Zohran Mamdani Wants Civilians to Replace Cops. Will It Work?

Civilian alternative crisis response: much more than you wanted to know.

Courtesy John Lamparski/Getty.

Zohran Mamdani, New York City’s likely next mayor, has a complicated history with the police. Once an enthusiastic supporter of “defund,” Mamdani has tempered his past radical views during the campaign, going so far as to issue a general apology to the NYPD.

But public repentance doesn’t mean the frontrunner doesn’t want big changes at One Police Plaza. His campaign has floated a 17-page proposal for a “Department of Community Safety,” which would shift many responsibilities currently allocated to the NYPD into the hands of civilians. On the debate stage last week, he highlighted one of his more controversial proposals: expanding the use of civilian alternative responders to certain 911 emergency calls.

“Our city is asking those same police officers who are being asked to respond to shootings, respond to murders, to also respond to [mental health] calls,” Mamdani said. “I trust the dispatchers who would be receiving these calls to make the determination as to whether there was any indication of violence. If there’s no indication of a threat of violence, then we would send the mental health experts and providers to respond to those same incidents. The reason I believe in the efficacy of this approach is because of the fact that it has been delivered elsewhere in the country.”

Source: Ba et al. 2025.

Mamdani’s plan is not unique to his campaign or New York. As the above chart shows, over 100 cities have launched “alternative crisis response programs,” almost all of them since the murder of George Floyd in 2020. New York City actually already has its own, called B-HEARD. But Mamdani has claimed that the program is underfunded, and has promised to expand it if elected.

Civilian alternative programs are controversial—a prominent police abolitionist has lauded Mamdani's plan, while one retired NYPD sergeant called it “probably the worst idea I’ve heard of in a long time.” But most coverage has failed to ask: what do we actually know about what civilian alternative response does? Are they a brilliant intervention, or a disaster waiting to happen?

Perhaps unsurprisingly, the answer is probably “neither.” The emerging research shows that civilians can pick up a meaningful but small fraction of the least challenging police calls. But that level maxes out quickly, such that they are unlikely to deal with the calls that create the most problems. Rather than big culture war questions, whether to adopt a civilian alternative program is ultimately about more boring questions of cost and benefit—questions to which the answers are rarely obvious.

Many cities, including New York City, have published reports on their new alternative responder programs. But these evaluations are limited in what they can tell us. They do not try to separate causation and correlation by comparing trends before and after the program was implemented in areas that did and did not take up the program.

There are three exceptions to this rule. One is Dee and Pyne 2022, published in Science Advances, and which evaluates the results of Denver’s STAR program. The other two are working papers: Ba et al. 2025, which evaluates Durham’s HEART program, and Davis et al. 2025, which evaluates the CAHOOTS program in Eugene, Ore., the nation’s oldest crisis response program.¹

Each of these evaluations uses a slightly different strategy to experimentally evaluate the effects of their programs.² Combined with some of the non-experimental evaluations, they imply a few stylized facts about alternative responder programs generally.

Responders do not handle that many calls: The upper-end estimates imply that alternative civilian responders can cover something like a low-double-digit percentage of all 911 calls. CAHOOTS, some have claimed, handles as many as 17 percent of calls. But that number seems flexible, with Davis et al.’s data showing CAHOOTS being involved in only about 9 percent of calls (including co-response). In Durham, HEART handles about 14 percent of calls, although only about 9 percent are in-person civilian responses without a co-responder.

Other figures are lower. Portland’s program, for example, reduced police call volume only by 3.2 percent. Denver STAR handled just 0.5 percent of all 911 calls. B-HEARD, New York City’s program, handled on the order of 15,000 calls last year—a vanishingly small fraction of the nine million 911 calls the city receives each year.

To be fair, many of these programs are pilots, restricted to fixed hours and of limited size. But even the big programs—CAHOOTS and HEART—are reliably covering less than 10 percent of calls. And it’s not obvious that the experience of Eugene, Ore. generalizes to New York City.

Civilians are mostly handing easy cases: civilians are never sent into situations where violence could emerge, but they also are not dispatched to, for example, property crime calls. Instead, the general pattern is that they handle low-level, non-serious calls. HEART, for example, covers “behavioral health issues such as intoxication, trespassing, and welfare checks.” Most of what STAR covers is mental health or substance use issues. 75 percent of what Portland’s program does is welfare checks.

Albuquerque Community Safety calls, Q3 of FY 2025

On the one hand, this is probably a boon to police officers, who would prefer not to be handling unsheltered individuals or doing welfare checks when they could actually be going after bad guys. On the other hand, it suggests a certain limit on how effective alternative responders can be—transporting homeless people is not exactly high priority work.

