No, a New Study Does Not ‘Lay To Rest’ the Debate Over Drug ‘Legalization’

Oregon and Washington both effectively decriminalized drug possession in early 2021 in two short-lived experiments. During these periods, overdose deaths rose sharply in both states. Nationally, they followed a broadly similar upward path during the same period.

Naturally, policy analysts seized on the experiments to tease out the effects of decriminalization from the factors driving the national trends.

While it’s clear that the decline in overdose deaths the proponents of decriminalization were hoping for did not occur, the evidence is mixed about whether decriminalization drove an increase in overdose deaths once you control for the fentanyl epidemic and COVID-19’s impact on social isolation and street encampments.

In a piece at City Journal titled “No, Seriously, Decriminalizing Drugs Kills People,” Manhattan Institute Senior Fellow Charles Fain Lehman argues that the debate is settled. A new study, he writes, shows “conclusively” that decriminalizing drugs caused overdose deaths “to explode.”

Lehman’s analysis is riddled with errors and misconceptions, starting with his claim that this study should “lay to rest the debate over legalization.” (Emphasis mine.) What was tried in Oregon and Washington has nothing to do with legalization. Decriminalization at the state level left all federal laws in place, and all laws against distribution. It only reduced penalties for possession of small amounts. Under legalization, drug users would be able to purchase their narcotics from established sources, mitigating the problem of black market drugs tainted with fentanyl. A grim reminder of the recurrent pattern is the 2019 vaping-injury outbreak, which hospitalized roughly 2,800 people and killed 68. The culprit was neither nicotine nor THC but vitamin E acetate, a cheap thickener that black market sellers cut into counterfeit cannabis cartridges; it showed up in victims’ lungs and almost nowhere else. Regulated, tested products weren’t the problem. The illicit supply was, exactly as it is with fentanyl.

Far from laying anything “to rest,” the study’s conclusions were mixed and nuanced, and other studies reached different conclusions. Lehman’s article is a case of a partisan selecting a single non-peer-reviewed working paper with equivocal findings from an even more equivocal literature and slamming the door on the debate.

The study in question is a National Bureau of Economic Research working paper, the first page of which carries the Bureau’s own disclaimer that it “ha[s] not been peer-reviewed.” Its central estimate rests on a synthetic control, meaning a weighted blend of other states stitched together to impersonate the Oregon and Washington that never happened. The authors, unlike Lehman, are scrupulous: They call their result a “reduced-form effect” of the “policy regime,” decline to identify any mechanism, list COVID and the timing of fentanyl’s arrival as live threats to their estimate, and explicitly warn against reading too much into short post-policy windows. Lehman has stripped the paper of qualifications and treated it like holy writ.

If decriminalization is what drove Oregon and Washington’s overdose deaths upward, then states that decriminalized nothing should not show the same surge. They do. By the Centers for Disease Control and Prevention’s own overdose counts, the per-capita increases from 2019 to 2023 in Alaska and West Virginia were as large as or larger than those in Oregon and Washington.

What happened in Alaska is particularly telling: The state saw one of the sharpest per-capita increases in the nation—and the biggest single-year jump of any state in 2023—yet it decriminalized nothing. It was simply the last place fentanyl deaths surged, with the sharp increase seen in 2020 after sweeping the country from east to west from 2013 to 2018. Washington and Oregon were a few months earlier in late 2019.

Ignoring geography, Oregon and Washington were among the top 10 percent of states in overdose death increases during the decriminalization period. That’s on the border of statistical significance. Depending on how you analyze the data, you could call it moderate evidence that decriminalization increased deaths or dismiss it as plausibly unrelated to decriminalization. (It is suggestive, not conclusive, evidence that decriminalization did not save lives.) That’s precisely why different studies are showing different results, and no credible researchers are claiming to “lay to rest the debate.”

