With violent crime up significantly in recent years, House Democrats have been working furiously to pass new crime legislation before the midterms. However, disagreements between factions in the party forced Nancy Pelosi on July 27 to postpone a vote on a slate of crime measures. A group of progressive lawmakers objected to provisions advanced by moderates providing municipalities with funds to hire more police, while members of the Congressional Black Caucus, many of whom support additional police funding, demanded new transparency and accountability requirements.

How Democratic leaders resolve these tensions and pass legislation is unclear. But there’s more than police funding at stake. A vote will determine the fate of community safety legislation in this package that all sides agree—more police or not—could improve public safety. Among the most important is a provision sponsored by Representative Katie Porter that would provide funds for states and municipalities to hire and train mental health professionals—instead of police officers—to handle nonviolent emergencies. The idea is that teams of clinicians and paramedics can free police to deal with more dangerous situations, lessen the chance of law enforcement shootings, and channel those struggling with addiction and mental illness into treatment instead of jail.

After the racial justice uprisings of 2020, a handful of cities, including New York, Washington, D.C., Austin, and San Francisco, started experimenting with this civilian responder model, first pioneered in Eugene, Oregon, in 1989. Progressives saw these programs as a component of their “defund the police” agenda, but in practice, they have been implemented without diverting funds from law enforcement and have largely been embraced by police departments. “Nobody became a cop because they were really excited about dealing with drunk people, because they really wanted to talk to a schizophrenic person who thinks the FBI is pursuing him,” says Keith Humphreys, a behavioral science researcher at Stanford. These programs are meant to relieve this burden.

But are they working? The data out of New York, San Francisco, and elsewhere has been promising, but until recently, no studies had examined the cause-and-effect relationship between such a program and crime, cost, and outcomes for people in crisis.

That changed in June when Stanford University researchers Thomas S. Dee and Jaymes Pyne published a study in Science Advances on a civilian responder program in Denver called Support Team Assisted Response, or STAR. The study found that STAR’s pilot program contributed to a 34 percent drop in low-level crime while reducing arrest rates and possibly saving the city money. These findings represent the best argument yet for federal funding of a novel approach to law enforcement that might reduce not only crime, but also the political temperature of the crime issue.


In May 2019, a year before calls to defund the police proliferated, a contingent of Denver lawmakers, officials, and police officers spent three days in Eugene doing ride-alongs with responders from their mental health program, called CAHOOTS (Crisis Assistance Helping Out on the Street). That November, the Denver police chief, Paul Pazen, presented a tentative outline to the city council, highlighting how STAR would be a welcome “force-multiplier” for law enforcement. The council approved $208,141 for a six-month trial.

Dee’s and Pyne’s study focused on that initial six-month pilot run, which began on June 1, 2020. A rotation of a few mental health professionals began working in shifts out of a single van, responding to behavioral health incidents in Denver’s downtown business district. In those six months, the STAR team responded to 748 incidents related to mental health crises, indecent exposure, homelessness, and substance use where police or 911 dispatchers deemed the threat of violence low. Their goal, says Carleigh Sailon, a social worker who helped found STAR, was “connecting folks to community resources and other ongoing support that can help to keep them out of crisis”—and out of the criminal justice system. In not one case did a STAR response result in an arrest.

Before the Stanford study of Denver’s pilot program, research on alternative policing models was mostly descriptive or correlational. In contrast, Dee told me that his study brought “tools of causal inference to bear”—that is, he wanted to tease out the isolated effect STAR had on crime.

Conveniently, STAR only operated during certain times of day, on certain days of the week, and in certain neighborhoods during the pilot period, making a natural experiment possible. Dee and Pyne used the granular, precinct-level crime data that Denver makes publicly available. By comparing rates of low-level crime in precincts where STAR was dispatched in the six-month pilot period to rates in those same precincts before the pilot, as well as rates in precincts STAR didn’t cover during the pilot period, Dee and Pyne concluded that the presence of STAR was associated with a 34 percent drop in recorded low-level crime.

But was there a genuine decrease in such crime—in incidents of indecent exposure, trespassing, and the like—or were STAR responders simply not recording such behavior as a crime the way police would? The drop, they concluded, was genuine: In precincts STAR covered, low-level crime fell even during days and hours when the program was not available. The finding was consistent with the theory that connecting low-level offenders—many of whom are homeless and mentally ill—with appropriate care instead of arresting them can lower recidivism. “If you’re having a psychotic break and the police show up, maybe you’re held in jail briefly, then you’re put back out in the community—you’re likely to re-offend,” Dee told me. “But if, instead, you receive health care, you may go on a better path that keeps you from re-offending.”

Stanford’s Humphreys, who was not involved in the study, told me he believes this finding was methodologically sound. A natural study outside the lab of this kind is never perfect, he noted, but in criminology, “evaluations of this rigor are rare.”

