Civilian alternative response systems are badly oversold.
The death of George Floyd was taken by many as indicative of a plague of murders by police, and hastened demands to reduce the rate of police contact with the public. It spurred demands for “alternative response models” — responses to people in crisis by professionals other than police — as a matter of “health equity” aligned with “public health values.” Across more than 80 million police-community contacts a year, fatal force occurs in just 0.00041% of 911 calls in the U.S.; however, according to the American Public Health Association, law enforcement violence is now a critical public health issue.
Any such incident is consequential, but the statistical rarity of the police use of fatal force complicates claims that lethal outcomes are representative of routine policing. Nevertheless, many reformers have taken up the call and are advancing their agenda.
In the eyes of these advocates and policymakers, the use of alternative responses is the cornerstone of efforts to reduce police involvement in private and public space. Medics and counselors will attend mental health crisis calls, outreach workers will attend victims of addiction and homelessness and credible messengers and violence interrupters will settle neighborhood beefs and reduce shootings. The goal is to make these programs “integral components of the public safety and wellness infrastructure.” In New York City, Mayor Zohran Mamdani’s new Office of Community Safety aims to expand these programs, even though the impact of a pilot program, B-HEARD, has not been assessed because the organizations responsible for implementation “do not, overall, collect sufficient evidence to assess the program’s effectiveness against its stated goals.”
Before New York and other cities double down on alternative response, they need to think hard about how it works — and doesn’t work — in practice.
The limits of diversion and the realities of implementation
Advocates of a public health approach to violence deploy carefully selected data to point out that crime control is a small part of policing. Clearly someone forgot to tell folk in Philadelphia. More than two-thirds of Philadelphia police patrol time spent on calls for service received from the public in a year is related to crime or quality of life or has the potential to be so related (68.4%). Using similar categories, Cynthia Lum and colleagues’ data showed the average across nine US cities was 60%.
Could the remaining non-crime calls be diverted to an alternative agency? As Loren Atherley, senior research scientist for the Seattle Police Department, has pointed out, the potential for large-scale diversion may be more limited than is often implied. The practical question is not whether some calls can be handled by non-police professionals, but how many can be diverted without introducing unacceptable risk. Atherley’s analysis of three years of data from Seattle suggests that about a quarter of police service hours could be diverted to non-police personnel without significant risk. He does, however, recognize what he calls the “Ratcliffe paradox” (kindly referencing a recent article of mine) in that determining whether a situation is appropriate for non-police response often requires an initial police presence. Risk assessment in dynamic environments is inherently uncertain, and public agencies understandably tend to err on the side of caution.
Existing alternative response programs are frequently placed on a pedestal with uncritical admiration. Eugene, Oregon’s CAHOOTS program is a prominent example, with claims that up to 23% of the calls in the emergency response system were dealt with by CAHOOTS (Crisis Assistance Helping Out On The Streets). Brian Stettin, who served as senior advisor on severe mental illness under former New York Mayor Eric Adams, reports this is misleading and estimates that the rate is more likely in the range of 3-8%. Whatever the real rate, it wasn’t enough to prevent the program from being shut down in 2025.
Context matters as well. I suspect many commentators (like me, writing from Center City, Philadelphia) reside in relatively safe, functional communities where many of the informal mechanisms of public life perform reliably — the “elephant of social control.” But to extrapolate a program like CAHOOTS from the affluent college town of Eugene, Ore., to somewhere like the Southwestern district of Baltimore is an altogether different proposition, where the elephant is barely on life support.
The practical question is not whether some calls can be diverted, but how many can be diverted without introducing unacceptable risk.
These constraints are compounded by decades of underinvestment in social services that have not kept pace with demand for services related to homelessness, mental health and addiction. This constrains the effectiveness of any response model, whether police-led or civilian. Institutional failures do not disappear; they are displaced. As David Thacher, professor of public policy and urban planning at the University of Michigan, observes, police are routinely called upon precisely because other systems have failed or are inaccessible, making them, in effect, managers of institutional breakdown.
