How Common Is Drugged Driving?


From a recent USA Today article by Jonathan Shorman comes an astounding (to me) set of facts about drugs and driving that certainly ought to be considered as part of the conversation about decriminalizing marijuana:*

Researchers examined data on more than 44,000 drivers in single-vehicle crashes who died between 1999 and 2009. They found that 24.9% tested positive for drugs and 37% had blood-alcohol levels in excess of 0.08, the legal limit. Fifty-eight percent had no alcohol in their systems; 5% had less than 0.08. The data were from a government database on traffic fatalities.

Study co-authors Eduardo Romano** and Robert Voas of the Pacific Institute for Research and Evaluation in Calverton, Md., say their study is one of the first to show the prevalence of drug use among fatally injured drivers. Among drivers who tested positive for drugs, 22% were positive for marijuana, 22% for stimulants and 9% for narcotics.

The risks of drunk driving (and walking) are well-established; it is good (though wildly sobering) to see some hard data on drugged driving.

*Before you start stomping up and down about the relative driving risk of marijuana versus alcohol, do read the first few paragraphs of this post.

** One of Romano’s research areas — along with applied economics, epidemiology, traffic safety — is “Latino issues.” I’ve never seen that listed as an academic field; I wonder how many non-Latinos are in it?


Do you think that alcohol should be outlawed because it can be implicated in so many accidents? I don't think so, and I think the same applies for marijuana.


Exactly. I'm all for legalizing marijuana, and making "drugged driving" a criminal offense.

Just like when I tell people I'm for legalizing marijuana, and they say "but I don't want my doctor to operate on me while he's stoned". Yeah, neither do I, but I also don't want him to operate on me drunk, yet there's not a huge push for a return to prohibition.


I'm wondering how much overlap there is in these numbers, i.e. how many drugged drivers are captured by the drunk driving number as well? Not to suggest anything about the safety of drugged driving, just wondering how to read this for crashes that involved only marijuana.


There's a problem with only looking at crashes: it's difficult if not impossible to estimate the effect a cause has just from those cases. Ashworth, Clinton, Meirowitz, and Ramsay (manuscript 2008) provide an excellent example that shows the fallacy:

"The medical examiner of a small city made a limited claim about the likelihood that the apparent association of consuming beverage X with death represents a real effect. This finding is based on the fact that there were 16 deaths in the previous week. All 16 were beverage-X-drinkers; none was a non-drinker.... This sounds good, until we learn that beverage X is water."

In probabilistic terms, the author seems to confuse Pr(marijuana|crash) with Pr(crash|marijuana). Those are only equal if Pr(crash)=Pr(marijuana), which strikes me as unlikely.


Thank you - this is exactly the point I was thinking of. The study would have to find a control group (or baseline group) of drivers in the same area/same conditions who did NOT crash and see what percent of them also tested positive for marijuana.


I wonder (and I was not able to access the paper to find out myself) about the threshold for testing positive for marijuana. What if the driver tests positive for marijuana, based on their drug use the day before, or the week before? It is my understanding that marijuana use is regularly detected up to a month after use, but the high/impairment rarely lasts that long...

Jay L

One problem with the study is that marijuana is fat-soluble and stays in the system for several weeks after ingestion, inflating the numbers. If a person crashes a car while sober, he/she would still test positive for the marijuana they smoked several days earlier. Blood alcohol content dissipates from the system much faster.

Don't get me wrong - I think drugged driving is a problem. I just don't necessarily buy the figures in the study.

Milton Recht

Actually, I would argue the risk of drunk driving is not well established. People who drink and drive are not the same type of people who do not drink and drive. Until you take a random sample of drivers (actually double blind is the best so you would have to give fake alcohol to some and hidden alcohol laced, non-alcoholic beverages to some of the others) and show that those who then drink real alcohol are more accident prone and less safe drivers, the increased accident rate maybe just a filtering and sorting affect of alcohol drinking. The higher rate of accidents may be due to the same personality traits that lead people to drink. Drinkers, even when sober, may just drive more carelessly and aggressively than non-drinking drivers do.

If drinkers are different psychologically from non-drinkers, the baseline to compare the increased accident rate is not against all drivers, but against the drivers with the same psychological profile as drinkers. Auto accidents and reckless driving occur among sober drivers also, but this accident-prone group maybe of the same type as drinkers who drive. The increased accident rate due to alcohol may be much lower than often cited. I have heard it said that the same 20 percent of drivers are responsible for 80 percent of accidents. If true, alcohol and impairment may not be the reason (or the sole reason) for the increase in auto accidents. The same traits that lead one to drink and drive may also lead one to drive more carelessly. It is not an absurd idea, since it is known that a subset of drunks will often act out aggressively and abusively and use alcohol as the excuse, when it known that they will act similarly even when they are sober.



