Education and Cannabis Use Share Genes — But in Opposite Directions for Use vs. Disorder | Research

The takeaway

Higher education shares genetic variants that promote cannabis use but protect against cannabis use disorder — a paradox where the same genetic background that leads to trying cannabis also buffers against addiction.

Genetics enthusiasts, addiction researchers, educators, and anyone curious about the biological basis of substance use risk.

What the researchers found

Education shared 84% of causal genetic variants with cannabis use disorder (CUD) but only 48% with cannabis use. The key distinction: variants linked to cannabis use mostly had concordant effects with education (71% same direction), while CUD variants mostly had discordant effects (only 38% same direction). Cognitive and non-cognitive components of education showed partially distinct patterns.

Why it matters

This explains a long-standing paradox: educated people are more likely to try cannabis but less likely to develop addiction. The genetic overlap suggests that traits enabling educational success (like cognitive flexibility and self-regulation) may both promote experimentation and protect against disorder.

The numbers in context

EA shared 48.07% of causal variants with cannabis use and 84.18% with CUD. For cannabis use: 71.42% concordant with cognitive EA, 65.56% with non-cognitive EA. For CUD: only 37.97% concordant with cognitive EA, 42.23% with non-cognitive EA. Functional enrichment varied across brain tissues.

How the study worked

Genomic analysis using bivariate causal mixture models, local genetic correlation analyses, and conditional/conjunctional false discovery rate analyses. Compared genetic architecture of educational attainment (cognitive and non-cognitive components) with alcohol consumption, AUD, cannabis use, and CUD using large genome-wide association study datasets.

What this study cannot tell us

Genomic studies identify statistical associations, not causal pathways. Primarily European-ancestry datasets limit generalizability. Education is a proxy for many socioeconomic factors. Genetic effects operate through complex environmental interactions.

How to read the evidence

Sophisticated genomic methodology using large datasets, but genetic associations are statistical and don't directly explain individual risk.

When this study was published

Published in 2026, using the latest available GWAS datasets for education and substance use traits.

The bigger picture

This genomic finding has real-world implications: the same people genetically inclined toward higher education may try cannabis at higher rates, but their genetic profile also protects against transitioning to problematic use. It challenges the 'gateway' narrative and suggests substance use trajectories are partly hardwired.

Questions still open

Could genetic profiling identify who is most at risk for transitioning from cannabis use to disorder? What specific cognitive or non-cognitive traits drive the protective effect? Do these genetic patterns hold across diverse populations?

Common questions

Are educated people more likely to use cannabis?

Genetically, yes — this study found that genetic variants promoting education also promote cannabis use (71% concordant direction). However, those same genetic backgrounds strongly protect against developing cannabis use disorder.

Is cannabis addiction genetic?

Partly. This study found that cannabis use disorder shares 84% of its causal genetic variants with educational attainment, though in mostly opposite directions. Genetics influences vulnerability but doesn't determine outcomes — environment matters enormously.

Read the original research

Education shares distinct genetic influences with substance use and disorder.

Psychological medicine, 56, e38

View on PubMed ↗Publisher’s paper ↗

Citation

Davis, Christal N; Khan, Yousef; Piserchia, Zachary; Gray, Joshua C; Kranzler, Henry R. (2026). Education shares distinct genetic influences with substance use and disorder.. Psychological medicine, 56, e38. https://doi.org/10.1017/S0033291726103353