Community Safety Beyond Policing — What Alternatives Exist

The United States spends approximately $115 billion annually on policing. The United Kingdom spends roughly £17 billion. These figures dwarf investment in most of what research identifies as the actual drivers of community safety. Yet debates about public safety are almost entirely debates about police: how many, what powers, what accountability, what technology.

This is a framing trap. It treats policing as synonymous with safety, which obscures the question of whether the resources invested in policing produce more safety than the same resources invested elsewhere. The evidence for that comparison is, at minimum, equivocal.

Policing does some things well. It deters certain categories of opportunistic crime when patrol presence is high. It provides a mechanism for reporting and recording crime. It can apprehend specific perpetrators after the fact. It offers a sense of visible authority that some communities find reassuring.

Policing does other things poorly. It is not well-designed for mental health emergencies (which constitute 10–20% of all 911 calls in most jurisdictions). It is not designed for neighbor disputes, domestic conflicts at early stages, or the kind of low-grade community breakdown that precedes serious violence. Its presence in communities with histories of racialized violence can undermine the social trust it is meant to protect. And incarceration — policing's primary downstream tool — consistently increases rather than decreases recidivism for most offense categories.

The alternatives described below are not proposals to eliminate policing. They are proposals to match tools to problems — a basic principle of any functional system.

CVI programs operate on the premise that violence is contagious. Like communicable disease, it spreads through social networks: someone is shot, their friends seek retaliation, bystanders normalize violence, younger community members learn that violence is the expected response to conflict. Interrupt enough transmission chains and you can shift these dynamics.

The canonical model is Cure Violence, founded by epidemiologist Gary Slutkin in Chicago in 2000. It employs "violence interrupters" — people with credibility in high-violence communities, typically because of their own histories with street violence or incarceration — who intervene directly in conflicts before they escalate. They mediate disputes, counsel individuals, and work to change norms. Independent evaluations have found reductions in shootings of 30–60% in areas where the program operates.

Similar programs: the Newark Community Street Team (New Jersey), B.R.A.V.E. (Baltimore), and Advance Peace (California) which targets the specific small number of individuals most likely to be involved in gun violence and provides them with stipends and intensive mentorship as alternatives to street economies.

What makes CVI work where policing often does not: trust. Violence interrupters can go where police cannot — not because of legal authority, but because community members talk to them. They operate before a crime has been committed, which means they can prevent it rather than respond to it.

The CAHOOTS model (Crisis Assistance Helping Out On The Streets) has operated in Eugene, Oregon since 1989. It dispatches teams of a medic and a crisis worker — no police officer — to calls involving mental health crises, intoxication, welfare checks, and similar situations. The team has no weapons. CAHOOTS handles about 24% of Eugene's 911 call volume and requires police backup in less than 1% of cases.

Denver's STAR (Support Team Assisted Response) program launched in 2020 with a mental health clinician and paramedic team. In its first six months, it responded to 748 calls with zero need for police backup and zero criminal complaints against responders.

The evidence from these programs is consistent: a large proportion of crisis calls do not require armed officers, and sending unarmed clinicians produces better outcomes for the people in crisis and frees police capacity for situations that actually require law enforcement authority.

The barriers to expansion are mostly institutional. Police unions sometimes resist programs they perceive as reducing their jurisdiction. Dispatch systems are not always designed to route calls to alternative responders. Funding is frequently precarious.

The adversarial criminal justice model positions the state against an individual. The victim is a witness, not a participant. The outcome is conviction or acquittal, followed in many cases by incarceration. The process rarely addresses what the harmed party actually needs or what caused the harmful behavior.

Restorative justice (RJ) is a different model. It is not one program — it encompasses a range of practices: victim-offender mediation, circle processes, community conferencing. What they share is this: the harmed party, the person who caused harm, and their respective supporters come together to answer three questions: What happened? Who was affected and how? What do we need to do to repair this harm?

The research on outcomes is consistently positive for nonviolent offenses and juvenile justice: - Higher victim satisfaction than conventional prosecution (typically 70–85% vs. 25–40%) - Lower recidivism rates in most study contexts - More consistent completion of restitution - Faster resolution

New Zealand has incorporated restorative approaches into its youth justice system since 1989 with measurable reductions in incarceration of young people. Several Canadian provinces use RJ circles for serious violent offenses in Indigenous communities. Oakland and Los Angeles have school-based RJ programs that have significantly reduced suspensions without increasing unsafe behavior.

