Legal, illegal and everything in between: how to judge cannabis policy

Few policy arguments are as old, or as stuck, as the one about cannabis. One camp says the law protects people, above all the young, from a harmful drug. The other says prohibition has failed, criminalised millions of people and handed a profitable trade to organised crime. Both camps can point to studies, and both have at times chosen the studies that suit them.

This series takes a different route. Across five articles it works through a critical comparative review of cannabis policy that asks a single question: judged against their own stated aims, what did prohibition and its alternatives actually achieve? It does not ask whether people ought to use cannabis. It asks whether the laws did what their supporters said they would, and at what cost. The review draws on peer-reviewed epidemiology, quasi-experimental policy evaluations, official statistics and treaty documents from the United Kingdom, the United States, Canada, Uruguay, Portugal, the Netherlands and Germany.

This first article lays the groundwork. It explains why "legal versus illegal" is too blunt a tool, what kinds of evidence can answer the question, which traps run through the whole literature, and how the world arrived at criminal prohibition in the first place.

One plant, many laws

The legal status of cannabis has changed more often than its pharmacology. For most of recorded history the plant was legal or governed by custom. Colonial states taxed it. From 1925 it was progressively brought under international criminal control, and since 1996 that control has been partially dismantled. The same plant, used by broadly similar populations, has been governed by sharply different regimes, sometimes within a single generation. The review calls this "something close to a natural experiment in drug policy". Societies cannot be put in a laboratory, but when one country changes its law and a similar neighbour does not, researchers get something approaching a controlled contrast. The words "close to" matter: conditions are never identical, and much of this article is about reading such contrasts honestly.

What each side promised

The test is internal: each regime is judged by its own promises. Prohibition promised to reduce use and protect health, especially among the young. The Single Convention on Narcotic Drugs 1961 frames control as a duty to protect "the health and welfare of mankind" (United Nations, 1961), and in practice prohibition pursued three operational aims: to reduce the number of people using cannabis, to deter young people from starting, and to suppress supply. Reform promised something different: to reduce criminalisation, to displace illicit markets and to allow the products themselves to be regulated.

A prohibitionist cannot fairly be held to the reformers' goals, nor a reformer to the prohibitionists'. But each can be held to their own. The review is a companion to a narrative history of cannabis that this site has already published as a separate series; the twentieth-century part of that story is in From Calcutta to Geneva. Here the aim is to move from description to evaluation.

Why "legal versus illegal" is too simple

The review argues that the binary of "legal" and "illegal" "hides the most important policy differences". The literature distinguishes at least six regimes (Room et al., 2010; Rolles and Murkin, 2016), and the review's typology sets out seven. They differ on three questions: is using cannabis a crime, is supplying it lawful, and who profits from supply? Treating them as one category is, the review warns, a common source of error in public debate.

Reading the table

A few rows are routinely confused.

• Depenalisation keeps possession as an offence but softens how it is handled, through low-level penalties or police discretion. The law on paper barely changes; what changes is what happens to someone caught with a small amount. The UK sat here between 2004 and 2009, when cannabis was in Class C: possession remained an offence throughout.
• Decriminalisation makes possession for personal use an administrative matter, handled outside the criminal courts, while supply remains a crime. Portugal did this for all drugs in 2001. It is often mistaken for legalisation, but under decriminalisation there is still no lawful way to buy the drug.
• Tolerated retail, the Dutch coffeeshop system since 1976, lets shops sell at the counter while wholesale supply stays illegal. The shop sells at the front door but has to buy from somewhere at the back, a gap the third article examines.
• Non-commercial and commercial legalisation both make possession lawful within limits and provide lawful supply. The difference is who supplies: the state, pharmacies, clubs or home growers in Uruguay and Germany, a licensed for-profit industry in Canada and many US states.

Why the distinctions matter

Evidence from one row cannot stand in for another. Findings about Portuguese decriminalisation cannot be transferred to Colorado's commercial market, and the reverse is also true (Kilmer, 2017; Hall and Lynskey, 2020). Decriminalisation changes what happens to users who are caught; it does not create shops, brands, advertising or a new industry. Commercial legalisation does all of those things.

