Does It Work?

Two questions sound identical but are not. First: do the programs housed inside Community Safety Departments produce measurable results? Four independent evaluations have examined this. Second: does the CSD structure itself produce better results than the same programs housed in separate agencies? On this question, no study exists.

A decision-maker evaluating whether to create a CSD is evaluating both questions simultaneously. This card lays out what the evidence supports, how strong the underlying sources are, and where the gaps remain.

The Safety Record

Across every documented CSD and its component programs, the safety record is consistent: zero or near-zero serious injuries to responders across hundreds of thousands of calls.

Albuquerque’s ACS has responded to more than 120,000 calls with police backup needed in fewer than 1% of cases.¹ Durham’s community response teams have never required police backup for a safety threat.² Minneapolis logged 16,000 calls with zero serious responder injuries.³ San Diego County handled 13,000 calls with 98% resolved without law enforcement.³ St. Petersburg’s crisis program completed more than 11,000 contacts (including dispatched calls, follow-ups, and officer referrals) with zero incidents, injuries, or life-threatening situations.⁴ Denver’s crisis program has never needed police backup.⁵

These are all program-reported figures. No independent evaluation has assessed responder safety as a primary outcome.

Critics argue that sending unarmed civilians to crisis situations is dangerous. The program-reported data from every documented CSD shows a different pattern. The safety record rests on dispatch screening: calls involving weapons, credible threats of violence, or active crime go to police before a civilian team is dispatched.¹

Crime and Arrest Reduction

Four independent peer-reviewed or working-paper studies have examined whether civilian-only crisis response reduces crime and arrests. All four point in the same direction.

Denver (Stanford University, 2022). Denver’s crisis program, which is not a CSD but uses the same civilian-only response model. Economists Thomas Dee and Jaymes Pyne conducted an independent evaluation and found that in areas where Denver’s crisis program operated, reports of low-level crimes dropped by 34% while violent crime rates remained steady.⁵ Dee described the mechanism: people who receive appropriate clinical treatment have fewer subsequent encounters with law enforcement, fewer detentions, and fewer hospitalizations. The Stanford evaluation also found that department responses cost roughly one-quarter of what criminal justice processing costs for the same situations.

Durham (National Bureau of Economic Research, 2025). Economists Bocar A. Ba, Patton Chen, Tony Cheng, Martha C. Eies, and Justin E. Holz evaluated Durham’s CSD using a difference-in-differences design and found that crime reports declined and arrests dropped.⁶ Response times decreased for both the department and police, with the improvement nearly twice as large for department responses. The researchers found that these effects “are driven by the program itself rather than changes in enforcement behavior.” They also found that the department “does not deter future calls,” meaning civilian teams handle emergencies without escalation and people continue using emergency services afterward.

Wayne State University (2025). Leonard Swanson, Sheryl Kubiak, and colleagues tracked outcomes across five Michigan communities over eleven months and found that individuals who received help from civilian-only teams were 45.2% less likely to be arrested than those who received a police-only response.⁷ The study compared civilian-only teams, co-responder models, and office-based crisis intervention. Only civilian-only response produced a statistically significant reduction in arrests. Co-responder programs produced arrest outcomes similar to police-only response and showed higher hospitalization rates.

Community Mental Health Journal (multi-study comparison, 2025). A comparison by Bakko, Swanson, Zettner, and colleagues confirmed the pattern: civilian-only response was the only intervention among those tested that produced significantly lower arrest incidence in the year following an initial crisis, compared to co-responder units and officer-based response.⁸ Civilian-only response also reduced jail admissions and emergency room admissions.

These findings apply to the mobile crisis component of CSDs. Whether the CSD structure amplifies or is incidental to these results has not been tested.

Police Workload Relief

The operational data from CSD cities consistently shows that civilian response frees officer time for law enforcement priorities. The figures below are reported by the programs themselves, not independently verified.

