Armed and Unprepared: The Cost of Asking Police to Do Everything

American police agencies function as the residual institution for needs that other systems leave unmet, including serious mental illness, homelessness, substance use, and poverty. Officers are selected, trained, and equipped primarily for law enforcement, and this mismatch imposes costs on people in crisis, officers, and public trust. Recent evaluations of civilian crisis response suggest that alternatives can divert a meaningful share of calls, but their reach is limited and their long-run effects remain uncertain. Here, I review the evidence and argue that the problem is difficult mainly because it is a distributional question about public obligation, not a technical question about dispatch.

What is the actual problem?

Policing is usually discussed as a response to crime. In practice, police are also the only public agency that is available at all hours, mobile, and legally empowered to intervene. When a family cannot reach a clinician, when a person sleeps outdoors, or when intoxication disrupts a street, the telephone number to reach a responder is 911. The consequence is a durable institutional default: need that is not met elsewhere is routed to the agency that cannot decline the call.

The consequence is a durable institutional default: need that is not met elsewhere is routed to the agency that cannot decline the call.

This arrangement is rarely defended on its merits. It persists because it is convenient and because the alternatives require sustained investment across agencies and budgets. The question I seek to address is therefore not whether police can perform social functions, which they plainly do, but whether a society should rely on them to do so and what that reliance reveals.

The Scale of Unmet Need

Homelessness offers a visible measure of the pressure on public space. The U.S. Department of Housing and Urban Development (HUD) reported that more than 770,000 people were homeless on a single night in January 2024, an increase of 18 percent over 2023 (HUD Exchange, 2024). HUD's 2025 report counted 745,652 people, of whom 266,320 were living unsheltered (U.S. Department of Housing and Urban Development, 2026). Unsheltered homelessness matters for policing because it places daily life in public view, where complaints and enforcement occur.

Mental illness appears in the criminal justice population at high rates. A Bureau of Justice Statistics survey found that 26 percent of jail inmates met a threshold for serious psychological distress in the prior 30 days, compared with 14 percent of prisoners (Bureau of Justice Statistics, 2017). Jails are the point where unmet need and arrest intersect, and the survey illustrates how a custodial setting becomes a de facto provider of care.

Jails are the point where unmet need and arrest intersect, and the survey illustrates how a custodial setting becomes a de facto provider of care.

There is a measurement problem that deserves emphasis. Police data capture a crisis only when a call is coded as such, and many behavioral health situations appear under categories such as disturbance, trespass, or welfare check. A 2026 scoping review notes growing attention to the volume of police contact with people experiencing mental health problems and to the time and resources spent attending these incidents ("Mental Health and Routine Police Work," 2026). Estimates of prevalence therefore depend heavily on definitions and coding practices, and no single national figure is reliable.

A Mismatch of Training and Task

Officers are trained primarily in legal authority, tactics, and procedure. Crisis intervention programs exist, and the Federal Bureau of Investigation's Law Enforcement Bulletin describes the Crisis Intervention Team model as a response to concerns that officers often lack appropriate procedures for de-escalating agitated individuals (FBI Law Enforcement Bulletin, 2017). The Bureau of Justice Assistance cautions, however, that training operates within a broader collaboration between police and mental health systems (Bureau of Justice Assistance, n.d.). A Council of State Governments Justice Center survey likewise documents variability in state standards for de-escalation and related training (Council of State Governments Justice Center, 2020).

The greater difficulty is that training addresses the encounter but not its disposition. An officer who resolves a crisis safely must still decide where the person goes next. If treatment capacity is absent, the options narrow to an emergency room, an arrest, or leaving the person where they were found. The limiting factor is the availability of services, and training cannot supply it.

The consequences of failure are severe. The Treatment Advocacy Center estimated that people with untreated serious mental illness are 16 times more likely to be killed during an encounter with law enforcement than other civilians, with a death rate of 0.0345 percent against 0.00211 percent (Fuller et al., 2015). These figures come from an advocacy organization, rest on imperfect data, and are now a decade old, so they should be read as indicative rather than definitive. The data gap is itself telling. PBS reported that federal law requires the collection of data on police force and deaths, including information on mental illness, but does not require departments to report how many people officers kill (PBS NewsHour, 2022).

Evidence on Alternative Responders

CAHOOTS (Crisis Assistance Helping Out On The Streets) in Eugene, Oregon, is the longest-running mobile crisis response program in the United States and has served as a template for many imitators. For years, its impact was gauged by informal estimates. A 2022 Eugene Police Department summary reported a diversion rate of police calls between 3 and 8 percent (OPB, 2026). Burton et al. (2026) applied a difference-in-differences design and estimated that 18 percent of calls were directly substituted for police response. They also reported that accounting for prevention raised the figure to 23 percent, indicating that civilian crisis teams can reduce police call volume through both direct substitution and incident prevention (Burton et al., 2026).

Dee and Pyne (2022), economists affiliated with Stanford University, evaluated Denver's Support Team Assisted Response (STAR) pilot, which dispatched mental health clinicians and paramedics to nonviolent calls. The program responded to 748 incidents in six months (Dee & Pyne, 2022). In precincts where STAR operated, reports of less serious crimes fell by 34 percent (KDVR, 2022). The authors calculated a direct cost of $151 per offense averted, against an estimated $646 for a minor criminal offense, and noted that these estimates exclude follow-up treatment and incarceration costs (Stanford Institute for Economic Policy Research, 2022). An Urban Institute review treats the program as a promising alternative while noting the attention CAHOOTS has received compared with other models (Urban Institute, 2023).

