Pot prohibitionists frequently warn that legalization will flood the roads with dangerously stoned drivers, leading to a surge in traffic fatalities. So far there is not much evidence of such a surge in Colorado or Washington, where marijuana was legalized in 2012. A new study may help explain why: It looks like marijuana's impact on traffic safety has been greatly exaggerated.
Reporters and anti-pot activists commonly warn that marijuana use doubles the risk of a car crash. Even if that were true, toking would pale in comparison to drinking as a road hazard, since research indicates that a blood alcohol concentration of 0.10 percent quintuples the risk of an accident. But according to an analysis that's about to be published by the journal Addiction, the increase in crash risk associated with marijuana use is roughly 20 percent to 30 percent, as opposed to the widely cited estimate of 92 percent.
That estimate comes from a 2012 BMJ meta-analysis by Canadian epidemiologist Mark Asbridge and two of his colleagues at Dalhousie University in Halifax. Ole Rogeberg and Rune Elvik—who work, respectively, at the Ragnar Frisch Center for Economic Research and the Norwegian Center for Transport Research, both in Oslo—did a new meta-analysis of the nine studies covered by the BMJ article, taking into account additional confounding variables. As they explain, "Using cannabis and driving under the influence are behaviours that are more common among young adults and males, groups with higher crash risks irrespective of use. Estimated odds ratios typically decline substantially after adjustments for such factors." That is what happened in this case, and it's the main reason Rogeberg and Elvik's results differ from Asbridge's.
Rogeberg and Elvik also took another look at a 2012 Epidemiologic Reviews article, likewise based on a meta-analysis of nine studies, in which Mu-Chen Li and other researchers at Columbia University's medical school estimated that marijuana use increases crash risk by 166 percent. Rogeberg and Elvik found that "the pooled studies report qualitatively different types of associations: Self-reported crashes in some past period for cannabis ever-users vs. never-users, self-reported crashes in a past period for those with self-reported intoxicated driving episodes in the past vs. those without, and acute intoxication amongst crash-involved and other motorists." They conclude that "the lack of clear study selection criteria…gives the resulting pooled estimate no meaningful interpretation."
In addition to reanalyzing the data from the earlier reviews, Rogeberg and Elvik did a fresh meta-analysis of 21 studies. The new analysis generated two estimates of the crash risk increase associated with marijuana use (based on two different statistical methods): 22 percent and 36 percent. "The replication of Asbridge et al. and Li et al.," Rogeberg and Elvik write, "indicates that their published pooled estimates substantially overestimated the effect of acute cannabis intoxication on crash risk, and that the pooled estimate presented by Li et al. is hard to interpret given the qualitatively different estimates pooled. The revised estimate from the studies used in Asbridge et al. was in line with the results from the expanded meta-analysis…lying in between the pooled odds from the mixed-effects model of 1.36…and the pooled odds from the PEESE meta-regression of 1.22."
Rogeberg and Elvik found "a statistically significant risk increase of low to moderate magnitude." By contrast, a 2015 National Highway Traffic Administration (NHTSA) study that was included in the meta-analysis found no statistically significant association between marijuana use and crash risk once the researchers adjusted for confounding variables. The explanation for this difference may be that the NHTSA analysis included any drivers who tested positive for active THC, whether or not they were still feeling the effects. Other studies may have done a better job of excluding sober drivers with residual THC. Rogeberg and Elvik say "the primary criterion for study inclusion was the quality of the information given about cannabis use, in particular if, based on the information given, there was reason to believe that cannabis had been used while driving or recently enough before driving for effects to persist."
The estimates from Rogeberg and Elvik's analysis are similar to the crash risk associated with blood alcohol concentrations of 0.05 percent or so, which is below the legal cutoff for drunk driving (0.08 percent). "The effect sizes found represent an average risk increase for those driving after the use of cannabis," they write. "Under a causal interpretation, this suggests that roughly 20-30% of traffic crashes involving cannabis use occur because of the cannabis use. By comparison, the comparable 'average' relative risk for accidents with fatalities after drinking alcohol has been estimated at 7.5, which would imply that around 85% of crashes involving alcohol occur because of alcohol."
There are a couple of additional complications. Research indicates that cannabis consumers, unlike drinkers, tend to be aware of their impairment, so it may be that the ones who end up driving after using marijuana are less stoned than the ones who choose not to do so. "If the low estimates are taken as evidence that 'driving after cannabis use' is unproblematic," Rogeberg and Elvik observe, "the new users would tend to be more impaired and have a higher cannabis-induced increase in crash risk."
Then again, the researchers add, "deciding to drive while intoxicated is a decision correlated with traits that predict higher crash risk independently of cannabis use: high speeds, close following, dangerous lane shifts and drunk driving." If some of the risk associated with cannabis use in Rogeberg and Elvik's analysis is actually due to those traits, their numbers may overestimate the role of marijuana intoxication.
Rogeberg and Elvik note that the impact of legalization on traffic fatalities will also depend on policies aimed at discouraging stoned driving and the extent to which marijuana substitutes for alcohol, a tendency that could reduce crashes because alcohol has a much more dramatic effect on driving ability. They conclude that "a low-to-moderate causal effect of acute cannabis intoxication on crash rates is likely to play a limited role in the overall policy picture surrounding cannabis legislation."
Yet that is not the way policy makers see it. Washington's marijuana legalization initiative included a per se rule that equates a THC blood level of five or more nanograms per milliliter with impairment, which effectively makes it illegal for many regular cannabis consumers to drive, even when they are perfectly capable of doing so safely. Colorado did not go quite that far, but legislators did authorize juries to convict people of driving under the influence based on nothing more than blood test results indicating a THC level that exceeds Washington's arbitrary limit. Defendants in Colorado can still try to persuade jurors that they were not actually impaired.
This month a Massachusetts legislative committee studying marijuana legalization, which voters in that state may approve in November, noted that "there is no well-accepted standard for determining driver impairment from marijuana intoxication." Its report nevertheless recommends "establishing a legal limit for THC blood concentration that would support at least a permissible inference standard in court"—the approach taken by Colorado. In fact, the committee says legalization should not be allowed to take effect until some such rule is established.
"Driving a vehicle while under the influence of marijuana is a significant public safety concern," the report says. "Fatal motor vehicle accidents attributable to marijuana-impaired operators tripled between 1999 and 2012." That claim carelessly conflates drivers who "test positive" for marijuana, including residual THC and nonpsychoactive metabolites, with impaired drivers. The same mistake underlies claims such as "marijuana is involved in one out of every nine fatal car crashes today," as the campaign against legalization in California asserts.
Politicians are in such a panic about stoned driving that they are willing to endorse legislation they concede has no scientific basis—legislation that is bound to result in the conviction of innocent drivers who pose no threat to the public. Advocates of that policy simply assume that requiring proof of actual impairment, as Massachusetts currently does, is too demanding in the face of the increase in cannabis consumption they anticipate as a result of legalization. This response is unfair, premature, and disproportionate given what we actually know about marijuana's impact on traffic safety.
This article originally appeared at Forbes.com.
Start your day with Reason. Get a daily brief of the most important stories and trends every weekday morning when you subscribe to Reason Roundup.