Drug Decriminalization Outcomes: Portugal Model and U.S. State Comparisons

Twenty twenty-four delivered the largest single-year percentage drop in American overdose deaths across the eleven-year data series running from twenty fourteen to twenty twenty-four — a twenty-six point two percent decline, per CDC provisional data. We've done our research on this one, and the facts are genuinely surprising. That drop arrived against a backdrop of radically different state-level policies: some jurisdictions decriminalizing, others recriminalizing, all of them fighting the same fentanyl-saturated supply. Meanwhile, Portugal has been running its health-led model since two thousand and one, and cut drug-induced deaths from seventy-six to ten across its first decade — though that number has climbed again since twenty eleven, a detail advocates rarely foreground. And Oregon, the boldest American experiment, saw fewer than six hundred of seven thousand three hundred and forty-nine citation recipients call the treatment hotline over three years — a one percent uptake rate that most national coverage has buried. So here is the question I want to put to this panel: given that the arrest data are settled and the overdose data are genuinely contested, what is the load-bearing variable that actually determines whether decriminalization saves lives — and are any of the current experiments actually testing it?

The one percent hotline uptake in Oregon isn’t a behavioral failure—it’s a structural signal. Portugal’s Dissuasion Commissions weren’t just health referrals; they were mandatory administrative encounters housed in the Ministry of Health, with the full weight of the state behind them. In the U.S., we’ve treated the health handoff as an opt-in service, not a load-bearing institution. The question isn’t whether people want treatment—it’s whether the system is designed so that encountering the state requires engaging with health services. That’s not a detail; it’s the entire theory of change. Portugal’s model worked because the legal reform was the delivery mechanism for a health-system intervention. We’ve replicated the delivery mechanism and omitted the intervention. The debate isn’t about decriminalization—it’s about whether we’re willing to build the institutional architecture that makes it meaningful.

The U.S. overdose decline of twenty-six point two percent between twenty twenty-three and twenty twenty-four arrived through naloxone saturation, expanded medications for opioid use disorder, and fentanyl supply shifts—none of which required decriminalization statutes. That suggests harm-reduction gains can emerge from entirely separate administrative channels when the health system itself is the intervention point. Portugal’s post-twenty-eleven rebound under documented budget and staffing pressures shows the same point in reverse: once reinvestment stops, the model erodes even with the legal framework intact. The experiments are therefore testing mismatched causal bundles, and five-year U.S. data cannot isolate whether the legal change itself matters once the health-system handoff capacity is held constant.

We are evaluating a policy sold partly as racial justice reform using completely colorblind metrics. State-level data on whether racial arrest disparities actually narrow post-reform are tracked inconsistently and largely absent from the quantitative record. I think that is not an oversight; it is a category error. When you remove the criminal penalty but leave the material conditions untouched, you are testing a legal abstraction against a structural reality. Ibn Khaldun argued centuries ago that social behavior tracks material cohesion and institutional trust, not statutory categories. Amartya Sen makes the same point in modern development terms: addiction sits downstream of capability deprivation and fractured social infrastructure. Decriminalization changes the administrative label; it does not rebuild the relational architecture that actually determines whether someone answers a hotline or walks into a clinic. Until we measure who the system reaches and who it silently drops, I think we are mistaking a paperwork reform for a social intervention.

The overlooked mechanism here is compulsion through contact, not coercion through punishment. Portugal’s commissions operate inside a medical bureaucracy that already held treatment capacity, so a person’s case entered a system where health actors had both mandate and resources to reach back. Oregon flipped the statute but lacked that second circuit. A citation points to a hotline, but without funded outreach, staffed clinics, or data linkage between justice and health records, the signal dies in transit. The one percent uptake isn’t evidence that people won’t seek help; it’s proof that the state isn’t organizationally equipped to follow up. Until that connective tissue exists, decriminalization can keep reducing arrests but will reliably yield null results on health outcomes.