This is why "legalisation" can mean quite different things in the same argument. Someone warning that it will bring aggressive marketing of high-strength products is usually thinking of the bottom row. Someone saying it will stop people being criminalised for possession may be thinking of the third row, which is not legalisation at all. Much public argument is two people defending different rows of this table while believing they disagree about the same policy.

What the review covers

The review focuses on the United Kingdom, where its author is based, and on the jurisdictions that provide the most informative comparisons: the United States, Canada, Uruguay, Portugal, the Netherlands and Germany. Between them they cover every row of the typology. It considers six outcome domains: prevalence and patterns of use; product potency; illicit markets and crime; criminal justice and equity; health; and medical access. Debates often fix on only one of these, but a policy can perform well on one and badly on another, and a fair assessment weighs them together.

In the UK, cultivation and possession of cannabis remain criminal offences, and growing it requires a Home Office licence even for patients with a prescription. This series is analytical and gives no guidance on obtaining, growing or using cannabis.

How the review weighs evidence

The review describes itself as "a critical comparative policy review, not a systematic review". A systematic review follows a fixed procedure: search terms set in advance, exhaustive database searches, explicit inclusion rules and a report of every study found. A critical review is more selective. Its sources came from targeted searches of peer-reviewed journals such as The Lancet Psychiatry, JAMA Pediatrics, Addiction and the British Journal of Criminology, official statistics from bodies such as UNODC, Health Canada and the Home Office, and statutory evaluations such as Germany's EKOCAN programme. That lets it range across epidemiology, criminology and treaty law, at the cost of depending on judgement; the review itself notes that it may underweight non-English research. One rule is worth highlighting: advocacy material from both prohibitionist and reform organisations was used only to locate claims that needed checking, never as evidence.

Evidence was weighted by how well it could separate the effect of a policy from background trends. The five tiers below, in descending order, recur throughout the series.

Tier 1: comparing places that changed with places that did not

The strongest evidence comes from quasi-experimental designs such as difference-in-differences and synthetic control (for example Anderson et al., 2019; Laqueur et al., 2020). They tackle a basic problem. If teenage use fell after a US state legalised, was that the law, or was use falling everywhere? To know what a law did, you need to know what would have happened without it.

Difference-in-differences measures the change over time in places that changed their law and the change over the same period in places that did not. Only the extra change in the reforming places, beyond what happened elsewhere, is attributed to the policy. One study discussed later compared Uruguayan and Chilean secondary-school students from 2007 to 2018 in exactly this way (Rivera-Aguirre et al., 2022). The method assumes that, without the reform, the two groups would have moved in parallel; if the reforming places were already on a different path, the comparison misleads.

Synthetic control helps when there is one reforming country and no obvious twin. Researchers build a weighted blend of several comparison countries that tracked the reforming country closely before the reform. This "synthetic" country shows what might have happened had the law not changed, and the gap that opens afterwards is the estimated effect. Laqueur et al. (2020), for example, found that perceived availability of cannabis among Uruguayan adolescents rose after legalisation: 58% observed against 51% in the synthetic comparison.

Neither is a true experiment. Governments do not assign laws at random, and the review notes that jurisdictions which legalised tended to have more liberal attitudes and rising use before reform. But these designs are the best available way of isolating a policy's effect.

Tier 2: whole populations over time

Next come population-wide administrative cohorts analysed with interrupted time series (for example Myran et al., 2022; Myran et al., 2025). An administrative cohort uses records health systems already keep, such as emergency visits or diagnoses, for an entire population; one Ontario study covered more than 13.5 million people.

An interrupted time series looks at a long run of measurements before and after a policy change and asks whether the change interrupted the pattern. Did the line jump at the moment of reform, or steepen, or carry on as before? Its great virtue is that it makes pre-existing trends visible. The Ontario schizophrenia study discussed later shows why: the authors stress that the rise they found was steady over the whole period, with no clear acceleration at the point of policy change (Myran et al., 2025). The weakness is that a single place has no built-in comparison group, so anything else that changed at the same moment can be confused with the law.

