1993: Fatal Crash Rate

Explain xkcd: It's 'cause you're dumb.
Jump to: navigation, search
Fatal Crash Rate
Fixating on this seems unhealty. But in general, the more likely I think a crash is, the less likely one becomes, which is a strange kind of reverse placebo effect.
Title text: Fixating on this seems unhealty. But in general, the more likely I think a crash is, the less likely one becomes, which is a strange kind of reverse placebo effect.


This is the second recent comic after 1990: Driving Cars on the subject of the dangers of cars. It combines general statistical correlations between age and safety improvements with fatal crashes. The graphs are:

  • Fatal car crash probability based on age: Young drivers are generally considered more reckless, which leads to more accidents (Randall is shown to have started this stage in the year 2000). Actuaries noted a spike in the death rates for teenage boys even before cars were invented. As drivers become comfortable with driving, internalizing the rules of the road, their accident probability quickly drops, but this decrease becomes less pronounced when the driver needs to adapt to new traffic patterns due to moving or changing schedules (2010). After driving for 20 years, accident probability reaches a minimum, but is shown to rise slightly in 2040, probably because Randall fears a midlife crisis. By 2050, aging starts to affect a driver's abilities (reflexes, concentration, eyesight, etc) so accident probability rises. The graph seems to be based fairly accurately on Massie and Cambell's 1993 paper 'analysis of Accident Rates by Age, Gender and time of Day ... ' itself taken from a 1990 survey. At the time the overall rate was 3.03 fatalities per 100 million vehicle miles - with rates of 9.21 for teenagers and 11.53 for those over 75. At the time those over 75 (born 1915 or earlier) may well not have been formally taught to drive. It would be interesting to see how this data changes with time.
  • Overall US fatal crash rate per mile traveled: This graph attempts to normalize these factors by correlating accident probability to how many miles the driver had driven by the time they died in an accident. However, the accident probability decreases with time as road traffic safety improves. The graph does a conservative estimate for future years, probably because improvements are, by nature, incremental -- which is why the graph has a slightly hyperbolic shape.
  • My miles traveled by car per year provides an estimate of miles traveled to be able to apply the second graph to himself. It shows a rise in the later half of the 2000s decade (indicating a job which requires a lot of driving) and a fall shortly after (indicating a job that doesn't require much driving). with a rise up to the present. The graph predicts either that this rise will continue, or will drop since this "depends on job, where I live etc."

The final graph, ostensibly the product of the three previous graphs' probabilities, shows that Randall worries that he will eventually be involved in a fatal car crash unless self driving cars take over, which he believes would eliminate car related fatalities. He is of the opinion that they will take over, but that they might not do so quickly enough to 'save' him from the spike of age-related fatalities in later life.

The comic includes three smaller line graphs along the top, and then a larger line graph, which is kind of a combination of the three smaller ones, at the bottom. A vertical dotted line is used on all these graphs to indicate "now", 2018; everything to the left of the graph has already happened (though the graphs are showing statistical history rather than actual history) and everything to the right is projected to happen, statistically.

The first smaller graph, labeled "My fatal car crash probability based on my age", shows the likelihood he'll be involved in a car crash at different ages. The line doesn't start until slightly before 2000, probably when he first learned how to drive and started driving himself. He's not including when he would have been a child and a passenger, just when he is the actual driver. The two most dangerous ages to be driving are generally when you've first learned how to drive (and haven't yet mastered the skills or gained learned reflexes) and then again at an elderly age when your reflexes are slower and your senses become more limited (narrow field of vision/loss of peripheral vision, worse hearing, etc.).

The middle smaller graph, labeled "Overall US fatal crash rate per mile traveled", lists how likely a fatal car crash is on a mile-by-mile basis, regardless of age. It used to be you were much more likely to have a fatal car crash in any given mile due to lack of safely features in cars in the 1970's. As more safely features were introduced and mandated, some to help prevent accidents (i.e. anti-lock brakes) and some to help make more of the accidents survivable (seat belts, air bags), overall safely has improved and is projected to continue improving.