Durham’s CSD has responded to more than 32,000 emergency calls and freed more than 10,000 officer hours (updated from earlier reports of 5,500 hours).⁹ Oklahoma City’s alternative response program contributed to a 57% reduction in police responses to mental health calls in a single year.¹⁰ New Orleans achieved a 36% reduction in police burden through its civilian crisis unit.³ Philadelphia’s network of mobile crisis teams now averages 20 mental health emergencies per day diverted from police.³

Albuquerque Police Commander Luke Languit credited the CSD with enabling crime reduction: “With 100,000 calls for service, the Albuquerque police officers that are out here in our community, they’re now able to focus more time on those felony crimes, and we’re able to bring down our crime rates because we have this third branch of public safety taking those calls for service for us.”¹¹

Police leadership in cities with established programs has endorsed the workload effect. Minneapolis Deputy Chief Erick Fors: “Feedback from the rank and file has been very positive.”¹² Durham Police Chief Patrice Andrews (retiring May 2026), despite officers who initially “chafed over the political fights that accompanied the program’s creation,” now reports “I’ve heard officers on the radio asking for the department’s teams. So there is that change.”²

Cost Effectiveness

Three levels of cost evidence exist, each with different implications for a budget director.

Per-case economics. The 2025 fiscal evaluation of Durham’s CSD by NBER economists found that the average department response costs $1,191 but generates estimated fiscal savings of $2,093 per call, a net savings of $902 per call.⁶ The researchers concluded “the program pays for itself through fiscal externalities” including reduced ER visits, fewer hospitalizations, less criminal justice involvement, and lower incarceration costs. When Durham residents were surveyed, 95% expressed willingness to pay for the program, valuing it at $102.91 per year, more than eight times the program’s per-resident cost.

The Denver crisis program evaluation found that criminal justice processing of the same situations costs roughly four times what a civilian crisis response costs.⁵ SAMHSA data compiled from multiple program reports shows civilian crisis programs cost 23% less per case than police response and reduce follow-up hospitalization costs by 79%.¹³

Program-level savings. Guilford County saved $400,000 and freed 170 staff hours by deploying a small team of social workers who reduced one group’s non-emergency 911 calls from 344 to four in a single month.¹⁴ Madison, Wisconsin’s crisis team handled over 3,500 calls in 2024 on a budget of $1.7 million, calls that would otherwise have gone to police.³

Liability exposure. Local governments representing the 25 largest police and sheriff’s departments paid out more than $3.2 billion over a recent decade to settle civil lawsuit claims related to police encounters, with a substantial share involving behavioral health situations.¹⁵ The DOJ’s investigation of the Phoenix Police Department documented officers who “fire Tasers at people with little or no warning” during behavioral health encounters and “fail to recognize that a person’s disability may impact whether they can understand commands.”¹⁶

What the cost evidence does not yet show. The NBER cost-benefit finding comes from one city with one program design. Whether the $902 net savings per call holds in cities with different cost structures, different Medicaid populations, or different program designs has not been tested. The SAMHSA data is broader but less methodologically rigorous.

Service Connection and the Revolving Door

CSD programs that include follow-up components show reductions in repeat 911 calls for the individuals they serve.

Guilford County’s targeted case management reduced one group’s non-emergency 911 calls from 344 to four in 30 days.¹⁴ St. Petersburg’s crisis program “never closes a client. They do as much follow-up as needed so that the client can get engaged with long-term services.”⁴ Durham’s care navigation segment follows up within 48 hours and provides ongoing case management.¹⁷ Austin provides up to 90 days of transitional crisis services.³

The causal chain here has documented links and undocumented ones. The documented links: CSD teams respond to calls, provide clinical intervention, and connect individuals to services (well-documented across multiple cities). The undocumented link: whether those service connections persist long enough to permanently reduce repeat crises. Short-term repeat-call reductions are documented. Long-term longitudinal data on whether people connected to services by these programs stay connected over years is still thin.

Public Trust

A Portland State University evaluation found that neighborhoods served by Portland’s crisis response department reported increased trust in emergency services.¹⁸ The NBER study of Durham found that the department “does not deter future calls,” meaning people continued to use 911 after civilian response, which the researchers interpreted as a trust indicator: civilian response encourages rather than discourages help-seeking.⁶

Polling shows 84% of voters agree CSDs “meet unmet needs because some people who need help are too scared to call 911 because they’re scared of the police” (polling data measures sentiment, not outcomes).¹⁹

What the Evidence Does Not Show

No study has evaluated the CSD as an institutional form. All the independently evaluated evidence cited above examines specific programs (Denver’s crisis program, Durham’s department) or specific response models (civilian-only vs. co-response). No researcher has tested whether housing these programs in a unified department produces better outcomes than housing the same programs in public health, fire, or police departments. The CSD governance model is a plausible organizational theory supported by operational experience, not a conclusion supported by independent evaluation.