Tier 3: case-control studies and odds ratios

Third come multi-site case-control studies (for example Di Forti et al., 2019). A case-control study starts from the outcome. Researchers recruit people with a condition, such as a first episode of psychosis, and a comparison group from the same population without it, then compare how often each group had been exposed to something, such as daily cannabis use. It is an efficient design for uncommon conditions, where following a population forward until enough cases appear would take decades.

Such studies produce an odds ratio. Odds express chance as a ratio: if 1 person in 5 has a characteristic, the odds are 1 to 4. An odds ratio compares the odds in two groups; 1 means no difference, above 1 means higher odds among the exposed, below 1 lower. When the outcome is uncommon, it is close to saying how many times as likely something is, which is how the review describes daily users in the EU-GEI study as about three times as likely as never-users to have a psychotic disorder (Di Forti et al., 2019). Odds ratios appear in other designs too: one US study reports an odds ratio of 0.92, which the review translates as an 8% decrease in the odds of high-school students using cannabis (Anderson et al., 2019).

Two cautions apply. Results reported after adjustment have been corrected statistically for other differences between groups, but only for differences that were measured, which is why a case-control study cannot by itself prove causation. And every estimate comes from a sample, so it carries uncertainty, usually expressed as a confidence interval: the range of values reasonably compatible with the data. A narrow interval means a precise estimate; an interval for an odds ratio that comfortably includes 1 means the data cannot rule out no difference. The review, like this series, mostly reports headline estimates, so it is worth remembering that each single number stands for a range.

Tier 4: repeated national surveys

Fourth come repeated national surveys (for example Health Canada, 2024; Caulkins, 2024), which ask large samples about their use year after year. They are indispensable for tracking how many people use cannabis, how often and where they buy it. But on their own they describe trends rather than explain them, and they depend on people answering honestly, a problem discussed below.

Tier 5: expert consensus and structured judgement

Last come expert consensus reports and structured expert judgement (for example National Academies of Sciences, Engineering, and Medicine, 2017; Nutt, King and Phillips, 2010). A consensus report has a committee grade the strength of existing evidence: an efficient guide, but only as good as the studies it summarises.

Structured judgement asks experts to score what cannot easily be measured. The best-known example is the multicriteria decision analysis by Nutt, King and Phillips (2010), which scored 20 drugs on 16 criteria of harm to users and to others. The method breaks a vague question, such as which drug is most harmful, into many specific ones, has a panel score each drug on each, weights the criteria and combines the results. Its strength is transparency: every judgement can be challenged. Its weaknesses, as the review notes, are that results depend on the judgement and weighting of a small expert panel, and that it scores harms as they occurred under existing laws, conflating the harms of a drug with those of its policy. The final article returns to what this ranking shows.

The ranking is not a claim that surveys or expert panels are worthless. It concerns one thing only: how well a study separates the effect of a policy from everything else going on. A useful rule of thumb for any headline about cannabis law is to ask what the claim was compared against.

Four traps that run through the evidence

The review identifies four methodological hazards that recur throughout this literature and shape every conclusion in the series.

Pre-existing trends. Use of cannabis was already rising in many countries before reform. A post-reform increase is not, by itself, evidence that reform caused it. If use was climbing at the same rate before and after a law changed, the law has not obviously done anything. The same applies in reverse when reformers claim credit for a fall that was already under way.

Reporting effects. When use becomes legal, people become more willing to report it, so surveys may show more use partly because people stopped hiding it. Caulkins (2024) explicitly notes that part of the long-run rise in reported daily use in the United States may reflect this, though not all of it. The honest conclusion is that such a rise is real in part, by an uncertain amount. This is one reason records that do not depend on self-report, such as hospital visits, carry particular weight.