The third smaller graph, labeled "My miles traveled by car each year", is a simple graph of the distance Randall has driven every year. As he approached 2010, he was driving a lot more then when he first started, then life circumstances presumably changed so his need to drive diminished a bit, and now it's slightly increasing again. He has no way to predict future life driving needs, however, so the graph converges after "now" to include both gradually increasing as well as gradually decreasing driving needs. At an advanced age he'll probably mostly stop driving.

The final, large graph, labeled "My estimated lifetime probability of being in a fatal car crash", combines these different factors into a smoother curve of gradually being safer (or at least not dying) while driving, with the possibility introduced, at an indeterminate time, that self-driving cars get to the point where they are both safe and widely adopted, at which point Randall expects the chance of a fatality to decrease to zero over a relatively short period of time (i.e. a decade). In the event the self-driving cars do not deliver in safely and/or are not widely adopted, the safety will gradually level off and then increase a bit near older age before dropping off again, but always with a distinct chance of fatality.

As the title text points out, fixating to this degree on a single source of danger is unhealthy. But the more Randall fixates on the danger of car crashes, the safer (or maybe the less) he drives, which reduces his chance of being in a fatal car crash. Note that Randall used to fixate on the danger of velociraptors, there is even an entire category based on his fear of them.

If one were to become a professional driver and drive at 50 mph for 8 hours a day, 200 days a year, for 50 years - one would drive about 4 million miles - so one's risk of dying in a car crash would be much less than 1%.


[Graphs are shown inside of a panel.]
[Graph 1:]
My fatal car crash probability based on my age
[Label at 2018:]
[Graph 2:]
Overall US fatal crash rate per mile traveled
General safety improvements
[Graph 3:]
My miles traveled by car per year
[Label after 2018:]
Depends on job, where I live, etc.
[Graph 4, below the previous graphs:]
My estimated lifetime probability of being in a fatal car crash
[Label pointing at late-2020s:]
Point at which self-driving cars become safe and widely adopted, making crashes rare (assuming that happens)
[Label pointing at a gray segment after late-2020s:]
Fatal crashes avoided
[Caption below the panel:]
It feels weird to look at car crash statistics and wonder whether we'll all be able to stop driving before I'm involved in a fatal crash.


  • The title text misspells "unhealthy".

comment.png add a comment! ⋅ comment.png add a topic (use sparingly)! ⋅ Icons-mini-action refresh blue.gif refresh comments!


Bad graphs, man. ~ ProphetZarquon (talk) 16:36, 14 May 2018 (UTC)
This implies that self-driving cars would have a fatality rate of zero. --Chrispugner (talk) 20:15, 14 May 2018 (UTC)

But even after "general safety improvements", the crash rate (top middle graph) is still not zero, so the bottom graph makes no sense... And how would you achieve a fatality rate of zero? Herobrine (talk) 03:47, 15 May 2018 (UTC)
According to Randall, a fatality rate of zero could only be achieved if self-driving cars worked well and if all cars on the road were self-driving cars. Assuming both conditions true, as they might be some day (or conversely might never be true), then it should be possible to minimize crashes to the point where fatalities would be statistically zero though certainly not absolutely zero, as malfunctions could happen. -boB (talk) 21:29, 15 May 2018 (UTC)
Obviously, you reach a fatality rate of zero when there is no human on the road or around. -- Hkmaly (talk) 01:10, 16 May 2018 (UTC)
Anything can be done by technology. :) Sweden had a lot of fatal train crashes, but after having introduced Automatic train control during the 80's, they completely vanished. (Three people have died on board Swedish trains since 1991, of which two were train employees. The accidents were due to non-rail vehicles being on or next to the track.) 10:38, 18 May 2018 (UTC)

"Fixating on this seems unhealty. But in general, the more likely I think a crash is, the less likely one becomes, which is a strange kind of reverse placebo effect." It will not be a reverse effect if "It feels weird to look at car crash statistics and wonder whether we'll all be able to stop driving before I'm involved in a fatal crash." is being done while driving. Gene Wirchenko [email protected] 04:37, 15 May 2018 (UTC)