The evidence base is concentrated. The same programs appear repeatedly across different analyses: Denver, Durham, Harris County, St. Petersburg, Austin, Minneapolis, Oklahoma City, Albuquerque, Guilford County. These are the programs with transparent data. Rigorous outcome data comes from roughly a dozen cities.

Longitudinal data is thin. Short-term outcomes are well-documented: calls resolved, officers freed, costs reduced, arrests avoided, crises de-escalated. Whether people connected to services through CSD teams stay connected over years is largely unknown.

The cities with CSDs are not representative. Albuquerque, Durham, Portland, Seattle, and Cambridge are mid-to-large cities with political environments favorable to the model. Whether CSD outcomes generalize to rural counties, conservative suburbs, or cities with different political dynamics is untested.

The fiscal sustainability question is unresolved. The Medicaid pathway provides 85% federal match for three years.²⁰ No state has completed the transition to its standard match rate, so how program funding holds up beyond the enhanced period is unknown. The ARPA funding cliff looms for programs built on American Rescue Plan Act dollars.

---

The Bottom Line

The programs housed inside Community Safety Departments produce measurable results: four independent evaluations show reduced crime, reduced arrests, and fiscal self-sustainability for civilian crisis response.⁵ ⁶ ⁷ ⁸ Operational data from every documented program shows a consistent safety record and police workload relief. The evidence is concentrated in a small number of well-documented cities, applies primarily to the mobile crisis component, and does not yet address whether the CSD governance structure itself produces better outcomes than the same programs housed elsewhere. The component-program evidence includes four independent evaluations all pointing in the same direction. The departmental-structure argument remains a plausible but unproven organizational theory.

---

Source Appendix

1. City of Albuquerque, “Albuquerque Community Safety Department Marks Four Years of Impact and Innovation,” press release, September 2025. 120,000+ calls, <1% police backup. https://www.cabq.gov/acs/news/albuquerque-community-safety-department-marks-four-years-of-impact-and-innovation

2. Tradeoffs / The Marshall Project, “How Durham Got Police Onboard with Unarmed Crisis Response,” May 2, 2025. 99% safety record, 25,000+ calls, Andrews quotes. https://tradeoffs.org/2025/05/02/how-durham-north-carolina-got-police-onboard-with-unarmed-crisis-response/

3. Program-reported safety and workload figures for Minneapolis, San Diego County, New Orleans, Philadelphia, and Madison compiled from city reports and local news coverage. Minneapolis (16,000 calls, zero serious injuries): City of Minneapolis BCR program data. San Diego County (13,000 calls, 98% resolved without law enforcement): San Diego County News Center, January 2025. New Orleans (36% reduction in police burden): City of New Orleans crisis unit data. Philadelphia (20 mental health emergencies per day diverted): Philadelphia Inquirer and city behavioral health reporting. Madison ($1.7M budget, 3,500+ calls): City of Madison CARES program data.

4. Results for America, “Non-police emergency response: St. Petersburg, FL” case study. 11,000+ contacts, 97% diversion rate, 5,000+ unique individuals, zero incidents. https://catalog.results4america.org/case-studies/non-police-emergency-response-st-petersburg-fl

5. Thomas S. Dee and Jaymes Pyne, “A community response approach to mental health and substance abuse crises reduced crime,” Science Advances, Vol. 8, Issue 23 (June 8, 2022). 34% crime reduction, costs 4x lower than criminal justice processing. https://pmc.ncbi.nlm.nih.gov/articles/PMC9176742/

6. Bocar A. Ba, Patton Chen, Tony Cheng, Martha C. Eies, and Justin E. Holz, NBER Working Paper No. 34344, October 2025. $1,191 cost, $2,093 savings, $902 net per call, “pays for itself through fiscal externalities,” crime reports declined, “does not deter future calls,” 95% willingness to pay at $102.91/year. https://www.nber.org/system/files/working_papers/w34344/w34344.pdf

7. Leonard Swanson, Catherine Zettner, Amy Watson, Melanie Hinojosa, Juliette Roddy, and Sheryl Kubiak, “Eleven-Month Arrest Outcomes Among Three Crisis Response Models in Michigan,” Psychiatric Research and Clinical Practice, Vol. 7, Issue 3, pp. 182-190 (May 2, 2025). IRR 0.548 (45.2% less likely arrested); co-response not significant. https://pmc.ncbi.nlm.nih.gov/articles/PMC12418740/