RJ does not work for all situations. It requires the person who caused harm to accept some responsibility, which makes it unsuitable for certain contested cases. It works poorly when power imbalances between parties are severe and unacknowledged. It can be co-opted by courts as a cost-cutting measure without adequate support for victims. These are implementation challenges, not inherent failures of the model.

CPTED (Crime Prevention Through Environmental Design) emerged in the 1970s through the work of criminologist C. Ray Jeffery and architect Oscar Newman. Its premise: the physical environment shapes the opportunity for crime. Design that increases natural surveillance (people can see what is happening), creates clear territorial definition (spaces have evident owners), and promotes legitimate activity reduces certain crime categories.

Specific applications: - Lighting improvements in dark corridors, parking areas, and alleyways. Multiple studies show that improved street lighting reduces crime in adjacent areas, with effects that extend beyond the lit area. - Removal of overgrown vegetation that creates concealment. - Activation of ground-floor spaces — retail, community use, active frontages — that put eyes on streets throughout the day. - Broken windows maintenance (keeping public spaces in good repair signals that they are monitored and cared for).

The "broken windows" theory has been badly misapplied to justify aggressive zero-tolerance policing of minor infractions. But the underlying environmental insight — that environments signal whether behavior is monitored — is sound. A clean, well-maintained public space is used more and is less likely to attract crime than an abandoned, neglected one. This is a design finding, not a policing strategy.

The strongest predictor of community safety is social cohesion — the density and quality of relationships between neighbors. A neighborhood where residents know each other, share information, and collectively care about the space they inhabit is one where unsafe behavior is noticed, named, and responded to before it escalates.

Mutual aid networks — organized systems for neighbors to help each other with food, transportation, childcare, health needs — are also informal safety networks. They are not designed as such, but they function that way because they build exactly the relational density that generates safety.

During COVID-19, thousands of mutual aid networks formed rapidly in neighborhoods across the world. Many have persisted. The research on their safety effects is sparse because it is difficult to isolate their impact, but the mechanism is well-established: people who know their neighbors notice when something is wrong, and people who have been helped by their neighbors are more likely to intervene when help is needed.

One specific mechanism: domestic violence. The research consistently shows that domestic abuse thrives in isolation. Neighbors who know each other, who have regular informal contact with each household, are more likely to notice warning signs and more likely to be trusted by victims as sources of help. Mutual aid networks that have contact with isolated households provide an informal early warning system that no patrol strategy can replicate.

For a subset of community harms — particularly low-level interpersonal conflict, nonviolent property damage, and first-offense nonviolent crimes — incarceration is an enormously expensive, deeply harmful intervention with poor outcomes. It removes people from families and communities, creates lasting economic disadvantage through criminal records, and produces high rates of reoffending.

Community-based accountability alternatives include: - Community accountability panels: Groups of trained community members who convene with the person who caused harm and the person harmed, reach agreement on what accountability looks like, and monitor its completion. Used extensively in some Scandinavian countries and in progressive prosecution jurisdictions in the US. - Transformative justice: A more radical framework that asks not just how to repair a specific harm but what conditions produced it, and what community changes would prevent recurrence. Associated with prison abolitionist movements but increasingly adopted in modified form by mainstream social service organizations. - Diversion programs: Pre-arrest or pre-prosecution diversion to social services, mental health treatment, substance abuse treatment, or community service. Well-run diversion programs have better recidivism outcomes than prosecution and dramatically lower costs.

The practical challenge is not identifying these alternatives — there is substantial evidence for all of them — but integrating them into functional systems. This requires:

Dispatch redesign. 911 systems must be able to route calls to appropriate responders. Currently, most systems route everything to police because that is the only 24/7 response option. Building alternative response infrastructure requires building alternative dispatch infrastructure.

Sustained funding. CVI programs, alternative response teams, and restorative justice programs are consistently funded through grants that expire. They cannot develop institutional knowledge and community trust on two-year grant cycles. Sustainable funding requires budget integration, which requires political will.

Inter-agency collaboration. The most effective safety systems involve police, mental health services, social services, housing, and community organizations working from shared information and shared protocol. These agencies rarely have functional working relationships or compatible data systems.

Community ownership. Programs designed and run by community members with lived experience of the problems they address consistently outperform those designed by external experts and imposed on communities. This is true for CVI programs, restorative justice, and mutual aid. It is less expensive to fund community-led solutions than to maintain externally managed institutions, and the outcomes are better.

Safety is not a product delivered by a government agency. It is a property of communities — produced by the quality of relationships, the condition of physical space, the availability of mental health and social services, and the economic conditions that shape whether people have enough. Policing is one tool in this system. It is not the most important one.