Here is the assumption the panel is now sharing that I want to put pressure on. We have collectively accepted that Portugal is the proof of concept and the U.S. is the failed replication. But EMCDDA data and Goulão's own statements document that after twenty eleven — when austerity hit and reinvestment stopped — Portugal's overdose numbers climbed back up, addiction programs lost staff, and budget constraints eroded capacity. Advocacy coverage froze Portugal at its best decade. What that means for our comparison is uncomfortable: we may be holding U.S. experiments to a standard that Portugal itself no longer meets. The honest framing is not Portugal versus Oregon. It is whether any jurisdiction, including Portugal, can sustain a health-system handoff without continuous political will and continuous reinvestment — and the answer, on the full twenty-five year arc, is genuinely uncertain.

The one percent uptake figure isn’t just a systems failure—it’s a measurement failure of the entire reform’s racial equity premise. Oregon’s data don’t break down who called the hotline by race, yet the policy was sold as a racial justice intervention. That silence isn’t incidental; it’s structural. When you remove criminal penalties but don’t track whether the health system actually reaches the populations most harmed by criminalization, you’re not testing decriminalization’s equity effects—you’re assuming they exist. The minimum viable racial outcome tracking system would require every citation to record race, every hotline call to log demographics, and every treatment referral to follow through with outcome data. Without that, we’re debating a policy’s impact on racial justice while having no idea whether it’s improving it, worsening it, or leaving it untouched. That’s not a data gap; it’s a design flaw in how we evaluate reform.

The US overdose drop of twenty-six point two percent between twenty twenty-three and twenty twenty-four came through naloxone distribution and medications for opioid use disorder expansion inside existing clinical and pharmacy channels, not through any new administrative citation system. That path bypassed the need for a single mandatory handoff point and still moved the aggregate numbers. Portugal’s model tied the legal change to one centralized health bureaucracy that already held both the mandate and the funding line; the US achieved partial movement by saturating multiple existing contact points instead. The structural question this raises is whether the health infrastructure must be assembled first and then paired with legal reform, or whether legal reform can be the forcing function that builds the infrastructure. Five-year state data cannot yet distinguish those sequences because no jurisdiction has run the capacity build without the statute change.

The overdose debate is stuck because difference-in-differences models, which isolate policy effects by comparing trends across jurisdictions, break down when the underlying drug supply is non-stationary. The NYU and CDC collaboration finds null effects; the Spencer preprint, which builds a statistical counterfactual to isolate Oregon’s trajectory, finds a twenty-three percent increase. That gap is not ideological. It is mathematical. When fentanyl penetration shifts monthly, before-and-after comparisons collapse unless you bake a supply-side toxicity covariate into the model. We are asking a legal variable to explain a toxicology curve. The same misalignment fractures the treatment argument. Sarah Clune’s research shows coercive diversion actually moves high-acuity, low-engagement cohorts, while voluntary architectures only convert people already leaning toward help. We have collapsed two distinct clinical populations into one binary debate. Until we stratify outcomes by baseline readiness and control for street-level toxicity, we are measuring statistical noise and calling it statecraft.

The deeper constraint isn’t just budget; it’s institutional architecture. Portugal’s commissions function because they sit inside a universal health system that can absorb and treat anyone who walks in. The United States has no equivalent intake layer. Even if funding were steady, a fragmented insurance system cannot guarantee treatment entry when addiction services depend on coverage status or county contracts. That’s why some analysts, from the Heritage group to India’s NDDTC, argue the model may be structurally incompatible. The real experiment isn’t whether decriminalization lowers overdoses—it’s whether a market-based health system can substitute for universal infrastructure in delivering population-level addiction care.

Here is what the twenty-six point two percent decline actually forces us to confront methodologically. The brief is clear that fentanyl market shifts, naloxone saturation, and MOUD expansion are plausible contributors to that drop — but it is equally clear that causal attribution in a non-stationary drug supply environment is genuinely hard. What that means is we now have two parallel puzzles with inverted uncertainty: the decriminalization experiments may be doing more than the null findings suggest if fentanyl penetration is suppressing the signal, and the twenty twenty-three to twenty twenty-four decline may have less to do with any specific intervention than with a supply-side market shift that nobody designed. We are possibly in a moment where the evidence on both sides is being inflated by the same confounder pointing in opposite directions — and the field has not reckoned with that symmetry at all.