Potency drift. The product itself changed across the period. Comparisons of "use" in 1975 and 2025 compare exposures to quite different substances (Freeman et al., 2021). A meta-analysis discussed in the next article found that the THC content of herbal cannabis rose by 0.29 percentage points a year from 1970 to 2017, and that of resin by 0.57 points a year (Freeman et al., 2021). A survey question about past-year use treats a weak product and a strong one as the same.

Short follow-up. Most commercial markets are less than a decade old. Arrests fall the moment possession stops being a crime, but effects that depend on cumulative exposure, such as dependence or psychosis, may not yet be visible. The absence of a visible harm in a young market is therefore provisional.

The review stresses that these hazards cut in both directions. They weaken claims that reform has been harmless, and they equally weaken claims that it has been disastrous. A reformer pointing to stable youth surveys must reckon with short follow-up; a prohibitionist pointing to rising adult use must reckon with pre-existing trends and reporting effects. Neither side can use the hazards only against the other.

A short history of legality and illegality

Criminal prohibition is a short episode in the long history of cannabis. The review divides that history into four broad periods; the companion history series covers them in detail.

The era of international criminal prohibition spans about seventy years. Since 1996 the system has fragmented, with countries choosing different rows of the typology for different reasons. The review draws out two features of this sequence that matter for evaluation.

Diplomacy and campaigning more than evidence

First, the move into prohibition was driven mainly by imperial diplomacy and moral campaigning, not by evidence of harm. The official inquiries of the period, from the Indian Hemp Drugs Commission of 1894 to the Wootton Report of 1968 (Advisory Committee on Drug Dependence, 1968), repeatedly reached more moderate conclusions than the laws that followed (Mills, 2003). British support for the 1925 Geneva controls owed more to imperial diplomacy than to evidence of harm, while in the United States the campaign that produced the Marihuana Tax Act of 1937 drew heavily on racialised imagery of Mexican migrants and Black musicians (Mills, 2003). The review places this within the sociological concept of moral panic (Cohen, 1972), in which a group is cast as a threat to social values and the reaction becomes disproportionate to the danger.

It is important to be precise about what this shows. As the review puts it, these origins do not by themselves show that prohibition was wrong; a law adopted for poor reasons can still turn out to work. What they do explain is why its legitimacy has been persistently contested and why its enforcement has fallen so unevenly, the subject of the next article. They also justify this series' method: if a policy was not founded on evidence, the fair test is how it has performed against its stated aims since.

Class B, Class C, Class B

Second, the prohibition era was never uniform. Under the Misuse of Drugs Act 1971, controlled drugs in the UK are placed in Class A, B or C, with different maximum penalties, and the Act set up the Advisory Council on the Misuse of Drugs (ACMD) to advise on classification. Cannabis was in Class B. In January 2004 it was moved to Class C, and in January 2009 back to Class B (Advisory Council on the Misuse of Drugs, 2008). In the terms of the typology, the UK spent five years in the depenalisation row before returning to full prohibition.

The return to Class B was made against the advice of the ACMD. The review concludes that the reversal shows how far classification decisions have been political rather than scientific (Advisory Council on the Misuse of Drugs, 2008; Nutt, King and Phillips, 2010). The authors of the multicriteria study described above likewise noted that their ranking of drug harms correlated poorly with the UK classification system (Nutt, King and Phillips, 2010).

There is a fair counter-argument. Ministers are elected and advisers are not, and a government may legitimately weigh public concern or caution about an uncertain future alongside expert evidence. Setting aside advice is not automatically wrong. But it does mean that a drug's legal class cannot be read as a scientific verdict on how harmful it is.

Where the review lands

The review's abstract summarises three conclusions, which the later articles examine one by one.

Prohibition did not achieve its central aim of suppressing use. Cannabis is now the most widely used controlled drug in the world, with an estimated 256 million users in 2024 (UNODC, 2026), and the THC content of illicit cannabis rose steadily under prohibition (Freeman et al., 2021). Its costs fell disproportionately on minority communities through arrests and stop-and-search (ACLU, 2020; Eastwood, Shiner and Bear, 2013).

