No, Weed Doesn't Make You Violent -- Despite the Scary Headlines

October 24, 2026

Photo by lilartsy via Unsplash.

A new meta-analysis out of King's College London is making the rounds with a headline number that sounds damning: cannabis users are roughly twice as likely to have committed an act of violence as non-users in the general population, and about 2.5 times as likely among people in psychiatric care. Researchers pooled data from 63 studies covering more than 265,000 people to get there. That's a serious amount of data, and it's not nothing.

The Daily Telegraph ran with the framing you'd expect -- cannabis users "twice as likely to be violent" -- the kind of headline built to travel fast and worry parents. Marijuana Moment republished a pushback op-ed from Simon C Moore, a public health researcher at Cardiff University, that picks apart what that number actually represents and what it doesn't. His argument isn't that the study is junk. It's that the real story here is about study design and how shaky evidence gets flattened into a scary multiplier, not a clean yes-or-no verdict on whether cannabis causes violent behavior.

Worth asking before any of the numbers get thrown around: when two things show up together a lot in the same population, does that mean one is causing the other, or just that they tend to travel together? That distinction does almost all the work in this story, and it's exactly the part that got dropped on the way to print.

What the King's College London Review Actually Found

Photo via Pexels.

The review itself was published in Psychological Medicine (DOI 10.1017/S0033291726105856), led by Giulia Trotta, Marta Di Forti, Robin Murray and colleagues at the Institute of Psychiatry, Psychology & Neuroscience, King's College London -- a group with a long track record studying cannabis and mental health, including Di Forti's prior work on high-potency cannabis and psychosis risk. The accompanying press release went out September 29, 2026, with funding credited to the NIHR's Maudsley Biomedical Research Centre and the Medical Research Council, two legitimate and well-established funders.

The scale of the review is genuinely impressive on paper. Sixty-three studies, split between general-population samples and people already in psychiatric care, covering upward of 265,000 individuals total. Pooling that much data is exactly how meta-analyses are supposed to work -- you combine smaller, noisier studies to get a more stable estimate than any single one could provide on its own.

The topline figures are what generated the headlines: roughly double the odds of violence among cannabis users in the general population, and about 2.5 times higher odds in psychiatric populations, where baseline rates of both substance use and violence run higher to begin with. To their credit, the authors themselves added an important caveat that got lost in most coverage -- much of the underlying research predates today's high-potency cannabis products, meaning the data may not reflect what's actually being consumed now. Their actual policy recommendation was narrow: build cannabis-use questions into violence-risk screening in psychiatric settings. That's a far more modest ask than the headlines implied.

The Problem: Most of the Studies Can't Tell Cause From Coincidence

Photo via Pexels.

Here's where the review gets shaky, and where Moore's rebuttal does its most useful work. Of the 63 studies included, 46 were cross-sectional -- meaning researchers took a single snapshot of a population at one point in time and looked for correlations. A cross-sectional study can tell you that cannabis use and violent behavior show up in the same people more often than chance would predict. It cannot tell you which came first, whether a third factor is driving both, or whether the relationship is even stable over time.

And those are precisely the studies that produced the strongest associations in the pooled data. That's not a coincidence -- it's a known pattern in this kind of research. Weaker study designs tend to generate bigger, scarier effect sizes, partly because they can't control for the messy reality of people's lives the way a longitudinal design can.

The longitudinal studies -- the ones that actually track the same people over months or years -- told a much less dramatic story. Those found about a 17% higher odds of violence among cannabis users, nowhere close to double. That gap between study types is, in Moore's view, the real headline the Telegraph buried in favor of the bigger number.

It's worth sitting with the analogy he raises: plenty of things correlate with violence in a one-time survey without causing it. Unemployment. Heavy drinking. Poverty. Take a snapshot of any struggling neighborhood and you'll find all of these clustered together with higher rates of conflict, without anyone seriously arguing that losing a job turns a person violent on its own.

Same Survey, Wildly Different Numbers

Using the same US survey data, reported odds ratios for violence varied wildly across analyses, ranging from 1.3 to 9.8, highlighting how analytical choices can drastically change the apparent strength of an association.

One detail in the review deserves more attention than it got: three separate analyses, all drawing on the same US national drug survey, produced wildly different odds ratios -- ranging from about 1.3 up to nearly 10. That's the identical source data producing results nearly eight times apart depending on how the question was asked.

The likely explanation isn't mysterious. Studies don't agree on basic definitions. One-time cannabis use, past-month use, and daily heavy use get lumped together as cannabis use in some papers and carefully separated in others. The same inconsistency plagues the violence side of the equation -- a shoving match, a bar fight, a robbery, a domestic incident, a sexual assault conviction can all end up coded as violence in the same category, even though they represent wildly different behaviors, contexts and severity.

When a review collapses that much variation into one pooled number, precision gets lost fast. And this isn't a minor footnote buried deep in a methods appendix -- it's reflected in the supplementary material, where the authors themselves rate the overall certainty of the evidence as very low. That rating is standard practice in evidence synthesis, and it's meant to be read alongside the topline figure, not ignored. It didn't make it into the Telegraph's coverage.

Once you know a single dataset can be sliced into results that differ by a factor of seven or eight depending on how researchers define their terms, the eye-catching twice as likely headline stops carrying much independent meaning. It's a number built on an average of very different things.

Why This Matters for Policy and Public Perception

Photo by Trac Vu via Unsplash.

None of this is academic hairsplitting. Alarmist cannabis causes violence framing has real downstream consequences -- it shows up in state and national legalization debates, in sentencing arguments, and in custody and family court decisions where a parent's cannabis use gets treated as evidence of a violence risk. Cannabis's legal status still varies enormously by state and by country, and readers trying to understand their own situation should check current local law rather than lean on a headline statistic to make assumptions about risk or legality.

The more plausible explanation for a lot of this association is confounding. Co-occurring mental illness, other substance use, poverty, and prior trauma are all strongly linked to both cannabis use and violent behavior independently. Populations carrying more of those risk factors will show higher rates of both, without cannabis itself doing the causal work.

That's what makes the KCL researchers' own stated recommendation more defensible than the way it got reported: screening for cannabis use as one factor among many in psychiatric violence-risk assessments is a reasonable, narrow clinical response to this data. A blanket claim that cannabis makes people violent is a different thing entirely, and the review doesn't actually support it.

Good science journalism on drug research means reporting the caveats -- study design, certainty ratings, the gap between cross-sectional and longitudinal findings -- right alongside the scariest number, not as an afterthought buried past the headline.

Strip away the headline multiplier and the honest takeaway from the KCL review is fairly modest: cannabis use and violent behavior show up together more often in certain populations, particularly in cross-sectional snapshots of people who often carry other serious risk factors too. That's a finding worth studying further. It is not evidence that lighting up turns someone violent, and the review's own authors never claimed that.

The next time a single number from a meta-analysis gets flattened into a tabloid headline, it's worth asking what got left out -- how many of the underlying studies could actually establish cause and effect, and what certainty rating the researchers themselves assigned to their own findings. In this case, both answers undercut the scary framing considerably.

Tightening up risk-assessment practices in psychiatric care, where clinicians are already weighing multiple factors for a specific clinical purpose, is a sensible response to data like this. Using the same data to justify broader stigma against cannabis users generally is not -- and it's not what the research, read carefully, actually supports.