The structural lesson from Portugal isn’t that you need a single bureaucracy—it’s that you need an administrative signal. Those 18 regional commissions weren’t just referral points; they were the state’s way of saying, “This is a health matter, not a criminal one.” The U.S. has no equivalent signal. When Oregon issues a citation, it’s still processed through the same courts that handle DUIs and petty theft. The paperwork may say “health assessment,” but the institutional grammar remains punitive. That mismatch isn’t a detail—it’s why the hotline sits unused. The system doesn’t just lack capacity; it lacks a clear institutional voice telling people, “You’re here because we care about your health.” Without that signal, the legal reform is just noise.

The twenty-six point two percent national drop between twenty twenty-three and twenty twenty-four arrived through channels that never required a new administrative signal at all. Naloxone and medications for opioid use disorder expanded inside existing pharmacy and clinic workflows, moving aggregate deaths without touching possession statutes. Portugal’s own rebound after twenty eleven under austerity shows that even a well-housed commission system loses ground once sustained funding and staffing erode. This suggests the active ingredient may be durable resource commitment rather than the ministry label itself. Five-year U.S. experiments cannot yet separate those effects because no jurisdiction has held the funding line constant long enough to observe whether the legal change adds anything once capacity is already present.

We are treating statutory reclassification as if it generates social trust, when historically it works the other way around. Ibn Khaldun observed that coercive systems only relax when material cohesion and institutional reciprocity are already secure. Amartya Sen frames it as capability: people engage with services when the state has already proven it can deliver housing, income stability, and clinical care. Portugal’s commissions succeeded because they plugged into a pre-existing civic covenant. The U.S. experiments inverted that sequence, assuming a citation could manufacture engagement in a fragmented, low-trust environment. That is why transferability remains genuinely contested. Public health advocates argue the health-first logic travels anywhere, while market-oriented analysts and clinical observers in India and Japan counter that without universal healthcare and baseline social homogeneity, the mechanism has nothing to grip. We are not debating drug policy. We are debating whether a state can legislate relational authority into existence.

The misread in both Portugal and Oregon is treating drug law as the intervention when the real causal agent is how the state manages human contact at moments of crisis. Portugal built a routine, face‑to‑face triage system housed in public health; Oregon substituted a paper citation and hotline. The difference is not moral clarity, it’s operational density. When clinical labor is the scarce resource, any model that requires sustained human engagement becomes brittle. The twenty‑five‑year question isn’t whether decriminalization works, it’s whether any government can keep financing the labor‑intensive relationship infrastructure that makes it work once the political spotlight moves on.

The racial equity measurement gap is the sharpest version of this problem. Decriminalization was sold substantially as a racial justice intervention — the Drug Policy Alliance and others made that argument explicitly. But the quantitative tracking infrastructure to test that claim barely exists. Oregon's citation data do not consistently break down by race. There is no standardized national framework requiring racial outcome tracking post-reform. That is not a minor data gap — it is a design flaw in the evaluation architecture itself. A policy justified partly on racial equity grounds, evaluated almost entirely on aggregate health metrics that cannot answer the racial equity question, is not being tested. It is being assumed. And the assumption may be wrong in either direction.

The hotline uptake figure isn’t just low—it’s a category error in how we measure success. We’re treating a phone call as the outcome when the real variable is whether the system can absorb the person who calls. Portugal’s commissions weren’t successful because they got high engagement numbers; they worked because they were embedded in a universal health system with guaranteed capacity. In the U.S., even if every citation recipient called the hotline tomorrow, most would hit insurance barriers, waitlists, or county funding gaps. The metric we should be tracking isn’t call volume—it’s what percentage of referred individuals actually receive same-week treatment. That number would reveal whether the infrastructure exists to make the legal reform meaningful. Right now, we’re grading the policy on intent, not function.

The fentanyl transition since twenty thirteen renders any before-after overdose comparison across twenty twenty to twenty twenty-four uninterpretable without explicit supply covariates. The NYU and CDC analysis and the synthetic-control preprint disagree on Oregon's trajectory because one embeds a monthly toxicity measure and the other does not; the gap is not values but model specification. That means the health-infrastructure question the panel is now debating sits on top of an outcome variable that is still mostly noise. Until the field isolates policy effects from the dominant confounder, claims that the U.S. lacks Portugal's handoff capacity remain structurally untestable rather than merely underfunded.

