Princeton Election Consortium

A first draft of electoral history. Since 2004

Train travel is incredibly safe

May 13th, 2015, 9:08am by Sam Wang



Television “news” media, as is often the case, is good at blowing an event out of proportion. In the wake of the Amtrak train crash (and hey, I am totally against train crashes!), we have this:


In contrast with this statement, the bar graph above shows actual transportation fatality risk in the United States, from an article by Ian Savage at Northwestern University. His data end in 2009, but note that the overall tendency is for travel to get safer over time:
From the work of Ian Savage, Northwestern University

Take a look: death rates have gone down for all forms of travel: car, train, airplane, and boat. Despite this, such gore is a staple of news broadcasts. However, maybe some good can yet come of this: there is a major need for repairs and upgrades in U.S. infrastructure. Train travel is one place to invest the dollars.

It is a shame that we are intensely visual primates. Actual threats like climate change are too abstract to appreciate, while on TV we get panic about Malaysian jets and train crashes.

Tags: Uncategorized

21 Comments so far ↓

  • Tom Jeffry

    Here in India, train travel is almost like a culture. I agree to Sam. Train is the safest way for long distance travel.

  • William S.

    Honestly, the surprise to me is that buses are *significantly* safer than cars. Is it because of better driver training, lower speed/less freeway travel, or something else?

  • bks

    Time to get off the train an on the bandwagon:

    Hillary to lose New Hampshire and Iowa to Bernie Sanders says GOP pollster
    http://atr.rollcall.com/pollster-to-christian-group-balance-ideology-electability/?dcz=emailalert

    –bks

  • Jacob Hartog

    I misread this as “Time Travel is Incredibly Safe,” which was a surprise to me, given various films I have watched on the topic.

  • Richard

    In the past I’ve seen three primary metrics for travel risks: fatalities by mile, fatalities by hour traveled, fatalities by trip. Given that different modalities will have differing risks (air travel, as pointed out, generally has risks while landing and take-off; car travel is mostly at risk when driving within a few miles of home), looking at all three measures is likely to be more robust than looking at one measure alone.

  • Eric Walker

    I always have a nagging sense that passenger-miles is a poor indicator of risk, especially for air travel. Almost all the risk is in the few minutes of takeoff and then of landing; the distance travelled in between is, I believe, almost irrelevant to fatality rates. Nor should the number of other passengers on board affect my risk in taking the flight.

    I suppose the question might best be posed as “What is my risk of death or serious injury in getting from A to B by this, that, or t’other mode of transportation?” I’m not sure there are readily available stats that can answer that for any arbitrary A and B (even restricting it to major markets.)

  • Neil

    Perhaps a better statistic to look at would be deaths per mile in rail accidents across countries. That way we would be comparing like to vaguely similar.

  • Joseph

    I’m no statistician, but the big difference between these different modes of transportation that leaps out at me is volume. I’d guess that the volume of cars vastly outnumbers the volume of trains, planes, buses, and ferries.

    In other words, moving a large quantity of people at a shot seems, by definition, to be safer.

    • Sam Wang

      I don’t understand your first comment. In regard to your second comment, one can define another measure, based on vehicles rather than people. Multiply each of these rates by the typical number of passengers per vehicle. For example, with the definitions above, Per-motorcycle and per-car death rates stay the same. Per-ferry, -bus, -rail, and -airplane rates go up by a factor of 10 to 100. In other words, it is a measure of fatal vehicle incidents per billion miles.

      With that new definition, it is apparent that per billion miles of travel, one is roughly equally likely to be part of (i.e. a passenger on the vehicle) a fatality incident by most modes of transport . The exceptions are motorcycle (again) and…ferry, where one can evidently expect a fair rate of harrowing incidents.

      The above reasoning is made more complicated by the fact that certain modes of transport cause deaths in large groups – most obviously, airplanes.

    • Matt McIrvin

      “one can define another measure, based on vehicles rather than people. Multiply each of these rates by the typical number of passengers per vehicle.”

      Is that right? In a fatal plane crash, usually all or most of the people on the plane die. In a fatal train crash, usually only a small fraction of the people on the train die. So if you’re going from a fatality rate to a per-vehicle-fatal-accident rate, shouldn’t there be different correction factors to allow for this? I should think that to get such a low fatality rate from planes the crash rate should be proportionately much lower than for trains.

  • DanF

    This answers my question of are we having more accidents (at least up to 2009) – but doesn’t address injury versus morbidity: http://www.census.gov/compendia/statab/2012/tables/12s1103.pdf

    Bounces up and down quite a bit year to year, but interestingly we’re not crashing into each other as much as we used to. So better tires, visibility, anti-lock brakes probably all a factor in this.

  • DanF

    I wonder what percentage of automobile fatalities are down due to medical advancements as opposed to technological safety advances. Are we having the same rate of accidents with injuries? If we aren’t, why? Are anti-lock brakes preventing more serious collisions (despite all of the txting-while-driving)? If we are having the same rate of injury accidents, do air bags save our brain cases more than doctors and first-responders getting better at keeping us alive? Or have we gotten significantly better at constructing crumple zones?

    I need answers! OK … I don’t *need* answers, but I am curious.

  • Matt McIrvin

    …The details in the Savage paper are fascinating: among other things, he says private general aviation has always been far riskier than scheduled commercial service, but general aviation has been getting both safer and less popular, possibly in part because of an increasingly urban population.

    The drop in aviation fatalities is even more amazing if you restrict your attention to large airliners. One of his charts has a strange, seemingly huge jump upward in fatality rates for smaller commercial flights in 1997, but it’s an artifact caused by the larger end of that range leaving the dataset as they got reclassified into the same regulatory regime as large airliners.

  • Matt McIrvin

    Some of these numbers are truly astounding. I remember people emphasizing, correctly, the great relative safety of air travel when I was a kid, but airliners were deathtraps in those days compared to now. Most amazing is the drop in fatal auto accidents: I remember thinking of that as something that would just inexorably get worse no matter what anyone did.

  • Matt McIrvin

    I recall an analysis of bicycle travel that suggested that, while the risk of accidental death was quite large, the reduced risk of death from heart disease more than made up for it.

  • Amitabh Lath

    Statistics say random events clump. So…are we waiting for the next couple of derailments?

    Seriously though, is there a version of the first plot that does things by trip-hours rather than trip-miles? I would suspect the values are more comparable.

  • Jillian W

    32,719 people killed in car accidents in 2013. Very scary.

    • Matt McIrvin

      The Ian Savage article closes by mentioning that about 90 people still die every day on US highways.

      Fatal car accidents are as visually grisly as train accidents, but they are so commonplace that they don’t get much coverage.