Last year, major headlines blared that fatal car crashes spiked on April 20—an internationally recognized day for cannabis protest—and put the blame on "4/20" rallies.
Well, it turns out the original study has been thoroughly debunked!
A new analysis shows there is no increase in fatal car accidents on April 20, and the original study was very obviously flawed.
Of course the original story generated massive international headlines, scare-mongering about how 4/20 was causing deadly car accidents. The follow-up study, however, got no media coverage at all.
Economics professor Chris Auld at the University of Victoria outlined the new study in a compelling twitter thread.
A rant about the state of statistical work in (clinical) medical journals: 4/20 "Weed day" edition.— Chris Auld (@Chris_Auld) February 23, 2019
Last year @JAMAInternalMed published a paper purporting to show that on April 20 traffic accidents spike, due to stoned drivers:https://t.co/1bp1faetir
The original study compared fatal traffic accidents after 4:20 p.m. on April 20 with that of the same time on dates a week ahead and following the celebration—April 13 and April 27. The study revealed a slightly higher rate on April 20, and jumped to the conclusion that the 4/20 celebrations were the cause.
Below is Auld's visual representation of the data from the original study.
This slight increase on April 20 was the basis of headlines like “4/20 doubles risk of fatal accidents" and “Fatal accidents see 'dramatic' spike after 4/20 celebrations".
Auld made another chart showing average car accidents for each day in April. You can see that the 20th is nothing special.
As Auld points out, comparing April 20 to April 13 and 27 is a sample size of three. Yet researchers and media treated this limited data as if it was proof positive that 4/20 was a big day for deadly car accidents.
University of Toronto professor Dr. Donald Redelmeier, one of the authors of the original study, made some audacious claims to the media.
"Our results suggest that drug use at 4/20 celebrations more than doubles the risk of a fatal crash,” he said, based entirely on this very limited data.
The new study also applied the same technique as the original study to every day of the year, comparing car accidents to the week before and the week after. They found April 20 doesn't stand out, but dates like Independence Day (July 4), Thanksgiving (October 14), and Christmas (December 25) do.
Even without this extra analysis, the original study had some other obvious problems, which were ignored when it came out. For example, states with no 4/20 celebrations, like Hawaii and Maine, had higher rates of fatal crashes on April 20, but states with substantial 4/20 events, like California and Colorado, actually showed slightly less.
The reality seems to be that there is a natural random variation in terms of daily car accidents, and any differences around April 20 seem to be within said deviation.
Another graph from Auld shows the average fatal accidents in April by day, from lowest to highest. Auld explains: "the paper compared the green bar (4/20) to the red bars (4/13 & 4/27) and concluded the green bar is unusually high."
Will the dozens of major media outlets which gave the original flawed study major headlines devote any attention at all to this new study? Will they correct the false message about April 20? Don't hold your breath!