Alex Berenson’s new book had a brief moment in the sun earlier this year, with favorable coverage in Mother Jones, a cordial review by Malcolm Gladwell in The New Yorker, and an essay from Berenson promoting the book in the New York Times, before being shouted down by a chorus of online critics. The critics had a point: This is a book that aims to marshal reams of statistical evidence documenting a link between marijuana and psychosis — and in turn one between marijuana and violence — and Berenson, to put it bluntly, is not particularly skilled when it comes to evaluating such evidence.
And yet, though he vastly overstates it, Berenson has a point too: There are signs that marijuana use can cause both temporary psychosis and full-blown schizophrenia, and we need to study this possibility as states debate whether and how to decriminalize or legalize the substance. We may ultimately decide that the risks of smoking pot are not serious enough to justify interfering in people’s personal decisions, much less incarcerating them. Indeed, on the basis of what we know now, that is my own view. But for all its flaws, Tell Your Children forces us to confront the idea that weed might not be as harmless as we think.
Ultimately, establishing marijuana’s connection to psychosis and violence boils down to the age-old quandary of how to distinguish causation from mere correlation. That there is a correlation between pot use and these bad outcomes is well established: Pot users, especially the heaviest users, are more likely to be schizophrenic or violent than non users. But the question of whether these relationships are causal is far trickier.
In a comprehensive and carefully written 2017 report reviewing the evidence on marijuana’s health effects, the National Academy of Medicine noted that when substance use and mental illness overlap, there can be three different reasons: 1) Substance use changes the brain in ways that cause mental illness; 2) mental illness causes people to use substances (say, to self-medicate); and 3) there are environmental or genetic influences that predispose people to both substance use and mental illness. “Although the precise explanation is still unclear,” the authors wrote, “it is reasonable to assume” that the reality is a mixture of these. In true academic form, the authors then called for more research to untangle this knot.
On the specific issue of pot and psychosis, while the report is quite confident in the correlation between the two, it rarely uses causal language and hedges when it does. For example: “Cannabis use is likely to increase the risk of developing schizophrenia and other psychoses” (emphasis mine). The authors also note that while pot use predicts the future development of psychosis — indicating that self-medication among the mentally ill isn’t the whole story — it’s also true that some societies have seen marijuana introduced to them without experiencing an increase in psychosis. In other words, while pot may indeed increase the risk of psychosis, we’re not quite sure it does, much less to what degree. And the degree is important too: A common estimate is that 1 percent of the population has schizophrenia, for example, and there’s a big difference between marijuana’s increasing that risk to 1.1 percent and increasing it to, say, 3 percent.
Berenson takes all that hedging and nuance and chucks it out the window at some stoner minding his own business on the sidewalk. An entire section of the book is called “Proof”; one chapter is entitled “Study after Study after Study.” And no, he doesn’t have access to some groundbreaking new scientific technique that justifies such confidence.
A good illustration of the problem is a piece of research published in The Lancet in 1987, a seminal work that Berenson quite rightly puts under the spotlight for several pages. Using a massive data set on Swedish military conscripts, it found that people who’d used pot at least 50 times were at six times the risk of developing schizophrenia of those who’d never used. It also found that the relationship persisted after adjusting for assorted control variables such as conscripts’ socioeconomic status and whether their parents were divorced.
But that’s not the end of the story. Among those who’d used ten times or more, the “relative risk” fell from about 4.6 to 2.3 times higher when the control variables were added. (The study does not report these results for the 50-times-plus group alone.) And while the remaining 2.3-fold difference might be there because pot really does cause schizophrenia, it also might be there because the control variables are inadequate. Asking someone whether any family members take medication for “nervous problems” (another variable from the study) is not the same as accounting for that person’s full genetic makeup; there is also no way to measure every relevant aspect of a person’s environment. Even some of the variables that are measured are not measured perfectly, because they are based on self-reports — and a variable measured with error can’t truly “control” the thing it represents.
We just don’t know whether all, none, or some — and if some, how much — of that 2.3-fold increase is the result of pot causing schizophrenia. And the fact that the relative risk falls so dramatically with controls is highly troubling, because it indicates that there’s a lot of statistical “confounding” in what we can measure.
