Neighbors, and What People Actually Do

The Campbell Collaboration review of neighborhood watch reviewed 19 studies with 43 evaluations and meta-analyzed 12 studies with 18 evaluations, reporting odds ratios interpreted as a 16 to 26% relative reduction in crime, with a majority of individual schemes indicating effectiveness. That is the headline, and quoting it alone misrepresents the review.

Because the same authors reported significant publication bias: among studies eligible for meta-analysis, 78% of those included showed positive results against only 28% of those excluded.

Comparison areas were rarely equivalent and sometimes not equivalent at all. Very few studies explained the mechanism by which a scheme was supposed to reduce crime. Effects were stronger in North America than in the UK, for reasons the review did not explain.

A systematic review's limitations section is part of its finding, and this one says: watch can work, and the published record probably overstates by how much.

There is also a documented failure mode. Watch schemes and their app-based successors have a well-documented record of being used to report neighbors and visitors as suspicious on the basis of race.

The largest neighborhood platform acknowledged this about its own crime-and-safety feature and redesigned it to require descriptions of behavior rather than appearance, reporting a substantial reduction afterward. That figure is the company's account of its own product.

One rule separates the version that works from the version that harms: report behavior, never appearance. What the person is doing, where, when, for how long. Observed is not the same as inferred.

For meeting the people who respond, the programs already exist. Citizens police academies run multi-week civilian courses. The research on them is mixed, finding that participants learn and that greater knowledge predicted lower concern about use of force, without translating into broader trust gains.

Coffee-with-a-cop events have no agenda, which is the point, and one attendance produces a name and a number. Volunteers in Police Service supports civilian units inside departments. And CERT, run by FEMA through local fire departments, teaches nine units including triage, light search and rescue, incident command and disaster psychology, with a free independent-study introduction available first.

Close with what disaster research found, because it contradicts almost every film. Mass panic is rare. The documented dominant pattern is prosocial — neighbors helping neighbors, emergent volunteer groups, orderly evacuation.

The myth has operational consequences: research on Hurricane Katrina coverage documented how assumptions about panic and disorder shaped the official response itself.

The honest limit is that this is descriptive sociology rather than an effectiveness trial, and no study was found quantifying how much a neighborhood network improves outcomes. The actionable version still fits on an index card: know the names of the households either side and opposite, know which are elderly, disabled or device-dependent, and make sure one of them knows the same about you.