Three cautions apply. First, reach is limited. A 2026 analysis cites estimates that STAR could divert about 3 percent of all police calls and that Portland Street Response diverts about 3.5 percent (Burton et al., 2026). Second, the studies concern specific places, hours, and call types, and results may not generalize. Third, the cost figures exclude the treatment and housing that must follow a successful diversion. The evidence supports the claim that civilian response is viable for a defined category of low-acuity calls. It does not show that such programs can replace the broader social role police now play. Civilian teams cannot fully replace police in this role, but they can take on a defined category of calls within a formal, well-resourced system.

Why the Problem Resists Resolution

Four features make this problem unusually difficult. First, the deficit is upstream. Redirecting a call changes who arrives but not what is available afterward. If outpatient psychiatric capacity, supportive housing, and substance use treatment are insufficient, the new responder meets the same dead end as the officer.

Second, the boundary between social and criminal problems is porous. The same incident may involve a person in crisis, a genuine threat, and a complaint from a neighbor. Any alternative must specify, in advance and under uncertainty, which calls exclude police, which include them, and who bears responsibility if a judgment proves wrong.

Third, costs and benefits fall in different budgets. Savings from fewer arrests and jail days accrue to the criminal justice system, while the expense of crisis teams and treatment falls on health and social service agencies. Without a mechanism that reconciles these accounts, no single actor has the incentive to fund the whole solution. This point is my own analytic inference and is not drawn from the sources above.

Fourth, public expectations conflict. Residents want a prompt response to disorder and distress, and they also want fewer arrests and less force. These demands can be reconciled only if a capable, always-available, non-police alternative exists, and building one is a long-term commitment of the kind that budgets rarely sustain.

Implications for Civic Life

The default responder is a statement about priorities, even if it was never chosen deliberately. A society that answers acute illness, deprivation, and addiction with an armed agent of the state has, in effect, defined these conditions as problems of order. Whether that definition is justified is a normative question, and it has practical consequences for who is arrested, who is treated, and how communities perceive the legitimacy of public institutions.

Several questions remain unresolved. Which categories of calls should civilians handle, and by what criteria? Who should finance the care that follows a crisis call, and how should savings and costs be shared across agencies? How much risk should communities accept while new models are tested, and who decides? What obligations do we owe officers asked to perform clinical and social work for which they were not prepared, and the people they encounter? The pilot evidence shows that other arrangements are possible. Whether they become the norm depends on collective choices that extend far beyond police departments.

This article serves as a starting point for a larger joint research initiative spearheaded by the CityGov Group in partnership with its academic partners.

References

Bureau of Justice Assistance. (n.d.). Training. Police-Mental Health Collaboration (PMHC) Toolkit. https://bja.ojp.gov/program/pmhc/training

Bureau of Justice Statistics. (2017). Indicators of mental health problems reported by prisoners and jail inmates, 2011-12. U.S. Department of Justice. https://bjs.ojp.gov/content/pub/pdf/imhprpji1112.pdf

Burton, N., Herbert, C., & Rohlfs, R. (2026). Quantifying CAHOOTS: Measuring a mobile crisis response program's impact on police response through call diversion and prevention. Science Advances, 12(39), eaeb8064. https://www.science.org/doi/10.1126/sciadv.aeb8064

Council of State Governments Justice Center. (2020). The variability in law enforcement state standards. https://csgjusticecenter.org/wp-content/uploads/2020/02/JC-LE-Survey.pdf

Dee, T. S., & Pyne, J. (2022). A community response approach to mental health and substance abuse crises reduced crime. Science Advances, 8. https://www.science.org/doi/10.1126/sciadv.abm2106

FBI Law Enforcement Bulletin. (2017, September 20). The importance of mental health training in law enforcement. https://leb.fbi.gov/articles/featured-articles/the-importance-of-mental-health-training-in-law-enforcement

Fuller, D. A., Lamb, H. R., Biasotti, M., & Snook, J. (2015). Overlooked in the undercounted: The role of mental illness in fatal law enforcement encounters. Treatment Advocacy Center. https://www.tac.org/wp-content/uploads/2023/11/Overlooked-in-the-Undercounted.pdf

HUD Exchange. (2024, December 30). HUD releases January 2024 Point-in-Time count report. https://www.hudexchange.info/news/hud-releases-2024-ahar-report/

KDVR. (2022, June 8). Study: Denver's STAR program reduced crime, costs. https://kdvr.com/news/local/study-denvers-star-program-reduced-crime-costs/

Mental health and routine police work: A systematic scoping review. (2026). Health & Justice. https://link.springer.com/article/10.1186/s40352-026-00405-4

OPB. (2026, September 27). UO researchers: CAHOOTS call diversion rate "substantially" greater than first estimated. https://www.opb.org/article/2026/09/27/cahoots-call-diversion-rate-substantially-greater-than-first-estimated/

PBS NewsHour. (2022, September 2). How some encounters between police and people with mental illness can turn tragic. https://www.pbs.org/newshour/health/how-some-encounters-between-police-and-people-with-mental-illness-can-turn-tragic

Stanford Institute for Economic Policy Research. (2022, June 8). Groundbreaking study shows benefits to reinventing responses to nonviolent 911 calls. https://siepr.stanford.edu/news/groundbreaking-study-shows-benefits-reinventing-responses-nonviolent-911-calls

Urban Institute. (2023). Understanding Denver's STAR program. https://www.urban.org/sites/default/files/2023-08/Understanding%20Denver%E2%80%99s%20STAR%20Program.pdf

U.S. Department of Housing and Urban Development. (2026, May 29). HUD releases 2025 Annual Homelessness Assessment Report to Congress. http://www.hud.gov/news/hud-no-26-037