Admirably, Berenson discusses the various limitations of the study. But then he just carries on, piling up more research that suffers from similar problems and arrives at similar conclusions and pretending it solves something. A whole lot of the studies he cites follow the same script: There’s a correlation between weed-smoking and either psychosis or violence; it doesn’t go away (but does shrink) when the researchers toss some control variables into the mix; this is treated as serious evidence of a causal relationship, with a caveat here and there but also strident assertions from Berenson that the pathway from pot to these bad outcomes is definitively established. And this pattern is even more grating with violence than it is with schizophrenia. People with criminal tendencies are more likely to do drugs? Who knew?
Berenson’s effort to analyze the issue geographically fares no better. Much of the online backlash earlier this year, for example, involved his claim that between 2013 and 2017, the four states that legalized recreational pot all saw above-average increases in violence. In his defense, it is a bit odd that all four pot states saw this. But this analysis is an incredibly crude exercise, especially since these four states implemented their laws at different points during this period, and it flies in the face of the fact that crime has generally fallen as pot use among young men, our most violent subpopulation, has ticked up since the turn of the century.
What’s particularly frustrating is that elsewhere in the book, Berenson has a professor friend run a more sophisticated analysis of geographic trends in opioid deaths as they relate to medical-marijuana laws and to marijuana use as measured by surveys; a study had claimed that medical-marijuana laws reduce opioid overdoses, but Berenson’s friend adds more data (as the original study ended with 2010) and finds no effect. You can do the same type of thing with crime rates, but there’s no such analysis to be found here. The same goes for the little-noticed but alarming recent upticks in “serious mental illness” in surveys and psychosis cases at hospitals that Berenson notes: Were they concentrated in states that legalized pot or saw particularly large increases in illegal use? The data exist, but Berenson doesn’t have his friend crunch them.
Maybe the idea came too late in the process. Berenson’s uncompensated research assistant, a political-science and law prof at New York University, gives a “long pause” after one request and our favorite buzzkillin’ journalist realizes he’s out of favors.
To do this topic justice, Berenson didn’t just need a numerate friend willing to donate some time; he needed a co-author or at least a paid helper who specializes in drug policy: someone familiar with the quirks of the relevant data, willing to run new analyses, and able to help him assess the various studies he writes about. He also needed footnotes citing all of the studies he refers to, and an appendix with the full methodology and results of the original analyses. Then again, if he’d done all that, he’d have written something a lot more similar to the aforementioned National Academy of Medicine report.
When we look to the future, other lines of evidence are rapidly emerging and could provide stronger answers. Berenson discusses a sophisticated technique called “Mendelian randomization” inaccurately, briefly, and vaguely. (He claims a study “rule[s] out” a genetic explanation for the marijuana–schizophrenia link despite the fact that different studies have reached different conclusions, and thanks to the lack of footnotes I’m not even sure which study of several in this vein he’s referring to.) To oversimplify, the technique involves looking to see whether people with genes for schizophrenia are more likely to use pot and whether people with genes for pot use are more likely to develop schizophrenia and trying to infer a causal relationship from there. Some early results suggest the effect might go both ways — i.e., schizophrenia causes pot use and pot use causes schizophrenia — but there’s much work left to be done, including figuring out which genes even to take into account in such analyses.
So where does this leave us? Berenson thinks it obvious that legalizing pot is a bad idea. As someone who changed his mind in the wake of the opioid epidemic about across-the-board drug legalization, I ought to be at least somewhat sympathetic. I am not. Easily available opioids set off an epidemic that killed nearly 48,000 Americans in 2017, almost triple the number who died by homicide of any kind; marijuana, by contrast, may increase the risk of schizophrenia from 1 percent or so to some higher number in heavy users, and also may occasionally trigger violent episodes (though in other cases one imagines it might prevent violent episodes through its more typical effect of mellowing users out).
You can tell your children not to smoke pot on the basis of these risks, certainly, and punish people who drive while high, get aggressive when they smoke, provide pot to minors, or switch to hard drugs such as cocaine. Reasonable people might even argue for a strictly limited form of legalization (as Jonathan Caulkins has in this magazine), or at least to tax marijuana heavily or regulate its potency. But to my mind this is simply not worth trampling on personal freedom over. Smoking marijuana is well within the range of risks we should allow adults to take.
I can’t recommend you read Tell Your Children; the National Academy of Medicine report, available for free online, provides a far more balanced and scientifically rigorous summary of the mental-health effects of pot. It must be said, however, that with his more, ahem, aggressive approach, Berenson has drawn attention to a significant potential risk of marijuana abuse, one that deserves close study and careful monitoring as pot laws inevitably continue to loosen.
This article appears as “Demon Weed?” in the March 25, 2019, print edition of National Review.