The entire result hinges on rejecting one modeling choice from a prior paper by drug policy researcher Michael Zoorob and colleagues, who found that once you control for fentanyl’s spread, decriminalization’s apparent effect disappears. How fragile is that hinge? In the new paper’s Table 1, when the authors apply Zoorob’s fentanyl control to Oregon, the estimated effect flips to a slight negative and loses all significance. The four-figure body count Lehman is selling exists only if you accept the authors’ preferred way of handling fentanyl and reject the other side’s approach. That is not a debate laid to rest. That is a debate resting entirely on a single contested knob.

Suppose we grant the deaths anyway. Who died? Overwhelmingly, longtime users. A 2023 survey of nearly 500 people who use drugs across eight Oregon counties found that just 1.5 percent had started after Measure 110 passed. The dead were not fresh recruits lured into addiction by the promise of a hundred-dollar ticket; they were people who had been using for years and decades, who met a newly lethal supply. Oregon’s fentanyl-involved overdose deaths nearly quadrupled between 2020 and 2022, and fentanyl was present in about two-thirds of the state’s overdose deaths. This is a story about potency—the same number of people using a drug that now kills them far more often—not about a policy manufacturing new addicts. Lehman’s morality tale requires new victims created by the law. The field data say they barely exist.

Since decriminalization did not create new addicts, the most plausible channel by which it could have cost lives is incarceration itself, meaning fewer users behind bars where drugs are harder to reach. But delaying an addict’s death by locking him up until he returns to the same lethal street supply carries a very different moral weight than saving a life by curing a disease or pulling a child from a burning building.

Lehman ignores the other side of the ledger. Washington’s drug possession arrests fell by 91 percent—from 9.2 to 0.8 per 100,000 residents per month—sparing on the order of 20,000 arrests, atop hundreds of thousands of prior convictions that the state moved to vacate after its Supreme Court voided the possession statute in State v. Blake. Those arrests fell most heavily on black and Native American residents, who were also, cruelly, the populations dying at the highest rates, which is precisely why racial justice organizations backed the policy.

A serious accounting must weigh the arrests, the records, the lost jobs, and the vacated convictions. Some of the arrested were innocent. Some arrests ruined lives. You can’t pile corpses on only one side of the scale and leave the other side empty.

The harm at the center of this study is dead drug users, most of whom would presumably have supported decriminalization. Citing “over 1,000 excess deaths” as the decisive strike against decriminalization only works if you have already assumed the paternalist’s premise that preventing people from harming themselves justifies coercing everyone else.

When Lehman folds in a “surge in crime” for good measure, note that the mortality study says nothing about crime, and that the arrest study he implicitly relies on found no statistically significant increase in violent or property offenses attributable to Measure 110.

Measure 110 legalized nothing. It decriminalized the act of possession while leaving the supply illegal, unregulated, and (the part that kills people) of unknown strength.

A smuggler faces the same penalty per shipment whether it holds heroin or something 50 times stronger, so the market evolves relentlessly toward the most potency per hidden ounce. That is the iron law of prohibition. A single 28-pound fentanyl seizure carries the dose equivalent of about 1,400 pounds of heroin. Fentanyl did not win because users demanded it. It won because prohibition rewards whatever is easiest to conceal.

The fentanyl death toll can’t be attributed to decriminalization. It is a product of the drug war. You cannot regulate what you outlaw. What kills is not the opioid molecule; it is not knowing whether the powder in front of you is a dose or a coffin. Supply a known quantity and the dying largely stops: Switzerland has dispensed pharmaceutical-grade heroin under medical supervision for three decades, with almost no overdose deaths among its patients and its drug fatalities cut by half.

Overdose remains rare among ordinary pain patients taking a labeled prescription. A regime that furnished a regulated, measured product would put that proposition to the test. Measure 110 did the reverse. It lifted the penalty for holding the poison while doing nothing about the poison. If it failed, it’s an indictment of half measures, not of legalization.

None of this means decriminalization worked, nor that the new paper is worthless. Extending the data and stress-testing the fentanyl measures is real work, and the raw post-2021 rise is undeniable in the numbers. But the reading is modest: a contested, non-peer-reviewed, two-sigma association, matched by states that decriminalized nothing, riding on a single modeling choice, among a population of established users killed by a fentanyl wave that respected no border. That is a finding worth arguing about. It is not a verdict.