are not grade-separated). On the other hand, if you really want to include everything then maybe we should also count industrial accidents, Golden Gate suicides (hey, they involve the road network), and what not…

    • Stephen Smith

      Jesus, that sounds pretty horrible! Though I wonder, after reading that story, how many of these crossing deaths are really more properly thought of as the local roads department’s problem. For example, here’s the third bulletpoint on the second web page:

      A complex signal pattern and double left-turn lane have the potential to back up cars trying to head south onto San Fernando Boulevard. The congestion means that despite “Keep Clear” signs, almost every red light traps one or more cars briefly on the track.

      Almost every red light?? Yikes!

      …and then there are the suicides, which the article doesn’t quantify, which aren’t exactly the fault of the railroad.

  8. Joseph E

    What these stats really show is that RIDING a train is very very safe, especially compared to travel by car (1 death per 0.2 billion miles) or plane (about 1 death per 2 billion miles).

    All of the FRA nonesense, which is meant to prevent train/train crashes, should be gotten rid of. Instead, we need to reduce the number of collisions between trains and cars and pedestrians at grade-crossings. That’s where the real danger is. (And we need to get people out of cars and into buses and trains)

  9. Pingback: Boston South Station’s Supposed Capacity Limit | Pedestrian Observations
  10. Pingback: Yes, Amtrak is Indeed Mismanaged | Pedestrian Observations
  11. Pingback: Racism and Accidents | Pedestrian Observations
  12. Pingback: FRA Stonewalling | Pedestrian Observations
  13. Pingback: FRA Rules Are Not Just Buff Strength | Pedestrian Observations
  14. Pingback: The Cost of Heavy Freight Trains | Pedestrian Observations
  15. Pingback: Cost of Comac C-919
  16. Pingback: Northeast Corridor HSR, 90% Cheaper | Pedestrian Observations
  17. Paul Druce

    Necro comment ahoy! It’s actually worse than that. FRA data from January 1991 through December 2011 shows 189 passenger fatalities, 17,054 passenger injuries, and 333,524 million passenger miles for an average fatality rate of one death for every 2.85 billion passenger kilometers. That’s railroad passengers only incidentally, if Amtrak slaughtered a bus full of widowed orphan nuns, they aren’t counted (pretty sure they didn’t though).

  18. DM

    It seems to me that many of the rail accidents in urban areas in France are suicides; a common euphemism is “grave passenger accident”. Are these taken into account in those statistics, and aren’t there cultural differences in their respect? (i.e. some cultures having more suicides, some covering them up, etc.)

    • Alon Levy

      I think the numbers are almost entirely passenger deaths – no suicides. Grade crossing accidents in which no passengers died are occasionally included in the post because the list sometimes includes them, but only a small minority of these are on the list.

  19. tobias b köhler

    Indeed a great part of rail related deaths are either suicides or track workers ….

    When we compare India and China to the USA we see that both India and China are mass transportation whereas long distance railway in the USA is a niche market. There aren’t that many passengers. So one horrible crash every couple of years accounts for a much greater share of the total passenger kilometers in the USA than it does in India or China. (I would like to see Russia in this comparison, another very important railway system!)

    • typhoidX

      That’s a good point. In that case, should a more significant comparison be the US’s primary mode of transportation (driving) and PRC’s primary mode of transportation (rail)? I haven’t done the research, but I’d imagine that statistic wouldn’t be very flattering for the US either…

      • incu

        I’m almost certain rail travel in the USA is safer than road travel in the USA, a lot of the safety measures that can be taken on railways are impossible on the road. If someone on a train or signalbox makes a mistake, there is usually a system to bring the train to a safe halt. If a driver of a road vehicle makes a mistake, he’s dead.
        On a wider scale, it would be good to compare the chance of survival of the traveller/citizen across all modes. I would assume in “developed” countries it is more likely to get killed in traffic, whereas in “developing” countries you will be more likely to die from illness that could be treated in more developed areas, or wouldn’t even be reason to worry thanks to better hygienic standards.

    • Yu Hai

      That’s actually not a very good point. Firstly lower occupancy means few people would die in an accident, the the reduction on fatalities is proportional to the reduction on total passenger-miles. Plus, the US commuter rail plays a crucial role in both passenger rail and fatalities in the US., and is very crowded. Either way this is by-no-means an excuse.

      Also I’ve taken many mainline and branch sections of the Siberian rail. Their condition is way better than that of the United States.

  20. typhoidX


    Very insightful blog post you got here. Just out of curiosity, can you tell us a bit more on how you got the passenger-KM/year figures? Is it a yearly average based on 20 years of accumulated data, or just data for the latest year? If we’re talking about a span of 20 years, I’d imagine there would be huge changes to that number over that time frame, especially in the PRC & India, where the demand for long distance commutes is rapidly growing along with rapidly growing economies. Consequently, if these figures are for the latest year only, the passenger-KM/death statistic may be off a bit, since the numerator would be inflated (especially for the PRC).

    • Alon Levy

      It’s data for the latest year; I didn’t find historical data.