There would be no point in making the study blind as you describe. It's not a pharmaceutical trial; the placebo effect is irrelevant (it doesn't matter whether people suck at driving because of the actual effect of alcohol, or because they think they're drunk). Also, instead of using a control group with random assignment, it would be better to test each driver both drunk and sober. I believe that such a study is underway at my university, using a driving simulator.


The numbers concerning MJ are suspect for two reasons. First has been cited above -- MJ and metabolites are detectable long after any psychoactive / psychomotor effects have dissipated.

Second is an excellent study on MJ and actual driving performance. Take a minute and read it -- I think many will be surprised. It shows that impairment from MJ is basically self-correcting.

One issue is multiple drug use. Often MJ and alcohol (OK, weed and beer) are used together, which confounds attempts to figure out what might be at fault in a crash.

Joel Upchurch

The paper is available here:


1. Notice: Fatal *single-vehicle* crashes. These are people who drove really fast into a tree. Presumably you are astounded that so many of them were sober.

2. According to the cited study, the 2007 National Roadside Survey estimated that 7.6% of nighttime drivers had been using marijuana. Only 6% of the dead drivers-into-trees tested positive for marijuana. It is not clear what you intend for this to contribute to the conversation about decriminalizing marijuana.


Here, in the Netherlands, they've pretty much solved to problem of showing whether a person have smoked recently, whether it's still active. It involves a saliva test instead of blood or urine sample.

My second point is that education is everything, everybody around me know you do not want to drive while drunk and/or stoned. First of all the insurance doesn't cover anything so if you fuck up, you're fucked. Secondly the term stoned didn't fall out of the sky, when you smoke you get a "stone like reaction time". and trust me, most stones don't move themselves. And finally I really don't understand how and why you can let 16 year old's drive, most accidents here involve drivers 18 - 21 (we are allowed to get our drivers licence after were 18)

I'm happy I live in a country where guns are considered more dangerous than plants.

BTW: our government is creating laws to ban foreigners (tourists) from coffeeshops


Eric M. Jones

Dude--Far out!....

Marijuana And Actual Driving Performance

U.S. Department of Transportation,
National Highway Traffic Safety Administration
(DOT HS 808 078), Final Report, November 1993

Marijuana's effects on actual driving performance were assessed in a series of three studies wherein dose-effect relationships were measured in actual driving situations that progressively approached reality. The first was conducted on a highway closed to other traffic. Subjects (24) were treated on separate occasions with THC 100, 200 and 300 g/kg, and placebo. They performed a 22-km road tracking test beginning 30 and 90 minutes after smoking. Their lateral position variability increased significantly after each THC dose relative to placebo in a dose-dependent manner for two hours after smoking. The second study was conducted on a highway in the presence of other traffic. Subjects (16) were treated with the same THC doses as before. They performed a 64-km road tracking test preceded and followed by 16-km car following tests. Results confirmed those of the previous study. Car following performance was only slightly impaired. The third study was conducted in high-density urban traffic. Separate groups of 16 subjects were treated with 100 g/kg THC and placebo; and, ethanol (mean BAC .034 g%) and placebo. Alcohol impaired performance relative to placebo but subjects did not perceive it. THC did not impair driving performance yet the subjects thought it had. These studies show that THC in single inhaled doses up to 300 g/kg has significant, yet not dramatic, dose-related impairing effects on driving performance.

But I remember that Easter week in 1969 when Sheikh came up with that righteous hash. Man we was blotto...we went to Chinatown to cure the munchies...I think...then where'd we go?...I don't remember.



I seriously doubt only 22% of Americans have marijuana in their bloodstream. The study in the original post would indicate that marijuana users are safer drivers than the general populace.


Also consider the fact that in many states within the US, all it takes for something to become a drug related crime or for a "positive" on a drug test is only a minuscule, and more than likely insignificant, amount. Therefore someone who had smoked hours prior and crashed his car b/c he was talking on his cellphone would be considered under the influence, when tested, just the same as someone smoking a joint while actually driving.


Duuude... in fatal car crashes, 58% were sober and 37% were blotto. Sobriety kills, man.


I think the usage rate of drugs in the United States must also be considered here. Unless I'm misunderstanding the information presented, it sounds like drivers in these accidents tested positive for exposure which could have occurred (at least in the case of marijuana) any time in the weeks prior to the accident. (Read: just because you test positive for drugs doesn't mean you were affected by them at the time of the accident.)

I know there is great disagreement on regular usage rates, but 25% is not outside of estimated range for regular (in the last month) use by the adult population. So, maybe these numbers aren't so shocking...


Oh. I should have read all those comments above mine first. Y'all get the idea now...


About half of crashes involving deaths are single car crashes. Typically "driver error". I think this study has more to say about the types of people involved in single car accidents than effects of substances. Which is to say-- if you are drunk or stoned, the intelligent thing to do is not drive. These are not people who had a beer or two and drove home-- these are people who would have been so impaired that they would have had trouble putting the keys in the ignition, and chose to drive anyway.