What reduces crime — and what doesn't.

Targeting — not presence — determines whether policing reduces crime.

The Kansas City Preventive Patrol Experiment (1972) found that increasing or decreasing random patrol had no measurable effect on crime rates, calls for service, or citizen fear. The Kansas City Gun Experiment (1992) targeted the same type of high-crime area but with directed patrol specifically focused on gun detection — and reduced gun crime 49%. The difference is not more or less policing; it is where and how policing is concentrated. Hot-spots research consistently finds that 80% of crime occurs in 3–5% of locations. Interventions targeted to those locations produce consistent reductions; diffuse interventions in the same geography do not.

Cognitive behavioral therapy reduces violent crime more reliably than deterrence or information.

The Becoming a Man experiment produced a 44–50% reduction in violent crime arrests among high-risk youth — one of the largest effects ever measured in a criminal justice intervention. Drug courts, which combine judicial supervision with mandatory treatment, produce 8–14 percentage point reductions in recidivism vs. standard prosecution. Both interventions target cognitive processes rather than information, deterrence, or punishment. D.A.R.E., which provides information about the consequences of drug use to youth who already know the consequences, produces no measurable reduction in drug use across 10-year follow-up. The evidence strongly favors mechanism-level targeting: what are the actual cognitive and behavioral processes driving risk?

Popular programs frequently fail rigorous evaluation — and continue regardless.

Three of the most widely adopted crime-reduction programs of the past 40 years — D.A.R.E. drug prevention, Scared Straight, and police body-worn cameras — have been subjected to rigorous randomized or quasi-experimental evaluation and found to produce null or negative effects. Scared Straight, a program that persisted for decades across US states, was found by a meta-analysis of 9 RCTs to increase re-arrest rates by 13% vs. controls. Police body cameras in Washington DC produced no measurable change in use of force or civilian complaints in a 2,224-officer randomized trial. Program popularity and political support are not correlated with program effectiveness.

Community violence intervention shows promise — but implementation quality drives the effect.

Cure Violence and the Group Violence Intervention model show significant reductions in shootings in some evaluations and null results in others. The best quasi-experimental evaluations (Baltimore, New Orleans, multi-city GVI) find 20–49% reductions in homicides in program areas. The heterogeneity across sites is large and likely reflects implementation quality: Violence Interrupters require deep community trust, which is difficult to scale uniformly. The most consistent finding is for the Group Violence Intervention's 'call-in' model, where law enforcement, community members, and social service providers jointly address the highest-risk individuals. The evidence is promising but not yet at the level of hot-spots policing or CBT programs.

D.A.R.E. — Drug Abuse Resistance Education

Multiple US cities · 1983

Information about drug consequences does not change adolescent drug use behavior. The program's mechanism (awareness) does not match the actual drivers of use (social identity, peer pressure, risk preference).

Scared Straight — Juvenile Awareness Project

New Jersey, USA · 1976

Confrontational exposure to prison culture may increase rather than decrease criminal identity among at-risk youth. The 'scared straight' mechanism is theoretically plausible but empirically harmful.

Police Body-Worn Cameras — Washington DC Randomized Trial

Washington, DC, USA · 2016

Surveillance accountability changes officer behavior in theory but not in the DC RCT. Footage review policies, activation protocols, and enforcement mechanisms may matter more than camera presence.

Kansas City Preventive Patrol Experiment

Kansas City, USA · 1972

Random preventive patrol — driving through an area — does not reduce crime. The deterrence mechanism requires specificity and concentration, not presence.