Skip to content

The sort of thing that gives technocratic reasoning a bad name

1. Freakonomics characterizes drunk driving as an example of “the human tendency to worry about rare problems that are unlikely to happen.”

2. The CDC reports, “Alcohol-impaired drivers are involved in about 1 in 3 crash deaths, resulting in nearly 11,000 deaths in 2009.”

No offense to the tenured faculty at the University of Chicago, but I’m going with the CDC on this one.

P.S. The Freakonomics blog deserves to be dinged another time, not just for claiming, based on implausible assumptions and making the all-else-equal fallacy that “drunk walking is 8 times more likely to result in your death than drunk driving” but for presenting this weak inference as a fact rather than as a speculation.

When doing “Freakonomics,” you can be counterintuitive, or you can be sensible, but it’s hard to be both. I mean, sure, sometimes you can be. But there’s a tradeoff, and in this case, they’re choosing to push the envelope on counterintuitiveness.


  1. Luis says:

    At the same time, CDC reporting sounds like alcohol *caused* 1/3 of the crash deaths. It is not difficult to imagine crash deaths that would have occurred even with sober drivers.

  2. Eli says:

    In fairness, Stephen Dubner is pretty explicit that drunk driving is very dangerous. His argument is that it’s relatively less dangerous than drunk walking, but he’s not trying to trivialize the dangers of drunk driving.
    I agree on every other point.

  3. Andrew McDowell says:

    In distinguishing drunk driving from drunk walking, it may be sufficient to note that a drunk driver is more likely than a drunk walker to kill other people by accident. However, I am wary of going with my gut instinct here, because for decades now, governments have been spending a great deal of money attempting to influence public attitudes to drunk driving – the CDC pages I got to following your link appear to have been written to persuade, rather than to provide objective information. Indeed, I have heard suggestions that the change in attitudes towards drunk driving make it a good precedent for other large campaigns of what is sometimes called public education.

  4. Leo Martins says:

    As I understood from your post (sorry, haven’t read the links) the CDC is talking about P(drunkdriver/crash death), not about P(crash death/drunkdriver) — which is low, according to freakonomics, right?

    PS: there are dozens of reasons why drunk driving is stupid, and there would be several unpleasant externalities in commending such behavior.

  5. Brian Newell says:

    No offense to the tenured faculty at Columbia either, but you might want to check where Steve Dubner (who writes most of the Freakonomics stuff) got his degree.

  6. Dave says:

    This has annoyed me too. The proper comparisons for drunk walking should be to taking a taxi or “sleep where you are.”

  7. Jim says:

    Thanks for posting this Andrew. They get away with too much flashy showmanship over there.

  8. Sebastian says:

    does anyone read Freakonomics anymore? I haven’t seen it cited in another blog for ages. Seems like moving away from the NYTs doomed them. Mostly good riddance I say – maybe Justin Wolfers can find a blog-home somewhere else.

    • Andrew says:


      I think you’re right. The Freakonomics blog seems to have moved away from Levitt and toward a Huffington-Post-style model of lots of low-impact posts by miscellaneous bloggers. I suspect some of the better bloggers on the site might be better off blogging on their own.

  9. Rob says:

    Why side with the CDC?

  10. jamdox says:

    The frekonomics guys have been hacks for years. Possibly always, but definitely since the global warming denial stuff in their second book.

    It seems like a fad in the social sciences to try to be counterintuitive for counterintuitiveness’ sake. The frekonomics guys take this to a new level, and produce proportionately crap work.

  11. jamdox says:

    Oh, the original post said what I said… read before posting I guess.

    Another freakonomics foul was a piece they did on some study supposedly showing that there is a lot less war in the modern era based on mortality statistics, but I heard no mention of controlling for the improvements in medical technology.

  12. mat roberts says:

    Part of the problem is measuring ‘deaths per mile’. Airlines like this measurement too. But if you follow it you find that flying in a space shuttle is safer than walking.

  13. txhoudini says:

    Dubner’s statistic was that you are 8 times more likely to die walking one mile drunk than if you were to drive one mile drunk.

    “Alcohol-impaired drivers are involved in about 1 in 3 crash deaths, resulting in nearly 11,000 deaths in 2009.” does not invalidate Dubner’s findings.

  14. Barrie says:

    @Andrew Gelman.

    Surely the all-else-being equal isn’t a fallacy in this case?

    I’ve not read it, but from the comments, including yours, in their example, it is equal, someone is choosing to walk home or drive home.