There were other important findings. In precincts where STAR operated, serious crime did not rise. This, Dee and Pyne wrote, serves as a rebuke of the “broken windows” theory, which holds that reduced police enforcement of low-level crime will cause a spike in serious crime. And Dee said that, while a full accounting of the costs associated with the STAR model will require further research, their study found that a STAR response costs just a quarter as much as a police response. Appropriate care meant savings for the city on unnecessary hospitalizations, arrests, and prosecution.

Sailon told me about one incident that showed how STAR can help those in distress in ways a traditional police response could not—while saving money. Early in the pilot period, she was dispatched to help a distressed man in a broken wheelchair. At the scene, she learned that he’d lost contact with his Veterans Affairs case manager during the pandemic and had lapsed into homelessness. As a public health expert, she knew who to contact, and the man was soon reintegrated at the VA. Police, lacking the relevant contacts and expertise, might have instead pursued a costlier option like calling an ambulance, she said. Plus, a police response to such calls would misuse their time. Pulling a police cruiser out of regular service “to deal with an issue that’s related to homelessness and poverty,” Sailon said, “is using a financial resource of the city in a way that it’s not really intended.”


After the pilot period concluded in December 2020, Denver extended STAR’s funding, and the program began adding more vans. After two years, STAR responses still have not led to a single arrest. With five vans and plans to have 10 in operation by the end of this summer, STAR will eventually cover all Denver precincts. Sailon, who moved from STAR field responses to a managerial role in Denver’s 911 dispatching center after the pilot, said the program would soon field 10,000 to 15,000 911 calls a year. (The city receives about 1.5 million 911 calls a year.)

How might civilian response systems like STAR, if deployed nationally, reduce the need for conventional policing? An October 2020 study from the liberal think tank Center for American Progress reviewed 911 data from eight cities (which did not include Denver) and found that community responders could handle between 21 and 38 percent of all 911 responses, depending on the city.

But Denver officials who advise other cities on building such programs say that figure is overly optimistic. The city’s internal estimate is that a fully built-out STAR could only respond to about 2.8 percent of 911 calls. STAR was designed as a marginal supplement to police, not a replacement for them, and it’ll stay that way. None of STAR’s $1.5 million budget for this year was diverted from the police department, and that’s not about to change. “I think that there’s the automatic assumption that these types of programs must be adversarial to police departments,” Sailon said, explaining that STAR’s success from the beginning was inextricable from its positive relationship with law enforcement. (A slightly different model is Eugene, where collaboration between police and CAHOOTS is high, but the team is funded with 2 percent of the police budget and responds to about 5 to 8 percent of 911 calls that police would otherwise handle.)

City officials say there’s a natural ceiling to STAR’s expansion based on what kinds of emergency calls can be expected to remain nonviolent. I asked Pazen, the Denver police chief, why unarmed civilian responders couldn’t eventually handle a more extensive range of call categories—traffic stops, shoplifting, and the like. He referenced an incident last year when a police officer was shot with a sawed-off shotgun while responding to a trespassing call. “If you send a STAR van to that call, and somebody gets shot or stabbed or killed, tell me what recruiting looks like in the future,” Pazen said. “Tell me who’s going to sign up for that.”


Is STAR’s success replicable across the country? It’s too soon to say. “​​I remain a committed and agnostic empiricist,” Dee said. “I’m going to want to see if other cities can do it.” There can be procedural obstacles: Seattle’s efforts to create such a program, for example, have so far amounted to “a lot of administrative shuffling” with little to show for it, as the Seattle blog PubliCola put it, and training 911 dispatchers has proved a headache in some cities. But in a recent op-ed, Dee argued that his study on STAR is “a strong argument for studying this innovation throughout the country,” and experts I spoke with agree. “Most of what we do around crime proceeds on far weaker evidence than what they’ve gathered here,” Humphreys said.

Expansion beyond a limited cluster of liberal cities will mean federal funding. “That’s what I’d like to see: Fund it through [the] DOJ,” Humphreys said. “You have Congress say, ‘We are going to fund 20 American cities to do this, replicate the Denver experience.’” This process has already begun: In December, the DOJ announced a $34 million package for cities to fund new behavioral health programs, about half of which was earmarked for programs like STAR. Further expansion could be achieved through Porter’s provision in the crime package before Congress or through other legislation that fulfills the president’s request for $15 billion in grants for local crisis services in his proposed 2023 budget.

Congress is a graveyard of failed public safety reforms, but with the potential to help police, save costs, and reduce minor crime, the alternative responder model has uncommonly universal appeal. A July poll found that even the majority of Republican voters favor federal funding for programs like STAR. Pazen underscored that there’s nothing partisan about the idea. “It doesn’t matter—red state, blue state, blue city, red city,” he said. “Better outcomes for people in crisis is a good thing.”

Will Norris

Will Norris is an intern at the Washington Monthly.