What hot spots policing actually showed.

The Kansas City Gun Experiment (1995) is widely cited as the origin of hot spots policing. Increased patrol in specific high-crime blocks reduced gun crime by 49% in target areas. The finding has been replicated more than 50 times across cities on five continents. It is among the most robust in all of applied criminology.

It is also routinely misapplied.

What the experiment actually measured

The original experiment targeted micro-locations — specific blocks, intersections, and addresses — not neighborhoods or precincts. Officers were assigned to patrol these precise locations at specific times of day, based on crime mapping that identified where and when gun incidents were most likely to occur.

The comparison was not "patrol these blocks more" versus "patrol them as usual." It was "saturate these exact locations during peak hours" versus "patrol as usual." The precision was the point.

What gets implemented

What cities implement when they say they are doing hot spots policing is usually something much broader: directing patrol resources toward "high-crime areas," which might mean zip codes, districts, or precincts. The geographic unit is larger by an order of magnitude. The time-of-day precision is absent. The underlying crime mapping may be months out of date.

This is not the intervention that was studied. Whether it works at that scale is a separate empirical question — one that has rarely been answered with the rigor of the original.

The general trap

Hot spots policing is an unusually clear example of a general problem in applying evidence from experiments to policy: the effective component of an intervention is usually narrower than the label on the box.

When a program gets labeled and adopted at scale, the label travels farther than the mechanism. What practitioners adopt is a category — "hot spots policing," "conditional cash transfers," "mentoring programs" — not a precise specification of what was actually done in the study.

The replication studies that exist are often testing the category, not the mechanism. When they find null results, the finding is ambiguous: did the mechanism fail, or was the mechanism never faithfully implemented?

A protocol proposal

One practical response to this problem is to require implementation fidelity checklists as part of the experiment documentation. Before a study result is treated as actionable, practitioners should be able to answer: what exactly was done, in what dosage, delivered by whom, to which population, in what setting?

The Experiment Society registry is beginning to add fidelity fields to submitted experiments for this reason. The question "does it replicate?" is less interesting than the question "under what conditions does it replicate?"

Next issue: Administrative friction as a policy choice — what a decade of simplification experiments teaches us.