      The numbers for China and India have other problems, which is that not everything gets on Wikipedia. For reference, if you include just China’s HSR, then there is historical numbers for passengers, and if you assume distance traveled per passenger is the same as on the legacy network you get a death rate of 1 per about 10 billion passenger-km. Since elsewhere HSR is safer than low-speed travel, it means that most likely China’s overall death rate is not 1 per 50+ billion passenger-km but that not all accidents are on that list.

      In the US, Japan, and the EU the passenger-km figure is increasing, but at a steady rather than meteoric rate.

      • typhoidX

        I actually just went through some World Bank data on rail passenger-KM traveled for all the aforementioned countries, your numbers are a bit off as a result of using only the latest data, but it is directionally correct. Assuming your death figures are accurate, China & Japan tops the safety stats at 1 death every 33.5 billion passenger-km traveled (Japan actually came out slightly ahead in my figures), whereas the US is somewhere on the bottom at 1 death every 1.2 billion. Granted the US is a very rough estimate since the WB data is incomplete.

        The Euro zone had an unusually high rate (1 death every 853 million), maybe the World Bank’s definition of Euro zone is different than yours?

        Here is the site I used:

        Anyway, the conclusion is that I have way too much time on my hands.

        • Alon Levy

          Go to pages 102 and 47 here, including both deaths and passenger-km. It works out to 1 in 5 billion passenger-km in the EU-27 in 2005-2009, the only range of years for which there’s full EU-27 data.

          The passenger-km figure I used for Japan includes both the JRs and the major private railways (though the headline number excludes subways). A lot of the figures you see around the Internet are JR-only, since the private railways don’t really provide intercity service, just commuter service and some long-range regional service. The second biggest fatal accident I’m counting for Japan is a collision between a JR train and a non-JR train.

  21. Pingback: High-speed railway links Northeast China
  22. Pingback: India drools over chinese HSR tech.
  23. Matt

    Per a Reuters article today, some 40 people die DAILY on their rail system. That’s well in excess of 2500 over 20 years. In fact, that would equal approximately 14,600. . .ANNUALLY. That includes folks who live near the tracks and use them as open toilets, as well as passengers falling off of the trains.

    • Matt

      The Reuters article is in reference to India’s rail system. . .sorry about the omission in my first comment.

    • Alon Levy

      As I said, I wouldn’t trust these numbers for developing countries. But, the higher numbers include much more than passenger deaths in accidents; in the US and Europe, most rail deaths (which are not accounted for in this post) are grade crossing accidents.

  24. Archie Leach

    Just so folks know, India has, by far-and-away, the highest railway deaths in the world at 15,000 per year. It’s not hard to see why: just go to youtube and type in India railways. In the videos you’ll see huge numbers of persons just wandering willy-nilly ON the tracks. Houses are built right to the edge of the tracks and most impoverished Indians use the rail corridors as open toilets due to the complete lack of outhouses and the such. In the stations persons jump off the platform to walk along the tracks or to cross the tracks. Then there’s the reality that trains in India are operated/ran with their doors open. Commuters trains don’t bother closing the coach doors and intercity trains doors are unlocked so passengers regularly open them to “feel the wind” with no fear of repercussions. In the rest of the world, if you’re found walking on the tracks or jumping off the platform, the transit/railway security is going after you! If you somehow open the locked door on a moving train, the train will stop at the next coming station (whether it was a stop or not) to have you arrested for tampering with safety equipment and for endangering other passengers. And commuter and metro/subway trains won’t even release their brake systems if there is an open coach door. India’s rail system VERY much needs reform. In comparison China’s rail systems are considered very safe because China has pretty much eliminated grade crossings along the mainlines. The only grade crossings to be found on China’s railways are those on short spur lines in economically underdeveloped areas. Also along the train corridors, high walls and other barriers cover the length of the rail corridor so the only humans you see along the tracks are rail workers.

  25. Pingback: Train Weight and Safety | Pedestrian Observations
  26. Me

    I wonder if China may under report train accidents given the control of media there. Seems bizarre to me that China would have a better safety record than Europe or US on things like this.

    India better than US or Europe also shocks me. Does the wikipedia list only include major rail accidents?? Maybe many smaller ones in India and China aren’t listed?

    • Matt Hutton

      I think the Chinese would really struggle to hide something big like a rail accident given how many foreigners and rich Chinese there are there.

      Maybe at the start of the period.

      • Alon Levy

        I’m pretty confident about the safety record for Chinese HSR, esp. since it’s been almost 8 years with perfect safety record since I wrote the post. I agree that the numbers could be underreported for the Chinese legacy network (the methodology is based on news reports that Wikipedia has heard of), and know for a fact the numbers are severely underreported for India, where fatal grade crossing accidents happen several times a day in Mumbai. That said, I doubt I’m missing big accidents, which would make international news – I’m probably missing the bunch of small crashes with 5 kills that are meticulously reported in the first world but not the third.

  27. Pingback: Why American Costs Are So High (Work-in-Progress) | Pedestrian Observations
  28. Pingback: American Construction Costs and National Stereotypes | Pedestrian Observations
  29. Pingback: The NTSB Wants American Trains to Be Less Safe | Pedestrian Observations
  30. Pingback: What Makes a Train Safe?

Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s

This site uses Akismet to reduce spam. Learn how your comment data is processed.