The strongest quasi-experimental studies show that decriminalisation and most forms of adult-use legalisation have not produced the sharp rises in adolescent use that opponents predicted (Hughes and Stevens, 2010; Anderson et al., 2019; Laqueur et al., 2020). The third article looks at the reform models, including their weaknesses.

Commercial legalisation carries real and measurable public-health costs among adults: rising daily use, more hospital care for cannabis use disorder, paediatric poisonings from edibles and a growing share of schizophrenia associated with heavy use (Caulkins, 2024; Myran et al., 2022; Myran et al., 2025). The fourth article examines these costs, with the caveats described above.

The evidence therefore supports neither a return to criminal prohibition nor unrestricted commercialisation. It points instead to regulatory design (potency, product form, marketing, pricing and the profit motive) as the variable that matters most, the theme of the final article.

What we know, and what we don't

Well established. Criminal prohibition, from 1925, is a short episode in a much longer history. "Legal" and "illegal" are not two policies but several distinct regimes, and evidence from one cannot simply be transferred to another (Kilmer, 2017; Hall and Lynskey, 2020). The official inquiries of the prohibition era repeatedly reached more moderate conclusions than the laws that followed (Mills, 2003). The UK moved cannabis to Class C in 2004 and back to Class B in 2009, against the advice of the ACMD.

Needs care. The strongest studies compare reforming places with similar ones that did not reform, but they assume the two would otherwise have moved in parallel, and reforming jurisdictions were not chosen at random. Pre-existing trends, reporting effects, potency drift and short follow-up weaken confident claims on both sides. Much of the best evidence comes from North America and may not transfer to the UK, which has a different illicit market, policing culture and health system.

Still open. Outcomes that depend on years of cumulative exposure may not yet be visible. Germany's final evaluation is not due until 2028, and the results of the Dutch controlled supply experiment were not available when the review was written, so conclusions about those models remain provisional.

This article is based on the critical review it expands; figures are reported as that review gives them and were not re-checked against every original study.

Next in the series: The record of prohibition tests criminal prohibition against its three aims, reducing use, deterring the young and suppressing supply, and looks at what it cost along the way, from rising potency to the uneven burden of arrests and stop-and-search.

Sources

ACLU (American Civil Liberties Union) (2020) A Tale of Two Countries: Racially Targeted Arrests in the Era of Marijuana Reform. New York: ACLU. Available at: https://www.aclu.org/publications/tale-two-countries-racially-targeted-arrests-era-marijuana-reform.

Advisory Committee on Drug Dependence (1968) Cannabis (the Wootton Report). London: HMSO.

Advisory Council on the Misuse of Drugs (2008) Cannabis: Classification and Public Health. London: Home Office.

Anderson, D.M., Hansen, B., Rees, D.I. and Sabia, J.J. (2019) 'Association of marijuana laws with teen marijuana use: new estimates from the Youth Risk Behavior Surveys', JAMA Pediatrics, 173(9), pp. 879–881. doi:10.1001/jamapediatrics.2019.1720.

Caulkins, J.P. (2024) 'Changes in self-reported cannabis use in the United States from 1979 to 2022', Addiction. doi:10.1111/add.16519.

Cohen, S. (1972) Folk Devils and Moral Panics. London: MacGibbon and Kee.

Di Forti, M., Quattrone, D., Freeman, T.P., Tripoli, G., Gayer-Anderson, C., Quigley, H. et al. (2019) 'The contribution of cannabis use to variation in the incidence of psychotic disorder across Europe (EU-GEI): a multicentre case-control study', The Lancet Psychiatry, 6(5), pp. 427–436. doi:10.1016/S2215-0366(19)30048-3.

Eastwood, N., Shiner, M. and Bear, D. (2013) The Numbers in Black and White: Ethnic Disparities in the Policing and Prosecution of Drug Offences in England and Wales. London: Release.

