Friends Don’t Let Friends Walk Drunk
I eagerly started reading SuperFreakonomics last night, and wasn’t long into before I encountered a rather eye-raising bit of traffic-related material (which is of course not a surprise).
The authors, Steven Levitt and Steven Dubner, write:
“Each year, more than 1,00 drunk pedestrians die in traffic accidents. They step off sidewalks into city streets; they lie down to rest on country roads; they make mad dashes across busy highways. Compared with the total number of people killed in alcohol-related traffic accidents each year — about 13,000—the number of drunk pedestrians is relatively small. But when you’re choosing whether to walk or driver, the overall number isn’t what counts. Here’s the relevant question: on a per-mile basis, is it more dangerous to drive drunk or walk drunk?
After running through some numbers, they find:
“Doing the match, you find that on a per-mile basis, a drunk walker is eight times more likely to get killed than a drunk driver.
They add a caveat that drunk walkers don’t kill other people, as drunk drivers do; but even factoring for that, “walking drunk leads to five times as many deaths per mile as driving drunk.”
As the jacket notes, the book asks “not only the tough questions, but the unexpected ones.” But having raised this rather startling statistic, it then moves on, leaving a number of interesting questions in its wake. The first is the problem with exposure data (not just how much but when and where), which for pedestrians is notoriously inaccurate, and for drivers only slightly better. Another thing that seemed worth raising immediately is the idea of causality; not that Levitt and Dubner have done this, but one might walk away from those numbers with the idea that all those pedestrians were running heedlessly drunk into the streets. But the statistic simply refers to the blood-alcohol content of pedestrians killed in traffic, so one could be standing on a street corner after having had a few drinks, and be killed by an out-of-control taxi (or a hit-and-run driver), and this would be coded as a pedestrian alcohol-impaired crash (One could also be walking to one’s car, of course, as many “pedestrians” in America are doing, although the authors present an either/or case of walking or driving).
And of course, the relationship in the book between drunk drivers and drunk walkers is perhaps less dramatic when viewed in light of the overall risk of walking versus driving. As a paper by John Pucher and Lewis Dijkstra notes, “it is much more dangerous to walk or cycle in American cities than to travel by car. Per kilometer traveled, pedestrians were 23 times more likely to get killed than car occupants in 2001 (140 vs 6 fatalities per billion kilometers), while bicyclists were 12 times more likely than car occupants to get killed (72 vs 6 fatalities per billion kilometers).”
This raises the question of how much more of a risk it is to walk drunk than to walk sober — which perhaps is the more interesting point of comparison here. But that question is of course complicated by any number of factors, like exposure — pedestrian deaths that involve alcohol are overrepresented at night, when visibility is a greater problem (and there are also more drunk drivers on the roads — and as this chart notes, a certain number of pedestrian fatalities involve alcohol impairment on both the part of the driver and the pedestrian). Another factor to consider, one not addressed by the authors, is the number of fatalities by .BAC level. Particularly in the category in which no driver alcohol impairment was implicated (but in other categories as well), there are many more pedestrians killed at at the .10+ BAC level than at the .01-.09 level (I’m not sure how this mathematically matches up with driver deaths by BAC level). So how much one has had to drink (which is probably associated with all sorts of other factors, like time of day, etc.) matters as well.
There’s a lot of other interesting things to consider here. One would be the mindset of severely intoxicated pedestrians versus drivers. One imagines, at least among a certain part of the intoxicated driving public, a certain behavioral adaptation (even if their performance per se cannot be modified), to avoid getting caught. As walking intoxicated is generally not against the law, is there greater risk-taking behavior at work? Certainly there are demographic factors to consider as well — it’s hard to imagine that drunk drivers and drunk walkers sync up in terms of life profile in most places in America. Then there’s the question of facilities — how many alcohol-impaired pedestrians were killed walking in unsafe places, places without sidewalks, etc.? (presumably stumbling through the French Quarter is safer than trying to leave a rural or exurban bar on foot, etc.). And I’m sure there are things I’ve left out that spring to your mind.
In any case, if it’s a sad commentary that driving drunk would be safer than walking drunk, it’s an even sadder commentary that it’s seemingly even more dangerous to walk sober than to drive sober.
