Study 1
Telep, Mitchell, and Weisburd (2014) conducted a randomized controlled trial to evaluate whether having officers move randomly from hot spot to hot spot and spend about 15 minutes in each area made an impact on crime incidents and calls for service, when compared with standard police patrol. The intervention was implemented in Sacramento, Calif., from Feb. 8, 2011, to May 8, 2011. In the planning phase, the Sacramento Police Department identified all citizen-generated calls for service for District 3 (downtown and midtown areas and the campus of California State University, Sacramento) and District 6 (the southeastern part of the city and the neighborhoods of Oak Park and Tahoe Park) from January 2009 through December 2010. The department excluded the following from the sample: 1) all noncrime-related calls, 2) calls geocoded to an intersection to create hot spots that were a street block in length, and 3) calls to certain high-call addresses that did not qualify as typical hot spots such as hospitals and the county courthouse. A total of 119,480 geocoded calls from District 3 and District 6 remained in the sample, but the department looked at the top 40 hot spots based on calls for service. For Part 1 crimes, all 2010 incidents were examined and, after intersection data and ineligible addresses were removed, a total of 7,479 incidents remained in the sample. The same criteria as calls for services were used, street segments were ranked, and the top 25 hot spots for Part 1 crimes not identified from the top 40 calls for service were added. The top 20 hot spots for soft crime incidents were assessed and removed from the sample, as they did not correlate highly with the calls for service and Part 1 crime hot spots.
After initial assessment, 52 hot spots were identified for further review. Two officers physically observed these areas to confirm that the hot spots were suitable for the experiment. Inclusion criteria at this stage included that no hot spot 1) was larger than one standard linear street block, 2) extended for more than one half-block from either side of an intersection, and 3) was within one standard linear block of another hot spot. Based on officer observations and the inclusion criteria, 42 hot spots were included in the experiment. To reduce variability between the 21 treatment hot spots (that received the intervention) and 21 control hot spots (that received standard police patrol), the hot spots were paired before randomization based on similarity in levels of calls for service, crime incidents, and similar physical appearance based on the initial observations. After pairing, a computerized random-number generator assigned hot spots to either the treatment or control group. Sixteen of the treatment group hot spots were in District 3, and five were in District 6. T-tests showed no statistically significant differences between the treatment and control hot spots in calls for service or Part 1 incidents in 2008, 2009, or 2010, suggesting that there were no baseline differences between the two study conditions. In 2010 (the preintervention period), the treatment hot spots had an average of 46.5 calls for service, 6.6 Part 1 crimes, and 2.8 soft crime incidents, while the control hot spots had an average of 40.6 calls for service, 4.5 Part 1 crimes, and 2.5 soft crime incidents.
The intervention was implemented 7 days a week, between the hours of 9:00 a.m. and 1:00 a.m. As part of this 90-day experiment, officers were assigned from one to six hot spots in their patrol areas and were given a random order, which shifted daily and varied by shift, to visit their assigned hot spots. Officers were not informed about which areas were designated as treatment or control hot spots. Each week, 10 percent of officer-initiated calls in the treatment area were randomly chosen and compared with automated vehicle locater data to determine whether officers were present in their hot spots for the same amount of time that their call logs indicated. Analyses suggested that officers complied with the experimental protocols, were always physically present in the treatment hot spots, and spent 12 to 16 minutes at a time in hot spots. The dosage levels of police presence were approximately 546 visits a week, with the lowest treatment week consisting of 432 visits and the highest 698. In general, there were 7,095 visits (averaged 78.8 visits per day) to 21 treatment group hot spots during the experiment; the total number of visits at each individual hot spot ranged from 223 to 559 visits. In total, 2,875 hours were spent on calls in the treatment hot spots versus 1,014 hours in the control hot spots.
The study examined three outcomes: 1) citizen-generated emergency calls for service to 911, per Sacramento Police Department databases; 2) Part 1 crime incidents based on the Uniform Crime Report classification of serious crimes (e.g., robbery, burglary, auto theft, and aggravated assault); and 3) soft crime incidents or less-serious criminal incidents related to disorder (public drunkenness, trespassing, and vandalism). A difference-in-differences analysis was used to compare the change in the three outcomes between the 90-day intervention period in 2011 and the same 90-day period in 2010 in the treatment hot spots versus the control hot spots. No subgroup analysis was conducted.
Study 2
Mitchell (2017) examined the randomized controlled trial of hot spots in Sacramento, Calif., conducted by Telep, Mitchell, and Weisburd (2014) (Study 1) for a new analysis of crime count decoupled by crime category (violent and property). The hot spot selection, methodology for matching, and randomization process were the same as described in Study 1 (above). The intervention was implemented with the same 21 treatment hot spots (where officers visited one to six high-crime and call-for-service hot spots for 15 minutes of high-visibility patrol multiple times per day) and 21 control hot spots (that received standard police patrol) in Sacramento, Calif., for 90 days from Feb. 8, 2011, to May 8, 2011.
In Study 1 (Telep, Mitchell, and Weisburd, 2014), the study authors examined three primary dependent variables: 1) calls for service, 2) Part 1 crime incidents, and 3) soft crime incidents during the study period in 2011 and the same period in 2010. In Study 2 (Mitchell, 2017), the study author used only the Part 1 crime incident data to examine violent and property crime incidents separately. T-tests showed no statistically significant differences between the treatment and control hot spots in Part 1 incidents in 2010, suggesting there were no baseline differences between the two study conditions. In 2010 (the baseline period), the treatment hot spots had an average of 6.6 Part 1 crimes, while the control hot spots had an average of 4.5 Part 1 crimes.
Rather than analyzing the data by pairs, the data were analyzed as a pooled analysis, making the statistical test more conservative. The analysis included difference-in-difference testing for Part 1 property and violent crime count change between the baseline 90-day period in 2010 and the 90-day experimental period in 2011, comparing the treatment hot spot changes with the control hot spots. No subgroup analysis was conducted.
