Neighborhood – Kelling et al. (1974)

Study Reference:

Kelling, G. L., Pate, A. M., Dieckman, D., & Brown, C. (1974). The Kansas City preventive patrol experiment: Technical report. Washington, DC: Police Foundation.

Location in the Matrix; Methodological Rigor; Outcome:

Neighborhood; General; Mostly Reactive; Moderately Rigorous; No evidence of an effect

What police practice or strategy was examined?

The study examined the impact of different levels of police patrol in Kansas City, MO. Three controlled levels of routine preventive patrol were compared across 15 beats. One set of beats, termed “reactive,” received no preventive patrol. Officers entered the area only in response to citizen calls for assistance, reducing police visibility in that area. In the second area, called “proactive,” police visibility was increased two to three times its usual level (by adding additional cars for patrol). In the third area, termed “control,” the normal level of patrol (one car per beat) was maintained. However, the geographical distribution of beats avoided clustering reactive beats together or at an unacceptable distance from proactive beats because such clustering could have resulted in lowered response time in the reactive beats.

How was the intervention evaluated?

The study randomly divided the three levels of patrol described above among 15 city police beats (5 beats in each category). The experiment ran successfully for 12 months. Findings examined the effect of experimental conditions on five categories of crimes traditionally considered to be deterrable through preventive patrol (burglary, auto theft, larceny-theft of auto accessories, robbery and vandalism) and on five other crime categories (including rape, assault, and other larcenies.) To measure the effects of the experimental conditions on crime, a victimization survey, departmental reported crime, departmental arrest data, and a survey of businesses were used.

What were the key findings?

– The victimization survey found no statistically significant differences in crime in any of the 69 comparisons made between reactive, control and proactive beats.

– Departmental reported crime showed only one statistically significant difference among 51 comparisons drawn between reactive, control and proactive beats (in the category of “other sex crimes”).

– Crimes which citizens and businessmen said they reported to the police showed statistically significant differences between reactive, control and proactive beats in only 5 of 48 comparisons (three vandalism and two residential burglary), and these differences showed no consistent pattern.

– Police arrests showed no statistically significant differences in the 27 comparisons made between reactive, control and proactive beats.

What were the implications for law enforcement?

The experiment revealed that the noncommitted time of the police officers (60 percent in the experiment) could be used for purposes other than routine patrol without any negative impact on public safety. While random patrol continues to be a common strategy used by police agencies, the results from Kansas City suggest that increasing or decreasing levels of random patrol does not have a substantial impact on crime.

Where can I find more information about this intervention, similar types of intervention, or related studies?