In 2012, a widely publicized study from the Dunedin longitudinal cohort claimed that persistent cannabis use starting in adolescence could result in long-term declines in IQ. The findings were alarming: individuals who began using cannabis in their teens and continued into adulthood appeared to suffer cognitive deficits that remained even after they stopped using.
“Meier et al.’s estimated effect of adolescent-onset cannabis use on IQ is likely biased, and the true effect could be zero.”
Ole Rogeberg, 2013
Headlines followed. Public health warnings intensified. Opponents of cannabis legalization found empirical backing. But just one year later, economist Ole Rogeberg published a sharply reasoned critique of the study’s conclusions, raising a fundamental question: What if the observed link between adolescent cannabis use and IQ decline wasn’t due to cannabis at all—but rather to underlying socioeconomic differences?
Using simulation models, theoretical insights, and prior literature, Rogeberg argued that the association between cannabis use and lower IQ may be an artifact of unaccounted socioeconomic confounding. His paper invites a reconsideration of how we interpret observational data in cannabis research—and more broadly, how socioeconomic context shapes outcomes we often attribute to personal behavior.
The Original Study: Cannabis and Declining IQ
The 2012 study by Meier et al. tracked over 1,000 individuals in New Zealand’s Dunedin cohort from birth into their late 30s. IQ was measured in childhood and again at age 38. Participants were categorized based on their frequency of cannabis use and whether they had ever met the criteria for cannabis dependence.
Meier and colleagues found that those who began using cannabis in adolescence and developed persistent dependence experienced significant drops in IQ by midlife—up to 8 IQ points. Importantly, these results held even after controlling for variables such as sex, other drug use, and mental health conditions.
The authors concluded that adolescent-onset cannabis use might have neurotoxic effects on the developing brain.
Rogeberg’s Response: A Confounding Alternative
Rogeberg’s critique doesn’t deny the data—but it questions the interpretation. He proposes that the IQ differences observed in the Meier study might be caused by socioeconomic status (SES) rather than cannabis use per se.
The key insight: low-SES individuals are more likely to start using cannabis early and heavily, and they also tend to have different IQ trajectories for reasons unrelated to cannabis. In other words, adolescent cannabis users may be overrepresented in environments that systematically limit cognitive development—through less stimulating home environments, lower educational opportunities, and fewer cognitive demands in adulthood.
The original study controlled for some confounding variables, but not for SES in a robust or integrated way. This opens the door for a competing explanation: the observed IQ decline reflects preexisting differences in environmental and social conditions, not a direct neurobiological effect of cannabis.
Understanding IQ Change: The Flynn–Dickens Model
Rogeberg draws upon the Flynn–Dickens model of IQ development, which offers a dynamic view of intelligence. In this framework, IQ is not static; it’s shaped by ongoing interaction between the individual and their environment. Higher-IQ individuals seek out or are placed into more cognitively enriching environments, which further boosts their intellectual abilities.
Crucially, low-SES individuals may start with similar childhood IQs but diverge over time due to environmental constraints. If they end up in less cognitively stimulating jobs or educational settings, their IQs may stagnate or decline relative to peers.
In this view, IQ decline among cannabis users might simply reflect the environments they are more likely to inhabit—not the physiological effects of cannabis on the brain.
Simulating the SES Effect
To test this hypothesis, Rogeberg created a simulation model with the following assumptions:
• SES predicts likelihood and intensity of cannabis use.
• Low-SES individuals temporarily boost their IQ during schooling but experience a decline later due to lower cognitive stimulation in adulthood.
• There is no causal effect of cannabis on IQ.
When run repeatedly, this model reproduced the same patterns observed in the Meier study, including the “dose-response” relationship between cannabis use and IQ loss. This suggests that SES-based environmental influences could entirely account for the IQ declines previously attributed to cannabis.
Supporting Evidence from Other Studies
Rogeberg strengthens his argument by pointing to other longitudinal studies:
• A Canadian study with a less socioeconomically diverse sample (mostly middle class) found no lasting IQ effects from cannabis use.
• Another large study found no evidence that cannabis use predicted cognitive decline over a ten-year period—after adjusting for education and minority status.
• Several studies suggest that cognitive decline is more pronounced in low-SES populations, independent of cannabis use.
Collectively, these findings undermine the assumption that cannabis alone explains the IQ changes observed in the Dunedin cohort.
Education, Not Intoxication?
Rogeberg also suggests that some of the IQ decline attributed to cannabis may in fact result from interrupted educational trajectories. Adolescents who begin using cannabis early may be more likely to disengage from school, which could reduce future cognitive development.
If this pathway is correct, the observed IQ losses may be indirect and preventable. They would stem not from cannabis itself but from reduced exposure to cognitively enriching environments—particularly education.
This mechanism would have significant implications. It shifts the policy conversation from cannabis’ potential neurotoxicity to the social and structural factors that influence youth behavior and opportunity.
The Policy Implications
The distinction between causation and correlation is more than academic. If cannabis directly harms the brain, the argument for strict youth prevention becomes one of biological protection. But if the real harm comes from social conditions and developmental detours, then the appropriate interventions may lie in education, family support, and community investment—not just drug policy.
This also has implications for interpreting cannabis research broadly. Observational studies often show correlations between cannabis use and various negative outcomes—from academic failure to mental health issues. But without accounting for deep, structural variables like SES, these associations risk overstating cannabis’ causal role.
The Scientific Standard: Better Models, Not Bigger Headlines
Rogeberg’s work is a reminder of how easily compelling results can lead to premature conclusions. It’s not that the original Dunedin study was flawed in data collection; rather, its analytical framework did not fully account for the complexities of human development and social stratification.
His critique underscores the need for rigorous causal modeling in cannabis research—especially when findings are used to shape public opinion or legislation. This includes integrating multivariate controls, testing for interactions, and considering long-term trajectories shaped by environment, not just individual behavior.
It also suggests that IQ should not be treated as a static or purely biological measure. Intelligence evolves with experience, and experiences are shaped by opportunity.
This critical reassessment of the cannabis-IQ link brings nuance to a charged topic. It challenges us to look beyond surface-level associations and engage with the deeper forces that shape human development. In doing so, it reframes the question from whether cannabis dulls the brain to whether society enables every brain to reach its potential.