![[del.icio.us]](http://www.howwedrive.com/wp-content/plugins/bookmarkify/delicious.png)
![[Digg]](http://www.howwedrive.com/wp-content/plugins/bookmarkify/digg.png)
![[Facebook]](http://www.howwedrive.com/wp-content/plugins/bookmarkify/facebook.png)
![[Google]](http://www.howwedrive.com/wp-content/plugins/bookmarkify/google.png)
![[MySpace]](http://www.howwedrive.com/wp-content/plugins/bookmarkify/myspace.png)
![[Slashdot]](http://www.howwedrive.com/wp-content/plugins/bookmarkify/slashdot.png)
![[StumbleUpon]](http://www.howwedrive.com/wp-content/plugins/bookmarkify/stumbleupon.png)
![[Yahoo!]](http://www.howwedrive.com/wp-content/plugins/bookmarkify/yahoo.png)
This entry was posted on Wednesday, October 7th, 2009 at 8:31 am and is filed under Traffic safety, Uncategorized. You can follow any responses to this entry through the RSS 2.0 feed. You can leave a response, or trackback from your own site.
| 
October 7th, 2009 at 9:34 am
Don’t you mean, “more dangerous to walk sober than to drive _drunk_”? Anyway, until pedestrians come equipped with roll bars and air bags, maybe motor vehicles should be banned from city streets.
October 7th, 2009 at 10:55 am
This statistical analysis makes me smack my forehead. The federal agency charged with ensuring the safety of our transportation system (NHTSA) is already obsessed with “solving” the problem of drunk pedestrian; their blame-the-victim mentality does not need additional encouragement. Meanwhile, the only thing NHTSA thinks drivers do wrong is failing to buckle up, or having one to many cocktails before getting behind the wheel. In the meantime, NHTSA will mandate airbags and stability control, while doing little or nothing to address the human factor and road design. Perhaps LaHood will help bend the bureaucracy; the Distracted Driving Summit symbolizes faint hope for many who work in bike-ped advocacy.
Happy International Walk to School Day! (October 7th)
October 7th, 2009 at 11:20 am
What’s the point of pedestrian statistics “per mile” when walkers put on so few miles compared to other modes of transportation. Wouldn’t it be better to compare “per hour” statistics?
October 7th, 2009 at 11:47 am
They’re using a *per mile* basis for comparison between vehicles and pedestrians. Calculating statistics on a per-mile basis is, I think, one of the biggest errors in the conventional American transportation paradigm. The results would look a bit different (and more favorable to the pedestrian) if they were calculated on a *per-trip* or even *per-hour* basis. Using *per-mile* in calculations assumes that driving down one mile of empty rural road is of the same “mobility” value as walking one mile along a busy New York street, past hundreds of potential destinations.
The *per mile* calculation that allows Dubner and Levitt to assert that driving drunk is less dangerous than walking drunk is the same misguided calculation that the federal government uses to undervalue the *accessibility* of denser urban areas, where numerous destinations are reachable within a relatively short distance without the need to drive, while spending money on expanding highways in exurban and rural areas to facilitate greater “mobility.” By this standard, it is therefore better to drive 12 miles to the nearest store rather than 3 miles, as “mobility” has increased.
October 7th, 2009 at 1:45 pm
It’s not a tough question, it’s a dumb question. Like Steve says the per-hour rate at least as import as per-mile rate, it’s just that the question is rigged to make per-hour irrelevant. But in the real world people first chose where they live, and drink. Once you’re at a bar 15 miles down a highway (or, in some regions of sprawl, 30 miles down the interstate), it is too late to pull out a copy of “Super Freakonomics” and decide if it’s safer to walk home. You either drive or pay for a $30+ cab. Whereas, if walking is even a possibility the distances are much lower and many more people are responsibly choosing to walk them. The suggestion that people living in cities or towns—the ones who can walk home after drinking—should drive instead, for Safety, is both destructive and asinine.
The per-mile data is only reflecting the fact that drunk (and sober) people walk shorter distances than they drive—generally, and also when they end up at the morgue. Amazing! Thanks as ever, Freakonomics, for that Hard Boiled Science.
October 7th, 2009 at 2:16 pm
Are deaths miles traveled a good comparison?
I would drive 5 miles but I wouldn’t walk 5 miles.
Maybe deaths per time spent traveling is a better comparison?
October 7th, 2009 at 3:05 pm
* the fact that drunk people are staggering home at the same time as drunk drivers may be a factor -especially at night. You’d need to compare danger to sober pedestrians journeying in the same areas at the same times.
* Per mile is a bogus measure, as it baises towards long-distance car/train journeys; per hour measures your risk per hour of use of a transport.
