Academics depict the peer-review process as the gold standard for intellectual honesty, ensuring published scholarly work is unbiased and accurate. But ideological conformity makes peer review a far thinner defense than advertised.
In January I published a book about the mental-health and violence risks of cannabis. Several dozen scholars signed a petition expressing in unison their objection to my work. Thus I’ve recently spent an inordinate amount of time reading papers seeking to prove that marijuana is a cure-all whose deleterious consequences are a figment of our collective imagination. The shoddiness of much of the work has shocked me.
Example: Driving deaths have risen more than 30% in Colorado and Washington, the first states to legalize marijuana for recreational use, since dispensaries opened there in 2014. That rise is more than double the national change. The Insurance Institute for Highway Safety reported in October that the number of vehicle accidents overall were up faster in legalized states than the rest of the nation.
Nonetheless, researchers have claimed legalization doesn’t increase driving deaths. Last year, Benjamin Hansen of the University of Oregon and two other researchers wrote a paper looking at the rise in vehicular deaths in which drivers had THC in their blood. They found such deaths rose in Colorado and Washington at rates “similar” to those in other states.
In fact, the data showed that legalization explains about half the increase in Colorado and Washington. But because the data were limited, the results had wide confidence intervals, which means it’s possible but not probable that the post legalization gap was the result of chance. A more honest way to report the results would have been to say that the paper had found a worrisome trend in cannabis-linked deaths, which more data might confirm.
Violent crime has also soared in the legalized states since 2013. Yet last month two criminologists claimed in a Seattle Times op-ed that they had found “no increase in violent crime that can be directly attributed to marijuana legalization.”
That formulation makes the statement a trivial truth. Without examining every murder, no one can say legalization has directly driven the increase.
The authors explain in the op-ed that they based their statement on a paper they co-wrote. They don’t name it, but the only published paper listing them as co-authors is a 2018 study called “Marijuana Legalization and Crime Clearance Rates.” It didn’t even examine whether crime had risen or fallen. It looked only at whether police were likelier to solve, or clear, crimes after legalization. But so what? If murders double from 100 to 200, and the police solve 50 the first year and 110 the second, the clearance rate has risen, but so has the crime rate.
Some of the research is comical. A paper published in March reported that cannabis use seemed to increase “satisfactory orgasms” in women. The paper divided subjects into three categories: nonusers, users who answered yes when asked if they used the drug “before sex,” and users who answered no. But it presented results on sexual satisfaction only from the last two categories—not from women who don’t use marijuana at all. The paper provides no data at all comparing users with nonusers.
I could offer a half-dozen other examples of dubious scientific practices—using nonstandard data sets, relying unnecessarily on “synthesized comparators” that are inherently vulnerable to manipulation, and “p-hacking” by looking at endless secondary outcomes until one pops up that reaches the threshold for statistical significance.
The tricks can be hard to find—and journalists, who are almost never trained in science or statistical analysis, often parrot the results unskeptically, especially when the findings confirm their own biases toward ideology or sensationalism. When car accidents and violent crime are involved, the results can be deadly.
Mr. Berenson is a former reporter for the New York Times and author of “Tell Your Children: The Truth About Marijuana, Mental Illness, and Violence.”