The purpose of this project was to conduct an experiment to study how a collaborative problem solving approach (PS) versus directed patrol (DP) versus standard policing practices (SPP) (the control group) differently impact crime in hot spots, but more importantly how the varied strategies impact residents' opinions about police, their neighborhoods, and their willingness to exert collective efficacy.
Seventy-one hot spots were chosen by identifying crime incidents between December 2010 and November 2011 using kernel density with Robert Cross and Getis Ord GI* and then assessing counts at street segments to identify crime concentrations in residential areas. A random number generator was used to randomly assign treatment status as problem solving approach or directed patrol to 40 hot spots, while the remaining 31 sites were designated as controls.
Twenty-two officers were assigned to the 20 problem solving (PS) sites for the 5-month treatment period and received training in the Scanning, Analysis, Response, and Evaluation (SARA) method of problem solving. Officers involved in this treatment submitted a written update and met monthly with the PI to discuss progress and adjust course. Paid observers documented activities twice per PS hot spot.
The objective for the direct patrol (DP) portion of the study was to double the time spent by officers at each hot spot, averaging around 4.5 hours weekly. Officers involved were presented with a memorandum about the study and the PI and project consultant rode along with officers during the first two weeks of the treatment. Automated vehicle location (AVL) data was used to assess treatment integrity. Officers were encouraged to conduct 11-15 minute patrols each targeted hour during specific "hot times." Dispatchers and officers recorded time spent and officer activities, and forty-one 4-hour blocks of systematic social observation were used to assess the reliability of officer activity data.
The second part of the study assessed community impact through person-to-person survey interviews with hot spot residents about police, perceptions of crime and safety, and willingness to exert collective efficacy. In the first wave, 985 surveys were completed; wave two produced 768 completed surveys, and wave three produced 1,098 surveys.
The study evaluated police officers' perceptions of implementing hot spot policing by colleting pre-treatment surveys from 151 officers in the St. Louis County Police Department and post-treatment surveys from 240 officers.
At wave 1: Random sample of addresses within each hot spot in St. Louis County. Person who answered the door, if agreeable, and at least 18 years old, was interviewed.
At waves 2-3: Completed interviews were re-contacted and a new random sample of addresses at each hot spot in St. Louis County was drawn to supplement. Person who answered the door, if agreeable, and at least 18 years old, was interviewed.
Longitudinal: Panel: Continuous
Longitudinal: Trend / Repeated Cross-section
Community impact: residents of designated hot spots in St. Louis County, Missouri.
Officer assessment: all officers in the St. Louis County, Missouri Police Department during 2012-2013.
coded on-site observation
This study includes six SPSS files.
- SCHIRA_aggregatecrimecfsdata.sav: This file includes 375 cases of crime recorded from December 2010 - May 2013, before, during, and after the 5 month treatment period. The 6 variables include designation of treatment group, the week of the incident, and aggregate calls for service and crimes for the designated week.
- SCHIRA_AVLdata.sav: This file includes 71 cases which correspond to the hot spots chosen for the study. The 25 variables include designation of treatment group, baseline average time spent at the hot spot, and total time spent at the hot spot for each week for the duration of the treatment.
- SCHIRA_cmtydata.sav: This file includes 2852 cases comprised of responses to all three waves of the community survey. The 101 variables include unique identifiers for respondent, address, and hot spot, as well as designation of treatment group, wave, North County precinct or not, survey format, length of respondent's residency, perception of crime, police performance, police involvement in the community, satisfaction with police, evaluation of neighborhood cohesion, legitimacy and trustworthiness of police, crime victimization, response to area crime, sense of safety in the area, and how often the respondent commits various misdemeanors. This dataset also includes demographic information about the respondent including marital status, age, race/ethnicity, level of education, annual income, and gender.
- SCHIRA_Observerdata.sav: This file contains 195 cases comprised of observer's report on officer behavior during patrol in the hot spots. The 19 variables include date, hot spot location, time of observation, and details about police behavior during the observed patrol.
- SCHIRA_OfficerActivitydata.sav: This file contains 3534 cases comprised of data about reported police officer direct patrol activity during the 5 month treatment. The 33 variables include questions about the patrol, including time, hot spot, type of patrol, officer activity and citizen interactions.
- SCHIRA_OfficerSurveyDataICPSR.sav: The 375 cases in this file contain responses to the pre- and post-treatment police officer survey. The 48 variables include questions to which officers responded on a scale of agreement to disagreement. Topics include the relationship between police officers and the communities that they serve, police legitimacy, law enforcement strategies, job satisfaction, length of employment, gender, and race/ethnicity. Questions specific to the post-treatment survey included specific services provided in the assigned hot spots, training attendance, feedback on aspects of the treatment including training, time allocation, supervisor support, community interactions, and changes to law-enforcement activity.
- Wave 1: 23.72% completion, 38.43% cooperation
- Wave 2: 22.41% completion, 42.17% cooperation
- Wave 3: 33.28% completion, 52.23% cooperation
A Likert-type scale was used.