* here in the UK we have less of a drunk driving problem on account of having lots of pubs in walking distance in our cities. I have two establishments reachable without crossing any roads at all, for which I am very grateful
October 7th, 2009 at 5:50 pm
Tom,
I am disappointed in this blog for the first time. SteveL appears to have beaten me to this, but you have fallen for the classic Randal O’Toole “per mile comparison” fallacy. The simple fact that walkers are more likely to start a short 1-3 mile trip than a 15-20 mile trip means that every time a drunk driver completes a long trip, he skews the deaths/mile statistic wildly in favor of car drivers. We can argue with little doubt that most drunk trips on foot are college students and urbanites walking very small distances. They probably complete many trips safely that would hardly register in a miles to miles comparison. Since many would object to a trip vs. trip comparison that would bias in favor of urban trips, I suppose an hour to hour comparison would be the best method.
October 8th, 2009 at 10:49 am
You mention “more than 1,00 drunk pedestrians.” Did you mean “100″ or “1,000″?
October 9th, 2009 at 1:37 pm
What the others said: a “per mile” comparison for walking/biking/driving is completely invalid, as not too many people walk 10, or 20 or 60 miles a a day. Slightly shocking that this blog would not make this clear. A per-hour comparison would make sense… but would not make cars seem safer and walking seem dangerous, which seems to be the point of the mis-interpretation.
October 9th, 2009 at 3:02 pm
I’m glad you’ve all chimed in the “per hour” note; as I mentioned in the original post, one of the main problems is exposure data — and one problem with that is just how to measure it (not to mention all the other issues, like individual risk versus aggregate risk). And yes, you might drive five miles but not walk five miles, but you also wouldn’t walk at 55 mph — so it’s not inconceivable to think the rates would actually sync up in some cases. In any case, the science and data on “per hour” crash rates is hardly conclusive and needs much further examination; as a paper by Sonkin and Edwards, et al, in the Journal of the Royal Society of Medicine notes, “in other studies, comparisons of fatalities per hour and for distance suggest that walking and cycling still have higher risks than car transport, but that the differential is not so great.” But the differential exists.
October 9th, 2009 at 9:09 pm
” one could be standing on a street corner after having had a few drinks, and be killed by an out-of-control taxi (or a hit-and-run driver), and this would be coded as a pedestrian alcohol-impaired crash”
I didn’t know there was any such code… but I didn’t know pedestrians could crash.
Perusing MMUCC (Model Minimum Uniform Crash Criteria, 3rd Ed., 2008, DOT HS 810 957) I see no such reference.
AFAIK, in the US, a guy could have a half a beer and be hit in a crosswalk by a stone sober driver and, if that half beer is mentioned to the investigating officer, we have an “alcohol related incident”.
October 12th, 2009 at 6:53 am
The per mile data is not completely invalid. The per hour rate doesnt take into account that a 3 mile journey will take a pedestrian an hour compared to a few minutes in a car. Neither measure is an absolute figure - they must BOTH be taken into account and this is not easy data to do that with.
In the UK per hour walking is safer than cycling. Per mile cycling is safer.
I would be tempted to read the book before commenting, these guys are pretty good a stats!
November 21st, 2009 at 2:41 am
Like they normalized per mile, shoudn’t they futher normalize drunk driving and drunk walking death by the average rates of deaths by traffic during walking or driving?
December 7th, 2009 at 3:55 pm
All of these posts arguing that per-mile data is invalid. While there are many problems with the statistics here, one has to examine the original reason for posing the question in the book, which is: If you live one mile from the party you get drunk at, is it safer to walk or drive home? In such a scenario, an analysis PER MILE is indeed the most valid statistic for comparison. Such an analysis is probably most relevant to city-dwellers, where a reasonable walking distance between bar/club/party/etc. and home is more likely.
Of course, the real statistical problems come in then if city-dwellers are indeed more likely to have such a choice. How does that affect the statistics comparing drunken deaths by walking/driving in urban versus rural areas? Certainly the chances of getting killed walking home drunk in a rural area are different from an urban area, and certainly the percentage of people who do walk home drunk in a rural area is different from that in an urban area. And that’s just the start of the difficulty, since city-dwellers would probably take a cab or public transport part of the way rather than walking all the way home, making the stated scenario in the book seem rather unlikely.
However, despite all of these problems, the per mile basis of analysis is NOT fundamentally flawed, as so many posts are claiming. It’s just only applicable to a certain situation as defined in the scenario of the book… the many problems come in evaluating the validity of that scenario and collating the correct statistics. But to answer that particular scenario, the per-mile analysis is the correct one.