Freeman, T.P., Craft, S., Wilson, J., Stylianou, S., ElSohly, M., Di Forti, M. and Lynskey, M.T. (2021) 'Changes in delta-9-tetrahydrocannabinol (THC) and cannabidiol (CBD) concentrations in cannabis over time: systematic review and meta-analysis', Addiction, 116(5), pp. 1000–1010. doi:10.1111/add.15253.

Hall, W. and Lynskey, M. (2020) 'Assessing the public health impacts of legalizing recreational cannabis use: the US experience', World Psychiatry, 19(2), pp. 179–186. doi:10.1002/wps.20735.

Health Canada (2024) Canadian Cannabis Survey 2024: Summary. Ottawa: Government of Canada. Available at: https://health-infobase.canada.ca/cannabis/.

Hughes, C.E. and Stevens, A. (2010) 'What can we learn from the Portuguese decriminalization of illicit drugs?', British Journal of Criminology, 50(6), pp. 999–1022. doi:10.1093/bjc/azq038.

Kilmer, B. (2017) 'Recreational cannabis — minimizing the health risks from legalization', New England Journal of Medicine, 376(8), pp. 705–707. doi:10.1056/NEJMp1614783.

Laqueur, H., Rivera-Aguirre, A., Shev, A., Castillo-Carniglia, A., Rudolph, K.E., Ramirez, J., Martins, S.S. and Cerdá, M. (2020) 'The impact of cannabis legalization in Uruguay on adolescent cannabis use', International Journal of Drug Policy, 80, 102748. doi:10.1016/j.drugpo.2020.102748.

Mills, J.H. (2003) Cannabis Britannica: Empire, Trade, and Prohibition 1800–1928. Oxford: Oxford University Press.

Myran, D.T., Cantor, N., Finkelstein, Y., Pugliese, M., Guttmann, A., Jesseman, R. and Tanuseputro, P. (2022) 'Unintentional pediatric cannabis exposures after legalization of recreational cannabis in Canada', JAMA Network Open, 5(1), e2142521. doi:10.1001/jamanetworkopen.2021.42521.

Myran, D.T. et al. (2025) 'Changes in incident schizophrenia diagnoses associated with cannabis use disorder after cannabis legalization', JAMA Network Open, February 2025. Available at: https://pubmed.ncbi.nlm.nih.gov/39903464/.

National Academies of Sciences, Engineering, and Medicine (2017) The Health Effects of Cannabis and Cannabinoids: The Current State of Evidence and Recommendations for Research. Washington, DC: The National Academies Press. Available at: https://www.ncbi.nlm.nih.gov/books/NBK425762/.

Nutt, D.J., King, L.A. and Phillips, L.D. (2010) 'Drug harms in the UK: a multicriteria decision analysis', The Lancet, 376(9752), pp. 1558–1565. doi:10.1016/S0140-6736(10)61462-6.

Rivera-Aguirre, A., Castillo-Carniglia, A., Laqueur, H.S., Rudolph, K.E., Martins, S.S., Ramirez, J., Queirolo, R. and Cerdá, M. (2022) 'Does recreational cannabis legalization change cannabis use patterns? Evidence from secondary school students in Uruguay', Addiction, 117(11), pp. 2866–2877. doi:10.1111/add.15913.

Rolles, S. and Murkin, G. (2016) How to Regulate Cannabis: A Practical Guide. 2nd edn. Bristol: Transform Drug Policy Foundation.

Room, R., Fischer, B., Hall, W., Lenton, S. and Reuter, P. (2010) Cannabis Policy: Moving Beyond Stalemate. Oxford: Oxford University Press.

Rosenthal, F. (1971) The Herb: Hashish versus Medieval Muslim Society. Leiden: E.J. Brill.

United Nations (1961) Single Convention on Narcotic Drugs, 1961. New York, 30 March 1961. United Nations Treaty Series, vol. 520.

UNODC (2026) World Drug Report 2026: Drug use. Vienna: United Nations Office on Drugs and Crime. Available at: https://data.unodc.org/wdr2026.