We are mistaking administrative design for social fabric. Portugal’s commissions rely on motivational interviewing and low-threshold services precisely because they operate inside a polity where state referrals still carry relational legitimacy. Sarah Clune’s finding that arrest sometimes catalyzes treatment entry is not a vindication of coercion; it is a structural signal. It shows that in fragmented, low-trust environments, the criminal system is often the only institution with consistent reach into marginalized lives. That is why analysts from India’s NDDTC to Japan’s JSAM treat Portugal’s model as structurally difficult to transfer. The barrier is not just insurance fragmentation or EU structural funding. It is whether a society can generate the civic reciprocity that makes a health referral function as care rather than surveillance. Until we measure relational trust alongside street-level drug toxicity, we will keep confusing legal architecture for governing capacity.

The most revealing comparison may not be Portugal versus Oregon, but Portugal versus California. One built a new public‑health apparatus to catch people as criminal enforcement receded; the other, through Proposition forty‑seven, linked sentencing reform to an explicit reinvestment stream for treatment and social programs. That pairing of legal relief with fiscal redirection is the only U.S. precedent that even approximates Portugal’s integrated logic. The unresolved question is whether that linkage can scale nationally without universal healthcare beneath it. If not, decriminalization remains fiscally hollow—it frees people from punishment but not from the structural conditions that made punishment their first point of contact.

California after Proposition forty-seven is genuinely the closest the U.S. has come to coupling legal relief with health investment — the brief confirms reduced prison populations, drug arrests, and fiscal savings redirected toward treatment. But the brief also confirms that analysts attribute the twenty twenty-three to twenty twenty-four overdose decline to naloxone, MOUD expansion, and fentanyl market shifts — not to any decriminalization statute. That gap is the tell. If California's reinvestment approach were sufficient, you would expect a cleaner signal. What you actually get is outcomes improving through harm reduction infrastructure that operates largely independently of the legal reform. Which means ChatGPT's California framing is right directionally but may be overstating the causal role of the fiscal linkage relative to the clinical infrastructure underneath it.

The structural blind spot isn’t just that U.S. reforms lack Portugal’s health infrastructure—it’s that they lack Portugal’s institutional grammar of reciprocity. In Portugal, the Dissuasion Commissions operate on an unspoken contract: the state offers care, and citizens engage because the system has historically delivered on housing, employment, and healthcare. That contract is visible in the commission’s design—no police presence, health professionals leading, and sanctions that feel administrative rather than punitive. The U.S. experiments inverted this logic. They removed the stick but didn’t replace it with a credible offer of care. Oregon’s hotline isn’t just a low-uptake service; it’s a symptom of a deeper mismatch. When the only institutional voice people hear is a citation from a system they associate with punishment, the reform’s signal gets lost in the noise of historical distrust. The real question isn’t whether decriminalization works—it’s whether any reform can work in a system where the state’s primary relationship with marginalized populations has been through control, not care. That’s not a funding gap. It’s a legitimacy gap. And no block grant can fix it.

The twenty twenty three to twenty twenty four decline shows measurable health improvement inside an incomplete architecture, driven by supply control and expanded treatment without any new relational contract. That undercuts the claim that civic reciprocity must precede gains. At the same time, the brief notes that outcomes beyond five years remain unobserved in any U.S. jurisdiction, so the durability of gains achieved through fragmented delivery is precisely what we cannot yet test. The Portugal model may still represent one stable equilibrium; it is not yet proven to be the only viable one.

We are treating a statute as if it can manufacture administrative capacity. Classical statecraft treats law as the outer shell; the machinery underneath does the actual governing. Oregon’s data show fewer than six hundred of seven thousand citation recipients called the health line over three years. That is not a verdict on individual motivation. It is what happens when you remove a penalty but leave no structured pathway into care. Portugal’s model works because the referral lands inside a universal health network where a clinician is actually waiting. Oregon optimized for low friction and got zero conversion because the bridge was never poured. Meanwhile, experts attribute the twenty twenty three to twenty twenty four national mortality decline to supply control, expanded treatment, and fentanyl market saturation—factors that operate entirely outside statutory reclassification. The country is already shifting outcomes through distributed clinical and market interventions while the policy debate remains locked on a legal switch that was never designed to function alone.

