Deadly Curve Tracked Fentanyl Supply, Not Statutes

When overdose curves move sharply, causation arguments harden just as fast—yet on Washington and Oregon’s decriminalization experiments, the strongest evidence now points to a simpler driver of mortality: the fentanyl wave, not the legal status of possession.

At a Glance

Multiple peer‑reviewed analyses find no detectable rise in fatal overdoses attributable to decriminalization in Oregon and Washington when the fentanyl surge is accounted for.
A high‑profile NBER working paper reports excess deaths after decriminalization relative to synthetic controls, but its attribution sits in tension with fentanyl’s rapid diffusion across both states.
Mechanism matters: decriminalization slashed possession arrests; fentanyl’s potency, ubiquity, and speed of market penetration drove mortality dynamics.
Policy timing collided with strained treatment capacity and COVID‑era disruptions, complicating any clean read on cause and effect.

What the best evidence does—and does not—show

Two large signals structure this debate. First, Oregon and Washington changed the legal consequences for drug possession in early 2021—Oregon through Measure 110’s statewide decriminalization and Washington through a short-lived de facto decriminalization after a state supreme court ruling, later modified by statute. Second, fatal overdoses climbed steeply across much of the United States in the same window as illicit fentanyl displaced heroin and contaminated non-opioid street supplies. Sorting attribution requires a credible counterfactual: what would deaths have looked like without the legal shifts, given fentanyl’s spread?

On that question, a JAMA Psychiatry cohort study using a synthetic-control design—the same general class of method used by critics—found no evidence that decriminalization increased fatal overdoses in either state during the first year after policy change. A follow-on analysis focused on Oregon’s drug market concluded that once the rapid introduction and diffusion of fentanyl are modeled, the apparent association between Measure 110 and overdose risk dissipates; the estimated effect flips direction and loses statistical significance. These are not advocacy briefs but peer‑reviewed efforts to isolate policy effects amid a moving market.

The counterclaim: excess deaths after decriminalization

An influential NBER working paper argues the opposite. Using synthetic-control estimates, the authors report that both Oregon’s and Washington’s regimes reduced arrests yet were followed by sustained increases in overdose mortality relative to their constructed counterfactuals—on the order of 1,186 excess deaths in Oregon and 1,895 in Washington from 2021 through 2023. This is a substantial signal and, on its face, a damning one. But method and mechanism must align. If fentanyl diffusion is uneven across time and geography, and if the model does not fully absorb that heterogeneity, then a policy dummy can soak up fentanyl’s momentum, creating the appearance of a causal policy effect where the causal agent is the drug supply itself.

That is precisely why subsequent analyses concentrate on fentanyl as a confounder with its own independent dynamics—supply-side shocks, wholesale price changes, and contamination of counterfeit pills. Oregon-focused studies that explicitly model the timing of fentanyl’s emergence find the policy signal vanishes once that market shift is in the equation. Put differently: the same mortality rise that the NBER team attributes to legal change is explained parsimoniously by a rapidly evolving illicit supply that would have increased deaths under either legal regime.

Mechanism: why fentanyl dominates the mortality curve

Decriminalization changes incentives at the margin: fewer arrests for possession, civil citations instead of charges, and intended redirection to services. It does not, by itself, alter the pharmacology of what people consume or the volatility of the supply chain. Fentanyl’s risk profile is structural—microgram‑level potency, short half-life prompting frequent dosing, and unpredictable adulteration that defeats user tolerance calibration. When fentanyl arrives quickly, fatalities rise quickly, particularly in places with thin treatment capacity and limited access to evidence-based interventions like methadone and buprenorphine. Oregon’s treatment infrastructure ranked poorly before Measure 110 and was slow to scale; Washington’s diversion capacity has also lagged needs in urban hotspots. Against that backdrop, it would be surprising if arrest policy, rather than supply potency, led the mortality dance.