This was evident in the extensive fieldwork colleagues and I undertook evaluating Philadelphia’s transit police opioid response and co-responder programs. A year-long randomized trial, largely in the subterranean network of tunnels that extend for blocks beneath the center of the city, examined a dedicated police SAVE team (Serving a Vulnerable Entity) aiding the vulnerable community. Shifts were randomized between the usual arrangement of officers working alone or in pairs, with shifts when they were accompanied by trained social work specialists. In terms of getting vulnerable people to accept transport to a shelter, triage facility or hospital, the combined social work-police teams outperformed police-only teams 41% to 32%; however, every team experienced how challenging it was to work with the city’s shelters and treatment facilities. Facility capacities are limited, users feel their rules are often restrictive and shelters are frequently perceived to be dangerous. People with extensive unmet needs often return to the street within days or hours. In some cases, police officers brought more practical experience to these encounters than their civilian counterparts, demonstrating a better understanding of the support ecosystem and tailoring more effective responses.
What if social workers replaced officers in a wide swath of interactions? We did not test that scenario, but the “doorman fallacy” is useful here. A hotel doorman who opens a lobby door can be easily replaced by a machine, but in the words of Rory Sutherland, “opening the door is only the notional role of a doorman; his other, less definable sources of value lie in a multiplicity of other functions, in addition to door-opening: taxi-hailing, security, vagrant discouragement, customer recognition, as well as in signalling the status of the hotel.” When not engaged in activities related to the vulnerable community, the social work specialists largely waited to be called, but the transit police officers performed a wide range of ancillary public safety functions such as helping with medical emergencies, tackling crime and disorder, guiding tourists and reassuring passengers. In an ironic twist, the SAVE initiative was cancelled by the police chief due to a lack of resources and a perceived need to increase the number of generalist officers available to handle whatever came up. The police chief needed more hotel doormen.
Would improving training within existing systems yield greater returns than building parallel ones?
Costs and constraints
Much of the discussion centers on an underlying assumption that police social service delivery to vulnerable communities is notably poorer than that of other providers. And debates about alternative responses often place significant weight on “lived experience” as a qualification for service providers. There is limited evidence here, and what evidence does exist is mixed. Research on interventions such as cognitive-behavioral therapy suggests that effectiveness is more closely associated with implementation quality and provider training than with experiential similarity between provider and client. In other words, people in recovery do not need a hand from someone with the same experiences, but instead access to well-trained, high-quality support structures.
In “The End of Policing,” sociologist Alex Vitale argues that improving police training is beside the point, because “The reality is that the police exist primarily as a system for managing and even producing inequality.” He adds, “American police receive a great deal of training.” If by that he means the amount of training time, his comment will surprise anyone who knows the topic. For example, at the time of writing, police training in South Carolina comprises four weeks at a local police station watching videos, eight weeks at a police academy, and a mandatory four weeks of post-academy field training (approximately 1,080 hours). The South Carolina Board of Cosmetology demands more training for someone who cuts and colors your hair (1,500 hours).
And yet still, most police encounters are resolved without incident, raising a different question: Would improving training within existing systems yield greater returns than building parallel ones? As New Yorkers will soon find out as the Department of Community Safety gets up and running, alternative response teams are not cost-neutral, and reallocating a portion of police funding can’t in and of itself support the infrastructure of an entirely new administrative entity, nor negate the need to sustain those functions in the police service. A new organization requires HR, systems and policies, accountability structures and sustained funding for vehicles, offices, training facilities and pension systems — all of which already exist in policing.
The value of alternative responders must also be evaluated not only in absolute terms, but relative to other possible costs. Specialization is expensive, and alternative responders have a narrow role. For all their imperfections, police, like hotel doormen, have a broad social function. In the process of their activities, built environment problems can be identified, crime opportunities reduced, crime victims reassured, confidential sources recruited and harmful social conditions ameliorated.
Given Americans’ propensity to want services and their hostility towards paying for them, increasing police training and recruitment might do more to improve a swath of social service response functions than the costly support necessary for a new, largely untested organization. Having police secure a situation and then wait for a team with scant additional capabilities to turn up seems less efficient than simply having the officers deal with the situation on arrival. Co-responder teams may seem an effective compromise, but at what point do the police officers gain sufficient experience and expertise to render the civilian co-responder redundant?
The current case for alternative response as a scalable, cost-effective replacement for the police in a social service role is grounded in weak analysis and skirts over potential adverse outcomes. Worse, these efforts may deflect attention from the potential benefits of improving law enforcement capability. Current efforts to institutionalize alternative response models as core components of public safety infrastructure may therefore be premature — not because the underlying goals are misguided, but because the operational, fiscal and institutional constraints are more substantial than is often understood and acknowledged.