The evidence now points to a deeper systems‑engineering question: what actually carries a person from contact to care? Portugal built redundancy into that handoff — legal referral, clinical intake, and social support all nested in one administrative chain. The U.S. substitutes optional links: a citation here, a hotline there, then county‑level eligibility checks. Each step loses an order of magnitude of people before treatment even begins. The striking discovery isn’t ideological at all; it’s throughput. Without mandated coordination between justice, health, and social services, a decriminalization law becomes a zero‑pressure pipe — open at both ends, delivering almost nothing.

The throughput framing is right but it stops one level short. The question underneath is not just whether the pipe delivers — it is what the pipe is delivering into. The CDC's cascade-of-care analysis found that only twenty-five percent of Americans who needed medication treatment for opioid use disorder actually received it, and that gap exists entirely independently of decriminalization. So even a perfectly designed handoff mechanism — mandatory, clinician-staffed, same-week intake — would be feeding into a treatment system already operating at roughly one-quarter of required capacity. That is not a design flaw in the legal reform. That is a structural ceiling on what any legal reform can achieve, regardless of how well the referral architecture is built.

The Portugal model didn’t just remove criminal penalties—it rewired the state’s institutional grammar around addiction. The key wasn’t the absence of punishment, but the presence of a health ministry that could absorb the caseload without discretionary gatekeeping. In the U.S., even when states reclassify possession, the administrative housing stays in justice systems that lack the mandate—or the budget—to guarantee treatment. That’s why Oregon’s hotline sits empty: it’s a referral into a system that still treats addiction as a discretionary service, not an entitlement. The real structural question isn’t whether decriminalization works, but whether any jurisdiction can sustain the labor-intensive relational work once the political spotlight fades. Portugal’s commissions succeeded because they were embedded in a welfare state where health access was already normalized. In the U.S., we’re asking a legal reform to do the work of a social contract that doesn’t exist. That’s not a policy failure—it’s a category mismatch.

The distinction that matters is not referral design but citizenship status of addiction treatment. Portugal embedded its commissions inside a universal entitlement that already treated health access as non-discretionary. The U.S. treatment system operates at a 25.1 percent MOUD receipt rate among those who need it, which means any handoff—mandatory or optional—still routes people into a rationed good whose baseline scarcity is set by insurance rules, not by statute. The policy debate therefore conflates the architecture of contact with the underlying definition of who holds a claim on care. Until that definition changes, every decriminalization statute tests a different underlying intervention than the one Portugal ran.

Classical statecraft treats law as the outer shell of governance, not the engine. When you lift a legal mechanism out of its native fiscal and clinical ecosystem, you do not get a weaker version of the original policy. You get a different policy entirely. The United States did not test decriminalization; it tested administrative decoupling. Oregon routed citations through a justice-adjacent pipeline while Portugal routed them through health ministry intake. That is not an implementation detail. It is a structural divergence that severs the causal chain at the point of referral. The Transform Drug Policy Foundation has argued for years that the legal switch is only a delivery vehicle for a comprehensive support system. Treating that coupling as optional guarantees the statute functions as a liability shield for the state rather than a care pathway for the user. We are grading two different experiments on the same curve.

The sharpest insight from this conversation is that no jurisdiction has yet tested decriminalization-plus-infrastructure as a unified intervention. Every U.S. experiment has tested one half of Portugal's theory and then attributed the outcome to the whole. The concrete takeaway: the one percent hotline uptake in Oregon is not a behavioral finding about people who use drugs. It is a structural finding about a state that removed a penalty without building a pathway. Five years of data tell us decriminalization reliably reduces criminal justice contact and probably cannot, by itself, reliably reduce overdose deaths. That is an honest answer, not a failure of ambition. Thank you for listening. As it happened; as it is.