One of the zombie lies that cannabis prohibitionists luuurrrve to shill is that legalization would inevitably result in mass carnage on the highways (if not also off of the highways as well) from stoned driving. These fearmongers can be every bit as hysterical as MADD is infamous for being in regards to alcohol. Studies to date on the matter have been mixed, but have been generally pointing in the direction of it being largely a non-problem overall, at least compared to the counterfactual of not legalizing cannabis. While it is true that cannabis can impair driving, it typically does so less than alcohol does, and regardless it does not necessarily follow that legalization would automatically result in more traffic casualties overall that would otherwise have occurred.
Enter the latest study by the Insurance Institute for Highway Safety (IIHS) using data from 2009-2019, the largest study yet on the matter. While it purportedly found that recreational legalization was associated with a small increase in both fatal and nonfatal injury crashes overall, there is far less here than meets the eye. For starters, there seemed to be no rhyme or reason to which states saw increases and which ones saw decreases (yes, some states actually saw decreases post-legalization) in such casualties, even after controlling for a whole host of potential confounders and including several comparison states (five legalization states and five non-legalization states from the same general region of the country). Such extreme heterogeneity strongly suggests that at least some, if not all, of the observed effect is spurious. More importantly, legalization of recreational cannabis use itself appeared to be responsible for nearly all of the observed effect, while legalization of retail sales (which occurred later, often a year or more later) had no significant effect (actually, a small decrease on average, go figure). You read that right--it is exactly backwards from what the standard prohibitionist narrative would claim, also suggesting a spurious effect. And finally, in any case the size of the effect (5.8% for injury crashes and 4.1% for fatal crashes) on average is basically too weak to conclusively prove a causal link from a single study alone, but especially from a study with the aforementioned issues like this one.
In other words, one can see that there is barely any signal in the noise, and that supposed signal may very well turn out to have been entirely noise all along.
So once again, legalization was NOT anywhere even remotely close to disaster after all. And most interestingly of all, the study authors note that, “the estimated effect of marijuana legalization on crash rates is only slightly lower than the estimated effects of lowering the legal drinking age in the United States from 21 to 18.” Which, as we at Twenty-One Debunked have noted time and again, are both nowhere near as large or scary as the prohibitionists and ageists want us to believe. And an honest interpretation of the aforementioned study would actually put it far closer to the famous Miron and Tetelbaum (2009) study than any pro-21 study.
UPDATE: Nevertheless, we skeptics do of course need to be careful not to give into the temptation to be "pseudocritics" here. Blithely, casually, and reflexively mocking and dismissing any paper for its apparent statistical noise or seemingly puny effect size while ignoring nuance, details, or context (often without even reading the paper in question) is intellectually lazy at best, and intellectually dishonest at worst. The above analysis, however, is NOT an example of this pitfall. Pointing out specific concrete issues or inconsistencies that seriously weaken or vitiate the conclusions, while keeping it all in context is not an example of this fallacy. Nor is pointing out, a fortiori in fact, the public policy implications or lack thereof (a second-order question that inherently goes beyond the science itself) that the prohibitionists so desperately wish for despite being a category error. And those who doggedly cling to even the flimsiest cherry-picked results and tiniest effect sizes, to push a thoroughly illiberal agenda while ignoring an honest cost-benefit analysis in the name of a perverted and Hobbesian version of the precautionary principle, are guilty of a categorically far worse intellectual failing than even the silliest pseudocritics are.
When we see that cannabis legalization and non-legalization states in recent years, and MLDA 18, 19, 20, and 21 states in the 1970s and 1980s, are all more or less within error bounds of each other overall in terms of traffic casualties, at least in the long run, it makes far more sense to err on the side of liberty.
And finally, to say that small effect sizes, especially with wide confidence intervals or large heterogeneity, are weak correlations and thus should be approached with caution before jumping to conclusions is NOT the same as saying they should be automatically dismissed out of hand. Those who argue otherwise don't seem to do nuance well, to put it mildly.