Civilian alternative response programs lower reports of and arrests for petty crimes: This is the big thing all the evaluations test for, and the effects are consistent and significant. In Denver, reports of STAR-eligible crimes fall 34 percent in STAR-treated areas; dispatching a civilian in Durham reduces the likelihood of a subsequent crime report by 57.5 percent. Comparing calls that are otherwise statistically equivalent, sending a civilian responder reduces the risk of arrest by 44 percent in Durham, and by 75 percent in Eugene.³

These big relative reductions may be obscuring small absolute reductions—in Durham, HEART response reduces crime reports by 9.2 percentage points and arrests by 2.8 percentage points. The CAHOOTS study also gives a little more insight into what’s driving this effect, noting that the difference is “almost entirely explained by involuntary holds (15 percentage points) and detoxification holds (5 percentage points), suggesting that CAHOOTS reduces arrests mostly by offering voluntary alternatives to non-criminal detention.”

In other words: when civilian alternatives are deployed or sent to the scene of otherwise similar calls, the result is reduced reports of petty crimes and a reduced likelihood of arrest. That’s good insofar as one thinks that petty offending should not generate arrests, either because it’s intrinsically wrong or because arrest can create unintended harms. But it’s not obvious that these calls were generating that many arrests to begin with, or that those arrests were unjustified—maybe someone needed an involuntary or detox hold.

Alternative response programs do not reduce or increase overall crime: That’s the other big takeaway from the Denver study—reports of non-STAR-eligible crime fall a non-significant 5 percent. That’s contrary to critics of civilian response programs, who argue that they will lead to increases in crime. The best experimental evidence we have right now say they do nothing at all to the level of major crime.

Over a certain threshold, alternative responders get less useful: This is the big takeaway from the CAHOOTS study. They use a complicated mechanism to estimate this, but in essence: the expansion of CAHOOTS into a new area/time is associated with a large increase in CAHOOTS activity and concurrent reduction in arrests. But trying to measure additional marginal CAHOOTS availability—by exploiting a measure of how busy the team is—shows diminishing marginal returns, with no additional reduction in arrests. And they show that in the absence of CAHOOTS, the majority of these calls would have gone unresponded-to—raising questions about how important they were in the first place.

Consistent with this, alternative responders don’t reduce use of force: One argument for sending in alternative responders is that reducing contacts with the police reduces risk of police violence. At least in Durham, however, there’s no significant effect on use of force. (Oddly, exposure to HEART increases the likelihood of follow-up 911 calls, while CAHOOTS reduces it—opposite results that each paper interprets as a positive sign for the model.)

If we were to tell a story that combines all these data points, it would go something like this: It is possible for civilians to pick up some fraction of 911 calls—generously, around 10 percent—without major incident. These are the least challenging calls, so it takes the burden off of cops to handle them. And people are less likely to get arrested from these calls, which is good insofar as arrest and jail time can be harmful.

At the same time, because they’re taking the least challenging calls, there’s an upper bound on how useful alternative responders can be—they can transport the homeless and do wellness checks, but they’re not going to get into serious situations. That upper bound is in part because it’s hard to know in advance which calls should go to whom: in a study of Philadelphia call data, public health-related 911 calls were confused for public safety calls (or vice-versa) about 20 percent of the time. But because of this, civilian alternatives are unlikely to get into the kind of situation where force is likely, so they can’t do much to reduce the direct harms of enforcing the law.

Which makes judgments about expanding B-HEARD—as Mamdani intends to—ultimately a problem of pragmatics, not ideology. NYPD officers cost a lot of money ($61,000 a year at the start, rising to $126,000 after five-and-a-half years of service plus lots of benefits). B-HEARD employees cost a lot too (between $82,000 and $113,000 depending on level). Separating B-HEARD from the NYPD, and moving it under Mamdani’s planned Department of Community Safety, will impose additional overhead costs. On the other hand, the program will prevent some number of arrests and reduce some amount of paperwork—those are dollars saved. How do these balance out?

These sound like boring questions (because they are). But civilian response is neither a revolutionary, life-saving alternative to the cops nor an anarchic proposal to end law and order. It’s mostly a shift in bureaucratic burden for dealing with petty issues in the everyday life of the city.

Which makes breathless denunciations of the idea from the right a bit silly. At the same time, it also challenges Mamdani’s presentation of the idea as a bold reform. It’s a pretty idea. Unfortunately, governing is more complicated than that.

Editor’s note: this article has been updated to reflect the fact that Dee and Pyne 2022 was published in Science Advances, not Science. We thank

Andy Wheeler

for the correction.

1. CAHOOTS actually stopped serving Eugene earlier this year.

2. I would explain these in detail, but they are sufficiently complicated that it would be hard to do so in a way that was both clear and concise.

3. HEART responses reduce arrests by 2.8 percent from a 5 percent baseline = 44 percent reduction. CAHOOTS responses reduce arrest probability by 24 percentage points relative to a control mean of 32 percent = 75 percent reduction.