Seattle’s jail diversion program resulted in more dead participants than employed ones: report

For more than a decade, governments across the United States have been told that Seattle’s Law Enforcement Assisted Diversion (LEAD) program produced remarkable results: participants were 58 percent less likely to be arrested, felony charges fell 39 percent, and employment and housing outcomes dramatically improved.

A new national review says those claims bear little resemblance to what actually happened.

The August 2026 National Retraction Brief, authored by Washington Policy Center Public Safety Solutions Policy Director Eric Zimmerman, alleges that LEAD’s foundational research was repeatedly misrepresented as the Seattle-born program expanded across the country, helping influence legislation, government grants, municipal budgets and other public and private funding in more than 20 states.

Among 176 LEAD participants examined for employment outcomes, 91 percent remained unemployed on average. The report contrasts those results with 2019 Bureau of Labor Statistics (BLS) research tracking men after release from at least six months of incarceration. According to the BLS comparison cited in the report, 33.9 percent were employed within one week of release, and 50.3 percent were employed by week ten.

According to the report, three times as many LEAD participants died as the estimated number who gained employment during the study period.

LEAD research has been promoted with claims that participants’ “odds of being housed increased 89%” or that participants were “89% more likely to obtain permanent housing during the follow-up.” Of 146 participants who were unhoused when they entered LEAD, an estimated average of 19 were permanently housed in a given month during the 18-month follow-up period, equivalent to approximately 13 percent of those entering without permanent housing.

Shelter placement performed somewhat better, with an estimated average of 31 participants appearing to be sheltered in a given month. Even combining shelter and permanent housing, however, the report calculates that LEAD achieved only a 35 percent success rate in connecting initially unhoused participants with either shelter or permanent housing.

There was another major limitation: that portion of the LEAD study did not use a control group. The report therefore compares the results with federal housing data from King County over roughly the same period. The county’s sheltered-homelessness rate declined slightly, from 70.5 percent in 2012 to 69.4 percent in 2014.

Zimmerman concludes that connecting participants to shelter and housing may nevertheless represent LEAD’s most plausible treatment effect beyond reducing contact with the criminal justice system, arguing the program could potentially be more accurately understood as a point-of-arrest intake method for “Housing First” programs.

Even on permanent housing, the number of participants who died was approximately half the estimated number entering permanent housing during the study period.

Despite the abysmal results of the program, LEAD was being promoted locally and nationally using dramatically different numbers, including that participants were “58% less likely to be arrested.” Zimmerman’s report argues that researchers reported an odds ratio that was subsequently promoted as a 58 percent reduction in arrests, crime, or recidivism.

The report identifies similar problems with claims that LEAD reduced felony charges by 39 percent and improved employment outcomes by 46 percent, arguing that promoters repeatedly confused mathematical concepts including odds, probability, and absolute versus relative change.

For arrests, the report identifies an unadjusted 22-percentage-point difference between LEAD participants and the control group after entry. Zimmerman cautions that even that difference cannot reliably be attributed to LEAD because the groups experienced different recruitment practices, policing patterns, timelines and geographic conditions. LEAD participants also appear to have benefited from reductions or dismissals of charges that affected subsequent warrant arrests in ways the control group did not.

According to the report, the King County Prosecuting Attorney’s Office was an “unblinded operational partner,” and prosecutors could take LEAD participation and progress into account when determining whether and when to file felony charges.

The report argues that the 11-percentage-point unadjusted difference in felony rates cannot be cleanly attributed to LEAD changing participants’ criminal behavior because the justice system itself treated LEAD participants differently.

Despite the problems, LEAD expanded far beyond Washington after its Seattle launch. Santa Fe, New Mexico, became one of the earliest jurisdictions to replicate the model, followed by Albany, New York, and Fayetteville, North Carolina. LEAD programs or pilots were subsequently established in jurisdictions including Los Angeles and San Francisco, California; Baltimore, Maryland; Washington, DC; Portland, Oregon; Huntington and Charleston, West Virginia; and communities in New Jersey and Maine.

North Carolina has continued expanding the model, announcing $1.5 million in funding in March 2026 for LEAD programs involving Fayetteville, Albemarle, and Greensboro, as well as the counties of Gaston, New Hanover, Robeson, and Harnett.

Statistics originating with Seattle’s program did not remain local as LEAD developed into a nationally replicated criminal justice model. Versions of the disputed claims appeared in materials from police departments, universities, prosecutors, state governments, medical and criminal justice publications and organizations promoting LEAD.

The report documents versions of the “58 percent” claim appearing in sources including the Seattle Police Department, University of Washington, LEAD Seattle/King County, Duke Law School, Michigan State Police, New Mexico State Legislature, National Association of Counties, American Journal of Public Health and King County Prosecuting Attorney’s Office.

Other sources described the finding even more broadly as LEAD participants having 58 percent fewer arrests, recidivism or crime, according to the report. Those sources included an American Journal of Criminal Justice article, New York State Department of Health materials, and organizations in Maryland and North Carolina.

The report traces the statistical problem back to the initial promotion of LEAD research in 2015, including press releases from the Seattle Police Department and University of Washington, saying a finding concerning the odds of arrest was subsequently translated into claims about probability and absolute percentage changes.

The result, Zimmerman argues, was that policymakers around the country were making decisions about an allegedly “evidence-based” criminal justice program using statistics that substantially exaggerated what the underlying research could prove.

The report says those differences could have been material to legislative votes, government appropriations and competitive grants, where stronger claims of program performance could affect funding decisions.