The arrest data from the NBER study are informative on this point: possession arrests dropped sharply after decriminalization, as designed, yet deaths rose; this shows implementation fidelity on enforcement, not causality for mortality. To argue the latter, one needs a channel: for example, if decriminalization increases initiation, frequency, or dangerous modes of consumption in a way that a fentanyl-aware model cannot explain. The fentanyl-focused analyses do not find such a channel; instead, they show the overdose curve tracking supply trends once those are measured in the model.

Why high-quality studies disagree

Disagreement here is not academic theater; it is about counterfactual construction under simultaneous shocks. Synthetic control methods are powerful when treated units are few and confounders stable. But they can misattribute effects if donor pools experience different fentanyl timings or if national shocks (COVID-era health care access, homelessness spikes, policing disruptions) intersect unevenly across states. Oregon’s and Washington’s policy windows coincide with fentanyl’s steep ascent westward, which complicates “clean” inference in any design that does not front‑load supply metrics.

Peer‑reviewed work that embeds fentanyl diffusion directly finds no policy‑driven mortality increase in Oregon and Washington during the examined period. The NBER working paper, notable and carefully executed, nevertheless leaves open the confounding channel because its core result can also be read as a fentanyl-timing artifact. When two methods of the same family yield divergent conclusions, domain mechanism adjudicates: the pharmacology and supply chain of fentanyl offer a more credible explanatory engine than possession law per se.

Context that shapes outcomes but doesn’t change the causal verdict

It is fair to say Oregon’s decriminalization rollout was troubled on public order and service delivery. Police departments were strained; referrals to treatment through citation hotlines underperformed; visible open-air use and homelessness made the policy politically radioactive. Washington’s brief de facto decriminalization and subsequent misdemeanor framework produced its own whiplash. These conditions matter for public safety and community trust. They also make clean evaluation harder because multiple systems were moving at once—law enforcement staffing, court backlogs, housing scarcity, and the overdose care continuum. None of that rescues a causal claim about mortality if the death curve is already explained by fentanyl’s diffusion.

The right lesson is not that decriminalization fixes overdose risk; it plainly does not. Nor is it that recriminalization will bend the mortality curve; jurisdictions that retained criminal penalties have seen the same fentanyl-driven surge. The lesson is narrower and more actionable: mortality tracks supply potency and treatment access far more than possession statutes. Any policy—criminal or civil—that fails to expand rapid, low‑threshold access to effective medications for opioid use disorder, distribute naloxone widely, and stabilize housing for people with substance use disorders will struggle against fentanyl’s baseline risk.

So, did drug decriminalization increase deaths in Oregon and Washington? At @reason, Aaron Brown responds to @CharlesFLehman of @ManhattanInst and @CityJournal https://t.co/JhC3JfSNx8

— Nick Gillespie (@nickgillespie) July 24, 2026

What this means going forward

For decision‑makers, three priorities follow from the evidence. First, evaluate overdose policy through the lens of supply potency and treatment penetration, not through arrest statistics. A state can cut possession arrests to near zero and still see deaths rise if fentanyl saturates the market. Conversely, a state can keep penalties and still lose ground absent robust MOUD access. Second, instrument policy with measurable pathways—link decriminalization or diversion to same‑day buprenorphine starts, hospital‑initiated methadone, assertive outreach, and post‑overdose engagement—so that the mechanism for harm reduction is visible in the data. Third, model fentanyl explicitly in any impact study; treating it as background noise guarantees misattribution.

The narrow claim at issue—did decriminalization itself cause more deaths in Washington and Oregon?—does not survive the strongest, fentanyl-aware analyses. Overdose mortality increased, but the weight of peer‑reviewed evidence attributes that rise to the fentanyl wave rather than to the legal status of possession. Policymakers should calibrate accordingly: center interventions on the drug supply’s lethality and the treatment system’s capacity, not on symbolic shifts between misdemeanor and citation.