8. Matthew Bakko, Leonard Swanson, Catherine Zettner, Kaitlyn Kok, Hosanna Fukuzawa, and Sheryl Kubiak, “A comparison of behavioral health crisis response models in meeting behavioral health goals and improving criminal legal diversion,” Community Mental Health Journal, January 3, 2025. DOI: 10.1007/s10597-024-01447-4. https://doi.org/10.1007/s10597-024-01447-4

9. City of Durham / ICMA, “City of Durham Wins ICMA 2025 Community Health & Safety Award.” 32,000+ calls, 10,000+ officer hours. https://icma.org/page/2025-community-health-safety-award-city-durham

10. KOSU, Sierra Pfeifer, “Oklahoma City plans to expand mental health crisis response with state opioid settlement money,” February 6, 2026. 57% reduction in mental health dispatches. https://www.kosu.org/health/2026-02-06/oklahoma-city-plans-to-expand-mental-health-crisis-response-with-state-opioid-settlement-money

11. Albuquerque Police Commander Luke Languit, remarks at ACS 100,000-call milestone press event, March 2025. https://www.cabq.gov/acs/news/albuquerque-community-safety-marks-100-000-calls-for-service-milestone

12. Minneapolis Deputy Chief Erick Fors, testimony at Minneapolis Public Safety Committee meeting, April 20, 2022. Reported in Southwest Connector, Jan Willms, June 21, 2022. https://swconnector.com/stories/alternative-to-police,6234

13. SAMHSA, “Crisis Services: Effectiveness, Cost-Effectiveness, and Funding Strategies,” Publication No. SMA14-4848. “23 percent lower average cost per case” and “reduced costs associated with inpatient hospitalization by approximately 79 percent.” https://library.samhsa.gov/sites/default/files/sma14-4848.pdf

14. NACo, “North Carolina county decreases 911 non-emergency calls,” November 2024. $400,000 savings, 170 staff hours, Justin Hargett. https://www.naco.org/news/north-carolina-county-decreases-911-non-emergency-calls-connecting-residents-services-they. WFMY: 344→4 figure. https://www.wfmynews2.com/article/news/local/guilford-countys-adult-care-team-non-emergency-ems-calls/83-e01ad2e6-3ad7-4354-b521-6f0ffb1850cb

15. Keith L. Alexander, Steven Rich, and Hannah Thacker, “The hidden costs of police misconduct,” The Washington Post, March 9, 2022. $3.2 billion across 25 departments over a decade. https://www.washingtonpost.com/investigations/interactive/2022/police-misconduct-repeated-settlements/

16. U.S. Department of Justice, Civil Rights Division, “Investigation of the City of Phoenix and the Phoenix Police Department,” June 2024. https://www.justice.gov/d9/2024-06/Phoenix%20Findings%20Report%20Final%20-%20Final%20508.pdf

17. CSG Justice Center, “Durham, NC — Expanding First Response Program Highlights,” updated December 2024. Care navigation 48-hour follow-up. https://csgjusticecenter.org/publications/expanding-first-response/program-highlights/durham-nc/

18. Portland State University Homelessness Research & Action Collaborative (Greg Townley, lead evaluator), Portland Street Response Year Two Evaluation, July 2023. Increased neighborhood trust. https://www.pdx.edu/homelessness/sites/homelessness.web.wdt.pdx.edu/files/2023-07/HRAC%20Portland%20Street%20Response%20Year%20Two%20Evaluation%20Report_FINAL%20FOR%20WEBSITE.pdf

19. Safer Cities, national survey of 2,400 registered voters. 84% agree on unmet needs from fear of police.

20. Milbank Memorial Fund, “Mobile Crisis Teams and Medicaid Funding,” June 2025. 21 states opted in by September 2024; enhanced 85% FMAP through March 2027. https://www.milbank.org/quarterly/opinions/mobile-crisis-teams-and-medicaid-funding-advancing-behavioral-health-crisis-response-across-the-united-states/. See also KFF: https://www.kff.org/medicaid/a-look-at-state-take-up-of-arpa-mobile-crisis-services-in-medicaid/