    If you’ve driven out of reasonable walking distance, then clearly all is not equal, and they’re deductions aren’t relevant to the situation. Or to put it another way, if walking isn’t an option, then the relative safety of walking or driving is irrelevant.

  15. Josh says:

    If there are 300,000,000 people in the country, and 11,000 traffic deaths invoking alcohol, that is 1 in 27,273 odds, which, is unlikely that you will find yourself in that position. That is the point, I think.

    • Andrew says:


      Less than 1000 kids die each year in swimming pools in the U.S., and Levitt had made a big deal about this one. So I don’t see how 11,000 deaths involving drunk driving can be dismissed as a “rare problem that is unlikely to happen.”

      • Barrie says:

        Is his point not that drunk driving isn’t dangerous, but that drunk walking is?

        • spencer says:

          But if Andrew is correct in saying that

          “Freakonomics characterizes drunk driving as an example of ‘the human tendency to worry about rare problems that are unlikely to happen,'”

          then that would suggest to me that he *is* suggesting that drunk driving isn’t dangerous, at least on a society-wide level.

          • Barrie says:

            Likelihood and acceptability are different things.

            I.e. it might not be likely, but the outcome when it does happen is unnacceptable, especially given it is both unnecessary and avoidable.

        • Martyn says:

          Yes that’s his point. As far as I can see he is saying that everyone knows that drunk driving is dangerous but look here’s something even more dangerous that people don’t know about. Rationally, if you don’t drink and drive then you shouldn’t drink and walk.

          If that is his argument, then there are a number of flaws. I’ll pick two:

          People are not discouraged from drink driving by fear of a fatal accident. It’s actually fear of a DUI citation, which is statistically far more likely, that puts them off. No such disincentive exists for walking.

          Walking is already more dangerous than driving, even without alcohol. David Spiegelhalter’s web site gives 17 miles per micromort for walking, as opposed to 250 miles for driving ( ). You may as well start a campaign against walking. I’m sure urban planners the world over will rejoice.

          • Barrie says:

            but surely the DUI citation exists due to the unnacceptable outcome of the risk?

            Campaign against walking? I doubt it’s the walking that’s creating the danger, more likely to be whatever killed the walkers, which I suspect is most likely bring us back to driving… but that’s a different topic!

    • C Ryan King says:

      x50 driving years. x6 incapacitating (highest severity grade) non-fatal injuries per fatality. 1/90 isn’t so hot.

      The walking bit is also fairly silly. I would say that almost all drinkers have walked around a fair bit while modestly intoxicated and not been injured. Claiming that drunk driving is 1/8 as risky as that conveys that there is essentially no risk. I am somewhat dubious that getting in a car while walking between bars fairly loaded would have been safer.

  16. Andy Brice says:

    So perhaps drunk walking is more likely to result in your death than drunk driving. But drunk driving is far more likely to result in someone else’s death.

    That’s largely why it’s illegal; to stop individuals with impaired judgment from endangering others’ lives, and make them responsible for their own actions.

  17. Stephen Boisvert says:

    Shouldn’t the denominator for the per comparisons be time? Exposure to danger is time contingent not distance contingent. Walking takes much more time than driving so of course per mile fatalities are going to be higher.

    Also fatalities are a very weak but convenient measure for the danger of drunk driving. Serious injuries would be more meaningful. Last time I looked the stats for driving accidents resulting in hospitalization were about 10X greater than the fatality numbers for North America. I suspect the multiplier would be a lot lower for walking.

  18. William Ockham says:

    When this subject first came up I pointed out that Dubner was just making stuff up. He made a completely unsupported assumption about the number of drunk miles walked per year. He also had a very inaccurate estimate for the total number of miles walked in the U.S. The government does what looks like a pretty good survey on all forms of transportation. They even track the times of day that people walk, drive, bike, etc. And of course they keep really good stats on pedestrian deaths.

    The real story is a lot less “freakonomics” than Dubner’s fairy tale. Walking in the daytime is really safe (drunk or not). Walking at night is dangerous, drunk or not. Walking at night in bad weather (rain, snow, and especially sleet) is just crazy. Compare the these stats to driving stats and you will see that Dubner is being an idiot. For driving, darkness and bad weather contribute slightly to increased fatalities, but drunkness is more significant. For walking, there is no solid evidence that drunkness is even a factor in increased fatality rates (although it probably has some effect) while darkness and bad weather are huge factors. If Dubner had spent a few hours on this instead of five minutes, he might have figured this out.