Public SafetyCommunity engagementMixed

Cure Violence / Group Violence Intervention — Community Violence Interruption

University of Illinois Chicago / National Network for Safe Communities (John Jay) · Multiple US cities · 2000

Summary

Community violence intervention represents a fundamentally different theory of gun violence: that most urban shootings are not random crimes but the predictable outcome of specific social networks, ongoing disputes, and retaliation cycles — and that these can be interrupted before they happen. The evidence is more heterogeneous than programs like BAM, with some city evaluations showing large effects and others showing null results. The variation likely reflects implementation quality: Violence Interrupters require deep community trust, and that trust is hard to standardize. The most consistent finding is for the Group Violence Intervention (call-in) model, which has been evaluated in more sites with more rigorous designs. The central lesson across both models is that the most effective violence reducers are often not police but credible community members who understand the logic of retaliation and can interrupt it.

Research question

"Does a community-based violence interruption model — using credible messengers (often formerly incarcerated individuals) to mediate conflicts before they become shootings — reduce gun violence?"

Methodology

Intervention

Two distinct but related approaches: (1) Cure Violence (formerly CeaseFire) uses trained Violence Interrupters — individuals with criminal histories who have credibility in high-violence communities — to detect and interrupt conflicts before they escalate. Interrupters mediate disputes, change social norms around violence, and connect individuals to services. (2) Group Violence Intervention (David Kennedy's model) convenes high-risk individuals in 'call-ins' with law enforcement, community members, and social service providers, offering services alongside a credible deterrence message ('the next shooting brings the full weight of law enforcement to this group').

Assignment

Quasi-experimental evaluations; several city-level evaluations using synthetic controls and difference-in-differences; no full city-level RCT exists; some individual program sites have been evaluated with regression discontinuity (program catchment boundaries); Cure Violence: Webster et al. Baltimore evaluation (2012); GVI: National Network for Safe Communities city evaluations

Sample size

City-level evaluations in Baltimore, New York, New Orleans, Oakland, Philadelphia; thousands of participants in GVI call-ins across 100+ cities

Primary outcome

Shootings and homicides in target areas; shooting victimization among high-risk individuals

Effect estimate

Cure Violence: Baltimore evaluation found homicides fell 34% and nonfatal shootings fell 21% in program sites vs. comparison areas; NYC results mixed (some sites positive, others null); New Orleans: 49% reduction in homicides in program areas. GVI: National multi-city evaluation (2019) found cities implementing GVI had 17% lower gun homicides than comparable non-GVI cities. Both programs show heterogeneous effects across sites.

Decision

Both models have been adopted by 100+ cities; GVI is used in cities ranging from Oakland to Glasgow to Puerto Rico; DOJ's COPS Office funds both approaches; American cities invested heavily in community violence intervention (CVI) during 2020–2022 crime increase; Biden administration provided $5B for CVI programs in the American Rescue Plan; debate continues about which program elements drive effects and how to measure program fidelity

Result

Mixed

Cure Violence: Baltimore evaluation found homicides fell 34% and nonfatal shootings fell 21% in program sites vs. comparison areas; NYC results mixed (some sites positive, others null); New Orleans: 49% reduction in homicides in program areas. GVI: National multi-city evaluation (2019) found cities implementing GVI had 17% lower gun homicides than comparable non-GVI cities. Both programs show heterogeneous effects across sites.

Evidence strength

Strong

Randomized controlled trial with large sample.

Replication status

Partially replicated

Institution

University of Illinois Chicago / National Network for Safe Communities (John Jay)

Location

Multiple US cities

Year

2000

Policy area

Public Safety

Mechanism

Community engagement